xref: /linux/drivers/net/ethernet/intel/e1000e/netdev.c (revision abacaf559950eec0d99d37ff6b92049409af5943)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Copyright(c) 1999 - 2018 Intel Corporation. */
3 
4 #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt
5 
6 #include <linux/module.h>
7 #include <linux/types.h>
8 #include <linux/init.h>
9 #include <linux/pci.h>
10 #include <linux/vmalloc.h>
11 #include <linux/pagemap.h>
12 #include <linux/delay.h>
13 #include <linux/netdevice.h>
14 #include <linux/interrupt.h>
15 #include <linux/tcp.h>
16 #include <linux/ipv6.h>
17 #include <linux/slab.h>
18 #include <net/checksum.h>
19 #include <net/ip6_checksum.h>
20 #include <linux/ethtool.h>
21 #include <linux/if_vlan.h>
22 #include <linux/cpu.h>
23 #include <linux/smp.h>
24 #include <linux/pm_qos.h>
25 #include <linux/pm_runtime.h>
26 #include <linux/prefetch.h>
27 #include <linux/suspend.h>
28 
29 #include "e1000.h"
30 #define CREATE_TRACE_POINTS
31 #include "e1000e_trace.h"
32 
33 char e1000e_driver_name[] = "e1000e";
34 
35 #define DEFAULT_MSG_ENABLE (NETIF_MSG_DRV|NETIF_MSG_PROBE|NETIF_MSG_LINK)
36 static int debug = -1;
37 module_param(debug, int, 0);
38 MODULE_PARM_DESC(debug, "Debug level (0=none,...,16=all)");
39 
40 static const struct e1000_info *e1000_info_tbl[] = {
41 	[board_82571]		= &e1000_82571_info,
42 	[board_82572]		= &e1000_82572_info,
43 	[board_82573]		= &e1000_82573_info,
44 	[board_82574]		= &e1000_82574_info,
45 	[board_82583]		= &e1000_82583_info,
46 	[board_80003es2lan]	= &e1000_es2_info,
47 	[board_ich8lan]		= &e1000_ich8_info,
48 	[board_ich9lan]		= &e1000_ich9_info,
49 	[board_ich10lan]	= &e1000_ich10_info,
50 	[board_pchlan]		= &e1000_pch_info,
51 	[board_pch2lan]		= &e1000_pch2_info,
52 	[board_pch_lpt]		= &e1000_pch_lpt_info,
53 	[board_pch_spt]		= &e1000_pch_spt_info,
54 	[board_pch_cnp]		= &e1000_pch_cnp_info,
55 	[board_pch_tgp]		= &e1000_pch_tgp_info,
56 	[board_pch_adp]		= &e1000_pch_adp_info,
57 	[board_pch_mtp]		= &e1000_pch_mtp_info,
58 	[board_pch_ptp]		= &e1000_pch_ptp_info,
59 };
60 
61 struct e1000_reg_info {
62 	u32 ofs;
63 	char *name;
64 };
65 
66 static const struct e1000_reg_info e1000_reg_info_tbl[] = {
67 	/* General Registers */
68 	{E1000_CTRL, "CTRL"},
69 	{E1000_STATUS, "STATUS"},
70 	{E1000_CTRL_EXT, "CTRL_EXT"},
71 
72 	/* Interrupt Registers */
73 	{E1000_ICR, "ICR"},
74 
75 	/* Rx Registers */
76 	{E1000_RCTL, "RCTL"},
77 	{E1000_RDLEN(0), "RDLEN"},
78 	{E1000_RDH(0), "RDH"},
79 	{E1000_RDT(0), "RDT"},
80 	{E1000_RDTR, "RDTR"},
81 	{E1000_RXDCTL(0), "RXDCTL"},
82 	{E1000_ERT, "ERT"},
83 	{E1000_RDBAL(0), "RDBAL"},
84 	{E1000_RDBAH(0), "RDBAH"},
85 	{E1000_RDFH, "RDFH"},
86 	{E1000_RDFT, "RDFT"},
87 	{E1000_RDFHS, "RDFHS"},
88 	{E1000_RDFTS, "RDFTS"},
89 	{E1000_RDFPC, "RDFPC"},
90 
91 	/* Tx Registers */
92 	{E1000_TCTL, "TCTL"},
93 	{E1000_TDBAL(0), "TDBAL"},
94 	{E1000_TDBAH(0), "TDBAH"},
95 	{E1000_TDLEN(0), "TDLEN"},
96 	{E1000_TDH(0), "TDH"},
97 	{E1000_TDT(0), "TDT"},
98 	{E1000_TIDV, "TIDV"},
99 	{E1000_TXDCTL(0), "TXDCTL"},
100 	{E1000_TADV, "TADV"},
101 	{E1000_TARC(0), "TARC"},
102 	{E1000_TDFH, "TDFH"},
103 	{E1000_TDFT, "TDFT"},
104 	{E1000_TDFHS, "TDFHS"},
105 	{E1000_TDFTS, "TDFTS"},
106 	{E1000_TDFPC, "TDFPC"},
107 
108 	/* List Terminator */
109 	{0, NULL}
110 };
111 
112 /**
113  * __ew32_prepare - prepare to write to MAC CSR register on certain parts
114  * @hw: pointer to the HW structure
115  *
116  * When updating the MAC CSR registers, the Manageability Engine (ME) could
117  * be accessing the registers at the same time.  Normally, this is handled in
118  * h/w by an arbiter but on some parts there is a bug that acknowledges Host
119  * accesses later than it should which could result in the register to have
120  * an incorrect value.  Workaround this by checking the FWSM register which
121  * has bit 24 set while ME is accessing MAC CSR registers, wait if it is set
122  * and try again a number of times.
123  **/
__ew32_prepare(struct e1000_hw * hw)124 static void __ew32_prepare(struct e1000_hw *hw)
125 {
126 	s32 i = E1000_ICH_FWSM_PCIM2PCI_COUNT;
127 
128 	while ((er32(FWSM) & E1000_ICH_FWSM_PCIM2PCI) && --i)
129 		udelay(50);
130 }
131 
__ew32(struct e1000_hw * hw,unsigned long reg,u32 val)132 void __ew32(struct e1000_hw *hw, unsigned long reg, u32 val)
133 {
134 	if (hw->adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
135 		__ew32_prepare(hw);
136 
137 	writel(val, hw->hw_addr + reg);
138 }
139 
140 /**
141  * e1000_regdump - register printout routine
142  * @hw: pointer to the HW structure
143  * @reginfo: pointer to the register info table
144  **/
e1000_regdump(struct e1000_hw * hw,struct e1000_reg_info * reginfo)145 static void e1000_regdump(struct e1000_hw *hw, struct e1000_reg_info *reginfo)
146 {
147 	int n = 0;
148 	char rname[16];
149 	u32 regs[8];
150 
151 	switch (reginfo->ofs) {
152 	case E1000_RXDCTL(0):
153 		for (n = 0; n < 2; n++)
154 			regs[n] = __er32(hw, E1000_RXDCTL(n));
155 		break;
156 	case E1000_TXDCTL(0):
157 		for (n = 0; n < 2; n++)
158 			regs[n] = __er32(hw, E1000_TXDCTL(n));
159 		break;
160 	case E1000_TARC(0):
161 		for (n = 0; n < 2; n++)
162 			regs[n] = __er32(hw, E1000_TARC(n));
163 		break;
164 	default:
165 		pr_info("%-15s %08x\n",
166 			reginfo->name, __er32(hw, reginfo->ofs));
167 		return;
168 	}
169 
170 	snprintf(rname, 16, "%s%s", reginfo->name, "[0-1]");
171 	pr_info("%-15s %08x %08x\n", rname, regs[0], regs[1]);
172 }
173 
e1000e_dump_ps_pages(struct e1000_adapter * adapter,struct e1000_buffer * bi)174 static void e1000e_dump_ps_pages(struct e1000_adapter *adapter,
175 				 struct e1000_buffer *bi)
176 {
177 	int i;
178 	struct e1000_ps_page *ps_page;
179 
180 	for (i = 0; i < adapter->rx_ps_pages; i++) {
181 		ps_page = &bi->ps_pages[i];
182 
183 		if (ps_page->page) {
184 			pr_info("packet dump for ps_page %d:\n", i);
185 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
186 				       16, 1, page_address(ps_page->page),
187 				       PAGE_SIZE, true);
188 		}
189 	}
190 }
191 
192 /**
193  * e1000e_dump - Print registers, Tx-ring and Rx-ring
194  * @adapter: board private structure
195  **/
e1000e_dump(struct e1000_adapter * adapter)196 static void e1000e_dump(struct e1000_adapter *adapter)
197 {
198 	struct net_device *netdev = adapter->netdev;
199 	struct e1000_hw *hw = &adapter->hw;
200 	struct e1000_reg_info *reginfo;
201 	struct e1000_ring *tx_ring = adapter->tx_ring;
202 	struct e1000_tx_desc *tx_desc;
203 	struct my_u0 {
204 		__le64 a;
205 		__le64 b;
206 	} *u0;
207 	struct e1000_buffer *buffer_info;
208 	struct e1000_ring *rx_ring = adapter->rx_ring;
209 	union e1000_rx_desc_packet_split *rx_desc_ps;
210 	union e1000_rx_desc_extended *rx_desc;
211 	struct my_u1 {
212 		__le64 a;
213 		__le64 b;
214 		__le64 c;
215 		__le64 d;
216 	} *u1;
217 	u32 staterr;
218 	int i = 0;
219 
220 	if (!netif_msg_hw(adapter))
221 		return;
222 
223 	/* Print netdevice Info */
224 	if (netdev) {
225 		dev_info(&adapter->pdev->dev, "Net device Info\n");
226 		pr_info("Device Name     state            trans_start\n");
227 		pr_info("%-15s %016lX %016lX\n", netdev->name,
228 			netdev->state, dev_trans_start(netdev));
229 	}
230 
231 	/* Print Registers */
232 	dev_info(&adapter->pdev->dev, "Register Dump\n");
233 	pr_info(" Register Name   Value\n");
234 	for (reginfo = (struct e1000_reg_info *)e1000_reg_info_tbl;
235 	     reginfo->name; reginfo++) {
236 		e1000_regdump(hw, reginfo);
237 	}
238 
239 	/* Print Tx Ring Summary */
240 	if (!netdev || !netif_running(netdev))
241 		return;
242 
243 	dev_info(&adapter->pdev->dev, "Tx Ring Summary\n");
244 	pr_info("Queue [NTU] [NTC] [bi(ntc)->dma  ] leng ntw timestamp\n");
245 	buffer_info = &tx_ring->buffer_info[tx_ring->next_to_clean];
246 	pr_info(" %5d %5X %5X %016llX %04X %3X %016llX\n",
247 		0, tx_ring->next_to_use, tx_ring->next_to_clean,
248 		(unsigned long long)buffer_info->dma,
249 		buffer_info->length,
250 		buffer_info->next_to_watch,
251 		(unsigned long long)buffer_info->time_stamp);
252 
253 	/* Print Tx Ring */
254 	if (!netif_msg_tx_done(adapter))
255 		goto rx_ring_summary;
256 
257 	dev_info(&adapter->pdev->dev, "Tx Ring Dump\n");
258 
259 	/* Transmit Descriptor Formats - DEXT[29] is 0 (Legacy) or 1 (Extended)
260 	 *
261 	 * Legacy Transmit Descriptor
262 	 *   +--------------------------------------------------------------+
263 	 * 0 |         Buffer Address [63:0] (Reserved on Write Back)       |
264 	 *   +--------------------------------------------------------------+
265 	 * 8 | Special  |    CSS     | Status |  CMD    |  CSO   |  Length  |
266 	 *   +--------------------------------------------------------------+
267 	 *   63       48 47        36 35    32 31     24 23    16 15        0
268 	 *
269 	 * Extended Context Descriptor (DTYP=0x0) for TSO or checksum offload
270 	 *   63      48 47    40 39       32 31             16 15    8 7      0
271 	 *   +----------------------------------------------------------------+
272 	 * 0 |  TUCSE  | TUCS0  |   TUCSS   |     IPCSE       | IPCS0 | IPCSS |
273 	 *   +----------------------------------------------------------------+
274 	 * 8 |   MSS   | HDRLEN | RSV | STA | TUCMD | DTYP |      PAYLEN      |
275 	 *   +----------------------------------------------------------------+
276 	 *   63      48 47    40 39 36 35 32 31   24 23  20 19                0
277 	 *
278 	 * Extended Data Descriptor (DTYP=0x1)
279 	 *   +----------------------------------------------------------------+
280 	 * 0 |                     Buffer Address [63:0]                      |
281 	 *   +----------------------------------------------------------------+
282 	 * 8 | VLAN tag |  POPTS  | Rsvd | Status | Command | DTYP |  DTALEN  |
283 	 *   +----------------------------------------------------------------+
284 	 *   63       48 47     40 39  36 35    32 31     24 23  20 19        0
285 	 */
286 	pr_info("Tl[desc]     [address 63:0  ] [SpeCssSCmCsLen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Legacy format\n");
287 	pr_info("Tc[desc]     [Ce CoCsIpceCoS] [MssHlRSCm0Plen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Context format\n");
288 	pr_info("Td[desc]     [address 63:0  ] [VlaPoRSCm1Dlen] [bi->dma       ] leng  ntw timestamp        bi->skb <-- Ext Data format\n");
289 	for (i = 0; tx_ring->desc && (i < tx_ring->count); i++) {
290 		const char *next_desc;
291 		tx_desc = E1000_TX_DESC(*tx_ring, i);
292 		buffer_info = &tx_ring->buffer_info[i];
293 		u0 = (struct my_u0 *)tx_desc;
294 		if (i == tx_ring->next_to_use && i == tx_ring->next_to_clean)
295 			next_desc = " NTC/U";
296 		else if (i == tx_ring->next_to_use)
297 			next_desc = " NTU";
298 		else if (i == tx_ring->next_to_clean)
299 			next_desc = " NTC";
300 		else
301 			next_desc = "";
302 		pr_info("T%c[0x%03X]    %016llX %016llX %016llX %04X  %3X %016llX %p%s\n",
303 			(!(le64_to_cpu(u0->b) & BIT(29)) ? 'l' :
304 			 ((le64_to_cpu(u0->b) & BIT(20)) ? 'd' : 'c')),
305 			i,
306 			(unsigned long long)le64_to_cpu(u0->a),
307 			(unsigned long long)le64_to_cpu(u0->b),
308 			(unsigned long long)buffer_info->dma,
309 			buffer_info->length, buffer_info->next_to_watch,
310 			(unsigned long long)buffer_info->time_stamp,
311 			buffer_info->skb, next_desc);
312 
313 		if (netif_msg_pktdata(adapter) && buffer_info->skb)
314 			print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS,
315 				       16, 1, buffer_info->skb->data,
316 				       buffer_info->skb->len, true);
317 	}
318 
319 	/* Print Rx Ring Summary */
320 rx_ring_summary:
321 	dev_info(&adapter->pdev->dev, "Rx Ring Summary\n");
322 	pr_info("Queue [NTU] [NTC]\n");
323 	pr_info(" %5d %5X %5X\n",
324 		0, rx_ring->next_to_use, rx_ring->next_to_clean);
325 
326 	/* Print Rx Ring */
327 	if (!netif_msg_rx_status(adapter))
328 		return;
329 
330 	dev_info(&adapter->pdev->dev, "Rx Ring Dump\n");
331 	switch (adapter->rx_ps_pages) {
332 	case 1:
333 	case 2:
334 	case 3:
335 		/* [Extended] Packet Split Receive Descriptor Format
336 		 *
337 		 *    +-----------------------------------------------------+
338 		 *  0 |                Buffer Address 0 [63:0]              |
339 		 *    +-----------------------------------------------------+
340 		 *  8 |                Buffer Address 1 [63:0]              |
341 		 *    +-----------------------------------------------------+
342 		 * 16 |                Buffer Address 2 [63:0]              |
343 		 *    +-----------------------------------------------------+
344 		 * 24 |                Buffer Address 3 [63:0]              |
345 		 *    +-----------------------------------------------------+
346 		 */
347 		pr_info("R  [desc]      [buffer 0 63:0 ] [buffer 1 63:0 ] [buffer 2 63:0 ] [buffer 3 63:0 ] [bi->dma       ] [bi->skb] <-- Ext Pkt Split format\n");
348 		/* [Extended] Receive Descriptor (Write-Back) Format
349 		 *
350 		 *   63       48 47    32 31     13 12    8 7    4 3        0
351 		 *   +------------------------------------------------------+
352 		 * 0 | Packet   | IP     |  Rsvd   | MRQ   | Rsvd | MRQ RSS |
353 		 *   | Checksum | Ident  |         | Queue |      |  Type   |
354 		 *   +------------------------------------------------------+
355 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
356 		 *   +------------------------------------------------------+
357 		 *   63       48 47    32 31            20 19               0
358 		 */
359 		pr_info("RWB[desc]      [ck ipid mrqhsh] [vl   l0 ee  es] [ l3  l2  l1 hs] [reserved      ] ---------------- [bi->skb] <-- Ext Rx Write-Back format\n");
360 		for (i = 0; i < rx_ring->count; i++) {
361 			const char *next_desc;
362 			buffer_info = &rx_ring->buffer_info[i];
363 			rx_desc_ps = E1000_RX_DESC_PS(*rx_ring, i);
364 			u1 = (struct my_u1 *)rx_desc_ps;
365 			staterr =
366 			    le32_to_cpu(rx_desc_ps->wb.middle.status_error);
367 
368 			if (i == rx_ring->next_to_use)
369 				next_desc = " NTU";
370 			else if (i == rx_ring->next_to_clean)
371 				next_desc = " NTC";
372 			else
373 				next_desc = "";
374 
375 			if (staterr & E1000_RXD_STAT_DD) {
376 				/* Descriptor Done */
377 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX ---------------- %p%s\n",
378 					"RWB", i,
379 					(unsigned long long)le64_to_cpu(u1->a),
380 					(unsigned long long)le64_to_cpu(u1->b),
381 					(unsigned long long)le64_to_cpu(u1->c),
382 					(unsigned long long)le64_to_cpu(u1->d),
383 					buffer_info->skb, next_desc);
384 			} else {
385 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %016llX %016llX %p%s\n",
386 					"R  ", i,
387 					(unsigned long long)le64_to_cpu(u1->a),
388 					(unsigned long long)le64_to_cpu(u1->b),
389 					(unsigned long long)le64_to_cpu(u1->c),
390 					(unsigned long long)le64_to_cpu(u1->d),
391 					(unsigned long long)buffer_info->dma,
392 					buffer_info->skb, next_desc);
393 
394 				if (netif_msg_pktdata(adapter))
395 					e1000e_dump_ps_pages(adapter,
396 							     buffer_info);
397 			}
398 		}
399 		break;
400 	default:
401 	case 0:
402 		/* Extended Receive Descriptor (Read) Format
403 		 *
404 		 *   +-----------------------------------------------------+
405 		 * 0 |                Buffer Address [63:0]                |
406 		 *   +-----------------------------------------------------+
407 		 * 8 |                      Reserved                       |
408 		 *   +-----------------------------------------------------+
409 		 */
410 		pr_info("R  [desc]      [buf addr 63:0 ] [reserved 63:0 ] [bi->dma       ] [bi->skb] <-- Ext (Read) format\n");
411 		/* Extended Receive Descriptor (Write-Back) Format
412 		 *
413 		 *   63       48 47    32 31    24 23            4 3        0
414 		 *   +------------------------------------------------------+
415 		 *   |     RSS Hash      |        |               |         |
416 		 * 0 +-------------------+  Rsvd  |   Reserved    | MRQ RSS |
417 		 *   | Packet   | IP     |        |               |  Type   |
418 		 *   | Checksum | Ident  |        |               |         |
419 		 *   +------------------------------------------------------+
420 		 * 8 | VLAN Tag | Length | Extended Error | Extended Status |
421 		 *   +------------------------------------------------------+
422 		 *   63       48 47    32 31            20 19               0
423 		 */
424 		pr_info("RWB[desc]      [cs ipid    mrq] [vt   ln xe  xs] [bi->skb] <-- Ext (Write-Back) format\n");
425 
426 		for (i = 0; i < rx_ring->count; i++) {
427 			const char *next_desc;
428 
429 			buffer_info = &rx_ring->buffer_info[i];
430 			rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
431 			u1 = (struct my_u1 *)rx_desc;
432 			staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
433 
434 			if (i == rx_ring->next_to_use)
435 				next_desc = " NTU";
436 			else if (i == rx_ring->next_to_clean)
437 				next_desc = " NTC";
438 			else
439 				next_desc = "";
440 
441 			if (staterr & E1000_RXD_STAT_DD) {
442 				/* Descriptor Done */
443 				pr_info("%s[0x%03X]     %016llX %016llX ---------------- %p%s\n",
444 					"RWB", i,
445 					(unsigned long long)le64_to_cpu(u1->a),
446 					(unsigned long long)le64_to_cpu(u1->b),
447 					buffer_info->skb, next_desc);
448 			} else {
449 				pr_info("%s[0x%03X]     %016llX %016llX %016llX %p%s\n",
450 					"R  ", i,
451 					(unsigned long long)le64_to_cpu(u1->a),
452 					(unsigned long long)le64_to_cpu(u1->b),
453 					(unsigned long long)buffer_info->dma,
454 					buffer_info->skb, next_desc);
455 
456 				if (netif_msg_pktdata(adapter) &&
457 				    buffer_info->skb)
458 					print_hex_dump(KERN_INFO, "",
459 						       DUMP_PREFIX_ADDRESS, 16,
460 						       1,
461 						       buffer_info->skb->data,
462 						       adapter->rx_buffer_len,
463 						       true);
464 			}
465 		}
466 	}
467 }
468 
469 /**
470  * e1000_desc_unused - calculate if we have unused descriptors
471  * @ring: pointer to ring struct to perform calculation on
472  **/
e1000_desc_unused(struct e1000_ring * ring)473 static int e1000_desc_unused(struct e1000_ring *ring)
474 {
475 	if (ring->next_to_clean > ring->next_to_use)
476 		return ring->next_to_clean - ring->next_to_use - 1;
477 
478 	return ring->count + ring->next_to_clean - ring->next_to_use - 1;
479 }
480 
481 /**
482  * e1000e_systim_to_hwtstamp - convert system time value to hw time stamp
483  * @adapter: board private structure
484  * @hwtstamps: time stamp structure to update
485  * @systim: unsigned 64bit system time value.
486  *
487  * Convert the system time value stored in the RX/TXSTMP registers into a
488  * hwtstamp which can be used by the upper level time stamping functions.
489  *
490  * The 'systim_lock' spinlock is used to protect the consistency of the
491  * system time value. This is needed because reading the 64 bit time
492  * value involves reading two 32 bit registers. The first read latches the
493  * value.
494  **/
e1000e_systim_to_hwtstamp(struct e1000_adapter * adapter,struct skb_shared_hwtstamps * hwtstamps,u64 systim)495 static void e1000e_systim_to_hwtstamp(struct e1000_adapter *adapter,
496 				      struct skb_shared_hwtstamps *hwtstamps,
497 				      u64 systim)
498 {
499 	u64 ns;
500 	unsigned long flags;
501 
502 	spin_lock_irqsave(&adapter->systim_lock, flags);
503 	ns = timecounter_cyc2time(&adapter->tc, systim);
504 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
505 
506 	memset(hwtstamps, 0, sizeof(*hwtstamps));
507 	hwtstamps->hwtstamp = ns_to_ktime(ns);
508 }
509 
510 /**
511  * e1000e_rx_hwtstamp - utility function which checks for Rx time stamp
512  * @adapter: board private structure
513  * @status: descriptor extended error and status field
514  * @skb: particular skb to include time stamp
515  *
516  * If the time stamp is valid, convert it into the timecounter ns value
517  * and store that result into the shhwtstamps structure which is passed
518  * up the network stack.
519  **/
e1000e_rx_hwtstamp(struct e1000_adapter * adapter,u32 status,struct sk_buff * skb)520 static void e1000e_rx_hwtstamp(struct e1000_adapter *adapter, u32 status,
521 			       struct sk_buff *skb)
522 {
523 	struct e1000_hw *hw = &adapter->hw;
524 	u64 rxstmp;
525 
526 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP) ||
527 	    !(status & E1000_RXDEXT_STATERR_TST) ||
528 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID))
529 		return;
530 
531 	/* The Rx time stamp registers contain the time stamp.  No other
532 	 * received packet will be time stamped until the Rx time stamp
533 	 * registers are read.  Because only one packet can be time stamped
534 	 * at a time, the register values must belong to this packet and
535 	 * therefore none of the other additional attributes need to be
536 	 * compared.
537 	 */
538 	rxstmp = (u64)er32(RXSTMPL);
539 	rxstmp |= (u64)er32(RXSTMPH) << 32;
540 	e1000e_systim_to_hwtstamp(adapter, skb_hwtstamps(skb), rxstmp);
541 
542 	adapter->flags2 &= ~FLAG2_CHECK_RX_HWTSTAMP;
543 }
544 
545 /**
546  * e1000_receive_skb - helper function to handle Rx indications
547  * @adapter: board private structure
548  * @netdev: pointer to netdev struct
549  * @staterr: descriptor extended error and status field as written by hardware
550  * @vlan: descriptor vlan field as written by hardware (no le/be conversion)
551  * @skb: pointer to sk_buff to be indicated to stack
552  **/
e1000_receive_skb(struct e1000_adapter * adapter,struct net_device * netdev,struct sk_buff * skb,u32 staterr,__le16 vlan)553 static void e1000_receive_skb(struct e1000_adapter *adapter,
554 			      struct net_device *netdev, struct sk_buff *skb,
555 			      u32 staterr, __le16 vlan)
556 {
557 	u16 tag = le16_to_cpu(vlan);
558 
559 	e1000e_rx_hwtstamp(adapter, staterr, skb);
560 
561 	skb->protocol = eth_type_trans(skb, netdev);
562 
563 	if (staterr & E1000_RXD_STAT_VP)
564 		__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q), tag);
565 
566 	napi_gro_receive(&adapter->napi, skb);
567 }
568 
569 /**
570  * e1000_rx_checksum - Receive Checksum Offload
571  * @adapter: board private structure
572  * @status_err: receive descriptor status and error fields
573  * @skb: socket buffer with received data
574  **/
e1000_rx_checksum(struct e1000_adapter * adapter,u32 status_err,struct sk_buff * skb)575 static void e1000_rx_checksum(struct e1000_adapter *adapter, u32 status_err,
576 			      struct sk_buff *skb)
577 {
578 	u16 status = (u16)status_err;
579 	u8 errors = (u8)(status_err >> 24);
580 
581 	skb_checksum_none_assert(skb);
582 
583 	/* Rx checksum disabled */
584 	if (!(adapter->netdev->features & NETIF_F_RXCSUM))
585 		return;
586 
587 	/* Ignore Checksum bit is set */
588 	if (status & E1000_RXD_STAT_IXSM)
589 		return;
590 
591 	/* TCP/UDP checksum error bit or IP checksum error bit is set */
592 	if (errors & (E1000_RXD_ERR_TCPE | E1000_RXD_ERR_IPE)) {
593 		/* let the stack verify checksum errors */
594 		adapter->hw_csum_err++;
595 		return;
596 	}
597 
598 	/* TCP/UDP Checksum has not been calculated */
599 	if (!(status & (E1000_RXD_STAT_TCPCS | E1000_RXD_STAT_UDPCS)))
600 		return;
601 
602 	/* It must be a TCP or UDP packet with a valid checksum */
603 	skb->ip_summed = CHECKSUM_UNNECESSARY;
604 	adapter->hw_csum_good++;
605 }
606 
e1000e_update_rdt_wa(struct e1000_ring * rx_ring,unsigned int i)607 static void e1000e_update_rdt_wa(struct e1000_ring *rx_ring, unsigned int i)
608 {
609 	struct e1000_adapter *adapter = rx_ring->adapter;
610 	struct e1000_hw *hw = &adapter->hw;
611 
612 	__ew32_prepare(hw);
613 	writel(i, rx_ring->tail);
614 
615 	if (unlikely(i != readl(rx_ring->tail))) {
616 		u32 rctl = er32(RCTL);
617 
618 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
619 		e_err("ME firmware caused invalid RDT - resetting\n");
620 		schedule_work(&adapter->reset_task);
621 	}
622 }
623 
e1000e_update_tdt_wa(struct e1000_ring * tx_ring,unsigned int i)624 static void e1000e_update_tdt_wa(struct e1000_ring *tx_ring, unsigned int i)
625 {
626 	struct e1000_adapter *adapter = tx_ring->adapter;
627 	struct e1000_hw *hw = &adapter->hw;
628 
629 	__ew32_prepare(hw);
630 	writel(i, tx_ring->tail);
631 
632 	if (unlikely(i != readl(tx_ring->tail))) {
633 		u32 tctl = er32(TCTL);
634 
635 		ew32(TCTL, tctl & ~E1000_TCTL_EN);
636 		e_err("ME firmware caused invalid TDT - resetting\n");
637 		schedule_work(&adapter->reset_task);
638 	}
639 }
640 
641 /**
642  * e1000_alloc_rx_buffers - Replace used receive buffers
643  * @rx_ring: Rx descriptor ring
644  * @cleaned_count: number to reallocate
645  * @gfp: flags for allocation
646  **/
e1000_alloc_rx_buffers(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)647 static void e1000_alloc_rx_buffers(struct e1000_ring *rx_ring,
648 				   int cleaned_count, gfp_t gfp)
649 {
650 	struct e1000_adapter *adapter = rx_ring->adapter;
651 	struct net_device *netdev = adapter->netdev;
652 	struct pci_dev *pdev = adapter->pdev;
653 	union e1000_rx_desc_extended *rx_desc;
654 	struct e1000_buffer *buffer_info;
655 	struct sk_buff *skb;
656 	unsigned int i;
657 	unsigned int bufsz = adapter->rx_buffer_len;
658 
659 	i = rx_ring->next_to_use;
660 	buffer_info = &rx_ring->buffer_info[i];
661 
662 	while (cleaned_count--) {
663 		skb = buffer_info->skb;
664 		if (skb) {
665 			skb_trim(skb, 0);
666 			goto map_skb;
667 		}
668 
669 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
670 		if (!skb) {
671 			/* Better luck next round */
672 			adapter->alloc_rx_buff_failed++;
673 			break;
674 		}
675 
676 		buffer_info->skb = skb;
677 map_skb:
678 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
679 						  adapter->rx_buffer_len,
680 						  DMA_FROM_DEVICE);
681 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
682 			dev_err(&pdev->dev, "Rx DMA map failed\n");
683 			adapter->rx_dma_failed++;
684 			break;
685 		}
686 
687 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
688 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
689 
690 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
691 			/* Force memory writes to complete before letting h/w
692 			 * know there are new descriptors to fetch.  (Only
693 			 * applicable for weak-ordered memory model archs,
694 			 * such as IA-64).
695 			 */
696 			wmb();
697 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
698 				e1000e_update_rdt_wa(rx_ring, i);
699 			else
700 				writel(i, rx_ring->tail);
701 		}
702 		i++;
703 		if (i == rx_ring->count)
704 			i = 0;
705 		buffer_info = &rx_ring->buffer_info[i];
706 	}
707 
708 	rx_ring->next_to_use = i;
709 }
710 
711 /**
712  * e1000_alloc_rx_buffers_ps - Replace used receive buffers; packet split
713  * @rx_ring: Rx descriptor ring
714  * @cleaned_count: number to reallocate
715  * @gfp: flags for allocation
716  **/
e1000_alloc_rx_buffers_ps(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)717 static void e1000_alloc_rx_buffers_ps(struct e1000_ring *rx_ring,
718 				      int cleaned_count, gfp_t gfp)
719 {
720 	struct e1000_adapter *adapter = rx_ring->adapter;
721 	struct net_device *netdev = adapter->netdev;
722 	struct pci_dev *pdev = adapter->pdev;
723 	union e1000_rx_desc_packet_split *rx_desc;
724 	struct e1000_buffer *buffer_info;
725 	struct e1000_ps_page *ps_page;
726 	struct sk_buff *skb;
727 	unsigned int i, j;
728 
729 	i = rx_ring->next_to_use;
730 	buffer_info = &rx_ring->buffer_info[i];
731 
732 	while (cleaned_count--) {
733 		rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
734 
735 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
736 			ps_page = &buffer_info->ps_pages[j];
737 			if (j >= adapter->rx_ps_pages) {
738 				/* all unused desc entries get hw null ptr */
739 				rx_desc->read.buffer_addr[j + 1] =
740 				    ~cpu_to_le64(0);
741 				continue;
742 			}
743 			if (!ps_page->page) {
744 				ps_page->page = alloc_page(gfp);
745 				if (!ps_page->page) {
746 					adapter->alloc_rx_buff_failed++;
747 					goto no_buffers;
748 				}
749 				ps_page->dma = dma_map_page(&pdev->dev,
750 							    ps_page->page,
751 							    0, PAGE_SIZE,
752 							    DMA_FROM_DEVICE);
753 				if (dma_mapping_error(&pdev->dev,
754 						      ps_page->dma)) {
755 					dev_err(&adapter->pdev->dev,
756 						"Rx DMA page map failed\n");
757 					adapter->rx_dma_failed++;
758 					goto no_buffers;
759 				}
760 			}
761 			/* Refresh the desc even if buffer_addrs
762 			 * didn't change because each write-back
763 			 * erases this info.
764 			 */
765 			rx_desc->read.buffer_addr[j + 1] =
766 			    cpu_to_le64(ps_page->dma);
767 		}
768 
769 		skb = __netdev_alloc_skb_ip_align(netdev, adapter->rx_ps_bsize0,
770 						  gfp);
771 
772 		if (!skb) {
773 			adapter->alloc_rx_buff_failed++;
774 			break;
775 		}
776 
777 		buffer_info->skb = skb;
778 		buffer_info->dma = dma_map_single(&pdev->dev, skb->data,
779 						  adapter->rx_ps_bsize0,
780 						  DMA_FROM_DEVICE);
781 		if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
782 			dev_err(&pdev->dev, "Rx DMA map failed\n");
783 			adapter->rx_dma_failed++;
784 			/* cleanup skb */
785 			dev_kfree_skb_any(skb);
786 			buffer_info->skb = NULL;
787 			break;
788 		}
789 
790 		rx_desc->read.buffer_addr[0] = cpu_to_le64(buffer_info->dma);
791 
792 		if (unlikely(!(i & (E1000_RX_BUFFER_WRITE - 1)))) {
793 			/* Force memory writes to complete before letting h/w
794 			 * know there are new descriptors to fetch.  (Only
795 			 * applicable for weak-ordered memory model archs,
796 			 * such as IA-64).
797 			 */
798 			wmb();
799 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
800 				e1000e_update_rdt_wa(rx_ring, i << 1);
801 			else
802 				writel(i << 1, rx_ring->tail);
803 		}
804 
805 		i++;
806 		if (i == rx_ring->count)
807 			i = 0;
808 		buffer_info = &rx_ring->buffer_info[i];
809 	}
810 
811 no_buffers:
812 	rx_ring->next_to_use = i;
813 }
814 
815 /**
816  * e1000_alloc_jumbo_rx_buffers - Replace used jumbo receive buffers
817  * @rx_ring: Rx descriptor ring
818  * @cleaned_count: number of buffers to allocate this pass
819  * @gfp: flags for allocation
820  **/
821 
e1000_alloc_jumbo_rx_buffers(struct e1000_ring * rx_ring,int cleaned_count,gfp_t gfp)822 static void e1000_alloc_jumbo_rx_buffers(struct e1000_ring *rx_ring,
823 					 int cleaned_count, gfp_t gfp)
824 {
825 	struct e1000_adapter *adapter = rx_ring->adapter;
826 	struct net_device *netdev = adapter->netdev;
827 	struct pci_dev *pdev = adapter->pdev;
828 	union e1000_rx_desc_extended *rx_desc;
829 	struct e1000_buffer *buffer_info;
830 	struct sk_buff *skb;
831 	unsigned int i;
832 	unsigned int bufsz = 256 - 16;	/* for skb_reserve */
833 
834 	i = rx_ring->next_to_use;
835 	buffer_info = &rx_ring->buffer_info[i];
836 
837 	while (cleaned_count--) {
838 		skb = buffer_info->skb;
839 		if (skb) {
840 			skb_trim(skb, 0);
841 			goto check_page;
842 		}
843 
844 		skb = __netdev_alloc_skb_ip_align(netdev, bufsz, gfp);
845 		if (unlikely(!skb)) {
846 			/* Better luck next round */
847 			adapter->alloc_rx_buff_failed++;
848 			break;
849 		}
850 
851 		buffer_info->skb = skb;
852 check_page:
853 		/* allocate a new page if necessary */
854 		if (!buffer_info->page) {
855 			buffer_info->page = alloc_page(gfp);
856 			if (unlikely(!buffer_info->page)) {
857 				adapter->alloc_rx_buff_failed++;
858 				break;
859 			}
860 		}
861 
862 		if (!buffer_info->dma) {
863 			buffer_info->dma = dma_map_page(&pdev->dev,
864 							buffer_info->page, 0,
865 							PAGE_SIZE,
866 							DMA_FROM_DEVICE);
867 			if (dma_mapping_error(&pdev->dev, buffer_info->dma)) {
868 				adapter->alloc_rx_buff_failed++;
869 				break;
870 			}
871 		}
872 
873 		rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
874 		rx_desc->read.buffer_addr = cpu_to_le64(buffer_info->dma);
875 
876 		if (unlikely(++i == rx_ring->count))
877 			i = 0;
878 		buffer_info = &rx_ring->buffer_info[i];
879 	}
880 
881 	if (likely(rx_ring->next_to_use != i)) {
882 		rx_ring->next_to_use = i;
883 		if (unlikely(i-- == 0))
884 			i = (rx_ring->count - 1);
885 
886 		/* Force memory writes to complete before letting h/w
887 		 * know there are new descriptors to fetch.  (Only
888 		 * applicable for weak-ordered memory model archs,
889 		 * such as IA-64).
890 		 */
891 		wmb();
892 		if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
893 			e1000e_update_rdt_wa(rx_ring, i);
894 		else
895 			writel(i, rx_ring->tail);
896 	}
897 }
898 
e1000_rx_hash(struct net_device * netdev,__le32 rss,struct sk_buff * skb)899 static inline void e1000_rx_hash(struct net_device *netdev, __le32 rss,
900 				 struct sk_buff *skb)
901 {
902 	if (netdev->features & NETIF_F_RXHASH)
903 		skb_set_hash(skb, le32_to_cpu(rss), PKT_HASH_TYPE_L3);
904 }
905 
906 /**
907  * e1000_clean_rx_irq - Send received data up the network stack
908  * @rx_ring: Rx descriptor ring
909  * @work_done: output parameter for indicating completed work
910  * @work_to_do: how many packets we can clean
911  *
912  * the return value indicates whether actual cleaning was done, there
913  * is no guarantee that everything was cleaned
914  **/
e1000_clean_rx_irq(struct e1000_ring * rx_ring,int * work_done,int work_to_do)915 static bool e1000_clean_rx_irq(struct e1000_ring *rx_ring, int *work_done,
916 			       int work_to_do)
917 {
918 	struct e1000_adapter *adapter = rx_ring->adapter;
919 	struct net_device *netdev = adapter->netdev;
920 	struct pci_dev *pdev = adapter->pdev;
921 	struct e1000_hw *hw = &adapter->hw;
922 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
923 	struct e1000_buffer *buffer_info, *next_buffer;
924 	u32 length, staterr;
925 	unsigned int i;
926 	int cleaned_count = 0;
927 	bool cleaned = false;
928 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
929 
930 	i = rx_ring->next_to_clean;
931 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
932 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
933 	buffer_info = &rx_ring->buffer_info[i];
934 
935 	while (staterr & E1000_RXD_STAT_DD) {
936 		struct sk_buff *skb;
937 
938 		if (*work_done >= work_to_do)
939 			break;
940 		(*work_done)++;
941 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
942 
943 		skb = buffer_info->skb;
944 		buffer_info->skb = NULL;
945 
946 		prefetch(skb->data - NET_IP_ALIGN);
947 
948 		i++;
949 		if (i == rx_ring->count)
950 			i = 0;
951 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
952 		prefetch(next_rxd);
953 
954 		next_buffer = &rx_ring->buffer_info[i];
955 
956 		cleaned = true;
957 		cleaned_count++;
958 		dma_unmap_single(&pdev->dev, buffer_info->dma,
959 				 adapter->rx_buffer_len, DMA_FROM_DEVICE);
960 		buffer_info->dma = 0;
961 
962 		length = le16_to_cpu(rx_desc->wb.upper.length);
963 
964 		/* !EOP means multiple descriptors were used to store a single
965 		 * packet, if that's the case we need to toss it.  In fact, we
966 		 * need to toss every packet with the EOP bit clear and the
967 		 * next frame that _does_ have the EOP bit set, as it is by
968 		 * definition only a frame fragment
969 		 */
970 		if (unlikely(!(staterr & E1000_RXD_STAT_EOP)))
971 			adapter->flags2 |= FLAG2_IS_DISCARDING;
972 
973 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
974 			/* All receives must fit into a single buffer */
975 			e_dbg("Receive packet consumed multiple buffers\n");
976 			/* recycle */
977 			buffer_info->skb = skb;
978 			if (staterr & E1000_RXD_STAT_EOP)
979 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
980 			goto next_desc;
981 		}
982 
983 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
984 			     !(netdev->features & NETIF_F_RXALL))) {
985 			/* recycle */
986 			buffer_info->skb = skb;
987 			goto next_desc;
988 		}
989 
990 		/* adjust length to remove Ethernet CRC */
991 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
992 			/* If configured to store CRC, don't subtract FCS,
993 			 * but keep the FCS bytes out of the total_rx_bytes
994 			 * counter
995 			 */
996 			if (netdev->features & NETIF_F_RXFCS)
997 				total_rx_bytes -= 4;
998 			else
999 				length -= 4;
1000 		}
1001 
1002 		total_rx_bytes += length;
1003 		total_rx_packets++;
1004 
1005 		/* code added for copybreak, this should improve
1006 		 * performance for small packets with large amounts
1007 		 * of reassembly being done in the stack
1008 		 */
1009 		if (length < copybreak) {
1010 			struct sk_buff *new_skb =
1011 				napi_alloc_skb(&adapter->napi, length);
1012 			if (new_skb) {
1013 				skb_copy_to_linear_data_offset(new_skb,
1014 							       -NET_IP_ALIGN,
1015 							       (skb->data -
1016 								NET_IP_ALIGN),
1017 							       (length +
1018 								NET_IP_ALIGN));
1019 				/* save the skb in buffer_info as good */
1020 				buffer_info->skb = skb;
1021 				skb = new_skb;
1022 			}
1023 			/* else just continue with the old one */
1024 		}
1025 		/* end copybreak code */
1026 		skb_put(skb, length);
1027 
1028 		/* Receive Checksum Offload */
1029 		e1000_rx_checksum(adapter, staterr, skb);
1030 
1031 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1032 
1033 		e1000_receive_skb(adapter, netdev, skb, staterr,
1034 				  rx_desc->wb.upper.vlan);
1035 
1036 next_desc:
1037 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1038 
1039 		/* return some buffers to hardware, one at a time is too slow */
1040 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1041 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1042 					      GFP_ATOMIC);
1043 			cleaned_count = 0;
1044 		}
1045 
1046 		/* use prefetched values */
1047 		rx_desc = next_rxd;
1048 		buffer_info = next_buffer;
1049 
1050 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1051 	}
1052 	rx_ring->next_to_clean = i;
1053 
1054 	cleaned_count = e1000_desc_unused(rx_ring);
1055 	if (cleaned_count)
1056 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1057 
1058 	adapter->total_rx_bytes += total_rx_bytes;
1059 	adapter->total_rx_packets += total_rx_packets;
1060 	return cleaned;
1061 }
1062 
e1000_put_txbuf(struct e1000_ring * tx_ring,struct e1000_buffer * buffer_info,bool drop)1063 static void e1000_put_txbuf(struct e1000_ring *tx_ring,
1064 			    struct e1000_buffer *buffer_info,
1065 			    bool drop)
1066 {
1067 	struct e1000_adapter *adapter = tx_ring->adapter;
1068 
1069 	if (buffer_info->dma) {
1070 		if (buffer_info->mapped_as_page)
1071 			dma_unmap_page(&adapter->pdev->dev, buffer_info->dma,
1072 				       buffer_info->length, DMA_TO_DEVICE);
1073 		else
1074 			dma_unmap_single(&adapter->pdev->dev, buffer_info->dma,
1075 					 buffer_info->length, DMA_TO_DEVICE);
1076 		buffer_info->dma = 0;
1077 	}
1078 	if (buffer_info->skb) {
1079 		if (drop)
1080 			dev_kfree_skb_any(buffer_info->skb);
1081 		else
1082 			dev_consume_skb_any(buffer_info->skb);
1083 		buffer_info->skb = NULL;
1084 	}
1085 	buffer_info->time_stamp = 0;
1086 }
1087 
e1000_print_hw_hang(struct work_struct * work)1088 static void e1000_print_hw_hang(struct work_struct *work)
1089 {
1090 	struct e1000_adapter *adapter = container_of(work,
1091 						     struct e1000_adapter,
1092 						     print_hang_task);
1093 	struct net_device *netdev = adapter->netdev;
1094 	struct e1000_ring *tx_ring = adapter->tx_ring;
1095 	unsigned int i = tx_ring->next_to_clean;
1096 	unsigned int eop = tx_ring->buffer_info[i].next_to_watch;
1097 	struct e1000_tx_desc *eop_desc = E1000_TX_DESC(*tx_ring, eop);
1098 	struct e1000_hw *hw = &adapter->hw;
1099 	u16 phy_status, phy_1000t_status, phy_ext_status;
1100 	u16 pci_status;
1101 
1102 	if (test_bit(__E1000_DOWN, &adapter->state))
1103 		return;
1104 
1105 	if (!adapter->tx_hang_recheck && (adapter->flags2 & FLAG2_DMA_BURST)) {
1106 		/* May be block on write-back, flush and detect again
1107 		 * flush pending descriptor writebacks to memory
1108 		 */
1109 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1110 		/* execute the writes immediately */
1111 		e1e_flush();
1112 		/* Due to rare timing issues, write to TIDV again to ensure
1113 		 * the write is successful
1114 		 */
1115 		ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
1116 		/* execute the writes immediately */
1117 		e1e_flush();
1118 		adapter->tx_hang_recheck = true;
1119 		return;
1120 	}
1121 	adapter->tx_hang_recheck = false;
1122 
1123 	if (er32(TDH(0)) == er32(TDT(0))) {
1124 		e_dbg("false hang detected, ignoring\n");
1125 		return;
1126 	}
1127 
1128 	/* Real hang detected */
1129 	netif_stop_queue(netdev);
1130 
1131 	e1e_rphy(hw, MII_BMSR, &phy_status);
1132 	e1e_rphy(hw, MII_STAT1000, &phy_1000t_status);
1133 	e1e_rphy(hw, MII_ESTATUS, &phy_ext_status);
1134 
1135 	pci_read_config_word(adapter->pdev, PCI_STATUS, &pci_status);
1136 
1137 	/* detected Hardware unit hang */
1138 	e_err("Detected Hardware Unit Hang:\n"
1139 	      "  TDH                  <%x>\n"
1140 	      "  TDT                  <%x>\n"
1141 	      "  next_to_use          <%x>\n"
1142 	      "  next_to_clean        <%x>\n"
1143 	      "buffer_info[next_to_clean]:\n"
1144 	      "  time_stamp           <%lx>\n"
1145 	      "  next_to_watch        <%x>\n"
1146 	      "  jiffies              <%lx>\n"
1147 	      "  next_to_watch.status <%x>\n"
1148 	      "MAC Status             <%x>\n"
1149 	      "PHY Status             <%x>\n"
1150 	      "PHY 1000BASE-T Status  <%x>\n"
1151 	      "PHY Extended Status    <%x>\n"
1152 	      "PCI Status             <%x>\n",
1153 	      readl(tx_ring->head), readl(tx_ring->tail), tx_ring->next_to_use,
1154 	      tx_ring->next_to_clean, tx_ring->buffer_info[eop].time_stamp,
1155 	      eop, jiffies, eop_desc->upper.fields.status, er32(STATUS),
1156 	      phy_status, phy_1000t_status, phy_ext_status, pci_status);
1157 
1158 	e1000e_dump(adapter);
1159 
1160 	/* Suggest workaround for known h/w issue */
1161 	if ((hw->mac.type == e1000_pchlan) && (er32(CTRL) & E1000_CTRL_TFCE))
1162 		e_err("Try turning off Tx pause (flow control) via ethtool\n");
1163 }
1164 
1165 /**
1166  * e1000e_tx_hwtstamp_work - check for Tx time stamp
1167  * @work: pointer to work struct
1168  *
1169  * This work function polls the TSYNCTXCTL valid bit to determine when a
1170  * timestamp has been taken for the current stored skb.  The timestamp must
1171  * be for this skb because only one such packet is allowed in the queue.
1172  */
e1000e_tx_hwtstamp_work(struct work_struct * work)1173 static void e1000e_tx_hwtstamp_work(struct work_struct *work)
1174 {
1175 	struct e1000_adapter *adapter = container_of(work, struct e1000_adapter,
1176 						     tx_hwtstamp_work);
1177 	struct e1000_hw *hw = &adapter->hw;
1178 
1179 	if (er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_VALID) {
1180 		struct sk_buff *skb = adapter->tx_hwtstamp_skb;
1181 		struct skb_shared_hwtstamps shhwtstamps;
1182 		u64 txstmp;
1183 
1184 		txstmp = er32(TXSTMPL);
1185 		txstmp |= (u64)er32(TXSTMPH) << 32;
1186 
1187 		e1000e_systim_to_hwtstamp(adapter, &shhwtstamps, txstmp);
1188 
1189 		/* Clear the global tx_hwtstamp_skb pointer and force writes
1190 		 * prior to notifying the stack of a Tx timestamp.
1191 		 */
1192 		adapter->tx_hwtstamp_skb = NULL;
1193 		wmb(); /* force write prior to skb_tstamp_tx */
1194 
1195 		skb_tstamp_tx(skb, &shhwtstamps);
1196 		dev_consume_skb_any(skb);
1197 	} else if (time_after(jiffies, adapter->tx_hwtstamp_start
1198 			      + adapter->tx_timeout_factor * HZ)) {
1199 		dev_kfree_skb_any(adapter->tx_hwtstamp_skb);
1200 		adapter->tx_hwtstamp_skb = NULL;
1201 		adapter->tx_hwtstamp_timeouts++;
1202 		e_warn("clearing Tx timestamp hang\n");
1203 	} else {
1204 		/* reschedule to check later */
1205 		schedule_work(&adapter->tx_hwtstamp_work);
1206 	}
1207 }
1208 
1209 /**
1210  * e1000_clean_tx_irq - Reclaim resources after transmit completes
1211  * @tx_ring: Tx descriptor ring
1212  *
1213  * the return value indicates whether actual cleaning was done, there
1214  * is no guarantee that everything was cleaned
1215  **/
e1000_clean_tx_irq(struct e1000_ring * tx_ring)1216 static bool e1000_clean_tx_irq(struct e1000_ring *tx_ring)
1217 {
1218 	struct e1000_adapter *adapter = tx_ring->adapter;
1219 	struct net_device *netdev = adapter->netdev;
1220 	struct e1000_hw *hw = &adapter->hw;
1221 	struct e1000_tx_desc *tx_desc, *eop_desc;
1222 	struct e1000_buffer *buffer_info;
1223 	unsigned int i, eop;
1224 	unsigned int count = 0;
1225 	unsigned int total_tx_bytes = 0, total_tx_packets = 0;
1226 	unsigned int bytes_compl = 0, pkts_compl = 0;
1227 
1228 	i = tx_ring->next_to_clean;
1229 	eop = tx_ring->buffer_info[i].next_to_watch;
1230 	eop_desc = E1000_TX_DESC(*tx_ring, eop);
1231 
1232 	while ((eop_desc->upper.data & cpu_to_le32(E1000_TXD_STAT_DD)) &&
1233 	       (count < tx_ring->count)) {
1234 		bool cleaned = false;
1235 
1236 		dma_rmb();		/* read buffer_info after eop_desc */
1237 		for (; !cleaned; count++) {
1238 			tx_desc = E1000_TX_DESC(*tx_ring, i);
1239 			buffer_info = &tx_ring->buffer_info[i];
1240 			cleaned = (i == eop);
1241 
1242 			if (cleaned) {
1243 				total_tx_packets += buffer_info->segs;
1244 				total_tx_bytes += buffer_info->bytecount;
1245 				if (buffer_info->skb) {
1246 					bytes_compl += buffer_info->skb->len;
1247 					pkts_compl++;
1248 				}
1249 			}
1250 
1251 			e1000_put_txbuf(tx_ring, buffer_info, false);
1252 			tx_desc->upper.data = 0;
1253 
1254 			i++;
1255 			if (i == tx_ring->count)
1256 				i = 0;
1257 		}
1258 
1259 		if (i == tx_ring->next_to_use)
1260 			break;
1261 		eop = tx_ring->buffer_info[i].next_to_watch;
1262 		eop_desc = E1000_TX_DESC(*tx_ring, eop);
1263 	}
1264 
1265 	tx_ring->next_to_clean = i;
1266 
1267 	netdev_completed_queue(netdev, pkts_compl, bytes_compl);
1268 
1269 #define TX_WAKE_THRESHOLD 32
1270 	if (count && netif_carrier_ok(netdev) &&
1271 	    e1000_desc_unused(tx_ring) >= TX_WAKE_THRESHOLD) {
1272 		/* Make sure that anybody stopping the queue after this
1273 		 * sees the new next_to_clean.
1274 		 */
1275 		smp_mb();
1276 
1277 		if (netif_queue_stopped(netdev) &&
1278 		    !(test_bit(__E1000_DOWN, &adapter->state))) {
1279 			netif_wake_queue(netdev);
1280 			++adapter->restart_queue;
1281 		}
1282 	}
1283 
1284 	if (adapter->detect_tx_hung) {
1285 		/* Detect a transmit hang in hardware, this serializes the
1286 		 * check with the clearing of time_stamp and movement of i
1287 		 */
1288 		adapter->detect_tx_hung = false;
1289 		if (tx_ring->buffer_info[i].time_stamp &&
1290 		    time_after(jiffies, tx_ring->buffer_info[i].time_stamp
1291 			       + (adapter->tx_timeout_factor * HZ)) &&
1292 		    !(er32(STATUS) & E1000_STATUS_TXOFF))
1293 			schedule_work(&adapter->print_hang_task);
1294 		else
1295 			adapter->tx_hang_recheck = false;
1296 	}
1297 	adapter->total_tx_bytes += total_tx_bytes;
1298 	adapter->total_tx_packets += total_tx_packets;
1299 	return count < tx_ring->count;
1300 }
1301 
1302 /**
1303  * e1000_clean_rx_irq_ps - Send received data up the network stack; packet split
1304  * @rx_ring: Rx descriptor ring
1305  * @work_done: output parameter for indicating completed work
1306  * @work_to_do: how many packets we can clean
1307  *
1308  * the return value indicates whether actual cleaning was done, there
1309  * is no guarantee that everything was cleaned
1310  **/
e1000_clean_rx_irq_ps(struct e1000_ring * rx_ring,int * work_done,int work_to_do)1311 static bool e1000_clean_rx_irq_ps(struct e1000_ring *rx_ring, int *work_done,
1312 				  int work_to_do)
1313 {
1314 	struct e1000_adapter *adapter = rx_ring->adapter;
1315 	struct e1000_hw *hw = &adapter->hw;
1316 	union e1000_rx_desc_packet_split *rx_desc, *next_rxd;
1317 	struct net_device *netdev = adapter->netdev;
1318 	struct pci_dev *pdev = adapter->pdev;
1319 	struct e1000_buffer *buffer_info, *next_buffer;
1320 	struct e1000_ps_page *ps_page;
1321 	struct sk_buff *skb;
1322 	unsigned int i, j;
1323 	u32 length, staterr;
1324 	int cleaned_count = 0;
1325 	bool cleaned = false;
1326 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1327 
1328 	i = rx_ring->next_to_clean;
1329 	rx_desc = E1000_RX_DESC_PS(*rx_ring, i);
1330 	staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1331 	buffer_info = &rx_ring->buffer_info[i];
1332 
1333 	while (staterr & E1000_RXD_STAT_DD) {
1334 		if (*work_done >= work_to_do)
1335 			break;
1336 		(*work_done)++;
1337 		skb = buffer_info->skb;
1338 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1339 
1340 		/* in the packet split case this is header only */
1341 		prefetch(skb->data - NET_IP_ALIGN);
1342 
1343 		i++;
1344 		if (i == rx_ring->count)
1345 			i = 0;
1346 		next_rxd = E1000_RX_DESC_PS(*rx_ring, i);
1347 		prefetch(next_rxd);
1348 
1349 		next_buffer = &rx_ring->buffer_info[i];
1350 
1351 		cleaned = true;
1352 		cleaned_count++;
1353 		dma_unmap_single(&pdev->dev, buffer_info->dma,
1354 				 adapter->rx_ps_bsize0, DMA_FROM_DEVICE);
1355 		buffer_info->dma = 0;
1356 
1357 		/* see !EOP comment in other Rx routine */
1358 		if (!(staterr & E1000_RXD_STAT_EOP))
1359 			adapter->flags2 |= FLAG2_IS_DISCARDING;
1360 
1361 		if (adapter->flags2 & FLAG2_IS_DISCARDING) {
1362 			e_dbg("Packet Split buffers didn't pick up the full packet\n");
1363 			dev_kfree_skb_irq(skb);
1364 			if (staterr & E1000_RXD_STAT_EOP)
1365 				adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1366 			goto next_desc;
1367 		}
1368 
1369 		if (unlikely((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1370 			     !(netdev->features & NETIF_F_RXALL))) {
1371 			dev_kfree_skb_irq(skb);
1372 			goto next_desc;
1373 		}
1374 
1375 		length = le16_to_cpu(rx_desc->wb.middle.length0);
1376 
1377 		if (!length) {
1378 			e_dbg("Last part of the packet spanning multiple descriptors\n");
1379 			dev_kfree_skb_irq(skb);
1380 			goto next_desc;
1381 		}
1382 
1383 		/* Good Receive */
1384 		skb_put(skb, length);
1385 
1386 		{
1387 			/* this looks ugly, but it seems compiler issues make
1388 			 * it more efficient than reusing j
1389 			 */
1390 			int l1 = le16_to_cpu(rx_desc->wb.upper.length[0]);
1391 
1392 			/* page alloc/put takes too long and effects small
1393 			 * packet throughput, so unsplit small packets and
1394 			 * save the alloc/put
1395 			 */
1396 			if (l1 && (l1 <= copybreak) &&
1397 			    ((length + l1) <= adapter->rx_ps_bsize0)) {
1398 				ps_page = &buffer_info->ps_pages[0];
1399 
1400 				dma_sync_single_for_cpu(&pdev->dev,
1401 							ps_page->dma,
1402 							PAGE_SIZE,
1403 							DMA_FROM_DEVICE);
1404 				memcpy(skb_tail_pointer(skb),
1405 				       page_address(ps_page->page), l1);
1406 				dma_sync_single_for_device(&pdev->dev,
1407 							   ps_page->dma,
1408 							   PAGE_SIZE,
1409 							   DMA_FROM_DEVICE);
1410 
1411 				/* remove the CRC */
1412 				if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1413 					if (!(netdev->features & NETIF_F_RXFCS))
1414 						l1 -= 4;
1415 				}
1416 
1417 				skb_put(skb, l1);
1418 				goto copydone;
1419 			}	/* if */
1420 		}
1421 
1422 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1423 			length = le16_to_cpu(rx_desc->wb.upper.length[j]);
1424 			if (!length)
1425 				break;
1426 
1427 			ps_page = &buffer_info->ps_pages[j];
1428 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1429 				       DMA_FROM_DEVICE);
1430 			ps_page->dma = 0;
1431 			skb_fill_page_desc(skb, j, ps_page->page, 0, length);
1432 			ps_page->page = NULL;
1433 			skb->len += length;
1434 			skb->data_len += length;
1435 			skb->truesize += PAGE_SIZE;
1436 		}
1437 
1438 		/* strip the ethernet crc, problem is we're using pages now so
1439 		 * this whole operation can get a little cpu intensive
1440 		 */
1441 		if (!(adapter->flags2 & FLAG2_CRC_STRIPPING)) {
1442 			if (!(netdev->features & NETIF_F_RXFCS))
1443 				pskb_trim(skb, skb->len - 4);
1444 		}
1445 
1446 copydone:
1447 		total_rx_bytes += skb->len;
1448 		total_rx_packets++;
1449 
1450 		e1000_rx_checksum(adapter, staterr, skb);
1451 
1452 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1453 
1454 		if (rx_desc->wb.upper.header_status &
1455 		    cpu_to_le16(E1000_RXDPS_HDRSTAT_HDRSP))
1456 			adapter->rx_hdr_split++;
1457 
1458 		e1000_receive_skb(adapter, netdev, skb, staterr,
1459 				  rx_desc->wb.middle.vlan);
1460 
1461 next_desc:
1462 		rx_desc->wb.middle.status_error &= cpu_to_le32(~0xFF);
1463 		buffer_info->skb = NULL;
1464 
1465 		/* return some buffers to hardware, one at a time is too slow */
1466 		if (cleaned_count >= E1000_RX_BUFFER_WRITE) {
1467 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1468 					      GFP_ATOMIC);
1469 			cleaned_count = 0;
1470 		}
1471 
1472 		/* use prefetched values */
1473 		rx_desc = next_rxd;
1474 		buffer_info = next_buffer;
1475 
1476 		staterr = le32_to_cpu(rx_desc->wb.middle.status_error);
1477 	}
1478 	rx_ring->next_to_clean = i;
1479 
1480 	cleaned_count = e1000_desc_unused(rx_ring);
1481 	if (cleaned_count)
1482 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1483 
1484 	adapter->total_rx_bytes += total_rx_bytes;
1485 	adapter->total_rx_packets += total_rx_packets;
1486 	return cleaned;
1487 }
1488 
e1000_consume_page(struct e1000_buffer * bi,struct sk_buff * skb,u16 length)1489 static void e1000_consume_page(struct e1000_buffer *bi, struct sk_buff *skb,
1490 			       u16 length)
1491 {
1492 	bi->page = NULL;
1493 	skb->len += length;
1494 	skb->data_len += length;
1495 	skb->truesize += PAGE_SIZE;
1496 }
1497 
1498 /**
1499  * e1000_clean_jumbo_rx_irq - Send received data up the network stack; legacy
1500  * @rx_ring: Rx descriptor ring
1501  * @work_done: output parameter for indicating completed work
1502  * @work_to_do: how many packets we can clean
1503  *
1504  * the return value indicates whether actual cleaning was done, there
1505  * is no guarantee that everything was cleaned
1506  **/
e1000_clean_jumbo_rx_irq(struct e1000_ring * rx_ring,int * work_done,int work_to_do)1507 static bool e1000_clean_jumbo_rx_irq(struct e1000_ring *rx_ring, int *work_done,
1508 				     int work_to_do)
1509 {
1510 	struct e1000_adapter *adapter = rx_ring->adapter;
1511 	struct net_device *netdev = adapter->netdev;
1512 	struct pci_dev *pdev = adapter->pdev;
1513 	union e1000_rx_desc_extended *rx_desc, *next_rxd;
1514 	struct e1000_buffer *buffer_info, *next_buffer;
1515 	u32 length, staterr;
1516 	unsigned int i;
1517 	int cleaned_count = 0;
1518 	bool cleaned = false;
1519 	unsigned int total_rx_bytes = 0, total_rx_packets = 0;
1520 	struct skb_shared_info *shinfo;
1521 
1522 	i = rx_ring->next_to_clean;
1523 	rx_desc = E1000_RX_DESC_EXT(*rx_ring, i);
1524 	staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1525 	buffer_info = &rx_ring->buffer_info[i];
1526 
1527 	while (staterr & E1000_RXD_STAT_DD) {
1528 		struct sk_buff *skb;
1529 
1530 		if (*work_done >= work_to_do)
1531 			break;
1532 		(*work_done)++;
1533 		dma_rmb();	/* read descriptor and rx_buffer_info after status DD */
1534 
1535 		skb = buffer_info->skb;
1536 		buffer_info->skb = NULL;
1537 
1538 		++i;
1539 		if (i == rx_ring->count)
1540 			i = 0;
1541 		next_rxd = E1000_RX_DESC_EXT(*rx_ring, i);
1542 		prefetch(next_rxd);
1543 
1544 		next_buffer = &rx_ring->buffer_info[i];
1545 
1546 		cleaned = true;
1547 		cleaned_count++;
1548 		dma_unmap_page(&pdev->dev, buffer_info->dma, PAGE_SIZE,
1549 			       DMA_FROM_DEVICE);
1550 		buffer_info->dma = 0;
1551 
1552 		length = le16_to_cpu(rx_desc->wb.upper.length);
1553 
1554 		/* errors is only valid for DD + EOP descriptors */
1555 		if (unlikely((staterr & E1000_RXD_STAT_EOP) &&
1556 			     ((staterr & E1000_RXDEXT_ERR_FRAME_ERR_MASK) &&
1557 			      !(netdev->features & NETIF_F_RXALL)))) {
1558 			/* recycle both page and skb */
1559 			buffer_info->skb = skb;
1560 			/* an error means any chain goes out the window too */
1561 			if (rx_ring->rx_skb_top)
1562 				dev_kfree_skb_irq(rx_ring->rx_skb_top);
1563 			rx_ring->rx_skb_top = NULL;
1564 			goto next_desc;
1565 		}
1566 #define rxtop (rx_ring->rx_skb_top)
1567 		if (!(staterr & E1000_RXD_STAT_EOP)) {
1568 			/* this descriptor is only the beginning (or middle) */
1569 			if (!rxtop) {
1570 				/* this is the beginning of a chain */
1571 				rxtop = skb;
1572 				skb_fill_page_desc(rxtop, 0, buffer_info->page,
1573 						   0, length);
1574 			} else {
1575 				/* this is the middle of a chain */
1576 				shinfo = skb_shinfo(rxtop);
1577 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1578 						   buffer_info->page, 0,
1579 						   length);
1580 				/* re-use the skb, only consumed the page */
1581 				buffer_info->skb = skb;
1582 			}
1583 			e1000_consume_page(buffer_info, rxtop, length);
1584 			goto next_desc;
1585 		} else {
1586 			if (rxtop) {
1587 				/* end of the chain */
1588 				shinfo = skb_shinfo(rxtop);
1589 				skb_fill_page_desc(rxtop, shinfo->nr_frags,
1590 						   buffer_info->page, 0,
1591 						   length);
1592 				/* re-use the current skb, we only consumed the
1593 				 * page
1594 				 */
1595 				buffer_info->skb = skb;
1596 				skb = rxtop;
1597 				rxtop = NULL;
1598 				e1000_consume_page(buffer_info, skb, length);
1599 			} else {
1600 				/* no chain, got EOP, this buf is the packet
1601 				 * copybreak to save the put_page/alloc_page
1602 				 */
1603 				if (length <= copybreak &&
1604 				    skb_tailroom(skb) >= length) {
1605 					memcpy(skb_tail_pointer(skb),
1606 					       page_address(buffer_info->page),
1607 					       length);
1608 					/* re-use the page, so don't erase
1609 					 * buffer_info->page
1610 					 */
1611 					skb_put(skb, length);
1612 				} else {
1613 					skb_fill_page_desc(skb, 0,
1614 							   buffer_info->page, 0,
1615 							   length);
1616 					e1000_consume_page(buffer_info, skb,
1617 							   length);
1618 				}
1619 			}
1620 		}
1621 
1622 		/* Receive Checksum Offload */
1623 		e1000_rx_checksum(adapter, staterr, skb);
1624 
1625 		e1000_rx_hash(netdev, rx_desc->wb.lower.hi_dword.rss, skb);
1626 
1627 		/* probably a little skewed due to removing CRC */
1628 		total_rx_bytes += skb->len;
1629 		total_rx_packets++;
1630 
1631 		/* eth type trans needs skb->data to point to something */
1632 		if (!pskb_may_pull(skb, ETH_HLEN)) {
1633 			e_err("pskb_may_pull failed.\n");
1634 			dev_kfree_skb_irq(skb);
1635 			goto next_desc;
1636 		}
1637 
1638 		e1000_receive_skb(adapter, netdev, skb, staterr,
1639 				  rx_desc->wb.upper.vlan);
1640 
1641 next_desc:
1642 		rx_desc->wb.upper.status_error &= cpu_to_le32(~0xFF);
1643 
1644 		/* return some buffers to hardware, one at a time is too slow */
1645 		if (unlikely(cleaned_count >= E1000_RX_BUFFER_WRITE)) {
1646 			adapter->alloc_rx_buf(rx_ring, cleaned_count,
1647 					      GFP_ATOMIC);
1648 			cleaned_count = 0;
1649 		}
1650 
1651 		/* use prefetched values */
1652 		rx_desc = next_rxd;
1653 		buffer_info = next_buffer;
1654 
1655 		staterr = le32_to_cpu(rx_desc->wb.upper.status_error);
1656 	}
1657 	rx_ring->next_to_clean = i;
1658 
1659 	cleaned_count = e1000_desc_unused(rx_ring);
1660 	if (cleaned_count)
1661 		adapter->alloc_rx_buf(rx_ring, cleaned_count, GFP_ATOMIC);
1662 
1663 	adapter->total_rx_bytes += total_rx_bytes;
1664 	adapter->total_rx_packets += total_rx_packets;
1665 	return cleaned;
1666 }
1667 
1668 /**
1669  * e1000_clean_rx_ring - Free Rx Buffers per Queue
1670  * @rx_ring: Rx descriptor ring
1671  **/
e1000_clean_rx_ring(struct e1000_ring * rx_ring)1672 static void e1000_clean_rx_ring(struct e1000_ring *rx_ring)
1673 {
1674 	struct e1000_adapter *adapter = rx_ring->adapter;
1675 	struct e1000_buffer *buffer_info;
1676 	struct e1000_ps_page *ps_page;
1677 	struct pci_dev *pdev = adapter->pdev;
1678 	unsigned int i, j;
1679 
1680 	/* Free all the Rx ring sk_buffs */
1681 	for (i = 0; i < rx_ring->count; i++) {
1682 		buffer_info = &rx_ring->buffer_info[i];
1683 		if (buffer_info->dma) {
1684 			if (adapter->clean_rx == e1000_clean_rx_irq)
1685 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1686 						 adapter->rx_buffer_len,
1687 						 DMA_FROM_DEVICE);
1688 			else if (adapter->clean_rx == e1000_clean_jumbo_rx_irq)
1689 				dma_unmap_page(&pdev->dev, buffer_info->dma,
1690 					       PAGE_SIZE, DMA_FROM_DEVICE);
1691 			else if (adapter->clean_rx == e1000_clean_rx_irq_ps)
1692 				dma_unmap_single(&pdev->dev, buffer_info->dma,
1693 						 adapter->rx_ps_bsize0,
1694 						 DMA_FROM_DEVICE);
1695 			buffer_info->dma = 0;
1696 		}
1697 
1698 		if (buffer_info->page) {
1699 			put_page(buffer_info->page);
1700 			buffer_info->page = NULL;
1701 		}
1702 
1703 		if (buffer_info->skb) {
1704 			dev_kfree_skb(buffer_info->skb);
1705 			buffer_info->skb = NULL;
1706 		}
1707 
1708 		for (j = 0; j < PS_PAGE_BUFFERS; j++) {
1709 			ps_page = &buffer_info->ps_pages[j];
1710 			if (!ps_page->page)
1711 				break;
1712 			dma_unmap_page(&pdev->dev, ps_page->dma, PAGE_SIZE,
1713 				       DMA_FROM_DEVICE);
1714 			ps_page->dma = 0;
1715 			put_page(ps_page->page);
1716 			ps_page->page = NULL;
1717 		}
1718 	}
1719 
1720 	/* there also may be some cached data from a chained receive */
1721 	if (rx_ring->rx_skb_top) {
1722 		dev_kfree_skb(rx_ring->rx_skb_top);
1723 		rx_ring->rx_skb_top = NULL;
1724 	}
1725 
1726 	/* Zero out the descriptor ring */
1727 	memset(rx_ring->desc, 0, rx_ring->size);
1728 
1729 	rx_ring->next_to_clean = 0;
1730 	rx_ring->next_to_use = 0;
1731 	adapter->flags2 &= ~FLAG2_IS_DISCARDING;
1732 }
1733 
e1000e_downshift_workaround(struct work_struct * work)1734 static void e1000e_downshift_workaround(struct work_struct *work)
1735 {
1736 	struct e1000_adapter *adapter = container_of(work,
1737 						     struct e1000_adapter,
1738 						     downshift_task);
1739 
1740 	if (test_bit(__E1000_DOWN, &adapter->state))
1741 		return;
1742 
1743 	e1000e_gig_downshift_workaround_ich8lan(&adapter->hw);
1744 }
1745 
1746 /**
1747  * e1000_intr_msi - Interrupt Handler
1748  * @irq: interrupt number
1749  * @data: pointer to a network interface device structure
1750  **/
e1000_intr_msi(int __always_unused irq,void * data)1751 static irqreturn_t e1000_intr_msi(int __always_unused irq, void *data)
1752 {
1753 	struct net_device *netdev = data;
1754 	struct e1000_adapter *adapter = netdev_priv(netdev);
1755 	struct e1000_hw *hw = &adapter->hw;
1756 	u32 icr = er32(ICR);
1757 
1758 	/* read ICR disables interrupts using IAM */
1759 	if (icr & E1000_ICR_LSC) {
1760 		hw->mac.get_link_status = true;
1761 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1762 		 * disconnect (LSC) before accessing any PHY registers
1763 		 */
1764 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1765 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1766 			schedule_work(&adapter->downshift_task);
1767 
1768 		/* 80003ES2LAN workaround-- For packet buffer work-around on
1769 		 * link down event; disable receives here in the ISR and reset
1770 		 * adapter in watchdog
1771 		 */
1772 		if (netif_carrier_ok(netdev) &&
1773 		    adapter->flags & FLAG_RX_NEEDS_RESTART) {
1774 			/* disable receives */
1775 			u32 rctl = er32(RCTL);
1776 
1777 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1778 			adapter->flags |= FLAG_RESTART_NOW;
1779 		}
1780 		/* guard against interrupt when we're going down */
1781 		if (!test_bit(__E1000_DOWN, &adapter->state))
1782 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1783 	}
1784 
1785 	/* Reset on uncorrectable ECC error */
1786 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1787 		u32 pbeccsts = er32(PBECCSTS);
1788 
1789 		adapter->corr_errors +=
1790 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1791 		adapter->uncorr_errors +=
1792 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1793 
1794 		/* Do the reset outside of interrupt context */
1795 		schedule_work(&adapter->reset_task);
1796 
1797 		/* return immediately since reset is imminent */
1798 		return IRQ_HANDLED;
1799 	}
1800 
1801 	if (napi_schedule_prep(&adapter->napi)) {
1802 		adapter->total_tx_bytes = 0;
1803 		adapter->total_tx_packets = 0;
1804 		adapter->total_rx_bytes = 0;
1805 		adapter->total_rx_packets = 0;
1806 		__napi_schedule(&adapter->napi);
1807 	}
1808 
1809 	return IRQ_HANDLED;
1810 }
1811 
1812 /**
1813  * e1000_intr - Interrupt Handler
1814  * @irq: interrupt number
1815  * @data: pointer to a network interface device structure
1816  **/
e1000_intr(int __always_unused irq,void * data)1817 static irqreturn_t e1000_intr(int __always_unused irq, void *data)
1818 {
1819 	struct net_device *netdev = data;
1820 	struct e1000_adapter *adapter = netdev_priv(netdev);
1821 	struct e1000_hw *hw = &adapter->hw;
1822 	u32 rctl, icr = er32(ICR);
1823 
1824 	if (!icr || test_bit(__E1000_DOWN, &adapter->state))
1825 		return IRQ_NONE;	/* Not our interrupt */
1826 
1827 	/* IMS will not auto-mask if INT_ASSERTED is not set, and if it is
1828 	 * not set, then the adapter didn't send an interrupt
1829 	 */
1830 	if (!(icr & E1000_ICR_INT_ASSERTED))
1831 		return IRQ_NONE;
1832 
1833 	/* Interrupt Auto-Mask...upon reading ICR,
1834 	 * interrupts are masked.  No need for the
1835 	 * IMC write
1836 	 */
1837 
1838 	if (icr & E1000_ICR_LSC) {
1839 		hw->mac.get_link_status = true;
1840 		/* ICH8 workaround-- Call gig speed drop workaround on cable
1841 		 * disconnect (LSC) before accessing any PHY registers
1842 		 */
1843 		if ((adapter->flags & FLAG_LSC_GIG_SPEED_DROP) &&
1844 		    (!(er32(STATUS) & E1000_STATUS_LU)))
1845 			schedule_work(&adapter->downshift_task);
1846 
1847 		/* 80003ES2LAN workaround--
1848 		 * For packet buffer work-around on link down event;
1849 		 * disable receives here in the ISR and
1850 		 * reset adapter in watchdog
1851 		 */
1852 		if (netif_carrier_ok(netdev) &&
1853 		    (adapter->flags & FLAG_RX_NEEDS_RESTART)) {
1854 			/* disable receives */
1855 			rctl = er32(RCTL);
1856 			ew32(RCTL, rctl & ~E1000_RCTL_EN);
1857 			adapter->flags |= FLAG_RESTART_NOW;
1858 		}
1859 		/* guard against interrupt when we're going down */
1860 		if (!test_bit(__E1000_DOWN, &adapter->state))
1861 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1862 	}
1863 
1864 	/* Reset on uncorrectable ECC error */
1865 	if ((icr & E1000_ICR_ECCER) && (hw->mac.type >= e1000_pch_lpt)) {
1866 		u32 pbeccsts = er32(PBECCSTS);
1867 
1868 		adapter->corr_errors +=
1869 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
1870 		adapter->uncorr_errors +=
1871 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
1872 
1873 		/* Do the reset outside of interrupt context */
1874 		schedule_work(&adapter->reset_task);
1875 
1876 		/* return immediately since reset is imminent */
1877 		return IRQ_HANDLED;
1878 	}
1879 
1880 	if (napi_schedule_prep(&adapter->napi)) {
1881 		adapter->total_tx_bytes = 0;
1882 		adapter->total_tx_packets = 0;
1883 		adapter->total_rx_bytes = 0;
1884 		adapter->total_rx_packets = 0;
1885 		__napi_schedule(&adapter->napi);
1886 	}
1887 
1888 	return IRQ_HANDLED;
1889 }
1890 
e1000_msix_other(int __always_unused irq,void * data)1891 static irqreturn_t e1000_msix_other(int __always_unused irq, void *data)
1892 {
1893 	struct net_device *netdev = data;
1894 	struct e1000_adapter *adapter = netdev_priv(netdev);
1895 	struct e1000_hw *hw = &adapter->hw;
1896 	u32 icr = er32(ICR);
1897 
1898 	if (icr & adapter->eiac_mask)
1899 		ew32(ICS, (icr & adapter->eiac_mask));
1900 
1901 	if (icr & E1000_ICR_LSC) {
1902 		hw->mac.get_link_status = true;
1903 		/* guard against interrupt when we're going down */
1904 		if (!test_bit(__E1000_DOWN, &adapter->state))
1905 			mod_timer(&adapter->watchdog_timer, jiffies + 1);
1906 	}
1907 
1908 	if (!test_bit(__E1000_DOWN, &adapter->state))
1909 		ew32(IMS, E1000_IMS_OTHER | IMS_OTHER_MASK);
1910 
1911 	return IRQ_HANDLED;
1912 }
1913 
e1000_intr_msix_tx(int __always_unused irq,void * data)1914 static irqreturn_t e1000_intr_msix_tx(int __always_unused irq, void *data)
1915 {
1916 	struct net_device *netdev = data;
1917 	struct e1000_adapter *adapter = netdev_priv(netdev);
1918 	struct e1000_hw *hw = &adapter->hw;
1919 	struct e1000_ring *tx_ring = adapter->tx_ring;
1920 
1921 	adapter->total_tx_bytes = 0;
1922 	adapter->total_tx_packets = 0;
1923 
1924 	if (!e1000_clean_tx_irq(tx_ring))
1925 		/* Ring was not completely cleaned, so fire another interrupt */
1926 		ew32(ICS, tx_ring->ims_val);
1927 
1928 	if (!test_bit(__E1000_DOWN, &adapter->state))
1929 		ew32(IMS, adapter->tx_ring->ims_val);
1930 
1931 	return IRQ_HANDLED;
1932 }
1933 
e1000_intr_msix_rx(int __always_unused irq,void * data)1934 static irqreturn_t e1000_intr_msix_rx(int __always_unused irq, void *data)
1935 {
1936 	struct net_device *netdev = data;
1937 	struct e1000_adapter *adapter = netdev_priv(netdev);
1938 	struct e1000_ring *rx_ring = adapter->rx_ring;
1939 
1940 	/* Write the ITR value calculated at the end of the
1941 	 * previous interrupt.
1942 	 */
1943 	if (rx_ring->set_itr) {
1944 		u32 itr = rx_ring->itr_val ?
1945 			  1000000000 / (rx_ring->itr_val * 256) : 0;
1946 
1947 		writel(itr, rx_ring->itr_register);
1948 		rx_ring->set_itr = 0;
1949 	}
1950 
1951 	if (napi_schedule_prep(&adapter->napi)) {
1952 		adapter->total_rx_bytes = 0;
1953 		adapter->total_rx_packets = 0;
1954 		__napi_schedule(&adapter->napi);
1955 	}
1956 	return IRQ_HANDLED;
1957 }
1958 
1959 /**
1960  * e1000_configure_msix - Configure MSI-X hardware
1961  * @adapter: board private structure
1962  *
1963  * e1000_configure_msix sets up the hardware to properly
1964  * generate MSI-X interrupts.
1965  **/
e1000_configure_msix(struct e1000_adapter * adapter)1966 static void e1000_configure_msix(struct e1000_adapter *adapter)
1967 {
1968 	struct e1000_hw *hw = &adapter->hw;
1969 	struct e1000_ring *rx_ring = adapter->rx_ring;
1970 	struct e1000_ring *tx_ring = adapter->tx_ring;
1971 	int vector = 0;
1972 	u32 ctrl_ext, ivar = 0;
1973 
1974 	adapter->eiac_mask = 0;
1975 
1976 	/* Workaround issue with spurious interrupts on 82574 in MSI-X mode */
1977 	if (hw->mac.type == e1000_82574) {
1978 		u32 rfctl = er32(RFCTL);
1979 
1980 		rfctl |= E1000_RFCTL_ACK_DIS;
1981 		ew32(RFCTL, rfctl);
1982 	}
1983 
1984 	/* Configure Rx vector */
1985 	rx_ring->ims_val = E1000_IMS_RXQ0;
1986 	adapter->eiac_mask |= rx_ring->ims_val;
1987 	if (rx_ring->itr_val)
1988 		writel(1000000000 / (rx_ring->itr_val * 256),
1989 		       rx_ring->itr_register);
1990 	else
1991 		writel(1, rx_ring->itr_register);
1992 	ivar = E1000_IVAR_INT_ALLOC_VALID | vector;
1993 
1994 	/* Configure Tx vector */
1995 	tx_ring->ims_val = E1000_IMS_TXQ0;
1996 	vector++;
1997 	if (tx_ring->itr_val)
1998 		writel(1000000000 / (tx_ring->itr_val * 256),
1999 		       tx_ring->itr_register);
2000 	else
2001 		writel(1, tx_ring->itr_register);
2002 	adapter->eiac_mask |= tx_ring->ims_val;
2003 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 8);
2004 
2005 	/* set vector for Other Causes, e.g. link changes */
2006 	vector++;
2007 	ivar |= ((E1000_IVAR_INT_ALLOC_VALID | vector) << 16);
2008 	if (rx_ring->itr_val)
2009 		writel(1000000000 / (rx_ring->itr_val * 256),
2010 		       hw->hw_addr + E1000_EITR_82574(vector));
2011 	else
2012 		writel(1, hw->hw_addr + E1000_EITR_82574(vector));
2013 
2014 	/* Cause Tx interrupts on every write back */
2015 	ivar |= BIT(31);
2016 
2017 	ew32(IVAR, ivar);
2018 
2019 	/* enable MSI-X PBA support */
2020 	ctrl_ext = er32(CTRL_EXT) & ~E1000_CTRL_EXT_IAME;
2021 	ctrl_ext |= E1000_CTRL_EXT_PBA_CLR | E1000_CTRL_EXT_EIAME;
2022 	ew32(CTRL_EXT, ctrl_ext);
2023 	e1e_flush();
2024 }
2025 
e1000e_reset_interrupt_capability(struct e1000_adapter * adapter)2026 void e1000e_reset_interrupt_capability(struct e1000_adapter *adapter)
2027 {
2028 	if (adapter->msix_entries) {
2029 		pci_disable_msix(adapter->pdev);
2030 		kfree(adapter->msix_entries);
2031 		adapter->msix_entries = NULL;
2032 	} else if (adapter->flags & FLAG_MSI_ENABLED) {
2033 		pci_disable_msi(adapter->pdev);
2034 		adapter->flags &= ~FLAG_MSI_ENABLED;
2035 	}
2036 }
2037 
2038 /**
2039  * e1000e_set_interrupt_capability - set MSI or MSI-X if supported
2040  * @adapter: board private structure
2041  *
2042  * Attempt to configure interrupts using the best available
2043  * capabilities of the hardware and kernel.
2044  **/
e1000e_set_interrupt_capability(struct e1000_adapter * adapter)2045 void e1000e_set_interrupt_capability(struct e1000_adapter *adapter)
2046 {
2047 	int err;
2048 	int i;
2049 
2050 	switch (adapter->int_mode) {
2051 	case E1000E_INT_MODE_MSIX:
2052 		if (adapter->flags & FLAG_HAS_MSIX) {
2053 			adapter->num_vectors = 3; /* RxQ0, TxQ0 and other */
2054 			adapter->msix_entries = kzalloc_objs(struct msix_entry,
2055 							     adapter->num_vectors);
2056 			if (adapter->msix_entries) {
2057 				struct e1000_adapter *a = adapter;
2058 
2059 				for (i = 0; i < adapter->num_vectors; i++)
2060 					adapter->msix_entries[i].entry = i;
2061 
2062 				err = pci_enable_msix_range(a->pdev,
2063 							    a->msix_entries,
2064 							    a->num_vectors,
2065 							    a->num_vectors);
2066 				if (err > 0)
2067 					return;
2068 			}
2069 			/* MSI-X failed, so fall through and try MSI */
2070 			e_err("Failed to initialize MSI-X interrupts.  Falling back to MSI interrupts.\n");
2071 			e1000e_reset_interrupt_capability(adapter);
2072 		}
2073 		adapter->int_mode = E1000E_INT_MODE_MSI;
2074 		fallthrough;
2075 	case E1000E_INT_MODE_MSI:
2076 		if (!pci_enable_msi(adapter->pdev)) {
2077 			adapter->flags |= FLAG_MSI_ENABLED;
2078 		} else {
2079 			adapter->int_mode = E1000E_INT_MODE_LEGACY;
2080 			e_err("Failed to initialize MSI interrupts.  Falling back to legacy interrupts.\n");
2081 		}
2082 		fallthrough;
2083 	case E1000E_INT_MODE_LEGACY:
2084 		/* Don't do anything; this is the system default */
2085 		break;
2086 	}
2087 
2088 	/* store the number of vectors being used */
2089 	adapter->num_vectors = 1;
2090 }
2091 
2092 /**
2093  * e1000_request_msix - Initialize MSI-X interrupts
2094  * @adapter: board private structure
2095  *
2096  * e1000_request_msix allocates MSI-X vectors and requests interrupts from the
2097  * kernel.
2098  **/
e1000_request_msix(struct e1000_adapter * adapter)2099 static int e1000_request_msix(struct e1000_adapter *adapter)
2100 {
2101 	struct net_device *netdev = adapter->netdev;
2102 	int err = 0, vector = 0;
2103 
2104 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2105 		snprintf(adapter->rx_ring->name,
2106 			 sizeof(adapter->rx_ring->name) - 1,
2107 			 "%.14s-rx-0", netdev->name);
2108 	else
2109 		memcpy(adapter->rx_ring->name, netdev->name, IFNAMSIZ);
2110 	err = request_irq(adapter->msix_entries[vector].vector,
2111 			  e1000_intr_msix_rx, 0, adapter->rx_ring->name,
2112 			  netdev);
2113 	if (err)
2114 		return err;
2115 	adapter->rx_ring->itr_register = adapter->hw.hw_addr +
2116 	    E1000_EITR_82574(vector);
2117 	adapter->rx_ring->itr_val = adapter->itr;
2118 	vector++;
2119 
2120 	if (strlen(netdev->name) < (IFNAMSIZ - 5))
2121 		snprintf(adapter->tx_ring->name,
2122 			 sizeof(adapter->tx_ring->name) - 1,
2123 			 "%.14s-tx-0", netdev->name);
2124 	else
2125 		memcpy(adapter->tx_ring->name, netdev->name, IFNAMSIZ);
2126 	err = request_irq(adapter->msix_entries[vector].vector,
2127 			  e1000_intr_msix_tx, 0, adapter->tx_ring->name,
2128 			  netdev);
2129 	if (err)
2130 		return err;
2131 	adapter->tx_ring->itr_register = adapter->hw.hw_addr +
2132 	    E1000_EITR_82574(vector);
2133 	adapter->tx_ring->itr_val = adapter->itr;
2134 	vector++;
2135 
2136 	err = request_irq(adapter->msix_entries[vector].vector,
2137 			  e1000_msix_other, 0, netdev->name, netdev);
2138 	if (err)
2139 		return err;
2140 
2141 	e1000_configure_msix(adapter);
2142 
2143 	return 0;
2144 }
2145 
2146 /**
2147  * e1000_request_irq - initialize interrupts
2148  * @adapter: board private structure
2149  *
2150  * Attempts to configure interrupts using the best available
2151  * capabilities of the hardware and kernel.
2152  **/
e1000_request_irq(struct e1000_adapter * adapter)2153 static int e1000_request_irq(struct e1000_adapter *adapter)
2154 {
2155 	struct net_device *netdev = adapter->netdev;
2156 	int err;
2157 
2158 	if (adapter->msix_entries) {
2159 		err = e1000_request_msix(adapter);
2160 		if (!err)
2161 			return err;
2162 		/* fall back to MSI */
2163 		e1000e_reset_interrupt_capability(adapter);
2164 		adapter->int_mode = E1000E_INT_MODE_MSI;
2165 		e1000e_set_interrupt_capability(adapter);
2166 	}
2167 	if (adapter->flags & FLAG_MSI_ENABLED) {
2168 		err = request_irq(adapter->pdev->irq, e1000_intr_msi, 0,
2169 				  netdev->name, netdev);
2170 		if (!err)
2171 			return err;
2172 
2173 		/* fall back to legacy interrupt */
2174 		e1000e_reset_interrupt_capability(adapter);
2175 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
2176 	}
2177 
2178 	err = request_irq(adapter->pdev->irq, e1000_intr, IRQF_SHARED,
2179 			  netdev->name, netdev);
2180 	if (err)
2181 		e_err("Unable to allocate interrupt, Error: %d\n", err);
2182 
2183 	return err;
2184 }
2185 
e1000_free_irq(struct e1000_adapter * adapter)2186 static void e1000_free_irq(struct e1000_adapter *adapter)
2187 {
2188 	struct net_device *netdev = adapter->netdev;
2189 
2190 	if (adapter->msix_entries) {
2191 		int vector = 0;
2192 
2193 		free_irq(adapter->msix_entries[vector].vector, netdev);
2194 		vector++;
2195 
2196 		free_irq(adapter->msix_entries[vector].vector, netdev);
2197 		vector++;
2198 
2199 		/* Other Causes interrupt vector */
2200 		free_irq(adapter->msix_entries[vector].vector, netdev);
2201 		return;
2202 	}
2203 
2204 	free_irq(adapter->pdev->irq, netdev);
2205 }
2206 
2207 /**
2208  * e1000_irq_disable - Mask off interrupt generation on the NIC
2209  * @adapter: board private structure
2210  **/
e1000_irq_disable(struct e1000_adapter * adapter)2211 static void e1000_irq_disable(struct e1000_adapter *adapter)
2212 {
2213 	struct e1000_hw *hw = &adapter->hw;
2214 
2215 	ew32(IMC, ~0);
2216 	if (adapter->msix_entries)
2217 		ew32(EIAC_82574, 0);
2218 	e1e_flush();
2219 
2220 	if (adapter->msix_entries) {
2221 		int i;
2222 
2223 		for (i = 0; i < adapter->num_vectors; i++)
2224 			synchronize_irq(adapter->msix_entries[i].vector);
2225 	} else {
2226 		synchronize_irq(adapter->pdev->irq);
2227 	}
2228 }
2229 
2230 /**
2231  * e1000_irq_enable - Enable default interrupt generation settings
2232  * @adapter: board private structure
2233  **/
e1000_irq_enable(struct e1000_adapter * adapter)2234 static void e1000_irq_enable(struct e1000_adapter *adapter)
2235 {
2236 	struct e1000_hw *hw = &adapter->hw;
2237 
2238 	if (adapter->msix_entries) {
2239 		ew32(EIAC_82574, adapter->eiac_mask & E1000_EIAC_MASK_82574);
2240 		ew32(IMS, adapter->eiac_mask | E1000_IMS_OTHER |
2241 		     IMS_OTHER_MASK);
2242 	} else if (hw->mac.type >= e1000_pch_lpt) {
2243 		ew32(IMS, IMS_ENABLE_MASK | E1000_IMS_ECCER);
2244 	} else {
2245 		ew32(IMS, IMS_ENABLE_MASK);
2246 	}
2247 	e1e_flush();
2248 }
2249 
2250 /**
2251  * e1000e_get_hw_control - get control of the h/w from f/w
2252  * @adapter: address of board private structure
2253  *
2254  * e1000e_get_hw_control sets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2255  * For ASF and Pass Through versions of f/w this means that
2256  * the driver is loaded. For AMT version (only with 82573)
2257  * of the f/w this means that the network i/f is open.
2258  **/
e1000e_get_hw_control(struct e1000_adapter * adapter)2259 void e1000e_get_hw_control(struct e1000_adapter *adapter)
2260 {
2261 	struct e1000_hw *hw = &adapter->hw;
2262 	u32 ctrl_ext;
2263 	u32 swsm;
2264 
2265 	/* Let firmware know the driver has taken over */
2266 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2267 		swsm = er32(SWSM);
2268 		ew32(SWSM, swsm | E1000_SWSM_DRV_LOAD);
2269 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2270 		ctrl_ext = er32(CTRL_EXT);
2271 		ew32(CTRL_EXT, ctrl_ext | E1000_CTRL_EXT_DRV_LOAD);
2272 	}
2273 }
2274 
2275 /**
2276  * e1000e_release_hw_control - release control of the h/w to f/w
2277  * @adapter: address of board private structure
2278  *
2279  * e1000e_release_hw_control resets {CTRL_EXT|SWSM}:DRV_LOAD bit.
2280  * For ASF and Pass Through versions of f/w this means that the
2281  * driver is no longer loaded. For AMT version (only with 82573) i
2282  * of the f/w this means that the network i/f is closed.
2283  *
2284  **/
e1000e_release_hw_control(struct e1000_adapter * adapter)2285 void e1000e_release_hw_control(struct e1000_adapter *adapter)
2286 {
2287 	struct e1000_hw *hw = &adapter->hw;
2288 	u32 ctrl_ext;
2289 	u32 swsm;
2290 
2291 	/* Let firmware taken over control of h/w */
2292 	if (adapter->flags & FLAG_HAS_SWSM_ON_LOAD) {
2293 		swsm = er32(SWSM);
2294 		ew32(SWSM, swsm & ~E1000_SWSM_DRV_LOAD);
2295 	} else if (adapter->flags & FLAG_HAS_CTRLEXT_ON_LOAD) {
2296 		ctrl_ext = er32(CTRL_EXT);
2297 		ew32(CTRL_EXT, ctrl_ext & ~E1000_CTRL_EXT_DRV_LOAD);
2298 	}
2299 }
2300 
2301 /**
2302  * e1000_alloc_ring_dma - allocate memory for a ring structure
2303  * @adapter: board private structure
2304  * @ring: ring struct for which to allocate dma
2305  **/
e1000_alloc_ring_dma(struct e1000_adapter * adapter,struct e1000_ring * ring)2306 static int e1000_alloc_ring_dma(struct e1000_adapter *adapter,
2307 				struct e1000_ring *ring)
2308 {
2309 	struct pci_dev *pdev = adapter->pdev;
2310 
2311 	ring->desc = dma_alloc_coherent(&pdev->dev, ring->size, &ring->dma,
2312 					GFP_KERNEL);
2313 	if (!ring->desc)
2314 		return -ENOMEM;
2315 
2316 	return 0;
2317 }
2318 
2319 /**
2320  * e1000e_setup_tx_resources - allocate Tx resources (Descriptors)
2321  * @tx_ring: Tx descriptor ring
2322  *
2323  * Return 0 on success, negative on failure
2324  **/
e1000e_setup_tx_resources(struct e1000_ring * tx_ring)2325 int e1000e_setup_tx_resources(struct e1000_ring *tx_ring)
2326 {
2327 	struct e1000_adapter *adapter = tx_ring->adapter;
2328 	int err = -ENOMEM, size;
2329 
2330 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2331 	tx_ring->buffer_info = vzalloc(size);
2332 	if (!tx_ring->buffer_info)
2333 		goto err;
2334 
2335 	/* round up to nearest 4K */
2336 	tx_ring->size = tx_ring->count * sizeof(struct e1000_tx_desc);
2337 	tx_ring->size = ALIGN(tx_ring->size, 4096);
2338 
2339 	err = e1000_alloc_ring_dma(adapter, tx_ring);
2340 	if (err)
2341 		goto err;
2342 
2343 	tx_ring->next_to_use = 0;
2344 	tx_ring->next_to_clean = 0;
2345 
2346 	return 0;
2347 err:
2348 	vfree(tx_ring->buffer_info);
2349 	e_err("Unable to allocate memory for the transmit descriptor ring\n");
2350 	return err;
2351 }
2352 
2353 /**
2354  * e1000e_setup_rx_resources - allocate Rx resources (Descriptors)
2355  * @rx_ring: Rx descriptor ring
2356  *
2357  * Returns 0 on success, negative on failure
2358  **/
e1000e_setup_rx_resources(struct e1000_ring * rx_ring)2359 int e1000e_setup_rx_resources(struct e1000_ring *rx_ring)
2360 {
2361 	struct e1000_adapter *adapter = rx_ring->adapter;
2362 	struct e1000_buffer *buffer_info;
2363 	int i, size, desc_len, err = -ENOMEM;
2364 
2365 	size = sizeof(struct e1000_buffer) * rx_ring->count;
2366 	rx_ring->buffer_info = vzalloc(size);
2367 	if (!rx_ring->buffer_info)
2368 		goto err;
2369 
2370 	for (i = 0; i < rx_ring->count; i++) {
2371 		buffer_info = &rx_ring->buffer_info[i];
2372 		buffer_info->ps_pages = kzalloc_objs(struct e1000_ps_page,
2373 						     PS_PAGE_BUFFERS);
2374 		if (!buffer_info->ps_pages)
2375 			goto err_pages;
2376 	}
2377 
2378 	desc_len = sizeof(union e1000_rx_desc_packet_split);
2379 
2380 	/* Round up to nearest 4K */
2381 	rx_ring->size = rx_ring->count * desc_len;
2382 	rx_ring->size = ALIGN(rx_ring->size, 4096);
2383 
2384 	err = e1000_alloc_ring_dma(adapter, rx_ring);
2385 	if (err)
2386 		goto err_pages;
2387 
2388 	rx_ring->next_to_clean = 0;
2389 	rx_ring->next_to_use = 0;
2390 	rx_ring->rx_skb_top = NULL;
2391 
2392 	return 0;
2393 
2394 err_pages:
2395 	for (i = 0; i < rx_ring->count; i++) {
2396 		buffer_info = &rx_ring->buffer_info[i];
2397 		kfree(buffer_info->ps_pages);
2398 	}
2399 err:
2400 	vfree(rx_ring->buffer_info);
2401 	e_err("Unable to allocate memory for the receive descriptor ring\n");
2402 	return err;
2403 }
2404 
2405 /**
2406  * e1000_clean_tx_ring - Free Tx Buffers
2407  * @tx_ring: Tx descriptor ring
2408  **/
e1000_clean_tx_ring(struct e1000_ring * tx_ring)2409 static void e1000_clean_tx_ring(struct e1000_ring *tx_ring)
2410 {
2411 	struct e1000_adapter *adapter = tx_ring->adapter;
2412 	struct e1000_buffer *buffer_info;
2413 	unsigned long size;
2414 	unsigned int i;
2415 
2416 	for (i = 0; i < tx_ring->count; i++) {
2417 		buffer_info = &tx_ring->buffer_info[i];
2418 		e1000_put_txbuf(tx_ring, buffer_info, false);
2419 	}
2420 
2421 	netdev_reset_queue(adapter->netdev);
2422 	size = sizeof(struct e1000_buffer) * tx_ring->count;
2423 	memset(tx_ring->buffer_info, 0, size);
2424 
2425 	memset(tx_ring->desc, 0, tx_ring->size);
2426 
2427 	tx_ring->next_to_use = 0;
2428 	tx_ring->next_to_clean = 0;
2429 }
2430 
2431 /**
2432  * e1000e_free_tx_resources - Free Tx Resources per Queue
2433  * @tx_ring: Tx descriptor ring
2434  *
2435  * Free all transmit software resources
2436  **/
e1000e_free_tx_resources(struct e1000_ring * tx_ring)2437 void e1000e_free_tx_resources(struct e1000_ring *tx_ring)
2438 {
2439 	struct e1000_adapter *adapter = tx_ring->adapter;
2440 	struct pci_dev *pdev = adapter->pdev;
2441 
2442 	e1000_clean_tx_ring(tx_ring);
2443 
2444 	vfree(tx_ring->buffer_info);
2445 	tx_ring->buffer_info = NULL;
2446 
2447 	dma_free_coherent(&pdev->dev, tx_ring->size, tx_ring->desc,
2448 			  tx_ring->dma);
2449 	tx_ring->desc = NULL;
2450 }
2451 
2452 /**
2453  * e1000e_free_rx_resources - Free Rx Resources
2454  * @rx_ring: Rx descriptor ring
2455  *
2456  * Free all receive software resources
2457  **/
e1000e_free_rx_resources(struct e1000_ring * rx_ring)2458 void e1000e_free_rx_resources(struct e1000_ring *rx_ring)
2459 {
2460 	struct e1000_adapter *adapter = rx_ring->adapter;
2461 	struct pci_dev *pdev = adapter->pdev;
2462 	int i;
2463 
2464 	e1000_clean_rx_ring(rx_ring);
2465 
2466 	for (i = 0; i < rx_ring->count; i++)
2467 		kfree(rx_ring->buffer_info[i].ps_pages);
2468 
2469 	vfree(rx_ring->buffer_info);
2470 	rx_ring->buffer_info = NULL;
2471 
2472 	dma_free_coherent(&pdev->dev, rx_ring->size, rx_ring->desc,
2473 			  rx_ring->dma);
2474 	rx_ring->desc = NULL;
2475 }
2476 
2477 /**
2478  * e1000_update_itr - update the dynamic ITR value based on statistics
2479  * @itr_setting: current adapter->itr
2480  * @packets: the number of packets during this measurement interval
2481  * @bytes: the number of bytes during this measurement interval
2482  *
2483  *      Stores a new ITR value based on packets and byte
2484  *      counts during the last interrupt.  The advantage of per interrupt
2485  *      computation is faster updates and more accurate ITR for the current
2486  *      traffic pattern.  Constants in this function were computed
2487  *      based on theoretical maximum wire speed and thresholds were set based
2488  *      on testing data as well as attempting to minimize response time
2489  *      while increasing bulk throughput.  This functionality is controlled
2490  *      by the InterruptThrottleRate module parameter.
2491  **/
e1000_update_itr(u16 itr_setting,int packets,int bytes)2492 static unsigned int e1000_update_itr(u16 itr_setting, int packets, int bytes)
2493 {
2494 	unsigned int retval = itr_setting;
2495 
2496 	if (packets == 0)
2497 		return itr_setting;
2498 
2499 	switch (itr_setting) {
2500 	case lowest_latency:
2501 		/* handle TSO and jumbo frames */
2502 		if (bytes / packets > 8000)
2503 			retval = bulk_latency;
2504 		else if ((packets < 5) && (bytes > 512))
2505 			retval = low_latency;
2506 		break;
2507 	case low_latency:	/* 50 usec aka 20000 ints/s */
2508 		if (bytes > 10000) {
2509 			/* this if handles the TSO accounting */
2510 			if (bytes / packets > 8000)
2511 				retval = bulk_latency;
2512 			else if ((packets < 10) || ((bytes / packets) > 1200))
2513 				retval = bulk_latency;
2514 			else if ((packets > 35))
2515 				retval = lowest_latency;
2516 		} else if (bytes / packets > 2000) {
2517 			retval = bulk_latency;
2518 		} else if (packets <= 2 && bytes < 512) {
2519 			retval = lowest_latency;
2520 		}
2521 		break;
2522 	case bulk_latency:	/* 250 usec aka 4000 ints/s */
2523 		if (bytes > 25000) {
2524 			if (packets > 35)
2525 				retval = low_latency;
2526 		} else if (bytes < 6000) {
2527 			retval = low_latency;
2528 		}
2529 		break;
2530 	}
2531 
2532 	return retval;
2533 }
2534 
e1000_set_itr(struct e1000_adapter * adapter)2535 static void e1000_set_itr(struct e1000_adapter *adapter)
2536 {
2537 	u16 current_itr;
2538 	u32 new_itr = adapter->itr;
2539 
2540 	/* for non-gigabit speeds, just fix the interrupt rate at 4000 */
2541 	if (adapter->link_speed != SPEED_1000) {
2542 		new_itr = 4000;
2543 		goto set_itr_now;
2544 	}
2545 
2546 	if (adapter->flags2 & FLAG2_DISABLE_AIM) {
2547 		new_itr = 0;
2548 		goto set_itr_now;
2549 	}
2550 
2551 	adapter->tx_itr = e1000_update_itr(adapter->tx_itr,
2552 					   adapter->total_tx_packets,
2553 					   adapter->total_tx_bytes);
2554 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2555 	if (adapter->itr_setting == 3 && adapter->tx_itr == lowest_latency)
2556 		adapter->tx_itr = low_latency;
2557 
2558 	adapter->rx_itr = e1000_update_itr(adapter->rx_itr,
2559 					   adapter->total_rx_packets,
2560 					   adapter->total_rx_bytes);
2561 	/* conservative mode (itr 3) eliminates the lowest_latency setting */
2562 	if (adapter->itr_setting == 3 && adapter->rx_itr == lowest_latency)
2563 		adapter->rx_itr = low_latency;
2564 
2565 	current_itr = max(adapter->rx_itr, adapter->tx_itr);
2566 
2567 	/* counts and packets in update_itr are dependent on these numbers */
2568 	switch (current_itr) {
2569 	case lowest_latency:
2570 		new_itr = 70000;
2571 		break;
2572 	case low_latency:
2573 		new_itr = 20000;	/* aka hwitr = ~200 */
2574 		break;
2575 	case bulk_latency:
2576 		new_itr = 4000;
2577 		break;
2578 	default:
2579 		break;
2580 	}
2581 
2582 set_itr_now:
2583 	if (new_itr != adapter->itr) {
2584 		/* this attempts to bias the interrupt rate towards Bulk
2585 		 * by adding intermediate steps when interrupt rate is
2586 		 * increasing
2587 		 */
2588 		new_itr = new_itr > adapter->itr ?
2589 		    min(adapter->itr + (new_itr >> 2), new_itr) : new_itr;
2590 		adapter->itr = new_itr;
2591 		adapter->rx_ring->itr_val = new_itr;
2592 		if (adapter->msix_entries)
2593 			adapter->rx_ring->set_itr = 1;
2594 		else
2595 			e1000e_write_itr(adapter, new_itr);
2596 	}
2597 }
2598 
2599 /**
2600  * e1000e_write_itr - write the ITR value to the appropriate registers
2601  * @adapter: address of board private structure
2602  * @itr: new ITR value to program
2603  *
2604  * e1000e_write_itr determines if the adapter is in MSI-X mode
2605  * and, if so, writes the EITR registers with the ITR value.
2606  * Otherwise, it writes the ITR value into the ITR register.
2607  **/
e1000e_write_itr(struct e1000_adapter * adapter,u32 itr)2608 void e1000e_write_itr(struct e1000_adapter *adapter, u32 itr)
2609 {
2610 	struct e1000_hw *hw = &adapter->hw;
2611 	u32 new_itr = itr ? 1000000000 / (itr * 256) : 0;
2612 
2613 	if (adapter->msix_entries) {
2614 		int vector;
2615 
2616 		for (vector = 0; vector < adapter->num_vectors; vector++)
2617 			writel(new_itr, hw->hw_addr + E1000_EITR_82574(vector));
2618 	} else {
2619 		ew32(ITR, new_itr);
2620 	}
2621 }
2622 
2623 /**
2624  * e1000_alloc_queues - Allocate memory for all rings
2625  * @adapter: board private structure to initialize
2626  **/
e1000_alloc_queues(struct e1000_adapter * adapter)2627 static int e1000_alloc_queues(struct e1000_adapter *adapter)
2628 {
2629 	int size = sizeof(struct e1000_ring);
2630 
2631 	adapter->tx_ring = kzalloc(size, GFP_KERNEL);
2632 	if (!adapter->tx_ring)
2633 		goto err;
2634 	adapter->tx_ring->count = adapter->tx_ring_count;
2635 	adapter->tx_ring->adapter = adapter;
2636 
2637 	adapter->rx_ring = kzalloc(size, GFP_KERNEL);
2638 	if (!adapter->rx_ring)
2639 		goto err;
2640 	adapter->rx_ring->count = adapter->rx_ring_count;
2641 	adapter->rx_ring->adapter = adapter;
2642 
2643 	return 0;
2644 err:
2645 	e_err("Unable to allocate memory for queues\n");
2646 	kfree(adapter->rx_ring);
2647 	kfree(adapter->tx_ring);
2648 	return -ENOMEM;
2649 }
2650 
2651 /**
2652  * e1000e_poll - NAPI Rx polling callback
2653  * @napi: struct associated with this polling callback
2654  * @budget: number of packets driver is allowed to process this poll
2655  **/
e1000e_poll(struct napi_struct * napi,int budget)2656 static int e1000e_poll(struct napi_struct *napi, int budget)
2657 {
2658 	struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter,
2659 						     napi);
2660 	struct e1000_hw *hw = &adapter->hw;
2661 	struct net_device *poll_dev = adapter->netdev;
2662 	int tx_cleaned = 1, work_done = 0;
2663 
2664 	adapter = netdev_priv(poll_dev);
2665 
2666 	if (!adapter->msix_entries ||
2667 	    (adapter->rx_ring->ims_val & adapter->tx_ring->ims_val))
2668 		tx_cleaned = e1000_clean_tx_irq(adapter->tx_ring);
2669 
2670 	adapter->clean_rx(adapter->rx_ring, &work_done, budget);
2671 
2672 	if (!tx_cleaned || work_done == budget)
2673 		return budget;
2674 
2675 	/* Exit the polling mode, but don't re-enable interrupts if stack might
2676 	 * poll us due to busy-polling
2677 	 */
2678 	if (likely(napi_complete_done(napi, work_done))) {
2679 		if (adapter->itr_setting & 3)
2680 			e1000_set_itr(adapter);
2681 		if (!test_bit(__E1000_DOWN, &adapter->state)) {
2682 			if (adapter->msix_entries)
2683 				ew32(IMS, adapter->rx_ring->ims_val);
2684 			else
2685 				e1000_irq_enable(adapter);
2686 		}
2687 	}
2688 
2689 	return work_done;
2690 }
2691 
e1000_vlan_rx_add_vid(struct net_device * netdev,__always_unused __be16 proto,u16 vid)2692 static int e1000_vlan_rx_add_vid(struct net_device *netdev,
2693 				 __always_unused __be16 proto, u16 vid)
2694 {
2695 	struct e1000_adapter *adapter = netdev_priv(netdev);
2696 	struct e1000_hw *hw = &adapter->hw;
2697 	u32 vfta, index;
2698 
2699 	/* don't update vlan cookie if already programmed */
2700 	if ((adapter->hw.mng_cookie.status &
2701 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2702 	    (vid == adapter->mng_vlan_id))
2703 		return 0;
2704 
2705 	/* add VID to filter table */
2706 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2707 		index = (vid >> 5) & 0x7F;
2708 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2709 		vfta |= BIT((vid & 0x1F));
2710 		hw->mac.ops.write_vfta(hw, index, vfta);
2711 	}
2712 
2713 	set_bit(vid, adapter->active_vlans);
2714 
2715 	return 0;
2716 }
2717 
e1000_vlan_rx_kill_vid(struct net_device * netdev,__always_unused __be16 proto,u16 vid)2718 static int e1000_vlan_rx_kill_vid(struct net_device *netdev,
2719 				  __always_unused __be16 proto, u16 vid)
2720 {
2721 	struct e1000_adapter *adapter = netdev_priv(netdev);
2722 	struct e1000_hw *hw = &adapter->hw;
2723 	u32 vfta, index;
2724 
2725 	if ((adapter->hw.mng_cookie.status &
2726 	     E1000_MNG_DHCP_COOKIE_STATUS_VLAN) &&
2727 	    (vid == adapter->mng_vlan_id)) {
2728 		/* release control to f/w */
2729 		e1000e_release_hw_control(adapter);
2730 		return 0;
2731 	}
2732 
2733 	/* remove VID from filter table */
2734 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2735 		index = (vid >> 5) & 0x7F;
2736 		vfta = E1000_READ_REG_ARRAY(hw, E1000_VFTA, index);
2737 		vfta &= ~BIT((vid & 0x1F));
2738 		hw->mac.ops.write_vfta(hw, index, vfta);
2739 	}
2740 
2741 	clear_bit(vid, adapter->active_vlans);
2742 
2743 	return 0;
2744 }
2745 
2746 /**
2747  * e1000e_vlan_filter_disable - helper to disable hw VLAN filtering
2748  * @adapter: board private structure to initialize
2749  **/
e1000e_vlan_filter_disable(struct e1000_adapter * adapter)2750 static void e1000e_vlan_filter_disable(struct e1000_adapter *adapter)
2751 {
2752 	struct net_device *netdev = adapter->netdev;
2753 	struct e1000_hw *hw = &adapter->hw;
2754 	u32 rctl;
2755 
2756 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2757 		/* disable VLAN receive filtering */
2758 		rctl = er32(RCTL);
2759 		rctl &= ~(E1000_RCTL_VFE | E1000_RCTL_CFIEN);
2760 		ew32(RCTL, rctl);
2761 
2762 		if (adapter->mng_vlan_id != E1000_MNG_VLAN_NONE) {
2763 			e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
2764 					       adapter->mng_vlan_id);
2765 			adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
2766 		}
2767 	}
2768 }
2769 
2770 /**
2771  * e1000e_vlan_filter_enable - helper to enable HW VLAN filtering
2772  * @adapter: board private structure to initialize
2773  **/
e1000e_vlan_filter_enable(struct e1000_adapter * adapter)2774 static void e1000e_vlan_filter_enable(struct e1000_adapter *adapter)
2775 {
2776 	struct e1000_hw *hw = &adapter->hw;
2777 	u32 rctl;
2778 
2779 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER) {
2780 		/* enable VLAN receive filtering */
2781 		rctl = er32(RCTL);
2782 		rctl |= E1000_RCTL_VFE;
2783 		rctl &= ~E1000_RCTL_CFIEN;
2784 		ew32(RCTL, rctl);
2785 	}
2786 }
2787 
2788 /**
2789  * e1000e_vlan_strip_disable - helper to disable HW VLAN stripping
2790  * @adapter: board private structure to initialize
2791  **/
e1000e_vlan_strip_disable(struct e1000_adapter * adapter)2792 static void e1000e_vlan_strip_disable(struct e1000_adapter *adapter)
2793 {
2794 	struct e1000_hw *hw = &adapter->hw;
2795 	u32 ctrl;
2796 
2797 	/* disable VLAN tag insert/strip */
2798 	ctrl = er32(CTRL);
2799 	ctrl &= ~E1000_CTRL_VME;
2800 	ew32(CTRL, ctrl);
2801 }
2802 
2803 /**
2804  * e1000e_vlan_strip_enable - helper to enable HW VLAN stripping
2805  * @adapter: board private structure to initialize
2806  **/
e1000e_vlan_strip_enable(struct e1000_adapter * adapter)2807 static void e1000e_vlan_strip_enable(struct e1000_adapter *adapter)
2808 {
2809 	struct e1000_hw *hw = &adapter->hw;
2810 	u32 ctrl;
2811 
2812 	/* enable VLAN tag insert/strip */
2813 	ctrl = er32(CTRL);
2814 	ctrl |= E1000_CTRL_VME;
2815 	ew32(CTRL, ctrl);
2816 }
2817 
e1000_update_mng_vlan(struct e1000_adapter * adapter)2818 static void e1000_update_mng_vlan(struct e1000_adapter *adapter)
2819 {
2820 	struct net_device *netdev = adapter->netdev;
2821 	u16 vid = adapter->hw.mng_cookie.vlan_id;
2822 	u16 old_vid = adapter->mng_vlan_id;
2823 
2824 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN) {
2825 		e1000_vlan_rx_add_vid(netdev, htons(ETH_P_8021Q), vid);
2826 		adapter->mng_vlan_id = vid;
2827 	}
2828 
2829 	if (old_vid != E1000_MNG_VLAN_NONE && vid != old_vid)
2830 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q), old_vid);
2831 }
2832 
e1000_restore_vlan(struct e1000_adapter * adapter)2833 static void e1000_restore_vlan(struct e1000_adapter *adapter)
2834 {
2835 	u16 vid;
2836 
2837 	e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), 0);
2838 
2839 	for_each_set_bit(vid, adapter->active_vlans, VLAN_N_VID)
2840 	    e1000_vlan_rx_add_vid(adapter->netdev, htons(ETH_P_8021Q), vid);
2841 }
2842 
e1000_init_manageability_pt(struct e1000_adapter * adapter)2843 static void e1000_init_manageability_pt(struct e1000_adapter *adapter)
2844 {
2845 	struct e1000_hw *hw = &adapter->hw;
2846 	u32 manc, manc2h, mdef, i, j;
2847 
2848 	if (!(adapter->flags & FLAG_MNG_PT_ENABLED))
2849 		return;
2850 
2851 	manc = er32(MANC);
2852 
2853 	/* enable receiving management packets to the host. this will probably
2854 	 * generate destination unreachable messages from the host OS, but
2855 	 * the packets will be handled on SMBUS
2856 	 */
2857 	manc |= E1000_MANC_EN_MNG2HOST;
2858 	manc2h = er32(MANC2H);
2859 
2860 	switch (hw->mac.type) {
2861 	default:
2862 		manc2h |= (E1000_MANC2H_PORT_623 | E1000_MANC2H_PORT_664);
2863 		break;
2864 	case e1000_82574:
2865 	case e1000_82583:
2866 		/* Check if IPMI pass-through decision filter already exists;
2867 		 * if so, enable it.
2868 		 */
2869 		for (i = 0, j = 0; i < 8; i++) {
2870 			mdef = er32(MDEF(i));
2871 
2872 			/* Ignore filters with anything other than IPMI ports */
2873 			if (mdef & ~(E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2874 				continue;
2875 
2876 			/* Enable this decision filter in MANC2H */
2877 			if (mdef)
2878 				manc2h |= BIT(i);
2879 
2880 			j |= mdef;
2881 		}
2882 
2883 		if (j == (E1000_MDEF_PORT_623 | E1000_MDEF_PORT_664))
2884 			break;
2885 
2886 		/* Create new decision filter in an empty filter */
2887 		for (i = 0, j = 0; i < 8; i++)
2888 			if (er32(MDEF(i)) == 0) {
2889 				ew32(MDEF(i), (E1000_MDEF_PORT_623 |
2890 					       E1000_MDEF_PORT_664));
2891 				manc2h |= BIT(1);
2892 				j++;
2893 				break;
2894 			}
2895 
2896 		if (!j)
2897 			e_warn("Unable to create IPMI pass-through filter\n");
2898 		break;
2899 	}
2900 
2901 	ew32(MANC2H, manc2h);
2902 	ew32(MANC, manc);
2903 }
2904 
2905 /**
2906  * e1000_configure_tx - Configure Transmit Unit after Reset
2907  * @adapter: board private structure
2908  *
2909  * Configure the Tx unit of the MAC after a reset.
2910  **/
e1000_configure_tx(struct e1000_adapter * adapter)2911 static void e1000_configure_tx(struct e1000_adapter *adapter)
2912 {
2913 	struct e1000_hw *hw = &adapter->hw;
2914 	struct e1000_ring *tx_ring = adapter->tx_ring;
2915 	u64 tdba;
2916 	u32 tdlen, tctl, tarc;
2917 
2918 	/* Setup the HW Tx Head and Tail descriptor pointers */
2919 	tdba = tx_ring->dma;
2920 	tdlen = tx_ring->count * sizeof(struct e1000_tx_desc);
2921 	ew32(TDBAL(0), (tdba & DMA_BIT_MASK(32)));
2922 	ew32(TDBAH(0), (tdba >> 32));
2923 	ew32(TDLEN(0), tdlen);
2924 	ew32(TDH(0), 0);
2925 	ew32(TDT(0), 0);
2926 	tx_ring->head = adapter->hw.hw_addr + E1000_TDH(0);
2927 	tx_ring->tail = adapter->hw.hw_addr + E1000_TDT(0);
2928 
2929 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
2930 		e1000e_update_tdt_wa(tx_ring, 0);
2931 
2932 	/* Set the Tx Interrupt Delay register */
2933 	ew32(TIDV, adapter->tx_int_delay);
2934 	/* Tx irq moderation */
2935 	ew32(TADV, adapter->tx_abs_int_delay);
2936 
2937 	if (adapter->flags2 & FLAG2_DMA_BURST) {
2938 		u32 txdctl = er32(TXDCTL(0));
2939 
2940 		txdctl &= ~(E1000_TXDCTL_PTHRESH | E1000_TXDCTL_HTHRESH |
2941 			    E1000_TXDCTL_WTHRESH);
2942 		/* set up some performance related parameters to encourage the
2943 		 * hardware to use the bus more efficiently in bursts, depends
2944 		 * on the tx_int_delay to be enabled,
2945 		 * wthresh = 1 ==> burst write is disabled to avoid Tx stalls
2946 		 * hthresh = 1 ==> prefetch when one or more available
2947 		 * pthresh = 0x1f ==> prefetch if internal cache 31 or less
2948 		 * BEWARE: this seems to work but should be considered first if
2949 		 * there are Tx hangs or other Tx related bugs
2950 		 */
2951 		txdctl |= E1000_TXDCTL_DMA_BURST_ENABLE;
2952 		ew32(TXDCTL(0), txdctl);
2953 	}
2954 	/* erratum work around: set txdctl the same for both queues */
2955 	ew32(TXDCTL(1), er32(TXDCTL(0)));
2956 
2957 	/* Program the Transmit Control Register */
2958 	tctl = er32(TCTL);
2959 	tctl &= ~E1000_TCTL_CT;
2960 	tctl |= E1000_TCTL_PSP | E1000_TCTL_RTLC |
2961 		(E1000_COLLISION_THRESHOLD << E1000_CT_SHIFT);
2962 
2963 	if (adapter->flags & FLAG_TARC_SPEED_MODE_BIT) {
2964 		tarc = er32(TARC(0));
2965 		/* set the speed mode bit, we'll clear it if we're not at
2966 		 * gigabit link later
2967 		 */
2968 #define SPEED_MODE_BIT BIT(21)
2969 		tarc |= SPEED_MODE_BIT;
2970 		ew32(TARC(0), tarc);
2971 	}
2972 
2973 	/* errata: program both queues to unweighted RR */
2974 	if (adapter->flags & FLAG_TARC_SET_BIT_ZERO) {
2975 		tarc = er32(TARC(0));
2976 		tarc |= 1;
2977 		ew32(TARC(0), tarc);
2978 		tarc = er32(TARC(1));
2979 		tarc |= 1;
2980 		ew32(TARC(1), tarc);
2981 	}
2982 
2983 	/* Setup Transmit Descriptor Settings for eop descriptor */
2984 	adapter->txd_cmd = E1000_TXD_CMD_EOP | E1000_TXD_CMD_IFCS;
2985 
2986 	/* only set IDE if we are delaying interrupts using the timers */
2987 	if (adapter->tx_int_delay)
2988 		adapter->txd_cmd |= E1000_TXD_CMD_IDE;
2989 
2990 	/* enable Report Status bit */
2991 	adapter->txd_cmd |= E1000_TXD_CMD_RS;
2992 
2993 	ew32(TCTL, tctl);
2994 
2995 	hw->mac.ops.config_collision_dist(hw);
2996 
2997 	/* SPT and KBL Si errata workaround to avoid data corruption */
2998 	if (hw->mac.type == e1000_pch_spt) {
2999 		u32 reg_val;
3000 
3001 		reg_val = er32(IOSFPC);
3002 		reg_val |= E1000_RCTL_RDMTS_HEX;
3003 		ew32(IOSFPC, reg_val);
3004 
3005 		reg_val = er32(TARC(0));
3006 		/* SPT and KBL Si errata workaround to avoid Tx hang.
3007 		 * Dropping the number of outstanding requests from
3008 		 * 3 to 2 in order to avoid a buffer overrun.
3009 		 */
3010 		reg_val &= ~E1000_TARC0_CB_MULTIQ_3_REQ;
3011 		reg_val |= E1000_TARC0_CB_MULTIQ_2_REQ;
3012 		ew32(TARC(0), reg_val);
3013 	}
3014 }
3015 
3016 #define PAGE_USE_COUNT(S) (((S) >> PAGE_SHIFT) + \
3017 			   (((S) & (PAGE_SIZE - 1)) ? 1 : 0))
3018 
3019 /**
3020  * e1000_setup_rctl - configure the receive control registers
3021  * @adapter: Board private structure
3022  **/
e1000_setup_rctl(struct e1000_adapter * adapter)3023 static void e1000_setup_rctl(struct e1000_adapter *adapter)
3024 {
3025 	struct e1000_hw *hw = &adapter->hw;
3026 	u32 rctl, rfctl;
3027 	u32 pages = 0;
3028 
3029 	/* Workaround Si errata on PCHx - configure jumbo frame flow.
3030 	 * If jumbo frames not set, program related MAC/PHY registers
3031 	 * to h/w defaults
3032 	 */
3033 	if (hw->mac.type >= e1000_pch2lan) {
3034 		s32 ret_val;
3035 
3036 		if (adapter->netdev->mtu > ETH_DATA_LEN)
3037 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, true);
3038 		else
3039 			ret_val = e1000_lv_jumbo_workaround_ich8lan(hw, false);
3040 
3041 		if (ret_val)
3042 			e_dbg("failed to enable|disable jumbo frame workaround mode\n");
3043 	}
3044 
3045 	/* Program MC offset vector base */
3046 	rctl = er32(RCTL);
3047 	rctl &= ~(3 << E1000_RCTL_MO_SHIFT);
3048 	rctl |= E1000_RCTL_EN | E1000_RCTL_BAM |
3049 	    E1000_RCTL_LBM_NO | E1000_RCTL_RDMTS_HALF |
3050 	    (adapter->hw.mac.mc_filter_type << E1000_RCTL_MO_SHIFT);
3051 
3052 	/* Do not Store bad packets */
3053 	rctl &= ~E1000_RCTL_SBP;
3054 
3055 	/* Enable Long Packet receive */
3056 	if (adapter->netdev->mtu <= ETH_DATA_LEN)
3057 		rctl &= ~E1000_RCTL_LPE;
3058 	else
3059 		rctl |= E1000_RCTL_LPE;
3060 
3061 	/* Some systems expect that the CRC is included in SMBUS traffic. The
3062 	 * hardware strips the CRC before sending to both SMBUS (BMC) and to
3063 	 * host memory when this is enabled
3064 	 */
3065 	if (adapter->flags2 & FLAG2_CRC_STRIPPING)
3066 		rctl |= E1000_RCTL_SECRC;
3067 
3068 	/* Workaround Si errata on 82577 PHY - configure IPG for jumbos */
3069 	if ((hw->phy.type == e1000_phy_82577) && (rctl & E1000_RCTL_LPE)) {
3070 		u16 phy_data;
3071 
3072 		e1e_rphy(hw, PHY_REG(770, 26), &phy_data);
3073 		phy_data &= 0xfff8;
3074 		phy_data |= BIT(2);
3075 		e1e_wphy(hw, PHY_REG(770, 26), phy_data);
3076 
3077 		e1e_rphy(hw, 22, &phy_data);
3078 		phy_data &= 0x0fff;
3079 		phy_data |= BIT(14);
3080 		e1e_wphy(hw, 0x10, 0x2823);
3081 		e1e_wphy(hw, 0x11, 0x0003);
3082 		e1e_wphy(hw, 22, phy_data);
3083 	}
3084 
3085 	/* Setup buffer sizes */
3086 	rctl &= ~E1000_RCTL_SZ_4096;
3087 	rctl |= E1000_RCTL_BSEX;
3088 	switch (adapter->rx_buffer_len) {
3089 	case 2048:
3090 	default:
3091 		rctl |= E1000_RCTL_SZ_2048;
3092 		rctl &= ~E1000_RCTL_BSEX;
3093 		break;
3094 	case 4096:
3095 		rctl |= E1000_RCTL_SZ_4096;
3096 		break;
3097 	case 8192:
3098 		rctl |= E1000_RCTL_SZ_8192;
3099 		break;
3100 	case 16384:
3101 		rctl |= E1000_RCTL_SZ_16384;
3102 		break;
3103 	}
3104 
3105 	/* Enable Extended Status in all Receive Descriptors */
3106 	rfctl = er32(RFCTL);
3107 	rfctl |= E1000_RFCTL_EXTEN;
3108 	ew32(RFCTL, rfctl);
3109 
3110 	/* 82571 and greater support packet-split where the protocol
3111 	 * header is placed in skb->data and the packet data is
3112 	 * placed in pages hanging off of skb_shinfo(skb)->nr_frags.
3113 	 * In the case of a non-split, skb->data is linearly filled,
3114 	 * followed by the page buffers.  Therefore, skb->data is
3115 	 * sized to hold the largest protocol header.
3116 	 *
3117 	 * allocations using alloc_page take too long for regular MTU
3118 	 * so only enable packet split for jumbo frames
3119 	 *
3120 	 * Using pages when the page size is greater than 16k wastes
3121 	 * a lot of memory, since we allocate 3 pages at all times
3122 	 * per packet.
3123 	 */
3124 	pages = PAGE_USE_COUNT(adapter->netdev->mtu);
3125 	if ((pages <= 3) && (PAGE_SIZE <= 16384) && (rctl & E1000_RCTL_LPE))
3126 		adapter->rx_ps_pages = pages;
3127 	else
3128 		adapter->rx_ps_pages = 0;
3129 
3130 	if (adapter->rx_ps_pages) {
3131 		u32 psrctl = 0;
3132 
3133 		/* Enable Packet split descriptors */
3134 		rctl |= E1000_RCTL_DTYP_PS;
3135 
3136 		psrctl |= adapter->rx_ps_bsize0 >> E1000_PSRCTL_BSIZE0_SHIFT;
3137 
3138 		switch (adapter->rx_ps_pages) {
3139 		case 3:
3140 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE3_SHIFT;
3141 			fallthrough;
3142 		case 2:
3143 			psrctl |= PAGE_SIZE << E1000_PSRCTL_BSIZE2_SHIFT;
3144 			fallthrough;
3145 		case 1:
3146 			psrctl |= PAGE_SIZE >> E1000_PSRCTL_BSIZE1_SHIFT;
3147 			break;
3148 		}
3149 
3150 		ew32(PSRCTL, psrctl);
3151 	}
3152 
3153 	/* This is useful for sniffing bad packets. */
3154 	if (adapter->netdev->features & NETIF_F_RXALL) {
3155 		/* UPE and MPE will be handled by normal PROMISC logic
3156 		 * in e1000e_set_rx_mode
3157 		 */
3158 		rctl |= (E1000_RCTL_SBP |	/* Receive bad packets */
3159 			 E1000_RCTL_BAM |	/* RX All Bcast Pkts */
3160 			 E1000_RCTL_PMCF);	/* RX All MAC Ctrl Pkts */
3161 
3162 		rctl &= ~(E1000_RCTL_VFE |	/* Disable VLAN filter */
3163 			  E1000_RCTL_DPF |	/* Allow filtered pause */
3164 			  E1000_RCTL_CFIEN);	/* Dis VLAN CFIEN Filter */
3165 		/* Do not mess with E1000_CTRL_VME, it affects transmit as well,
3166 		 * and that breaks VLANs.
3167 		 */
3168 	}
3169 
3170 	ew32(RCTL, rctl);
3171 	/* just started the receive unit, no need to restart */
3172 	adapter->flags &= ~FLAG_RESTART_NOW;
3173 }
3174 
3175 /**
3176  * e1000_configure_rx - Configure Receive Unit after Reset
3177  * @adapter: board private structure
3178  *
3179  * Configure the Rx unit of the MAC after a reset.
3180  **/
e1000_configure_rx(struct e1000_adapter * adapter)3181 static void e1000_configure_rx(struct e1000_adapter *adapter)
3182 {
3183 	struct e1000_hw *hw = &adapter->hw;
3184 	struct e1000_ring *rx_ring = adapter->rx_ring;
3185 	u64 rdba;
3186 	u32 rdlen, rctl, rxcsum, ctrl_ext;
3187 
3188 	if (adapter->rx_ps_pages) {
3189 		/* this is a 32 byte descriptor */
3190 		rdlen = rx_ring->count *
3191 		    sizeof(union e1000_rx_desc_packet_split);
3192 		adapter->clean_rx = e1000_clean_rx_irq_ps;
3193 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers_ps;
3194 	} else if (adapter->netdev->mtu > ETH_FRAME_LEN + ETH_FCS_LEN) {
3195 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3196 		adapter->clean_rx = e1000_clean_jumbo_rx_irq;
3197 		adapter->alloc_rx_buf = e1000_alloc_jumbo_rx_buffers;
3198 	} else {
3199 		rdlen = rx_ring->count * sizeof(union e1000_rx_desc_extended);
3200 		adapter->clean_rx = e1000_clean_rx_irq;
3201 		adapter->alloc_rx_buf = e1000_alloc_rx_buffers;
3202 	}
3203 
3204 	/* disable receives while setting up the descriptors */
3205 	rctl = er32(RCTL);
3206 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
3207 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
3208 	e1e_flush();
3209 	usleep_range(10000, 11000);
3210 
3211 	if (adapter->flags2 & FLAG2_DMA_BURST) {
3212 		/* set the writeback threshold (only takes effect if the RDTR
3213 		 * is set). set GRAN=1 and write back up to 0x4 worth, and
3214 		 * enable prefetching of 0x20 Rx descriptors
3215 		 * granularity = 01
3216 		 * wthresh = 04,
3217 		 * hthresh = 04,
3218 		 * pthresh = 0x20
3219 		 */
3220 		ew32(RXDCTL(0), E1000_RXDCTL_DMA_BURST_ENABLE);
3221 		ew32(RXDCTL(1), E1000_RXDCTL_DMA_BURST_ENABLE);
3222 	}
3223 
3224 	/* set the Receive Delay Timer Register */
3225 	ew32(RDTR, adapter->rx_int_delay);
3226 
3227 	/* irq moderation */
3228 	ew32(RADV, adapter->rx_abs_int_delay);
3229 	if ((adapter->itr_setting != 0) && (adapter->itr != 0))
3230 		e1000e_write_itr(adapter, adapter->itr);
3231 
3232 	ctrl_ext = er32(CTRL_EXT);
3233 	/* Auto-Mask interrupts upon ICR access */
3234 	ctrl_ext |= E1000_CTRL_EXT_IAME;
3235 	ew32(IAM, 0xffffffff);
3236 	ew32(CTRL_EXT, ctrl_ext);
3237 	e1e_flush();
3238 
3239 	/* Setup the HW Rx Head and Tail Descriptor Pointers and
3240 	 * the Base and Length of the Rx Descriptor Ring
3241 	 */
3242 	rdba = rx_ring->dma;
3243 	ew32(RDBAL(0), (rdba & DMA_BIT_MASK(32)));
3244 	ew32(RDBAH(0), (rdba >> 32));
3245 	ew32(RDLEN(0), rdlen);
3246 	ew32(RDH(0), 0);
3247 	ew32(RDT(0), 0);
3248 	rx_ring->head = adapter->hw.hw_addr + E1000_RDH(0);
3249 	rx_ring->tail = adapter->hw.hw_addr + E1000_RDT(0);
3250 
3251 	if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
3252 		e1000e_update_rdt_wa(rx_ring, 0);
3253 
3254 	/* Enable Receive Checksum Offload for TCP and UDP */
3255 	rxcsum = er32(RXCSUM);
3256 	if (adapter->netdev->features & NETIF_F_RXCSUM)
3257 		rxcsum |= E1000_RXCSUM_TUOFL;
3258 	else
3259 		rxcsum &= ~E1000_RXCSUM_TUOFL;
3260 	ew32(RXCSUM, rxcsum);
3261 
3262 	/* With jumbo frames, excessive C-state transition latencies result
3263 	 * in dropped transactions.
3264 	 */
3265 	if (adapter->netdev->mtu > ETH_DATA_LEN) {
3266 		u32 lat =
3267 		    ((er32(PBA) & E1000_PBA_RXA_MASK) * 1024 -
3268 		     adapter->max_frame_size) * 8 / 1000;
3269 
3270 		if (adapter->flags & FLAG_IS_ICH) {
3271 			u32 rxdctl = er32(RXDCTL(0));
3272 
3273 			ew32(RXDCTL(0), rxdctl | 0x3 | BIT(8));
3274 		}
3275 
3276 		dev_info(&adapter->pdev->dev,
3277 			 "Some CPU C-states have been disabled in order to enable jumbo frames\n");
3278 		cpu_latency_qos_update_request(&adapter->pm_qos_req, lat);
3279 	} else {
3280 		cpu_latency_qos_update_request(&adapter->pm_qos_req,
3281 					       PM_QOS_DEFAULT_VALUE);
3282 	}
3283 
3284 	/* Enable Receives */
3285 	ew32(RCTL, rctl);
3286 }
3287 
3288 /**
3289  * e1000e_write_mc_addr_list - write multicast addresses to MTA
3290  * @netdev: network interface device structure
3291  *
3292  * Writes multicast address list to the MTA hash table.
3293  * Returns: -ENOMEM on failure
3294  *                0 on no addresses written
3295  *                X on writing X addresses to MTA
3296  */
e1000e_write_mc_addr_list(struct net_device * netdev)3297 static int e1000e_write_mc_addr_list(struct net_device *netdev)
3298 {
3299 	struct e1000_adapter *adapter = netdev_priv(netdev);
3300 	struct e1000_hw *hw = &adapter->hw;
3301 	struct netdev_hw_addr *ha;
3302 	u8 *mta_list;
3303 	int i;
3304 
3305 	if (netdev_mc_empty(netdev)) {
3306 		/* nothing to program, so clear mc list */
3307 		hw->mac.ops.update_mc_addr_list(hw, NULL, 0);
3308 		return 0;
3309 	}
3310 
3311 	mta_list = kcalloc(netdev_mc_count(netdev), ETH_ALEN, GFP_ATOMIC);
3312 	if (!mta_list)
3313 		return -ENOMEM;
3314 
3315 	/* update_mc_addr_list expects a packed array of only addresses. */
3316 	i = 0;
3317 	netdev_for_each_mc_addr(ha, netdev)
3318 	    memcpy(mta_list + (i++ * ETH_ALEN), ha->addr, ETH_ALEN);
3319 
3320 	hw->mac.ops.update_mc_addr_list(hw, mta_list, i);
3321 	kfree(mta_list);
3322 
3323 	return netdev_mc_count(netdev);
3324 }
3325 
3326 /**
3327  * e1000e_write_uc_addr_list - write unicast addresses to RAR table
3328  * @netdev: network interface device structure
3329  *
3330  * Writes unicast address list to the RAR table.
3331  * Returns: -ENOMEM on failure/insufficient address space
3332  *                0 on no addresses written
3333  *                X on writing X addresses to the RAR table
3334  **/
e1000e_write_uc_addr_list(struct net_device * netdev)3335 static int e1000e_write_uc_addr_list(struct net_device *netdev)
3336 {
3337 	struct e1000_adapter *adapter = netdev_priv(netdev);
3338 	struct e1000_hw *hw = &adapter->hw;
3339 	unsigned int rar_entries;
3340 	int count = 0;
3341 
3342 	rar_entries = hw->mac.ops.rar_get_count(hw);
3343 
3344 	/* save a rar entry for our hardware address */
3345 	rar_entries--;
3346 
3347 	/* save a rar entry for the LAA workaround */
3348 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA)
3349 		rar_entries--;
3350 
3351 	/* return ENOMEM indicating insufficient memory for addresses */
3352 	if (netdev_uc_count(netdev) > rar_entries)
3353 		return -ENOMEM;
3354 
3355 	if (!netdev_uc_empty(netdev) && rar_entries) {
3356 		struct netdev_hw_addr *ha;
3357 
3358 		/* write the addresses in reverse order to avoid write
3359 		 * combining
3360 		 */
3361 		netdev_for_each_uc_addr(ha, netdev) {
3362 			int ret_val;
3363 
3364 			if (!rar_entries)
3365 				break;
3366 			ret_val = hw->mac.ops.rar_set(hw, ha->addr, rar_entries--);
3367 			if (ret_val < 0)
3368 				return -ENOMEM;
3369 			count++;
3370 		}
3371 	}
3372 
3373 	/* zero out the remaining RAR entries not used above */
3374 	for (; rar_entries > 0; rar_entries--) {
3375 		ew32(RAH(rar_entries), 0);
3376 		ew32(RAL(rar_entries), 0);
3377 	}
3378 	e1e_flush();
3379 
3380 	return count;
3381 }
3382 
3383 /**
3384  * e1000e_set_rx_mode - secondary unicast, Multicast and Promiscuous mode set
3385  * @netdev: network interface device structure
3386  *
3387  * The ndo_set_rx_mode entry point is called whenever the unicast or multicast
3388  * address list or the network interface flags are updated.  This routine is
3389  * responsible for configuring the hardware for proper unicast, multicast,
3390  * promiscuous mode, and all-multi behavior.
3391  **/
e1000e_set_rx_mode(struct net_device * netdev)3392 static void e1000e_set_rx_mode(struct net_device *netdev)
3393 {
3394 	struct e1000_adapter *adapter = netdev_priv(netdev);
3395 	struct e1000_hw *hw = &adapter->hw;
3396 	u32 rctl;
3397 
3398 	if (pm_runtime_suspended(netdev->dev.parent))
3399 		return;
3400 
3401 	/* Check for Promiscuous and All Multicast modes */
3402 	rctl = er32(RCTL);
3403 
3404 	/* clear the affected bits */
3405 	rctl &= ~(E1000_RCTL_UPE | E1000_RCTL_MPE);
3406 
3407 	if (netdev->flags & IFF_PROMISC) {
3408 		rctl |= (E1000_RCTL_UPE | E1000_RCTL_MPE);
3409 		/* Do not hardware filter VLANs in promisc mode */
3410 		e1000e_vlan_filter_disable(adapter);
3411 	} else {
3412 		int count;
3413 
3414 		if (netdev->flags & IFF_ALLMULTI) {
3415 			rctl |= E1000_RCTL_MPE;
3416 		} else {
3417 			/* Write addresses to the MTA, if the attempt fails
3418 			 * then we should just turn on promiscuous mode so
3419 			 * that we can at least receive multicast traffic
3420 			 */
3421 			count = e1000e_write_mc_addr_list(netdev);
3422 			if (count < 0)
3423 				rctl |= E1000_RCTL_MPE;
3424 		}
3425 		e1000e_vlan_filter_enable(adapter);
3426 		/* Write addresses to available RAR registers, if there is not
3427 		 * sufficient space to store all the addresses then enable
3428 		 * unicast promiscuous mode
3429 		 */
3430 		count = e1000e_write_uc_addr_list(netdev);
3431 		if (count < 0)
3432 			rctl |= E1000_RCTL_UPE;
3433 	}
3434 
3435 	ew32(RCTL, rctl);
3436 
3437 	if (netdev->features & NETIF_F_HW_VLAN_CTAG_RX)
3438 		e1000e_vlan_strip_enable(adapter);
3439 	else
3440 		e1000e_vlan_strip_disable(adapter);
3441 }
3442 
e1000e_setup_rss_hash(struct e1000_adapter * adapter)3443 static void e1000e_setup_rss_hash(struct e1000_adapter *adapter)
3444 {
3445 	struct e1000_hw *hw = &adapter->hw;
3446 	u32 mrqc, rxcsum;
3447 	u32 rss_key[10];
3448 	int i;
3449 
3450 	netdev_rss_key_fill(rss_key, sizeof(rss_key));
3451 	for (i = 0; i < 10; i++)
3452 		ew32(RSSRK(i), rss_key[i]);
3453 
3454 	/* Direct all traffic to queue 0 */
3455 	for (i = 0; i < 32; i++)
3456 		ew32(RETA(i), 0);
3457 
3458 	/* Disable raw packet checksumming so that RSS hash is placed in
3459 	 * descriptor on writeback.
3460 	 */
3461 	rxcsum = er32(RXCSUM);
3462 	rxcsum |= E1000_RXCSUM_PCSD;
3463 
3464 	ew32(RXCSUM, rxcsum);
3465 
3466 	mrqc = (E1000_MRQC_RSS_FIELD_IPV4 |
3467 		E1000_MRQC_RSS_FIELD_IPV4_TCP |
3468 		E1000_MRQC_RSS_FIELD_IPV6 |
3469 		E1000_MRQC_RSS_FIELD_IPV6_TCP |
3470 		E1000_MRQC_RSS_FIELD_IPV6_TCP_EX);
3471 
3472 	ew32(MRQC, mrqc);
3473 }
3474 
3475 /**
3476  * e1000e_get_base_timinca - get default SYSTIM time increment attributes
3477  * @adapter: board private structure
3478  * @timinca: pointer to returned time increment attributes
3479  *
3480  * Get attributes for incrementing the System Time Register SYSTIML/H at
3481  * the default base frequency, and set the cyclecounter shift value.
3482  **/
e1000e_get_base_timinca(struct e1000_adapter * adapter,u32 * timinca)3483 s32 e1000e_get_base_timinca(struct e1000_adapter *adapter, u32 *timinca)
3484 {
3485 	struct e1000_hw *hw = &adapter->hw;
3486 	u32 incvalue, incperiod, shift;
3487 
3488 	/* Make sure clock is enabled on I217/I218/I219  before checking
3489 	 * the frequency
3490 	 */
3491 	if ((hw->mac.type >= e1000_pch_lpt) &&
3492 	    !(er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) &&
3493 	    !(er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_ENABLED)) {
3494 		u32 fextnvm7 = er32(FEXTNVM7);
3495 
3496 		if (!(fextnvm7 & BIT(0))) {
3497 			ew32(FEXTNVM7, fextnvm7 | BIT(0));
3498 			e1e_flush();
3499 		}
3500 	}
3501 
3502 	switch (hw->mac.type) {
3503 	case e1000_pch2lan:
3504 		/* Stable 96MHz frequency */
3505 		incperiod = INCPERIOD_96MHZ;
3506 		incvalue = INCVALUE_96MHZ;
3507 		shift = INCVALUE_SHIFT_96MHZ;
3508 		adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3509 		break;
3510 	case e1000_pch_lpt:
3511 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3512 			/* Stable 96MHz frequency */
3513 			incperiod = INCPERIOD_96MHZ;
3514 			incvalue = INCVALUE_96MHZ;
3515 			shift = INCVALUE_SHIFT_96MHZ;
3516 			adapter->cc.shift = shift + INCPERIOD_SHIFT_96MHZ;
3517 		} else {
3518 			/* Stable 25MHz frequency */
3519 			incperiod = INCPERIOD_25MHZ;
3520 			incvalue = INCVALUE_25MHZ;
3521 			shift = INCVALUE_SHIFT_25MHZ;
3522 			adapter->cc.shift = shift;
3523 		}
3524 		break;
3525 	case e1000_pch_spt:
3526 		/* Stable 24MHz frequency */
3527 		incperiod = INCPERIOD_24MHZ;
3528 		incvalue = INCVALUE_24MHZ;
3529 		shift = INCVALUE_SHIFT_24MHZ;
3530 		adapter->cc.shift = shift;
3531 		break;
3532 	case e1000_pch_cnp:
3533 	case e1000_pch_tgp:
3534 	case e1000_pch_adp:
3535 	case e1000_pch_nvp:
3536 		if (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_SYSCFI) {
3537 			/* Stable 24MHz frequency */
3538 			incperiod = INCPERIOD_24MHZ;
3539 			incvalue = INCVALUE_24MHZ;
3540 			shift = INCVALUE_SHIFT_24MHZ;
3541 			adapter->cc.shift = shift;
3542 		} else {
3543 			/* Stable 38400KHz frequency */
3544 			incperiod = INCPERIOD_38400KHZ;
3545 			incvalue = INCVALUE_38400KHZ;
3546 			shift = INCVALUE_SHIFT_38400KHZ;
3547 			adapter->cc.shift = shift;
3548 		}
3549 		break;
3550 	case e1000_pch_mtp:
3551 	case e1000_pch_lnp:
3552 	case e1000_pch_ptp:
3553 		/* System firmware can misreport this value, so set it to a
3554 		 * stable 38400KHz frequency.
3555 		 */
3556 		incperiod = INCPERIOD_38400KHZ;
3557 		incvalue = INCVALUE_38400KHZ;
3558 		shift = INCVALUE_SHIFT_38400KHZ;
3559 		adapter->cc.shift = shift;
3560 		break;
3561 	case e1000_82574:
3562 	case e1000_82583:
3563 		/* Stable 25MHz frequency */
3564 		incperiod = INCPERIOD_25MHZ;
3565 		incvalue = INCVALUE_25MHZ;
3566 		shift = INCVALUE_SHIFT_25MHZ;
3567 		adapter->cc.shift = shift;
3568 		break;
3569 	default:
3570 		return -EINVAL;
3571 	}
3572 
3573 	*timinca = ((incperiod << E1000_TIMINCA_INCPERIOD_SHIFT) |
3574 		    ((incvalue << shift) & E1000_TIMINCA_INCVALUE_MASK));
3575 
3576 	return 0;
3577 }
3578 
3579 /**
3580  * e1000e_config_hwtstamp - configure the hwtstamp registers and enable/disable
3581  * @adapter: board private structure
3582  * @config: timestamp configuration
3583  * @extack: netlink extended ACK for error report
3584  *
3585  * Outgoing time stamping can be enabled and disabled. Play nice and
3586  * disable it when requested, although it shouldn't cause any overhead
3587  * when no packet needs it. At most one packet in the queue may be
3588  * marked for time stamping, otherwise it would be impossible to tell
3589  * for sure to which packet the hardware time stamp belongs.
3590  *
3591  * Incoming time stamping has to be configured via the hardware filters.
3592  * Not all combinations are supported, in particular event type has to be
3593  * specified. Matching the kind of event packet is not supported, with the
3594  * exception of "all V2 events regardless of level 2 or 4".
3595  **/
e1000e_config_hwtstamp(struct e1000_adapter * adapter,struct kernel_hwtstamp_config * config,struct netlink_ext_ack * extack)3596 static int e1000e_config_hwtstamp(struct e1000_adapter *adapter,
3597 				  struct kernel_hwtstamp_config *config,
3598 				  struct netlink_ext_ack *extack)
3599 {
3600 	struct e1000_hw *hw = &adapter->hw;
3601 	u32 tsync_tx_ctl = E1000_TSYNCTXCTL_ENABLED;
3602 	u32 tsync_rx_ctl = E1000_TSYNCRXCTL_ENABLED;
3603 	u32 rxmtrl = 0;
3604 	u16 rxudp = 0;
3605 	bool is_l4 = false;
3606 	bool is_l2 = false;
3607 	u32 regval;
3608 
3609 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
3610 		NL_SET_ERR_MSG(extack, "No HW timestamp support");
3611 		return -EINVAL;
3612 	}
3613 
3614 	switch (config->tx_type) {
3615 	case HWTSTAMP_TX_OFF:
3616 		tsync_tx_ctl = 0;
3617 		break;
3618 	case HWTSTAMP_TX_ON:
3619 		break;
3620 	default:
3621 		NL_SET_ERR_MSG(extack, "Unsupported TX HW timestamp type");
3622 		return -ERANGE;
3623 	}
3624 
3625 	switch (config->rx_filter) {
3626 	case HWTSTAMP_FILTER_NONE:
3627 		tsync_rx_ctl = 0;
3628 		break;
3629 	case HWTSTAMP_FILTER_PTP_V1_L4_SYNC:
3630 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3631 		rxmtrl = E1000_RXMTRL_PTP_V1_SYNC_MESSAGE;
3632 		is_l4 = true;
3633 		break;
3634 	case HWTSTAMP_FILTER_PTP_V1_L4_DELAY_REQ:
3635 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L4_V1;
3636 		rxmtrl = E1000_RXMTRL_PTP_V1_DELAY_REQ_MESSAGE;
3637 		is_l4 = true;
3638 		break;
3639 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
3640 		/* Also time stamps V2 L2 Path Delay Request/Response */
3641 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3642 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3643 		is_l2 = true;
3644 		break;
3645 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
3646 		/* Also time stamps V2 L2 Path Delay Request/Response. */
3647 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_V2;
3648 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3649 		is_l2 = true;
3650 		break;
3651 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
3652 		/* Hardware cannot filter just V2 L4 Sync messages */
3653 		fallthrough;
3654 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
3655 		/* Also time stamps V2 Path Delay Request/Response. */
3656 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3657 		rxmtrl = E1000_RXMTRL_PTP_V2_SYNC_MESSAGE;
3658 		is_l2 = true;
3659 		is_l4 = true;
3660 		break;
3661 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
3662 		/* Hardware cannot filter just V2 L4 Delay Request messages */
3663 		fallthrough;
3664 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
3665 		/* Also time stamps V2 Path Delay Request/Response. */
3666 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_L2_L4_V2;
3667 		rxmtrl = E1000_RXMTRL_PTP_V2_DELAY_REQ_MESSAGE;
3668 		is_l2 = true;
3669 		is_l4 = true;
3670 		break;
3671 	case HWTSTAMP_FILTER_PTP_V2_L4_EVENT:
3672 	case HWTSTAMP_FILTER_PTP_V2_L2_EVENT:
3673 		/* Hardware cannot filter just V2 L4 or L2 Event messages */
3674 		fallthrough;
3675 	case HWTSTAMP_FILTER_PTP_V2_EVENT:
3676 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_EVENT_V2;
3677 		config->rx_filter = HWTSTAMP_FILTER_PTP_V2_EVENT;
3678 		is_l2 = true;
3679 		is_l4 = true;
3680 		break;
3681 	case HWTSTAMP_FILTER_PTP_V1_L4_EVENT:
3682 		/* For V1, the hardware can only filter Sync messages or
3683 		 * Delay Request messages but not both so fall-through to
3684 		 * time stamp all packets.
3685 		 */
3686 		fallthrough;
3687 	case HWTSTAMP_FILTER_NTP_ALL:
3688 	case HWTSTAMP_FILTER_ALL:
3689 		is_l2 = true;
3690 		is_l4 = true;
3691 		tsync_rx_ctl |= E1000_TSYNCRXCTL_TYPE_ALL;
3692 		config->rx_filter = HWTSTAMP_FILTER_ALL;
3693 		break;
3694 	default:
3695 		NL_SET_ERR_MSG(extack, "Unsupported RX HW timestamp filter");
3696 		return -ERANGE;
3697 	}
3698 
3699 	adapter->hwtstamp_config = *config;
3700 
3701 	/* enable/disable Tx h/w time stamping */
3702 	regval = er32(TSYNCTXCTL);
3703 	regval &= ~E1000_TSYNCTXCTL_ENABLED;
3704 	regval |= tsync_tx_ctl;
3705 	ew32(TSYNCTXCTL, regval);
3706 	if ((er32(TSYNCTXCTL) & E1000_TSYNCTXCTL_ENABLED) !=
3707 	    (regval & E1000_TSYNCTXCTL_ENABLED)) {
3708 		NL_SET_ERR_MSG(extack,
3709 			       "Timesync Tx Control register not set as expected");
3710 		return -EAGAIN;
3711 	}
3712 
3713 	/* enable/disable Rx h/w time stamping */
3714 	regval = er32(TSYNCRXCTL);
3715 	regval &= ~(E1000_TSYNCRXCTL_ENABLED | E1000_TSYNCRXCTL_TYPE_MASK);
3716 	regval |= tsync_rx_ctl;
3717 	ew32(TSYNCRXCTL, regval);
3718 	if ((er32(TSYNCRXCTL) & (E1000_TSYNCRXCTL_ENABLED |
3719 				 E1000_TSYNCRXCTL_TYPE_MASK)) !=
3720 	    (regval & (E1000_TSYNCRXCTL_ENABLED |
3721 		       E1000_TSYNCRXCTL_TYPE_MASK))) {
3722 		NL_SET_ERR_MSG(extack,
3723 			       "Timesync Rx Control register not set as expected");
3724 		return -EAGAIN;
3725 	}
3726 
3727 	/* L2: define ethertype filter for time stamped packets */
3728 	if (is_l2)
3729 		rxmtrl |= ETH_P_1588;
3730 
3731 	/* define which PTP packets get time stamped */
3732 	ew32(RXMTRL, rxmtrl);
3733 
3734 	/* Filter by destination port */
3735 	if (is_l4) {
3736 		rxudp = PTP_EV_PORT;
3737 		cpu_to_be16s(&rxudp);
3738 	}
3739 	ew32(RXUDP, rxudp);
3740 
3741 	e1e_flush();
3742 
3743 	/* Clear TSYNCRXCTL_VALID & TSYNCTXCTL_VALID bit */
3744 	er32(RXSTMPH);
3745 	er32(TXSTMPH);
3746 
3747 	return 0;
3748 }
3749 
3750 /**
3751  * e1000_configure - configure the hardware for Rx and Tx
3752  * @adapter: private board structure
3753  **/
e1000_configure(struct e1000_adapter * adapter)3754 static void e1000_configure(struct e1000_adapter *adapter)
3755 {
3756 	struct e1000_ring *rx_ring = adapter->rx_ring;
3757 
3758 	e1000e_set_rx_mode(adapter->netdev);
3759 
3760 	e1000_restore_vlan(adapter);
3761 	e1000_init_manageability_pt(adapter);
3762 
3763 	e1000_configure_tx(adapter);
3764 
3765 	if (adapter->netdev->features & NETIF_F_RXHASH)
3766 		e1000e_setup_rss_hash(adapter);
3767 	e1000_setup_rctl(adapter);
3768 	e1000_configure_rx(adapter);
3769 	adapter->alloc_rx_buf(rx_ring, e1000_desc_unused(rx_ring), GFP_KERNEL);
3770 }
3771 
3772 /**
3773  * e1000e_power_up_phy - restore link in case the phy was powered down
3774  * @adapter: address of board private structure
3775  *
3776  * The phy may be powered down to save power and turn off link when the
3777  * driver is unloaded and wake on lan is not enabled (among others)
3778  * *** this routine MUST be followed by a call to e1000e_reset ***
3779  **/
e1000e_power_up_phy(struct e1000_adapter * adapter)3780 void e1000e_power_up_phy(struct e1000_adapter *adapter)
3781 {
3782 	if (adapter->hw.phy.ops.power_up)
3783 		adapter->hw.phy.ops.power_up(&adapter->hw);
3784 
3785 	adapter->hw.mac.ops.setup_link(&adapter->hw);
3786 }
3787 
3788 /**
3789  * e1000_power_down_phy - Power down the PHY
3790  * @adapter: board private structure
3791  *
3792  * Power down the PHY so no link is implied when interface is down.
3793  * The PHY cannot be powered down if management or WoL is active.
3794  */
e1000_power_down_phy(struct e1000_adapter * adapter)3795 static void e1000_power_down_phy(struct e1000_adapter *adapter)
3796 {
3797 	if (adapter->hw.phy.ops.power_down)
3798 		adapter->hw.phy.ops.power_down(&adapter->hw);
3799 }
3800 
3801 /**
3802  * e1000_flush_tx_ring - remove all descriptors from the tx_ring
3803  * @adapter: board private structure
3804  *
3805  * We want to clear all pending descriptors from the TX ring.
3806  * zeroing happens when the HW reads the regs. We  assign the ring itself as
3807  * the data of the next descriptor. We don't care about the data we are about
3808  * to reset the HW.
3809  */
e1000_flush_tx_ring(struct e1000_adapter * adapter)3810 static void e1000_flush_tx_ring(struct e1000_adapter *adapter)
3811 {
3812 	struct e1000_hw *hw = &adapter->hw;
3813 	struct e1000_ring *tx_ring = adapter->tx_ring;
3814 	struct e1000_tx_desc *tx_desc = NULL;
3815 	u32 tdt, tctl, txd_lower = E1000_TXD_CMD_IFCS;
3816 	u16 size = 512;
3817 
3818 	tctl = er32(TCTL);
3819 	ew32(TCTL, tctl | E1000_TCTL_EN);
3820 	tdt = er32(TDT(0));
3821 	BUG_ON(tdt != tx_ring->next_to_use);
3822 	tx_desc =  E1000_TX_DESC(*tx_ring, tx_ring->next_to_use);
3823 	tx_desc->buffer_addr = cpu_to_le64(tx_ring->dma);
3824 
3825 	tx_desc->lower.data = cpu_to_le32(txd_lower | size);
3826 	tx_desc->upper.data = 0;
3827 	/* flush descriptors to memory before notifying the HW */
3828 	wmb();
3829 	tx_ring->next_to_use++;
3830 	if (tx_ring->next_to_use == tx_ring->count)
3831 		tx_ring->next_to_use = 0;
3832 	ew32(TDT(0), tx_ring->next_to_use);
3833 	usleep_range(200, 250);
3834 }
3835 
3836 /**
3837  * e1000_flush_rx_ring - remove all descriptors from the rx_ring
3838  * @adapter: board private structure
3839  *
3840  * Mark all descriptors in the RX ring as consumed and disable the rx ring
3841  */
e1000_flush_rx_ring(struct e1000_adapter * adapter)3842 static void e1000_flush_rx_ring(struct e1000_adapter *adapter)
3843 {
3844 	u32 rctl, rxdctl;
3845 	struct e1000_hw *hw = &adapter->hw;
3846 
3847 	rctl = er32(RCTL);
3848 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3849 	e1e_flush();
3850 	usleep_range(100, 150);
3851 
3852 	rxdctl = er32(RXDCTL(0));
3853 	/* zero the lower 14 bits (prefetch and host thresholds) */
3854 	rxdctl &= 0xffffc000;
3855 
3856 	/* update thresholds: prefetch threshold to 31, host threshold to 1
3857 	 * and make sure the granularity is "descriptors" and not "cache lines"
3858 	 */
3859 	rxdctl |= (0x1F | BIT(8) | E1000_RXDCTL_THRESH_UNIT_DESC);
3860 
3861 	ew32(RXDCTL(0), rxdctl);
3862 	/* momentarily enable the RX ring for the changes to take effect */
3863 	ew32(RCTL, rctl | E1000_RCTL_EN);
3864 	e1e_flush();
3865 	usleep_range(100, 150);
3866 	ew32(RCTL, rctl & ~E1000_RCTL_EN);
3867 }
3868 
3869 /**
3870  * e1000_flush_desc_rings - remove all descriptors from the descriptor rings
3871  * @adapter: board private structure
3872  *
3873  * In i219, the descriptor rings must be emptied before resetting the HW
3874  * or before changing the device state to D3 during runtime (runtime PM).
3875  *
3876  * Failure to do this will cause the HW to enter a unit hang state which can
3877  * only be released by PCI reset on the device
3878  *
3879  */
3880 
e1000_flush_desc_rings(struct e1000_adapter * adapter)3881 static void e1000_flush_desc_rings(struct e1000_adapter *adapter)
3882 {
3883 	u16 hang_state;
3884 	u32 fext_nvm11, tdlen;
3885 	struct e1000_hw *hw = &adapter->hw;
3886 
3887 	/* First, disable MULR fix in FEXTNVM11 */
3888 	fext_nvm11 = er32(FEXTNVM11);
3889 	fext_nvm11 |= E1000_FEXTNVM11_DISABLE_MULR_FIX;
3890 	ew32(FEXTNVM11, fext_nvm11);
3891 	/* do nothing if we're not in faulty state, or if the queue is empty */
3892 	tdlen = er32(TDLEN(0));
3893 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3894 			     &hang_state);
3895 	if (!(hang_state & FLUSH_DESC_REQUIRED) || !tdlen)
3896 		return;
3897 	e1000_flush_tx_ring(adapter);
3898 	/* recheck, maybe the fault is caused by the rx ring */
3899 	pci_read_config_word(adapter->pdev, PCICFG_DESC_RING_STATUS,
3900 			     &hang_state);
3901 	if (hang_state & FLUSH_DESC_REQUIRED)
3902 		e1000_flush_rx_ring(adapter);
3903 }
3904 
3905 /**
3906  * e1000e_systim_reset - reset the timesync registers after a hardware reset
3907  * @adapter: board private structure
3908  *
3909  * When the MAC is reset, all hardware bits for timesync will be reset to the
3910  * default values. This function will restore the settings last in place.
3911  * Since the clock SYSTIME registers are reset, we will simply restore the
3912  * cyclecounter to the kernel real clock time.
3913  **/
e1000e_systim_reset(struct e1000_adapter * adapter)3914 static void e1000e_systim_reset(struct e1000_adapter *adapter)
3915 {
3916 	struct ptp_clock_info *info = &adapter->ptp_clock_info;
3917 	struct e1000_hw *hw = &adapter->hw;
3918 	struct netlink_ext_ack extack = {};
3919 	unsigned long flags;
3920 	u32 timinca;
3921 	s32 ret_val;
3922 
3923 	if (!(adapter->flags & FLAG_HAS_HW_TIMESTAMP))
3924 		return;
3925 
3926 	if (info->adjfine) {
3927 		/* restore the previous ptp frequency delta */
3928 		ret_val = info->adjfine(info, adapter->ptp_delta);
3929 	} else {
3930 		/* set the default base frequency if no adjustment possible */
3931 		ret_val = e1000e_get_base_timinca(adapter, &timinca);
3932 		if (!ret_val)
3933 			ew32(TIMINCA, timinca);
3934 	}
3935 
3936 	if (ret_val) {
3937 		dev_warn(&adapter->pdev->dev,
3938 			 "Failed to restore TIMINCA clock rate delta: %d\n",
3939 			 ret_val);
3940 		return;
3941 	}
3942 
3943 	/* reset the systim ns time counter */
3944 	spin_lock_irqsave(&adapter->systim_lock, flags);
3945 	timecounter_init(&adapter->tc, &adapter->cc,
3946 			 ktime_to_ns(ktime_get_real()));
3947 	spin_unlock_irqrestore(&adapter->systim_lock, flags);
3948 
3949 	/* restore the previous hwtstamp configuration settings */
3950 	ret_val = e1000e_config_hwtstamp(adapter, &adapter->hwtstamp_config,
3951 					 &extack);
3952 	if (ret_val) {
3953 		if (extack._msg)
3954 			e_err("%s\n", extack._msg);
3955 	}
3956 }
3957 
3958 /**
3959  * e1000e_reset - bring the hardware into a known good state
3960  * @adapter: board private structure
3961  *
3962  * This function boots the hardware and enables some settings that
3963  * require a configuration cycle of the hardware - those cannot be
3964  * set/changed during runtime. After reset the device needs to be
3965  * properly configured for Rx, Tx etc.
3966  */
e1000e_reset(struct e1000_adapter * adapter)3967 void e1000e_reset(struct e1000_adapter *adapter)
3968 {
3969 	struct e1000_mac_info *mac = &adapter->hw.mac;
3970 	struct e1000_fc_info *fc = &adapter->hw.fc;
3971 	struct e1000_hw *hw = &adapter->hw;
3972 	u32 tx_space, min_tx_space, min_rx_space;
3973 	u32 pba = adapter->pba;
3974 	u16 hwm;
3975 
3976 	/* reset Packet Buffer Allocation to default */
3977 	ew32(PBA, pba);
3978 
3979 	if (adapter->max_frame_size > (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN)) {
3980 		/* To maintain wire speed transmits, the Tx FIFO should be
3981 		 * large enough to accommodate two full transmit packets,
3982 		 * rounded up to the next 1KB and expressed in KB.  Likewise,
3983 		 * the Rx FIFO should be large enough to accommodate at least
3984 		 * one full receive packet and is similarly rounded up and
3985 		 * expressed in KB.
3986 		 */
3987 		pba = er32(PBA);
3988 		/* upper 16 bits has Tx packet buffer allocation size in KB */
3989 		tx_space = pba >> 16;
3990 		/* lower 16 bits has Rx packet buffer allocation size in KB */
3991 		pba &= 0xffff;
3992 		/* the Tx fifo also stores 16 bytes of information about the Tx
3993 		 * but don't include ethernet FCS because hardware appends it
3994 		 */
3995 		min_tx_space = (adapter->max_frame_size +
3996 				sizeof(struct e1000_tx_desc) - ETH_FCS_LEN) * 2;
3997 		min_tx_space = ALIGN(min_tx_space, 1024);
3998 		min_tx_space >>= 10;
3999 		/* software strips receive CRC, so leave room for it */
4000 		min_rx_space = adapter->max_frame_size;
4001 		min_rx_space = ALIGN(min_rx_space, 1024);
4002 		min_rx_space >>= 10;
4003 
4004 		/* If current Tx allocation is less than the min Tx FIFO size,
4005 		 * and the min Tx FIFO size is less than the current Rx FIFO
4006 		 * allocation, take space away from current Rx allocation
4007 		 */
4008 		if ((tx_space < min_tx_space) &&
4009 		    ((min_tx_space - tx_space) < pba)) {
4010 			pba -= min_tx_space - tx_space;
4011 
4012 			/* if short on Rx space, Rx wins and must trump Tx
4013 			 * adjustment
4014 			 */
4015 			if (pba < min_rx_space)
4016 				pba = min_rx_space;
4017 		}
4018 
4019 		ew32(PBA, pba);
4020 	}
4021 
4022 	/* flow control settings
4023 	 *
4024 	 * The high water mark must be low enough to fit one full frame
4025 	 * (or the size used for early receive) above it in the Rx FIFO.
4026 	 * Set it to the lower of:
4027 	 * - 90% of the Rx FIFO size, and
4028 	 * - the full Rx FIFO size minus one full frame
4029 	 */
4030 	if (adapter->flags & FLAG_DISABLE_FC_PAUSE_TIME)
4031 		fc->pause_time = 0xFFFF;
4032 	else
4033 		fc->pause_time = E1000_FC_PAUSE_TIME;
4034 	fc->send_xon = true;
4035 	fc->current_mode = fc->requested_mode;
4036 
4037 	switch (hw->mac.type) {
4038 	case e1000_ich9lan:
4039 	case e1000_ich10lan:
4040 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4041 			pba = 14;
4042 			ew32(PBA, pba);
4043 			fc->high_water = 0x2800;
4044 			fc->low_water = fc->high_water - 8;
4045 			break;
4046 		}
4047 		fallthrough;
4048 	default:
4049 		hwm = min(((pba << 10) * 9 / 10),
4050 			  ((pba << 10) - adapter->max_frame_size));
4051 
4052 		fc->high_water = hwm & E1000_FCRTH_RTH;	/* 8-byte granularity */
4053 		fc->low_water = fc->high_water - 8;
4054 		break;
4055 	case e1000_pchlan:
4056 		/* Workaround PCH LOM adapter hangs with certain network
4057 		 * loads.  If hangs persist, try disabling Tx flow control.
4058 		 */
4059 		if (adapter->netdev->mtu > ETH_DATA_LEN) {
4060 			fc->high_water = 0x3500;
4061 			fc->low_water = 0x1500;
4062 		} else {
4063 			fc->high_water = 0x5000;
4064 			fc->low_water = 0x3000;
4065 		}
4066 		fc->refresh_time = 0x1000;
4067 		break;
4068 	case e1000_pch2lan:
4069 	case e1000_pch_lpt:
4070 	case e1000_pch_spt:
4071 	case e1000_pch_cnp:
4072 	case e1000_pch_tgp:
4073 	case e1000_pch_adp:
4074 	case e1000_pch_mtp:
4075 	case e1000_pch_lnp:
4076 	case e1000_pch_ptp:
4077 	case e1000_pch_nvp:
4078 		fc->refresh_time = 0xFFFF;
4079 		fc->pause_time = 0xFFFF;
4080 
4081 		if (adapter->netdev->mtu <= ETH_DATA_LEN) {
4082 			fc->high_water = 0x05C20;
4083 			fc->low_water = 0x05048;
4084 			break;
4085 		}
4086 
4087 		pba = 14;
4088 		ew32(PBA, pba);
4089 		fc->high_water = ((pba << 10) * 9 / 10) & E1000_FCRTH_RTH;
4090 		fc->low_water = ((pba << 10) * 8 / 10) & E1000_FCRTL_RTL;
4091 		break;
4092 	}
4093 
4094 	/* Alignment of Tx data is on an arbitrary byte boundary with the
4095 	 * maximum size per Tx descriptor limited only to the transmit
4096 	 * allocation of the packet buffer minus 96 bytes with an upper
4097 	 * limit of 24KB due to receive synchronization limitations.
4098 	 */
4099 	adapter->tx_fifo_limit = min_t(u32, ((er32(PBA) >> 16) << 10) - 96,
4100 				       24 << 10);
4101 
4102 	/* Disable Adaptive Interrupt Moderation if 2 full packets cannot
4103 	 * fit in receive buffer.
4104 	 */
4105 	if (adapter->itr_setting & 0x3) {
4106 		if ((adapter->max_frame_size * 2) > (pba << 10)) {
4107 			if (!(adapter->flags2 & FLAG2_DISABLE_AIM)) {
4108 				dev_info(&adapter->pdev->dev,
4109 					 "Interrupt Throttle Rate off\n");
4110 				adapter->flags2 |= FLAG2_DISABLE_AIM;
4111 				e1000e_write_itr(adapter, 0);
4112 			}
4113 		} else if (adapter->flags2 & FLAG2_DISABLE_AIM) {
4114 			dev_info(&adapter->pdev->dev,
4115 				 "Interrupt Throttle Rate on\n");
4116 			adapter->flags2 &= ~FLAG2_DISABLE_AIM;
4117 			adapter->itr = 20000;
4118 			e1000e_write_itr(adapter, adapter->itr);
4119 		}
4120 	}
4121 
4122 	if (hw->mac.type >= e1000_pch_spt)
4123 		e1000_flush_desc_rings(adapter);
4124 	/* Allow time for pending master requests to run */
4125 	mac->ops.reset_hw(hw);
4126 
4127 	/* For parts with AMT enabled, let the firmware know
4128 	 * that the network interface is in control
4129 	 */
4130 	if (adapter->flags & FLAG_HAS_AMT)
4131 		e1000e_get_hw_control(adapter);
4132 
4133 	ew32(WUC, 0);
4134 
4135 	if (mac->ops.init_hw(hw))
4136 		e_err("Hardware Error\n");
4137 
4138 	e1000_update_mng_vlan(adapter);
4139 
4140 	/* Enable h/w to recognize an 802.1Q VLAN Ethernet packet */
4141 	ew32(VET, ETH_P_8021Q);
4142 
4143 	e1000e_reset_adaptive(hw);
4144 
4145 	/* restore systim and hwtstamp settings */
4146 	e1000e_systim_reset(adapter);
4147 
4148 	/* Set EEE advertisement as appropriate */
4149 	if (adapter->flags2 & FLAG2_HAS_EEE) {
4150 		s32 ret_val;
4151 		u16 adv_addr;
4152 
4153 		switch (hw->phy.type) {
4154 		case e1000_phy_82579:
4155 			adv_addr = I82579_EEE_ADVERTISEMENT;
4156 			break;
4157 		case e1000_phy_i217:
4158 			adv_addr = I217_EEE_ADVERTISEMENT;
4159 			break;
4160 		default:
4161 			dev_err(&adapter->pdev->dev,
4162 				"Invalid PHY type setting EEE advertisement\n");
4163 			return;
4164 		}
4165 
4166 		ret_val = hw->phy.ops.acquire(hw);
4167 		if (ret_val) {
4168 			dev_err(&adapter->pdev->dev,
4169 				"EEE advertisement - unable to acquire PHY\n");
4170 			return;
4171 		}
4172 
4173 		e1000_write_emi_reg_locked(hw, adv_addr,
4174 					   hw->dev_spec.ich8lan.eee_disable ?
4175 					   0 : adapter->eee_advert);
4176 
4177 		hw->phy.ops.release(hw);
4178 	}
4179 
4180 	if (!netif_running(adapter->netdev) &&
4181 	    !test_bit(__E1000_TESTING, &adapter->state))
4182 		e1000_power_down_phy(adapter);
4183 
4184 	e1000_get_phy_info(hw);
4185 
4186 	if ((adapter->flags & FLAG_HAS_SMART_POWER_DOWN) &&
4187 	    !(adapter->flags & FLAG_SMART_POWER_DOWN)) {
4188 		u16 phy_data = 0;
4189 		/* speed up time to link by disabling smart power down, ignore
4190 		 * the return value of this function because there is nothing
4191 		 * different we would do if it failed
4192 		 */
4193 		e1e_rphy(hw, IGP02E1000_PHY_POWER_MGMT, &phy_data);
4194 		phy_data &= ~IGP02E1000_PM_SPD;
4195 		e1e_wphy(hw, IGP02E1000_PHY_POWER_MGMT, phy_data);
4196 	}
4197 	if (hw->mac.type >= e1000_pch_spt && adapter->int_mode == 0) {
4198 		u32 reg;
4199 
4200 		/* Fextnvm7 @ 0xe4[2] = 1 */
4201 		reg = er32(FEXTNVM7);
4202 		reg |= E1000_FEXTNVM7_SIDE_CLK_UNGATE;
4203 		ew32(FEXTNVM7, reg);
4204 		/* Fextnvm9 @ 0x5bb4[13:12] = 11 */
4205 		reg = er32(FEXTNVM9);
4206 		reg |= E1000_FEXTNVM9_IOSFSB_CLKGATE_DIS |
4207 		       E1000_FEXTNVM9_IOSFSB_CLKREQ_DIS;
4208 		ew32(FEXTNVM9, reg);
4209 	}
4210 
4211 }
4212 
4213 /**
4214  * e1000e_trigger_lsc - trigger an LSC interrupt
4215  * @adapter: board private structure
4216  *
4217  * Fire a link status change interrupt to start the watchdog.
4218  **/
e1000e_trigger_lsc(struct e1000_adapter * adapter)4219 static void e1000e_trigger_lsc(struct e1000_adapter *adapter)
4220 {
4221 	struct e1000_hw *hw = &adapter->hw;
4222 
4223 	if (adapter->msix_entries)
4224 		ew32(ICS, E1000_ICS_LSC | E1000_ICS_OTHER);
4225 	else
4226 		ew32(ICS, E1000_ICS_LSC);
4227 }
4228 
e1000e_up(struct e1000_adapter * adapter)4229 void e1000e_up(struct e1000_adapter *adapter)
4230 {
4231 	/* hardware has been reset, we need to reload some things */
4232 	e1000_configure(adapter);
4233 
4234 	clear_bit(__E1000_DOWN, &adapter->state);
4235 
4236 	if (adapter->msix_entries)
4237 		e1000_configure_msix(adapter);
4238 	e1000_irq_enable(adapter);
4239 
4240 	/* Tx queue started by watchdog timer when link is up */
4241 
4242 	e1000e_trigger_lsc(adapter);
4243 }
4244 
e1000e_flush_descriptors(struct e1000_adapter * adapter)4245 static void e1000e_flush_descriptors(struct e1000_adapter *adapter)
4246 {
4247 	struct e1000_hw *hw = &adapter->hw;
4248 
4249 	if (!(adapter->flags2 & FLAG2_DMA_BURST))
4250 		return;
4251 
4252 	/* flush pending descriptor writebacks to memory */
4253 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4254 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4255 
4256 	/* execute the writes immediately */
4257 	e1e_flush();
4258 
4259 	/* due to rare timing issues, write to TIDV/RDTR again to ensure the
4260 	 * write is successful
4261 	 */
4262 	ew32(TIDV, adapter->tx_int_delay | E1000_TIDV_FPD);
4263 	ew32(RDTR, adapter->rx_int_delay | E1000_RDTR_FPD);
4264 
4265 	/* execute the writes immediately */
4266 	e1e_flush();
4267 }
4268 
4269 static void e1000e_update_stats(struct e1000_adapter *adapter);
4270 
4271 /**
4272  * e1000e_down - quiesce the device and optionally reset the hardware
4273  * @adapter: board private structure
4274  * @reset: boolean flag to reset the hardware or not
4275  */
e1000e_down(struct e1000_adapter * adapter,bool reset)4276 void e1000e_down(struct e1000_adapter *adapter, bool reset)
4277 {
4278 	struct net_device *netdev = adapter->netdev;
4279 	struct e1000_hw *hw = &adapter->hw;
4280 	u32 tctl, rctl;
4281 
4282 	/* signal that we're down so the interrupt handler does not
4283 	 * reschedule our watchdog timer
4284 	 */
4285 	set_bit(__E1000_DOWN, &adapter->state);
4286 
4287 	netif_carrier_off(netdev);
4288 
4289 	/* disable receives in the hardware */
4290 	rctl = er32(RCTL);
4291 	if (!(adapter->flags2 & FLAG2_NO_DISABLE_RX))
4292 		ew32(RCTL, rctl & ~E1000_RCTL_EN);
4293 	/* flush and sleep below */
4294 
4295 	netif_stop_queue(netdev);
4296 
4297 	/* disable transmits in the hardware */
4298 	tctl = er32(TCTL);
4299 	tctl &= ~E1000_TCTL_EN;
4300 	ew32(TCTL, tctl);
4301 
4302 	/* flush both disables and wait for them to finish */
4303 	e1e_flush();
4304 	usleep_range(10000, 11000);
4305 
4306 	e1000_irq_disable(adapter);
4307 
4308 	napi_synchronize(&adapter->napi);
4309 
4310 	timer_delete_sync(&adapter->watchdog_timer);
4311 	timer_delete_sync(&adapter->phy_info_timer);
4312 
4313 	spin_lock(&adapter->stats64_lock);
4314 	e1000e_update_stats(adapter);
4315 	spin_unlock(&adapter->stats64_lock);
4316 
4317 	e1000e_flush_descriptors(adapter);
4318 
4319 	adapter->link_speed = 0;
4320 	adapter->link_duplex = 0;
4321 
4322 	/* Disable Si errata workaround on PCHx for jumbo frame flow */
4323 	if ((hw->mac.type >= e1000_pch2lan) &&
4324 	    (adapter->netdev->mtu > ETH_DATA_LEN) &&
4325 	    e1000_lv_jumbo_workaround_ich8lan(hw, false))
4326 		e_dbg("failed to disable jumbo frame workaround mode\n");
4327 
4328 	if (!pci_channel_offline(adapter->pdev)) {
4329 		if (reset)
4330 			e1000e_reset(adapter);
4331 		else if (hw->mac.type >= e1000_pch_spt)
4332 			e1000_flush_desc_rings(adapter);
4333 	}
4334 	e1000_clean_tx_ring(adapter->tx_ring);
4335 	e1000_clean_rx_ring(adapter->rx_ring);
4336 }
4337 
e1000e_reinit_locked(struct e1000_adapter * adapter)4338 void e1000e_reinit_locked(struct e1000_adapter *adapter)
4339 {
4340 	might_sleep();
4341 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
4342 		usleep_range(1000, 1100);
4343 	e1000e_down(adapter, true);
4344 	e1000e_up(adapter);
4345 	clear_bit(__E1000_RESETTING, &adapter->state);
4346 }
4347 
4348 /**
4349  * e1000e_sanitize_systim - sanitize raw cycle counter reads
4350  * @hw: pointer to the HW structure
4351  * @systim: PHC time value read, sanitized and returned
4352  * @sts: structure to hold system time before and after reading SYSTIML,
4353  * may be NULL
4354  *
4355  * Errata for 82574/82583 possible bad bits read from SYSTIMH/L:
4356  * check to see that the time is incrementing at a reasonable
4357  * rate and is a multiple of incvalue.
4358  **/
e1000e_sanitize_systim(struct e1000_hw * hw,u64 systim,struct ptp_system_timestamp * sts)4359 static u64 e1000e_sanitize_systim(struct e1000_hw *hw, u64 systim,
4360 				  struct ptp_system_timestamp *sts)
4361 {
4362 	u64 time_delta, rem, temp;
4363 	u64 systim_next;
4364 	u32 incvalue;
4365 	int i;
4366 
4367 	incvalue = er32(TIMINCA) & E1000_TIMINCA_INCVALUE_MASK;
4368 	for (i = 0; i < E1000_MAX_82574_SYSTIM_REREADS; i++) {
4369 		/* latch SYSTIMH on read of SYSTIML */
4370 		ptp_read_system_prets(sts);
4371 		systim_next = (u64)er32(SYSTIML);
4372 		ptp_read_system_postts(sts);
4373 		systim_next |= (u64)er32(SYSTIMH) << 32;
4374 
4375 		time_delta = systim_next - systim;
4376 		temp = time_delta;
4377 		/* VMWare users have seen incvalue of zero, don't div / 0 */
4378 		rem = incvalue ? do_div(temp, incvalue) : (time_delta != 0);
4379 
4380 		systim = systim_next;
4381 
4382 		if ((time_delta < E1000_82574_SYSTIM_EPSILON) && (rem == 0))
4383 			break;
4384 	}
4385 
4386 	return systim;
4387 }
4388 
4389 /**
4390  * e1000e_read_systim - read SYSTIM register
4391  * @adapter: board private structure
4392  * @sts: structure which will contain system time before and after reading
4393  * SYSTIML, may be NULL
4394  **/
e1000e_read_systim(struct e1000_adapter * adapter,struct ptp_system_timestamp * sts)4395 u64 e1000e_read_systim(struct e1000_adapter *adapter,
4396 		       struct ptp_system_timestamp *sts)
4397 {
4398 	struct e1000_hw *hw = &adapter->hw;
4399 	u32 systimel, systimel_2, systimeh;
4400 	u64 systim;
4401 	/* SYSTIMH latching upon SYSTIML read does not work well.
4402 	 * This means that if SYSTIML overflows after we read it but before
4403 	 * we read SYSTIMH, the value of SYSTIMH has been incremented and we
4404 	 * will experience a huge non linear increment in the systime value
4405 	 * to fix that we test for overflow and if true, we re-read systime.
4406 	 */
4407 	ptp_read_system_prets(sts);
4408 	systimel = er32(SYSTIML);
4409 	ptp_read_system_postts(sts);
4410 	systimeh = er32(SYSTIMH);
4411 	/* Is systimel is so large that overflow is possible? */
4412 	if (systimel >= (u32)0xffffffff - E1000_TIMINCA_INCVALUE_MASK) {
4413 		ptp_read_system_prets(sts);
4414 		systimel_2 = er32(SYSTIML);
4415 		ptp_read_system_postts(sts);
4416 		if (systimel > systimel_2) {
4417 			/* There was an overflow, read again SYSTIMH, and use
4418 			 * systimel_2
4419 			 */
4420 			systimeh = er32(SYSTIMH);
4421 			systimel = systimel_2;
4422 		}
4423 	}
4424 	systim = (u64)systimel;
4425 	systim |= (u64)systimeh << 32;
4426 
4427 	if (adapter->flags2 & FLAG2_CHECK_SYSTIM_OVERFLOW)
4428 		systim = e1000e_sanitize_systim(hw, systim, sts);
4429 
4430 	return systim;
4431 }
4432 
4433 /**
4434  * e1000e_cyclecounter_read - read raw cycle counter (used by time counter)
4435  * @cc: cyclecounter structure
4436  **/
e1000e_cyclecounter_read(struct cyclecounter * cc)4437 static u64 e1000e_cyclecounter_read(struct cyclecounter *cc)
4438 {
4439 	struct e1000_adapter *adapter = container_of(cc, struct e1000_adapter,
4440 						     cc);
4441 
4442 	return e1000e_read_systim(adapter, NULL);
4443 }
4444 
4445 /**
4446  * e1000_sw_init - Initialize general software structures (struct e1000_adapter)
4447  * @adapter: board private structure to initialize
4448  *
4449  * e1000_sw_init initializes the Adapter private data structure.
4450  * Fields are initialized based on PCI device information and
4451  * OS network device settings (MTU size).
4452  **/
e1000_sw_init(struct e1000_adapter * adapter)4453 static int e1000_sw_init(struct e1000_adapter *adapter)
4454 {
4455 	struct net_device *netdev = adapter->netdev;
4456 
4457 	adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
4458 	adapter->rx_ps_bsize0 = 128;
4459 	adapter->max_frame_size = netdev->mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
4460 	adapter->min_frame_size = ETH_ZLEN + ETH_FCS_LEN;
4461 	adapter->tx_ring_count = E1000_DEFAULT_TXD;
4462 	adapter->rx_ring_count = E1000_DEFAULT_RXD;
4463 
4464 	spin_lock_init(&adapter->stats64_lock);
4465 
4466 	e1000e_set_interrupt_capability(adapter);
4467 
4468 	if (e1000_alloc_queues(adapter))
4469 		return -ENOMEM;
4470 
4471 	/* Setup hardware time stamping cyclecounter */
4472 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
4473 		adapter->cc.read = e1000e_cyclecounter_read;
4474 		adapter->cc.mask = CYCLECOUNTER_MASK(64);
4475 		adapter->cc.mult = 1;
4476 		/* cc.shift set in e1000e_get_base_tininca() */
4477 
4478 		spin_lock_init(&adapter->systim_lock);
4479 		INIT_WORK(&adapter->tx_hwtstamp_work, e1000e_tx_hwtstamp_work);
4480 	}
4481 
4482 	/* Explicitly disable IRQ since the NIC can be in any state. */
4483 	e1000_irq_disable(adapter);
4484 
4485 	set_bit(__E1000_DOWN, &adapter->state);
4486 	return 0;
4487 }
4488 
4489 /**
4490  * e1000_intr_msi_test - Interrupt Handler
4491  * @irq: interrupt number
4492  * @data: pointer to a network interface device structure
4493  **/
e1000_intr_msi_test(int __always_unused irq,void * data)4494 static irqreturn_t e1000_intr_msi_test(int __always_unused irq, void *data)
4495 {
4496 	struct net_device *netdev = data;
4497 	struct e1000_adapter *adapter = netdev_priv(netdev);
4498 	struct e1000_hw *hw = &adapter->hw;
4499 	u32 icr = er32(ICR);
4500 
4501 	e_dbg("icr is %08X\n", icr);
4502 	if (icr & E1000_ICR_RXSEQ) {
4503 		adapter->flags &= ~FLAG_MSI_TEST_FAILED;
4504 		/* Force memory writes to complete before acknowledging the
4505 		 * interrupt is handled.
4506 		 */
4507 		wmb();
4508 	}
4509 
4510 	return IRQ_HANDLED;
4511 }
4512 
4513 /**
4514  * e1000_test_msi_interrupt - Returns 0 for successful test
4515  * @adapter: board private struct
4516  *
4517  * code flow taken from tg3.c
4518  **/
e1000_test_msi_interrupt(struct e1000_adapter * adapter)4519 static int e1000_test_msi_interrupt(struct e1000_adapter *adapter)
4520 {
4521 	struct net_device *netdev = adapter->netdev;
4522 	struct e1000_hw *hw = &adapter->hw;
4523 	int err;
4524 
4525 	/* poll_enable hasn't been called yet, so don't need disable */
4526 	/* clear any pending events */
4527 	er32(ICR);
4528 
4529 	/* free the real vector and request a test handler */
4530 	e1000_free_irq(adapter);
4531 	e1000e_reset_interrupt_capability(adapter);
4532 
4533 	/* Assume that the test fails, if it succeeds then the test
4534 	 * MSI irq handler will unset this flag
4535 	 */
4536 	adapter->flags |= FLAG_MSI_TEST_FAILED;
4537 
4538 	err = pci_enable_msi(adapter->pdev);
4539 	if (err)
4540 		goto msi_test_failed;
4541 
4542 	err = request_irq(adapter->pdev->irq, e1000_intr_msi_test, 0,
4543 			  netdev->name, netdev);
4544 	if (err) {
4545 		pci_disable_msi(adapter->pdev);
4546 		goto msi_test_failed;
4547 	}
4548 
4549 	/* Force memory writes to complete before enabling and firing an
4550 	 * interrupt.
4551 	 */
4552 	wmb();
4553 
4554 	e1000_irq_enable(adapter);
4555 
4556 	/* fire an unusual interrupt on the test handler */
4557 	ew32(ICS, E1000_ICS_RXSEQ);
4558 	e1e_flush();
4559 	msleep(100);
4560 
4561 	e1000_irq_disable(adapter);
4562 
4563 	rmb();			/* read flags after interrupt has been fired */
4564 
4565 	if (adapter->flags & FLAG_MSI_TEST_FAILED) {
4566 		adapter->int_mode = E1000E_INT_MODE_LEGACY;
4567 		e_info("MSI interrupt test failed, using legacy interrupt.\n");
4568 	} else {
4569 		e_dbg("MSI interrupt test succeeded!\n");
4570 	}
4571 
4572 	free_irq(adapter->pdev->irq, netdev);
4573 	pci_disable_msi(adapter->pdev);
4574 
4575 msi_test_failed:
4576 	e1000e_set_interrupt_capability(adapter);
4577 	return e1000_request_irq(adapter);
4578 }
4579 
4580 /**
4581  * e1000_test_msi - Returns 0 if MSI test succeeds or INTx mode is restored
4582  * @adapter: board private struct
4583  *
4584  * code flow taken from tg3.c, called with e1000 interrupts disabled.
4585  **/
e1000_test_msi(struct e1000_adapter * adapter)4586 static int e1000_test_msi(struct e1000_adapter *adapter)
4587 {
4588 	int err;
4589 	u16 pci_cmd;
4590 
4591 	if (!(adapter->flags & FLAG_MSI_ENABLED))
4592 		return 0;
4593 
4594 	/* disable SERR in case the MSI write causes a master abort */
4595 	pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4596 	if (pci_cmd & PCI_COMMAND_SERR)
4597 		pci_write_config_word(adapter->pdev, PCI_COMMAND,
4598 				      pci_cmd & ~PCI_COMMAND_SERR);
4599 
4600 	err = e1000_test_msi_interrupt(adapter);
4601 
4602 	/* re-enable SERR */
4603 	if (pci_cmd & PCI_COMMAND_SERR) {
4604 		pci_read_config_word(adapter->pdev, PCI_COMMAND, &pci_cmd);
4605 		pci_cmd |= PCI_COMMAND_SERR;
4606 		pci_write_config_word(adapter->pdev, PCI_COMMAND, pci_cmd);
4607 	}
4608 
4609 	return err;
4610 }
4611 
4612 /**
4613  * e1000e_open - Called when a network interface is made active
4614  * @netdev: network interface device structure
4615  *
4616  * Returns 0 on success, negative value on failure
4617  *
4618  * The open entry point is called when a network interface is made
4619  * active by the system (IFF_UP).  At this point all resources needed
4620  * for transmit and receive operations are allocated, the interrupt
4621  * handler is registered with the OS, the watchdog timer is started,
4622  * and the stack is notified that the interface is ready.
4623  **/
e1000e_open(struct net_device * netdev)4624 int e1000e_open(struct net_device *netdev)
4625 {
4626 	struct e1000_adapter *adapter = netdev_priv(netdev);
4627 	struct e1000_hw *hw = &adapter->hw;
4628 	struct pci_dev *pdev = adapter->pdev;
4629 	int err;
4630 	int irq;
4631 
4632 	/* disallow open during test */
4633 	if (test_bit(__E1000_TESTING, &adapter->state))
4634 		return -EBUSY;
4635 
4636 	pm_runtime_get_sync(&pdev->dev);
4637 
4638 	netif_carrier_off(netdev);
4639 	netif_stop_queue(netdev);
4640 
4641 	/* allocate transmit descriptors */
4642 	err = e1000e_setup_tx_resources(adapter->tx_ring);
4643 	if (err)
4644 		goto err_setup_tx;
4645 
4646 	/* allocate receive descriptors */
4647 	err = e1000e_setup_rx_resources(adapter->rx_ring);
4648 	if (err)
4649 		goto err_setup_rx;
4650 
4651 	/* If AMT is enabled, let the firmware know that the network
4652 	 * interface is now open and reset the part to a known state.
4653 	 */
4654 	if (adapter->flags & FLAG_HAS_AMT) {
4655 		e1000e_get_hw_control(adapter);
4656 		e1000e_reset(adapter);
4657 	}
4658 
4659 	e1000e_power_up_phy(adapter);
4660 
4661 	adapter->mng_vlan_id = E1000_MNG_VLAN_NONE;
4662 	if ((adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN))
4663 		e1000_update_mng_vlan(adapter);
4664 
4665 	/* DMA latency requirement to workaround jumbo issue */
4666 	cpu_latency_qos_add_request(&adapter->pm_qos_req, PM_QOS_DEFAULT_VALUE);
4667 
4668 	/* before we allocate an interrupt, we must be ready to handle it.
4669 	 * Setting DEBUG_SHIRQ in the kernel makes it fire an interrupt
4670 	 * as soon as we call pci_request_irq, so we have to setup our
4671 	 * clean_rx handler before we do so.
4672 	 */
4673 	e1000_configure(adapter);
4674 
4675 	err = e1000_request_irq(adapter);
4676 	if (err)
4677 		goto err_req_irq;
4678 
4679 	/* Work around PCIe errata with MSI interrupts causing some chipsets to
4680 	 * ignore e1000e MSI messages, which means we need to test our MSI
4681 	 * interrupt now
4682 	 */
4683 	if (adapter->int_mode != E1000E_INT_MODE_LEGACY) {
4684 		err = e1000_test_msi(adapter);
4685 		if (err) {
4686 			e_err("Interrupt allocation failed\n");
4687 			goto err_req_irq;
4688 		}
4689 	}
4690 
4691 	/* From here on the code is the same as e1000e_up() */
4692 	clear_bit(__E1000_DOWN, &adapter->state);
4693 
4694 	if (adapter->int_mode == E1000E_INT_MODE_MSIX)
4695 		irq = adapter->msix_entries[0].vector;
4696 	else
4697 		irq = adapter->pdev->irq;
4698 
4699 	netif_napi_set_irq(&adapter->napi, irq);
4700 	napi_enable(&adapter->napi);
4701 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, &adapter->napi);
4702 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, &adapter->napi);
4703 
4704 	e1000_irq_enable(adapter);
4705 
4706 	adapter->tx_hang_recheck = false;
4707 
4708 	hw->mac.get_link_status = true;
4709 	pm_runtime_put(&pdev->dev);
4710 
4711 	e1000e_trigger_lsc(adapter);
4712 
4713 	return 0;
4714 
4715 err_req_irq:
4716 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4717 	e1000e_release_hw_control(adapter);
4718 	e1000_power_down_phy(adapter);
4719 	e1000e_free_rx_resources(adapter->rx_ring);
4720 err_setup_rx:
4721 	e1000e_free_tx_resources(adapter->tx_ring);
4722 err_setup_tx:
4723 	e1000e_reset(adapter);
4724 	pm_runtime_put_sync(&pdev->dev);
4725 
4726 	return err;
4727 }
4728 
4729 /**
4730  * e1000e_close - Disables a network interface
4731  * @netdev: network interface device structure
4732  *
4733  * Returns 0, this is not allowed to fail
4734  *
4735  * The close entry point is called when an interface is de-activated
4736  * by the OS.  The hardware is still under the drivers control, but
4737  * needs to be disabled.  A global MAC reset is issued to stop the
4738  * hardware, and all transmit and receive resources are freed.
4739  **/
e1000e_close(struct net_device * netdev)4740 int e1000e_close(struct net_device *netdev)
4741 {
4742 	struct e1000_adapter *adapter = netdev_priv(netdev);
4743 	struct pci_dev *pdev = adapter->pdev;
4744 	int count = E1000_CHECK_RESET_COUNT;
4745 
4746 	while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
4747 		usleep_range(10000, 11000);
4748 
4749 	WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
4750 
4751 	pm_runtime_get_sync(&pdev->dev);
4752 
4753 	if (netif_device_present(netdev)) {
4754 		e1000e_down(adapter, true);
4755 		e1000_free_irq(adapter);
4756 
4757 		/* Link status message must follow this format */
4758 		netdev_info(netdev, "NIC Link is Down\n");
4759 	}
4760 
4761 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_RX, NULL);
4762 	netif_queue_set_napi(netdev, 0, NETDEV_QUEUE_TYPE_TX, NULL);
4763 	napi_disable(&adapter->napi);
4764 
4765 	e1000e_free_tx_resources(adapter->tx_ring);
4766 	e1000e_free_rx_resources(adapter->rx_ring);
4767 
4768 	/* kill manageability vlan ID if supported, but not if a vlan with
4769 	 * the same ID is registered on the host OS (let 8021q kill it)
4770 	 */
4771 	if (adapter->hw.mng_cookie.status & E1000_MNG_DHCP_COOKIE_STATUS_VLAN)
4772 		e1000_vlan_rx_kill_vid(netdev, htons(ETH_P_8021Q),
4773 				       adapter->mng_vlan_id);
4774 
4775 	/* If AMT is enabled, let the firmware know that the network
4776 	 * interface is now closed
4777 	 */
4778 	if ((adapter->flags & FLAG_HAS_AMT) &&
4779 	    !test_bit(__E1000_TESTING, &adapter->state))
4780 		e1000e_release_hw_control(adapter);
4781 
4782 	cpu_latency_qos_remove_request(&adapter->pm_qos_req);
4783 
4784 	pm_runtime_put_sync(&pdev->dev);
4785 
4786 	return 0;
4787 }
4788 
4789 /**
4790  * e1000_set_mac - Change the Ethernet Address of the NIC
4791  * @netdev: network interface device structure
4792  * @p: pointer to an address structure
4793  *
4794  * Returns 0 on success, negative on failure
4795  **/
e1000_set_mac(struct net_device * netdev,void * p)4796 static int e1000_set_mac(struct net_device *netdev, void *p)
4797 {
4798 	struct e1000_adapter *adapter = netdev_priv(netdev);
4799 	struct e1000_hw *hw = &adapter->hw;
4800 	struct sockaddr *addr = p;
4801 
4802 	if (!is_valid_ether_addr(addr->sa_data))
4803 		return -EADDRNOTAVAIL;
4804 
4805 	eth_hw_addr_set(netdev, addr->sa_data);
4806 	memcpy(adapter->hw.mac.addr, addr->sa_data, netdev->addr_len);
4807 
4808 	hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr, 0);
4809 
4810 	if (adapter->flags & FLAG_RESET_OVERWRITES_LAA) {
4811 		/* activate the work around */
4812 		e1000e_set_laa_state_82571(&adapter->hw, 1);
4813 
4814 		/* Hold a copy of the LAA in RAR[14] This is done so that
4815 		 * between the time RAR[0] gets clobbered  and the time it
4816 		 * gets fixed (in e1000_watchdog), the actual LAA is in one
4817 		 * of the RARs and no incoming packets directed to this port
4818 		 * are dropped. Eventually the LAA will be in RAR[0] and
4819 		 * RAR[14]
4820 		 */
4821 		hw->mac.ops.rar_set(&adapter->hw, adapter->hw.mac.addr,
4822 				    adapter->hw.mac.rar_entry_count - 1);
4823 	}
4824 
4825 	return 0;
4826 }
4827 
4828 /**
4829  * e1000e_update_phy_task - work thread to update phy
4830  * @work: pointer to our work struct
4831  *
4832  * this worker thread exists because we must acquire a
4833  * semaphore to read the phy, which we could msleep while
4834  * waiting for it, and we can't msleep in a timer.
4835  **/
e1000e_update_phy_task(struct work_struct * work)4836 static void e1000e_update_phy_task(struct work_struct *work)
4837 {
4838 	struct e1000_adapter *adapter = container_of(work,
4839 						     struct e1000_adapter,
4840 						     update_phy_task);
4841 	struct e1000_hw *hw = &adapter->hw;
4842 
4843 	if (test_bit(__E1000_DOWN, &adapter->state))
4844 		return;
4845 
4846 	e1000_get_phy_info(hw);
4847 
4848 	/* Enable EEE on 82579 after link up */
4849 	if (hw->phy.type >= e1000_phy_82579)
4850 		e1000_set_eee_pchlan(hw);
4851 }
4852 
4853 /**
4854  * e1000_update_phy_info - timre call-back to update PHY info
4855  * @t: pointer to timer_list containing private info adapter
4856  *
4857  * Need to wait a few seconds after link up to get diagnostic information from
4858  * the phy
4859  **/
e1000_update_phy_info(struct timer_list * t)4860 static void e1000_update_phy_info(struct timer_list *t)
4861 {
4862 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
4863 							   phy_info_timer);
4864 
4865 	if (test_bit(__E1000_DOWN, &adapter->state))
4866 		return;
4867 
4868 	schedule_work(&adapter->update_phy_task);
4869 }
4870 
4871 /**
4872  * e1000e_update_phy_stats - Update the PHY statistics counters
4873  * @adapter: board private structure
4874  *
4875  * Read/clear the upper 16-bit PHY registers and read/accumulate lower
4876  **/
e1000e_update_phy_stats(struct e1000_adapter * adapter)4877 static void e1000e_update_phy_stats(struct e1000_adapter *adapter)
4878 {
4879 	struct e1000_hw *hw = &adapter->hw;
4880 	s32 ret_val;
4881 	u16 phy_data;
4882 
4883 	ret_val = hw->phy.ops.acquire(hw);
4884 	if (ret_val)
4885 		return;
4886 
4887 	/* A page set is expensive so check if already on desired page.
4888 	 * If not, set to the page with the PHY status registers.
4889 	 */
4890 	hw->phy.addr = 1;
4891 	ret_val = e1000e_read_phy_reg_mdic(hw, IGP01E1000_PHY_PAGE_SELECT,
4892 					   &phy_data);
4893 	if (ret_val)
4894 		goto release;
4895 	if (phy_data != (HV_STATS_PAGE << IGP_PAGE_SHIFT)) {
4896 		ret_val = hw->phy.ops.set_page(hw,
4897 					       HV_STATS_PAGE << IGP_PAGE_SHIFT);
4898 		if (ret_val)
4899 			goto release;
4900 	}
4901 
4902 	/* Single Collision Count */
4903 	hw->phy.ops.read_reg_page(hw, HV_SCC_UPPER, &phy_data);
4904 	ret_val = hw->phy.ops.read_reg_page(hw, HV_SCC_LOWER, &phy_data);
4905 	if (!ret_val)
4906 		adapter->stats.scc += phy_data;
4907 
4908 	/* Excessive Collision Count */
4909 	hw->phy.ops.read_reg_page(hw, HV_ECOL_UPPER, &phy_data);
4910 	ret_val = hw->phy.ops.read_reg_page(hw, HV_ECOL_LOWER, &phy_data);
4911 	if (!ret_val)
4912 		adapter->stats.ecol += phy_data;
4913 
4914 	/* Multiple Collision Count */
4915 	hw->phy.ops.read_reg_page(hw, HV_MCC_UPPER, &phy_data);
4916 	ret_val = hw->phy.ops.read_reg_page(hw, HV_MCC_LOWER, &phy_data);
4917 	if (!ret_val)
4918 		adapter->stats.mcc += phy_data;
4919 
4920 	/* Late Collision Count */
4921 	hw->phy.ops.read_reg_page(hw, HV_LATECOL_UPPER, &phy_data);
4922 	ret_val = hw->phy.ops.read_reg_page(hw, HV_LATECOL_LOWER, &phy_data);
4923 	if (!ret_val)
4924 		adapter->stats.latecol += phy_data;
4925 
4926 	/* Collision Count - also used for adaptive IFS */
4927 	hw->phy.ops.read_reg_page(hw, HV_COLC_UPPER, &phy_data);
4928 	ret_val = hw->phy.ops.read_reg_page(hw, HV_COLC_LOWER, &phy_data);
4929 	if (!ret_val)
4930 		hw->mac.collision_delta = phy_data;
4931 
4932 	/* Defer Count */
4933 	hw->phy.ops.read_reg_page(hw, HV_DC_UPPER, &phy_data);
4934 	ret_val = hw->phy.ops.read_reg_page(hw, HV_DC_LOWER, &phy_data);
4935 	if (!ret_val)
4936 		adapter->stats.dc += phy_data;
4937 
4938 	/* Transmit with no CRS */
4939 	hw->phy.ops.read_reg_page(hw, HV_TNCRS_UPPER, &phy_data);
4940 	ret_val = hw->phy.ops.read_reg_page(hw, HV_TNCRS_LOWER, &phy_data);
4941 	if (!ret_val)
4942 		adapter->stats.tncrs += phy_data;
4943 
4944 release:
4945 	hw->phy.ops.release(hw);
4946 }
4947 
4948 /**
4949  * e1000e_update_stats - Update the board statistics counters
4950  * @adapter: board private structure
4951  **/
e1000e_update_stats(struct e1000_adapter * adapter)4952 static void e1000e_update_stats(struct e1000_adapter *adapter)
4953 {
4954 	struct net_device *netdev = adapter->netdev;
4955 	struct e1000_hw *hw = &adapter->hw;
4956 	struct pci_dev *pdev = adapter->pdev;
4957 
4958 	/* Prevent stats update while adapter is being reset, or if the pci
4959 	 * connection is down.
4960 	 */
4961 	if (adapter->link_speed == 0)
4962 		return;
4963 	if (pci_channel_offline(pdev))
4964 		return;
4965 
4966 	adapter->stats.crcerrs += er32(CRCERRS);
4967 	adapter->stats.gprc += er32(GPRC);
4968 	adapter->stats.gorc += er32(GORCL);
4969 	er32(GORCH);		/* Clear gorc */
4970 	adapter->stats.bprc += er32(BPRC);
4971 	adapter->stats.mprc += er32(MPRC);
4972 	adapter->stats.roc += er32(ROC);
4973 
4974 	adapter->stats.mpc += er32(MPC);
4975 
4976 	/* Half-duplex statistics */
4977 	if (adapter->link_duplex == HALF_DUPLEX) {
4978 		if (adapter->flags2 & FLAG2_HAS_PHY_STATS) {
4979 			e1000e_update_phy_stats(adapter);
4980 		} else {
4981 			adapter->stats.scc += er32(SCC);
4982 			adapter->stats.ecol += er32(ECOL);
4983 			adapter->stats.mcc += er32(MCC);
4984 			adapter->stats.latecol += er32(LATECOL);
4985 			adapter->stats.dc += er32(DC);
4986 
4987 			hw->mac.collision_delta = er32(COLC);
4988 
4989 			if ((hw->mac.type != e1000_82574) &&
4990 			    (hw->mac.type != e1000_82583))
4991 				adapter->stats.tncrs += er32(TNCRS);
4992 		}
4993 		adapter->stats.colc += hw->mac.collision_delta;
4994 	}
4995 
4996 	adapter->stats.xonrxc += er32(XONRXC);
4997 	adapter->stats.xontxc += er32(XONTXC);
4998 	adapter->stats.xoffrxc += er32(XOFFRXC);
4999 	adapter->stats.xofftxc += er32(XOFFTXC);
5000 	adapter->stats.gptc += er32(GPTC);
5001 	adapter->stats.gotc += er32(GOTCL);
5002 	er32(GOTCH);		/* Clear gotc */
5003 	adapter->stats.rnbc += er32(RNBC);
5004 	adapter->stats.ruc += er32(RUC);
5005 
5006 	adapter->stats.mptc += er32(MPTC);
5007 	adapter->stats.bptc += er32(BPTC);
5008 
5009 	/* used for adaptive IFS */
5010 
5011 	hw->mac.tx_packet_delta = er32(TPT);
5012 	adapter->stats.tpt += hw->mac.tx_packet_delta;
5013 
5014 	adapter->stats.algnerrc += er32(ALGNERRC);
5015 	adapter->stats.rxerrc += er32(RXERRC);
5016 	adapter->stats.cexterr += er32(CEXTERR);
5017 	adapter->stats.tsctc += er32(TSCTC);
5018 	adapter->stats.tsctfc += er32(TSCTFC);
5019 
5020 	/* Fill out the OS statistics structure */
5021 	netdev->stats.multicast = adapter->stats.mprc;
5022 	netdev->stats.collisions = adapter->stats.colc;
5023 
5024 	/* Rx Errors */
5025 
5026 	/* RLEC on some newer hardware can be incorrect so build
5027 	 * our own version based on RUC and ROC
5028 	 */
5029 	netdev->stats.rx_errors = adapter->stats.rxerrc +
5030 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
5031 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
5032 	netdev->stats.rx_length_errors = adapter->stats.ruc +
5033 	    adapter->stats.roc;
5034 	netdev->stats.rx_crc_errors = adapter->stats.crcerrs;
5035 	netdev->stats.rx_frame_errors = adapter->stats.algnerrc;
5036 	netdev->stats.rx_missed_errors = adapter->stats.mpc;
5037 
5038 	/* Tx Errors */
5039 	netdev->stats.tx_errors = adapter->stats.ecol + adapter->stats.latecol;
5040 	netdev->stats.tx_aborted_errors = adapter->stats.ecol;
5041 	netdev->stats.tx_window_errors = adapter->stats.latecol;
5042 	netdev->stats.tx_carrier_errors = adapter->stats.tncrs;
5043 
5044 	/* Tx Dropped needs to be maintained elsewhere */
5045 
5046 	/* Management Stats */
5047 	adapter->stats.mgptc += er32(MGTPTC);
5048 	adapter->stats.mgprc += er32(MGTPRC);
5049 	adapter->stats.mgpdc += er32(MGTPDC);
5050 
5051 	/* Correctable ECC Errors */
5052 	if (hw->mac.type >= e1000_pch_lpt) {
5053 		u32 pbeccsts = er32(PBECCSTS);
5054 
5055 		adapter->corr_errors +=
5056 		    pbeccsts & E1000_PBECCSTS_CORR_ERR_CNT_MASK;
5057 		adapter->uncorr_errors +=
5058 		    FIELD_GET(E1000_PBECCSTS_UNCORR_ERR_CNT_MASK, pbeccsts);
5059 	}
5060 }
5061 
5062 /**
5063  * e1000_phy_read_status - Update the PHY register status snapshot
5064  * @adapter: board private structure
5065  **/
e1000_phy_read_status(struct e1000_adapter * adapter)5066 static void e1000_phy_read_status(struct e1000_adapter *adapter)
5067 {
5068 	struct e1000_hw *hw = &adapter->hw;
5069 	struct e1000_phy_regs *phy = &adapter->phy_regs;
5070 
5071 	if (!pm_runtime_suspended((&adapter->pdev->dev)->parent) &&
5072 	    (er32(STATUS) & E1000_STATUS_LU) &&
5073 	    (adapter->hw.phy.media_type == e1000_media_type_copper)) {
5074 		int ret_val;
5075 
5076 		ret_val = e1e_rphy(hw, MII_BMCR, &phy->bmcr);
5077 		ret_val |= e1e_rphy(hw, MII_BMSR, &phy->bmsr);
5078 		ret_val |= e1e_rphy(hw, MII_ADVERTISE, &phy->advertise);
5079 		ret_val |= e1e_rphy(hw, MII_LPA, &phy->lpa);
5080 		ret_val |= e1e_rphy(hw, MII_EXPANSION, &phy->expansion);
5081 		ret_val |= e1e_rphy(hw, MII_CTRL1000, &phy->ctrl1000);
5082 		ret_val |= e1e_rphy(hw, MII_STAT1000, &phy->stat1000);
5083 		ret_val |= e1e_rphy(hw, MII_ESTATUS, &phy->estatus);
5084 		if (ret_val)
5085 			e_warn("Error reading PHY register\n");
5086 	} else {
5087 		/* Do not read PHY registers if link is not up
5088 		 * Set values to typical power-on defaults
5089 		 */
5090 		phy->bmcr = (BMCR_SPEED1000 | BMCR_ANENABLE | BMCR_FULLDPLX);
5091 		phy->bmsr = (BMSR_100FULL | BMSR_100HALF | BMSR_10FULL |
5092 			     BMSR_10HALF | BMSR_ESTATEN | BMSR_ANEGCAPABLE |
5093 			     BMSR_ERCAP);
5094 		phy->advertise = (ADVERTISE_PAUSE_ASYM | ADVERTISE_PAUSE_CAP |
5095 				  ADVERTISE_ALL | ADVERTISE_CSMA);
5096 		phy->lpa = 0;
5097 		phy->expansion = EXPANSION_ENABLENPAGE;
5098 		phy->ctrl1000 = ADVERTISE_1000FULL;
5099 		phy->stat1000 = 0;
5100 		phy->estatus = (ESTATUS_1000_TFULL | ESTATUS_1000_THALF);
5101 	}
5102 }
5103 
e1000_print_link_info(struct e1000_adapter * adapter)5104 static void e1000_print_link_info(struct e1000_adapter *adapter)
5105 {
5106 	struct e1000_hw *hw = &adapter->hw;
5107 	u32 ctrl = er32(CTRL);
5108 
5109 	/* Link status message must follow this format for user tools */
5110 	netdev_info(adapter->netdev,
5111 		    "NIC Link is Up %d Mbps %s Duplex, Flow Control: %s\n",
5112 		    adapter->link_speed,
5113 		    adapter->link_duplex == FULL_DUPLEX ? "Full" : "Half",
5114 		    (ctrl & E1000_CTRL_TFCE) && (ctrl & E1000_CTRL_RFCE) ? "Rx/Tx" :
5115 		    (ctrl & E1000_CTRL_RFCE) ? "Rx" :
5116 		    (ctrl & E1000_CTRL_TFCE) ? "Tx" : "None");
5117 }
5118 
e1000e_has_link(struct e1000_adapter * adapter)5119 static bool e1000e_has_link(struct e1000_adapter *adapter)
5120 {
5121 	struct e1000_hw *hw = &adapter->hw;
5122 	bool link_active = false;
5123 	s32 ret_val = 0;
5124 
5125 	/* get_link_status is set on LSC (link status) interrupt or
5126 	 * Rx sequence error interrupt.  get_link_status will stay
5127 	 * true until the check_for_link establishes link
5128 	 * for copper adapters ONLY
5129 	 */
5130 	switch (hw->phy.media_type) {
5131 	case e1000_media_type_copper:
5132 		if (hw->mac.get_link_status) {
5133 			ret_val = hw->mac.ops.check_for_link(hw);
5134 			link_active = !hw->mac.get_link_status;
5135 		} else {
5136 			link_active = true;
5137 		}
5138 		break;
5139 	case e1000_media_type_fiber:
5140 		ret_val = hw->mac.ops.check_for_link(hw);
5141 		link_active = !!(er32(STATUS) & E1000_STATUS_LU);
5142 		break;
5143 	case e1000_media_type_internal_serdes:
5144 		ret_val = hw->mac.ops.check_for_link(hw);
5145 		link_active = hw->mac.serdes_has_link;
5146 		break;
5147 	default:
5148 	case e1000_media_type_unknown:
5149 		break;
5150 	}
5151 
5152 	if ((ret_val == -E1000_ERR_PHY) && (hw->phy.type == e1000_phy_igp_3) &&
5153 	    (er32(CTRL) & E1000_PHY_CTRL_GBE_DISABLE)) {
5154 		/* See e1000_kmrn_lock_loss_workaround_ich8lan() */
5155 		e_info("Gigabit has been disabled, downgrading speed\n");
5156 	}
5157 
5158 	return link_active;
5159 }
5160 
e1000e_enable_receives(struct e1000_adapter * adapter)5161 static void e1000e_enable_receives(struct e1000_adapter *adapter)
5162 {
5163 	/* make sure the receive unit is started */
5164 	if ((adapter->flags & FLAG_RX_NEEDS_RESTART) &&
5165 	    (adapter->flags & FLAG_RESTART_NOW)) {
5166 		struct e1000_hw *hw = &adapter->hw;
5167 		u32 rctl = er32(RCTL);
5168 
5169 		ew32(RCTL, rctl | E1000_RCTL_EN);
5170 		adapter->flags &= ~FLAG_RESTART_NOW;
5171 	}
5172 }
5173 
e1000e_check_82574_phy_workaround(struct e1000_adapter * adapter)5174 static void e1000e_check_82574_phy_workaround(struct e1000_adapter *adapter)
5175 {
5176 	struct e1000_hw *hw = &adapter->hw;
5177 
5178 	/* With 82574 controllers, PHY needs to be checked periodically
5179 	 * for hung state and reset, if two calls return true
5180 	 */
5181 	if (e1000_check_phy_82574(hw))
5182 		adapter->phy_hang_count++;
5183 	else
5184 		adapter->phy_hang_count = 0;
5185 
5186 	if (adapter->phy_hang_count > 1) {
5187 		adapter->phy_hang_count = 0;
5188 		e_dbg("PHY appears hung - resetting\n");
5189 		schedule_work(&adapter->reset_task);
5190 	}
5191 }
5192 
5193 /**
5194  * e1000_watchdog - Timer Call-back
5195  * @t: pointer to timer_list containing private info adapter
5196  **/
e1000_watchdog(struct timer_list * t)5197 static void e1000_watchdog(struct timer_list *t)
5198 {
5199 	struct e1000_adapter *adapter = timer_container_of(adapter, t,
5200 							   watchdog_timer);
5201 
5202 	/* Do the rest outside of interrupt context */
5203 	schedule_work(&adapter->watchdog_task);
5204 
5205 	/* TODO: make this use queue_delayed_work() */
5206 }
5207 
e1000_watchdog_task(struct work_struct * work)5208 static void e1000_watchdog_task(struct work_struct *work)
5209 {
5210 	struct e1000_adapter *adapter = container_of(work,
5211 						     struct e1000_adapter,
5212 						     watchdog_task);
5213 	struct net_device *netdev = adapter->netdev;
5214 	struct e1000_mac_info *mac = &adapter->hw.mac;
5215 	struct e1000_phy_info *phy = &adapter->hw.phy;
5216 	struct e1000_ring *tx_ring = adapter->tx_ring;
5217 	u32 dmoff_exit_timeout = 100, tries = 0;
5218 	struct e1000_hw *hw = &adapter->hw;
5219 	u32 link, tctl, pcim_state;
5220 
5221 	if (test_bit(__E1000_DOWN, &adapter->state))
5222 		return;
5223 
5224 	link = e1000e_has_link(adapter);
5225 	if ((netif_carrier_ok(netdev)) && link) {
5226 		/* Cancel scheduled suspend requests. */
5227 		pm_runtime_resume(netdev->dev.parent);
5228 
5229 		e1000e_enable_receives(adapter);
5230 		goto link_up;
5231 	}
5232 
5233 	if ((e1000e_enable_tx_pkt_filtering(hw)) &&
5234 	    (adapter->mng_vlan_id != adapter->hw.mng_cookie.vlan_id))
5235 		e1000_update_mng_vlan(adapter);
5236 
5237 	if (link) {
5238 		if (!netif_carrier_ok(netdev)) {
5239 			bool txb2b = true;
5240 
5241 			/* Cancel scheduled suspend requests. */
5242 			pm_runtime_resume(netdev->dev.parent);
5243 
5244 			/* Checking if MAC is in DMoff state*/
5245 			if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID) {
5246 				pcim_state = er32(STATUS);
5247 				while (pcim_state & E1000_STATUS_PCIM_STATE) {
5248 					if (tries++ == dmoff_exit_timeout) {
5249 						e_dbg("Error in exiting dmoff\n");
5250 						break;
5251 					}
5252 					usleep_range(10000, 20000);
5253 					pcim_state = er32(STATUS);
5254 
5255 					/* Checking if MAC exited DMoff state */
5256 					if (!(pcim_state & E1000_STATUS_PCIM_STATE))
5257 						e1000_phy_hw_reset(&adapter->hw);
5258 				}
5259 			}
5260 
5261 			/* update snapshot of PHY registers on LSC */
5262 			e1000_phy_read_status(adapter);
5263 			mac->ops.get_link_up_info(&adapter->hw,
5264 						  &adapter->link_speed,
5265 						  &adapter->link_duplex);
5266 			e1000_print_link_info(adapter);
5267 
5268 			/* check if SmartSpeed worked */
5269 			e1000e_check_downshift(hw);
5270 			if (phy->speed_downgraded)
5271 				netdev_warn(netdev,
5272 					    "Link Speed was downgraded by SmartSpeed\n");
5273 
5274 			/* On supported PHYs, check for duplex mismatch only
5275 			 * if link has autonegotiated at 10/100 half
5276 			 */
5277 			if ((hw->phy.type == e1000_phy_igp_3 ||
5278 			     hw->phy.type == e1000_phy_bm) &&
5279 			    hw->mac.autoneg &&
5280 			    (adapter->link_speed == SPEED_10 ||
5281 			     adapter->link_speed == SPEED_100) &&
5282 			    (adapter->link_duplex == HALF_DUPLEX)) {
5283 				u16 autoneg_exp;
5284 
5285 				e1e_rphy(hw, MII_EXPANSION, &autoneg_exp);
5286 
5287 				if (!(autoneg_exp & EXPANSION_NWAY))
5288 					e_info("Autonegotiated half duplex but link partner cannot autoneg.  Try forcing full duplex if link gets many collisions.\n");
5289 			}
5290 
5291 			/* adjust timeout factor according to speed/duplex */
5292 			adapter->tx_timeout_factor = 1;
5293 			switch (adapter->link_speed) {
5294 			case SPEED_10:
5295 				txb2b = false;
5296 				adapter->tx_timeout_factor = 16;
5297 				break;
5298 			case SPEED_100:
5299 				txb2b = false;
5300 				adapter->tx_timeout_factor = 10;
5301 				break;
5302 			}
5303 
5304 			/* workaround: re-program speed mode bit after
5305 			 * link-up event
5306 			 */
5307 			if ((adapter->flags & FLAG_TARC_SPEED_MODE_BIT) &&
5308 			    !txb2b) {
5309 				u32 tarc0;
5310 
5311 				tarc0 = er32(TARC(0));
5312 				tarc0 &= ~SPEED_MODE_BIT;
5313 				ew32(TARC(0), tarc0);
5314 			}
5315 
5316 			/* enable transmits in the hardware, need to do this
5317 			 * after setting TARC(0)
5318 			 */
5319 			tctl = er32(TCTL);
5320 			tctl |= E1000_TCTL_EN;
5321 			ew32(TCTL, tctl);
5322 
5323 			/* Perform any post-link-up configuration before
5324 			 * reporting link up.
5325 			 */
5326 			if (phy->ops.cfg_on_link_up)
5327 				phy->ops.cfg_on_link_up(hw);
5328 
5329 			netif_wake_queue(netdev);
5330 			netif_carrier_on(netdev);
5331 
5332 			if (!test_bit(__E1000_DOWN, &adapter->state))
5333 				mod_timer(&adapter->phy_info_timer,
5334 					  round_jiffies(jiffies + 2 * HZ));
5335 		}
5336 	} else {
5337 		if (netif_carrier_ok(netdev)) {
5338 			adapter->link_speed = 0;
5339 			adapter->link_duplex = 0;
5340 			/* Link status message must follow this format */
5341 			netdev_info(netdev, "NIC Link is Down\n");
5342 			netif_carrier_off(netdev);
5343 			netif_stop_queue(netdev);
5344 			if (!test_bit(__E1000_DOWN, &adapter->state))
5345 				mod_timer(&adapter->phy_info_timer,
5346 					  round_jiffies(jiffies + 2 * HZ));
5347 
5348 			/* 8000ES2LAN requires a Rx packet buffer work-around
5349 			 * on link down event; reset the controller to flush
5350 			 * the Rx packet buffer.
5351 			 */
5352 			if (adapter->flags & FLAG_RX_NEEDS_RESTART)
5353 				adapter->flags |= FLAG_RESTART_NOW;
5354 			else
5355 				pm_schedule_suspend(netdev->dev.parent,
5356 						    LINK_TIMEOUT);
5357 		}
5358 	}
5359 
5360 link_up:
5361 	spin_lock(&adapter->stats64_lock);
5362 	e1000e_update_stats(adapter);
5363 
5364 	mac->tx_packet_delta = adapter->stats.tpt - adapter->tpt_old;
5365 	adapter->tpt_old = adapter->stats.tpt;
5366 	mac->collision_delta = adapter->stats.colc - adapter->colc_old;
5367 	adapter->colc_old = adapter->stats.colc;
5368 
5369 	adapter->gorc = adapter->stats.gorc - adapter->gorc_old;
5370 	adapter->gorc_old = adapter->stats.gorc;
5371 	adapter->gotc = adapter->stats.gotc - adapter->gotc_old;
5372 	adapter->gotc_old = adapter->stats.gotc;
5373 	spin_unlock(&adapter->stats64_lock);
5374 
5375 	/* If the link is lost the controller stops DMA, but
5376 	 * if there is queued Tx work it cannot be done.  So
5377 	 * reset the controller to flush the Tx packet buffers.
5378 	 */
5379 	if (!netif_carrier_ok(netdev) &&
5380 	    (e1000_desc_unused(tx_ring) + 1 < tx_ring->count))
5381 		adapter->flags |= FLAG_RESTART_NOW;
5382 
5383 	/* If reset is necessary, do it outside of interrupt context. */
5384 	if (adapter->flags & FLAG_RESTART_NOW) {
5385 		schedule_work(&adapter->reset_task);
5386 		/* return immediately since reset is imminent */
5387 		return;
5388 	}
5389 
5390 	e1000e_update_adaptive(&adapter->hw);
5391 
5392 	/* Simple mode for Interrupt Throttle Rate (ITR) */
5393 	if (adapter->itr_setting == 4) {
5394 		/* Symmetric Tx/Rx gets a reduced ITR=2000;
5395 		 * Total asymmetrical Tx or Rx gets ITR=8000;
5396 		 * everyone else is between 2000-8000.
5397 		 */
5398 		u32 goc = (adapter->gotc + adapter->gorc) / 10000;
5399 		u32 dif = (adapter->gotc > adapter->gorc ?
5400 			   adapter->gotc - adapter->gorc :
5401 			   adapter->gorc - adapter->gotc) / 10000;
5402 		u32 itr = goc > 0 ? (dif * 6000 / goc + 2000) : 8000;
5403 
5404 		e1000e_write_itr(adapter, itr);
5405 	}
5406 
5407 	/* Cause software interrupt to ensure Rx ring is cleaned */
5408 	if (adapter->msix_entries)
5409 		ew32(ICS, adapter->rx_ring->ims_val);
5410 	else
5411 		ew32(ICS, E1000_ICS_RXDMT0);
5412 
5413 	/* flush pending descriptors to memory before detecting Tx hang */
5414 	e1000e_flush_descriptors(adapter);
5415 
5416 	/* Force detection of hung controller every watchdog period */
5417 	adapter->detect_tx_hung = true;
5418 
5419 	/* With 82571 controllers, LAA may be overwritten due to controller
5420 	 * reset from the other port. Set the appropriate LAA in RAR[0]
5421 	 */
5422 	if (e1000e_get_laa_state_82571(hw))
5423 		hw->mac.ops.rar_set(hw, adapter->hw.mac.addr, 0);
5424 
5425 	if (adapter->flags2 & FLAG2_CHECK_PHY_HANG)
5426 		e1000e_check_82574_phy_workaround(adapter);
5427 
5428 	/* Clear valid timestamp stuck in RXSTMPL/H due to a Rx error */
5429 	if (adapter->hwtstamp_config.rx_filter != HWTSTAMP_FILTER_NONE) {
5430 		if ((adapter->flags2 & FLAG2_CHECK_RX_HWTSTAMP) &&
5431 		    (er32(TSYNCRXCTL) & E1000_TSYNCRXCTL_VALID)) {
5432 			er32(RXSTMPH);
5433 			adapter->rx_hwtstamp_cleared++;
5434 		} else {
5435 			adapter->flags2 |= FLAG2_CHECK_RX_HWTSTAMP;
5436 		}
5437 	}
5438 
5439 	/* Reset the timer */
5440 	if (!test_bit(__E1000_DOWN, &adapter->state))
5441 		mod_timer(&adapter->watchdog_timer,
5442 			  round_jiffies(jiffies + 2 * HZ));
5443 }
5444 
5445 #define E1000_TX_FLAGS_CSUM		0x00000001
5446 #define E1000_TX_FLAGS_VLAN		0x00000002
5447 #define E1000_TX_FLAGS_TSO		0x00000004
5448 #define E1000_TX_FLAGS_IPV4		0x00000008
5449 #define E1000_TX_FLAGS_NO_FCS		0x00000010
5450 #define E1000_TX_FLAGS_HWTSTAMP		0x00000020
5451 #define E1000_TX_FLAGS_VLAN_MASK	0xffff0000
5452 #define E1000_TX_FLAGS_VLAN_SHIFT	16
5453 
e1000_tso(struct e1000_ring * tx_ring,struct sk_buff * skb,__be16 protocol)5454 static int e1000_tso(struct e1000_ring *tx_ring, struct sk_buff *skb,
5455 		     __be16 protocol)
5456 {
5457 	struct e1000_context_desc *context_desc;
5458 	struct e1000_buffer *buffer_info;
5459 	unsigned int i;
5460 	u32 cmd_length = 0;
5461 	u16 ipcse = 0, mss;
5462 	u8 ipcss, ipcso, tucss, tucso, hdr_len;
5463 	int err;
5464 
5465 	if (!skb_is_gso(skb))
5466 		return 0;
5467 
5468 	err = skb_cow_head(skb, 0);
5469 	if (err < 0)
5470 		return err;
5471 
5472 	hdr_len = skb_tcp_all_headers(skb);
5473 	mss = skb_shinfo(skb)->gso_size;
5474 	if (protocol == htons(ETH_P_IP)) {
5475 		struct iphdr *iph = ip_hdr(skb);
5476 		iph->tot_len = 0;
5477 		iph->check = 0;
5478 		tcp_hdr(skb)->check = ~csum_tcpudp_magic(iph->saddr, iph->daddr,
5479 							 0, IPPROTO_TCP, 0);
5480 		cmd_length = E1000_TXD_CMD_IP;
5481 		ipcse = skb_transport_offset(skb) - 1;
5482 	} else if (skb_is_gso_v6(skb)) {
5483 		tcp_v6_gso_csum_prep(skb);
5484 		ipcse = 0;
5485 	}
5486 	ipcss = skb_network_offset(skb);
5487 	ipcso = (void *)&(ip_hdr(skb)->check) - (void *)skb->data;
5488 	tucss = skb_transport_offset(skb);
5489 	tucso = (void *)&(tcp_hdr(skb)->check) - (void *)skb->data;
5490 
5491 	cmd_length |= (E1000_TXD_CMD_DEXT | E1000_TXD_CMD_TSE |
5492 		       E1000_TXD_CMD_TCP | (skb->len - (hdr_len)));
5493 
5494 	i = tx_ring->next_to_use;
5495 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5496 	buffer_info = &tx_ring->buffer_info[i];
5497 
5498 	context_desc->lower_setup.ip_fields.ipcss = ipcss;
5499 	context_desc->lower_setup.ip_fields.ipcso = ipcso;
5500 	context_desc->lower_setup.ip_fields.ipcse = cpu_to_le16(ipcse);
5501 	context_desc->upper_setup.tcp_fields.tucss = tucss;
5502 	context_desc->upper_setup.tcp_fields.tucso = tucso;
5503 	context_desc->upper_setup.tcp_fields.tucse = 0;
5504 	context_desc->tcp_seg_setup.fields.mss = cpu_to_le16(mss);
5505 	context_desc->tcp_seg_setup.fields.hdr_len = hdr_len;
5506 	context_desc->cmd_and_length = cpu_to_le32(cmd_length);
5507 
5508 	buffer_info->time_stamp = jiffies;
5509 	buffer_info->next_to_watch = i;
5510 
5511 	i++;
5512 	if (i == tx_ring->count)
5513 		i = 0;
5514 	tx_ring->next_to_use = i;
5515 
5516 	return 1;
5517 }
5518 
e1000_tx_csum(struct e1000_ring * tx_ring,struct sk_buff * skb,__be16 protocol)5519 static bool e1000_tx_csum(struct e1000_ring *tx_ring, struct sk_buff *skb,
5520 			  __be16 protocol)
5521 {
5522 	struct e1000_adapter *adapter = tx_ring->adapter;
5523 	struct e1000_context_desc *context_desc;
5524 	struct e1000_buffer *buffer_info;
5525 	unsigned int i;
5526 	u8 css;
5527 	u32 cmd_len = E1000_TXD_CMD_DEXT;
5528 
5529 	if (skb->ip_summed != CHECKSUM_PARTIAL)
5530 		return false;
5531 
5532 	switch (protocol) {
5533 	case cpu_to_be16(ETH_P_IP):
5534 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
5535 			cmd_len |= E1000_TXD_CMD_TCP;
5536 		break;
5537 	case cpu_to_be16(ETH_P_IPV6):
5538 		/* XXX not handling all IPV6 headers */
5539 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
5540 			cmd_len |= E1000_TXD_CMD_TCP;
5541 		break;
5542 	default:
5543 		if (unlikely(net_ratelimit()))
5544 			e_warn("checksum_partial proto=%x!\n",
5545 			       be16_to_cpu(protocol));
5546 		break;
5547 	}
5548 
5549 	css = skb_checksum_start_offset(skb);
5550 
5551 	i = tx_ring->next_to_use;
5552 	buffer_info = &tx_ring->buffer_info[i];
5553 	context_desc = E1000_CONTEXT_DESC(*tx_ring, i);
5554 
5555 	context_desc->lower_setup.ip_config = 0;
5556 	context_desc->upper_setup.tcp_fields.tucss = css;
5557 	context_desc->upper_setup.tcp_fields.tucso = css + skb->csum_offset;
5558 	context_desc->upper_setup.tcp_fields.tucse = 0;
5559 	context_desc->tcp_seg_setup.data = 0;
5560 	context_desc->cmd_and_length = cpu_to_le32(cmd_len);
5561 
5562 	buffer_info->time_stamp = jiffies;
5563 	buffer_info->next_to_watch = i;
5564 
5565 	i++;
5566 	if (i == tx_ring->count)
5567 		i = 0;
5568 	tx_ring->next_to_use = i;
5569 
5570 	return true;
5571 }
5572 
e1000_tx_map(struct e1000_ring * tx_ring,struct sk_buff * skb,unsigned int first,unsigned int max_per_txd,unsigned int nr_frags)5573 static int e1000_tx_map(struct e1000_ring *tx_ring, struct sk_buff *skb,
5574 			unsigned int first, unsigned int max_per_txd,
5575 			unsigned int nr_frags)
5576 {
5577 	struct e1000_adapter *adapter = tx_ring->adapter;
5578 	struct pci_dev *pdev = adapter->pdev;
5579 	struct e1000_buffer *buffer_info;
5580 	unsigned int len = skb_headlen(skb);
5581 	unsigned int offset = 0, size, count = 0, i;
5582 	unsigned int f, bytecount, segs;
5583 
5584 	i = tx_ring->next_to_use;
5585 
5586 	while (len) {
5587 		buffer_info = &tx_ring->buffer_info[i];
5588 		size = min(len, max_per_txd);
5589 
5590 		buffer_info->length = size;
5591 		buffer_info->time_stamp = jiffies;
5592 		buffer_info->next_to_watch = i;
5593 		buffer_info->dma = dma_map_single(&pdev->dev,
5594 						  skb->data + offset,
5595 						  size, DMA_TO_DEVICE);
5596 		buffer_info->mapped_as_page = false;
5597 		if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5598 			goto dma_error;
5599 
5600 		len -= size;
5601 		offset += size;
5602 		count++;
5603 
5604 		if (len) {
5605 			i++;
5606 			if (i == tx_ring->count)
5607 				i = 0;
5608 		}
5609 	}
5610 
5611 	for (f = 0; f < nr_frags; f++) {
5612 		const skb_frag_t *frag = &skb_shinfo(skb)->frags[f];
5613 
5614 		len = skb_frag_size(frag);
5615 		offset = 0;
5616 
5617 		while (len) {
5618 			i++;
5619 			if (i == tx_ring->count)
5620 				i = 0;
5621 
5622 			buffer_info = &tx_ring->buffer_info[i];
5623 			size = min(len, max_per_txd);
5624 
5625 			buffer_info->length = size;
5626 			buffer_info->time_stamp = jiffies;
5627 			buffer_info->next_to_watch = i;
5628 			buffer_info->dma = skb_frag_dma_map(&pdev->dev, frag,
5629 							    offset, size,
5630 							    DMA_TO_DEVICE);
5631 			buffer_info->mapped_as_page = true;
5632 			if (dma_mapping_error(&pdev->dev, buffer_info->dma))
5633 				goto dma_error;
5634 
5635 			len -= size;
5636 			offset += size;
5637 			count++;
5638 		}
5639 	}
5640 
5641 	segs = skb_shinfo(skb)->gso_segs ? : 1;
5642 	/* multiply data chunks by size of headers */
5643 	bytecount = ((segs - 1) * skb_headlen(skb)) + skb->len;
5644 
5645 	tx_ring->buffer_info[i].skb = skb;
5646 	tx_ring->buffer_info[i].segs = segs;
5647 	tx_ring->buffer_info[i].bytecount = bytecount;
5648 	tx_ring->buffer_info[first].next_to_watch = i;
5649 
5650 	return count;
5651 
5652 dma_error:
5653 	dev_err(&pdev->dev, "Tx DMA map failed\n");
5654 	buffer_info->dma = 0;
5655 	if (count)
5656 		count--;
5657 
5658 	while (count--) {
5659 		if (i == 0)
5660 			i += tx_ring->count;
5661 		i--;
5662 		buffer_info = &tx_ring->buffer_info[i];
5663 		e1000_put_txbuf(tx_ring, buffer_info, true);
5664 	}
5665 
5666 	return 0;
5667 }
5668 
e1000_tx_queue(struct e1000_ring * tx_ring,int tx_flags,int count)5669 static void e1000_tx_queue(struct e1000_ring *tx_ring, int tx_flags, int count)
5670 {
5671 	struct e1000_adapter *adapter = tx_ring->adapter;
5672 	struct e1000_tx_desc *tx_desc = NULL;
5673 	struct e1000_buffer *buffer_info;
5674 	u32 txd_upper = 0, txd_lower = E1000_TXD_CMD_IFCS;
5675 	unsigned int i;
5676 
5677 	if (tx_flags & E1000_TX_FLAGS_TSO) {
5678 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D |
5679 		    E1000_TXD_CMD_TSE;
5680 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5681 
5682 		if (tx_flags & E1000_TX_FLAGS_IPV4)
5683 			txd_upper |= E1000_TXD_POPTS_IXSM << 8;
5684 	}
5685 
5686 	if (tx_flags & E1000_TX_FLAGS_CSUM) {
5687 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5688 		txd_upper |= E1000_TXD_POPTS_TXSM << 8;
5689 	}
5690 
5691 	if (tx_flags & E1000_TX_FLAGS_VLAN) {
5692 		txd_lower |= E1000_TXD_CMD_VLE;
5693 		txd_upper |= (tx_flags & E1000_TX_FLAGS_VLAN_MASK);
5694 	}
5695 
5696 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5697 		txd_lower &= ~(E1000_TXD_CMD_IFCS);
5698 
5699 	if (unlikely(tx_flags & E1000_TX_FLAGS_HWTSTAMP)) {
5700 		txd_lower |= E1000_TXD_CMD_DEXT | E1000_TXD_DTYP_D;
5701 		txd_upper |= E1000_TXD_EXTCMD_TSTAMP;
5702 	}
5703 
5704 	i = tx_ring->next_to_use;
5705 
5706 	do {
5707 		buffer_info = &tx_ring->buffer_info[i];
5708 		tx_desc = E1000_TX_DESC(*tx_ring, i);
5709 		tx_desc->buffer_addr = cpu_to_le64(buffer_info->dma);
5710 		tx_desc->lower.data = cpu_to_le32(txd_lower |
5711 						  buffer_info->length);
5712 		tx_desc->upper.data = cpu_to_le32(txd_upper);
5713 
5714 		i++;
5715 		if (i == tx_ring->count)
5716 			i = 0;
5717 	} while (--count > 0);
5718 
5719 	tx_desc->lower.data |= cpu_to_le32(adapter->txd_cmd);
5720 
5721 	/* txd_cmd re-enables FCS, so we'll re-disable it here as desired. */
5722 	if (unlikely(tx_flags & E1000_TX_FLAGS_NO_FCS))
5723 		tx_desc->lower.data &= ~(cpu_to_le32(E1000_TXD_CMD_IFCS));
5724 
5725 	/* Force memory writes to complete before letting h/w
5726 	 * know there are new descriptors to fetch.  (Only
5727 	 * applicable for weak-ordered memory model archs,
5728 	 * such as IA-64).
5729 	 */
5730 	wmb();
5731 
5732 	tx_ring->next_to_use = i;
5733 }
5734 
5735 #define MINIMUM_DHCP_PACKET_SIZE 282
e1000_transfer_dhcp_info(struct e1000_adapter * adapter,struct sk_buff * skb)5736 static int e1000_transfer_dhcp_info(struct e1000_adapter *adapter,
5737 				    struct sk_buff *skb)
5738 {
5739 	struct e1000_hw *hw = &adapter->hw;
5740 	u16 length, offset;
5741 
5742 	if (skb_vlan_tag_present(skb) &&
5743 	    !((skb_vlan_tag_get(skb) == adapter->hw.mng_cookie.vlan_id) &&
5744 	      (adapter->hw.mng_cookie.status &
5745 	       E1000_MNG_DHCP_COOKIE_STATUS_VLAN)))
5746 		return 0;
5747 
5748 	if (skb->len <= MINIMUM_DHCP_PACKET_SIZE)
5749 		return 0;
5750 
5751 	if (((struct ethhdr *)skb->data)->h_proto != htons(ETH_P_IP))
5752 		return 0;
5753 
5754 	{
5755 		const struct iphdr *ip = (struct iphdr *)((u8 *)skb->data + 14);
5756 		struct udphdr *udp;
5757 
5758 		if (ip->protocol != IPPROTO_UDP)
5759 			return 0;
5760 
5761 		udp = (struct udphdr *)((u8 *)ip + (ip->ihl << 2));
5762 		if (ntohs(udp->dest) != 67)
5763 			return 0;
5764 
5765 		offset = (u8 *)udp + 8 - skb->data;
5766 		length = skb->len - offset;
5767 		return e1000e_mng_write_dhcp_info(hw, (u8 *)udp + 8, length);
5768 	}
5769 
5770 	return 0;
5771 }
5772 
__e1000_maybe_stop_tx(struct e1000_ring * tx_ring,int size)5773 static int __e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5774 {
5775 	struct e1000_adapter *adapter = tx_ring->adapter;
5776 
5777 	netif_stop_queue(adapter->netdev);
5778 	/* Herbert's original patch had:
5779 	 *  smp_mb__after_netif_stop_queue();
5780 	 * but since that doesn't exist yet, just open code it.
5781 	 */
5782 	smp_mb();
5783 
5784 	/* We need to check again in a case another CPU has just
5785 	 * made room available.
5786 	 */
5787 	if (e1000_desc_unused(tx_ring) < size)
5788 		return -EBUSY;
5789 
5790 	/* A reprieve! */
5791 	netif_start_queue(adapter->netdev);
5792 	++adapter->restart_queue;
5793 	return 0;
5794 }
5795 
e1000_maybe_stop_tx(struct e1000_ring * tx_ring,int size)5796 static int e1000_maybe_stop_tx(struct e1000_ring *tx_ring, int size)
5797 {
5798 	BUG_ON(size > tx_ring->count);
5799 
5800 	if (e1000_desc_unused(tx_ring) >= size)
5801 		return 0;
5802 	return __e1000_maybe_stop_tx(tx_ring, size);
5803 }
5804 
e1000_xmit_frame(struct sk_buff * skb,struct net_device * netdev)5805 static netdev_tx_t e1000_xmit_frame(struct sk_buff *skb,
5806 				    struct net_device *netdev)
5807 {
5808 	struct e1000_adapter *adapter = netdev_priv(netdev);
5809 	struct e1000_ring *tx_ring = adapter->tx_ring;
5810 	unsigned int first;
5811 	unsigned int tx_flags = 0;
5812 	unsigned int len = skb_headlen(skb);
5813 	unsigned int nr_frags;
5814 	unsigned int mss;
5815 	int count = 0;
5816 	int tso;
5817 	unsigned int f;
5818 	__be16 protocol = vlan_get_protocol(skb);
5819 
5820 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5821 		dev_kfree_skb_any(skb);
5822 		return NETDEV_TX_OK;
5823 	}
5824 
5825 	if (skb->len <= 0) {
5826 		dev_kfree_skb_any(skb);
5827 		return NETDEV_TX_OK;
5828 	}
5829 
5830 	/* The minimum packet size with TCTL.PSP set is 17 bytes so
5831 	 * pad skb in order to meet this minimum size requirement
5832 	 */
5833 	if (skb_put_padto(skb, 17))
5834 		return NETDEV_TX_OK;
5835 
5836 	mss = skb_shinfo(skb)->gso_size;
5837 	if (mss) {
5838 		u8 hdr_len;
5839 
5840 		/* TSO Workaround for 82571/2/3 Controllers -- if skb->data
5841 		 * points to just header, pull a few bytes of payload from
5842 		 * frags into skb->data
5843 		 */
5844 		hdr_len = skb_tcp_all_headers(skb);
5845 		/* we do this workaround for ES2LAN, but it is un-necessary,
5846 		 * avoiding it could save a lot of cycles
5847 		 */
5848 		if (skb->data_len && (hdr_len == len)) {
5849 			unsigned int pull_size;
5850 
5851 			pull_size = min_t(unsigned int, 4, skb->data_len);
5852 			if (!__pskb_pull_tail(skb, pull_size)) {
5853 				e_err("__pskb_pull_tail failed.\n");
5854 				dev_kfree_skb_any(skb);
5855 				return NETDEV_TX_OK;
5856 			}
5857 			len = skb_headlen(skb);
5858 		}
5859 	}
5860 
5861 	/* reserve a descriptor for the offload context */
5862 	if ((mss) || (skb->ip_summed == CHECKSUM_PARTIAL))
5863 		count++;
5864 	count++;
5865 
5866 	count += DIV_ROUND_UP(len, adapter->tx_fifo_limit);
5867 
5868 	nr_frags = skb_shinfo(skb)->nr_frags;
5869 	for (f = 0; f < nr_frags; f++)
5870 		count += DIV_ROUND_UP(skb_frag_size(&skb_shinfo(skb)->frags[f]),
5871 				      adapter->tx_fifo_limit);
5872 
5873 	if (adapter->hw.mac.tx_pkt_filtering)
5874 		e1000_transfer_dhcp_info(adapter, skb);
5875 
5876 	/* need: count + 2 desc gap to keep tail from touching
5877 	 * head, otherwise try next time
5878 	 */
5879 	if (e1000_maybe_stop_tx(tx_ring, count + 2))
5880 		return NETDEV_TX_BUSY;
5881 
5882 	if (skb_vlan_tag_present(skb)) {
5883 		tx_flags |= E1000_TX_FLAGS_VLAN;
5884 		tx_flags |= (skb_vlan_tag_get(skb) <<
5885 			     E1000_TX_FLAGS_VLAN_SHIFT);
5886 	}
5887 
5888 	first = tx_ring->next_to_use;
5889 
5890 	tso = e1000_tso(tx_ring, skb, protocol);
5891 	if (tso < 0) {
5892 		dev_kfree_skb_any(skb);
5893 		return NETDEV_TX_OK;
5894 	}
5895 
5896 	if (tso)
5897 		tx_flags |= E1000_TX_FLAGS_TSO;
5898 	else if (e1000_tx_csum(tx_ring, skb, protocol))
5899 		tx_flags |= E1000_TX_FLAGS_CSUM;
5900 
5901 	/* Old method was to assume IPv4 packet by default if TSO was enabled.
5902 	 * 82571 hardware supports TSO capabilities for IPv6 as well...
5903 	 * no longer assume, we must.
5904 	 */
5905 	if (protocol == htons(ETH_P_IP))
5906 		tx_flags |= E1000_TX_FLAGS_IPV4;
5907 
5908 	if (unlikely(skb->no_fcs))
5909 		tx_flags |= E1000_TX_FLAGS_NO_FCS;
5910 
5911 	/* if count is 0 then mapping error has occurred */
5912 	count = e1000_tx_map(tx_ring, skb, first, adapter->tx_fifo_limit,
5913 			     nr_frags);
5914 	if (count) {
5915 		if (unlikely(skb_shinfo(skb)->tx_flags & SKBTX_HW_TSTAMP) &&
5916 		    (adapter->flags & FLAG_HAS_HW_TIMESTAMP)) {
5917 			if (!adapter->tx_hwtstamp_skb) {
5918 				skb_shinfo(skb)->tx_flags |= SKBTX_IN_PROGRESS;
5919 				tx_flags |= E1000_TX_FLAGS_HWTSTAMP;
5920 				adapter->tx_hwtstamp_skb = skb_get(skb);
5921 				adapter->tx_hwtstamp_start = jiffies;
5922 				schedule_work(&adapter->tx_hwtstamp_work);
5923 			} else {
5924 				adapter->tx_hwtstamp_skipped++;
5925 			}
5926 		}
5927 
5928 		skb_tx_timestamp(skb);
5929 
5930 		netdev_sent_queue(netdev, skb->len);
5931 		e1000_tx_queue(tx_ring, tx_flags, count);
5932 		/* Make sure there is space in the ring for the next send. */
5933 		e1000_maybe_stop_tx(tx_ring,
5934 				    ((MAX_SKB_FRAGS + 1) *
5935 				     DIV_ROUND_UP(PAGE_SIZE,
5936 						  adapter->tx_fifo_limit) + 4));
5937 
5938 		if (!netdev_xmit_more() ||
5939 		    netif_xmit_stopped(netdev_get_tx_queue(netdev, 0))) {
5940 			if (adapter->flags2 & FLAG2_PCIM2PCI_ARBITER_WA)
5941 				e1000e_update_tdt_wa(tx_ring,
5942 						     tx_ring->next_to_use);
5943 			else
5944 				writel(tx_ring->next_to_use, tx_ring->tail);
5945 		}
5946 	} else {
5947 		dev_kfree_skb_any(skb);
5948 		tx_ring->buffer_info[first].time_stamp = 0;
5949 		tx_ring->next_to_use = first;
5950 	}
5951 
5952 	return NETDEV_TX_OK;
5953 }
5954 
5955 /**
5956  * e1000_tx_timeout - Respond to a Tx Hang
5957  * @netdev: network interface device structure
5958  * @txqueue: index of the hung queue (unused)
5959  **/
e1000_tx_timeout(struct net_device * netdev,unsigned int __always_unused txqueue)5960 static void e1000_tx_timeout(struct net_device *netdev, unsigned int __always_unused txqueue)
5961 {
5962 	struct e1000_adapter *adapter = netdev_priv(netdev);
5963 
5964 	/* Do the reset outside of interrupt context */
5965 	adapter->tx_timeout_count++;
5966 	schedule_work(&adapter->reset_task);
5967 }
5968 
e1000_reset_task(struct work_struct * work)5969 static void e1000_reset_task(struct work_struct *work)
5970 {
5971 	struct e1000_adapter *adapter;
5972 	adapter = container_of(work, struct e1000_adapter, reset_task);
5973 
5974 	rtnl_lock();
5975 	/* don't run the task if already down */
5976 	if (test_bit(__E1000_DOWN, &adapter->state)) {
5977 		rtnl_unlock();
5978 		return;
5979 	}
5980 
5981 	if (!(adapter->flags & FLAG_RESTART_NOW)) {
5982 		e1000e_dump(adapter);
5983 		e_err("Reset adapter unexpectedly\n");
5984 	}
5985 	e1000e_reinit_locked(adapter);
5986 	rtnl_unlock();
5987 }
5988 
5989 /**
5990  * e1000e_get_stats64 - Get System Network Statistics
5991  * @netdev: network interface device structure
5992  * @stats: rtnl_link_stats64 pointer
5993  *
5994  * Returns the address of the device statistics structure.
5995  **/
e1000e_get_stats64(struct net_device * netdev,struct rtnl_link_stats64 * stats)5996 void e1000e_get_stats64(struct net_device *netdev,
5997 			struct rtnl_link_stats64 *stats)
5998 {
5999 	struct e1000_adapter *adapter = netdev_priv(netdev);
6000 
6001 	spin_lock(&adapter->stats64_lock);
6002 	e1000e_update_stats(adapter);
6003 	/* Fill out the OS statistics structure */
6004 	stats->rx_bytes = adapter->stats.gorc;
6005 	stats->rx_packets = adapter->stats.gprc;
6006 	stats->tx_bytes = adapter->stats.gotc;
6007 	stats->tx_packets = adapter->stats.gptc;
6008 	stats->multicast = adapter->stats.mprc;
6009 	stats->collisions = adapter->stats.colc;
6010 
6011 	/* Rx Errors */
6012 
6013 	/* RLEC on some newer hardware can be incorrect so build
6014 	 * our own version based on RUC and ROC
6015 	 */
6016 	stats->rx_errors = adapter->stats.rxerrc +
6017 	    adapter->stats.crcerrs + adapter->stats.algnerrc +
6018 	    adapter->stats.ruc + adapter->stats.roc + adapter->stats.cexterr;
6019 	stats->rx_length_errors = adapter->stats.ruc + adapter->stats.roc;
6020 	stats->rx_crc_errors = adapter->stats.crcerrs;
6021 	stats->rx_frame_errors = adapter->stats.algnerrc;
6022 	stats->rx_missed_errors = adapter->stats.mpc;
6023 
6024 	/* Tx Errors */
6025 	stats->tx_errors = adapter->stats.ecol + adapter->stats.latecol;
6026 	stats->tx_aborted_errors = adapter->stats.ecol;
6027 	stats->tx_window_errors = adapter->stats.latecol;
6028 	stats->tx_carrier_errors = adapter->stats.tncrs;
6029 
6030 	/* Tx Dropped needs to be maintained elsewhere */
6031 
6032 	spin_unlock(&adapter->stats64_lock);
6033 }
6034 
6035 /**
6036  * e1000_change_mtu - Change the Maximum Transfer Unit
6037  * @netdev: network interface device structure
6038  * @new_mtu: new value for maximum frame size
6039  *
6040  * Returns 0 on success, negative on failure
6041  **/
e1000_change_mtu(struct net_device * netdev,int new_mtu)6042 static int e1000_change_mtu(struct net_device *netdev, int new_mtu)
6043 {
6044 	struct e1000_adapter *adapter = netdev_priv(netdev);
6045 	int max_frame = new_mtu + VLAN_ETH_HLEN + ETH_FCS_LEN;
6046 
6047 	/* Jumbo frame support */
6048 	if ((new_mtu > ETH_DATA_LEN) &&
6049 	    !(adapter->flags & FLAG_HAS_JUMBO_FRAMES)) {
6050 		e_err("Jumbo Frames not supported.\n");
6051 		return -EINVAL;
6052 	}
6053 
6054 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
6055 	if ((adapter->hw.mac.type >= e1000_pch2lan) &&
6056 	    !(adapter->flags2 & FLAG2_CRC_STRIPPING) &&
6057 	    (new_mtu > ETH_DATA_LEN)) {
6058 		e_err("Jumbo Frames not supported on this device when CRC stripping is disabled.\n");
6059 		return -EINVAL;
6060 	}
6061 
6062 	while (test_and_set_bit(__E1000_RESETTING, &adapter->state))
6063 		usleep_range(1000, 1100);
6064 	/* e1000e_down -> e1000e_reset dependent on max_frame_size & mtu */
6065 	adapter->max_frame_size = max_frame;
6066 	netdev_dbg(netdev, "changing MTU from %d to %d\n",
6067 		   netdev->mtu, new_mtu);
6068 	WRITE_ONCE(netdev->mtu, new_mtu);
6069 
6070 	pm_runtime_get_sync(netdev->dev.parent);
6071 
6072 	if (netif_running(netdev))
6073 		e1000e_down(adapter, true);
6074 
6075 	/* NOTE: netdev_alloc_skb reserves 16 bytes, and typically NET_IP_ALIGN
6076 	 * means we reserve 2 more, this pushes us to allocate from the next
6077 	 * larger slab size.
6078 	 * i.e. RXBUFFER_2048 --> size-4096 slab
6079 	 * However with the new *_jumbo_rx* routines, jumbo receives will use
6080 	 * fragmented skbs
6081 	 */
6082 
6083 	if (max_frame <= 2048)
6084 		adapter->rx_buffer_len = 2048;
6085 	else
6086 		adapter->rx_buffer_len = 4096;
6087 
6088 	/* adjust allocation if LPE protects us, and we aren't using SBP */
6089 	if (max_frame <= (VLAN_ETH_FRAME_LEN + ETH_FCS_LEN))
6090 		adapter->rx_buffer_len = VLAN_ETH_FRAME_LEN + ETH_FCS_LEN;
6091 
6092 	if (netif_running(netdev))
6093 		e1000e_up(adapter);
6094 	else
6095 		e1000e_reset(adapter);
6096 
6097 	pm_runtime_put_sync(netdev->dev.parent);
6098 
6099 	clear_bit(__E1000_RESETTING, &adapter->state);
6100 
6101 	return 0;
6102 }
6103 
e1000_ioctl(struct net_device * netdev,struct ifreq * ifr,int cmd)6104 static int e1000_ioctl(struct net_device *netdev, struct ifreq *ifr, int cmd)
6105 {
6106 	struct e1000_adapter *adapter = netdev_priv(netdev);
6107 	struct mii_ioctl_data *data = if_mii(ifr);
6108 
6109 	if (adapter->hw.phy.media_type != e1000_media_type_copper)
6110 		return -EOPNOTSUPP;
6111 
6112 	switch (cmd) {
6113 	case SIOCGMIIPHY:
6114 		data->phy_id = adapter->hw.phy.addr;
6115 		break;
6116 	case SIOCGMIIREG:
6117 		e1000_phy_read_status(adapter);
6118 
6119 		switch (data->reg_num & 0x1F) {
6120 		case MII_BMCR:
6121 			data->val_out = adapter->phy_regs.bmcr;
6122 			break;
6123 		case MII_BMSR:
6124 			data->val_out = adapter->phy_regs.bmsr;
6125 			break;
6126 		case MII_PHYSID1:
6127 			data->val_out = (adapter->hw.phy.id >> 16);
6128 			break;
6129 		case MII_PHYSID2:
6130 			data->val_out = (adapter->hw.phy.id & 0xFFFF);
6131 			break;
6132 		case MII_ADVERTISE:
6133 			data->val_out = adapter->phy_regs.advertise;
6134 			break;
6135 		case MII_LPA:
6136 			data->val_out = adapter->phy_regs.lpa;
6137 			break;
6138 		case MII_EXPANSION:
6139 			data->val_out = adapter->phy_regs.expansion;
6140 			break;
6141 		case MII_CTRL1000:
6142 			data->val_out = adapter->phy_regs.ctrl1000;
6143 			break;
6144 		case MII_STAT1000:
6145 			data->val_out = adapter->phy_regs.stat1000;
6146 			break;
6147 		case MII_ESTATUS:
6148 			data->val_out = adapter->phy_regs.estatus;
6149 			break;
6150 		default:
6151 			return -EIO;
6152 		}
6153 		break;
6154 	case SIOCSMIIREG:
6155 	default:
6156 		return -EOPNOTSUPP;
6157 	}
6158 	return 0;
6159 }
6160 
6161 /**
6162  * e1000e_hwtstamp_set - control hardware time stamping
6163  * @netdev: network interface device structure
6164  * @config: timestamp configuration
6165  * @extack: netlink extended ACK report
6166  *
6167  * Outgoing time stamping can be enabled and disabled. Play nice and
6168  * disable it when requested, although it shouldn't cause any overhead
6169  * when no packet needs it. At most one packet in the queue may be
6170  * marked for time stamping, otherwise it would be impossible to tell
6171  * for sure to which packet the hardware time stamp belongs.
6172  *
6173  * Incoming time stamping has to be configured via the hardware filters.
6174  * Not all combinations are supported, in particular event type has to be
6175  * specified. Matching the kind of event packet is not supported, with the
6176  * exception of "all V2 events regardless of level 2 or 4".
6177  **/
e1000e_hwtstamp_set(struct net_device * netdev,struct kernel_hwtstamp_config * config,struct netlink_ext_ack * extack)6178 static int e1000e_hwtstamp_set(struct net_device *netdev,
6179 			       struct kernel_hwtstamp_config *config,
6180 			       struct netlink_ext_ack *extack)
6181 {
6182 	struct e1000_adapter *adapter = netdev_priv(netdev);
6183 	int ret_val;
6184 
6185 	ret_val = e1000e_config_hwtstamp(adapter, config, extack);
6186 	if (ret_val)
6187 		return ret_val;
6188 
6189 	switch (config->rx_filter) {
6190 	case HWTSTAMP_FILTER_PTP_V2_L4_SYNC:
6191 	case HWTSTAMP_FILTER_PTP_V2_L2_SYNC:
6192 	case HWTSTAMP_FILTER_PTP_V2_SYNC:
6193 	case HWTSTAMP_FILTER_PTP_V2_L4_DELAY_REQ:
6194 	case HWTSTAMP_FILTER_PTP_V2_L2_DELAY_REQ:
6195 	case HWTSTAMP_FILTER_PTP_V2_DELAY_REQ:
6196 		/* With V2 type filters which specify a Sync or Delay Request,
6197 		 * Path Delay Request/Response messages are also time stamped
6198 		 * by hardware so notify the caller the requested packets plus
6199 		 * some others are time stamped.
6200 		 */
6201 		config->rx_filter = HWTSTAMP_FILTER_SOME;
6202 		break;
6203 	default:
6204 		break;
6205 	}
6206 
6207 	return 0;
6208 }
6209 
e1000e_hwtstamp_get(struct net_device * netdev,struct kernel_hwtstamp_config * kernel_config)6210 static int e1000e_hwtstamp_get(struct net_device *netdev,
6211 			       struct kernel_hwtstamp_config *kernel_config)
6212 {
6213 	struct e1000_adapter *adapter = netdev_priv(netdev);
6214 
6215 	*kernel_config = adapter->hwtstamp_config;
6216 
6217 	return 0;
6218 }
6219 
e1000_init_phy_wakeup(struct e1000_adapter * adapter,u32 wufc)6220 static int e1000_init_phy_wakeup(struct e1000_adapter *adapter, u32 wufc)
6221 {
6222 	struct e1000_hw *hw = &adapter->hw;
6223 	u32 i, mac_reg, wuc;
6224 	u16 phy_reg, wuc_enable;
6225 	int retval;
6226 
6227 	/* copy MAC RARs to PHY RARs */
6228 	e1000_copy_rx_addrs_to_phy_ich8lan(hw);
6229 
6230 	retval = hw->phy.ops.acquire(hw);
6231 	if (retval) {
6232 		e_err("Could not acquire PHY\n");
6233 		return retval;
6234 	}
6235 
6236 	/* Enable access to wakeup registers on and set page to BM_WUC_PAGE */
6237 	retval = e1000_enable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6238 	if (retval)
6239 		goto release;
6240 
6241 	/* copy MAC MTA to PHY MTA - only needed for pchlan */
6242 	for (i = 0; i < adapter->hw.mac.mta_reg_count; i++) {
6243 		mac_reg = E1000_READ_REG_ARRAY(hw, E1000_MTA, i);
6244 		hw->phy.ops.write_reg_page(hw, BM_MTA(i),
6245 					   (u16)(mac_reg & 0xFFFF));
6246 		hw->phy.ops.write_reg_page(hw, BM_MTA(i) + 1,
6247 					   (u16)((mac_reg >> 16) & 0xFFFF));
6248 	}
6249 
6250 	/* configure PHY Rx Control register */
6251 	hw->phy.ops.read_reg_page(&adapter->hw, BM_RCTL, &phy_reg);
6252 	mac_reg = er32(RCTL);
6253 	if (mac_reg & E1000_RCTL_UPE)
6254 		phy_reg |= BM_RCTL_UPE;
6255 	if (mac_reg & E1000_RCTL_MPE)
6256 		phy_reg |= BM_RCTL_MPE;
6257 	phy_reg &= ~(BM_RCTL_MO_MASK);
6258 	if (mac_reg & E1000_RCTL_MO_3)
6259 		phy_reg |= (FIELD_GET(E1000_RCTL_MO_3, mac_reg)
6260 			    << BM_RCTL_MO_SHIFT);
6261 	if (mac_reg & E1000_RCTL_BAM)
6262 		phy_reg |= BM_RCTL_BAM;
6263 	if (mac_reg & E1000_RCTL_PMCF)
6264 		phy_reg |= BM_RCTL_PMCF;
6265 	mac_reg = er32(CTRL);
6266 	if (mac_reg & E1000_CTRL_RFCE)
6267 		phy_reg |= BM_RCTL_RFCE;
6268 	hw->phy.ops.write_reg_page(&adapter->hw, BM_RCTL, phy_reg);
6269 
6270 	wuc = E1000_WUC_PME_EN;
6271 	if (wufc & (E1000_WUFC_MAG | E1000_WUFC_LNKC))
6272 		wuc |= E1000_WUC_APME;
6273 
6274 	/* enable PHY wakeup in MAC register */
6275 	ew32(WUFC, wufc);
6276 	ew32(WUC, (E1000_WUC_PHY_WAKE | E1000_WUC_APMPME |
6277 		   E1000_WUC_PME_STATUS | wuc));
6278 
6279 	/* configure and enable PHY wakeup in PHY registers */
6280 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUFC, wufc);
6281 	hw->phy.ops.write_reg_page(&adapter->hw, BM_WUC, wuc);
6282 
6283 	/* activate PHY wakeup */
6284 	wuc_enable |= BM_WUC_ENABLE_BIT | BM_WUC_HOST_WU_BIT;
6285 	retval = e1000_disable_phy_wakeup_reg_access_bm(hw, &wuc_enable);
6286 	if (retval)
6287 		e_err("Could not set PHY Host Wakeup bit\n");
6288 release:
6289 	hw->phy.ops.release(hw);
6290 
6291 	return retval;
6292 }
6293 
e1000e_flush_lpic(struct pci_dev * pdev)6294 static void e1000e_flush_lpic(struct pci_dev *pdev)
6295 {
6296 	struct net_device *netdev = pci_get_drvdata(pdev);
6297 	struct e1000_adapter *adapter = netdev_priv(netdev);
6298 	struct e1000_hw *hw = &adapter->hw;
6299 	u32 ret_val;
6300 
6301 	pm_runtime_get_sync(netdev->dev.parent);
6302 
6303 	ret_val = hw->phy.ops.acquire(hw);
6304 	if (ret_val)
6305 		goto fl_out;
6306 
6307 	pr_info("EEE TX LPI TIMER: %08X\n",
6308 		er32(LPIC) >> E1000_LPIC_LPIET_SHIFT);
6309 
6310 	hw->phy.ops.release(hw);
6311 
6312 fl_out:
6313 	pm_runtime_put_sync(netdev->dev.parent);
6314 }
6315 
6316 /* S0ix implementation */
e1000e_s0ix_entry_flow(struct e1000_adapter * adapter)6317 static void e1000e_s0ix_entry_flow(struct e1000_adapter *adapter)
6318 {
6319 	struct e1000_hw *hw = &adapter->hw;
6320 	u32 mac_data;
6321 	u16 phy_data;
6322 
6323 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6324 	    hw->mac.type >= e1000_pch_adp) {
6325 		/* Request ME configure the device for S0ix */
6326 		mac_data = er32(H2ME);
6327 		mac_data |= E1000_H2ME_START_DPG;
6328 		mac_data &= ~E1000_H2ME_EXIT_DPG;
6329 		trace_e1000e_trace_mac_register(mac_data);
6330 		ew32(H2ME, mac_data);
6331 	} else {
6332 		/* Request driver configure the device to S0ix */
6333 		/* Disable the periodic inband message,
6334 		 * don't request PCIe clock in K1 page770_17[10:9] = 10b
6335 		 */
6336 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6337 		phy_data &= ~HV_PM_CTRL_K1_CLK_REQ;
6338 		phy_data |= BIT(10);
6339 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6340 
6341 		/* Make sure we don't exit K1 every time a new packet arrives
6342 		 * 772_29[5] = 1 CS_Mode_Stay_In_K1
6343 		 */
6344 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6345 		phy_data |= BIT(5);
6346 		e1e_wphy(hw, I217_CGFREG, phy_data);
6347 
6348 		/* Change the MAC/PHY interface to SMBus
6349 		 * Force the SMBus in PHY page769_23[0] = 1
6350 		 * Force the SMBus in MAC CTRL_EXT[11] = 1
6351 		 */
6352 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6353 		phy_data |= CV_SMB_CTRL_FORCE_SMBUS;
6354 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6355 		mac_data = er32(CTRL_EXT);
6356 		mac_data |= E1000_CTRL_EXT_FORCE_SMBUS;
6357 		ew32(CTRL_EXT, mac_data);
6358 
6359 		/* DFT control: PHY bit: page769_20[0] = 1
6360 		 * page769_20[7] - PHY PLL stop
6361 		 * page769_20[8] - PHY go to the electrical idle
6362 		 * page769_20[9] - PHY serdes disable
6363 		 * Gate PPW via EXTCNF_CTRL - set 0x0F00[7] = 1
6364 		 */
6365 		e1e_rphy(hw, I82579_DFT_CTRL, &phy_data);
6366 		phy_data |= BIT(0);
6367 		phy_data |= BIT(7);
6368 		phy_data |= BIT(8);
6369 		phy_data |= BIT(9);
6370 		e1e_wphy(hw, I82579_DFT_CTRL, phy_data);
6371 
6372 		mac_data = er32(EXTCNF_CTRL);
6373 		mac_data |= E1000_EXTCNF_CTRL_GATE_PHY_CFG;
6374 		ew32(EXTCNF_CTRL, mac_data);
6375 
6376 		/* Disable disconnected cable conditioning for Power Gating */
6377 		mac_data = er32(DPGFR);
6378 		mac_data |= BIT(2);
6379 		ew32(DPGFR, mac_data);
6380 
6381 		/* Enable the Dynamic Clock Gating in the DMA and MAC */
6382 		mac_data = er32(CTRL_EXT);
6383 		mac_data |= E1000_CTRL_EXT_DMA_DYN_CLK_EN;
6384 		ew32(CTRL_EXT, mac_data);
6385 	}
6386 
6387 	/* Enable the Dynamic Power Gating in the MAC */
6388 	mac_data = er32(FEXTNVM7);
6389 	mac_data |= BIT(22);
6390 	ew32(FEXTNVM7, mac_data);
6391 
6392 	/* Don't wake from dynamic Power Gating with clock request */
6393 	mac_data = er32(FEXTNVM12);
6394 	mac_data |= BIT(12);
6395 	ew32(FEXTNVM12, mac_data);
6396 
6397 	/* Ungate PGCB clock */
6398 	mac_data = er32(FEXTNVM9);
6399 	mac_data &= ~BIT(28);
6400 	ew32(FEXTNVM9, mac_data);
6401 
6402 	/* Enable K1 off to enable mPHY Power Gating */
6403 	mac_data = er32(FEXTNVM6);
6404 	mac_data |= BIT(31);
6405 	ew32(FEXTNVM6, mac_data);
6406 
6407 	/* Enable mPHY power gating for any link and speed */
6408 	mac_data = er32(FEXTNVM8);
6409 	mac_data |= BIT(9);
6410 	ew32(FEXTNVM8, mac_data);
6411 
6412 	/* No MAC DPG gating SLP_S0 in modern standby
6413 	 * Switch the logic of the lanphypc to use PMC counter
6414 	 */
6415 	mac_data = er32(FEXTNVM5);
6416 	mac_data |= BIT(7);
6417 	ew32(FEXTNVM5, mac_data);
6418 
6419 	/* Disable the time synchronization clock */
6420 	mac_data = er32(FEXTNVM7);
6421 	mac_data |= BIT(31);
6422 	mac_data &= ~BIT(0);
6423 	ew32(FEXTNVM7, mac_data);
6424 
6425 	/* Dynamic Power Gating Enable */
6426 	mac_data = er32(CTRL_EXT);
6427 	mac_data |= BIT(3);
6428 	ew32(CTRL_EXT, mac_data);
6429 
6430 	/* Check MAC Tx/Rx packet buffer pointers.
6431 	 * Reset MAC Tx/Rx packet buffer pointers to suppress any
6432 	 * pending traffic indication that would prevent power gating.
6433 	 */
6434 	mac_data = er32(TDFH);
6435 	if (mac_data)
6436 		ew32(TDFH, 0);
6437 	mac_data = er32(TDFT);
6438 	if (mac_data)
6439 		ew32(TDFT, 0);
6440 	mac_data = er32(TDFHS);
6441 	if (mac_data)
6442 		ew32(TDFHS, 0);
6443 	mac_data = er32(TDFTS);
6444 	if (mac_data)
6445 		ew32(TDFTS, 0);
6446 	mac_data = er32(TDFPC);
6447 	if (mac_data)
6448 		ew32(TDFPC, 0);
6449 	mac_data = er32(RDFH);
6450 	if (mac_data)
6451 		ew32(RDFH, 0);
6452 	mac_data = er32(RDFT);
6453 	if (mac_data)
6454 		ew32(RDFT, 0);
6455 	mac_data = er32(RDFHS);
6456 	if (mac_data)
6457 		ew32(RDFHS, 0);
6458 	mac_data = er32(RDFTS);
6459 	if (mac_data)
6460 		ew32(RDFTS, 0);
6461 	mac_data = er32(RDFPC);
6462 	if (mac_data)
6463 		ew32(RDFPC, 0);
6464 }
6465 
e1000e_s0ix_exit_flow(struct e1000_adapter * adapter)6466 static void e1000e_s0ix_exit_flow(struct e1000_adapter *adapter)
6467 {
6468 	struct e1000_hw *hw = &adapter->hw;
6469 	bool firmware_bug = false;
6470 	u32 mac_data;
6471 	u16 phy_data;
6472 	u32 i = 0;
6473 
6474 	if (er32(FWSM) & E1000_ICH_FWSM_FW_VALID &&
6475 	    hw->mac.type >= e1000_pch_adp) {
6476 		/* Keep the GPT clock enabled for CSME */
6477 		mac_data = er32(FEXTNVM);
6478 		mac_data |= BIT(3);
6479 		ew32(FEXTNVM, mac_data);
6480 		/* Request ME unconfigure the device from S0ix */
6481 		mac_data = er32(H2ME);
6482 		mac_data &= ~E1000_H2ME_START_DPG;
6483 		mac_data |= E1000_H2ME_EXIT_DPG;
6484 		trace_e1000e_trace_mac_register(mac_data);
6485 		ew32(H2ME, mac_data);
6486 
6487 		/* Poll up to 2.5 seconds for ME to unconfigure DPG.
6488 		 * If this takes more than 1 second, show a warning indicating a
6489 		 * firmware bug
6490 		 */
6491 		while (!(er32(EXFWSM) & E1000_EXFWSM_DPG_EXIT_DONE)) {
6492 			if (i > 100 && !firmware_bug)
6493 				firmware_bug = true;
6494 
6495 			if (i++ == 250) {
6496 				e_dbg("Timeout (firmware bug): %d msec\n",
6497 				      i * 10);
6498 				break;
6499 			}
6500 
6501 			usleep_range(10000, 11000);
6502 		}
6503 		if (firmware_bug)
6504 			e_warn("DPG_EXIT_DONE took %d msec. This is a firmware bug\n",
6505 			       i * 10);
6506 		else
6507 			e_dbg("DPG_EXIT_DONE cleared after %d msec\n", i * 10);
6508 	} else {
6509 		/* Request driver unconfigure the device from S0ix */
6510 
6511 		/* Cancel disable disconnected cable conditioning
6512 		 * for Power Gating
6513 		 */
6514 		mac_data = er32(DPGFR);
6515 		mac_data &= ~BIT(2);
6516 		ew32(DPGFR, mac_data);
6517 
6518 		/* Disable the Dynamic Clock Gating in the DMA and MAC */
6519 		mac_data = er32(CTRL_EXT);
6520 		mac_data &= 0xFFF7FFFF;
6521 		ew32(CTRL_EXT, mac_data);
6522 
6523 		/* Enable the periodic inband message,
6524 		 * Request PCIe clock in K1 page770_17[10:9] =01b
6525 		 */
6526 		e1e_rphy(hw, HV_PM_CTRL, &phy_data);
6527 		phy_data &= 0xFBFF;
6528 		phy_data |= HV_PM_CTRL_K1_CLK_REQ;
6529 		e1e_wphy(hw, HV_PM_CTRL, phy_data);
6530 
6531 		/* Return back configuration
6532 		 * 772_29[5] = 0 CS_Mode_Stay_In_K1
6533 		 */
6534 		e1e_rphy(hw, I217_CGFREG, &phy_data);
6535 		phy_data &= 0xFFDF;
6536 		e1e_wphy(hw, I217_CGFREG, phy_data);
6537 
6538 		/* Change the MAC/PHY interface to Kumeran
6539 		 * Unforce the SMBus in PHY page769_23[0] = 0
6540 		 * Unforce the SMBus in MAC CTRL_EXT[11] = 0
6541 		 */
6542 		e1e_rphy(hw, CV_SMB_CTRL, &phy_data);
6543 		phy_data &= ~CV_SMB_CTRL_FORCE_SMBUS;
6544 		e1e_wphy(hw, CV_SMB_CTRL, phy_data);
6545 		mac_data = er32(CTRL_EXT);
6546 		mac_data &= ~E1000_CTRL_EXT_FORCE_SMBUS;
6547 		ew32(CTRL_EXT, mac_data);
6548 	}
6549 
6550 	/* Disable Dynamic Power Gating */
6551 	mac_data = er32(CTRL_EXT);
6552 	mac_data &= 0xFFFFFFF7;
6553 	ew32(CTRL_EXT, mac_data);
6554 
6555 	/* Enable the time synchronization clock */
6556 	mac_data = er32(FEXTNVM7);
6557 	mac_data &= ~BIT(31);
6558 	mac_data |= BIT(0);
6559 	ew32(FEXTNVM7, mac_data);
6560 
6561 	/* Disable the Dynamic Power Gating in the MAC */
6562 	mac_data = er32(FEXTNVM7);
6563 	mac_data &= 0xFFBFFFFF;
6564 	ew32(FEXTNVM7, mac_data);
6565 
6566 	/* Disable mPHY power gating for any link and speed */
6567 	mac_data = er32(FEXTNVM8);
6568 	mac_data &= ~BIT(9);
6569 	ew32(FEXTNVM8, mac_data);
6570 
6571 	/* Disable K1 off */
6572 	mac_data = er32(FEXTNVM6);
6573 	mac_data &= ~BIT(31);
6574 	ew32(FEXTNVM6, mac_data);
6575 
6576 	/* Disable Ungate PGCB clock */
6577 	mac_data = er32(FEXTNVM9);
6578 	mac_data |= BIT(28);
6579 	ew32(FEXTNVM9, mac_data);
6580 
6581 	/* Cancel not waking from dynamic
6582 	 * Power Gating with clock request
6583 	 */
6584 	mac_data = er32(FEXTNVM12);
6585 	mac_data &= ~BIT(12);
6586 	ew32(FEXTNVM12, mac_data);
6587 
6588 	/* Revert the lanphypc logic to use the internal Gbe counter
6589 	 * and not the PMC counter
6590 	 */
6591 	mac_data = er32(FEXTNVM5);
6592 	mac_data &= 0xFFFFFF7F;
6593 	ew32(FEXTNVM5, mac_data);
6594 }
6595 
e1000e_pm_freeze(struct device * dev)6596 static int e1000e_pm_freeze(struct device *dev)
6597 {
6598 	struct net_device *netdev = dev_get_drvdata(dev);
6599 	struct e1000_adapter *adapter = netdev_priv(netdev);
6600 	bool present;
6601 
6602 	rtnl_lock();
6603 
6604 	present = netif_device_present(netdev);
6605 	netif_device_detach(netdev);
6606 
6607 	if (present && netif_running(netdev)) {
6608 		int count = E1000_CHECK_RESET_COUNT;
6609 
6610 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
6611 			usleep_range(10000, 11000);
6612 
6613 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
6614 
6615 		/* Quiesce the device without resetting the hardware */
6616 		e1000e_down(adapter, false);
6617 		e1000_free_irq(adapter);
6618 	}
6619 	rtnl_unlock();
6620 
6621 	e1000e_reset_interrupt_capability(adapter);
6622 
6623 	/* Allow time for pending master requests to run */
6624 	e1000e_disable_pcie_master(&adapter->hw);
6625 
6626 	return 0;
6627 }
6628 
__e1000_shutdown(struct pci_dev * pdev,bool runtime)6629 static int __e1000_shutdown(struct pci_dev *pdev, bool runtime)
6630 {
6631 	struct net_device *netdev = pci_get_drvdata(pdev);
6632 	struct e1000_adapter *adapter = netdev_priv(netdev);
6633 	struct e1000_hw *hw = &adapter->hw;
6634 	u32 ctrl, ctrl_ext, rctl, status, wufc;
6635 	int retval = 0;
6636 
6637 	/* Runtime suspend should only enable wakeup for link changes */
6638 	if (runtime)
6639 		wufc = E1000_WUFC_LNKC;
6640 	else if (device_may_wakeup(&pdev->dev))
6641 		wufc = adapter->wol;
6642 	else
6643 		wufc = 0;
6644 
6645 	status = er32(STATUS);
6646 	if (status & E1000_STATUS_LU)
6647 		wufc &= ~E1000_WUFC_LNKC;
6648 
6649 	if (wufc) {
6650 		e1000_setup_rctl(adapter);
6651 		e1000e_set_rx_mode(netdev);
6652 
6653 		/* turn on all-multi mode if wake on multicast is enabled */
6654 		if (wufc & E1000_WUFC_MC) {
6655 			rctl = er32(RCTL);
6656 			rctl |= E1000_RCTL_MPE;
6657 			ew32(RCTL, rctl);
6658 		}
6659 
6660 		ctrl = er32(CTRL);
6661 		ctrl |= E1000_CTRL_ADVD3WUC;
6662 		if (!(adapter->flags2 & FLAG2_HAS_PHY_WAKEUP))
6663 			ctrl |= E1000_CTRL_EN_PHY_PWR_MGMT;
6664 		ew32(CTRL, ctrl);
6665 
6666 		if (adapter->hw.phy.media_type == e1000_media_type_fiber ||
6667 		    adapter->hw.phy.media_type ==
6668 		    e1000_media_type_internal_serdes) {
6669 			/* keep the laser running in D3 */
6670 			ctrl_ext = er32(CTRL_EXT);
6671 			ctrl_ext |= E1000_CTRL_EXT_SDP3_DATA;
6672 			ew32(CTRL_EXT, ctrl_ext);
6673 		}
6674 
6675 		if (!runtime)
6676 			e1000e_power_up_phy(adapter);
6677 
6678 		if (adapter->flags & FLAG_IS_ICH)
6679 			e1000_suspend_workarounds_ich8lan(&adapter->hw);
6680 
6681 		if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6682 			/* enable wakeup by the PHY */
6683 			retval = e1000_init_phy_wakeup(adapter, wufc);
6684 			if (retval) {
6685 				e_err("Failed to enable wakeup\n");
6686 				goto skip_phy_configurations;
6687 			}
6688 		} else {
6689 			/* enable wakeup by the MAC */
6690 			ew32(WUFC, wufc);
6691 			ew32(WUC, E1000_WUC_PME_EN);
6692 		}
6693 	} else {
6694 		ew32(WUC, 0);
6695 		ew32(WUFC, 0);
6696 
6697 		e1000_power_down_phy(adapter);
6698 	}
6699 
6700 	if (adapter->hw.phy.type == e1000_phy_igp_3) {
6701 		e1000e_igp3_phy_powerdown_workaround_ich8lan(&adapter->hw);
6702 	} else if (hw->mac.type >= e1000_pch_lpt) {
6703 		if (wufc && !(wufc & (E1000_WUFC_EX | E1000_WUFC_MC | E1000_WUFC_BC))) {
6704 			/* ULP does not support wake from unicast, multicast
6705 			 * or broadcast.
6706 			 */
6707 			retval = e1000_enable_ulp_lpt_lp(hw, !runtime);
6708 			if (retval) {
6709 				e_err("Failed to enable ULP\n");
6710 				goto skip_phy_configurations;
6711 			}
6712 		}
6713 	}
6714 
6715 	/* Ensure that the appropriate bits are set in LPI_CTRL
6716 	 * for EEE in Sx
6717 	 */
6718 	if ((hw->phy.type >= e1000_phy_i217) &&
6719 	    adapter->eee_advert && hw->dev_spec.ich8lan.eee_lp_ability) {
6720 		u16 lpi_ctrl = 0;
6721 
6722 		retval = hw->phy.ops.acquire(hw);
6723 		if (!retval) {
6724 			retval = e1e_rphy_locked(hw, I82579_LPI_CTRL,
6725 						 &lpi_ctrl);
6726 			if (!retval) {
6727 				if (adapter->eee_advert &
6728 				    hw->dev_spec.ich8lan.eee_lp_ability &
6729 				    I82579_EEE_100_SUPPORTED)
6730 					lpi_ctrl |= I82579_LPI_CTRL_100_ENABLE;
6731 				if (adapter->eee_advert &
6732 				    hw->dev_spec.ich8lan.eee_lp_ability &
6733 				    I82579_EEE_1000_SUPPORTED)
6734 					lpi_ctrl |= I82579_LPI_CTRL_1000_ENABLE;
6735 
6736 				retval = e1e_wphy_locked(hw, I82579_LPI_CTRL,
6737 							 lpi_ctrl);
6738 			}
6739 		}
6740 		hw->phy.ops.release(hw);
6741 	}
6742 
6743 skip_phy_configurations:
6744 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
6745 	 * would have already happened in close and is redundant.
6746 	 */
6747 	e1000e_release_hw_control(adapter);
6748 
6749 	pci_clear_master(pdev);
6750 
6751 	/* The pci-e switch on some quad port adapters will report a
6752 	 * correctable error when the MAC transitions from D0 to D3.  To
6753 	 * prevent this we need to mask off the correctable errors on the
6754 	 * downstream port of the pci-e switch.
6755 	 *
6756 	 * We don't have the associated upstream bridge while assigning
6757 	 * the PCI device into guest. For example, the KVM on power is
6758 	 * one of the cases.
6759 	 */
6760 	if (adapter->flags & FLAG_IS_QUAD_PORT) {
6761 		struct pci_dev *us_dev = pdev->bus->self;
6762 		u16 devctl;
6763 
6764 		if (!us_dev)
6765 			return 0;
6766 
6767 		pcie_capability_read_word(us_dev, PCI_EXP_DEVCTL, &devctl);
6768 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL,
6769 					   (devctl & ~PCI_EXP_DEVCTL_CERE));
6770 
6771 		pci_save_state(pdev);
6772 		pci_prepare_to_sleep(pdev);
6773 
6774 		pcie_capability_write_word(us_dev, PCI_EXP_DEVCTL, devctl);
6775 	}
6776 
6777 	return 0;
6778 }
6779 
6780 /**
6781  * __e1000e_disable_aspm - Disable ASPM states
6782  * @pdev: pointer to PCI device struct
6783  * @state: bit-mask of ASPM states to disable
6784  * @locked: indication if this context holds pci_bus_sem locked.
6785  *
6786  * Some devices *must* have certain ASPM states disabled per hardware errata.
6787  **/
__e1000e_disable_aspm(struct pci_dev * pdev,u16 state,int locked)6788 static void __e1000e_disable_aspm(struct pci_dev *pdev, u16 state, int locked)
6789 {
6790 	struct pci_dev *parent = pdev->bus->self;
6791 	u16 aspm_dis_mask = 0;
6792 	u16 pdev_aspmc, parent_aspmc;
6793 
6794 	switch (state) {
6795 	case PCIE_LINK_STATE_L0S:
6796 	case PCIE_LINK_STATE_L0S | PCIE_LINK_STATE_L1:
6797 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L0S;
6798 		fallthrough; /* can't have L1 without L0s */
6799 	case PCIE_LINK_STATE_L1:
6800 		aspm_dis_mask |= PCI_EXP_LNKCTL_ASPM_L1;
6801 		break;
6802 	default:
6803 		return;
6804 	}
6805 
6806 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6807 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6808 
6809 	if (parent) {
6810 		pcie_capability_read_word(parent, PCI_EXP_LNKCTL,
6811 					  &parent_aspmc);
6812 		parent_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6813 	}
6814 
6815 	/* Nothing to do if the ASPM states to be disabled already are */
6816 	if (!(pdev_aspmc & aspm_dis_mask) &&
6817 	    (!parent || !(parent_aspmc & aspm_dis_mask)))
6818 		return;
6819 
6820 	dev_info(&pdev->dev, "Disabling ASPM %s %s\n",
6821 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L0S) ?
6822 		 "L0s" : "",
6823 		 (aspm_dis_mask & pdev_aspmc & PCI_EXP_LNKCTL_ASPM_L1) ?
6824 		 "L1" : "");
6825 
6826 #ifdef CONFIG_PCIEASPM
6827 	if (locked)
6828 		pci_disable_link_state_locked(pdev, state);
6829 	else
6830 		pci_disable_link_state(pdev, state);
6831 
6832 	/* Double-check ASPM control.  If not disabled by the above, the
6833 	 * BIOS is preventing that from happening (or CONFIG_PCIEASPM is
6834 	 * not enabled); override by writing PCI config space directly.
6835 	 */
6836 	pcie_capability_read_word(pdev, PCI_EXP_LNKCTL, &pdev_aspmc);
6837 	pdev_aspmc &= PCI_EXP_LNKCTL_ASPMC;
6838 
6839 	if (!(aspm_dis_mask & pdev_aspmc))
6840 		return;
6841 #endif
6842 
6843 	/* Both device and parent should have the same ASPM setting.
6844 	 * Disable ASPM in downstream component first and then upstream.
6845 	 */
6846 	pcie_capability_clear_word(pdev, PCI_EXP_LNKCTL, aspm_dis_mask);
6847 
6848 	if (parent)
6849 		pcie_capability_clear_word(parent, PCI_EXP_LNKCTL,
6850 					   aspm_dis_mask);
6851 }
6852 
6853 /**
6854  * e1000e_disable_aspm - Disable ASPM states.
6855  * @pdev: pointer to PCI device struct
6856  * @state: bit-mask of ASPM states to disable
6857  *
6858  * This function acquires the pci_bus_sem!
6859  * Some devices *must* have certain ASPM states disabled per hardware errata.
6860  **/
e1000e_disable_aspm(struct pci_dev * pdev,u16 state)6861 static void e1000e_disable_aspm(struct pci_dev *pdev, u16 state)
6862 {
6863 	__e1000e_disable_aspm(pdev, state, 0);
6864 }
6865 
6866 /**
6867  * e1000e_disable_aspm_locked - Disable ASPM states.
6868  * @pdev: pointer to PCI device struct
6869  * @state: bit-mask of ASPM states to disable
6870  *
6871  * This function must be called with pci_bus_sem acquired!
6872  * Some devices *must* have certain ASPM states disabled per hardware errata.
6873  **/
e1000e_disable_aspm_locked(struct pci_dev * pdev,u16 state)6874 static void e1000e_disable_aspm_locked(struct pci_dev *pdev, u16 state)
6875 {
6876 	__e1000e_disable_aspm(pdev, state, 1);
6877 }
6878 
e1000e_pm_thaw(struct device * dev)6879 static int e1000e_pm_thaw(struct device *dev)
6880 {
6881 	struct net_device *netdev = dev_get_drvdata(dev);
6882 	struct e1000_adapter *adapter = netdev_priv(netdev);
6883 	int rc = 0;
6884 
6885 	e1000e_set_interrupt_capability(adapter);
6886 
6887 	rtnl_lock();
6888 	if (netif_running(netdev)) {
6889 		rc = e1000_request_irq(adapter);
6890 		if (rc)
6891 			goto err_irq;
6892 
6893 		e1000e_up(adapter);
6894 	}
6895 
6896 	netif_device_attach(netdev);
6897 err_irq:
6898 	rtnl_unlock();
6899 
6900 	return rc;
6901 }
6902 
__e1000_resume(struct pci_dev * pdev)6903 static int __e1000_resume(struct pci_dev *pdev)
6904 {
6905 	struct net_device *netdev = pci_get_drvdata(pdev);
6906 	struct e1000_adapter *adapter = netdev_priv(netdev);
6907 	struct e1000_hw *hw = &adapter->hw;
6908 	u16 aspm_disable_flag = 0;
6909 
6910 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
6911 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
6912 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
6913 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
6914 	if (aspm_disable_flag)
6915 		e1000e_disable_aspm(pdev, aspm_disable_flag);
6916 
6917 	pci_set_master(pdev);
6918 
6919 	if (hw->mac.type >= e1000_pch2lan)
6920 		e1000_resume_workarounds_pchlan(&adapter->hw);
6921 
6922 	e1000e_power_up_phy(adapter);
6923 
6924 	/* report the system wakeup cause from S3/S4 */
6925 	if (adapter->flags2 & FLAG2_HAS_PHY_WAKEUP) {
6926 		u16 phy_data;
6927 
6928 		e1e_rphy(&adapter->hw, BM_WUS, &phy_data);
6929 		if (phy_data) {
6930 			e_info("PHY Wakeup cause - %s\n",
6931 			       phy_data & E1000_WUS_EX ? "Unicast Packet" :
6932 			       phy_data & E1000_WUS_MC ? "Multicast Packet" :
6933 			       phy_data & E1000_WUS_BC ? "Broadcast Packet" :
6934 			       phy_data & E1000_WUS_MAG ? "Magic Packet" :
6935 			       phy_data & E1000_WUS_LNKC ?
6936 			       "Link Status Change" : "other");
6937 		}
6938 		e1e_wphy(&adapter->hw, BM_WUS, ~0);
6939 	} else {
6940 		u32 wus = er32(WUS);
6941 
6942 		if (wus) {
6943 			e_info("MAC Wakeup cause - %s\n",
6944 			       wus & E1000_WUS_EX ? "Unicast Packet" :
6945 			       wus & E1000_WUS_MC ? "Multicast Packet" :
6946 			       wus & E1000_WUS_BC ? "Broadcast Packet" :
6947 			       wus & E1000_WUS_MAG ? "Magic Packet" :
6948 			       wus & E1000_WUS_LNKC ? "Link Status Change" :
6949 			       "other");
6950 		}
6951 		ew32(WUS, ~0);
6952 	}
6953 
6954 	e1000e_reset(adapter);
6955 
6956 	e1000_init_manageability_pt(adapter);
6957 
6958 	/* If the controller has AMT, do not set DRV_LOAD until the interface
6959 	 * is up.  For all other cases, let the f/w know that the h/w is now
6960 	 * under the control of the driver.
6961 	 */
6962 	if (!(adapter->flags & FLAG_HAS_AMT))
6963 		e1000e_get_hw_control(adapter);
6964 
6965 	return 0;
6966 }
6967 
e1000e_pm_prepare(struct device * dev)6968 static int e1000e_pm_prepare(struct device *dev)
6969 {
6970 	return pm_runtime_suspended(dev) &&
6971 		pm_suspend_via_firmware();
6972 }
6973 
e1000e_pm_suspend(struct device * dev)6974 static int e1000e_pm_suspend(struct device *dev)
6975 {
6976 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6977 	struct e1000_adapter *adapter = netdev_priv(netdev);
6978 	struct pci_dev *pdev = to_pci_dev(dev);
6979 	int rc;
6980 
6981 	e1000e_flush_lpic(pdev);
6982 
6983 	e1000e_pm_freeze(dev);
6984 
6985 	rc = __e1000_shutdown(pdev, false);
6986 	if (!rc) {
6987 		/* Introduce S0ix implementation */
6988 		if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
6989 			e1000e_s0ix_entry_flow(adapter);
6990 	}
6991 
6992 	return 0;
6993 }
6994 
e1000e_pm_resume(struct device * dev)6995 static int e1000e_pm_resume(struct device *dev)
6996 {
6997 	struct net_device *netdev = pci_get_drvdata(to_pci_dev(dev));
6998 	struct e1000_adapter *adapter = netdev_priv(netdev);
6999 	struct pci_dev *pdev = to_pci_dev(dev);
7000 	int rc;
7001 
7002 	/* Introduce S0ix implementation */
7003 	if (adapter->flags2 & FLAG2_ENABLE_S0IX_FLOWS)
7004 		e1000e_s0ix_exit_flow(adapter);
7005 
7006 	rc = __e1000_resume(pdev);
7007 	if (rc)
7008 		return rc;
7009 
7010 	return e1000e_pm_thaw(dev);
7011 }
7012 
e1000e_pm_runtime_idle(struct device * dev)7013 static __maybe_unused int e1000e_pm_runtime_idle(struct device *dev)
7014 {
7015 	struct net_device *netdev = dev_get_drvdata(dev);
7016 	struct e1000_adapter *adapter = netdev_priv(netdev);
7017 	u16 eee_lp;
7018 
7019 	eee_lp = adapter->hw.dev_spec.ich8lan.eee_lp_ability;
7020 
7021 	if (!e1000e_has_link(adapter)) {
7022 		adapter->hw.dev_spec.ich8lan.eee_lp_ability = eee_lp;
7023 		pm_schedule_suspend(dev, 5 * MSEC_PER_SEC);
7024 	}
7025 
7026 	return -EBUSY;
7027 }
7028 
e1000e_pm_runtime_resume(struct device * dev)7029 static int e1000e_pm_runtime_resume(struct device *dev)
7030 {
7031 	struct pci_dev *pdev = to_pci_dev(dev);
7032 	struct net_device *netdev = pci_get_drvdata(pdev);
7033 	struct e1000_adapter *adapter = netdev_priv(netdev);
7034 	int rc;
7035 
7036 	pdev->pme_poll = true;
7037 
7038 	rc = __e1000_resume(pdev);
7039 	if (rc)
7040 		return rc;
7041 
7042 	if (netdev->flags & IFF_UP)
7043 		e1000e_up(adapter);
7044 
7045 	return rc;
7046 }
7047 
e1000e_pm_runtime_suspend(struct device * dev)7048 static int e1000e_pm_runtime_suspend(struct device *dev)
7049 {
7050 	struct pci_dev *pdev = to_pci_dev(dev);
7051 	struct net_device *netdev = pci_get_drvdata(pdev);
7052 	struct e1000_adapter *adapter = netdev_priv(netdev);
7053 
7054 	if (netdev->flags & IFF_UP) {
7055 		int count = E1000_CHECK_RESET_COUNT;
7056 
7057 		while (test_bit(__E1000_RESETTING, &adapter->state) && count--)
7058 			usleep_range(10000, 11000);
7059 
7060 		WARN_ON(test_bit(__E1000_RESETTING, &adapter->state));
7061 
7062 		/* Down the device without resetting the hardware */
7063 		e1000e_down(adapter, false);
7064 	}
7065 
7066 	if (__e1000_shutdown(pdev, true)) {
7067 		e1000e_pm_runtime_resume(dev);
7068 		return -EBUSY;
7069 	}
7070 
7071 	return 0;
7072 }
7073 
e1000_shutdown(struct pci_dev * pdev)7074 static void e1000_shutdown(struct pci_dev *pdev)
7075 {
7076 	e1000e_flush_lpic(pdev);
7077 
7078 	e1000e_pm_freeze(&pdev->dev);
7079 
7080 	__e1000_shutdown(pdev, false);
7081 }
7082 
7083 #ifdef CONFIG_NET_POLL_CONTROLLER
7084 
e1000_intr_msix(int __always_unused irq,void * data)7085 static irqreturn_t e1000_intr_msix(int __always_unused irq, void *data)
7086 {
7087 	struct net_device *netdev = data;
7088 	struct e1000_adapter *adapter = netdev_priv(netdev);
7089 
7090 	if (adapter->msix_entries) {
7091 		int vector, msix_irq;
7092 
7093 		vector = 0;
7094 		msix_irq = adapter->msix_entries[vector].vector;
7095 		if (disable_hardirq(msix_irq))
7096 			e1000_intr_msix_rx(msix_irq, netdev);
7097 		enable_irq(msix_irq);
7098 
7099 		vector++;
7100 		msix_irq = adapter->msix_entries[vector].vector;
7101 		if (disable_hardirq(msix_irq))
7102 			e1000_intr_msix_tx(msix_irq, netdev);
7103 		enable_irq(msix_irq);
7104 
7105 		vector++;
7106 		msix_irq = adapter->msix_entries[vector].vector;
7107 		if (disable_hardirq(msix_irq))
7108 			e1000_msix_other(msix_irq, netdev);
7109 		enable_irq(msix_irq);
7110 	}
7111 
7112 	return IRQ_HANDLED;
7113 }
7114 
7115 /**
7116  * e1000_netpoll
7117  * @netdev: network interface device structure
7118  *
7119  * Polling 'interrupt' - used by things like netconsole to send skbs
7120  * without having to re-enable interrupts. It's not called while
7121  * the interrupt routine is executing.
7122  */
e1000_netpoll(struct net_device * netdev)7123 static void e1000_netpoll(struct net_device *netdev)
7124 {
7125 	struct e1000_adapter *adapter = netdev_priv(netdev);
7126 
7127 	switch (adapter->int_mode) {
7128 	case E1000E_INT_MODE_MSIX:
7129 		e1000_intr_msix(adapter->pdev->irq, netdev);
7130 		break;
7131 	case E1000E_INT_MODE_MSI:
7132 		if (disable_hardirq(adapter->pdev->irq))
7133 			e1000_intr_msi(adapter->pdev->irq, netdev);
7134 		enable_irq(adapter->pdev->irq);
7135 		break;
7136 	default:		/* E1000E_INT_MODE_LEGACY */
7137 		if (disable_hardirq(adapter->pdev->irq))
7138 			e1000_intr(adapter->pdev->irq, netdev);
7139 		enable_irq(adapter->pdev->irq);
7140 		break;
7141 	}
7142 }
7143 #endif
7144 
7145 /**
7146  * e1000_io_error_detected - called when PCI error is detected
7147  * @pdev: Pointer to PCI device
7148  * @state: The current pci connection state
7149  *
7150  * This function is called after a PCI bus error affecting
7151  * this device has been detected.
7152  */
e1000_io_error_detected(struct pci_dev * pdev,pci_channel_state_t state)7153 static pci_ers_result_t e1000_io_error_detected(struct pci_dev *pdev,
7154 						pci_channel_state_t state)
7155 {
7156 	e1000e_pm_freeze(&pdev->dev);
7157 
7158 	if (state == pci_channel_io_perm_failure)
7159 		return PCI_ERS_RESULT_DISCONNECT;
7160 
7161 	pci_disable_device(pdev);
7162 
7163 	/* Request a slot reset. */
7164 	return PCI_ERS_RESULT_NEED_RESET;
7165 }
7166 
7167 /**
7168  * e1000_io_slot_reset - called after the pci bus has been reset.
7169  * @pdev: Pointer to PCI device
7170  *
7171  * Restart the card from scratch, as if from a cold-boot. Implementation
7172  * resembles the first-half of the e1000e_pm_resume routine.
7173  */
e1000_io_slot_reset(struct pci_dev * pdev)7174 static pci_ers_result_t e1000_io_slot_reset(struct pci_dev *pdev)
7175 {
7176 	struct net_device *netdev = pci_get_drvdata(pdev);
7177 	struct e1000_adapter *adapter = netdev_priv(netdev);
7178 	struct e1000_hw *hw = &adapter->hw;
7179 	u16 aspm_disable_flag = 0;
7180 	int err;
7181 	pci_ers_result_t result;
7182 
7183 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L0S)
7184 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7185 	if (adapter->flags2 & FLAG2_DISABLE_ASPM_L1)
7186 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7187 	if (aspm_disable_flag)
7188 		e1000e_disable_aspm_locked(pdev, aspm_disable_flag);
7189 
7190 	err = pci_enable_device_mem(pdev);
7191 	if (err) {
7192 		dev_err(&pdev->dev,
7193 			"Cannot re-enable PCI device after reset.\n");
7194 		result = PCI_ERS_RESULT_DISCONNECT;
7195 	} else {
7196 		pci_restore_state(pdev);
7197 		pci_set_master(pdev);
7198 
7199 		pci_enable_wake(pdev, PCI_D3hot, 0);
7200 		pci_enable_wake(pdev, PCI_D3cold, 0);
7201 
7202 		e1000e_reset(adapter);
7203 		ew32(WUS, ~0);
7204 		result = PCI_ERS_RESULT_RECOVERED;
7205 	}
7206 
7207 	return result;
7208 }
7209 
7210 /**
7211  * e1000_io_resume - called when traffic can start flowing again.
7212  * @pdev: Pointer to PCI device
7213  *
7214  * This callback is called when the error recovery driver tells us that
7215  * its OK to resume normal operation. Implementation resembles the
7216  * second-half of the e1000e_pm_resume routine.
7217  */
e1000_io_resume(struct pci_dev * pdev)7218 static void e1000_io_resume(struct pci_dev *pdev)
7219 {
7220 	struct net_device *netdev = pci_get_drvdata(pdev);
7221 	struct e1000_adapter *adapter = netdev_priv(netdev);
7222 
7223 	e1000_init_manageability_pt(adapter);
7224 
7225 	e1000e_pm_thaw(&pdev->dev);
7226 
7227 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7228 	 * is up.  For all other cases, let the f/w know that the h/w is now
7229 	 * under the control of the driver.
7230 	 */
7231 	if (!(adapter->flags & FLAG_HAS_AMT))
7232 		e1000e_get_hw_control(adapter);
7233 }
7234 
e1000_print_device_info(struct e1000_adapter * adapter)7235 static void e1000_print_device_info(struct e1000_adapter *adapter)
7236 {
7237 	struct e1000_hw *hw = &adapter->hw;
7238 	struct net_device *netdev = adapter->netdev;
7239 	u32 ret_val;
7240 	u8 pba_str[E1000_PBANUM_LENGTH];
7241 
7242 	/* print bus type/speed/width info */
7243 	e_info("(PCI Express:2.5GT/s:%s) %pM\n",
7244 	       /* bus width */
7245 	       ((hw->bus.width == e1000_bus_width_pcie_x4) ? "Width x4" :
7246 		"Width x1"),
7247 	       /* MAC address */
7248 	       netdev->dev_addr);
7249 	e_info("Intel(R) PRO/%s Network Connection\n",
7250 	       (hw->phy.type == e1000_phy_ife) ? "10/100" : "1000");
7251 	ret_val = e1000_read_pba_string_generic(hw, pba_str,
7252 						E1000_PBANUM_LENGTH);
7253 	if (ret_val)
7254 		strscpy((char *)pba_str, "Unknown", sizeof(pba_str));
7255 	e_info("MAC: %d, PHY: %d, PBA No: %s\n",
7256 	       hw->mac.type, hw->phy.type, pba_str);
7257 }
7258 
e1000_eeprom_checks(struct e1000_adapter * adapter)7259 static void e1000_eeprom_checks(struct e1000_adapter *adapter)
7260 {
7261 	struct e1000_hw *hw = &adapter->hw;
7262 	int ret_val;
7263 	u16 buf = 0;
7264 
7265 	if (hw->mac.type != e1000_82573)
7266 		return;
7267 
7268 	ret_val = e1000_read_nvm(hw, NVM_INIT_CONTROL2_REG, 1, &buf);
7269 	le16_to_cpus(&buf);
7270 	if (!ret_val && (!(buf & BIT(0)))) {
7271 		/* Deep Smart Power Down (DSPD) */
7272 		dev_warn(&adapter->pdev->dev,
7273 			 "Warning: detected DSPD enabled in EEPROM\n");
7274 	}
7275 }
7276 
e1000_fix_features(struct net_device * netdev,netdev_features_t features)7277 static netdev_features_t e1000_fix_features(struct net_device *netdev,
7278 					    netdev_features_t features)
7279 {
7280 	struct e1000_adapter *adapter = netdev_priv(netdev);
7281 	struct e1000_hw *hw = &adapter->hw;
7282 
7283 	/* Jumbo frame workaround on 82579 and newer requires CRC be stripped */
7284 	if ((hw->mac.type >= e1000_pch2lan) && (netdev->mtu > ETH_DATA_LEN))
7285 		features &= ~NETIF_F_RXFCS;
7286 
7287 	/* Since there is no support for separate Rx/Tx vlan accel
7288 	 * enable/disable make sure Tx flag is always in same state as Rx.
7289 	 */
7290 	if (features & NETIF_F_HW_VLAN_CTAG_RX)
7291 		features |= NETIF_F_HW_VLAN_CTAG_TX;
7292 	else
7293 		features &= ~NETIF_F_HW_VLAN_CTAG_TX;
7294 
7295 	return features;
7296 }
7297 
e1000_set_features(struct net_device * netdev,netdev_features_t features)7298 static int e1000_set_features(struct net_device *netdev,
7299 			      netdev_features_t features)
7300 {
7301 	struct e1000_adapter *adapter = netdev_priv(netdev);
7302 	netdev_features_t changed = features ^ netdev->features;
7303 
7304 	if (changed & (NETIF_F_TSO | NETIF_F_TSO6))
7305 		adapter->flags |= FLAG_TSO_FORCE;
7306 
7307 	if (!(changed & (NETIF_F_HW_VLAN_CTAG_RX | NETIF_F_HW_VLAN_CTAG_TX |
7308 			 NETIF_F_RXCSUM | NETIF_F_RXHASH | NETIF_F_RXFCS |
7309 			 NETIF_F_RXALL)))
7310 		return 0;
7311 
7312 	if (changed & NETIF_F_RXFCS) {
7313 		if (features & NETIF_F_RXFCS) {
7314 			adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7315 		} else {
7316 			/* We need to take it back to defaults, which might mean
7317 			 * stripping is still disabled at the adapter level.
7318 			 */
7319 			if (adapter->flags2 & FLAG2_DFLT_CRC_STRIPPING)
7320 				adapter->flags2 |= FLAG2_CRC_STRIPPING;
7321 			else
7322 				adapter->flags2 &= ~FLAG2_CRC_STRIPPING;
7323 		}
7324 	}
7325 
7326 	netdev->features = features;
7327 
7328 	if (netif_running(netdev))
7329 		e1000e_reinit_locked(adapter);
7330 	else
7331 		e1000e_reset(adapter);
7332 
7333 	return 1;
7334 }
7335 
7336 static const struct net_device_ops e1000e_netdev_ops = {
7337 	.ndo_open		= e1000e_open,
7338 	.ndo_stop		= e1000e_close,
7339 	.ndo_start_xmit		= e1000_xmit_frame,
7340 	.ndo_get_stats64	= e1000e_get_stats64,
7341 	.ndo_set_rx_mode	= e1000e_set_rx_mode,
7342 	.ndo_set_mac_address	= e1000_set_mac,
7343 	.ndo_change_mtu		= e1000_change_mtu,
7344 	.ndo_eth_ioctl		= e1000_ioctl,
7345 	.ndo_tx_timeout		= e1000_tx_timeout,
7346 	.ndo_validate_addr	= eth_validate_addr,
7347 
7348 	.ndo_vlan_rx_add_vid	= e1000_vlan_rx_add_vid,
7349 	.ndo_vlan_rx_kill_vid	= e1000_vlan_rx_kill_vid,
7350 #ifdef CONFIG_NET_POLL_CONTROLLER
7351 	.ndo_poll_controller	= e1000_netpoll,
7352 #endif
7353 	.ndo_set_features	= e1000_set_features,
7354 	.ndo_fix_features	= e1000_fix_features,
7355 	.ndo_features_check	= passthru_features_check,
7356 	.ndo_hwtstamp_get	= e1000e_hwtstamp_get,
7357 	.ndo_hwtstamp_set	= e1000e_hwtstamp_set,
7358 };
7359 
7360 /**
7361  * e1000_probe - Device Initialization Routine
7362  * @pdev: PCI device information struct
7363  * @ent: entry in e1000_pci_tbl
7364  *
7365  * Returns 0 on success, negative on failure
7366  *
7367  * e1000_probe initializes an adapter identified by a pci_dev structure.
7368  * The OS initialization, configuring of the adapter private structure,
7369  * and a hardware reset occur.
7370  **/
e1000_probe(struct pci_dev * pdev,const struct pci_device_id * ent)7371 static int e1000_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
7372 {
7373 	struct net_device *netdev;
7374 	struct e1000_adapter *adapter;
7375 	struct e1000_hw *hw;
7376 	const struct e1000_info *ei = e1000_info_tbl[ent->driver_data];
7377 	resource_size_t mmio_start, mmio_len;
7378 	resource_size_t flash_start, flash_len;
7379 	static int cards_found;
7380 	u16 aspm_disable_flag = 0;
7381 	u16 eeprom_data = 0;
7382 	u16 eeprom_apme_mask = E1000_EEPROM_APME;
7383 	int bars, i, err;
7384 	s32 ret_val = 0;
7385 
7386 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L0S)
7387 		aspm_disable_flag = PCIE_LINK_STATE_L0S;
7388 	if (ei->flags2 & FLAG2_DISABLE_ASPM_L1)
7389 		aspm_disable_flag |= PCIE_LINK_STATE_L1;
7390 	if (aspm_disable_flag)
7391 		e1000e_disable_aspm(pdev, aspm_disable_flag);
7392 
7393 	err = pci_enable_device_mem(pdev);
7394 	if (err)
7395 		return err;
7396 
7397 	err = dma_set_mask_and_coherent(&pdev->dev, DMA_BIT_MASK(64));
7398 	if (err) {
7399 		dev_err(&pdev->dev,
7400 			"No usable DMA configuration, aborting\n");
7401 		goto err_dma;
7402 	}
7403 
7404 	bars = pci_select_bars(pdev, IORESOURCE_MEM);
7405 	err = pci_request_selected_regions_exclusive(pdev, bars,
7406 						     e1000e_driver_name);
7407 	if (err)
7408 		goto err_pci_reg;
7409 
7410 	pci_set_master(pdev);
7411 	/* PCI config space info */
7412 	err = pci_save_state(pdev);
7413 	if (err)
7414 		goto err_alloc_etherdev;
7415 
7416 	err = -ENOMEM;
7417 	netdev = alloc_etherdev(sizeof(struct e1000_adapter));
7418 	if (!netdev)
7419 		goto err_alloc_etherdev;
7420 
7421 	SET_NETDEV_DEV(netdev, &pdev->dev);
7422 
7423 	netdev->irq = pdev->irq;
7424 
7425 	pci_set_drvdata(pdev, netdev);
7426 	adapter = netdev_priv(netdev);
7427 	hw = &adapter->hw;
7428 	adapter->netdev = netdev;
7429 	adapter->pdev = pdev;
7430 	adapter->ei = ei;
7431 	adapter->pba = ei->pba;
7432 	adapter->flags = ei->flags;
7433 	adapter->flags2 = ei->flags2;
7434 	adapter->hw.adapter = adapter;
7435 	adapter->hw.mac.type = ei->mac;
7436 	adapter->max_hw_frame_size = ei->max_hw_frame_size;
7437 	adapter->msg_enable = netif_msg_init(debug, DEFAULT_MSG_ENABLE);
7438 
7439 	mmio_start = pci_resource_start(pdev, 0);
7440 	mmio_len = pci_resource_len(pdev, 0);
7441 
7442 	err = -EIO;
7443 	adapter->hw.hw_addr = ioremap(mmio_start, mmio_len);
7444 	if (!adapter->hw.hw_addr)
7445 		goto err_ioremap;
7446 
7447 	if ((adapter->flags & FLAG_HAS_FLASH) &&
7448 	    (pci_resource_flags(pdev, 1) & IORESOURCE_MEM) &&
7449 	    (hw->mac.type < e1000_pch_spt)) {
7450 		flash_start = pci_resource_start(pdev, 1);
7451 		flash_len = pci_resource_len(pdev, 1);
7452 		adapter->hw.flash_address = ioremap(flash_start, flash_len);
7453 		if (!adapter->hw.flash_address)
7454 			goto err_flashmap;
7455 	}
7456 
7457 	/* Set default EEE advertisement */
7458 	if (adapter->flags2 & FLAG2_HAS_EEE)
7459 		adapter->eee_advert = MDIO_EEE_100TX | MDIO_EEE_1000T;
7460 
7461 	/* construct the net_device struct */
7462 	netdev->netdev_ops = &e1000e_netdev_ops;
7463 	e1000e_set_ethtool_ops(netdev);
7464 	netdev->watchdog_timeo = 5 * HZ;
7465 	netif_napi_add(netdev, &adapter->napi, e1000e_poll);
7466 	strscpy(netdev->name, pci_name(pdev), sizeof(netdev->name));
7467 
7468 	netdev->mem_start = mmio_start;
7469 	netdev->mem_end = mmio_start + mmio_len;
7470 
7471 	adapter->bd_number = cards_found++;
7472 
7473 	e1000e_check_options(adapter);
7474 
7475 	/* setup adapter struct */
7476 	err = e1000_sw_init(adapter);
7477 	if (err)
7478 		goto err_sw_init;
7479 
7480 	memcpy(&hw->mac.ops, ei->mac_ops, sizeof(hw->mac.ops));
7481 	memcpy(&hw->nvm.ops, ei->nvm_ops, sizeof(hw->nvm.ops));
7482 	memcpy(&hw->phy.ops, ei->phy_ops, sizeof(hw->phy.ops));
7483 
7484 	err = ei->get_variants(adapter);
7485 	if (err)
7486 		goto err_hw_init;
7487 
7488 	if ((adapter->flags & FLAG_IS_ICH) &&
7489 	    (adapter->flags & FLAG_READ_ONLY_NVM) &&
7490 	    (hw->mac.type < e1000_pch_spt))
7491 		e1000e_write_protect_nvm_ich8lan(&adapter->hw);
7492 
7493 	hw->mac.ops.get_bus_info(&adapter->hw);
7494 
7495 	adapter->hw.phy.autoneg_wait_to_complete = 0;
7496 
7497 	/* Copper options */
7498 	if (adapter->hw.phy.media_type == e1000_media_type_copper) {
7499 		adapter->hw.phy.mdix = AUTO_ALL_MODES;
7500 		adapter->hw.phy.disable_polarity_correction = 0;
7501 		adapter->hw.phy.ms_type = e1000_ms_hw_default;
7502 	}
7503 
7504 	if (hw->phy.ops.check_reset_block && hw->phy.ops.check_reset_block(hw))
7505 		dev_info(&pdev->dev,
7506 			 "PHY reset is blocked due to SOL/IDER session.\n");
7507 
7508 	/* Set initial default active device features */
7509 	netdev->features = (NETIF_F_SG |
7510 			    NETIF_F_HW_VLAN_CTAG_RX |
7511 			    NETIF_F_HW_VLAN_CTAG_TX |
7512 			    NETIF_F_TSO |
7513 			    NETIF_F_TSO6 |
7514 			    NETIF_F_RXHASH |
7515 			    NETIF_F_RXCSUM |
7516 			    NETIF_F_HW_CSUM);
7517 
7518 	/* disable TSO for pcie and 10/100 speeds to avoid
7519 	 * some hardware issues and for i219 to fix transfer
7520 	 * speed being capped at 60%
7521 	 */
7522 	if (!(adapter->flags & FLAG_TSO_FORCE)) {
7523 		switch (adapter->link_speed) {
7524 		case SPEED_10:
7525 		case SPEED_100:
7526 			e_info("10/100 speed: disabling TSO\n");
7527 			netdev->features &= ~NETIF_F_TSO;
7528 			netdev->features &= ~NETIF_F_TSO6;
7529 			break;
7530 		case SPEED_1000:
7531 			netdev->features |= NETIF_F_TSO;
7532 			netdev->features |= NETIF_F_TSO6;
7533 			break;
7534 		default:
7535 			/* oops */
7536 			break;
7537 		}
7538 		if (hw->mac.type == e1000_pch_spt) {
7539 			netdev->features &= ~NETIF_F_TSO;
7540 			netdev->features &= ~NETIF_F_TSO6;
7541 		}
7542 	}
7543 
7544 	/* Set user-changeable features (subset of all device features) */
7545 	netdev->hw_features = netdev->features;
7546 	netdev->hw_features |= NETIF_F_RXFCS;
7547 	netdev->priv_flags |= IFF_SUPP_NOFCS;
7548 	netdev->hw_features |= NETIF_F_RXALL;
7549 
7550 	if (adapter->flags & FLAG_HAS_HW_VLAN_FILTER)
7551 		netdev->features |= NETIF_F_HW_VLAN_CTAG_FILTER;
7552 
7553 	netdev->vlan_features |= (NETIF_F_SG |
7554 				  NETIF_F_TSO |
7555 				  NETIF_F_TSO6 |
7556 				  NETIF_F_HW_CSUM);
7557 
7558 	netdev->priv_flags |= IFF_UNICAST_FLT;
7559 
7560 	netdev->features |= NETIF_F_HIGHDMA;
7561 	netdev->vlan_features |= NETIF_F_HIGHDMA;
7562 
7563 	/* MTU range: 68 - max_hw_frame_size */
7564 	netdev->min_mtu = ETH_MIN_MTU;
7565 	netdev->max_mtu = adapter->max_hw_frame_size -
7566 			  (VLAN_ETH_HLEN + ETH_FCS_LEN);
7567 
7568 	if (e1000e_enable_mng_pass_thru(&adapter->hw))
7569 		adapter->flags |= FLAG_MNG_PT_ENABLED;
7570 
7571 	/* before reading the NVM, reset the controller to
7572 	 * put the device in a known good starting state
7573 	 */
7574 	adapter->hw.mac.ops.reset_hw(&adapter->hw);
7575 
7576 	/* systems with ASPM and others may see the checksum fail on the first
7577 	 * attempt. Let's give it a few tries
7578 	 */
7579 	for (i = 0;; i++) {
7580 		if (e1000_validate_nvm_checksum(&adapter->hw) >= 0)
7581 			break;
7582 		if (i == 2) {
7583 			dev_err(&pdev->dev, "The NVM Checksum Is Not Valid\n");
7584 			err = -EIO;
7585 			goto err_eeprom;
7586 		}
7587 	}
7588 
7589 	e1000_eeprom_checks(adapter);
7590 
7591 	/* copy the MAC address */
7592 	if (e1000e_read_mac_addr(&adapter->hw))
7593 		dev_err(&pdev->dev,
7594 			"NVM Read Error while reading MAC address\n");
7595 
7596 	eth_hw_addr_set(netdev, adapter->hw.mac.addr);
7597 
7598 	if (!is_valid_ether_addr(netdev->dev_addr)) {
7599 		dev_err(&pdev->dev, "Invalid MAC Address: %pM\n",
7600 			netdev->dev_addr);
7601 		err = -EIO;
7602 		goto err_eeprom;
7603 	}
7604 
7605 	timer_setup(&adapter->watchdog_timer, e1000_watchdog, 0);
7606 	timer_setup(&adapter->phy_info_timer, e1000_update_phy_info, 0);
7607 
7608 	INIT_WORK(&adapter->reset_task, e1000_reset_task);
7609 	INIT_WORK(&adapter->watchdog_task, e1000_watchdog_task);
7610 	INIT_WORK(&adapter->downshift_task, e1000e_downshift_workaround);
7611 	INIT_WORK(&adapter->update_phy_task, e1000e_update_phy_task);
7612 	INIT_WORK(&adapter->print_hang_task, e1000_print_hw_hang);
7613 
7614 	/* Initialize link parameters. User can change them with ethtool */
7615 	adapter->hw.mac.autoneg = 1;
7616 	adapter->fc_autoneg = true;
7617 	adapter->hw.fc.requested_mode = e1000_fc_default;
7618 	adapter->hw.fc.current_mode = e1000_fc_default;
7619 	adapter->hw.phy.autoneg_advertised = 0x2f;
7620 
7621 	/* Initial Wake on LAN setting - If APM wake is enabled in
7622 	 * the EEPROM, enable the ACPI Magic Packet filter
7623 	 */
7624 	if (adapter->flags & FLAG_APME_IN_WUC) {
7625 		/* APME bit in EEPROM is mapped to WUC.APME */
7626 		eeprom_data = er32(WUC);
7627 		eeprom_apme_mask = E1000_WUC_APME;
7628 		if ((hw->mac.type > e1000_ich10lan) &&
7629 		    (eeprom_data & E1000_WUC_PHY_WAKE))
7630 			adapter->flags2 |= FLAG2_HAS_PHY_WAKEUP;
7631 	} else if (adapter->flags & FLAG_APME_IN_CTRL3) {
7632 		if (adapter->flags & FLAG_APME_CHECK_PORT_B &&
7633 		    (adapter->hw.bus.func == 1))
7634 			ret_val = e1000_read_nvm(&adapter->hw,
7635 					      NVM_INIT_CONTROL3_PORT_B,
7636 					      1, &eeprom_data);
7637 		else
7638 			ret_val = e1000_read_nvm(&adapter->hw,
7639 					      NVM_INIT_CONTROL3_PORT_A,
7640 					      1, &eeprom_data);
7641 	}
7642 
7643 	/* fetch WoL from EEPROM */
7644 	if (ret_val)
7645 		e_dbg("NVM read error getting WoL initial values: %d\n", ret_val);
7646 	else if (eeprom_data & eeprom_apme_mask)
7647 		adapter->eeprom_wol |= E1000_WUFC_MAG;
7648 
7649 	/* now that we have the eeprom settings, apply the special cases
7650 	 * where the eeprom may be wrong or the board simply won't support
7651 	 * wake on lan on a particular port
7652 	 */
7653 	if (!(adapter->flags & FLAG_HAS_WOL))
7654 		adapter->eeprom_wol = 0;
7655 
7656 	/* initialize the wol settings based on the eeprom settings */
7657 	adapter->wol = adapter->eeprom_wol;
7658 
7659 	/* make sure adapter isn't asleep if manageability is enabled */
7660 	if (adapter->wol || (adapter->flags & FLAG_MNG_PT_ENABLED) ||
7661 	    (hw->mac.ops.check_mng_mode(hw)))
7662 		device_wakeup_enable(&pdev->dev);
7663 
7664 	/* save off EEPROM version number */
7665 	ret_val = e1000_read_nvm(&adapter->hw, 5, 1, &adapter->eeprom_vers);
7666 
7667 	if (ret_val) {
7668 		e_dbg("NVM read error getting EEPROM version: %d\n", ret_val);
7669 		adapter->eeprom_vers = 0;
7670 	}
7671 
7672 	/* init PTP hardware clock */
7673 	e1000e_ptp_init(adapter);
7674 
7675 	if (hw->mac.type >= e1000_pch_mtp)
7676 		adapter->flags2 |= FLAG2_DISABLE_K1;
7677 
7678 	/* reset the hardware with the new settings */
7679 	e1000e_reset(adapter);
7680 
7681 	/* If the controller has AMT, do not set DRV_LOAD until the interface
7682 	 * is up.  For all other cases, let the f/w know that the h/w is now
7683 	 * under the control of the driver.
7684 	 */
7685 	if (!(adapter->flags & FLAG_HAS_AMT))
7686 		e1000e_get_hw_control(adapter);
7687 
7688 	if (hw->mac.type >= e1000_pch_cnp)
7689 		adapter->flags2 |= FLAG2_ENABLE_S0IX_FLOWS;
7690 
7691 	strscpy(netdev->name, "eth%d", sizeof(netdev->name));
7692 	err = register_netdev(netdev);
7693 	if (err)
7694 		goto err_register;
7695 
7696 	/* carrier off reporting is important to ethtool even BEFORE open */
7697 	netif_carrier_off(netdev);
7698 
7699 	e1000_print_device_info(adapter);
7700 
7701 	dev_pm_set_driver_flags(&pdev->dev, DPM_FLAG_SMART_PREPARE);
7702 
7703 	if (pci_dev_run_wake(pdev))
7704 		pm_runtime_put_noidle(&pdev->dev);
7705 
7706 	return 0;
7707 
7708 err_register:
7709 	if (!(adapter->flags & FLAG_HAS_AMT))
7710 		e1000e_release_hw_control(adapter);
7711 err_eeprom:
7712 	if (hw->phy.ops.check_reset_block && !hw->phy.ops.check_reset_block(hw))
7713 		e1000_phy_hw_reset(&adapter->hw);
7714 err_hw_init:
7715 	kfree(adapter->tx_ring);
7716 	kfree(adapter->rx_ring);
7717 err_sw_init:
7718 	if ((adapter->hw.flash_address) && (hw->mac.type < e1000_pch_spt))
7719 		iounmap(adapter->hw.flash_address);
7720 	e1000e_reset_interrupt_capability(adapter);
7721 err_flashmap:
7722 	iounmap(adapter->hw.hw_addr);
7723 err_ioremap:
7724 	free_netdev(netdev);
7725 err_alloc_etherdev:
7726 	pci_release_mem_regions(pdev);
7727 err_pci_reg:
7728 err_dma:
7729 	pci_disable_device(pdev);
7730 	return err;
7731 }
7732 
7733 /**
7734  * e1000_remove - Device Removal Routine
7735  * @pdev: PCI device information struct
7736  *
7737  * e1000_remove is called by the PCI subsystem to alert the driver
7738  * that it should release a PCI device.  This could be caused by a
7739  * Hot-Plug event, or because the driver is going to be removed from
7740  * memory.
7741  **/
e1000_remove(struct pci_dev * pdev)7742 static void e1000_remove(struct pci_dev *pdev)
7743 {
7744 	struct net_device *netdev = pci_get_drvdata(pdev);
7745 	struct e1000_adapter *adapter = netdev_priv(netdev);
7746 
7747 	e1000e_ptp_remove(adapter);
7748 
7749 	/* The timers may be rescheduled, so explicitly disable them
7750 	 * from being rescheduled.
7751 	 */
7752 	set_bit(__E1000_DOWN, &adapter->state);
7753 	timer_delete_sync(&adapter->watchdog_timer);
7754 	timer_delete_sync(&adapter->phy_info_timer);
7755 
7756 	cancel_work_sync(&adapter->reset_task);
7757 	cancel_work_sync(&adapter->watchdog_task);
7758 	cancel_work_sync(&adapter->downshift_task);
7759 	cancel_work_sync(&adapter->update_phy_task);
7760 	cancel_work_sync(&adapter->print_hang_task);
7761 
7762 	if (adapter->flags & FLAG_HAS_HW_TIMESTAMP) {
7763 		cancel_work_sync(&adapter->tx_hwtstamp_work);
7764 		if (adapter->tx_hwtstamp_skb) {
7765 			dev_consume_skb_any(adapter->tx_hwtstamp_skb);
7766 			adapter->tx_hwtstamp_skb = NULL;
7767 		}
7768 	}
7769 
7770 	unregister_netdev(netdev);
7771 
7772 	if (pci_dev_run_wake(pdev))
7773 		pm_runtime_get_noresume(&pdev->dev);
7774 
7775 	/* Release control of h/w to f/w.  If f/w is AMT enabled, this
7776 	 * would have already happened in close and is redundant.
7777 	 */
7778 	e1000e_release_hw_control(adapter);
7779 
7780 	e1000e_reset_interrupt_capability(adapter);
7781 	kfree(adapter->tx_ring);
7782 	kfree(adapter->rx_ring);
7783 
7784 	iounmap(adapter->hw.hw_addr);
7785 	if ((adapter->hw.flash_address) &&
7786 	    (adapter->hw.mac.type < e1000_pch_spt))
7787 		iounmap(adapter->hw.flash_address);
7788 	pci_release_mem_regions(pdev);
7789 
7790 	free_netdev(netdev);
7791 
7792 	pci_disable_device(pdev);
7793 }
7794 
7795 /* PCI Error Recovery (ERS) */
7796 static const struct pci_error_handlers e1000_err_handler = {
7797 	.error_detected = e1000_io_error_detected,
7798 	.slot_reset = e1000_io_slot_reset,
7799 	.resume = e1000_io_resume,
7800 };
7801 
7802 static const struct pci_device_id e1000_pci_tbl[] = {
7803 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_COPPER), board_82571 },
7804 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_FIBER), board_82571 },
7805 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER), board_82571 },
7806 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_COPPER_LP),
7807 	  board_82571 },
7808 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_QUAD_FIBER), board_82571 },
7809 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES), board_82571 },
7810 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_DUAL), board_82571 },
7811 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571EB_SERDES_QUAD), board_82571 },
7812 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82571PT_QUAD_COPPER), board_82571 },
7813 
7814 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI), board_82572 },
7815 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_COPPER), board_82572 },
7816 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_FIBER), board_82572 },
7817 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82572EI_SERDES), board_82572 },
7818 
7819 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E), board_82573 },
7820 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573E_IAMT), board_82573 },
7821 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82573L), board_82573 },
7822 
7823 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574L), board_82574 },
7824 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82574LA), board_82574 },
7825 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_82583V), board_82583 },
7826 
7827 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_DPT),
7828 	  board_80003es2lan },
7829 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_COPPER_SPT),
7830 	  board_80003es2lan },
7831 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_DPT),
7832 	  board_80003es2lan },
7833 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_80003ES2LAN_SERDES_SPT),
7834 	  board_80003es2lan },
7835 
7836 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE), board_ich8lan },
7837 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_G), board_ich8lan },
7838 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IFE_GT), board_ich8lan },
7839 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_AMT), board_ich8lan },
7840 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_C), board_ich8lan },
7841 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M), board_ich8lan },
7842 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_IGP_M_AMT), board_ich8lan },
7843 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH8_82567V_3), board_ich8lan },
7844 
7845 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE), board_ich9lan },
7846 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_G), board_ich9lan },
7847 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IFE_GT), board_ich9lan },
7848 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_AMT), board_ich9lan },
7849 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_C), board_ich9lan },
7850 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_BM), board_ich9lan },
7851 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M), board_ich9lan },
7852 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_AMT), board_ich9lan },
7853 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH9_IGP_M_V), board_ich9lan },
7854 
7855 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LM), board_ich9lan },
7856 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_LF), board_ich9lan },
7857 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_R_BM_V), board_ich9lan },
7858 
7859 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LM), board_ich10lan },
7860 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_LF), board_ich10lan },
7861 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_ICH10_D_BM_V), board_ich10lan },
7862 
7863 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LM), board_pchlan },
7864 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_M_HV_LC), board_pchlan },
7865 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DM), board_pchlan },
7866 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_D_HV_DC), board_pchlan },
7867 
7868 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_LM), board_pch2lan },
7869 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH2_LV_V), board_pch2lan },
7870 
7871 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_LM), board_pch_lpt },
7872 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPT_I217_V), board_pch_lpt },
7873 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_LM), board_pch_lpt },
7874 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LPTLP_I218_V), board_pch_lpt },
7875 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM2), board_pch_lpt },
7876 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V2), board_pch_lpt },
7877 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_LM3), board_pch_lpt },
7878 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_I218_V3), board_pch_lpt },
7879 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM), board_pch_spt },
7880 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V), board_pch_spt },
7881 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM2), board_pch_spt },
7882 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V2), board_pch_spt },
7883 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LBG_I219_LM3), board_pch_spt },
7884 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM4), board_pch_spt },
7885 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V4), board_pch_spt },
7886 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_LM5), board_pch_spt },
7887 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_SPT_I219_V5), board_pch_spt },
7888 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM6), board_pch_cnp },
7889 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V6), board_pch_cnp },
7890 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_LM7), board_pch_cnp },
7891 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CNP_I219_V7), board_pch_cnp },
7892 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM8), board_pch_cnp },
7893 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V8), board_pch_cnp },
7894 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_LM9), board_pch_cnp },
7895 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ICP_I219_V9), board_pch_cnp },
7896 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM10), board_pch_cnp },
7897 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V10), board_pch_cnp },
7898 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM11), board_pch_cnp },
7899 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V11), board_pch_cnp },
7900 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_LM12), board_pch_spt },
7901 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_CMP_I219_V12), board_pch_spt },
7902 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM13), board_pch_tgp },
7903 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V13), board_pch_tgp },
7904 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM14), board_pch_tgp },
7905 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V14), board_pch_tgp },
7906 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_LM15), board_pch_tgp },
7907 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_TGP_I219_V15), board_pch_tgp },
7908 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM23), board_pch_adp },
7909 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V23), board_pch_adp },
7910 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM16), board_pch_adp },
7911 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V16), board_pch_adp },
7912 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM17), board_pch_adp },
7913 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V17), board_pch_adp },
7914 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_LM22), board_pch_adp },
7915 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_RPL_I219_V22), board_pch_adp },
7916 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_LM19), board_pch_adp },
7917 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ADP_I219_V19), board_pch_adp },
7918 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_LM18), board_pch_mtp },
7919 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_MTP_I219_V18), board_pch_mtp },
7920 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM20), board_pch_mtp },
7921 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V20), board_pch_mtp },
7922 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_LM21), board_pch_mtp },
7923 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_LNP_I219_V21), board_pch_mtp },
7924 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_LM24), board_pch_mtp },
7925 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_ARL_I219_V24), board_pch_mtp },
7926 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM25), board_pch_ptp },
7927 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V25), board_pch_ptp },
7928 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_LM27), board_pch_ptp },
7929 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_PTP_I219_V27), board_pch_ptp },
7930 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_LM29), board_pch_ptp },
7931 	{ PCI_VDEVICE(INTEL, E1000_DEV_ID_PCH_NVL_I219_V29), board_pch_ptp },
7932 
7933 	{ 0, 0, 0, 0, 0, 0, 0 }	/* terminate list */
7934 };
7935 MODULE_DEVICE_TABLE(pci, e1000_pci_tbl);
7936 
7937 static const struct dev_pm_ops e1000e_pm_ops = {
7938 	.prepare	= e1000e_pm_prepare,
7939 	.suspend	= e1000e_pm_suspend,
7940 	.resume		= e1000e_pm_resume,
7941 	.freeze		= e1000e_pm_freeze,
7942 	.thaw		= e1000e_pm_thaw,
7943 	.poweroff	= e1000e_pm_suspend,
7944 	.restore	= e1000e_pm_resume,
7945 	RUNTIME_PM_OPS(e1000e_pm_runtime_suspend, e1000e_pm_runtime_resume,
7946 		       e1000e_pm_runtime_idle)
7947 };
7948 
7949 /* PCI Device API Driver */
7950 static struct pci_driver e1000_driver = {
7951 	.name     = e1000e_driver_name,
7952 	.id_table = e1000_pci_tbl,
7953 	.probe    = e1000_probe,
7954 	.remove   = e1000_remove,
7955 	.driver.pm = pm_ptr(&e1000e_pm_ops),
7956 	.shutdown = e1000_shutdown,
7957 	.err_handler = &e1000_err_handler
7958 };
7959 
7960 /**
7961  * e1000_init_module - Driver Registration Routine
7962  *
7963  * e1000_init_module is the first routine called when the driver is
7964  * loaded. All it does is register with the PCI subsystem.
7965  **/
e1000_init_module(void)7966 static int __init e1000_init_module(void)
7967 {
7968 	pr_info("Intel(R) PRO/1000 Network Driver\n");
7969 	pr_info("Copyright(c) 1999 - 2015 Intel Corporation.\n");
7970 
7971 	return pci_register_driver(&e1000_driver);
7972 }
7973 module_init(e1000_init_module);
7974 
7975 /**
7976  * e1000_exit_module - Driver Exit Cleanup Routine
7977  *
7978  * e1000_exit_module is called just before the driver is removed
7979  * from memory.
7980  **/
e1000_exit_module(void)7981 static void __exit e1000_exit_module(void)
7982 {
7983 	pci_unregister_driver(&e1000_driver);
7984 }
7985 module_exit(e1000_exit_module);
7986 
7987 MODULE_DESCRIPTION("Intel(R) PRO/1000 Network Driver");
7988 MODULE_LICENSE("GPL v2");
7989 
7990 /* netdev.c */
7991