xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 86919f9dd2dbb88bbbe4e381dc8e6b9e9a1b9eb5)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/if_arp.h>
63 #include <linux/in.h>
64 #include <linux/ip.h>
65 #include <linux/ipv6.h>
66 #include <linux/irq.h>
67 #include <linux/kthread.h>
68 #include <linux/seq_file.h>
69 #include <linux/interrupt.h>
70 #include <net/net_namespace.h>
71 #include <asm/hvcall.h>
72 #include <linux/atomic.h>
73 #include <asm/vio.h>
74 #include <asm/iommu.h>
75 #include <linux/uaccess.h>
76 #include <asm/firmware.h>
77 #include <linux/workqueue.h>
78 #include <linux/if_vlan.h>
79 #include <linux/utsname.h>
80 
81 #include "ibmvnic.h"
82 
83 static const char ibmvnic_driver_name[] = "ibmvnic";
84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
85 
86 MODULE_AUTHOR("Santiago Leon");
87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
88 MODULE_LICENSE("GPL");
89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
90 
91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
92 static int ibmvnic_remove(struct vio_dev *);
93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
99 		       union sub_crq *sub_crq);
100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
102 static int enable_scrq_irq(struct ibmvnic_adapter *,
103 			   struct ibmvnic_sub_crq_queue *);
104 static int disable_scrq_irq(struct ibmvnic_adapter *,
105 			    struct ibmvnic_sub_crq_queue *);
106 static int pending_scrq(struct ibmvnic_adapter *,
107 			struct ibmvnic_sub_crq_queue *);
108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
109 					struct ibmvnic_sub_crq_queue *);
110 static int ibmvnic_poll(struct napi_struct *napi, int data);
111 static void send_map_query(struct ibmvnic_adapter *adapter);
112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
114 static int send_login(struct ibmvnic_adapter *adapter);
115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
116 static int init_sub_crqs(struct ibmvnic_adapter *);
117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
118 static int ibmvnic_init(struct ibmvnic_adapter *);
119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
120 static void release_crq_queue(struct ibmvnic_adapter *);
121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
123 
124 struct ibmvnic_stat {
125 	char name[ETH_GSTRING_LEN];
126 	int offset;
127 };
128 
129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
130 			     offsetof(struct ibmvnic_statistics, stat))
131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
132 
133 static const struct ibmvnic_stat ibmvnic_stats[] = {
134 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
135 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
136 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
137 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
138 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
139 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
140 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
141 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
142 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
143 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
144 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
145 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
146 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
147 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
148 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
149 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
150 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
151 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
152 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
153 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
154 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
155 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
156 };
157 
158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
159 			  unsigned long length, unsigned long *number,
160 			  unsigned long *irq)
161 {
162 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
163 	long rc;
164 
165 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
166 	*number = retbuf[0];
167 	*irq = retbuf[1];
168 
169 	return rc;
170 }
171 
172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
173 				struct ibmvnic_long_term_buff *ltb, int size)
174 {
175 	struct device *dev = &adapter->vdev->dev;
176 	int rc;
177 
178 	ltb->size = size;
179 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
180 				       GFP_KERNEL);
181 
182 	if (!ltb->buff) {
183 		dev_err(dev, "Couldn't alloc long term buffer\n");
184 		return -ENOMEM;
185 	}
186 	ltb->map_id = adapter->map_id;
187 	adapter->map_id++;
188 
189 	init_completion(&adapter->fw_done);
190 	rc = send_request_map(adapter, ltb->addr,
191 			      ltb->size, ltb->map_id);
192 	if (rc) {
193 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
194 		return rc;
195 	}
196 	wait_for_completion(&adapter->fw_done);
197 
198 	if (adapter->fw_done_rc) {
199 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
200 			adapter->fw_done_rc);
201 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
202 		return -1;
203 	}
204 	return 0;
205 }
206 
207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
208 				struct ibmvnic_long_term_buff *ltb)
209 {
210 	struct device *dev = &adapter->vdev->dev;
211 
212 	if (!ltb->buff)
213 		return;
214 
215 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
216 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
217 		send_request_unmap(adapter, ltb->map_id);
218 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
219 }
220 
221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
222 				struct ibmvnic_long_term_buff *ltb)
223 {
224 	int rc;
225 
226 	memset(ltb->buff, 0, ltb->size);
227 
228 	init_completion(&adapter->fw_done);
229 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
230 	if (rc)
231 		return rc;
232 	wait_for_completion(&adapter->fw_done);
233 
234 	if (adapter->fw_done_rc) {
235 		dev_info(&adapter->vdev->dev,
236 			 "Reset failed, attempting to free and reallocate buffer\n");
237 		free_long_term_buff(adapter, ltb);
238 		return alloc_long_term_buff(adapter, ltb, ltb->size);
239 	}
240 	return 0;
241 }
242 
243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
244 {
245 	int i;
246 
247 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
248 	     i++)
249 		adapter->rx_pool[i].active = 0;
250 }
251 
252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
253 			      struct ibmvnic_rx_pool *pool)
254 {
255 	int count = pool->size - atomic_read(&pool->available);
256 	struct device *dev = &adapter->vdev->dev;
257 	int buffers_added = 0;
258 	unsigned long lpar_rc;
259 	union sub_crq sub_crq;
260 	struct sk_buff *skb;
261 	unsigned int offset;
262 	dma_addr_t dma_addr;
263 	unsigned char *dst;
264 	u64 *handle_array;
265 	int shift = 0;
266 	int index;
267 	int i;
268 
269 	if (!pool->active)
270 		return;
271 
272 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
273 				      be32_to_cpu(adapter->login_rsp_buf->
274 				      off_rxadd_subcrqs));
275 
276 	for (i = 0; i < count; ++i) {
277 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
278 		if (!skb) {
279 			dev_err(dev, "Couldn't replenish rx buff\n");
280 			adapter->replenish_no_mem++;
281 			break;
282 		}
283 
284 		index = pool->free_map[pool->next_free];
285 
286 		if (pool->rx_buff[index].skb)
287 			dev_err(dev, "Inconsistent free_map!\n");
288 
289 		/* Copy the skb to the long term mapped DMA buffer */
290 		offset = index * pool->buff_size;
291 		dst = pool->long_term_buff.buff + offset;
292 		memset(dst, 0, pool->buff_size);
293 		dma_addr = pool->long_term_buff.addr + offset;
294 		pool->rx_buff[index].data = dst;
295 
296 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
297 		pool->rx_buff[index].dma = dma_addr;
298 		pool->rx_buff[index].skb = skb;
299 		pool->rx_buff[index].pool_index = pool->index;
300 		pool->rx_buff[index].size = pool->buff_size;
301 
302 		memset(&sub_crq, 0, sizeof(sub_crq));
303 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
304 		sub_crq.rx_add.correlator =
305 		    cpu_to_be64((u64)&pool->rx_buff[index]);
306 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
307 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
308 
309 		/* The length field of the sCRQ is defined to be 24 bits so the
310 		 * buffer size needs to be left shifted by a byte before it is
311 		 * converted to big endian to prevent the last byte from being
312 		 * truncated.
313 		 */
314 #ifdef __LITTLE_ENDIAN__
315 		shift = 8;
316 #endif
317 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
318 
319 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
320 				      &sub_crq);
321 		if (lpar_rc != H_SUCCESS)
322 			goto failure;
323 
324 		buffers_added++;
325 		adapter->replenish_add_buff_success++;
326 		pool->next_free = (pool->next_free + 1) % pool->size;
327 	}
328 	atomic_add(buffers_added, &pool->available);
329 	return;
330 
331 failure:
332 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
333 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
334 	pool->free_map[pool->next_free] = index;
335 	pool->rx_buff[index].skb = NULL;
336 
337 	dev_kfree_skb_any(skb);
338 	adapter->replenish_add_buff_failure++;
339 	atomic_add(buffers_added, &pool->available);
340 
341 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
342 		/* Disable buffer pool replenishment and report carrier off if
343 		 * queue is closed or pending failover.
344 		 * Firmware guarantees that a signal will be sent to the
345 		 * driver, triggering a reset.
346 		 */
347 		deactivate_rx_pools(adapter);
348 		netif_carrier_off(adapter->netdev);
349 	}
350 }
351 
352 static void replenish_pools(struct ibmvnic_adapter *adapter)
353 {
354 	int i;
355 
356 	adapter->replenish_task_cycles++;
357 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
358 	     i++) {
359 		if (adapter->rx_pool[i].active)
360 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
361 	}
362 }
363 
364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
365 {
366 	kfree(adapter->tx_stats_buffers);
367 	kfree(adapter->rx_stats_buffers);
368 	adapter->tx_stats_buffers = NULL;
369 	adapter->rx_stats_buffers = NULL;
370 }
371 
372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
373 {
374 	adapter->tx_stats_buffers =
375 				kcalloc(IBMVNIC_MAX_QUEUES,
376 					sizeof(struct ibmvnic_tx_queue_stats),
377 					GFP_KERNEL);
378 	if (!adapter->tx_stats_buffers)
379 		return -ENOMEM;
380 
381 	adapter->rx_stats_buffers =
382 				kcalloc(IBMVNIC_MAX_QUEUES,
383 					sizeof(struct ibmvnic_rx_queue_stats),
384 					GFP_KERNEL);
385 	if (!adapter->rx_stats_buffers)
386 		return -ENOMEM;
387 
388 	return 0;
389 }
390 
391 static void release_stats_token(struct ibmvnic_adapter *adapter)
392 {
393 	struct device *dev = &adapter->vdev->dev;
394 
395 	if (!adapter->stats_token)
396 		return;
397 
398 	dma_unmap_single(dev, adapter->stats_token,
399 			 sizeof(struct ibmvnic_statistics),
400 			 DMA_FROM_DEVICE);
401 	adapter->stats_token = 0;
402 }
403 
404 static int init_stats_token(struct ibmvnic_adapter *adapter)
405 {
406 	struct device *dev = &adapter->vdev->dev;
407 	dma_addr_t stok;
408 
409 	stok = dma_map_single(dev, &adapter->stats,
410 			      sizeof(struct ibmvnic_statistics),
411 			      DMA_FROM_DEVICE);
412 	if (dma_mapping_error(dev, stok)) {
413 		dev_err(dev, "Couldn't map stats buffer\n");
414 		return -1;
415 	}
416 
417 	adapter->stats_token = stok;
418 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
419 	return 0;
420 }
421 
422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
423 {
424 	struct ibmvnic_rx_pool *rx_pool;
425 	int rx_scrqs;
426 	int i, j, rc;
427 	u64 *size_array;
428 
429 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
430 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
431 
432 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
433 	for (i = 0; i < rx_scrqs; i++) {
434 		rx_pool = &adapter->rx_pool[i];
435 
436 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
437 
438 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
439 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
440 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
441 			alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
442 					     rx_pool->size *
443 					     rx_pool->buff_size);
444 		} else {
445 			rc = reset_long_term_buff(adapter,
446 						  &rx_pool->long_term_buff);
447 		}
448 
449 		if (rc)
450 			return rc;
451 
452 		for (j = 0; j < rx_pool->size; j++)
453 			rx_pool->free_map[j] = j;
454 
455 		memset(rx_pool->rx_buff, 0,
456 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
457 
458 		atomic_set(&rx_pool->available, 0);
459 		rx_pool->next_alloc = 0;
460 		rx_pool->next_free = 0;
461 		rx_pool->active = 1;
462 	}
463 
464 	return 0;
465 }
466 
467 static void release_rx_pools(struct ibmvnic_adapter *adapter)
468 {
469 	struct ibmvnic_rx_pool *rx_pool;
470 	int i, j;
471 
472 	if (!adapter->rx_pool)
473 		return;
474 
475 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
476 		rx_pool = &adapter->rx_pool[i];
477 
478 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
479 
480 		kfree(rx_pool->free_map);
481 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
482 
483 		if (!rx_pool->rx_buff)
484 			continue;
485 
486 		for (j = 0; j < rx_pool->size; j++) {
487 			if (rx_pool->rx_buff[j].skb) {
488 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
489 				rx_pool->rx_buff[j].skb = NULL;
490 			}
491 		}
492 
493 		kfree(rx_pool->rx_buff);
494 	}
495 
496 	kfree(adapter->rx_pool);
497 	adapter->rx_pool = NULL;
498 	adapter->num_active_rx_pools = 0;
499 }
500 
501 static int init_rx_pools(struct net_device *netdev)
502 {
503 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
504 	struct device *dev = &adapter->vdev->dev;
505 	struct ibmvnic_rx_pool *rx_pool;
506 	int rxadd_subcrqs;
507 	u64 *size_array;
508 	int i, j;
509 
510 	rxadd_subcrqs =
511 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
512 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
513 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
514 
515 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
516 				   sizeof(struct ibmvnic_rx_pool),
517 				   GFP_KERNEL);
518 	if (!adapter->rx_pool) {
519 		dev_err(dev, "Failed to allocate rx pools\n");
520 		return -1;
521 	}
522 
523 	adapter->num_active_rx_pools = rxadd_subcrqs;
524 
525 	for (i = 0; i < rxadd_subcrqs; i++) {
526 		rx_pool = &adapter->rx_pool[i];
527 
528 		netdev_dbg(adapter->netdev,
529 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
530 			   i, adapter->req_rx_add_entries_per_subcrq,
531 			   be64_to_cpu(size_array[i]));
532 
533 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
534 		rx_pool->index = i;
535 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
536 		rx_pool->active = 1;
537 
538 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
539 					    GFP_KERNEL);
540 		if (!rx_pool->free_map) {
541 			release_rx_pools(adapter);
542 			return -1;
543 		}
544 
545 		rx_pool->rx_buff = kcalloc(rx_pool->size,
546 					   sizeof(struct ibmvnic_rx_buff),
547 					   GFP_KERNEL);
548 		if (!rx_pool->rx_buff) {
549 			dev_err(dev, "Couldn't alloc rx buffers\n");
550 			release_rx_pools(adapter);
551 			return -1;
552 		}
553 
554 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
555 					 rx_pool->size * rx_pool->buff_size)) {
556 			release_rx_pools(adapter);
557 			return -1;
558 		}
559 
560 		for (j = 0; j < rx_pool->size; ++j)
561 			rx_pool->free_map[j] = j;
562 
563 		atomic_set(&rx_pool->available, 0);
564 		rx_pool->next_alloc = 0;
565 		rx_pool->next_free = 0;
566 	}
567 
568 	return 0;
569 }
570 
571 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
572 			     struct ibmvnic_tx_pool *tx_pool)
573 {
574 	int rc, i;
575 
576 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
577 	if (rc)
578 		return rc;
579 
580 	memset(tx_pool->tx_buff, 0,
581 	       tx_pool->num_buffers *
582 	       sizeof(struct ibmvnic_tx_buff));
583 
584 	for (i = 0; i < tx_pool->num_buffers; i++)
585 		tx_pool->free_map[i] = i;
586 
587 	tx_pool->consumer_index = 0;
588 	tx_pool->producer_index = 0;
589 
590 	return 0;
591 }
592 
593 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
594 {
595 	int tx_scrqs;
596 	int i, rc;
597 
598 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
599 	for (i = 0; i < tx_scrqs; i++) {
600 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
601 		if (rc)
602 			return rc;
603 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
604 		if (rc)
605 			return rc;
606 	}
607 
608 	return 0;
609 }
610 
611 static void release_vpd_data(struct ibmvnic_adapter *adapter)
612 {
613 	if (!adapter->vpd)
614 		return;
615 
616 	kfree(adapter->vpd->buff);
617 	kfree(adapter->vpd);
618 
619 	adapter->vpd = NULL;
620 }
621 
622 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
623 				struct ibmvnic_tx_pool *tx_pool)
624 {
625 	kfree(tx_pool->tx_buff);
626 	kfree(tx_pool->free_map);
627 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
628 }
629 
630 static void release_tx_pools(struct ibmvnic_adapter *adapter)
631 {
632 	int i;
633 
634 	if (!adapter->tx_pool)
635 		return;
636 
637 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
638 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
639 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
640 	}
641 
642 	kfree(adapter->tx_pool);
643 	adapter->tx_pool = NULL;
644 	kfree(adapter->tso_pool);
645 	adapter->tso_pool = NULL;
646 	adapter->num_active_tx_pools = 0;
647 }
648 
649 static int init_one_tx_pool(struct net_device *netdev,
650 			    struct ibmvnic_tx_pool *tx_pool,
651 			    int num_entries, int buf_size)
652 {
653 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
654 	int i;
655 
656 	tx_pool->tx_buff = kcalloc(num_entries,
657 				   sizeof(struct ibmvnic_tx_buff),
658 				   GFP_KERNEL);
659 	if (!tx_pool->tx_buff)
660 		return -1;
661 
662 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
663 				 num_entries * buf_size))
664 		return -1;
665 
666 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
667 	if (!tx_pool->free_map)
668 		return -1;
669 
670 	for (i = 0; i < num_entries; i++)
671 		tx_pool->free_map[i] = i;
672 
673 	tx_pool->consumer_index = 0;
674 	tx_pool->producer_index = 0;
675 	tx_pool->num_buffers = num_entries;
676 	tx_pool->buf_size = buf_size;
677 
678 	return 0;
679 }
680 
681 static int init_tx_pools(struct net_device *netdev)
682 {
683 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
684 	int tx_subcrqs;
685 	int i, rc;
686 
687 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
688 	adapter->tx_pool = kcalloc(tx_subcrqs,
689 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
690 	if (!adapter->tx_pool)
691 		return -1;
692 
693 	adapter->tso_pool = kcalloc(tx_subcrqs,
694 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
695 	if (!adapter->tso_pool)
696 		return -1;
697 
698 	adapter->num_active_tx_pools = tx_subcrqs;
699 
700 	for (i = 0; i < tx_subcrqs; i++) {
701 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
702 				      adapter->req_tx_entries_per_subcrq,
703 				      adapter->req_mtu + VLAN_HLEN);
704 		if (rc) {
705 			release_tx_pools(adapter);
706 			return rc;
707 		}
708 
709 		init_one_tx_pool(netdev, &adapter->tso_pool[i],
710 				 IBMVNIC_TSO_BUFS,
711 				 IBMVNIC_TSO_BUF_SZ);
712 		if (rc) {
713 			release_tx_pools(adapter);
714 			return rc;
715 		}
716 	}
717 
718 	return 0;
719 }
720 
721 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
722 {
723 	int i;
724 
725 	if (adapter->napi_enabled)
726 		return;
727 
728 	for (i = 0; i < adapter->req_rx_queues; i++)
729 		napi_enable(&adapter->napi[i]);
730 
731 	adapter->napi_enabled = true;
732 }
733 
734 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
735 {
736 	int i;
737 
738 	if (!adapter->napi_enabled)
739 		return;
740 
741 	for (i = 0; i < adapter->req_rx_queues; i++) {
742 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
743 		napi_disable(&adapter->napi[i]);
744 	}
745 
746 	adapter->napi_enabled = false;
747 }
748 
749 static int init_napi(struct ibmvnic_adapter *adapter)
750 {
751 	int i;
752 
753 	adapter->napi = kcalloc(adapter->req_rx_queues,
754 				sizeof(struct napi_struct), GFP_KERNEL);
755 	if (!adapter->napi)
756 		return -ENOMEM;
757 
758 	for (i = 0; i < adapter->req_rx_queues; i++) {
759 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
760 		netif_napi_add(adapter->netdev, &adapter->napi[i],
761 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
762 	}
763 
764 	adapter->num_active_rx_napi = adapter->req_rx_queues;
765 	return 0;
766 }
767 
768 static void release_napi(struct ibmvnic_adapter *adapter)
769 {
770 	int i;
771 
772 	if (!adapter->napi)
773 		return;
774 
775 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
776 		if (&adapter->napi[i]) {
777 			netdev_dbg(adapter->netdev,
778 				   "Releasing napi[%d]\n", i);
779 			netif_napi_del(&adapter->napi[i]);
780 		}
781 	}
782 
783 	kfree(adapter->napi);
784 	adapter->napi = NULL;
785 	adapter->num_active_rx_napi = 0;
786 	adapter->napi_enabled = false;
787 }
788 
789 static int ibmvnic_login(struct net_device *netdev)
790 {
791 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
792 	unsigned long timeout = msecs_to_jiffies(30000);
793 	int retry_count = 0;
794 	bool retry;
795 	int rc;
796 
797 	do {
798 		retry = false;
799 		if (retry_count > IBMVNIC_MAX_QUEUES) {
800 			netdev_warn(netdev, "Login attempts exceeded\n");
801 			return -1;
802 		}
803 
804 		adapter->init_done_rc = 0;
805 		reinit_completion(&adapter->init_done);
806 		rc = send_login(adapter);
807 		if (rc) {
808 			netdev_warn(netdev, "Unable to login\n");
809 			return rc;
810 		}
811 
812 		if (!wait_for_completion_timeout(&adapter->init_done,
813 						 timeout)) {
814 			netdev_warn(netdev, "Login timed out\n");
815 			return -1;
816 		}
817 
818 		if (adapter->init_done_rc == PARTIALSUCCESS) {
819 			retry_count++;
820 			release_sub_crqs(adapter, 1);
821 
822 			retry = true;
823 			netdev_dbg(netdev,
824 				   "Received partial success, retrying...\n");
825 			adapter->init_done_rc = 0;
826 			reinit_completion(&adapter->init_done);
827 			send_cap_queries(adapter);
828 			if (!wait_for_completion_timeout(&adapter->init_done,
829 							 timeout)) {
830 				netdev_warn(netdev,
831 					    "Capabilities query timed out\n");
832 				return -1;
833 			}
834 
835 			rc = init_sub_crqs(adapter);
836 			if (rc) {
837 				netdev_warn(netdev,
838 					    "SCRQ initialization failed\n");
839 				return -1;
840 			}
841 
842 			rc = init_sub_crq_irqs(adapter);
843 			if (rc) {
844 				netdev_warn(netdev,
845 					    "SCRQ irq initialization failed\n");
846 				return -1;
847 			}
848 		} else if (adapter->init_done_rc) {
849 			netdev_warn(netdev, "Adapter login failed\n");
850 			return -1;
851 		}
852 	} while (retry);
853 
854 	/* handle pending MAC address changes after successful login */
855 	if (adapter->mac_change_pending) {
856 		__ibmvnic_set_mac(netdev, &adapter->desired.mac);
857 		adapter->mac_change_pending = false;
858 	}
859 
860 	return 0;
861 }
862 
863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
864 {
865 	kfree(adapter->login_buf);
866 	adapter->login_buf = NULL;
867 }
868 
869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
870 {
871 	kfree(adapter->login_rsp_buf);
872 	adapter->login_rsp_buf = NULL;
873 }
874 
875 static void release_resources(struct ibmvnic_adapter *adapter)
876 {
877 	release_vpd_data(adapter);
878 
879 	release_tx_pools(adapter);
880 	release_rx_pools(adapter);
881 
882 	release_napi(adapter);
883 	release_login_rsp_buffer(adapter);
884 }
885 
886 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
887 {
888 	struct net_device *netdev = adapter->netdev;
889 	unsigned long timeout = msecs_to_jiffies(30000);
890 	union ibmvnic_crq crq;
891 	bool resend;
892 	int rc;
893 
894 	netdev_dbg(netdev, "setting link state %d\n", link_state);
895 
896 	memset(&crq, 0, sizeof(crq));
897 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
898 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
899 	crq.logical_link_state.link_state = link_state;
900 
901 	do {
902 		resend = false;
903 
904 		reinit_completion(&adapter->init_done);
905 		rc = ibmvnic_send_crq(adapter, &crq);
906 		if (rc) {
907 			netdev_err(netdev, "Failed to set link state\n");
908 			return rc;
909 		}
910 
911 		if (!wait_for_completion_timeout(&adapter->init_done,
912 						 timeout)) {
913 			netdev_err(netdev, "timeout setting link state\n");
914 			return -1;
915 		}
916 
917 		if (adapter->init_done_rc == 1) {
918 			/* Partuial success, delay and re-send */
919 			mdelay(1000);
920 			resend = true;
921 		} else if (adapter->init_done_rc) {
922 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
923 				    adapter->init_done_rc);
924 			return adapter->init_done_rc;
925 		}
926 	} while (resend);
927 
928 	return 0;
929 }
930 
931 static int set_real_num_queues(struct net_device *netdev)
932 {
933 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
934 	int rc;
935 
936 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
937 		   adapter->req_tx_queues, adapter->req_rx_queues);
938 
939 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
940 	if (rc) {
941 		netdev_err(netdev, "failed to set the number of tx queues\n");
942 		return rc;
943 	}
944 
945 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
946 	if (rc)
947 		netdev_err(netdev, "failed to set the number of rx queues\n");
948 
949 	return rc;
950 }
951 
952 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
953 {
954 	struct device *dev = &adapter->vdev->dev;
955 	union ibmvnic_crq crq;
956 	int len = 0;
957 	int rc;
958 
959 	if (adapter->vpd->buff)
960 		len = adapter->vpd->len;
961 
962 	init_completion(&adapter->fw_done);
963 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
964 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
965 	rc = ibmvnic_send_crq(adapter, &crq);
966 	if (rc)
967 		return rc;
968 	wait_for_completion(&adapter->fw_done);
969 
970 	if (!adapter->vpd->len)
971 		return -ENODATA;
972 
973 	if (!adapter->vpd->buff)
974 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
975 	else if (adapter->vpd->len != len)
976 		adapter->vpd->buff =
977 			krealloc(adapter->vpd->buff,
978 				 adapter->vpd->len, GFP_KERNEL);
979 
980 	if (!adapter->vpd->buff) {
981 		dev_err(dev, "Could allocate VPD buffer\n");
982 		return -ENOMEM;
983 	}
984 
985 	adapter->vpd->dma_addr =
986 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
987 			       DMA_FROM_DEVICE);
988 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
989 		dev_err(dev, "Could not map VPD buffer\n");
990 		kfree(adapter->vpd->buff);
991 		adapter->vpd->buff = NULL;
992 		return -ENOMEM;
993 	}
994 
995 	reinit_completion(&adapter->fw_done);
996 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
997 	crq.get_vpd.cmd = GET_VPD;
998 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
999 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1000 	rc = ibmvnic_send_crq(adapter, &crq);
1001 	if (rc) {
1002 		kfree(adapter->vpd->buff);
1003 		adapter->vpd->buff = NULL;
1004 		return rc;
1005 	}
1006 	wait_for_completion(&adapter->fw_done);
1007 
1008 	return 0;
1009 }
1010 
1011 static int init_resources(struct ibmvnic_adapter *adapter)
1012 {
1013 	struct net_device *netdev = adapter->netdev;
1014 	int rc;
1015 
1016 	rc = set_real_num_queues(netdev);
1017 	if (rc)
1018 		return rc;
1019 
1020 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1021 	if (!adapter->vpd)
1022 		return -ENOMEM;
1023 
1024 	/* Vital Product Data (VPD) */
1025 	rc = ibmvnic_get_vpd(adapter);
1026 	if (rc) {
1027 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1028 		return rc;
1029 	}
1030 
1031 	adapter->map_id = 1;
1032 
1033 	rc = init_napi(adapter);
1034 	if (rc)
1035 		return rc;
1036 
1037 	send_map_query(adapter);
1038 
1039 	rc = init_rx_pools(netdev);
1040 	if (rc)
1041 		return rc;
1042 
1043 	rc = init_tx_pools(netdev);
1044 	return rc;
1045 }
1046 
1047 static int __ibmvnic_open(struct net_device *netdev)
1048 {
1049 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1050 	enum vnic_state prev_state = adapter->state;
1051 	int i, rc;
1052 
1053 	adapter->state = VNIC_OPENING;
1054 	replenish_pools(adapter);
1055 	ibmvnic_napi_enable(adapter);
1056 
1057 	/* We're ready to receive frames, enable the sub-crq interrupts and
1058 	 * set the logical link state to up
1059 	 */
1060 	for (i = 0; i < adapter->req_rx_queues; i++) {
1061 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1062 		if (prev_state == VNIC_CLOSED)
1063 			enable_irq(adapter->rx_scrq[i]->irq);
1064 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1065 	}
1066 
1067 	for (i = 0; i < adapter->req_tx_queues; i++) {
1068 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1069 		if (prev_state == VNIC_CLOSED)
1070 			enable_irq(adapter->tx_scrq[i]->irq);
1071 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1072 	}
1073 
1074 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1075 	if (rc) {
1076 		for (i = 0; i < adapter->req_rx_queues; i++)
1077 			napi_disable(&adapter->napi[i]);
1078 		release_resources(adapter);
1079 		return rc;
1080 	}
1081 
1082 	netif_tx_start_all_queues(netdev);
1083 
1084 	if (prev_state == VNIC_CLOSED) {
1085 		for (i = 0; i < adapter->req_rx_queues; i++)
1086 			napi_schedule(&adapter->napi[i]);
1087 	}
1088 
1089 	adapter->state = VNIC_OPEN;
1090 	return rc;
1091 }
1092 
1093 static int ibmvnic_open(struct net_device *netdev)
1094 {
1095 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1096 	int rc;
1097 
1098 	/* If device failover is pending, just set device state and return.
1099 	 * Device operation will be handled by reset routine.
1100 	 */
1101 	if (adapter->failover_pending) {
1102 		adapter->state = VNIC_OPEN;
1103 		return 0;
1104 	}
1105 
1106 	if (adapter->state != VNIC_CLOSED) {
1107 		rc = ibmvnic_login(netdev);
1108 		if (rc)
1109 			return rc;
1110 
1111 		rc = init_resources(adapter);
1112 		if (rc) {
1113 			netdev_err(netdev, "failed to initialize resources\n");
1114 			release_resources(adapter);
1115 			return rc;
1116 		}
1117 	}
1118 
1119 	rc = __ibmvnic_open(netdev);
1120 	netif_carrier_on(netdev);
1121 
1122 	return rc;
1123 }
1124 
1125 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1126 {
1127 	struct ibmvnic_rx_pool *rx_pool;
1128 	struct ibmvnic_rx_buff *rx_buff;
1129 	u64 rx_entries;
1130 	int rx_scrqs;
1131 	int i, j;
1132 
1133 	if (!adapter->rx_pool)
1134 		return;
1135 
1136 	rx_scrqs = adapter->num_active_rx_pools;
1137 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1138 
1139 	/* Free any remaining skbs in the rx buffer pools */
1140 	for (i = 0; i < rx_scrqs; i++) {
1141 		rx_pool = &adapter->rx_pool[i];
1142 		if (!rx_pool || !rx_pool->rx_buff)
1143 			continue;
1144 
1145 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1146 		for (j = 0; j < rx_entries; j++) {
1147 			rx_buff = &rx_pool->rx_buff[j];
1148 			if (rx_buff && rx_buff->skb) {
1149 				dev_kfree_skb_any(rx_buff->skb);
1150 				rx_buff->skb = NULL;
1151 			}
1152 		}
1153 	}
1154 }
1155 
1156 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1157 			      struct ibmvnic_tx_pool *tx_pool)
1158 {
1159 	struct ibmvnic_tx_buff *tx_buff;
1160 	u64 tx_entries;
1161 	int i;
1162 
1163 	if (!tx_pool || !tx_pool->tx_buff)
1164 		return;
1165 
1166 	tx_entries = tx_pool->num_buffers;
1167 
1168 	for (i = 0; i < tx_entries; i++) {
1169 		tx_buff = &tx_pool->tx_buff[i];
1170 		if (tx_buff && tx_buff->skb) {
1171 			dev_kfree_skb_any(tx_buff->skb);
1172 			tx_buff->skb = NULL;
1173 		}
1174 	}
1175 }
1176 
1177 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1178 {
1179 	int tx_scrqs;
1180 	int i;
1181 
1182 	if (!adapter->tx_pool || !adapter->tso_pool)
1183 		return;
1184 
1185 	tx_scrqs = adapter->num_active_tx_pools;
1186 
1187 	/* Free any remaining skbs in the tx buffer pools */
1188 	for (i = 0; i < tx_scrqs; i++) {
1189 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1190 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1191 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1192 	}
1193 }
1194 
1195 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1196 {
1197 	struct net_device *netdev = adapter->netdev;
1198 	int i;
1199 
1200 	if (adapter->tx_scrq) {
1201 		for (i = 0; i < adapter->req_tx_queues; i++)
1202 			if (adapter->tx_scrq[i]->irq) {
1203 				netdev_dbg(netdev,
1204 					   "Disabling tx_scrq[%d] irq\n", i);
1205 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1206 				disable_irq(adapter->tx_scrq[i]->irq);
1207 			}
1208 	}
1209 
1210 	if (adapter->rx_scrq) {
1211 		for (i = 0; i < adapter->req_rx_queues; i++) {
1212 			if (adapter->rx_scrq[i]->irq) {
1213 				netdev_dbg(netdev,
1214 					   "Disabling rx_scrq[%d] irq\n", i);
1215 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1216 				disable_irq(adapter->rx_scrq[i]->irq);
1217 			}
1218 		}
1219 	}
1220 }
1221 
1222 static void ibmvnic_cleanup(struct net_device *netdev)
1223 {
1224 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1225 
1226 	/* ensure that transmissions are stopped if called by do_reset */
1227 	if (adapter->resetting)
1228 		netif_tx_disable(netdev);
1229 	else
1230 		netif_tx_stop_all_queues(netdev);
1231 
1232 	ibmvnic_napi_disable(adapter);
1233 	ibmvnic_disable_irqs(adapter);
1234 
1235 	clean_rx_pools(adapter);
1236 	clean_tx_pools(adapter);
1237 }
1238 
1239 static int __ibmvnic_close(struct net_device *netdev)
1240 {
1241 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1242 	int rc = 0;
1243 
1244 	adapter->state = VNIC_CLOSING;
1245 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1246 	if (rc)
1247 		return rc;
1248 	adapter->state = VNIC_CLOSED;
1249 	return 0;
1250 }
1251 
1252 static int ibmvnic_close(struct net_device *netdev)
1253 {
1254 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1255 	int rc;
1256 
1257 	/* If device failover is pending, just set device state and return.
1258 	 * Device operation will be handled by reset routine.
1259 	 */
1260 	if (adapter->failover_pending) {
1261 		adapter->state = VNIC_CLOSED;
1262 		return 0;
1263 	}
1264 
1265 	rc = __ibmvnic_close(netdev);
1266 	ibmvnic_cleanup(netdev);
1267 
1268 	return rc;
1269 }
1270 
1271 /**
1272  * build_hdr_data - creates L2/L3/L4 header data buffer
1273  * @hdr_field - bitfield determining needed headers
1274  * @skb - socket buffer
1275  * @hdr_len - array of header lengths
1276  * @tot_len - total length of data
1277  *
1278  * Reads hdr_field to determine which headers are needed by firmware.
1279  * Builds a buffer containing these headers.  Saves individual header
1280  * lengths and total buffer length to be used to build descriptors.
1281  */
1282 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1283 			  int *hdr_len, u8 *hdr_data)
1284 {
1285 	int len = 0;
1286 	u8 *hdr;
1287 
1288 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1289 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1290 	else
1291 		hdr_len[0] = sizeof(struct ethhdr);
1292 
1293 	if (skb->protocol == htons(ETH_P_IP)) {
1294 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1295 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1296 			hdr_len[2] = tcp_hdrlen(skb);
1297 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1298 			hdr_len[2] = sizeof(struct udphdr);
1299 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1300 		hdr_len[1] = sizeof(struct ipv6hdr);
1301 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1302 			hdr_len[2] = tcp_hdrlen(skb);
1303 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1304 			hdr_len[2] = sizeof(struct udphdr);
1305 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1306 		hdr_len[1] = arp_hdr_len(skb->dev);
1307 		hdr_len[2] = 0;
1308 	}
1309 
1310 	memset(hdr_data, 0, 120);
1311 	if ((hdr_field >> 6) & 1) {
1312 		hdr = skb_mac_header(skb);
1313 		memcpy(hdr_data, hdr, hdr_len[0]);
1314 		len += hdr_len[0];
1315 	}
1316 
1317 	if ((hdr_field >> 5) & 1) {
1318 		hdr = skb_network_header(skb);
1319 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1320 		len += hdr_len[1];
1321 	}
1322 
1323 	if ((hdr_field >> 4) & 1) {
1324 		hdr = skb_transport_header(skb);
1325 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1326 		len += hdr_len[2];
1327 	}
1328 	return len;
1329 }
1330 
1331 /**
1332  * create_hdr_descs - create header and header extension descriptors
1333  * @hdr_field - bitfield determining needed headers
1334  * @data - buffer containing header data
1335  * @len - length of data buffer
1336  * @hdr_len - array of individual header lengths
1337  * @scrq_arr - descriptor array
1338  *
1339  * Creates header and, if needed, header extension descriptors and
1340  * places them in a descriptor array, scrq_arr
1341  */
1342 
1343 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1344 			    union sub_crq *scrq_arr)
1345 {
1346 	union sub_crq hdr_desc;
1347 	int tmp_len = len;
1348 	int num_descs = 0;
1349 	u8 *data, *cur;
1350 	int tmp;
1351 
1352 	while (tmp_len > 0) {
1353 		cur = hdr_data + len - tmp_len;
1354 
1355 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1356 		if (cur != hdr_data) {
1357 			data = hdr_desc.hdr_ext.data;
1358 			tmp = tmp_len > 29 ? 29 : tmp_len;
1359 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1360 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1361 			hdr_desc.hdr_ext.len = tmp;
1362 		} else {
1363 			data = hdr_desc.hdr.data;
1364 			tmp = tmp_len > 24 ? 24 : tmp_len;
1365 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1366 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1367 			hdr_desc.hdr.len = tmp;
1368 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1369 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1370 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1371 			hdr_desc.hdr.flag = hdr_field << 1;
1372 		}
1373 		memcpy(data, cur, tmp);
1374 		tmp_len -= tmp;
1375 		*scrq_arr = hdr_desc;
1376 		scrq_arr++;
1377 		num_descs++;
1378 	}
1379 
1380 	return num_descs;
1381 }
1382 
1383 /**
1384  * build_hdr_descs_arr - build a header descriptor array
1385  * @skb - socket buffer
1386  * @num_entries - number of descriptors to be sent
1387  * @subcrq - first TX descriptor
1388  * @hdr_field - bit field determining which headers will be sent
1389  *
1390  * This function will build a TX descriptor array with applicable
1391  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1392  */
1393 
1394 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1395 				int *num_entries, u8 hdr_field)
1396 {
1397 	int hdr_len[3] = {0, 0, 0};
1398 	int tot_len;
1399 	u8 *hdr_data = txbuff->hdr_data;
1400 
1401 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1402 				 txbuff->hdr_data);
1403 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1404 			 txbuff->indir_arr + 1);
1405 }
1406 
1407 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1408 				    struct net_device *netdev)
1409 {
1410 	/* For some backing devices, mishandling of small packets
1411 	 * can result in a loss of connection or TX stall. Device
1412 	 * architects recommend that no packet should be smaller
1413 	 * than the minimum MTU value provided to the driver, so
1414 	 * pad any packets to that length
1415 	 */
1416 	if (skb->len < netdev->min_mtu)
1417 		return skb_put_padto(skb, netdev->min_mtu);
1418 
1419 	return 0;
1420 }
1421 
1422 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1423 {
1424 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1425 	int queue_num = skb_get_queue_mapping(skb);
1426 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1427 	struct device *dev = &adapter->vdev->dev;
1428 	struct ibmvnic_tx_buff *tx_buff = NULL;
1429 	struct ibmvnic_sub_crq_queue *tx_scrq;
1430 	struct ibmvnic_tx_pool *tx_pool;
1431 	unsigned int tx_send_failed = 0;
1432 	unsigned int tx_map_failed = 0;
1433 	unsigned int tx_dropped = 0;
1434 	unsigned int tx_packets = 0;
1435 	unsigned int tx_bytes = 0;
1436 	dma_addr_t data_dma_addr;
1437 	struct netdev_queue *txq;
1438 	unsigned long lpar_rc;
1439 	union sub_crq tx_crq;
1440 	unsigned int offset;
1441 	int num_entries = 1;
1442 	unsigned char *dst;
1443 	u64 *handle_array;
1444 	int index = 0;
1445 	u8 proto = 0;
1446 	netdev_tx_t ret = NETDEV_TX_OK;
1447 
1448 	if (adapter->resetting) {
1449 		if (!netif_subqueue_stopped(netdev, skb))
1450 			netif_stop_subqueue(netdev, queue_num);
1451 		dev_kfree_skb_any(skb);
1452 
1453 		tx_send_failed++;
1454 		tx_dropped++;
1455 		ret = NETDEV_TX_OK;
1456 		goto out;
1457 	}
1458 
1459 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1460 		tx_dropped++;
1461 		tx_send_failed++;
1462 		ret = NETDEV_TX_OK;
1463 		goto out;
1464 	}
1465 	if (skb_is_gso(skb))
1466 		tx_pool = &adapter->tso_pool[queue_num];
1467 	else
1468 		tx_pool = &adapter->tx_pool[queue_num];
1469 
1470 	tx_scrq = adapter->tx_scrq[queue_num];
1471 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1472 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1473 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1474 
1475 	index = tx_pool->free_map[tx_pool->consumer_index];
1476 
1477 	if (index == IBMVNIC_INVALID_MAP) {
1478 		dev_kfree_skb_any(skb);
1479 		tx_send_failed++;
1480 		tx_dropped++;
1481 		ret = NETDEV_TX_OK;
1482 		goto out;
1483 	}
1484 
1485 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1486 
1487 	offset = index * tx_pool->buf_size;
1488 	dst = tx_pool->long_term_buff.buff + offset;
1489 	memset(dst, 0, tx_pool->buf_size);
1490 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1491 
1492 	if (skb_shinfo(skb)->nr_frags) {
1493 		int cur, i;
1494 
1495 		/* Copy the head */
1496 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1497 		cur = skb_headlen(skb);
1498 
1499 		/* Copy the frags */
1500 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1501 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1502 
1503 			memcpy(dst + cur,
1504 			       page_address(skb_frag_page(frag)) +
1505 			       frag->page_offset, skb_frag_size(frag));
1506 			cur += skb_frag_size(frag);
1507 		}
1508 	} else {
1509 		skb_copy_from_linear_data(skb, dst, skb->len);
1510 	}
1511 
1512 	tx_pool->consumer_index =
1513 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1514 
1515 	tx_buff = &tx_pool->tx_buff[index];
1516 	tx_buff->skb = skb;
1517 	tx_buff->data_dma[0] = data_dma_addr;
1518 	tx_buff->data_len[0] = skb->len;
1519 	tx_buff->index = index;
1520 	tx_buff->pool_index = queue_num;
1521 	tx_buff->last_frag = true;
1522 
1523 	memset(&tx_crq, 0, sizeof(tx_crq));
1524 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1525 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1526 	tx_crq.v1.n_crq_elem = 1;
1527 	tx_crq.v1.n_sge = 1;
1528 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1529 
1530 	if (skb_is_gso(skb))
1531 		tx_crq.v1.correlator =
1532 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1533 	else
1534 		tx_crq.v1.correlator = cpu_to_be32(index);
1535 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1536 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1537 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1538 
1539 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1540 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1541 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1542 	}
1543 
1544 	if (skb->protocol == htons(ETH_P_IP)) {
1545 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1546 		proto = ip_hdr(skb)->protocol;
1547 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1548 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1549 		proto = ipv6_hdr(skb)->nexthdr;
1550 	}
1551 
1552 	if (proto == IPPROTO_TCP)
1553 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1554 	else if (proto == IPPROTO_UDP)
1555 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1556 
1557 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1558 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1559 		hdrs += 2;
1560 	}
1561 	if (skb_is_gso(skb)) {
1562 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1563 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1564 		hdrs += 2;
1565 	}
1566 	/* determine if l2/3/4 headers are sent to firmware */
1567 	if ((*hdrs >> 7) & 1) {
1568 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1569 		tx_crq.v1.n_crq_elem = num_entries;
1570 		tx_buff->num_entries = num_entries;
1571 		tx_buff->indir_arr[0] = tx_crq;
1572 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1573 						    sizeof(tx_buff->indir_arr),
1574 						    DMA_TO_DEVICE);
1575 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1576 			dev_kfree_skb_any(skb);
1577 			tx_buff->skb = NULL;
1578 			if (!firmware_has_feature(FW_FEATURE_CMO))
1579 				dev_err(dev, "tx: unable to map descriptor array\n");
1580 			tx_map_failed++;
1581 			tx_dropped++;
1582 			ret = NETDEV_TX_OK;
1583 			goto tx_err_out;
1584 		}
1585 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1586 					       (u64)tx_buff->indir_dma,
1587 					       (u64)num_entries);
1588 	} else {
1589 		tx_buff->num_entries = num_entries;
1590 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1591 				      &tx_crq);
1592 	}
1593 	if (lpar_rc != H_SUCCESS) {
1594 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1595 			dev_err_ratelimited(dev, "tx: send failed\n");
1596 		dev_kfree_skb_any(skb);
1597 		tx_buff->skb = NULL;
1598 
1599 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1600 			/* Disable TX and report carrier off if queue is closed
1601 			 * or pending failover.
1602 			 * Firmware guarantees that a signal will be sent to the
1603 			 * driver, triggering a reset or some other action.
1604 			 */
1605 			netif_tx_stop_all_queues(netdev);
1606 			netif_carrier_off(netdev);
1607 		}
1608 
1609 		tx_send_failed++;
1610 		tx_dropped++;
1611 		ret = NETDEV_TX_OK;
1612 		goto tx_err_out;
1613 	}
1614 
1615 	if (atomic_add_return(num_entries, &tx_scrq->used)
1616 					>= adapter->req_tx_entries_per_subcrq) {
1617 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1618 		netif_stop_subqueue(netdev, queue_num);
1619 	}
1620 
1621 	tx_packets++;
1622 	tx_bytes += skb->len;
1623 	txq->trans_start = jiffies;
1624 	ret = NETDEV_TX_OK;
1625 	goto out;
1626 
1627 tx_err_out:
1628 	/* roll back consumer index and map array*/
1629 	if (tx_pool->consumer_index == 0)
1630 		tx_pool->consumer_index =
1631 			tx_pool->num_buffers - 1;
1632 	else
1633 		tx_pool->consumer_index--;
1634 	tx_pool->free_map[tx_pool->consumer_index] = index;
1635 out:
1636 	netdev->stats.tx_dropped += tx_dropped;
1637 	netdev->stats.tx_bytes += tx_bytes;
1638 	netdev->stats.tx_packets += tx_packets;
1639 	adapter->tx_send_failed += tx_send_failed;
1640 	adapter->tx_map_failed += tx_map_failed;
1641 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1642 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1643 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1644 
1645 	return ret;
1646 }
1647 
1648 static void ibmvnic_set_multi(struct net_device *netdev)
1649 {
1650 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1651 	struct netdev_hw_addr *ha;
1652 	union ibmvnic_crq crq;
1653 
1654 	memset(&crq, 0, sizeof(crq));
1655 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1656 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1657 
1658 	if (netdev->flags & IFF_PROMISC) {
1659 		if (!adapter->promisc_supported)
1660 			return;
1661 	} else {
1662 		if (netdev->flags & IFF_ALLMULTI) {
1663 			/* Accept all multicast */
1664 			memset(&crq, 0, sizeof(crq));
1665 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1666 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1667 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1668 			ibmvnic_send_crq(adapter, &crq);
1669 		} else if (netdev_mc_empty(netdev)) {
1670 			/* Reject all multicast */
1671 			memset(&crq, 0, sizeof(crq));
1672 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1673 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1674 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1675 			ibmvnic_send_crq(adapter, &crq);
1676 		} else {
1677 			/* Accept one or more multicast(s) */
1678 			netdev_for_each_mc_addr(ha, netdev) {
1679 				memset(&crq, 0, sizeof(crq));
1680 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1681 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1682 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1683 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1684 						ha->addr);
1685 				ibmvnic_send_crq(adapter, &crq);
1686 			}
1687 		}
1688 	}
1689 }
1690 
1691 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1692 {
1693 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1694 	struct sockaddr *addr = p;
1695 	union ibmvnic_crq crq;
1696 	int rc;
1697 
1698 	if (!is_valid_ether_addr(addr->sa_data))
1699 		return -EADDRNOTAVAIL;
1700 
1701 	memset(&crq, 0, sizeof(crq));
1702 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1703 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1704 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1705 
1706 	init_completion(&adapter->fw_done);
1707 	rc = ibmvnic_send_crq(adapter, &crq);
1708 	if (rc)
1709 		return rc;
1710 	wait_for_completion(&adapter->fw_done);
1711 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1712 	return adapter->fw_done_rc ? -EIO : 0;
1713 }
1714 
1715 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1716 {
1717 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1718 	struct sockaddr *addr = p;
1719 	int rc;
1720 
1721 	if (adapter->state == VNIC_PROBED) {
1722 		memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1723 		adapter->mac_change_pending = true;
1724 		return 0;
1725 	}
1726 
1727 	rc = __ibmvnic_set_mac(netdev, addr);
1728 
1729 	return rc;
1730 }
1731 
1732 /**
1733  * do_reset returns zero if we are able to keep processing reset events, or
1734  * non-zero if we hit a fatal error and must halt.
1735  */
1736 static int do_reset(struct ibmvnic_adapter *adapter,
1737 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1738 {
1739 	u64 old_num_rx_queues, old_num_tx_queues;
1740 	u64 old_num_rx_slots, old_num_tx_slots;
1741 	struct net_device *netdev = adapter->netdev;
1742 	int i, rc;
1743 
1744 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1745 		   rwi->reset_reason);
1746 
1747 	netif_carrier_off(netdev);
1748 	adapter->reset_reason = rwi->reset_reason;
1749 
1750 	old_num_rx_queues = adapter->req_rx_queues;
1751 	old_num_tx_queues = adapter->req_tx_queues;
1752 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1753 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1754 
1755 	ibmvnic_cleanup(netdev);
1756 
1757 	if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1758 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1759 		rc = __ibmvnic_close(netdev);
1760 		if (rc)
1761 			return rc;
1762 	}
1763 
1764 	if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1765 	    adapter->wait_for_reset) {
1766 		release_resources(adapter);
1767 		release_sub_crqs(adapter, 1);
1768 		release_crq_queue(adapter);
1769 	}
1770 
1771 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1772 		/* remove the closed state so when we call open it appears
1773 		 * we are coming from the probed state.
1774 		 */
1775 		adapter->state = VNIC_PROBED;
1776 
1777 		if (adapter->wait_for_reset) {
1778 			rc = init_crq_queue(adapter);
1779 		} else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1780 			rc = ibmvnic_reenable_crq_queue(adapter);
1781 			release_sub_crqs(adapter, 1);
1782 		} else {
1783 			rc = ibmvnic_reset_crq(adapter);
1784 			if (!rc)
1785 				rc = vio_enable_interrupts(adapter->vdev);
1786 		}
1787 
1788 		if (rc) {
1789 			netdev_err(adapter->netdev,
1790 				   "Couldn't initialize crq. rc=%d\n", rc);
1791 			return rc;
1792 		}
1793 
1794 		rc = ibmvnic_reset_init(adapter);
1795 		if (rc)
1796 			return IBMVNIC_INIT_FAILED;
1797 
1798 		/* If the adapter was in PROBE state prior to the reset,
1799 		 * exit here.
1800 		 */
1801 		if (reset_state == VNIC_PROBED)
1802 			return 0;
1803 
1804 		rc = ibmvnic_login(netdev);
1805 		if (rc) {
1806 			adapter->state = reset_state;
1807 			return rc;
1808 		}
1809 
1810 		if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1811 		    adapter->wait_for_reset) {
1812 			rc = init_resources(adapter);
1813 			if (rc)
1814 				return rc;
1815 		} else if (adapter->req_rx_queues != old_num_rx_queues ||
1816 			   adapter->req_tx_queues != old_num_tx_queues ||
1817 			   adapter->req_rx_add_entries_per_subcrq !=
1818 							old_num_rx_slots ||
1819 			   adapter->req_tx_entries_per_subcrq !=
1820 							old_num_tx_slots) {
1821 			release_rx_pools(adapter);
1822 			release_tx_pools(adapter);
1823 			release_napi(adapter);
1824 			release_vpd_data(adapter);
1825 
1826 			rc = init_resources(adapter);
1827 			if (rc)
1828 				return rc;
1829 
1830 		} else {
1831 			rc = reset_tx_pools(adapter);
1832 			if (rc)
1833 				return rc;
1834 
1835 			rc = reset_rx_pools(adapter);
1836 			if (rc)
1837 				return rc;
1838 		}
1839 		ibmvnic_disable_irqs(adapter);
1840 	}
1841 	adapter->state = VNIC_CLOSED;
1842 
1843 	if (reset_state == VNIC_CLOSED)
1844 		return 0;
1845 
1846 	rc = __ibmvnic_open(netdev);
1847 	if (rc) {
1848 		if (list_empty(&adapter->rwi_list))
1849 			adapter->state = VNIC_CLOSED;
1850 		else
1851 			adapter->state = reset_state;
1852 
1853 		return 0;
1854 	}
1855 
1856 	/* kick napi */
1857 	for (i = 0; i < adapter->req_rx_queues; i++)
1858 		napi_schedule(&adapter->napi[i]);
1859 
1860 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1861 	    adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1862 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
1863 
1864 	netif_carrier_on(netdev);
1865 
1866 	return 0;
1867 }
1868 
1869 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1870 			 struct ibmvnic_rwi *rwi, u32 reset_state)
1871 {
1872 	struct net_device *netdev = adapter->netdev;
1873 	int rc;
1874 
1875 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1876 		   rwi->reset_reason);
1877 
1878 	netif_carrier_off(netdev);
1879 	adapter->reset_reason = rwi->reset_reason;
1880 
1881 	ibmvnic_cleanup(netdev);
1882 	release_resources(adapter);
1883 	release_sub_crqs(adapter, 0);
1884 	release_crq_queue(adapter);
1885 
1886 	/* remove the closed state so when we call open it appears
1887 	 * we are coming from the probed state.
1888 	 */
1889 	adapter->state = VNIC_PROBED;
1890 
1891 	rc = init_crq_queue(adapter);
1892 	if (rc) {
1893 		netdev_err(adapter->netdev,
1894 			   "Couldn't initialize crq. rc=%d\n", rc);
1895 		return rc;
1896 	}
1897 
1898 	rc = ibmvnic_init(adapter);
1899 	if (rc)
1900 		return rc;
1901 
1902 	/* If the adapter was in PROBE state prior to the reset,
1903 	 * exit here.
1904 	 */
1905 	if (reset_state == VNIC_PROBED)
1906 		return 0;
1907 
1908 	rc = ibmvnic_login(netdev);
1909 	if (rc) {
1910 		adapter->state = VNIC_PROBED;
1911 		return 0;
1912 	}
1913 
1914 	rc = init_resources(adapter);
1915 	if (rc)
1916 		return rc;
1917 
1918 	ibmvnic_disable_irqs(adapter);
1919 	adapter->state = VNIC_CLOSED;
1920 
1921 	if (reset_state == VNIC_CLOSED)
1922 		return 0;
1923 
1924 	rc = __ibmvnic_open(netdev);
1925 	if (rc) {
1926 		if (list_empty(&adapter->rwi_list))
1927 			adapter->state = VNIC_CLOSED;
1928 		else
1929 			adapter->state = reset_state;
1930 
1931 		return 0;
1932 	}
1933 
1934 	netif_carrier_on(netdev);
1935 
1936 	return 0;
1937 }
1938 
1939 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1940 {
1941 	struct ibmvnic_rwi *rwi;
1942 	unsigned long flags;
1943 
1944 	spin_lock_irqsave(&adapter->rwi_lock, flags);
1945 
1946 	if (!list_empty(&adapter->rwi_list)) {
1947 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1948 				       list);
1949 		list_del(&rwi->list);
1950 	} else {
1951 		rwi = NULL;
1952 	}
1953 
1954 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
1955 	return rwi;
1956 }
1957 
1958 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1959 {
1960 	struct ibmvnic_rwi *rwi;
1961 
1962 	rwi = get_next_rwi(adapter);
1963 	while (rwi) {
1964 		kfree(rwi);
1965 		rwi = get_next_rwi(adapter);
1966 	}
1967 }
1968 
1969 static void __ibmvnic_reset(struct work_struct *work)
1970 {
1971 	struct ibmvnic_rwi *rwi;
1972 	struct ibmvnic_adapter *adapter;
1973 	struct net_device *netdev;
1974 	bool we_lock_rtnl = false;
1975 	u32 reset_state;
1976 	int rc = 0;
1977 
1978 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1979 	netdev = adapter->netdev;
1980 
1981 	/* netif_set_real_num_xx_queues needs to take rtnl lock here
1982 	 * unless wait_for_reset is set, in which case the rtnl lock
1983 	 * has already been taken before initializing the reset
1984 	 */
1985 	if (!adapter->wait_for_reset) {
1986 		rtnl_lock();
1987 		we_lock_rtnl = true;
1988 	}
1989 	reset_state = adapter->state;
1990 
1991 	rwi = get_next_rwi(adapter);
1992 	while (rwi) {
1993 		if (adapter->force_reset_recovery) {
1994 			adapter->force_reset_recovery = false;
1995 			rc = do_hard_reset(adapter, rwi, reset_state);
1996 		} else {
1997 			rc = do_reset(adapter, rwi, reset_state);
1998 		}
1999 		kfree(rwi);
2000 		if (rc && rc != IBMVNIC_INIT_FAILED &&
2001 		    !adapter->force_reset_recovery)
2002 			break;
2003 
2004 		rwi = get_next_rwi(adapter);
2005 	}
2006 
2007 	if (adapter->wait_for_reset) {
2008 		adapter->wait_for_reset = false;
2009 		adapter->reset_done_rc = rc;
2010 		complete(&adapter->reset_done);
2011 	}
2012 
2013 	if (rc) {
2014 		netdev_dbg(adapter->netdev, "Reset failed\n");
2015 		free_all_rwi(adapter);
2016 	}
2017 
2018 	adapter->resetting = false;
2019 	if (we_lock_rtnl)
2020 		rtnl_unlock();
2021 }
2022 
2023 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2024 			 enum ibmvnic_reset_reason reason)
2025 {
2026 	struct list_head *entry, *tmp_entry;
2027 	struct ibmvnic_rwi *rwi, *tmp;
2028 	struct net_device *netdev = adapter->netdev;
2029 	unsigned long flags;
2030 	int ret;
2031 
2032 	if (adapter->state == VNIC_REMOVING ||
2033 	    adapter->state == VNIC_REMOVED ||
2034 	    adapter->failover_pending) {
2035 		ret = EBUSY;
2036 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2037 		goto err;
2038 	}
2039 
2040 	if (adapter->state == VNIC_PROBING) {
2041 		netdev_warn(netdev, "Adapter reset during probe\n");
2042 		ret = adapter->init_done_rc = EAGAIN;
2043 		goto err;
2044 	}
2045 
2046 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2047 
2048 	list_for_each(entry, &adapter->rwi_list) {
2049 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2050 		if (tmp->reset_reason == reason) {
2051 			netdev_dbg(netdev, "Skipping matching reset\n");
2052 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2053 			ret = EBUSY;
2054 			goto err;
2055 		}
2056 	}
2057 
2058 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2059 	if (!rwi) {
2060 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2061 		ibmvnic_close(netdev);
2062 		ret = ENOMEM;
2063 		goto err;
2064 	}
2065 	/* if we just received a transport event,
2066 	 * flush reset queue and process this reset
2067 	 */
2068 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2069 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2070 			list_del(entry);
2071 	}
2072 	rwi->reset_reason = reason;
2073 	list_add_tail(&rwi->list, &adapter->rwi_list);
2074 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2075 	adapter->resetting = true;
2076 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2077 	schedule_work(&adapter->ibmvnic_reset);
2078 
2079 	return 0;
2080 err:
2081 	if (adapter->wait_for_reset)
2082 		adapter->wait_for_reset = false;
2083 	return -ret;
2084 }
2085 
2086 static void ibmvnic_tx_timeout(struct net_device *dev)
2087 {
2088 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2089 
2090 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2091 }
2092 
2093 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2094 				  struct ibmvnic_rx_buff *rx_buff)
2095 {
2096 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2097 
2098 	rx_buff->skb = NULL;
2099 
2100 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2101 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2102 
2103 	atomic_dec(&pool->available);
2104 }
2105 
2106 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2107 {
2108 	struct net_device *netdev = napi->dev;
2109 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2110 	int scrq_num = (int)(napi - adapter->napi);
2111 	int frames_processed = 0;
2112 
2113 restart_poll:
2114 	while (frames_processed < budget) {
2115 		struct sk_buff *skb;
2116 		struct ibmvnic_rx_buff *rx_buff;
2117 		union sub_crq *next;
2118 		u32 length;
2119 		u16 offset;
2120 		u8 flags = 0;
2121 
2122 		if (unlikely(adapter->resetting &&
2123 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2124 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2125 			napi_complete_done(napi, frames_processed);
2126 			return frames_processed;
2127 		}
2128 
2129 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2130 			break;
2131 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2132 		rx_buff =
2133 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2134 							  rx_comp.correlator);
2135 		/* do error checking */
2136 		if (next->rx_comp.rc) {
2137 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2138 				   be16_to_cpu(next->rx_comp.rc));
2139 			/* free the entry */
2140 			next->rx_comp.first = 0;
2141 			dev_kfree_skb_any(rx_buff->skb);
2142 			remove_buff_from_pool(adapter, rx_buff);
2143 			continue;
2144 		} else if (!rx_buff->skb) {
2145 			/* free the entry */
2146 			next->rx_comp.first = 0;
2147 			remove_buff_from_pool(adapter, rx_buff);
2148 			continue;
2149 		}
2150 
2151 		length = be32_to_cpu(next->rx_comp.len);
2152 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2153 		flags = next->rx_comp.flags;
2154 		skb = rx_buff->skb;
2155 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2156 					length);
2157 
2158 		/* VLAN Header has been stripped by the system firmware and
2159 		 * needs to be inserted by the driver
2160 		 */
2161 		if (adapter->rx_vlan_header_insertion &&
2162 		    (flags & IBMVNIC_VLAN_STRIPPED))
2163 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2164 					       ntohs(next->rx_comp.vlan_tci));
2165 
2166 		/* free the entry */
2167 		next->rx_comp.first = 0;
2168 		remove_buff_from_pool(adapter, rx_buff);
2169 
2170 		skb_put(skb, length);
2171 		skb->protocol = eth_type_trans(skb, netdev);
2172 		skb_record_rx_queue(skb, scrq_num);
2173 
2174 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2175 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2176 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2177 		}
2178 
2179 		length = skb->len;
2180 		napi_gro_receive(napi, skb); /* send it up */
2181 		netdev->stats.rx_packets++;
2182 		netdev->stats.rx_bytes += length;
2183 		adapter->rx_stats_buffers[scrq_num].packets++;
2184 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2185 		frames_processed++;
2186 	}
2187 
2188 	if (adapter->state != VNIC_CLOSING)
2189 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2190 
2191 	if (frames_processed < budget) {
2192 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2193 		napi_complete_done(napi, frames_processed);
2194 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2195 		    napi_reschedule(napi)) {
2196 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2197 			goto restart_poll;
2198 		}
2199 	}
2200 	return frames_processed;
2201 }
2202 
2203 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2204 {
2205 	int rc, ret;
2206 
2207 	adapter->fallback.mtu = adapter->req_mtu;
2208 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2209 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2210 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2211 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2212 
2213 	init_completion(&adapter->reset_done);
2214 	adapter->wait_for_reset = true;
2215 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2216 	if (rc)
2217 		return rc;
2218 	wait_for_completion(&adapter->reset_done);
2219 
2220 	ret = 0;
2221 	if (adapter->reset_done_rc) {
2222 		ret = -EIO;
2223 		adapter->desired.mtu = adapter->fallback.mtu;
2224 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2225 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2226 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2227 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2228 
2229 		init_completion(&adapter->reset_done);
2230 		adapter->wait_for_reset = true;
2231 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2232 		if (rc)
2233 			return ret;
2234 		wait_for_completion(&adapter->reset_done);
2235 	}
2236 	adapter->wait_for_reset = false;
2237 
2238 	return ret;
2239 }
2240 
2241 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2242 {
2243 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2244 
2245 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2246 
2247 	return wait_for_reset(adapter);
2248 }
2249 
2250 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2251 						struct net_device *dev,
2252 						netdev_features_t features)
2253 {
2254 	/* Some backing hardware adapters can not
2255 	 * handle packets with a MSS less than 224
2256 	 * or with only one segment.
2257 	 */
2258 	if (skb_is_gso(skb)) {
2259 		if (skb_shinfo(skb)->gso_size < 224 ||
2260 		    skb_shinfo(skb)->gso_segs == 1)
2261 			features &= ~NETIF_F_GSO_MASK;
2262 	}
2263 
2264 	return features;
2265 }
2266 
2267 static const struct net_device_ops ibmvnic_netdev_ops = {
2268 	.ndo_open		= ibmvnic_open,
2269 	.ndo_stop		= ibmvnic_close,
2270 	.ndo_start_xmit		= ibmvnic_xmit,
2271 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2272 	.ndo_set_mac_address	= ibmvnic_set_mac,
2273 	.ndo_validate_addr	= eth_validate_addr,
2274 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2275 	.ndo_change_mtu		= ibmvnic_change_mtu,
2276 	.ndo_features_check     = ibmvnic_features_check,
2277 };
2278 
2279 /* ethtool functions */
2280 
2281 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2282 				      struct ethtool_link_ksettings *cmd)
2283 {
2284 	u32 supported, advertising;
2285 
2286 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2287 			  SUPPORTED_FIBRE);
2288 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2289 			    ADVERTISED_FIBRE);
2290 	cmd->base.speed = SPEED_1000;
2291 	cmd->base.duplex = DUPLEX_FULL;
2292 	cmd->base.port = PORT_FIBRE;
2293 	cmd->base.phy_address = 0;
2294 	cmd->base.autoneg = AUTONEG_ENABLE;
2295 
2296 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2297 						supported);
2298 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2299 						advertising);
2300 
2301 	return 0;
2302 }
2303 
2304 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2305 				struct ethtool_drvinfo *info)
2306 {
2307 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2308 
2309 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2310 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2311 	strlcpy(info->fw_version, adapter->fw_version,
2312 		sizeof(info->fw_version));
2313 }
2314 
2315 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2316 {
2317 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2318 
2319 	return adapter->msg_enable;
2320 }
2321 
2322 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2323 {
2324 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2325 
2326 	adapter->msg_enable = data;
2327 }
2328 
2329 static u32 ibmvnic_get_link(struct net_device *netdev)
2330 {
2331 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2332 
2333 	/* Don't need to send a query because we request a logical link up at
2334 	 * init and then we wait for link state indications
2335 	 */
2336 	return adapter->logical_link_state;
2337 }
2338 
2339 static void ibmvnic_get_ringparam(struct net_device *netdev,
2340 				  struct ethtool_ringparam *ring)
2341 {
2342 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2343 
2344 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2345 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2346 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2347 	} else {
2348 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2349 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2350 	}
2351 	ring->rx_mini_max_pending = 0;
2352 	ring->rx_jumbo_max_pending = 0;
2353 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2354 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2355 	ring->rx_mini_pending = 0;
2356 	ring->rx_jumbo_pending = 0;
2357 }
2358 
2359 static int ibmvnic_set_ringparam(struct net_device *netdev,
2360 				 struct ethtool_ringparam *ring)
2361 {
2362 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2363 	int ret;
2364 
2365 	ret = 0;
2366 	adapter->desired.rx_entries = ring->rx_pending;
2367 	adapter->desired.tx_entries = ring->tx_pending;
2368 
2369 	ret = wait_for_reset(adapter);
2370 
2371 	if (!ret &&
2372 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2373 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2374 		netdev_info(netdev,
2375 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2376 			    ring->rx_pending, ring->tx_pending,
2377 			    adapter->req_rx_add_entries_per_subcrq,
2378 			    adapter->req_tx_entries_per_subcrq);
2379 	return ret;
2380 }
2381 
2382 static void ibmvnic_get_channels(struct net_device *netdev,
2383 				 struct ethtool_channels *channels)
2384 {
2385 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2386 
2387 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2388 		channels->max_rx = adapter->max_rx_queues;
2389 		channels->max_tx = adapter->max_tx_queues;
2390 	} else {
2391 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2392 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2393 	}
2394 
2395 	channels->max_other = 0;
2396 	channels->max_combined = 0;
2397 	channels->rx_count = adapter->req_rx_queues;
2398 	channels->tx_count = adapter->req_tx_queues;
2399 	channels->other_count = 0;
2400 	channels->combined_count = 0;
2401 }
2402 
2403 static int ibmvnic_set_channels(struct net_device *netdev,
2404 				struct ethtool_channels *channels)
2405 {
2406 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2407 	int ret;
2408 
2409 	ret = 0;
2410 	adapter->desired.rx_queues = channels->rx_count;
2411 	adapter->desired.tx_queues = channels->tx_count;
2412 
2413 	ret = wait_for_reset(adapter);
2414 
2415 	if (!ret &&
2416 	    (adapter->req_rx_queues != channels->rx_count ||
2417 	     adapter->req_tx_queues != channels->tx_count))
2418 		netdev_info(netdev,
2419 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2420 			    channels->rx_count, channels->tx_count,
2421 			    adapter->req_rx_queues, adapter->req_tx_queues);
2422 	return ret;
2423 
2424 }
2425 
2426 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2427 {
2428 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2429 	int i;
2430 
2431 	switch (stringset) {
2432 	case ETH_SS_STATS:
2433 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2434 				i++, data += ETH_GSTRING_LEN)
2435 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2436 
2437 		for (i = 0; i < adapter->req_tx_queues; i++) {
2438 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2439 			data += ETH_GSTRING_LEN;
2440 
2441 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2442 			data += ETH_GSTRING_LEN;
2443 
2444 			snprintf(data, ETH_GSTRING_LEN,
2445 				 "tx%d_dropped_packets", i);
2446 			data += ETH_GSTRING_LEN;
2447 		}
2448 
2449 		for (i = 0; i < adapter->req_rx_queues; i++) {
2450 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2451 			data += ETH_GSTRING_LEN;
2452 
2453 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2454 			data += ETH_GSTRING_LEN;
2455 
2456 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2457 			data += ETH_GSTRING_LEN;
2458 		}
2459 		break;
2460 
2461 	case ETH_SS_PRIV_FLAGS:
2462 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2463 			strcpy(data + i * ETH_GSTRING_LEN,
2464 			       ibmvnic_priv_flags[i]);
2465 		break;
2466 	default:
2467 		return;
2468 	}
2469 }
2470 
2471 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2472 {
2473 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2474 
2475 	switch (sset) {
2476 	case ETH_SS_STATS:
2477 		return ARRAY_SIZE(ibmvnic_stats) +
2478 		       adapter->req_tx_queues * NUM_TX_STATS +
2479 		       adapter->req_rx_queues * NUM_RX_STATS;
2480 	case ETH_SS_PRIV_FLAGS:
2481 		return ARRAY_SIZE(ibmvnic_priv_flags);
2482 	default:
2483 		return -EOPNOTSUPP;
2484 	}
2485 }
2486 
2487 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2488 				      struct ethtool_stats *stats, u64 *data)
2489 {
2490 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2491 	union ibmvnic_crq crq;
2492 	int i, j;
2493 	int rc;
2494 
2495 	memset(&crq, 0, sizeof(crq));
2496 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2497 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2498 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2499 	crq.request_statistics.len =
2500 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2501 
2502 	/* Wait for data to be written */
2503 	init_completion(&adapter->stats_done);
2504 	rc = ibmvnic_send_crq(adapter, &crq);
2505 	if (rc)
2506 		return;
2507 	wait_for_completion(&adapter->stats_done);
2508 
2509 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2510 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2511 						ibmvnic_stats[i].offset));
2512 
2513 	for (j = 0; j < adapter->req_tx_queues; j++) {
2514 		data[i] = adapter->tx_stats_buffers[j].packets;
2515 		i++;
2516 		data[i] = adapter->tx_stats_buffers[j].bytes;
2517 		i++;
2518 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2519 		i++;
2520 	}
2521 
2522 	for (j = 0; j < adapter->req_rx_queues; j++) {
2523 		data[i] = adapter->rx_stats_buffers[j].packets;
2524 		i++;
2525 		data[i] = adapter->rx_stats_buffers[j].bytes;
2526 		i++;
2527 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2528 		i++;
2529 	}
2530 }
2531 
2532 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2533 {
2534 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2535 
2536 	return adapter->priv_flags;
2537 }
2538 
2539 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2540 {
2541 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2542 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2543 
2544 	if (which_maxes)
2545 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2546 	else
2547 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2548 
2549 	return 0;
2550 }
2551 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2552 	.get_drvinfo		= ibmvnic_get_drvinfo,
2553 	.get_msglevel		= ibmvnic_get_msglevel,
2554 	.set_msglevel		= ibmvnic_set_msglevel,
2555 	.get_link		= ibmvnic_get_link,
2556 	.get_ringparam		= ibmvnic_get_ringparam,
2557 	.set_ringparam		= ibmvnic_set_ringparam,
2558 	.get_channels		= ibmvnic_get_channels,
2559 	.set_channels		= ibmvnic_set_channels,
2560 	.get_strings            = ibmvnic_get_strings,
2561 	.get_sset_count         = ibmvnic_get_sset_count,
2562 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2563 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2564 	.get_priv_flags		= ibmvnic_get_priv_flags,
2565 	.set_priv_flags		= ibmvnic_set_priv_flags,
2566 };
2567 
2568 /* Routines for managing CRQs/sCRQs  */
2569 
2570 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2571 				   struct ibmvnic_sub_crq_queue *scrq)
2572 {
2573 	int rc;
2574 
2575 	if (scrq->irq) {
2576 		free_irq(scrq->irq, scrq);
2577 		irq_dispose_mapping(scrq->irq);
2578 		scrq->irq = 0;
2579 	}
2580 
2581 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2582 	atomic_set(&scrq->used, 0);
2583 	scrq->cur = 0;
2584 
2585 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2586 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2587 	return rc;
2588 }
2589 
2590 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2591 {
2592 	int i, rc;
2593 
2594 	for (i = 0; i < adapter->req_tx_queues; i++) {
2595 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2596 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2597 		if (rc)
2598 			return rc;
2599 	}
2600 
2601 	for (i = 0; i < adapter->req_rx_queues; i++) {
2602 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2603 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2604 		if (rc)
2605 			return rc;
2606 	}
2607 
2608 	return rc;
2609 }
2610 
2611 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2612 				  struct ibmvnic_sub_crq_queue *scrq,
2613 				  bool do_h_free)
2614 {
2615 	struct device *dev = &adapter->vdev->dev;
2616 	long rc;
2617 
2618 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2619 
2620 	if (do_h_free) {
2621 		/* Close the sub-crqs */
2622 		do {
2623 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2624 						adapter->vdev->unit_address,
2625 						scrq->crq_num);
2626 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2627 
2628 		if (rc) {
2629 			netdev_err(adapter->netdev,
2630 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2631 				   scrq->crq_num, rc);
2632 		}
2633 	}
2634 
2635 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2636 			 DMA_BIDIRECTIONAL);
2637 	free_pages((unsigned long)scrq->msgs, 2);
2638 	kfree(scrq);
2639 }
2640 
2641 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2642 							*adapter)
2643 {
2644 	struct device *dev = &adapter->vdev->dev;
2645 	struct ibmvnic_sub_crq_queue *scrq;
2646 	int rc;
2647 
2648 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2649 	if (!scrq)
2650 		return NULL;
2651 
2652 	scrq->msgs =
2653 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2654 	if (!scrq->msgs) {
2655 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2656 		goto zero_page_failed;
2657 	}
2658 
2659 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2660 					 DMA_BIDIRECTIONAL);
2661 	if (dma_mapping_error(dev, scrq->msg_token)) {
2662 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2663 		goto map_failed;
2664 	}
2665 
2666 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2667 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2668 
2669 	if (rc == H_RESOURCE)
2670 		rc = ibmvnic_reset_crq(adapter);
2671 
2672 	if (rc == H_CLOSED) {
2673 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2674 	} else if (rc) {
2675 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2676 		goto reg_failed;
2677 	}
2678 
2679 	scrq->adapter = adapter;
2680 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2681 	spin_lock_init(&scrq->lock);
2682 
2683 	netdev_dbg(adapter->netdev,
2684 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2685 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2686 
2687 	return scrq;
2688 
2689 reg_failed:
2690 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2691 			 DMA_BIDIRECTIONAL);
2692 map_failed:
2693 	free_pages((unsigned long)scrq->msgs, 2);
2694 zero_page_failed:
2695 	kfree(scrq);
2696 
2697 	return NULL;
2698 }
2699 
2700 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2701 {
2702 	int i;
2703 
2704 	if (adapter->tx_scrq) {
2705 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2706 			if (!adapter->tx_scrq[i])
2707 				continue;
2708 
2709 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2710 				   i);
2711 			if (adapter->tx_scrq[i]->irq) {
2712 				free_irq(adapter->tx_scrq[i]->irq,
2713 					 adapter->tx_scrq[i]);
2714 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2715 				adapter->tx_scrq[i]->irq = 0;
2716 			}
2717 
2718 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2719 					      do_h_free);
2720 		}
2721 
2722 		kfree(adapter->tx_scrq);
2723 		adapter->tx_scrq = NULL;
2724 		adapter->num_active_tx_scrqs = 0;
2725 	}
2726 
2727 	if (adapter->rx_scrq) {
2728 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2729 			if (!adapter->rx_scrq[i])
2730 				continue;
2731 
2732 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2733 				   i);
2734 			if (adapter->rx_scrq[i]->irq) {
2735 				free_irq(adapter->rx_scrq[i]->irq,
2736 					 adapter->rx_scrq[i]);
2737 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2738 				adapter->rx_scrq[i]->irq = 0;
2739 			}
2740 
2741 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2742 					      do_h_free);
2743 		}
2744 
2745 		kfree(adapter->rx_scrq);
2746 		adapter->rx_scrq = NULL;
2747 		adapter->num_active_rx_scrqs = 0;
2748 	}
2749 }
2750 
2751 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2752 			    struct ibmvnic_sub_crq_queue *scrq)
2753 {
2754 	struct device *dev = &adapter->vdev->dev;
2755 	unsigned long rc;
2756 
2757 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2758 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2759 	if (rc)
2760 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2761 			scrq->hw_irq, rc);
2762 	return rc;
2763 }
2764 
2765 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2766 			   struct ibmvnic_sub_crq_queue *scrq)
2767 {
2768 	struct device *dev = &adapter->vdev->dev;
2769 	unsigned long rc;
2770 
2771 	if (scrq->hw_irq > 0x100000000ULL) {
2772 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2773 		return 1;
2774 	}
2775 
2776 	if (adapter->resetting &&
2777 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
2778 		u64 val = (0xff000000) | scrq->hw_irq;
2779 
2780 		rc = plpar_hcall_norets(H_EOI, val);
2781 		if (rc)
2782 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2783 				val, rc);
2784 	}
2785 
2786 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2787 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2788 	if (rc)
2789 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2790 			scrq->hw_irq, rc);
2791 	return rc;
2792 }
2793 
2794 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2795 			       struct ibmvnic_sub_crq_queue *scrq)
2796 {
2797 	struct device *dev = &adapter->vdev->dev;
2798 	struct ibmvnic_tx_pool *tx_pool;
2799 	struct ibmvnic_tx_buff *txbuff;
2800 	union sub_crq *next;
2801 	int index;
2802 	int i, j;
2803 	u8 *first;
2804 
2805 restart_loop:
2806 	while (pending_scrq(adapter, scrq)) {
2807 		unsigned int pool = scrq->pool_index;
2808 		int num_entries = 0;
2809 
2810 		next = ibmvnic_next_scrq(adapter, scrq);
2811 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2812 			if (next->tx_comp.rcs[i]) {
2813 				dev_err(dev, "tx error %x\n",
2814 					next->tx_comp.rcs[i]);
2815 				continue;
2816 			}
2817 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2818 			if (index & IBMVNIC_TSO_POOL_MASK) {
2819 				tx_pool = &adapter->tso_pool[pool];
2820 				index &= ~IBMVNIC_TSO_POOL_MASK;
2821 			} else {
2822 				tx_pool = &adapter->tx_pool[pool];
2823 			}
2824 
2825 			txbuff = &tx_pool->tx_buff[index];
2826 
2827 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2828 				if (!txbuff->data_dma[j])
2829 					continue;
2830 
2831 				txbuff->data_dma[j] = 0;
2832 			}
2833 			/* if sub_crq was sent indirectly */
2834 			first = &txbuff->indir_arr[0].generic.first;
2835 			if (*first == IBMVNIC_CRQ_CMD) {
2836 				dma_unmap_single(dev, txbuff->indir_dma,
2837 						 sizeof(txbuff->indir_arr),
2838 						 DMA_TO_DEVICE);
2839 				*first = 0;
2840 			}
2841 
2842 			if (txbuff->last_frag) {
2843 				dev_kfree_skb_any(txbuff->skb);
2844 				txbuff->skb = NULL;
2845 			}
2846 
2847 			num_entries += txbuff->num_entries;
2848 
2849 			tx_pool->free_map[tx_pool->producer_index] = index;
2850 			tx_pool->producer_index =
2851 				(tx_pool->producer_index + 1) %
2852 					tx_pool->num_buffers;
2853 		}
2854 		/* remove tx_comp scrq*/
2855 		next->tx_comp.first = 0;
2856 
2857 		if (atomic_sub_return(num_entries, &scrq->used) <=
2858 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2859 		    __netif_subqueue_stopped(adapter->netdev,
2860 					     scrq->pool_index)) {
2861 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2862 			netdev_dbg(adapter->netdev, "Started queue %d\n",
2863 				   scrq->pool_index);
2864 		}
2865 	}
2866 
2867 	enable_scrq_irq(adapter, scrq);
2868 
2869 	if (pending_scrq(adapter, scrq)) {
2870 		disable_scrq_irq(adapter, scrq);
2871 		goto restart_loop;
2872 	}
2873 
2874 	return 0;
2875 }
2876 
2877 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2878 {
2879 	struct ibmvnic_sub_crq_queue *scrq = instance;
2880 	struct ibmvnic_adapter *adapter = scrq->adapter;
2881 
2882 	disable_scrq_irq(adapter, scrq);
2883 	ibmvnic_complete_tx(adapter, scrq);
2884 
2885 	return IRQ_HANDLED;
2886 }
2887 
2888 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2889 {
2890 	struct ibmvnic_sub_crq_queue *scrq = instance;
2891 	struct ibmvnic_adapter *adapter = scrq->adapter;
2892 
2893 	/* When booting a kdump kernel we can hit pending interrupts
2894 	 * prior to completing driver initialization.
2895 	 */
2896 	if (unlikely(adapter->state != VNIC_OPEN))
2897 		return IRQ_NONE;
2898 
2899 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2900 
2901 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2902 		disable_scrq_irq(adapter, scrq);
2903 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2904 	}
2905 
2906 	return IRQ_HANDLED;
2907 }
2908 
2909 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2910 {
2911 	struct device *dev = &adapter->vdev->dev;
2912 	struct ibmvnic_sub_crq_queue *scrq;
2913 	int i = 0, j = 0;
2914 	int rc = 0;
2915 
2916 	for (i = 0; i < adapter->req_tx_queues; i++) {
2917 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2918 			   i);
2919 		scrq = adapter->tx_scrq[i];
2920 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2921 
2922 		if (!scrq->irq) {
2923 			rc = -EINVAL;
2924 			dev_err(dev, "Error mapping irq\n");
2925 			goto req_tx_irq_failed;
2926 		}
2927 
2928 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2929 				 0, "ibmvnic_tx", scrq);
2930 
2931 		if (rc) {
2932 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2933 				scrq->irq, rc);
2934 			irq_dispose_mapping(scrq->irq);
2935 			goto req_tx_irq_failed;
2936 		}
2937 	}
2938 
2939 	for (i = 0; i < adapter->req_rx_queues; i++) {
2940 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2941 			   i);
2942 		scrq = adapter->rx_scrq[i];
2943 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2944 		if (!scrq->irq) {
2945 			rc = -EINVAL;
2946 			dev_err(dev, "Error mapping irq\n");
2947 			goto req_rx_irq_failed;
2948 		}
2949 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2950 				 0, "ibmvnic_rx", scrq);
2951 		if (rc) {
2952 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2953 				scrq->irq, rc);
2954 			irq_dispose_mapping(scrq->irq);
2955 			goto req_rx_irq_failed;
2956 		}
2957 	}
2958 	return rc;
2959 
2960 req_rx_irq_failed:
2961 	for (j = 0; j < i; j++) {
2962 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2963 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2964 	}
2965 	i = adapter->req_tx_queues;
2966 req_tx_irq_failed:
2967 	for (j = 0; j < i; j++) {
2968 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2969 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2970 	}
2971 	release_sub_crqs(adapter, 1);
2972 	return rc;
2973 }
2974 
2975 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2976 {
2977 	struct device *dev = &adapter->vdev->dev;
2978 	struct ibmvnic_sub_crq_queue **allqueues;
2979 	int registered_queues = 0;
2980 	int total_queues;
2981 	int more = 0;
2982 	int i;
2983 
2984 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2985 
2986 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2987 	if (!allqueues)
2988 		return -1;
2989 
2990 	for (i = 0; i < total_queues; i++) {
2991 		allqueues[i] = init_sub_crq_queue(adapter);
2992 		if (!allqueues[i]) {
2993 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2994 			break;
2995 		}
2996 		registered_queues++;
2997 	}
2998 
2999 	/* Make sure we were able to register the minimum number of queues */
3000 	if (registered_queues <
3001 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3002 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3003 		goto tx_failed;
3004 	}
3005 
3006 	/* Distribute the failed allocated queues*/
3007 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3008 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3009 		switch (i % 3) {
3010 		case 0:
3011 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3012 				adapter->req_rx_queues--;
3013 			else
3014 				more++;
3015 			break;
3016 		case 1:
3017 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3018 				adapter->req_tx_queues--;
3019 			else
3020 				more++;
3021 			break;
3022 		}
3023 	}
3024 
3025 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3026 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3027 	if (!adapter->tx_scrq)
3028 		goto tx_failed;
3029 
3030 	for (i = 0; i < adapter->req_tx_queues; i++) {
3031 		adapter->tx_scrq[i] = allqueues[i];
3032 		adapter->tx_scrq[i]->pool_index = i;
3033 		adapter->num_active_tx_scrqs++;
3034 	}
3035 
3036 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3037 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3038 	if (!adapter->rx_scrq)
3039 		goto rx_failed;
3040 
3041 	for (i = 0; i < adapter->req_rx_queues; i++) {
3042 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3043 		adapter->rx_scrq[i]->scrq_num = i;
3044 		adapter->num_active_rx_scrqs++;
3045 	}
3046 
3047 	kfree(allqueues);
3048 	return 0;
3049 
3050 rx_failed:
3051 	kfree(adapter->tx_scrq);
3052 	adapter->tx_scrq = NULL;
3053 tx_failed:
3054 	for (i = 0; i < registered_queues; i++)
3055 		release_sub_crq_queue(adapter, allqueues[i], 1);
3056 	kfree(allqueues);
3057 	return -1;
3058 }
3059 
3060 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3061 {
3062 	struct device *dev = &adapter->vdev->dev;
3063 	union ibmvnic_crq crq;
3064 	int max_entries;
3065 
3066 	if (!retry) {
3067 		/* Sub-CRQ entries are 32 byte long */
3068 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3069 
3070 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3071 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3072 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3073 			return;
3074 		}
3075 
3076 		if (adapter->desired.mtu)
3077 			adapter->req_mtu = adapter->desired.mtu;
3078 		else
3079 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3080 
3081 		if (!adapter->desired.tx_entries)
3082 			adapter->desired.tx_entries =
3083 					adapter->max_tx_entries_per_subcrq;
3084 		if (!adapter->desired.rx_entries)
3085 			adapter->desired.rx_entries =
3086 					adapter->max_rx_add_entries_per_subcrq;
3087 
3088 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3089 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3090 
3091 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3092 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3093 			adapter->desired.tx_entries = max_entries;
3094 		}
3095 
3096 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3097 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3098 			adapter->desired.rx_entries = max_entries;
3099 		}
3100 
3101 		if (adapter->desired.tx_entries)
3102 			adapter->req_tx_entries_per_subcrq =
3103 					adapter->desired.tx_entries;
3104 		else
3105 			adapter->req_tx_entries_per_subcrq =
3106 					adapter->max_tx_entries_per_subcrq;
3107 
3108 		if (adapter->desired.rx_entries)
3109 			adapter->req_rx_add_entries_per_subcrq =
3110 					adapter->desired.rx_entries;
3111 		else
3112 			adapter->req_rx_add_entries_per_subcrq =
3113 					adapter->max_rx_add_entries_per_subcrq;
3114 
3115 		if (adapter->desired.tx_queues)
3116 			adapter->req_tx_queues =
3117 					adapter->desired.tx_queues;
3118 		else
3119 			adapter->req_tx_queues =
3120 					adapter->opt_tx_comp_sub_queues;
3121 
3122 		if (adapter->desired.rx_queues)
3123 			adapter->req_rx_queues =
3124 					adapter->desired.rx_queues;
3125 		else
3126 			adapter->req_rx_queues =
3127 					adapter->opt_rx_comp_queues;
3128 
3129 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3130 	}
3131 
3132 	memset(&crq, 0, sizeof(crq));
3133 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3134 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3135 
3136 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3137 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3138 	atomic_inc(&adapter->running_cap_crqs);
3139 	ibmvnic_send_crq(adapter, &crq);
3140 
3141 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3142 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3143 	atomic_inc(&adapter->running_cap_crqs);
3144 	ibmvnic_send_crq(adapter, &crq);
3145 
3146 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3147 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3148 	atomic_inc(&adapter->running_cap_crqs);
3149 	ibmvnic_send_crq(adapter, &crq);
3150 
3151 	crq.request_capability.capability =
3152 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3153 	crq.request_capability.number =
3154 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3155 	atomic_inc(&adapter->running_cap_crqs);
3156 	ibmvnic_send_crq(adapter, &crq);
3157 
3158 	crq.request_capability.capability =
3159 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3160 	crq.request_capability.number =
3161 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3162 	atomic_inc(&adapter->running_cap_crqs);
3163 	ibmvnic_send_crq(adapter, &crq);
3164 
3165 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3166 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3167 	atomic_inc(&adapter->running_cap_crqs);
3168 	ibmvnic_send_crq(adapter, &crq);
3169 
3170 	if (adapter->netdev->flags & IFF_PROMISC) {
3171 		if (adapter->promisc_supported) {
3172 			crq.request_capability.capability =
3173 			    cpu_to_be16(PROMISC_REQUESTED);
3174 			crq.request_capability.number = cpu_to_be64(1);
3175 			atomic_inc(&adapter->running_cap_crqs);
3176 			ibmvnic_send_crq(adapter, &crq);
3177 		}
3178 	} else {
3179 		crq.request_capability.capability =
3180 		    cpu_to_be16(PROMISC_REQUESTED);
3181 		crq.request_capability.number = cpu_to_be64(0);
3182 		atomic_inc(&adapter->running_cap_crqs);
3183 		ibmvnic_send_crq(adapter, &crq);
3184 	}
3185 }
3186 
3187 static int pending_scrq(struct ibmvnic_adapter *adapter,
3188 			struct ibmvnic_sub_crq_queue *scrq)
3189 {
3190 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3191 
3192 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3193 		return 1;
3194 	else
3195 		return 0;
3196 }
3197 
3198 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3199 					struct ibmvnic_sub_crq_queue *scrq)
3200 {
3201 	union sub_crq *entry;
3202 	unsigned long flags;
3203 
3204 	spin_lock_irqsave(&scrq->lock, flags);
3205 	entry = &scrq->msgs[scrq->cur];
3206 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3207 		if (++scrq->cur == scrq->size)
3208 			scrq->cur = 0;
3209 	} else {
3210 		entry = NULL;
3211 	}
3212 	spin_unlock_irqrestore(&scrq->lock, flags);
3213 
3214 	return entry;
3215 }
3216 
3217 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3218 {
3219 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3220 	union ibmvnic_crq *crq;
3221 
3222 	crq = &queue->msgs[queue->cur];
3223 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3224 		if (++queue->cur == queue->size)
3225 			queue->cur = 0;
3226 	} else {
3227 		crq = NULL;
3228 	}
3229 
3230 	return crq;
3231 }
3232 
3233 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3234 {
3235 	switch (rc) {
3236 	case H_PARAMETER:
3237 		dev_warn_ratelimited(dev,
3238 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3239 				     func, rc);
3240 		break;
3241 	case H_CLOSED:
3242 		dev_warn_ratelimited(dev,
3243 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3244 				     func, rc);
3245 		break;
3246 	default:
3247 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3248 		break;
3249 	}
3250 }
3251 
3252 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3253 		       union sub_crq *sub_crq)
3254 {
3255 	unsigned int ua = adapter->vdev->unit_address;
3256 	struct device *dev = &adapter->vdev->dev;
3257 	u64 *u64_crq = (u64 *)sub_crq;
3258 	int rc;
3259 
3260 	netdev_dbg(adapter->netdev,
3261 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3262 		   (unsigned long int)cpu_to_be64(remote_handle),
3263 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3264 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3265 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3266 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3267 
3268 	/* Make sure the hypervisor sees the complete request */
3269 	mb();
3270 
3271 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3272 				cpu_to_be64(remote_handle),
3273 				cpu_to_be64(u64_crq[0]),
3274 				cpu_to_be64(u64_crq[1]),
3275 				cpu_to_be64(u64_crq[2]),
3276 				cpu_to_be64(u64_crq[3]));
3277 
3278 	if (rc)
3279 		print_subcrq_error(dev, rc, __func__);
3280 
3281 	return rc;
3282 }
3283 
3284 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3285 				u64 remote_handle, u64 ioba, u64 num_entries)
3286 {
3287 	unsigned int ua = adapter->vdev->unit_address;
3288 	struct device *dev = &adapter->vdev->dev;
3289 	int rc;
3290 
3291 	/* Make sure the hypervisor sees the complete request */
3292 	mb();
3293 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3294 				cpu_to_be64(remote_handle),
3295 				ioba, num_entries);
3296 
3297 	if (rc)
3298 		print_subcrq_error(dev, rc, __func__);
3299 
3300 	return rc;
3301 }
3302 
3303 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3304 			    union ibmvnic_crq *crq)
3305 {
3306 	unsigned int ua = adapter->vdev->unit_address;
3307 	struct device *dev = &adapter->vdev->dev;
3308 	u64 *u64_crq = (u64 *)crq;
3309 	int rc;
3310 
3311 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3312 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3313 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3314 
3315 	if (!adapter->crq.active &&
3316 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3317 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3318 		return -EINVAL;
3319 	}
3320 
3321 	/* Make sure the hypervisor sees the complete request */
3322 	mb();
3323 
3324 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3325 				cpu_to_be64(u64_crq[0]),
3326 				cpu_to_be64(u64_crq[1]));
3327 
3328 	if (rc) {
3329 		if (rc == H_CLOSED) {
3330 			dev_warn(dev, "CRQ Queue closed\n");
3331 			if (adapter->resetting)
3332 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3333 		}
3334 
3335 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3336 	}
3337 
3338 	return rc;
3339 }
3340 
3341 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3342 {
3343 	union ibmvnic_crq crq;
3344 
3345 	memset(&crq, 0, sizeof(crq));
3346 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3347 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3348 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3349 
3350 	return ibmvnic_send_crq(adapter, &crq);
3351 }
3352 
3353 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3354 {
3355 	union ibmvnic_crq crq;
3356 
3357 	memset(&crq, 0, sizeof(crq));
3358 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3359 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3360 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3361 
3362 	return ibmvnic_send_crq(adapter, &crq);
3363 }
3364 
3365 struct vnic_login_client_data {
3366 	u8	type;
3367 	__be16	len;
3368 	char	name[];
3369 } __packed;
3370 
3371 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3372 {
3373 	int len;
3374 
3375 	/* Calculate the amount of buffer space needed for the
3376 	 * vnic client data in the login buffer. There are four entries,
3377 	 * OS name, LPAR name, device name, and a null last entry.
3378 	 */
3379 	len = 4 * sizeof(struct vnic_login_client_data);
3380 	len += 6; /* "Linux" plus NULL */
3381 	len += strlen(utsname()->nodename) + 1;
3382 	len += strlen(adapter->netdev->name) + 1;
3383 
3384 	return len;
3385 }
3386 
3387 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3388 				 struct vnic_login_client_data *vlcd)
3389 {
3390 	const char *os_name = "Linux";
3391 	int len;
3392 
3393 	/* Type 1 - LPAR OS */
3394 	vlcd->type = 1;
3395 	len = strlen(os_name) + 1;
3396 	vlcd->len = cpu_to_be16(len);
3397 	strncpy(vlcd->name, os_name, len);
3398 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3399 
3400 	/* Type 2 - LPAR name */
3401 	vlcd->type = 2;
3402 	len = strlen(utsname()->nodename) + 1;
3403 	vlcd->len = cpu_to_be16(len);
3404 	strncpy(vlcd->name, utsname()->nodename, len);
3405 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3406 
3407 	/* Type 3 - device name */
3408 	vlcd->type = 3;
3409 	len = strlen(adapter->netdev->name) + 1;
3410 	vlcd->len = cpu_to_be16(len);
3411 	strncpy(vlcd->name, adapter->netdev->name, len);
3412 }
3413 
3414 static int send_login(struct ibmvnic_adapter *adapter)
3415 {
3416 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3417 	struct ibmvnic_login_buffer *login_buffer;
3418 	struct device *dev = &adapter->vdev->dev;
3419 	dma_addr_t rsp_buffer_token;
3420 	dma_addr_t buffer_token;
3421 	size_t rsp_buffer_size;
3422 	union ibmvnic_crq crq;
3423 	size_t buffer_size;
3424 	__be64 *tx_list_p;
3425 	__be64 *rx_list_p;
3426 	int client_data_len;
3427 	struct vnic_login_client_data *vlcd;
3428 	int i;
3429 
3430 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3431 		netdev_err(adapter->netdev,
3432 			   "RX or TX queues are not allocated, device login failed\n");
3433 		return -1;
3434 	}
3435 
3436 	release_login_rsp_buffer(adapter);
3437 	client_data_len = vnic_client_data_len(adapter);
3438 
3439 	buffer_size =
3440 	    sizeof(struct ibmvnic_login_buffer) +
3441 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3442 	    client_data_len;
3443 
3444 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3445 	if (!login_buffer)
3446 		goto buf_alloc_failed;
3447 
3448 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3449 				      DMA_TO_DEVICE);
3450 	if (dma_mapping_error(dev, buffer_token)) {
3451 		dev_err(dev, "Couldn't map login buffer\n");
3452 		goto buf_map_failed;
3453 	}
3454 
3455 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3456 			  sizeof(u64) * adapter->req_tx_queues +
3457 			  sizeof(u64) * adapter->req_rx_queues +
3458 			  sizeof(u64) * adapter->req_rx_queues +
3459 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3460 
3461 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3462 	if (!login_rsp_buffer)
3463 		goto buf_rsp_alloc_failed;
3464 
3465 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3466 					  rsp_buffer_size, DMA_FROM_DEVICE);
3467 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3468 		dev_err(dev, "Couldn't map login rsp buffer\n");
3469 		goto buf_rsp_map_failed;
3470 	}
3471 
3472 	adapter->login_buf = login_buffer;
3473 	adapter->login_buf_token = buffer_token;
3474 	adapter->login_buf_sz = buffer_size;
3475 	adapter->login_rsp_buf = login_rsp_buffer;
3476 	adapter->login_rsp_buf_token = rsp_buffer_token;
3477 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3478 
3479 	login_buffer->len = cpu_to_be32(buffer_size);
3480 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3481 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3482 	login_buffer->off_txcomp_subcrqs =
3483 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3484 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3485 	login_buffer->off_rxcomp_subcrqs =
3486 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3487 			sizeof(u64) * adapter->req_tx_queues);
3488 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3489 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3490 
3491 	tx_list_p = (__be64 *)((char *)login_buffer +
3492 				      sizeof(struct ibmvnic_login_buffer));
3493 	rx_list_p = (__be64 *)((char *)login_buffer +
3494 				      sizeof(struct ibmvnic_login_buffer) +
3495 				      sizeof(u64) * adapter->req_tx_queues);
3496 
3497 	for (i = 0; i < adapter->req_tx_queues; i++) {
3498 		if (adapter->tx_scrq[i]) {
3499 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3500 						   crq_num);
3501 		}
3502 	}
3503 
3504 	for (i = 0; i < adapter->req_rx_queues; i++) {
3505 		if (adapter->rx_scrq[i]) {
3506 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3507 						   crq_num);
3508 		}
3509 	}
3510 
3511 	/* Insert vNIC login client data */
3512 	vlcd = (struct vnic_login_client_data *)
3513 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3514 	login_buffer->client_data_offset =
3515 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3516 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3517 
3518 	vnic_add_client_data(adapter, vlcd);
3519 
3520 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3521 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3522 		netdev_dbg(adapter->netdev, "%016lx\n",
3523 			   ((unsigned long int *)(adapter->login_buf))[i]);
3524 	}
3525 
3526 	memset(&crq, 0, sizeof(crq));
3527 	crq.login.first = IBMVNIC_CRQ_CMD;
3528 	crq.login.cmd = LOGIN;
3529 	crq.login.ioba = cpu_to_be32(buffer_token);
3530 	crq.login.len = cpu_to_be32(buffer_size);
3531 	ibmvnic_send_crq(adapter, &crq);
3532 
3533 	return 0;
3534 
3535 buf_rsp_map_failed:
3536 	kfree(login_rsp_buffer);
3537 buf_rsp_alloc_failed:
3538 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3539 buf_map_failed:
3540 	kfree(login_buffer);
3541 buf_alloc_failed:
3542 	return -1;
3543 }
3544 
3545 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3546 			    u32 len, u8 map_id)
3547 {
3548 	union ibmvnic_crq crq;
3549 
3550 	memset(&crq, 0, sizeof(crq));
3551 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3552 	crq.request_map.cmd = REQUEST_MAP;
3553 	crq.request_map.map_id = map_id;
3554 	crq.request_map.ioba = cpu_to_be32(addr);
3555 	crq.request_map.len = cpu_to_be32(len);
3556 	return ibmvnic_send_crq(adapter, &crq);
3557 }
3558 
3559 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3560 {
3561 	union ibmvnic_crq crq;
3562 
3563 	memset(&crq, 0, sizeof(crq));
3564 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3565 	crq.request_unmap.cmd = REQUEST_UNMAP;
3566 	crq.request_unmap.map_id = map_id;
3567 	return ibmvnic_send_crq(adapter, &crq);
3568 }
3569 
3570 static void send_map_query(struct ibmvnic_adapter *adapter)
3571 {
3572 	union ibmvnic_crq crq;
3573 
3574 	memset(&crq, 0, sizeof(crq));
3575 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3576 	crq.query_map.cmd = QUERY_MAP;
3577 	ibmvnic_send_crq(adapter, &crq);
3578 }
3579 
3580 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3581 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3582 {
3583 	union ibmvnic_crq crq;
3584 
3585 	atomic_set(&adapter->running_cap_crqs, 0);
3586 	memset(&crq, 0, sizeof(crq));
3587 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3588 	crq.query_capability.cmd = QUERY_CAPABILITY;
3589 
3590 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3591 	atomic_inc(&adapter->running_cap_crqs);
3592 	ibmvnic_send_crq(adapter, &crq);
3593 
3594 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3595 	atomic_inc(&adapter->running_cap_crqs);
3596 	ibmvnic_send_crq(adapter, &crq);
3597 
3598 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3599 	atomic_inc(&adapter->running_cap_crqs);
3600 	ibmvnic_send_crq(adapter, &crq);
3601 
3602 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3603 	atomic_inc(&adapter->running_cap_crqs);
3604 	ibmvnic_send_crq(adapter, &crq);
3605 
3606 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3607 	atomic_inc(&adapter->running_cap_crqs);
3608 	ibmvnic_send_crq(adapter, &crq);
3609 
3610 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3611 	atomic_inc(&adapter->running_cap_crqs);
3612 	ibmvnic_send_crq(adapter, &crq);
3613 
3614 	crq.query_capability.capability =
3615 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3616 	atomic_inc(&adapter->running_cap_crqs);
3617 	ibmvnic_send_crq(adapter, &crq);
3618 
3619 	crq.query_capability.capability =
3620 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3621 	atomic_inc(&adapter->running_cap_crqs);
3622 	ibmvnic_send_crq(adapter, &crq);
3623 
3624 	crq.query_capability.capability =
3625 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3626 	atomic_inc(&adapter->running_cap_crqs);
3627 	ibmvnic_send_crq(adapter, &crq);
3628 
3629 	crq.query_capability.capability =
3630 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3631 	atomic_inc(&adapter->running_cap_crqs);
3632 	ibmvnic_send_crq(adapter, &crq);
3633 
3634 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3635 	atomic_inc(&adapter->running_cap_crqs);
3636 	ibmvnic_send_crq(adapter, &crq);
3637 
3638 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3639 	atomic_inc(&adapter->running_cap_crqs);
3640 	ibmvnic_send_crq(adapter, &crq);
3641 
3642 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3643 	atomic_inc(&adapter->running_cap_crqs);
3644 	ibmvnic_send_crq(adapter, &crq);
3645 
3646 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3647 	atomic_inc(&adapter->running_cap_crqs);
3648 	ibmvnic_send_crq(adapter, &crq);
3649 
3650 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3651 	atomic_inc(&adapter->running_cap_crqs);
3652 	ibmvnic_send_crq(adapter, &crq);
3653 
3654 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3655 	atomic_inc(&adapter->running_cap_crqs);
3656 	ibmvnic_send_crq(adapter, &crq);
3657 
3658 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3659 	atomic_inc(&adapter->running_cap_crqs);
3660 	ibmvnic_send_crq(adapter, &crq);
3661 
3662 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3663 	atomic_inc(&adapter->running_cap_crqs);
3664 	ibmvnic_send_crq(adapter, &crq);
3665 
3666 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3667 	atomic_inc(&adapter->running_cap_crqs);
3668 	ibmvnic_send_crq(adapter, &crq);
3669 
3670 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3671 	atomic_inc(&adapter->running_cap_crqs);
3672 	ibmvnic_send_crq(adapter, &crq);
3673 
3674 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3675 	atomic_inc(&adapter->running_cap_crqs);
3676 	ibmvnic_send_crq(adapter, &crq);
3677 
3678 	crq.query_capability.capability =
3679 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3680 	atomic_inc(&adapter->running_cap_crqs);
3681 	ibmvnic_send_crq(adapter, &crq);
3682 
3683 	crq.query_capability.capability =
3684 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3685 	atomic_inc(&adapter->running_cap_crqs);
3686 	ibmvnic_send_crq(adapter, &crq);
3687 
3688 	crq.query_capability.capability =
3689 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3690 	atomic_inc(&adapter->running_cap_crqs);
3691 	ibmvnic_send_crq(adapter, &crq);
3692 
3693 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3694 	atomic_inc(&adapter->running_cap_crqs);
3695 	ibmvnic_send_crq(adapter, &crq);
3696 }
3697 
3698 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3699 				struct ibmvnic_adapter *adapter)
3700 {
3701 	struct device *dev = &adapter->vdev->dev;
3702 
3703 	if (crq->get_vpd_size_rsp.rc.code) {
3704 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3705 			crq->get_vpd_size_rsp.rc.code);
3706 		complete(&adapter->fw_done);
3707 		return;
3708 	}
3709 
3710 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3711 	complete(&adapter->fw_done);
3712 }
3713 
3714 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3715 			   struct ibmvnic_adapter *adapter)
3716 {
3717 	struct device *dev = &adapter->vdev->dev;
3718 	unsigned char *substr = NULL;
3719 	u8 fw_level_len = 0;
3720 
3721 	memset(adapter->fw_version, 0, 32);
3722 
3723 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3724 			 DMA_FROM_DEVICE);
3725 
3726 	if (crq->get_vpd_rsp.rc.code) {
3727 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3728 			crq->get_vpd_rsp.rc.code);
3729 		goto complete;
3730 	}
3731 
3732 	/* get the position of the firmware version info
3733 	 * located after the ASCII 'RM' substring in the buffer
3734 	 */
3735 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3736 	if (!substr) {
3737 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3738 		goto complete;
3739 	}
3740 
3741 	/* get length of firmware level ASCII substring */
3742 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3743 		fw_level_len = *(substr + 2);
3744 	} else {
3745 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3746 		goto complete;
3747 	}
3748 
3749 	/* copy firmware version string from vpd into adapter */
3750 	if ((substr + 3 + fw_level_len) <
3751 	    (adapter->vpd->buff + adapter->vpd->len)) {
3752 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3753 	} else {
3754 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3755 	}
3756 
3757 complete:
3758 	if (adapter->fw_version[0] == '\0')
3759 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3760 	complete(&adapter->fw_done);
3761 }
3762 
3763 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3764 {
3765 	struct device *dev = &adapter->vdev->dev;
3766 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3767 	union ibmvnic_crq crq;
3768 	int i;
3769 
3770 	dma_unmap_single(dev, adapter->ip_offload_tok,
3771 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3772 
3773 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3774 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3775 		netdev_dbg(adapter->netdev, "%016lx\n",
3776 			   ((unsigned long int *)(buf))[i]);
3777 
3778 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3779 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3780 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3781 		   buf->tcp_ipv4_chksum);
3782 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3783 		   buf->tcp_ipv6_chksum);
3784 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3785 		   buf->udp_ipv4_chksum);
3786 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3787 		   buf->udp_ipv6_chksum);
3788 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3789 		   buf->large_tx_ipv4);
3790 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3791 		   buf->large_tx_ipv6);
3792 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3793 		   buf->large_rx_ipv4);
3794 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3795 		   buf->large_rx_ipv6);
3796 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3797 		   buf->max_ipv4_header_size);
3798 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3799 		   buf->max_ipv6_header_size);
3800 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3801 		   buf->max_tcp_header_size);
3802 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3803 		   buf->max_udp_header_size);
3804 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3805 		   buf->max_large_tx_size);
3806 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3807 		   buf->max_large_rx_size);
3808 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3809 		   buf->ipv6_extension_header);
3810 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3811 		   buf->tcp_pseudosum_req);
3812 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3813 		   buf->num_ipv6_ext_headers);
3814 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3815 		   buf->off_ipv6_ext_headers);
3816 
3817 	adapter->ip_offload_ctrl_tok =
3818 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
3819 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3820 
3821 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3822 		dev_err(dev, "Couldn't map ip offload control buffer\n");
3823 		return;
3824 	}
3825 
3826 	adapter->ip_offload_ctrl.len =
3827 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3828 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3829 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3830 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3831 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3832 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3833 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3834 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3835 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3836 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3837 
3838 	/* large_rx disabled for now, additional features needed */
3839 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3840 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3841 
3842 	adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3843 
3844 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3845 		adapter->netdev->features |= NETIF_F_IP_CSUM;
3846 
3847 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3848 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3849 
3850 	if ((adapter->netdev->features &
3851 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3852 		adapter->netdev->features |= NETIF_F_RXCSUM;
3853 
3854 	if (buf->large_tx_ipv4)
3855 		adapter->netdev->features |= NETIF_F_TSO;
3856 	if (buf->large_tx_ipv6)
3857 		adapter->netdev->features |= NETIF_F_TSO6;
3858 
3859 	adapter->netdev->hw_features |= adapter->netdev->features;
3860 
3861 	memset(&crq, 0, sizeof(crq));
3862 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3863 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3864 	crq.control_ip_offload.len =
3865 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3866 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3867 	ibmvnic_send_crq(adapter, &crq);
3868 }
3869 
3870 static const char *ibmvnic_fw_err_cause(u16 cause)
3871 {
3872 	switch (cause) {
3873 	case ADAPTER_PROBLEM:
3874 		return "adapter problem";
3875 	case BUS_PROBLEM:
3876 		return "bus problem";
3877 	case FW_PROBLEM:
3878 		return "firmware problem";
3879 	case DD_PROBLEM:
3880 		return "device driver problem";
3881 	case EEH_RECOVERY:
3882 		return "EEH recovery";
3883 	case FW_UPDATED:
3884 		return "firmware updated";
3885 	case LOW_MEMORY:
3886 		return "low Memory";
3887 	default:
3888 		return "unknown";
3889 	}
3890 }
3891 
3892 static void handle_error_indication(union ibmvnic_crq *crq,
3893 				    struct ibmvnic_adapter *adapter)
3894 {
3895 	struct device *dev = &adapter->vdev->dev;
3896 	u16 cause;
3897 
3898 	cause = be16_to_cpu(crq->error_indication.error_cause);
3899 
3900 	dev_warn_ratelimited(dev,
3901 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
3902 			     crq->error_indication.flags
3903 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3904 			     ibmvnic_fw_err_cause(cause));
3905 
3906 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3907 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3908 	else
3909 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3910 }
3911 
3912 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3913 				 struct ibmvnic_adapter *adapter)
3914 {
3915 	struct net_device *netdev = adapter->netdev;
3916 	struct device *dev = &adapter->vdev->dev;
3917 	long rc;
3918 
3919 	rc = crq->change_mac_addr_rsp.rc.code;
3920 	if (rc) {
3921 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3922 		goto out;
3923 	}
3924 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3925 	       ETH_ALEN);
3926 out:
3927 	complete(&adapter->fw_done);
3928 	return rc;
3929 }
3930 
3931 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3932 				   struct ibmvnic_adapter *adapter)
3933 {
3934 	struct device *dev = &adapter->vdev->dev;
3935 	u64 *req_value;
3936 	char *name;
3937 
3938 	atomic_dec(&adapter->running_cap_crqs);
3939 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3940 	case REQ_TX_QUEUES:
3941 		req_value = &adapter->req_tx_queues;
3942 		name = "tx";
3943 		break;
3944 	case REQ_RX_QUEUES:
3945 		req_value = &adapter->req_rx_queues;
3946 		name = "rx";
3947 		break;
3948 	case REQ_RX_ADD_QUEUES:
3949 		req_value = &adapter->req_rx_add_queues;
3950 		name = "rx_add";
3951 		break;
3952 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3953 		req_value = &adapter->req_tx_entries_per_subcrq;
3954 		name = "tx_entries_per_subcrq";
3955 		break;
3956 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3957 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3958 		name = "rx_add_entries_per_subcrq";
3959 		break;
3960 	case REQ_MTU:
3961 		req_value = &adapter->req_mtu;
3962 		name = "mtu";
3963 		break;
3964 	case PROMISC_REQUESTED:
3965 		req_value = &adapter->promisc;
3966 		name = "promisc";
3967 		break;
3968 	default:
3969 		dev_err(dev, "Got invalid cap request rsp %d\n",
3970 			crq->request_capability.capability);
3971 		return;
3972 	}
3973 
3974 	switch (crq->request_capability_rsp.rc.code) {
3975 	case SUCCESS:
3976 		break;
3977 	case PARTIALSUCCESS:
3978 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3979 			 *req_value,
3980 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3981 					       number), name);
3982 
3983 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3984 		    REQ_MTU) {
3985 			pr_err("mtu of %llu is not supported. Reverting.\n",
3986 			       *req_value);
3987 			*req_value = adapter->fallback.mtu;
3988 		} else {
3989 			*req_value =
3990 				be64_to_cpu(crq->request_capability_rsp.number);
3991 		}
3992 
3993 		ibmvnic_send_req_caps(adapter, 1);
3994 		return;
3995 	default:
3996 		dev_err(dev, "Error %d in request cap rsp\n",
3997 			crq->request_capability_rsp.rc.code);
3998 		return;
3999 	}
4000 
4001 	/* Done receiving requested capabilities, query IP offload support */
4002 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4003 		union ibmvnic_crq newcrq;
4004 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4005 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4006 		    &adapter->ip_offload_buf;
4007 
4008 		adapter->wait_capability = false;
4009 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4010 							 buf_sz,
4011 							 DMA_FROM_DEVICE);
4012 
4013 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4014 			if (!firmware_has_feature(FW_FEATURE_CMO))
4015 				dev_err(dev, "Couldn't map offload buffer\n");
4016 			return;
4017 		}
4018 
4019 		memset(&newcrq, 0, sizeof(newcrq));
4020 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4021 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4022 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4023 		newcrq.query_ip_offload.ioba =
4024 		    cpu_to_be32(adapter->ip_offload_tok);
4025 
4026 		ibmvnic_send_crq(adapter, &newcrq);
4027 	}
4028 }
4029 
4030 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4031 			    struct ibmvnic_adapter *adapter)
4032 {
4033 	struct device *dev = &adapter->vdev->dev;
4034 	struct net_device *netdev = adapter->netdev;
4035 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4036 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4037 	int i;
4038 
4039 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4040 			 DMA_TO_DEVICE);
4041 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4042 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4043 
4044 	/* If the number of queues requested can't be allocated by the
4045 	 * server, the login response will return with code 1. We will need
4046 	 * to resend the login buffer with fewer queues requested.
4047 	 */
4048 	if (login_rsp_crq->generic.rc.code) {
4049 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4050 		complete(&adapter->init_done);
4051 		return 0;
4052 	}
4053 
4054 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4055 
4056 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4057 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4058 		netdev_dbg(adapter->netdev, "%016lx\n",
4059 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4060 	}
4061 
4062 	/* Sanity checks */
4063 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4064 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4065 	     adapter->req_rx_add_queues !=
4066 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4067 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4068 		ibmvnic_remove(adapter->vdev);
4069 		return -EIO;
4070 	}
4071 	release_login_buffer(adapter);
4072 	complete(&adapter->init_done);
4073 
4074 	return 0;
4075 }
4076 
4077 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4078 				     struct ibmvnic_adapter *adapter)
4079 {
4080 	struct device *dev = &adapter->vdev->dev;
4081 	long rc;
4082 
4083 	rc = crq->request_unmap_rsp.rc.code;
4084 	if (rc)
4085 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4086 }
4087 
4088 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4089 				 struct ibmvnic_adapter *adapter)
4090 {
4091 	struct net_device *netdev = adapter->netdev;
4092 	struct device *dev = &adapter->vdev->dev;
4093 	long rc;
4094 
4095 	rc = crq->query_map_rsp.rc.code;
4096 	if (rc) {
4097 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4098 		return;
4099 	}
4100 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4101 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4102 		   crq->query_map_rsp.free_pages);
4103 }
4104 
4105 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4106 				 struct ibmvnic_adapter *adapter)
4107 {
4108 	struct net_device *netdev = adapter->netdev;
4109 	struct device *dev = &adapter->vdev->dev;
4110 	long rc;
4111 
4112 	atomic_dec(&adapter->running_cap_crqs);
4113 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4114 		   atomic_read(&adapter->running_cap_crqs));
4115 	rc = crq->query_capability.rc.code;
4116 	if (rc) {
4117 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4118 		goto out;
4119 	}
4120 
4121 	switch (be16_to_cpu(crq->query_capability.capability)) {
4122 	case MIN_TX_QUEUES:
4123 		adapter->min_tx_queues =
4124 		    be64_to_cpu(crq->query_capability.number);
4125 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4126 			   adapter->min_tx_queues);
4127 		break;
4128 	case MIN_RX_QUEUES:
4129 		adapter->min_rx_queues =
4130 		    be64_to_cpu(crq->query_capability.number);
4131 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4132 			   adapter->min_rx_queues);
4133 		break;
4134 	case MIN_RX_ADD_QUEUES:
4135 		adapter->min_rx_add_queues =
4136 		    be64_to_cpu(crq->query_capability.number);
4137 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4138 			   adapter->min_rx_add_queues);
4139 		break;
4140 	case MAX_TX_QUEUES:
4141 		adapter->max_tx_queues =
4142 		    be64_to_cpu(crq->query_capability.number);
4143 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4144 			   adapter->max_tx_queues);
4145 		break;
4146 	case MAX_RX_QUEUES:
4147 		adapter->max_rx_queues =
4148 		    be64_to_cpu(crq->query_capability.number);
4149 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4150 			   adapter->max_rx_queues);
4151 		break;
4152 	case MAX_RX_ADD_QUEUES:
4153 		adapter->max_rx_add_queues =
4154 		    be64_to_cpu(crq->query_capability.number);
4155 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4156 			   adapter->max_rx_add_queues);
4157 		break;
4158 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4159 		adapter->min_tx_entries_per_subcrq =
4160 		    be64_to_cpu(crq->query_capability.number);
4161 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4162 			   adapter->min_tx_entries_per_subcrq);
4163 		break;
4164 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4165 		adapter->min_rx_add_entries_per_subcrq =
4166 		    be64_to_cpu(crq->query_capability.number);
4167 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4168 			   adapter->min_rx_add_entries_per_subcrq);
4169 		break;
4170 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4171 		adapter->max_tx_entries_per_subcrq =
4172 		    be64_to_cpu(crq->query_capability.number);
4173 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4174 			   adapter->max_tx_entries_per_subcrq);
4175 		break;
4176 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4177 		adapter->max_rx_add_entries_per_subcrq =
4178 		    be64_to_cpu(crq->query_capability.number);
4179 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4180 			   adapter->max_rx_add_entries_per_subcrq);
4181 		break;
4182 	case TCP_IP_OFFLOAD:
4183 		adapter->tcp_ip_offload =
4184 		    be64_to_cpu(crq->query_capability.number);
4185 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4186 			   adapter->tcp_ip_offload);
4187 		break;
4188 	case PROMISC_SUPPORTED:
4189 		adapter->promisc_supported =
4190 		    be64_to_cpu(crq->query_capability.number);
4191 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4192 			   adapter->promisc_supported);
4193 		break;
4194 	case MIN_MTU:
4195 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4196 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4197 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4198 		break;
4199 	case MAX_MTU:
4200 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4201 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4202 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4203 		break;
4204 	case MAX_MULTICAST_FILTERS:
4205 		adapter->max_multicast_filters =
4206 		    be64_to_cpu(crq->query_capability.number);
4207 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4208 			   adapter->max_multicast_filters);
4209 		break;
4210 	case VLAN_HEADER_INSERTION:
4211 		adapter->vlan_header_insertion =
4212 		    be64_to_cpu(crq->query_capability.number);
4213 		if (adapter->vlan_header_insertion)
4214 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4215 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4216 			   adapter->vlan_header_insertion);
4217 		break;
4218 	case RX_VLAN_HEADER_INSERTION:
4219 		adapter->rx_vlan_header_insertion =
4220 		    be64_to_cpu(crq->query_capability.number);
4221 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4222 			   adapter->rx_vlan_header_insertion);
4223 		break;
4224 	case MAX_TX_SG_ENTRIES:
4225 		adapter->max_tx_sg_entries =
4226 		    be64_to_cpu(crq->query_capability.number);
4227 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4228 			   adapter->max_tx_sg_entries);
4229 		break;
4230 	case RX_SG_SUPPORTED:
4231 		adapter->rx_sg_supported =
4232 		    be64_to_cpu(crq->query_capability.number);
4233 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4234 			   adapter->rx_sg_supported);
4235 		break;
4236 	case OPT_TX_COMP_SUB_QUEUES:
4237 		adapter->opt_tx_comp_sub_queues =
4238 		    be64_to_cpu(crq->query_capability.number);
4239 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4240 			   adapter->opt_tx_comp_sub_queues);
4241 		break;
4242 	case OPT_RX_COMP_QUEUES:
4243 		adapter->opt_rx_comp_queues =
4244 		    be64_to_cpu(crq->query_capability.number);
4245 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4246 			   adapter->opt_rx_comp_queues);
4247 		break;
4248 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4249 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4250 		    be64_to_cpu(crq->query_capability.number);
4251 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4252 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4253 		break;
4254 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4255 		adapter->opt_tx_entries_per_subcrq =
4256 		    be64_to_cpu(crq->query_capability.number);
4257 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4258 			   adapter->opt_tx_entries_per_subcrq);
4259 		break;
4260 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4261 		adapter->opt_rxba_entries_per_subcrq =
4262 		    be64_to_cpu(crq->query_capability.number);
4263 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4264 			   adapter->opt_rxba_entries_per_subcrq);
4265 		break;
4266 	case TX_RX_DESC_REQ:
4267 		adapter->tx_rx_desc_req = crq->query_capability.number;
4268 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4269 			   adapter->tx_rx_desc_req);
4270 		break;
4271 
4272 	default:
4273 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4274 			   crq->query_capability.capability);
4275 	}
4276 
4277 out:
4278 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4279 		adapter->wait_capability = false;
4280 		ibmvnic_send_req_caps(adapter, 0);
4281 	}
4282 }
4283 
4284 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4285 			       struct ibmvnic_adapter *adapter)
4286 {
4287 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4288 	struct net_device *netdev = adapter->netdev;
4289 	struct device *dev = &adapter->vdev->dev;
4290 	u64 *u64_crq = (u64 *)crq;
4291 	long rc;
4292 
4293 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4294 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4295 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4296 	switch (gen_crq->first) {
4297 	case IBMVNIC_CRQ_INIT_RSP:
4298 		switch (gen_crq->cmd) {
4299 		case IBMVNIC_CRQ_INIT:
4300 			dev_info(dev, "Partner initialized\n");
4301 			adapter->from_passive_init = true;
4302 			adapter->failover_pending = false;
4303 			if (!completion_done(&adapter->init_done)) {
4304 				complete(&adapter->init_done);
4305 				adapter->init_done_rc = -EIO;
4306 			}
4307 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4308 			break;
4309 		case IBMVNIC_CRQ_INIT_COMPLETE:
4310 			dev_info(dev, "Partner initialization complete\n");
4311 			adapter->crq.active = true;
4312 			send_version_xchg(adapter);
4313 			break;
4314 		default:
4315 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4316 		}
4317 		return;
4318 	case IBMVNIC_CRQ_XPORT_EVENT:
4319 		netif_carrier_off(netdev);
4320 		adapter->crq.active = false;
4321 		if (adapter->resetting)
4322 			adapter->force_reset_recovery = true;
4323 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4324 			dev_info(dev, "Migrated, re-enabling adapter\n");
4325 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4326 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4327 			dev_info(dev, "Backing device failover detected\n");
4328 			adapter->failover_pending = true;
4329 		} else {
4330 			/* The adapter lost the connection */
4331 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4332 				gen_crq->cmd);
4333 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4334 		}
4335 		return;
4336 	case IBMVNIC_CRQ_CMD_RSP:
4337 		break;
4338 	default:
4339 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4340 			gen_crq->first);
4341 		return;
4342 	}
4343 
4344 	switch (gen_crq->cmd) {
4345 	case VERSION_EXCHANGE_RSP:
4346 		rc = crq->version_exchange_rsp.rc.code;
4347 		if (rc) {
4348 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4349 			break;
4350 		}
4351 		dev_info(dev, "Partner protocol version is %d\n",
4352 			 crq->version_exchange_rsp.version);
4353 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4354 		    ibmvnic_version)
4355 			ibmvnic_version =
4356 			    be16_to_cpu(crq->version_exchange_rsp.version);
4357 		send_cap_queries(adapter);
4358 		break;
4359 	case QUERY_CAPABILITY_RSP:
4360 		handle_query_cap_rsp(crq, adapter);
4361 		break;
4362 	case QUERY_MAP_RSP:
4363 		handle_query_map_rsp(crq, adapter);
4364 		break;
4365 	case REQUEST_MAP_RSP:
4366 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4367 		complete(&adapter->fw_done);
4368 		break;
4369 	case REQUEST_UNMAP_RSP:
4370 		handle_request_unmap_rsp(crq, adapter);
4371 		break;
4372 	case REQUEST_CAPABILITY_RSP:
4373 		handle_request_cap_rsp(crq, adapter);
4374 		break;
4375 	case LOGIN_RSP:
4376 		netdev_dbg(netdev, "Got Login Response\n");
4377 		handle_login_rsp(crq, adapter);
4378 		break;
4379 	case LOGICAL_LINK_STATE_RSP:
4380 		netdev_dbg(netdev,
4381 			   "Got Logical Link State Response, state: %d rc: %d\n",
4382 			   crq->logical_link_state_rsp.link_state,
4383 			   crq->logical_link_state_rsp.rc.code);
4384 		adapter->logical_link_state =
4385 		    crq->logical_link_state_rsp.link_state;
4386 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4387 		complete(&adapter->init_done);
4388 		break;
4389 	case LINK_STATE_INDICATION:
4390 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4391 		adapter->phys_link_state =
4392 		    crq->link_state_indication.phys_link_state;
4393 		adapter->logical_link_state =
4394 		    crq->link_state_indication.logical_link_state;
4395 		break;
4396 	case CHANGE_MAC_ADDR_RSP:
4397 		netdev_dbg(netdev, "Got MAC address change Response\n");
4398 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4399 		break;
4400 	case ERROR_INDICATION:
4401 		netdev_dbg(netdev, "Got Error Indication\n");
4402 		handle_error_indication(crq, adapter);
4403 		break;
4404 	case REQUEST_STATISTICS_RSP:
4405 		netdev_dbg(netdev, "Got Statistics Response\n");
4406 		complete(&adapter->stats_done);
4407 		break;
4408 	case QUERY_IP_OFFLOAD_RSP:
4409 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4410 		handle_query_ip_offload_rsp(adapter);
4411 		break;
4412 	case MULTICAST_CTRL_RSP:
4413 		netdev_dbg(netdev, "Got multicast control Response\n");
4414 		break;
4415 	case CONTROL_IP_OFFLOAD_RSP:
4416 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4417 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4418 				 sizeof(adapter->ip_offload_ctrl),
4419 				 DMA_TO_DEVICE);
4420 		complete(&adapter->init_done);
4421 		break;
4422 	case COLLECT_FW_TRACE_RSP:
4423 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4424 		complete(&adapter->fw_done);
4425 		break;
4426 	case GET_VPD_SIZE_RSP:
4427 		handle_vpd_size_rsp(crq, adapter);
4428 		break;
4429 	case GET_VPD_RSP:
4430 		handle_vpd_rsp(crq, adapter);
4431 		break;
4432 	default:
4433 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4434 			   gen_crq->cmd);
4435 	}
4436 }
4437 
4438 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4439 {
4440 	struct ibmvnic_adapter *adapter = instance;
4441 
4442 	tasklet_schedule(&adapter->tasklet);
4443 	return IRQ_HANDLED;
4444 }
4445 
4446 static void ibmvnic_tasklet(void *data)
4447 {
4448 	struct ibmvnic_adapter *adapter = data;
4449 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4450 	union ibmvnic_crq *crq;
4451 	unsigned long flags;
4452 	bool done = false;
4453 
4454 	spin_lock_irqsave(&queue->lock, flags);
4455 	while (!done) {
4456 		/* Pull all the valid messages off the CRQ */
4457 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4458 			ibmvnic_handle_crq(crq, adapter);
4459 			crq->generic.first = 0;
4460 		}
4461 
4462 		/* remain in tasklet until all
4463 		 * capabilities responses are received
4464 		 */
4465 		if (!adapter->wait_capability)
4466 			done = true;
4467 	}
4468 	/* if capabilities CRQ's were sent in this tasklet, the following
4469 	 * tasklet must wait until all responses are received
4470 	 */
4471 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4472 		adapter->wait_capability = true;
4473 	spin_unlock_irqrestore(&queue->lock, flags);
4474 }
4475 
4476 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4477 {
4478 	struct vio_dev *vdev = adapter->vdev;
4479 	int rc;
4480 
4481 	do {
4482 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4483 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4484 
4485 	if (rc)
4486 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4487 
4488 	return rc;
4489 }
4490 
4491 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4492 {
4493 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4494 	struct device *dev = &adapter->vdev->dev;
4495 	struct vio_dev *vdev = adapter->vdev;
4496 	int rc;
4497 
4498 	/* Close the CRQ */
4499 	do {
4500 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4501 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4502 
4503 	/* Clean out the queue */
4504 	memset(crq->msgs, 0, PAGE_SIZE);
4505 	crq->cur = 0;
4506 	crq->active = false;
4507 
4508 	/* And re-open it again */
4509 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4510 				crq->msg_token, PAGE_SIZE);
4511 
4512 	if (rc == H_CLOSED)
4513 		/* Adapter is good, but other end is not ready */
4514 		dev_warn(dev, "Partner adapter not ready\n");
4515 	else if (rc != 0)
4516 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4517 
4518 	return rc;
4519 }
4520 
4521 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4522 {
4523 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4524 	struct vio_dev *vdev = adapter->vdev;
4525 	long rc;
4526 
4527 	if (!crq->msgs)
4528 		return;
4529 
4530 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4531 	free_irq(vdev->irq, adapter);
4532 	tasklet_kill(&adapter->tasklet);
4533 	do {
4534 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4535 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4536 
4537 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4538 			 DMA_BIDIRECTIONAL);
4539 	free_page((unsigned long)crq->msgs);
4540 	crq->msgs = NULL;
4541 	crq->active = false;
4542 }
4543 
4544 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4545 {
4546 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4547 	struct device *dev = &adapter->vdev->dev;
4548 	struct vio_dev *vdev = adapter->vdev;
4549 	int rc, retrc = -ENOMEM;
4550 
4551 	if (crq->msgs)
4552 		return 0;
4553 
4554 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4555 	/* Should we allocate more than one page? */
4556 
4557 	if (!crq->msgs)
4558 		return -ENOMEM;
4559 
4560 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4561 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4562 					DMA_BIDIRECTIONAL);
4563 	if (dma_mapping_error(dev, crq->msg_token))
4564 		goto map_failed;
4565 
4566 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4567 				crq->msg_token, PAGE_SIZE);
4568 
4569 	if (rc == H_RESOURCE)
4570 		/* maybe kexecing and resource is busy. try a reset */
4571 		rc = ibmvnic_reset_crq(adapter);
4572 	retrc = rc;
4573 
4574 	if (rc == H_CLOSED) {
4575 		dev_warn(dev, "Partner adapter not ready\n");
4576 	} else if (rc) {
4577 		dev_warn(dev, "Error %d opening adapter\n", rc);
4578 		goto reg_crq_failed;
4579 	}
4580 
4581 	retrc = 0;
4582 
4583 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4584 		     (unsigned long)adapter);
4585 
4586 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4587 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4588 			 adapter);
4589 	if (rc) {
4590 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4591 			vdev->irq, rc);
4592 		goto req_irq_failed;
4593 	}
4594 
4595 	rc = vio_enable_interrupts(vdev);
4596 	if (rc) {
4597 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4598 		goto req_irq_failed;
4599 	}
4600 
4601 	crq->cur = 0;
4602 	spin_lock_init(&crq->lock);
4603 
4604 	return retrc;
4605 
4606 req_irq_failed:
4607 	tasklet_kill(&adapter->tasklet);
4608 	do {
4609 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4610 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4611 reg_crq_failed:
4612 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4613 map_failed:
4614 	free_page((unsigned long)crq->msgs);
4615 	crq->msgs = NULL;
4616 	return retrc;
4617 }
4618 
4619 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4620 {
4621 	struct device *dev = &adapter->vdev->dev;
4622 	unsigned long timeout = msecs_to_jiffies(30000);
4623 	u64 old_num_rx_queues, old_num_tx_queues;
4624 	int rc;
4625 
4626 	adapter->from_passive_init = false;
4627 
4628 	old_num_rx_queues = adapter->req_rx_queues;
4629 	old_num_tx_queues = adapter->req_tx_queues;
4630 
4631 	init_completion(&adapter->init_done);
4632 	adapter->init_done_rc = 0;
4633 	ibmvnic_send_crq_init(adapter);
4634 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4635 		dev_err(dev, "Initialization sequence timed out\n");
4636 		return -1;
4637 	}
4638 
4639 	if (adapter->init_done_rc) {
4640 		release_crq_queue(adapter);
4641 		return adapter->init_done_rc;
4642 	}
4643 
4644 	if (adapter->from_passive_init) {
4645 		adapter->state = VNIC_OPEN;
4646 		adapter->from_passive_init = false;
4647 		return -1;
4648 	}
4649 
4650 	if (adapter->resetting && !adapter->wait_for_reset &&
4651 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
4652 		if (adapter->req_rx_queues != old_num_rx_queues ||
4653 		    adapter->req_tx_queues != old_num_tx_queues) {
4654 			release_sub_crqs(adapter, 0);
4655 			rc = init_sub_crqs(adapter);
4656 		} else {
4657 			rc = reset_sub_crq_queues(adapter);
4658 		}
4659 	} else {
4660 		rc = init_sub_crqs(adapter);
4661 	}
4662 
4663 	if (rc) {
4664 		dev_err(dev, "Initialization of sub crqs failed\n");
4665 		release_crq_queue(adapter);
4666 		return rc;
4667 	}
4668 
4669 	rc = init_sub_crq_irqs(adapter);
4670 	if (rc) {
4671 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4672 		release_crq_queue(adapter);
4673 	}
4674 
4675 	return rc;
4676 }
4677 
4678 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4679 {
4680 	struct device *dev = &adapter->vdev->dev;
4681 	unsigned long timeout = msecs_to_jiffies(30000);
4682 	int rc;
4683 
4684 	adapter->from_passive_init = false;
4685 
4686 	init_completion(&adapter->init_done);
4687 	adapter->init_done_rc = 0;
4688 	ibmvnic_send_crq_init(adapter);
4689 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4690 		dev_err(dev, "Initialization sequence timed out\n");
4691 		return -1;
4692 	}
4693 
4694 	if (adapter->init_done_rc) {
4695 		release_crq_queue(adapter);
4696 		return adapter->init_done_rc;
4697 	}
4698 
4699 	if (adapter->from_passive_init) {
4700 		adapter->state = VNIC_OPEN;
4701 		adapter->from_passive_init = false;
4702 		return -1;
4703 	}
4704 
4705 	rc = init_sub_crqs(adapter);
4706 	if (rc) {
4707 		dev_err(dev, "Initialization of sub crqs failed\n");
4708 		release_crq_queue(adapter);
4709 		return rc;
4710 	}
4711 
4712 	rc = init_sub_crq_irqs(adapter);
4713 	if (rc) {
4714 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4715 		release_crq_queue(adapter);
4716 	}
4717 
4718 	return rc;
4719 }
4720 
4721 static struct device_attribute dev_attr_failover;
4722 
4723 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4724 {
4725 	struct ibmvnic_adapter *adapter;
4726 	struct net_device *netdev;
4727 	unsigned char *mac_addr_p;
4728 	int rc;
4729 
4730 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4731 		dev->unit_address);
4732 
4733 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4734 							VETH_MAC_ADDR, NULL);
4735 	if (!mac_addr_p) {
4736 		dev_err(&dev->dev,
4737 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4738 			__FILE__, __LINE__);
4739 		return 0;
4740 	}
4741 
4742 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4743 				   IBMVNIC_MAX_QUEUES);
4744 	if (!netdev)
4745 		return -ENOMEM;
4746 
4747 	adapter = netdev_priv(netdev);
4748 	adapter->state = VNIC_PROBING;
4749 	dev_set_drvdata(&dev->dev, netdev);
4750 	adapter->vdev = dev;
4751 	adapter->netdev = netdev;
4752 
4753 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
4754 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4755 	netdev->irq = dev->irq;
4756 	netdev->netdev_ops = &ibmvnic_netdev_ops;
4757 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4758 	SET_NETDEV_DEV(netdev, &dev->dev);
4759 
4760 	spin_lock_init(&adapter->stats_lock);
4761 
4762 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4763 	INIT_LIST_HEAD(&adapter->rwi_list);
4764 	spin_lock_init(&adapter->rwi_lock);
4765 	adapter->resetting = false;
4766 
4767 	adapter->mac_change_pending = false;
4768 
4769 	do {
4770 		rc = init_crq_queue(adapter);
4771 		if (rc) {
4772 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4773 				rc);
4774 			goto ibmvnic_init_fail;
4775 		}
4776 
4777 		rc = ibmvnic_init(adapter);
4778 		if (rc && rc != EAGAIN)
4779 			goto ibmvnic_init_fail;
4780 	} while (rc == EAGAIN);
4781 
4782 	rc = init_stats_buffers(adapter);
4783 	if (rc)
4784 		goto ibmvnic_init_fail;
4785 
4786 	rc = init_stats_token(adapter);
4787 	if (rc)
4788 		goto ibmvnic_stats_fail;
4789 
4790 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4791 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4792 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4793 
4794 	rc = device_create_file(&dev->dev, &dev_attr_failover);
4795 	if (rc)
4796 		goto ibmvnic_dev_file_err;
4797 
4798 	netif_carrier_off(netdev);
4799 	rc = register_netdev(netdev);
4800 	if (rc) {
4801 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4802 		goto ibmvnic_register_fail;
4803 	}
4804 	dev_info(&dev->dev, "ibmvnic registered\n");
4805 
4806 	adapter->state = VNIC_PROBED;
4807 
4808 	adapter->wait_for_reset = false;
4809 
4810 	return 0;
4811 
4812 ibmvnic_register_fail:
4813 	device_remove_file(&dev->dev, &dev_attr_failover);
4814 
4815 ibmvnic_dev_file_err:
4816 	release_stats_token(adapter);
4817 
4818 ibmvnic_stats_fail:
4819 	release_stats_buffers(adapter);
4820 
4821 ibmvnic_init_fail:
4822 	release_sub_crqs(adapter, 1);
4823 	release_crq_queue(adapter);
4824 	free_netdev(netdev);
4825 
4826 	return rc;
4827 }
4828 
4829 static int ibmvnic_remove(struct vio_dev *dev)
4830 {
4831 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
4832 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4833 
4834 	adapter->state = VNIC_REMOVING;
4835 	rtnl_lock();
4836 	unregister_netdevice(netdev);
4837 
4838 	release_resources(adapter);
4839 	release_sub_crqs(adapter, 1);
4840 	release_crq_queue(adapter);
4841 
4842 	release_stats_token(adapter);
4843 	release_stats_buffers(adapter);
4844 
4845 	adapter->state = VNIC_REMOVED;
4846 
4847 	rtnl_unlock();
4848 	device_remove_file(&dev->dev, &dev_attr_failover);
4849 	free_netdev(netdev);
4850 	dev_set_drvdata(&dev->dev, NULL);
4851 
4852 	return 0;
4853 }
4854 
4855 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4856 			      const char *buf, size_t count)
4857 {
4858 	struct net_device *netdev = dev_get_drvdata(dev);
4859 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4860 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4861 	__be64 session_token;
4862 	long rc;
4863 
4864 	if (!sysfs_streq(buf, "1"))
4865 		return -EINVAL;
4866 
4867 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4868 			 H_GET_SESSION_TOKEN, 0, 0, 0);
4869 	if (rc) {
4870 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4871 			   rc);
4872 		return -EINVAL;
4873 	}
4874 
4875 	session_token = (__be64)retbuf[0];
4876 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4877 		   be64_to_cpu(session_token));
4878 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4879 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
4880 	if (rc) {
4881 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4882 			   rc);
4883 		return -EINVAL;
4884 	}
4885 
4886 	return count;
4887 }
4888 
4889 static DEVICE_ATTR_WO(failover);
4890 
4891 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4892 {
4893 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4894 	struct ibmvnic_adapter *adapter;
4895 	struct iommu_table *tbl;
4896 	unsigned long ret = 0;
4897 	int i;
4898 
4899 	tbl = get_iommu_table_base(&vdev->dev);
4900 
4901 	/* netdev inits at probe time along with the structures we need below*/
4902 	if (!netdev)
4903 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4904 
4905 	adapter = netdev_priv(netdev);
4906 
4907 	ret += PAGE_SIZE; /* the crq message queue */
4908 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4909 
4910 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4911 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4912 
4913 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4914 	     i++)
4915 		ret += adapter->rx_pool[i].size *
4916 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4917 
4918 	return ret;
4919 }
4920 
4921 static int ibmvnic_resume(struct device *dev)
4922 {
4923 	struct net_device *netdev = dev_get_drvdata(dev);
4924 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4925 
4926 	if (adapter->state != VNIC_OPEN)
4927 		return 0;
4928 
4929 	tasklet_schedule(&adapter->tasklet);
4930 
4931 	return 0;
4932 }
4933 
4934 static const struct vio_device_id ibmvnic_device_table[] = {
4935 	{"network", "IBM,vnic"},
4936 	{"", "" }
4937 };
4938 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4939 
4940 static const struct dev_pm_ops ibmvnic_pm_ops = {
4941 	.resume = ibmvnic_resume
4942 };
4943 
4944 static struct vio_driver ibmvnic_driver = {
4945 	.id_table       = ibmvnic_device_table,
4946 	.probe          = ibmvnic_probe,
4947 	.remove         = ibmvnic_remove,
4948 	.get_desired_dma = ibmvnic_get_desired_dma,
4949 	.name		= ibmvnic_driver_name,
4950 	.pm		= &ibmvnic_pm_ops,
4951 };
4952 
4953 /* module functions */
4954 static int __init ibmvnic_module_init(void)
4955 {
4956 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4957 		IBMVNIC_DRIVER_VERSION);
4958 
4959 	return vio_register_driver(&ibmvnic_driver);
4960 }
4961 
4962 static void __exit ibmvnic_module_exit(void)
4963 {
4964 	vio_unregister_driver(&ibmvnic_driver);
4965 }
4966 
4967 module_init(ibmvnic_module_init);
4968 module_exit(ibmvnic_module_exit);
4969