xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision e3b9f1e81de2083f359bacd2a94bf1c024f2ede0)
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 *);
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 void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
113 static void send_request_unmap(struct ibmvnic_adapter *, u8);
114 static void 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 void release_crq_queue(struct ibmvnic_adapter *);
120 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
121 
122 struct ibmvnic_stat {
123 	char name[ETH_GSTRING_LEN];
124 	int offset;
125 };
126 
127 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
128 			     offsetof(struct ibmvnic_statistics, stat))
129 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
130 
131 static const struct ibmvnic_stat ibmvnic_stats[] = {
132 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
133 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
134 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
135 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
136 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
137 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
138 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
139 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
140 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
141 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
142 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
143 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
144 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
145 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
146 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
147 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
148 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
149 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
150 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
151 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
152 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
153 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
154 };
155 
156 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
157 			  unsigned long length, unsigned long *number,
158 			  unsigned long *irq)
159 {
160 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
161 	long rc;
162 
163 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
164 	*number = retbuf[0];
165 	*irq = retbuf[1];
166 
167 	return rc;
168 }
169 
170 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
171 				struct ibmvnic_long_term_buff *ltb, int size)
172 {
173 	struct device *dev = &adapter->vdev->dev;
174 
175 	ltb->size = size;
176 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
177 				       GFP_KERNEL);
178 
179 	if (!ltb->buff) {
180 		dev_err(dev, "Couldn't alloc long term buffer\n");
181 		return -ENOMEM;
182 	}
183 	ltb->map_id = adapter->map_id;
184 	adapter->map_id++;
185 
186 	init_completion(&adapter->fw_done);
187 	send_request_map(adapter, ltb->addr,
188 			 ltb->size, ltb->map_id);
189 	wait_for_completion(&adapter->fw_done);
190 
191 	if (adapter->fw_done_rc) {
192 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
193 			adapter->fw_done_rc);
194 		return -1;
195 	}
196 	return 0;
197 }
198 
199 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
200 				struct ibmvnic_long_term_buff *ltb)
201 {
202 	struct device *dev = &adapter->vdev->dev;
203 
204 	if (!ltb->buff)
205 		return;
206 
207 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
208 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
209 		send_request_unmap(adapter, ltb->map_id);
210 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
211 }
212 
213 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
214 				struct ibmvnic_long_term_buff *ltb)
215 {
216 	memset(ltb->buff, 0, ltb->size);
217 
218 	init_completion(&adapter->fw_done);
219 	send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
220 	wait_for_completion(&adapter->fw_done);
221 
222 	if (adapter->fw_done_rc) {
223 		dev_info(&adapter->vdev->dev,
224 			 "Reset failed, attempting to free and reallocate buffer\n");
225 		free_long_term_buff(adapter, ltb);
226 		return alloc_long_term_buff(adapter, ltb, ltb->size);
227 	}
228 	return 0;
229 }
230 
231 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
232 {
233 	int i;
234 
235 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
236 	     i++)
237 		adapter->rx_pool[i].active = 0;
238 }
239 
240 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
241 			      struct ibmvnic_rx_pool *pool)
242 {
243 	int count = pool->size - atomic_read(&pool->available);
244 	struct device *dev = &adapter->vdev->dev;
245 	int buffers_added = 0;
246 	unsigned long lpar_rc;
247 	union sub_crq sub_crq;
248 	struct sk_buff *skb;
249 	unsigned int offset;
250 	dma_addr_t dma_addr;
251 	unsigned char *dst;
252 	u64 *handle_array;
253 	int shift = 0;
254 	int index;
255 	int i;
256 
257 	if (!pool->active)
258 		return;
259 
260 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
261 				      be32_to_cpu(adapter->login_rsp_buf->
262 				      off_rxadd_subcrqs));
263 
264 	for (i = 0; i < count; ++i) {
265 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
266 		if (!skb) {
267 			dev_err(dev, "Couldn't replenish rx buff\n");
268 			adapter->replenish_no_mem++;
269 			break;
270 		}
271 
272 		index = pool->free_map[pool->next_free];
273 
274 		if (pool->rx_buff[index].skb)
275 			dev_err(dev, "Inconsistent free_map!\n");
276 
277 		/* Copy the skb to the long term mapped DMA buffer */
278 		offset = index * pool->buff_size;
279 		dst = pool->long_term_buff.buff + offset;
280 		memset(dst, 0, pool->buff_size);
281 		dma_addr = pool->long_term_buff.addr + offset;
282 		pool->rx_buff[index].data = dst;
283 
284 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
285 		pool->rx_buff[index].dma = dma_addr;
286 		pool->rx_buff[index].skb = skb;
287 		pool->rx_buff[index].pool_index = pool->index;
288 		pool->rx_buff[index].size = pool->buff_size;
289 
290 		memset(&sub_crq, 0, sizeof(sub_crq));
291 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
292 		sub_crq.rx_add.correlator =
293 		    cpu_to_be64((u64)&pool->rx_buff[index]);
294 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
295 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
296 
297 		/* The length field of the sCRQ is defined to be 24 bits so the
298 		 * buffer size needs to be left shifted by a byte before it is
299 		 * converted to big endian to prevent the last byte from being
300 		 * truncated.
301 		 */
302 #ifdef __LITTLE_ENDIAN__
303 		shift = 8;
304 #endif
305 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
306 
307 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
308 				      &sub_crq);
309 		if (lpar_rc != H_SUCCESS)
310 			goto failure;
311 
312 		buffers_added++;
313 		adapter->replenish_add_buff_success++;
314 		pool->next_free = (pool->next_free + 1) % pool->size;
315 	}
316 	atomic_add(buffers_added, &pool->available);
317 	return;
318 
319 failure:
320 	dev_info(dev, "replenish pools failure\n");
321 	pool->free_map[pool->next_free] = index;
322 	pool->rx_buff[index].skb = NULL;
323 	if (!dma_mapping_error(dev, dma_addr))
324 		dma_unmap_single(dev, dma_addr, pool->buff_size,
325 				 DMA_FROM_DEVICE);
326 
327 	dev_kfree_skb_any(skb);
328 	adapter->replenish_add_buff_failure++;
329 	atomic_add(buffers_added, &pool->available);
330 
331 	if (lpar_rc == H_CLOSED) {
332 		/* Disable buffer pool replenishment and report carrier off if
333 		 * queue is closed. Firmware guarantees that a signal will
334 		 * be sent to the driver, triggering a reset.
335 		 */
336 		deactivate_rx_pools(adapter);
337 		netif_carrier_off(adapter->netdev);
338 	}
339 }
340 
341 static void replenish_pools(struct ibmvnic_adapter *adapter)
342 {
343 	int i;
344 
345 	adapter->replenish_task_cycles++;
346 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
347 	     i++) {
348 		if (adapter->rx_pool[i].active)
349 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
350 	}
351 }
352 
353 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
354 {
355 	kfree(adapter->tx_stats_buffers);
356 	kfree(adapter->rx_stats_buffers);
357 	adapter->tx_stats_buffers = NULL;
358 	adapter->rx_stats_buffers = NULL;
359 }
360 
361 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
362 {
363 	adapter->tx_stats_buffers =
364 				kcalloc(adapter->req_tx_queues,
365 					sizeof(struct ibmvnic_tx_queue_stats),
366 					GFP_KERNEL);
367 	if (!adapter->tx_stats_buffers)
368 		return -ENOMEM;
369 
370 	adapter->rx_stats_buffers =
371 				kcalloc(adapter->req_rx_queues,
372 					sizeof(struct ibmvnic_rx_queue_stats),
373 					GFP_KERNEL);
374 	if (!adapter->rx_stats_buffers)
375 		return -ENOMEM;
376 
377 	return 0;
378 }
379 
380 static void release_stats_token(struct ibmvnic_adapter *adapter)
381 {
382 	struct device *dev = &adapter->vdev->dev;
383 
384 	if (!adapter->stats_token)
385 		return;
386 
387 	dma_unmap_single(dev, adapter->stats_token,
388 			 sizeof(struct ibmvnic_statistics),
389 			 DMA_FROM_DEVICE);
390 	adapter->stats_token = 0;
391 }
392 
393 static int init_stats_token(struct ibmvnic_adapter *adapter)
394 {
395 	struct device *dev = &adapter->vdev->dev;
396 	dma_addr_t stok;
397 
398 	stok = dma_map_single(dev, &adapter->stats,
399 			      sizeof(struct ibmvnic_statistics),
400 			      DMA_FROM_DEVICE);
401 	if (dma_mapping_error(dev, stok)) {
402 		dev_err(dev, "Couldn't map stats buffer\n");
403 		return -1;
404 	}
405 
406 	adapter->stats_token = stok;
407 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
408 	return 0;
409 }
410 
411 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
412 {
413 	struct ibmvnic_rx_pool *rx_pool;
414 	int rx_scrqs;
415 	int i, j, rc;
416 	u64 *size_array;
417 
418 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
419 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
420 
421 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
422 	for (i = 0; i < rx_scrqs; i++) {
423 		rx_pool = &adapter->rx_pool[i];
424 
425 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
426 
427 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
428 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
429 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
430 			alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
431 					     rx_pool->size *
432 					     rx_pool->buff_size);
433 		} else {
434 			rc = reset_long_term_buff(adapter,
435 						  &rx_pool->long_term_buff);
436 		}
437 
438 		if (rc)
439 			return rc;
440 
441 		for (j = 0; j < rx_pool->size; j++)
442 			rx_pool->free_map[j] = j;
443 
444 		memset(rx_pool->rx_buff, 0,
445 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
446 
447 		atomic_set(&rx_pool->available, 0);
448 		rx_pool->next_alloc = 0;
449 		rx_pool->next_free = 0;
450 		rx_pool->active = 1;
451 	}
452 
453 	return 0;
454 }
455 
456 static void release_rx_pools(struct ibmvnic_adapter *adapter)
457 {
458 	struct ibmvnic_rx_pool *rx_pool;
459 	int i, j;
460 
461 	if (!adapter->rx_pool)
462 		return;
463 
464 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
465 		rx_pool = &adapter->rx_pool[i];
466 
467 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
468 
469 		kfree(rx_pool->free_map);
470 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
471 
472 		if (!rx_pool->rx_buff)
473 			continue;
474 
475 		for (j = 0; j < rx_pool->size; j++) {
476 			if (rx_pool->rx_buff[j].skb) {
477 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
478 				rx_pool->rx_buff[i].skb = NULL;
479 			}
480 		}
481 
482 		kfree(rx_pool->rx_buff);
483 	}
484 
485 	kfree(adapter->rx_pool);
486 	adapter->rx_pool = NULL;
487 	adapter->num_active_rx_pools = 0;
488 }
489 
490 static int init_rx_pools(struct net_device *netdev)
491 {
492 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
493 	struct device *dev = &adapter->vdev->dev;
494 	struct ibmvnic_rx_pool *rx_pool;
495 	int rxadd_subcrqs;
496 	u64 *size_array;
497 	int i, j;
498 
499 	rxadd_subcrqs =
500 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
501 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
502 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
503 
504 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
505 				   sizeof(struct ibmvnic_rx_pool),
506 				   GFP_KERNEL);
507 	if (!adapter->rx_pool) {
508 		dev_err(dev, "Failed to allocate rx pools\n");
509 		return -1;
510 	}
511 
512 	adapter->num_active_rx_pools = 0;
513 
514 	for (i = 0; i < rxadd_subcrqs; i++) {
515 		rx_pool = &adapter->rx_pool[i];
516 
517 		netdev_dbg(adapter->netdev,
518 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
519 			   i, adapter->req_rx_add_entries_per_subcrq,
520 			   be64_to_cpu(size_array[i]));
521 
522 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
523 		rx_pool->index = i;
524 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
525 		rx_pool->active = 1;
526 
527 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
528 					    GFP_KERNEL);
529 		if (!rx_pool->free_map) {
530 			release_rx_pools(adapter);
531 			return -1;
532 		}
533 
534 		rx_pool->rx_buff = kcalloc(rx_pool->size,
535 					   sizeof(struct ibmvnic_rx_buff),
536 					   GFP_KERNEL);
537 		if (!rx_pool->rx_buff) {
538 			dev_err(dev, "Couldn't alloc rx buffers\n");
539 			release_rx_pools(adapter);
540 			return -1;
541 		}
542 
543 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
544 					 rx_pool->size * rx_pool->buff_size)) {
545 			release_rx_pools(adapter);
546 			return -1;
547 		}
548 
549 		for (j = 0; j < rx_pool->size; ++j)
550 			rx_pool->free_map[j] = j;
551 
552 		atomic_set(&rx_pool->available, 0);
553 		rx_pool->next_alloc = 0;
554 		rx_pool->next_free = 0;
555 	}
556 
557 	adapter->num_active_rx_pools = rxadd_subcrqs;
558 
559 	return 0;
560 }
561 
562 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
563 {
564 	struct ibmvnic_tx_pool *tx_pool;
565 	int tx_scrqs;
566 	int i, j, rc;
567 
568 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
569 	for (i = 0; i < tx_scrqs; i++) {
570 		netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
571 
572 		tx_pool = &adapter->tx_pool[i];
573 
574 		rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
575 		if (rc)
576 			return rc;
577 
578 		rc = reset_long_term_buff(adapter, &tx_pool->tso_ltb);
579 		if (rc)
580 			return rc;
581 
582 		memset(tx_pool->tx_buff, 0,
583 		       adapter->req_tx_entries_per_subcrq *
584 		       sizeof(struct ibmvnic_tx_buff));
585 
586 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
587 			tx_pool->free_map[j] = j;
588 
589 		tx_pool->consumer_index = 0;
590 		tx_pool->producer_index = 0;
591 		tx_pool->tso_index = 0;
592 	}
593 
594 	return 0;
595 }
596 
597 static void release_vpd_data(struct ibmvnic_adapter *adapter)
598 {
599 	if (!adapter->vpd)
600 		return;
601 
602 	kfree(adapter->vpd->buff);
603 	kfree(adapter->vpd);
604 
605 	adapter->vpd = NULL;
606 }
607 
608 static void release_tx_pools(struct ibmvnic_adapter *adapter)
609 {
610 	struct ibmvnic_tx_pool *tx_pool;
611 	int i;
612 
613 	if (!adapter->tx_pool)
614 		return;
615 
616 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
617 		netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
618 		tx_pool = &adapter->tx_pool[i];
619 		kfree(tx_pool->tx_buff);
620 		free_long_term_buff(adapter, &tx_pool->long_term_buff);
621 		free_long_term_buff(adapter, &tx_pool->tso_ltb);
622 		kfree(tx_pool->free_map);
623 	}
624 
625 	kfree(adapter->tx_pool);
626 	adapter->tx_pool = NULL;
627 	adapter->num_active_tx_pools = 0;
628 }
629 
630 static int init_tx_pools(struct net_device *netdev)
631 {
632 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
633 	struct device *dev = &adapter->vdev->dev;
634 	struct ibmvnic_tx_pool *tx_pool;
635 	int tx_subcrqs;
636 	int i, j;
637 
638 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
639 	adapter->tx_pool = kcalloc(tx_subcrqs,
640 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
641 	if (!adapter->tx_pool)
642 		return -1;
643 
644 	adapter->num_active_tx_pools = 0;
645 
646 	for (i = 0; i < tx_subcrqs; i++) {
647 		tx_pool = &adapter->tx_pool[i];
648 
649 		netdev_dbg(adapter->netdev,
650 			   "Initializing tx_pool[%d], %lld buffs\n",
651 			   i, adapter->req_tx_entries_per_subcrq);
652 
653 		tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
654 					   sizeof(struct ibmvnic_tx_buff),
655 					   GFP_KERNEL);
656 		if (!tx_pool->tx_buff) {
657 			dev_err(dev, "tx pool buffer allocation failed\n");
658 			release_tx_pools(adapter);
659 			return -1;
660 		}
661 
662 		if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
663 					 adapter->req_tx_entries_per_subcrq *
664 					 adapter->req_mtu)) {
665 			release_tx_pools(adapter);
666 			return -1;
667 		}
668 
669 		/* alloc TSO ltb */
670 		if (alloc_long_term_buff(adapter, &tx_pool->tso_ltb,
671 					 IBMVNIC_TSO_BUFS *
672 					 IBMVNIC_TSO_BUF_SZ)) {
673 			release_tx_pools(adapter);
674 			return -1;
675 		}
676 
677 		tx_pool->tso_index = 0;
678 
679 		tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
680 					    sizeof(int), GFP_KERNEL);
681 		if (!tx_pool->free_map) {
682 			release_tx_pools(adapter);
683 			return -1;
684 		}
685 
686 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
687 			tx_pool->free_map[j] = j;
688 
689 		tx_pool->consumer_index = 0;
690 		tx_pool->producer_index = 0;
691 	}
692 
693 	adapter->num_active_tx_pools = tx_subcrqs;
694 
695 	return 0;
696 }
697 
698 static void release_error_buffers(struct ibmvnic_adapter *adapter)
699 {
700 	struct device *dev = &adapter->vdev->dev;
701 	struct ibmvnic_error_buff *error_buff, *tmp;
702 	unsigned long flags;
703 
704 	spin_lock_irqsave(&adapter->error_list_lock, flags);
705 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
706 		list_del(&error_buff->list);
707 		dma_unmap_single(dev, error_buff->dma, error_buff->len,
708 				 DMA_FROM_DEVICE);
709 		kfree(error_buff->buff);
710 		kfree(error_buff);
711 	}
712 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
713 }
714 
715 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
716 {
717 	int i;
718 
719 	if (adapter->napi_enabled)
720 		return;
721 
722 	for (i = 0; i < adapter->req_rx_queues; i++)
723 		napi_enable(&adapter->napi[i]);
724 
725 	adapter->napi_enabled = true;
726 }
727 
728 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
729 {
730 	int i;
731 
732 	if (!adapter->napi_enabled)
733 		return;
734 
735 	for (i = 0; i < adapter->req_rx_queues; i++) {
736 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
737 		napi_disable(&adapter->napi[i]);
738 	}
739 
740 	adapter->napi_enabled = false;
741 }
742 
743 static int ibmvnic_login(struct net_device *netdev)
744 {
745 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
746 	unsigned long timeout = msecs_to_jiffies(30000);
747 	struct device *dev = &adapter->vdev->dev;
748 	int rc;
749 
750 	do {
751 		if (adapter->renegotiate) {
752 			adapter->renegotiate = false;
753 			release_sub_crqs(adapter);
754 
755 			reinit_completion(&adapter->init_done);
756 			send_cap_queries(adapter);
757 			if (!wait_for_completion_timeout(&adapter->init_done,
758 							 timeout)) {
759 				dev_err(dev, "Capabilities query timeout\n");
760 				return -1;
761 			}
762 			rc = init_sub_crqs(adapter);
763 			if (rc) {
764 				dev_err(dev,
765 					"Initialization of SCRQ's failed\n");
766 				return -1;
767 			}
768 			rc = init_sub_crq_irqs(adapter);
769 			if (rc) {
770 				dev_err(dev,
771 					"Initialization of SCRQ's irqs failed\n");
772 				return -1;
773 			}
774 		}
775 
776 		reinit_completion(&adapter->init_done);
777 		send_login(adapter);
778 		if (!wait_for_completion_timeout(&adapter->init_done,
779 						 timeout)) {
780 			dev_err(dev, "Login timeout\n");
781 			return -1;
782 		}
783 	} while (adapter->renegotiate);
784 
785 	/* handle pending MAC address changes after successful login */
786 	if (adapter->mac_change_pending) {
787 		__ibmvnic_set_mac(netdev, &adapter->desired.mac);
788 		adapter->mac_change_pending = false;
789 	}
790 
791 	return 0;
792 }
793 
794 static void release_login_buffer(struct ibmvnic_adapter *adapter)
795 {
796 	kfree(adapter->login_buf);
797 	adapter->login_buf = NULL;
798 }
799 
800 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
801 {
802 	kfree(adapter->login_rsp_buf);
803 	adapter->login_rsp_buf = NULL;
804 }
805 
806 static void release_resources(struct ibmvnic_adapter *adapter)
807 {
808 	int i;
809 
810 	release_vpd_data(adapter);
811 
812 	release_tx_pools(adapter);
813 	release_rx_pools(adapter);
814 
815 	release_stats_token(adapter);
816 	release_stats_buffers(adapter);
817 	release_error_buffers(adapter);
818 
819 	if (adapter->napi) {
820 		for (i = 0; i < adapter->req_rx_queues; i++) {
821 			if (&adapter->napi[i]) {
822 				netdev_dbg(adapter->netdev,
823 					   "Releasing napi[%d]\n", i);
824 				netif_napi_del(&adapter->napi[i]);
825 			}
826 		}
827 	}
828 	kfree(adapter->napi);
829 	adapter->napi = NULL;
830 
831 	release_login_rsp_buffer(adapter);
832 }
833 
834 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
835 {
836 	struct net_device *netdev = adapter->netdev;
837 	unsigned long timeout = msecs_to_jiffies(30000);
838 	union ibmvnic_crq crq;
839 	bool resend;
840 	int rc;
841 
842 	netdev_dbg(netdev, "setting link state %d\n", link_state);
843 
844 	memset(&crq, 0, sizeof(crq));
845 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
846 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
847 	crq.logical_link_state.link_state = link_state;
848 
849 	do {
850 		resend = false;
851 
852 		reinit_completion(&adapter->init_done);
853 		rc = ibmvnic_send_crq(adapter, &crq);
854 		if (rc) {
855 			netdev_err(netdev, "Failed to set link state\n");
856 			return rc;
857 		}
858 
859 		if (!wait_for_completion_timeout(&adapter->init_done,
860 						 timeout)) {
861 			netdev_err(netdev, "timeout setting link state\n");
862 			return -1;
863 		}
864 
865 		if (adapter->init_done_rc == 1) {
866 			/* Partuial success, delay and re-send */
867 			mdelay(1000);
868 			resend = true;
869 		}
870 	} while (resend);
871 
872 	return 0;
873 }
874 
875 static int set_real_num_queues(struct net_device *netdev)
876 {
877 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
878 	int rc;
879 
880 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
881 		   adapter->req_tx_queues, adapter->req_rx_queues);
882 
883 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
884 	if (rc) {
885 		netdev_err(netdev, "failed to set the number of tx queues\n");
886 		return rc;
887 	}
888 
889 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
890 	if (rc)
891 		netdev_err(netdev, "failed to set the number of rx queues\n");
892 
893 	return rc;
894 }
895 
896 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
897 {
898 	struct device *dev = &adapter->vdev->dev;
899 	union ibmvnic_crq crq;
900 	int len = 0;
901 
902 	if (adapter->vpd->buff)
903 		len = adapter->vpd->len;
904 
905 	init_completion(&adapter->fw_done);
906 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
907 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
908 	ibmvnic_send_crq(adapter, &crq);
909 	wait_for_completion(&adapter->fw_done);
910 
911 	if (!adapter->vpd->len)
912 		return -ENODATA;
913 
914 	if (!adapter->vpd->buff)
915 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
916 	else if (adapter->vpd->len != len)
917 		adapter->vpd->buff =
918 			krealloc(adapter->vpd->buff,
919 				 adapter->vpd->len, GFP_KERNEL);
920 
921 	if (!adapter->vpd->buff) {
922 		dev_err(dev, "Could allocate VPD buffer\n");
923 		return -ENOMEM;
924 	}
925 
926 	adapter->vpd->dma_addr =
927 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
928 			       DMA_FROM_DEVICE);
929 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
930 		dev_err(dev, "Could not map VPD buffer\n");
931 		kfree(adapter->vpd->buff);
932 		adapter->vpd->buff = NULL;
933 		return -ENOMEM;
934 	}
935 
936 	reinit_completion(&adapter->fw_done);
937 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
938 	crq.get_vpd.cmd = GET_VPD;
939 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
940 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
941 	ibmvnic_send_crq(adapter, &crq);
942 	wait_for_completion(&adapter->fw_done);
943 
944 	return 0;
945 }
946 
947 static int init_resources(struct ibmvnic_adapter *adapter)
948 {
949 	struct net_device *netdev = adapter->netdev;
950 	int i, rc;
951 
952 	rc = set_real_num_queues(netdev);
953 	if (rc)
954 		return rc;
955 
956 	rc = init_stats_buffers(adapter);
957 	if (rc)
958 		return rc;
959 
960 	rc = init_stats_token(adapter);
961 	if (rc)
962 		return rc;
963 
964 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
965 	if (!adapter->vpd)
966 		return -ENOMEM;
967 
968 	/* Vital Product Data (VPD) */
969 	rc = ibmvnic_get_vpd(adapter);
970 	if (rc) {
971 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
972 		return rc;
973 	}
974 
975 	adapter->map_id = 1;
976 	adapter->napi = kcalloc(adapter->req_rx_queues,
977 				sizeof(struct napi_struct), GFP_KERNEL);
978 	if (!adapter->napi)
979 		return -ENOMEM;
980 
981 	for (i = 0; i < adapter->req_rx_queues; i++) {
982 		netdev_dbg(netdev, "Adding napi[%d]\n", i);
983 		netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
984 			       NAPI_POLL_WEIGHT);
985 	}
986 
987 	send_map_query(adapter);
988 
989 	rc = init_rx_pools(netdev);
990 	if (rc)
991 		return rc;
992 
993 	rc = init_tx_pools(netdev);
994 	return rc;
995 }
996 
997 static int __ibmvnic_open(struct net_device *netdev)
998 {
999 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1000 	enum vnic_state prev_state = adapter->state;
1001 	int i, rc;
1002 
1003 	adapter->state = VNIC_OPENING;
1004 	replenish_pools(adapter);
1005 	ibmvnic_napi_enable(adapter);
1006 
1007 	/* We're ready to receive frames, enable the sub-crq interrupts and
1008 	 * set the logical link state to up
1009 	 */
1010 	for (i = 0; i < adapter->req_rx_queues; i++) {
1011 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1012 		if (prev_state == VNIC_CLOSED)
1013 			enable_irq(adapter->rx_scrq[i]->irq);
1014 		else
1015 			enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1016 	}
1017 
1018 	for (i = 0; i < adapter->req_tx_queues; i++) {
1019 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1020 		if (prev_state == VNIC_CLOSED)
1021 			enable_irq(adapter->tx_scrq[i]->irq);
1022 		else
1023 			enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1024 	}
1025 
1026 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1027 	if (rc) {
1028 		for (i = 0; i < adapter->req_rx_queues; i++)
1029 			napi_disable(&adapter->napi[i]);
1030 		release_resources(adapter);
1031 		return rc;
1032 	}
1033 
1034 	netif_tx_start_all_queues(netdev);
1035 
1036 	if (prev_state == VNIC_CLOSED) {
1037 		for (i = 0; i < adapter->req_rx_queues; i++)
1038 			napi_schedule(&adapter->napi[i]);
1039 	}
1040 
1041 	adapter->state = VNIC_OPEN;
1042 	return rc;
1043 }
1044 
1045 static int ibmvnic_open(struct net_device *netdev)
1046 {
1047 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1048 	int rc;
1049 
1050 	mutex_lock(&adapter->reset_lock);
1051 
1052 	if (adapter->state != VNIC_CLOSED) {
1053 		rc = ibmvnic_login(netdev);
1054 		if (rc) {
1055 			mutex_unlock(&adapter->reset_lock);
1056 			return rc;
1057 		}
1058 
1059 		rc = init_resources(adapter);
1060 		if (rc) {
1061 			netdev_err(netdev, "failed to initialize resources\n");
1062 			release_resources(adapter);
1063 			mutex_unlock(&adapter->reset_lock);
1064 			return rc;
1065 		}
1066 	}
1067 
1068 	rc = __ibmvnic_open(netdev);
1069 	netif_carrier_on(netdev);
1070 
1071 	mutex_unlock(&adapter->reset_lock);
1072 
1073 	return rc;
1074 }
1075 
1076 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1077 {
1078 	struct ibmvnic_rx_pool *rx_pool;
1079 	u64 rx_entries;
1080 	int rx_scrqs;
1081 	int i, j;
1082 
1083 	if (!adapter->rx_pool)
1084 		return;
1085 
1086 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
1087 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1088 
1089 	/* Free any remaining skbs in the rx buffer pools */
1090 	for (i = 0; i < rx_scrqs; i++) {
1091 		rx_pool = &adapter->rx_pool[i];
1092 		if (!rx_pool)
1093 			continue;
1094 
1095 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1096 		for (j = 0; j < rx_entries; j++) {
1097 			if (rx_pool->rx_buff[j].skb) {
1098 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
1099 				rx_pool->rx_buff[j].skb = NULL;
1100 			}
1101 		}
1102 	}
1103 }
1104 
1105 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1106 {
1107 	struct ibmvnic_tx_pool *tx_pool;
1108 	u64 tx_entries;
1109 	int tx_scrqs;
1110 	int i, j;
1111 
1112 	if (!adapter->tx_pool)
1113 		return;
1114 
1115 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
1116 	tx_entries = adapter->req_tx_entries_per_subcrq;
1117 
1118 	/* Free any remaining skbs in the tx buffer pools */
1119 	for (i = 0; i < tx_scrqs; i++) {
1120 		tx_pool = &adapter->tx_pool[i];
1121 		if (!tx_pool)
1122 			continue;
1123 
1124 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1125 		for (j = 0; j < tx_entries; j++) {
1126 			if (tx_pool->tx_buff[j].skb) {
1127 				dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
1128 				tx_pool->tx_buff[j].skb = NULL;
1129 			}
1130 		}
1131 	}
1132 }
1133 
1134 static int __ibmvnic_close(struct net_device *netdev)
1135 {
1136 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1137 	int rc = 0;
1138 	int i;
1139 
1140 	adapter->state = VNIC_CLOSING;
1141 
1142 	/* ensure that transmissions are stopped if called by do_reset */
1143 	if (adapter->resetting)
1144 		netif_tx_disable(netdev);
1145 	else
1146 		netif_tx_stop_all_queues(netdev);
1147 
1148 	ibmvnic_napi_disable(adapter);
1149 
1150 	if (adapter->tx_scrq) {
1151 		for (i = 0; i < adapter->req_tx_queues; i++)
1152 			if (adapter->tx_scrq[i]->irq) {
1153 				netdev_dbg(adapter->netdev,
1154 					   "Disabling tx_scrq[%d] irq\n", i);
1155 				disable_irq(adapter->tx_scrq[i]->irq);
1156 			}
1157 	}
1158 
1159 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1160 	if (rc)
1161 		return rc;
1162 
1163 	if (adapter->rx_scrq) {
1164 		for (i = 0; i < adapter->req_rx_queues; i++) {
1165 			int retries = 10;
1166 
1167 			while (pending_scrq(adapter, adapter->rx_scrq[i])) {
1168 				retries--;
1169 				mdelay(100);
1170 
1171 				if (retries == 0)
1172 					break;
1173 			}
1174 
1175 			if (adapter->rx_scrq[i]->irq) {
1176 				netdev_dbg(adapter->netdev,
1177 					   "Disabling rx_scrq[%d] irq\n", i);
1178 				disable_irq(adapter->rx_scrq[i]->irq);
1179 			}
1180 		}
1181 	}
1182 	clean_rx_pools(adapter);
1183 	clean_tx_pools(adapter);
1184 	adapter->state = VNIC_CLOSED;
1185 	return rc;
1186 }
1187 
1188 static int ibmvnic_close(struct net_device *netdev)
1189 {
1190 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1191 	int rc;
1192 
1193 	mutex_lock(&adapter->reset_lock);
1194 	rc = __ibmvnic_close(netdev);
1195 	mutex_unlock(&adapter->reset_lock);
1196 
1197 	return rc;
1198 }
1199 
1200 /**
1201  * build_hdr_data - creates L2/L3/L4 header data buffer
1202  * @hdr_field - bitfield determining needed headers
1203  * @skb - socket buffer
1204  * @hdr_len - array of header lengths
1205  * @tot_len - total length of data
1206  *
1207  * Reads hdr_field to determine which headers are needed by firmware.
1208  * Builds a buffer containing these headers.  Saves individual header
1209  * lengths and total buffer length to be used to build descriptors.
1210  */
1211 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1212 			  int *hdr_len, u8 *hdr_data)
1213 {
1214 	int len = 0;
1215 	u8 *hdr;
1216 
1217 	hdr_len[0] = sizeof(struct ethhdr);
1218 
1219 	if (skb->protocol == htons(ETH_P_IP)) {
1220 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1221 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1222 			hdr_len[2] = tcp_hdrlen(skb);
1223 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1224 			hdr_len[2] = sizeof(struct udphdr);
1225 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1226 		hdr_len[1] = sizeof(struct ipv6hdr);
1227 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1228 			hdr_len[2] = tcp_hdrlen(skb);
1229 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1230 			hdr_len[2] = sizeof(struct udphdr);
1231 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1232 		hdr_len[1] = arp_hdr_len(skb->dev);
1233 		hdr_len[2] = 0;
1234 	}
1235 
1236 	memset(hdr_data, 0, 120);
1237 	if ((hdr_field >> 6) & 1) {
1238 		hdr = skb_mac_header(skb);
1239 		memcpy(hdr_data, hdr, hdr_len[0]);
1240 		len += hdr_len[0];
1241 	}
1242 
1243 	if ((hdr_field >> 5) & 1) {
1244 		hdr = skb_network_header(skb);
1245 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1246 		len += hdr_len[1];
1247 	}
1248 
1249 	if ((hdr_field >> 4) & 1) {
1250 		hdr = skb_transport_header(skb);
1251 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1252 		len += hdr_len[2];
1253 	}
1254 	return len;
1255 }
1256 
1257 /**
1258  * create_hdr_descs - create header and header extension descriptors
1259  * @hdr_field - bitfield determining needed headers
1260  * @data - buffer containing header data
1261  * @len - length of data buffer
1262  * @hdr_len - array of individual header lengths
1263  * @scrq_arr - descriptor array
1264  *
1265  * Creates header and, if needed, header extension descriptors and
1266  * places them in a descriptor array, scrq_arr
1267  */
1268 
1269 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1270 			    union sub_crq *scrq_arr)
1271 {
1272 	union sub_crq hdr_desc;
1273 	int tmp_len = len;
1274 	int num_descs = 0;
1275 	u8 *data, *cur;
1276 	int tmp;
1277 
1278 	while (tmp_len > 0) {
1279 		cur = hdr_data + len - tmp_len;
1280 
1281 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1282 		if (cur != hdr_data) {
1283 			data = hdr_desc.hdr_ext.data;
1284 			tmp = tmp_len > 29 ? 29 : tmp_len;
1285 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1286 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1287 			hdr_desc.hdr_ext.len = tmp;
1288 		} else {
1289 			data = hdr_desc.hdr.data;
1290 			tmp = tmp_len > 24 ? 24 : tmp_len;
1291 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1292 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1293 			hdr_desc.hdr.len = tmp;
1294 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1295 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1296 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1297 			hdr_desc.hdr.flag = hdr_field << 1;
1298 		}
1299 		memcpy(data, cur, tmp);
1300 		tmp_len -= tmp;
1301 		*scrq_arr = hdr_desc;
1302 		scrq_arr++;
1303 		num_descs++;
1304 	}
1305 
1306 	return num_descs;
1307 }
1308 
1309 /**
1310  * build_hdr_descs_arr - build a header descriptor array
1311  * @skb - socket buffer
1312  * @num_entries - number of descriptors to be sent
1313  * @subcrq - first TX descriptor
1314  * @hdr_field - bit field determining which headers will be sent
1315  *
1316  * This function will build a TX descriptor array with applicable
1317  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1318  */
1319 
1320 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1321 				int *num_entries, u8 hdr_field)
1322 {
1323 	int hdr_len[3] = {0, 0, 0};
1324 	int tot_len;
1325 	u8 *hdr_data = txbuff->hdr_data;
1326 
1327 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1328 				 txbuff->hdr_data);
1329 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1330 			 txbuff->indir_arr + 1);
1331 }
1332 
1333 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1334 {
1335 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1336 	int queue_num = skb_get_queue_mapping(skb);
1337 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1338 	struct device *dev = &adapter->vdev->dev;
1339 	struct ibmvnic_tx_buff *tx_buff = NULL;
1340 	struct ibmvnic_sub_crq_queue *tx_scrq;
1341 	struct ibmvnic_tx_pool *tx_pool;
1342 	unsigned int tx_send_failed = 0;
1343 	unsigned int tx_map_failed = 0;
1344 	unsigned int tx_dropped = 0;
1345 	unsigned int tx_packets = 0;
1346 	unsigned int tx_bytes = 0;
1347 	dma_addr_t data_dma_addr;
1348 	struct netdev_queue *txq;
1349 	unsigned long lpar_rc;
1350 	union sub_crq tx_crq;
1351 	unsigned int offset;
1352 	int num_entries = 1;
1353 	unsigned char *dst;
1354 	u64 *handle_array;
1355 	int index = 0;
1356 	u8 proto = 0;
1357 	int ret = 0;
1358 
1359 	if (adapter->resetting) {
1360 		if (!netif_subqueue_stopped(netdev, skb))
1361 			netif_stop_subqueue(netdev, queue_num);
1362 		dev_kfree_skb_any(skb);
1363 
1364 		tx_send_failed++;
1365 		tx_dropped++;
1366 		ret = NETDEV_TX_OK;
1367 		goto out;
1368 	}
1369 
1370 	tx_pool = &adapter->tx_pool[queue_num];
1371 	tx_scrq = adapter->tx_scrq[queue_num];
1372 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1373 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1374 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1375 
1376 	index = tx_pool->free_map[tx_pool->consumer_index];
1377 
1378 	if (skb_is_gso(skb)) {
1379 		offset = tx_pool->tso_index * IBMVNIC_TSO_BUF_SZ;
1380 		dst = tx_pool->tso_ltb.buff + offset;
1381 		memset(dst, 0, IBMVNIC_TSO_BUF_SZ);
1382 		data_dma_addr = tx_pool->tso_ltb.addr + offset;
1383 		tx_pool->tso_index++;
1384 		if (tx_pool->tso_index == IBMVNIC_TSO_BUFS)
1385 			tx_pool->tso_index = 0;
1386 	} else {
1387 		offset = index * adapter->req_mtu;
1388 		dst = tx_pool->long_term_buff.buff + offset;
1389 		memset(dst, 0, adapter->req_mtu);
1390 		data_dma_addr = tx_pool->long_term_buff.addr + offset;
1391 	}
1392 
1393 	if (skb_shinfo(skb)->nr_frags) {
1394 		int cur, i;
1395 
1396 		/* Copy the head */
1397 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1398 		cur = skb_headlen(skb);
1399 
1400 		/* Copy the frags */
1401 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1402 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1403 
1404 			memcpy(dst + cur,
1405 			       page_address(skb_frag_page(frag)) +
1406 			       frag->page_offset, skb_frag_size(frag));
1407 			cur += skb_frag_size(frag);
1408 		}
1409 	} else {
1410 		skb_copy_from_linear_data(skb, dst, skb->len);
1411 	}
1412 
1413 	tx_pool->consumer_index =
1414 	    (tx_pool->consumer_index + 1) %
1415 		adapter->req_tx_entries_per_subcrq;
1416 
1417 	tx_buff = &tx_pool->tx_buff[index];
1418 	tx_buff->skb = skb;
1419 	tx_buff->data_dma[0] = data_dma_addr;
1420 	tx_buff->data_len[0] = skb->len;
1421 	tx_buff->index = index;
1422 	tx_buff->pool_index = queue_num;
1423 	tx_buff->last_frag = true;
1424 
1425 	memset(&tx_crq, 0, sizeof(tx_crq));
1426 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1427 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1428 	tx_crq.v1.n_crq_elem = 1;
1429 	tx_crq.v1.n_sge = 1;
1430 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1431 	tx_crq.v1.correlator = cpu_to_be32(index);
1432 	if (skb_is_gso(skb))
1433 		tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->tso_ltb.map_id);
1434 	else
1435 		tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1436 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1437 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1438 
1439 	if (adapter->vlan_header_insertion) {
1440 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1441 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1442 	}
1443 
1444 	if (skb->protocol == htons(ETH_P_IP)) {
1445 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1446 		proto = ip_hdr(skb)->protocol;
1447 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1448 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1449 		proto = ipv6_hdr(skb)->nexthdr;
1450 	}
1451 
1452 	if (proto == IPPROTO_TCP)
1453 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1454 	else if (proto == IPPROTO_UDP)
1455 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1456 
1457 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1458 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1459 		hdrs += 2;
1460 	}
1461 	if (skb_is_gso(skb)) {
1462 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1463 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1464 		hdrs += 2;
1465 	}
1466 	/* determine if l2/3/4 headers are sent to firmware */
1467 	if ((*hdrs >> 7) & 1) {
1468 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1469 		tx_crq.v1.n_crq_elem = num_entries;
1470 		tx_buff->indir_arr[0] = tx_crq;
1471 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1472 						    sizeof(tx_buff->indir_arr),
1473 						    DMA_TO_DEVICE);
1474 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1475 			dev_kfree_skb_any(skb);
1476 			tx_buff->skb = NULL;
1477 			if (!firmware_has_feature(FW_FEATURE_CMO))
1478 				dev_err(dev, "tx: unable to map descriptor array\n");
1479 			tx_map_failed++;
1480 			tx_dropped++;
1481 			ret = NETDEV_TX_OK;
1482 			goto out;
1483 		}
1484 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1485 					       (u64)tx_buff->indir_dma,
1486 					       (u64)num_entries);
1487 	} else {
1488 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1489 				      &tx_crq);
1490 	}
1491 	if (lpar_rc != H_SUCCESS) {
1492 		dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1493 
1494 		if (tx_pool->consumer_index == 0)
1495 			tx_pool->consumer_index =
1496 				adapter->req_tx_entries_per_subcrq - 1;
1497 		else
1498 			tx_pool->consumer_index--;
1499 
1500 		dev_kfree_skb_any(skb);
1501 		tx_buff->skb = NULL;
1502 
1503 		if (lpar_rc == H_CLOSED) {
1504 			/* Disable TX and report carrier off if queue is closed.
1505 			 * Firmware guarantees that a signal will be sent to the
1506 			 * driver, triggering a reset or some other action.
1507 			 */
1508 			netif_tx_stop_all_queues(netdev);
1509 			netif_carrier_off(netdev);
1510 		}
1511 
1512 		tx_send_failed++;
1513 		tx_dropped++;
1514 		ret = NETDEV_TX_OK;
1515 		goto out;
1516 	}
1517 
1518 	if (atomic_inc_return(&tx_scrq->used)
1519 					>= adapter->req_tx_entries_per_subcrq) {
1520 		netdev_info(netdev, "Stopping queue %d\n", queue_num);
1521 		netif_stop_subqueue(netdev, queue_num);
1522 	}
1523 
1524 	tx_packets++;
1525 	tx_bytes += skb->len;
1526 	txq->trans_start = jiffies;
1527 	ret = NETDEV_TX_OK;
1528 
1529 out:
1530 	netdev->stats.tx_dropped += tx_dropped;
1531 	netdev->stats.tx_bytes += tx_bytes;
1532 	netdev->stats.tx_packets += tx_packets;
1533 	adapter->tx_send_failed += tx_send_failed;
1534 	adapter->tx_map_failed += tx_map_failed;
1535 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1536 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1537 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1538 
1539 	return ret;
1540 }
1541 
1542 static void ibmvnic_set_multi(struct net_device *netdev)
1543 {
1544 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1545 	struct netdev_hw_addr *ha;
1546 	union ibmvnic_crq crq;
1547 
1548 	memset(&crq, 0, sizeof(crq));
1549 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1550 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1551 
1552 	if (netdev->flags & IFF_PROMISC) {
1553 		if (!adapter->promisc_supported)
1554 			return;
1555 	} else {
1556 		if (netdev->flags & IFF_ALLMULTI) {
1557 			/* Accept all multicast */
1558 			memset(&crq, 0, sizeof(crq));
1559 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1560 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1561 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1562 			ibmvnic_send_crq(adapter, &crq);
1563 		} else if (netdev_mc_empty(netdev)) {
1564 			/* Reject all multicast */
1565 			memset(&crq, 0, sizeof(crq));
1566 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1567 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1568 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1569 			ibmvnic_send_crq(adapter, &crq);
1570 		} else {
1571 			/* Accept one or more multicast(s) */
1572 			netdev_for_each_mc_addr(ha, netdev) {
1573 				memset(&crq, 0, sizeof(crq));
1574 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1575 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1576 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1577 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1578 						ha->addr);
1579 				ibmvnic_send_crq(adapter, &crq);
1580 			}
1581 		}
1582 	}
1583 }
1584 
1585 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1586 {
1587 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1588 	struct sockaddr *addr = p;
1589 	union ibmvnic_crq crq;
1590 
1591 	if (!is_valid_ether_addr(addr->sa_data))
1592 		return -EADDRNOTAVAIL;
1593 
1594 	memset(&crq, 0, sizeof(crq));
1595 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1596 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1597 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1598 
1599 	init_completion(&adapter->fw_done);
1600 	ibmvnic_send_crq(adapter, &crq);
1601 	wait_for_completion(&adapter->fw_done);
1602 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1603 	return adapter->fw_done_rc ? -EIO : 0;
1604 }
1605 
1606 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1607 {
1608 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1609 	struct sockaddr *addr = p;
1610 	int rc;
1611 
1612 	if (adapter->state == VNIC_PROBED) {
1613 		memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1614 		adapter->mac_change_pending = true;
1615 		return 0;
1616 	}
1617 
1618 	rc = __ibmvnic_set_mac(netdev, addr);
1619 
1620 	return rc;
1621 }
1622 
1623 /**
1624  * do_reset returns zero if we are able to keep processing reset events, or
1625  * non-zero if we hit a fatal error and must halt.
1626  */
1627 static int do_reset(struct ibmvnic_adapter *adapter,
1628 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1629 {
1630 	u64 old_num_rx_queues, old_num_tx_queues;
1631 	struct net_device *netdev = adapter->netdev;
1632 	int i, rc;
1633 
1634 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1635 		   rwi->reset_reason);
1636 
1637 	netif_carrier_off(netdev);
1638 	adapter->reset_reason = rwi->reset_reason;
1639 
1640 	old_num_rx_queues = adapter->req_rx_queues;
1641 	old_num_tx_queues = adapter->req_tx_queues;
1642 
1643 	if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1644 		rc = ibmvnic_reenable_crq_queue(adapter);
1645 		if (rc)
1646 			return 0;
1647 	}
1648 
1649 	rc = __ibmvnic_close(netdev);
1650 	if (rc)
1651 		return rc;
1652 
1653 	if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1654 	    adapter->wait_for_reset) {
1655 		release_resources(adapter);
1656 		release_sub_crqs(adapter);
1657 		release_crq_queue(adapter);
1658 	}
1659 
1660 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1661 		/* remove the closed state so when we call open it appears
1662 		 * we are coming from the probed state.
1663 		 */
1664 		adapter->state = VNIC_PROBED;
1665 
1666 		rc = ibmvnic_init(adapter);
1667 		if (rc)
1668 			return IBMVNIC_INIT_FAILED;
1669 
1670 		/* If the adapter was in PROBE state prior to the reset,
1671 		 * exit here.
1672 		 */
1673 		if (reset_state == VNIC_PROBED)
1674 			return 0;
1675 
1676 		rc = ibmvnic_login(netdev);
1677 		if (rc) {
1678 			adapter->state = VNIC_PROBED;
1679 			return 0;
1680 		}
1681 
1682 		if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1683 		    adapter->wait_for_reset) {
1684 			rc = init_resources(adapter);
1685 			if (rc)
1686 				return rc;
1687 		} else if (adapter->req_rx_queues != old_num_rx_queues ||
1688 			   adapter->req_tx_queues != old_num_tx_queues) {
1689 			adapter->map_id = 1;
1690 			release_rx_pools(adapter);
1691 			release_tx_pools(adapter);
1692 			init_rx_pools(netdev);
1693 			init_tx_pools(netdev);
1694 		} else {
1695 			rc = reset_tx_pools(adapter);
1696 			if (rc)
1697 				return rc;
1698 
1699 			rc = reset_rx_pools(adapter);
1700 			if (rc)
1701 				return rc;
1702 
1703 			if (reset_state == VNIC_CLOSED)
1704 				return 0;
1705 		}
1706 	}
1707 
1708 	rc = __ibmvnic_open(netdev);
1709 	if (rc) {
1710 		if (list_empty(&adapter->rwi_list))
1711 			adapter->state = VNIC_CLOSED;
1712 		else
1713 			adapter->state = reset_state;
1714 
1715 		return 0;
1716 	}
1717 
1718 	/* kick napi */
1719 	for (i = 0; i < adapter->req_rx_queues; i++)
1720 		napi_schedule(&adapter->napi[i]);
1721 
1722 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1723 		netdev_notify_peers(netdev);
1724 
1725 	netif_carrier_on(netdev);
1726 
1727 	return 0;
1728 }
1729 
1730 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1731 {
1732 	struct ibmvnic_rwi *rwi;
1733 
1734 	mutex_lock(&adapter->rwi_lock);
1735 
1736 	if (!list_empty(&adapter->rwi_list)) {
1737 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1738 				       list);
1739 		list_del(&rwi->list);
1740 	} else {
1741 		rwi = NULL;
1742 	}
1743 
1744 	mutex_unlock(&adapter->rwi_lock);
1745 	return rwi;
1746 }
1747 
1748 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1749 {
1750 	struct ibmvnic_rwi *rwi;
1751 
1752 	rwi = get_next_rwi(adapter);
1753 	while (rwi) {
1754 		kfree(rwi);
1755 		rwi = get_next_rwi(adapter);
1756 	}
1757 }
1758 
1759 static void __ibmvnic_reset(struct work_struct *work)
1760 {
1761 	struct ibmvnic_rwi *rwi;
1762 	struct ibmvnic_adapter *adapter;
1763 	struct net_device *netdev;
1764 	u32 reset_state;
1765 	int rc = 0;
1766 
1767 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1768 	netdev = adapter->netdev;
1769 
1770 	mutex_lock(&adapter->reset_lock);
1771 	adapter->resetting = true;
1772 	reset_state = adapter->state;
1773 
1774 	rwi = get_next_rwi(adapter);
1775 	while (rwi) {
1776 		rc = do_reset(adapter, rwi, reset_state);
1777 		kfree(rwi);
1778 		if (rc && rc != IBMVNIC_INIT_FAILED)
1779 			break;
1780 
1781 		rwi = get_next_rwi(adapter);
1782 	}
1783 
1784 	if (adapter->wait_for_reset) {
1785 		adapter->wait_for_reset = false;
1786 		adapter->reset_done_rc = rc;
1787 		complete(&adapter->reset_done);
1788 	}
1789 
1790 	if (rc) {
1791 		netdev_dbg(adapter->netdev, "Reset failed\n");
1792 		free_all_rwi(adapter);
1793 		mutex_unlock(&adapter->reset_lock);
1794 		return;
1795 	}
1796 
1797 	adapter->resetting = false;
1798 	mutex_unlock(&adapter->reset_lock);
1799 }
1800 
1801 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1802 			  enum ibmvnic_reset_reason reason)
1803 {
1804 	struct ibmvnic_rwi *rwi, *tmp;
1805 	struct net_device *netdev = adapter->netdev;
1806 	struct list_head *entry;
1807 
1808 	if (adapter->state == VNIC_REMOVING ||
1809 	    adapter->state == VNIC_REMOVED) {
1810 		netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1811 		return;
1812 	}
1813 
1814 	if (adapter->state == VNIC_PROBING) {
1815 		netdev_warn(netdev, "Adapter reset during probe\n");
1816 		adapter->init_done_rc = EAGAIN;
1817 		return;
1818 	}
1819 
1820 	mutex_lock(&adapter->rwi_lock);
1821 
1822 	list_for_each(entry, &adapter->rwi_list) {
1823 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
1824 		if (tmp->reset_reason == reason) {
1825 			netdev_dbg(netdev, "Skipping matching reset\n");
1826 			mutex_unlock(&adapter->rwi_lock);
1827 			return;
1828 		}
1829 	}
1830 
1831 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1832 	if (!rwi) {
1833 		mutex_unlock(&adapter->rwi_lock);
1834 		ibmvnic_close(netdev);
1835 		return;
1836 	}
1837 
1838 	rwi->reset_reason = reason;
1839 	list_add_tail(&rwi->list, &adapter->rwi_list);
1840 	mutex_unlock(&adapter->rwi_lock);
1841 
1842 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1843 	schedule_work(&adapter->ibmvnic_reset);
1844 }
1845 
1846 static void ibmvnic_tx_timeout(struct net_device *dev)
1847 {
1848 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1849 
1850 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1851 }
1852 
1853 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1854 				  struct ibmvnic_rx_buff *rx_buff)
1855 {
1856 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1857 
1858 	rx_buff->skb = NULL;
1859 
1860 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1861 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1862 
1863 	atomic_dec(&pool->available);
1864 }
1865 
1866 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1867 {
1868 	struct net_device *netdev = napi->dev;
1869 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1870 	int scrq_num = (int)(napi - adapter->napi);
1871 	int frames_processed = 0;
1872 
1873 restart_poll:
1874 	while (frames_processed < budget) {
1875 		struct sk_buff *skb;
1876 		struct ibmvnic_rx_buff *rx_buff;
1877 		union sub_crq *next;
1878 		u32 length;
1879 		u16 offset;
1880 		u8 flags = 0;
1881 
1882 		if (unlikely(adapter->resetting &&
1883 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
1884 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1885 			napi_complete_done(napi, frames_processed);
1886 			return frames_processed;
1887 		}
1888 
1889 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1890 			break;
1891 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1892 		rx_buff =
1893 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1894 							  rx_comp.correlator);
1895 		/* do error checking */
1896 		if (next->rx_comp.rc) {
1897 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1898 				   be16_to_cpu(next->rx_comp.rc));
1899 			/* free the entry */
1900 			next->rx_comp.first = 0;
1901 			dev_kfree_skb_any(rx_buff->skb);
1902 			remove_buff_from_pool(adapter, rx_buff);
1903 			continue;
1904 		} else if (!rx_buff->skb) {
1905 			/* free the entry */
1906 			next->rx_comp.first = 0;
1907 			remove_buff_from_pool(adapter, rx_buff);
1908 			continue;
1909 		}
1910 
1911 		length = be32_to_cpu(next->rx_comp.len);
1912 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
1913 		flags = next->rx_comp.flags;
1914 		skb = rx_buff->skb;
1915 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
1916 					length);
1917 
1918 		/* VLAN Header has been stripped by the system firmware and
1919 		 * needs to be inserted by the driver
1920 		 */
1921 		if (adapter->rx_vlan_header_insertion &&
1922 		    (flags & IBMVNIC_VLAN_STRIPPED))
1923 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1924 					       ntohs(next->rx_comp.vlan_tci));
1925 
1926 		/* free the entry */
1927 		next->rx_comp.first = 0;
1928 		remove_buff_from_pool(adapter, rx_buff);
1929 
1930 		skb_put(skb, length);
1931 		skb->protocol = eth_type_trans(skb, netdev);
1932 		skb_record_rx_queue(skb, scrq_num);
1933 
1934 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1935 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1936 			skb->ip_summed = CHECKSUM_UNNECESSARY;
1937 		}
1938 
1939 		length = skb->len;
1940 		napi_gro_receive(napi, skb); /* send it up */
1941 		netdev->stats.rx_packets++;
1942 		netdev->stats.rx_bytes += length;
1943 		adapter->rx_stats_buffers[scrq_num].packets++;
1944 		adapter->rx_stats_buffers[scrq_num].bytes += length;
1945 		frames_processed++;
1946 	}
1947 
1948 	if (adapter->state != VNIC_CLOSING)
1949 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1950 
1951 	if (frames_processed < budget) {
1952 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1953 		napi_complete_done(napi, frames_processed);
1954 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1955 		    napi_reschedule(napi)) {
1956 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1957 			goto restart_poll;
1958 		}
1959 	}
1960 	return frames_processed;
1961 }
1962 
1963 #ifdef CONFIG_NET_POLL_CONTROLLER
1964 static void ibmvnic_netpoll_controller(struct net_device *dev)
1965 {
1966 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1967 	int i;
1968 
1969 	replenish_pools(netdev_priv(dev));
1970 	for (i = 0; i < adapter->req_rx_queues; i++)
1971 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1972 				     adapter->rx_scrq[i]);
1973 }
1974 #endif
1975 
1976 static int wait_for_reset(struct ibmvnic_adapter *adapter)
1977 {
1978 	adapter->fallback.mtu = adapter->req_mtu;
1979 	adapter->fallback.rx_queues = adapter->req_rx_queues;
1980 	adapter->fallback.tx_queues = adapter->req_tx_queues;
1981 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
1982 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
1983 
1984 	init_completion(&adapter->reset_done);
1985 	ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1986 	adapter->wait_for_reset = true;
1987 	wait_for_completion(&adapter->reset_done);
1988 
1989 	if (adapter->reset_done_rc) {
1990 		adapter->desired.mtu = adapter->fallback.mtu;
1991 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
1992 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
1993 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
1994 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
1995 
1996 		init_completion(&adapter->reset_done);
1997 		ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
1998 		wait_for_completion(&adapter->reset_done);
1999 	}
2000 	adapter->wait_for_reset = false;
2001 
2002 	return adapter->reset_done_rc;
2003 }
2004 
2005 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2006 {
2007 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2008 
2009 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2010 
2011 	return wait_for_reset(adapter);
2012 }
2013 
2014 static const struct net_device_ops ibmvnic_netdev_ops = {
2015 	.ndo_open		= ibmvnic_open,
2016 	.ndo_stop		= ibmvnic_close,
2017 	.ndo_start_xmit		= ibmvnic_xmit,
2018 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2019 	.ndo_set_mac_address	= ibmvnic_set_mac,
2020 	.ndo_validate_addr	= eth_validate_addr,
2021 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2022 #ifdef CONFIG_NET_POLL_CONTROLLER
2023 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
2024 #endif
2025 	.ndo_change_mtu		= ibmvnic_change_mtu,
2026 };
2027 
2028 /* ethtool functions */
2029 
2030 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2031 				      struct ethtool_link_ksettings *cmd)
2032 {
2033 	u32 supported, advertising;
2034 
2035 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2036 			  SUPPORTED_FIBRE);
2037 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2038 			    ADVERTISED_FIBRE);
2039 	cmd->base.speed = SPEED_1000;
2040 	cmd->base.duplex = DUPLEX_FULL;
2041 	cmd->base.port = PORT_FIBRE;
2042 	cmd->base.phy_address = 0;
2043 	cmd->base.autoneg = AUTONEG_ENABLE;
2044 
2045 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2046 						supported);
2047 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2048 						advertising);
2049 
2050 	return 0;
2051 }
2052 
2053 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2054 				struct ethtool_drvinfo *info)
2055 {
2056 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2057 
2058 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2059 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2060 	strlcpy(info->fw_version, adapter->fw_version,
2061 		sizeof(info->fw_version));
2062 }
2063 
2064 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2065 {
2066 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2067 
2068 	return adapter->msg_enable;
2069 }
2070 
2071 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2072 {
2073 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2074 
2075 	adapter->msg_enable = data;
2076 }
2077 
2078 static u32 ibmvnic_get_link(struct net_device *netdev)
2079 {
2080 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2081 
2082 	/* Don't need to send a query because we request a logical link up at
2083 	 * init and then we wait for link state indications
2084 	 */
2085 	return adapter->logical_link_state;
2086 }
2087 
2088 static void ibmvnic_get_ringparam(struct net_device *netdev,
2089 				  struct ethtool_ringparam *ring)
2090 {
2091 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2092 
2093 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2094 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2095 	ring->rx_mini_max_pending = 0;
2096 	ring->rx_jumbo_max_pending = 0;
2097 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2098 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2099 	ring->rx_mini_pending = 0;
2100 	ring->rx_jumbo_pending = 0;
2101 }
2102 
2103 static int ibmvnic_set_ringparam(struct net_device *netdev,
2104 				 struct ethtool_ringparam *ring)
2105 {
2106 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2107 
2108 	if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2109 	    ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2110 		netdev_err(netdev, "Invalid request.\n");
2111 		netdev_err(netdev, "Max tx buffers = %llu\n",
2112 			   adapter->max_rx_add_entries_per_subcrq);
2113 		netdev_err(netdev, "Max rx buffers = %llu\n",
2114 			   adapter->max_tx_entries_per_subcrq);
2115 		return -EINVAL;
2116 	}
2117 
2118 	adapter->desired.rx_entries = ring->rx_pending;
2119 	adapter->desired.tx_entries = ring->tx_pending;
2120 
2121 	return wait_for_reset(adapter);
2122 }
2123 
2124 static void ibmvnic_get_channels(struct net_device *netdev,
2125 				 struct ethtool_channels *channels)
2126 {
2127 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2128 
2129 	channels->max_rx = adapter->max_rx_queues;
2130 	channels->max_tx = adapter->max_tx_queues;
2131 	channels->max_other = 0;
2132 	channels->max_combined = 0;
2133 	channels->rx_count = adapter->req_rx_queues;
2134 	channels->tx_count = adapter->req_tx_queues;
2135 	channels->other_count = 0;
2136 	channels->combined_count = 0;
2137 }
2138 
2139 static int ibmvnic_set_channels(struct net_device *netdev,
2140 				struct ethtool_channels *channels)
2141 {
2142 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2143 
2144 	adapter->desired.rx_queues = channels->rx_count;
2145 	adapter->desired.tx_queues = channels->tx_count;
2146 
2147 	return wait_for_reset(adapter);
2148 }
2149 
2150 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2151 {
2152 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2153 	int i;
2154 
2155 	if (stringset != ETH_SS_STATS)
2156 		return;
2157 
2158 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2159 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2160 
2161 	for (i = 0; i < adapter->req_tx_queues; i++) {
2162 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2163 		data += ETH_GSTRING_LEN;
2164 
2165 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2166 		data += ETH_GSTRING_LEN;
2167 
2168 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2169 		data += ETH_GSTRING_LEN;
2170 	}
2171 
2172 	for (i = 0; i < adapter->req_rx_queues; i++) {
2173 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2174 		data += ETH_GSTRING_LEN;
2175 
2176 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2177 		data += ETH_GSTRING_LEN;
2178 
2179 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2180 		data += ETH_GSTRING_LEN;
2181 	}
2182 }
2183 
2184 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2185 {
2186 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2187 
2188 	switch (sset) {
2189 	case ETH_SS_STATS:
2190 		return ARRAY_SIZE(ibmvnic_stats) +
2191 		       adapter->req_tx_queues * NUM_TX_STATS +
2192 		       adapter->req_rx_queues * NUM_RX_STATS;
2193 	default:
2194 		return -EOPNOTSUPP;
2195 	}
2196 }
2197 
2198 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2199 				      struct ethtool_stats *stats, u64 *data)
2200 {
2201 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2202 	union ibmvnic_crq crq;
2203 	int i, j;
2204 
2205 	memset(&crq, 0, sizeof(crq));
2206 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2207 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2208 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2209 	crq.request_statistics.len =
2210 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2211 
2212 	/* Wait for data to be written */
2213 	init_completion(&adapter->stats_done);
2214 	ibmvnic_send_crq(adapter, &crq);
2215 	wait_for_completion(&adapter->stats_done);
2216 
2217 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2218 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2219 						ibmvnic_stats[i].offset));
2220 
2221 	for (j = 0; j < adapter->req_tx_queues; j++) {
2222 		data[i] = adapter->tx_stats_buffers[j].packets;
2223 		i++;
2224 		data[i] = adapter->tx_stats_buffers[j].bytes;
2225 		i++;
2226 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2227 		i++;
2228 	}
2229 
2230 	for (j = 0; j < adapter->req_rx_queues; j++) {
2231 		data[i] = adapter->rx_stats_buffers[j].packets;
2232 		i++;
2233 		data[i] = adapter->rx_stats_buffers[j].bytes;
2234 		i++;
2235 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2236 		i++;
2237 	}
2238 }
2239 
2240 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2241 	.get_drvinfo		= ibmvnic_get_drvinfo,
2242 	.get_msglevel		= ibmvnic_get_msglevel,
2243 	.set_msglevel		= ibmvnic_set_msglevel,
2244 	.get_link		= ibmvnic_get_link,
2245 	.get_ringparam		= ibmvnic_get_ringparam,
2246 	.set_ringparam		= ibmvnic_set_ringparam,
2247 	.get_channels		= ibmvnic_get_channels,
2248 	.set_channels		= ibmvnic_set_channels,
2249 	.get_strings            = ibmvnic_get_strings,
2250 	.get_sset_count         = ibmvnic_get_sset_count,
2251 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2252 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2253 };
2254 
2255 /* Routines for managing CRQs/sCRQs  */
2256 
2257 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2258 				   struct ibmvnic_sub_crq_queue *scrq)
2259 {
2260 	int rc;
2261 
2262 	if (scrq->irq) {
2263 		free_irq(scrq->irq, scrq);
2264 		irq_dispose_mapping(scrq->irq);
2265 		scrq->irq = 0;
2266 	}
2267 
2268 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2269 	scrq->cur = 0;
2270 
2271 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2272 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2273 	return rc;
2274 }
2275 
2276 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2277 {
2278 	int i, rc;
2279 
2280 	for (i = 0; i < adapter->req_tx_queues; i++) {
2281 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2282 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2283 		if (rc)
2284 			return rc;
2285 	}
2286 
2287 	for (i = 0; i < adapter->req_rx_queues; i++) {
2288 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2289 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2290 		if (rc)
2291 			return rc;
2292 	}
2293 
2294 	return rc;
2295 }
2296 
2297 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2298 				  struct ibmvnic_sub_crq_queue *scrq)
2299 {
2300 	struct device *dev = &adapter->vdev->dev;
2301 	long rc;
2302 
2303 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2304 
2305 	/* Close the sub-crqs */
2306 	do {
2307 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2308 					adapter->vdev->unit_address,
2309 					scrq->crq_num);
2310 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2311 
2312 	if (rc) {
2313 		netdev_err(adapter->netdev,
2314 			   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2315 			   scrq->crq_num, rc);
2316 	}
2317 
2318 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2319 			 DMA_BIDIRECTIONAL);
2320 	free_pages((unsigned long)scrq->msgs, 2);
2321 	kfree(scrq);
2322 }
2323 
2324 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2325 							*adapter)
2326 {
2327 	struct device *dev = &adapter->vdev->dev;
2328 	struct ibmvnic_sub_crq_queue *scrq;
2329 	int rc;
2330 
2331 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2332 	if (!scrq)
2333 		return NULL;
2334 
2335 	scrq->msgs =
2336 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2337 	if (!scrq->msgs) {
2338 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2339 		goto zero_page_failed;
2340 	}
2341 
2342 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2343 					 DMA_BIDIRECTIONAL);
2344 	if (dma_mapping_error(dev, scrq->msg_token)) {
2345 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2346 		goto map_failed;
2347 	}
2348 
2349 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2350 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2351 
2352 	if (rc == H_RESOURCE)
2353 		rc = ibmvnic_reset_crq(adapter);
2354 
2355 	if (rc == H_CLOSED) {
2356 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2357 	} else if (rc) {
2358 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2359 		goto reg_failed;
2360 	}
2361 
2362 	scrq->adapter = adapter;
2363 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2364 	spin_lock_init(&scrq->lock);
2365 
2366 	netdev_dbg(adapter->netdev,
2367 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2368 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2369 
2370 	return scrq;
2371 
2372 reg_failed:
2373 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2374 			 DMA_BIDIRECTIONAL);
2375 map_failed:
2376 	free_pages((unsigned long)scrq->msgs, 2);
2377 zero_page_failed:
2378 	kfree(scrq);
2379 
2380 	return NULL;
2381 }
2382 
2383 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
2384 {
2385 	int i;
2386 
2387 	if (adapter->tx_scrq) {
2388 		for (i = 0; i < adapter->req_tx_queues; i++) {
2389 			if (!adapter->tx_scrq[i])
2390 				continue;
2391 
2392 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2393 				   i);
2394 			if (adapter->tx_scrq[i]->irq) {
2395 				free_irq(adapter->tx_scrq[i]->irq,
2396 					 adapter->tx_scrq[i]);
2397 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2398 				adapter->tx_scrq[i]->irq = 0;
2399 			}
2400 
2401 			release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2402 		}
2403 
2404 		kfree(adapter->tx_scrq);
2405 		adapter->tx_scrq = NULL;
2406 	}
2407 
2408 	if (adapter->rx_scrq) {
2409 		for (i = 0; i < adapter->req_rx_queues; i++) {
2410 			if (!adapter->rx_scrq[i])
2411 				continue;
2412 
2413 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2414 				   i);
2415 			if (adapter->rx_scrq[i]->irq) {
2416 				free_irq(adapter->rx_scrq[i]->irq,
2417 					 adapter->rx_scrq[i]);
2418 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2419 				adapter->rx_scrq[i]->irq = 0;
2420 			}
2421 
2422 			release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2423 		}
2424 
2425 		kfree(adapter->rx_scrq);
2426 		adapter->rx_scrq = NULL;
2427 	}
2428 }
2429 
2430 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2431 			    struct ibmvnic_sub_crq_queue *scrq)
2432 {
2433 	struct device *dev = &adapter->vdev->dev;
2434 	unsigned long rc;
2435 
2436 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2437 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2438 	if (rc)
2439 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2440 			scrq->hw_irq, rc);
2441 	return rc;
2442 }
2443 
2444 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2445 			   struct ibmvnic_sub_crq_queue *scrq)
2446 {
2447 	struct device *dev = &adapter->vdev->dev;
2448 	unsigned long rc;
2449 
2450 	if (scrq->hw_irq > 0x100000000ULL) {
2451 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2452 		return 1;
2453 	}
2454 
2455 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2456 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2457 	if (rc)
2458 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2459 			scrq->hw_irq, rc);
2460 	return rc;
2461 }
2462 
2463 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2464 			       struct ibmvnic_sub_crq_queue *scrq)
2465 {
2466 	struct device *dev = &adapter->vdev->dev;
2467 	struct ibmvnic_tx_buff *txbuff;
2468 	union sub_crq *next;
2469 	int index;
2470 	int i, j;
2471 	u8 first;
2472 
2473 restart_loop:
2474 	while (pending_scrq(adapter, scrq)) {
2475 		unsigned int pool = scrq->pool_index;
2476 
2477 		next = ibmvnic_next_scrq(adapter, scrq);
2478 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2479 			if (next->tx_comp.rcs[i]) {
2480 				dev_err(dev, "tx error %x\n",
2481 					next->tx_comp.rcs[i]);
2482 				continue;
2483 			}
2484 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2485 			txbuff = &adapter->tx_pool[pool].tx_buff[index];
2486 
2487 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2488 				if (!txbuff->data_dma[j])
2489 					continue;
2490 
2491 				txbuff->data_dma[j] = 0;
2492 			}
2493 			/* if sub_crq was sent indirectly */
2494 			first = txbuff->indir_arr[0].generic.first;
2495 			if (first == IBMVNIC_CRQ_CMD) {
2496 				dma_unmap_single(dev, txbuff->indir_dma,
2497 						 sizeof(txbuff->indir_arr),
2498 						 DMA_TO_DEVICE);
2499 			}
2500 
2501 			if (txbuff->last_frag) {
2502 				dev_kfree_skb_any(txbuff->skb);
2503 				txbuff->skb = NULL;
2504 			}
2505 
2506 			adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2507 						     producer_index] = index;
2508 			adapter->tx_pool[pool].producer_index =
2509 			    (adapter->tx_pool[pool].producer_index + 1) %
2510 			    adapter->req_tx_entries_per_subcrq;
2511 		}
2512 		/* remove tx_comp scrq*/
2513 		next->tx_comp.first = 0;
2514 
2515 		if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2516 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2517 		    __netif_subqueue_stopped(adapter->netdev,
2518 					     scrq->pool_index)) {
2519 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2520 			netdev_info(adapter->netdev, "Started queue %d\n",
2521 				    scrq->pool_index);
2522 		}
2523 	}
2524 
2525 	enable_scrq_irq(adapter, scrq);
2526 
2527 	if (pending_scrq(adapter, scrq)) {
2528 		disable_scrq_irq(adapter, scrq);
2529 		goto restart_loop;
2530 	}
2531 
2532 	return 0;
2533 }
2534 
2535 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2536 {
2537 	struct ibmvnic_sub_crq_queue *scrq = instance;
2538 	struct ibmvnic_adapter *adapter = scrq->adapter;
2539 
2540 	disable_scrq_irq(adapter, scrq);
2541 	ibmvnic_complete_tx(adapter, scrq);
2542 
2543 	return IRQ_HANDLED;
2544 }
2545 
2546 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2547 {
2548 	struct ibmvnic_sub_crq_queue *scrq = instance;
2549 	struct ibmvnic_adapter *adapter = scrq->adapter;
2550 
2551 	/* When booting a kdump kernel we can hit pending interrupts
2552 	 * prior to completing driver initialization.
2553 	 */
2554 	if (unlikely(adapter->state != VNIC_OPEN))
2555 		return IRQ_NONE;
2556 
2557 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2558 
2559 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2560 		disable_scrq_irq(adapter, scrq);
2561 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2562 	}
2563 
2564 	return IRQ_HANDLED;
2565 }
2566 
2567 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2568 {
2569 	struct device *dev = &adapter->vdev->dev;
2570 	struct ibmvnic_sub_crq_queue *scrq;
2571 	int i = 0, j = 0;
2572 	int rc = 0;
2573 
2574 	for (i = 0; i < adapter->req_tx_queues; i++) {
2575 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2576 			   i);
2577 		scrq = adapter->tx_scrq[i];
2578 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2579 
2580 		if (!scrq->irq) {
2581 			rc = -EINVAL;
2582 			dev_err(dev, "Error mapping irq\n");
2583 			goto req_tx_irq_failed;
2584 		}
2585 
2586 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2587 				 0, "ibmvnic_tx", scrq);
2588 
2589 		if (rc) {
2590 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2591 				scrq->irq, rc);
2592 			irq_dispose_mapping(scrq->irq);
2593 			goto req_rx_irq_failed;
2594 		}
2595 	}
2596 
2597 	for (i = 0; i < adapter->req_rx_queues; i++) {
2598 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2599 			   i);
2600 		scrq = adapter->rx_scrq[i];
2601 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2602 		if (!scrq->irq) {
2603 			rc = -EINVAL;
2604 			dev_err(dev, "Error mapping irq\n");
2605 			goto req_rx_irq_failed;
2606 		}
2607 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2608 				 0, "ibmvnic_rx", scrq);
2609 		if (rc) {
2610 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2611 				scrq->irq, rc);
2612 			irq_dispose_mapping(scrq->irq);
2613 			goto req_rx_irq_failed;
2614 		}
2615 	}
2616 	return rc;
2617 
2618 req_rx_irq_failed:
2619 	for (j = 0; j < i; j++) {
2620 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2621 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2622 	}
2623 	i = adapter->req_tx_queues;
2624 req_tx_irq_failed:
2625 	for (j = 0; j < i; j++) {
2626 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2627 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2628 	}
2629 	release_sub_crqs(adapter);
2630 	return rc;
2631 }
2632 
2633 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2634 {
2635 	struct device *dev = &adapter->vdev->dev;
2636 	struct ibmvnic_sub_crq_queue **allqueues;
2637 	int registered_queues = 0;
2638 	int total_queues;
2639 	int more = 0;
2640 	int i;
2641 
2642 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2643 
2644 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2645 	if (!allqueues)
2646 		return -1;
2647 
2648 	for (i = 0; i < total_queues; i++) {
2649 		allqueues[i] = init_sub_crq_queue(adapter);
2650 		if (!allqueues[i]) {
2651 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2652 			break;
2653 		}
2654 		registered_queues++;
2655 	}
2656 
2657 	/* Make sure we were able to register the minimum number of queues */
2658 	if (registered_queues <
2659 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2660 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2661 		goto tx_failed;
2662 	}
2663 
2664 	/* Distribute the failed allocated queues*/
2665 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2666 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2667 		switch (i % 3) {
2668 		case 0:
2669 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2670 				adapter->req_rx_queues--;
2671 			else
2672 				more++;
2673 			break;
2674 		case 1:
2675 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2676 				adapter->req_tx_queues--;
2677 			else
2678 				more++;
2679 			break;
2680 		}
2681 	}
2682 
2683 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2684 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2685 	if (!adapter->tx_scrq)
2686 		goto tx_failed;
2687 
2688 	for (i = 0; i < adapter->req_tx_queues; i++) {
2689 		adapter->tx_scrq[i] = allqueues[i];
2690 		adapter->tx_scrq[i]->pool_index = i;
2691 	}
2692 
2693 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2694 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2695 	if (!adapter->rx_scrq)
2696 		goto rx_failed;
2697 
2698 	for (i = 0; i < adapter->req_rx_queues; i++) {
2699 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2700 		adapter->rx_scrq[i]->scrq_num = i;
2701 	}
2702 
2703 	kfree(allqueues);
2704 	return 0;
2705 
2706 rx_failed:
2707 	kfree(adapter->tx_scrq);
2708 	adapter->tx_scrq = NULL;
2709 tx_failed:
2710 	for (i = 0; i < registered_queues; i++)
2711 		release_sub_crq_queue(adapter, allqueues[i]);
2712 	kfree(allqueues);
2713 	return -1;
2714 }
2715 
2716 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2717 {
2718 	struct device *dev = &adapter->vdev->dev;
2719 	union ibmvnic_crq crq;
2720 	int max_entries;
2721 
2722 	if (!retry) {
2723 		/* Sub-CRQ entries are 32 byte long */
2724 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2725 
2726 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
2727 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
2728 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2729 			return;
2730 		}
2731 
2732 		if (adapter->desired.mtu)
2733 			adapter->req_mtu = adapter->desired.mtu;
2734 		else
2735 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2736 
2737 		if (!adapter->desired.tx_entries)
2738 			adapter->desired.tx_entries =
2739 					adapter->max_tx_entries_per_subcrq;
2740 		if (!adapter->desired.rx_entries)
2741 			adapter->desired.rx_entries =
2742 					adapter->max_rx_add_entries_per_subcrq;
2743 
2744 		max_entries = IBMVNIC_MAX_LTB_SIZE /
2745 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2746 
2747 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2748 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2749 			adapter->desired.tx_entries = max_entries;
2750 		}
2751 
2752 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2753 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2754 			adapter->desired.rx_entries = max_entries;
2755 		}
2756 
2757 		if (adapter->desired.tx_entries)
2758 			adapter->req_tx_entries_per_subcrq =
2759 					adapter->desired.tx_entries;
2760 		else
2761 			adapter->req_tx_entries_per_subcrq =
2762 					adapter->max_tx_entries_per_subcrq;
2763 
2764 		if (adapter->desired.rx_entries)
2765 			adapter->req_rx_add_entries_per_subcrq =
2766 					adapter->desired.rx_entries;
2767 		else
2768 			adapter->req_rx_add_entries_per_subcrq =
2769 					adapter->max_rx_add_entries_per_subcrq;
2770 
2771 		if (adapter->desired.tx_queues)
2772 			adapter->req_tx_queues =
2773 					adapter->desired.tx_queues;
2774 		else
2775 			adapter->req_tx_queues =
2776 					adapter->opt_tx_comp_sub_queues;
2777 
2778 		if (adapter->desired.rx_queues)
2779 			adapter->req_rx_queues =
2780 					adapter->desired.rx_queues;
2781 		else
2782 			adapter->req_rx_queues =
2783 					adapter->opt_rx_comp_queues;
2784 
2785 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2786 	}
2787 
2788 	memset(&crq, 0, sizeof(crq));
2789 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
2790 	crq.request_capability.cmd = REQUEST_CAPABILITY;
2791 
2792 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2793 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2794 	atomic_inc(&adapter->running_cap_crqs);
2795 	ibmvnic_send_crq(adapter, &crq);
2796 
2797 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2798 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2799 	atomic_inc(&adapter->running_cap_crqs);
2800 	ibmvnic_send_crq(adapter, &crq);
2801 
2802 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2803 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2804 	atomic_inc(&adapter->running_cap_crqs);
2805 	ibmvnic_send_crq(adapter, &crq);
2806 
2807 	crq.request_capability.capability =
2808 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2809 	crq.request_capability.number =
2810 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2811 	atomic_inc(&adapter->running_cap_crqs);
2812 	ibmvnic_send_crq(adapter, &crq);
2813 
2814 	crq.request_capability.capability =
2815 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2816 	crq.request_capability.number =
2817 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2818 	atomic_inc(&adapter->running_cap_crqs);
2819 	ibmvnic_send_crq(adapter, &crq);
2820 
2821 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2822 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2823 	atomic_inc(&adapter->running_cap_crqs);
2824 	ibmvnic_send_crq(adapter, &crq);
2825 
2826 	if (adapter->netdev->flags & IFF_PROMISC) {
2827 		if (adapter->promisc_supported) {
2828 			crq.request_capability.capability =
2829 			    cpu_to_be16(PROMISC_REQUESTED);
2830 			crq.request_capability.number = cpu_to_be64(1);
2831 			atomic_inc(&adapter->running_cap_crqs);
2832 			ibmvnic_send_crq(adapter, &crq);
2833 		}
2834 	} else {
2835 		crq.request_capability.capability =
2836 		    cpu_to_be16(PROMISC_REQUESTED);
2837 		crq.request_capability.number = cpu_to_be64(0);
2838 		atomic_inc(&adapter->running_cap_crqs);
2839 		ibmvnic_send_crq(adapter, &crq);
2840 	}
2841 }
2842 
2843 static int pending_scrq(struct ibmvnic_adapter *adapter,
2844 			struct ibmvnic_sub_crq_queue *scrq)
2845 {
2846 	union sub_crq *entry = &scrq->msgs[scrq->cur];
2847 
2848 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2849 		return 1;
2850 	else
2851 		return 0;
2852 }
2853 
2854 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2855 					struct ibmvnic_sub_crq_queue *scrq)
2856 {
2857 	union sub_crq *entry;
2858 	unsigned long flags;
2859 
2860 	spin_lock_irqsave(&scrq->lock, flags);
2861 	entry = &scrq->msgs[scrq->cur];
2862 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2863 		if (++scrq->cur == scrq->size)
2864 			scrq->cur = 0;
2865 	} else {
2866 		entry = NULL;
2867 	}
2868 	spin_unlock_irqrestore(&scrq->lock, flags);
2869 
2870 	return entry;
2871 }
2872 
2873 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2874 {
2875 	struct ibmvnic_crq_queue *queue = &adapter->crq;
2876 	union ibmvnic_crq *crq;
2877 
2878 	crq = &queue->msgs[queue->cur];
2879 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2880 		if (++queue->cur == queue->size)
2881 			queue->cur = 0;
2882 	} else {
2883 		crq = NULL;
2884 	}
2885 
2886 	return crq;
2887 }
2888 
2889 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2890 		       union sub_crq *sub_crq)
2891 {
2892 	unsigned int ua = adapter->vdev->unit_address;
2893 	struct device *dev = &adapter->vdev->dev;
2894 	u64 *u64_crq = (u64 *)sub_crq;
2895 	int rc;
2896 
2897 	netdev_dbg(adapter->netdev,
2898 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2899 		   (unsigned long int)cpu_to_be64(remote_handle),
2900 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2901 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
2902 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
2903 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
2904 
2905 	/* Make sure the hypervisor sees the complete request */
2906 	mb();
2907 
2908 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2909 				cpu_to_be64(remote_handle),
2910 				cpu_to_be64(u64_crq[0]),
2911 				cpu_to_be64(u64_crq[1]),
2912 				cpu_to_be64(u64_crq[2]),
2913 				cpu_to_be64(u64_crq[3]));
2914 
2915 	if (rc) {
2916 		if (rc == H_CLOSED)
2917 			dev_warn(dev, "CRQ Queue closed\n");
2918 		dev_err(dev, "Send error (rc=%d)\n", rc);
2919 	}
2920 
2921 	return rc;
2922 }
2923 
2924 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2925 				u64 remote_handle, u64 ioba, u64 num_entries)
2926 {
2927 	unsigned int ua = adapter->vdev->unit_address;
2928 	struct device *dev = &adapter->vdev->dev;
2929 	int rc;
2930 
2931 	/* Make sure the hypervisor sees the complete request */
2932 	mb();
2933 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2934 				cpu_to_be64(remote_handle),
2935 				ioba, num_entries);
2936 
2937 	if (rc) {
2938 		if (rc == H_CLOSED)
2939 			dev_warn(dev, "CRQ Queue closed\n");
2940 		dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2941 	}
2942 
2943 	return rc;
2944 }
2945 
2946 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2947 			    union ibmvnic_crq *crq)
2948 {
2949 	unsigned int ua = adapter->vdev->unit_address;
2950 	struct device *dev = &adapter->vdev->dev;
2951 	u64 *u64_crq = (u64 *)crq;
2952 	int rc;
2953 
2954 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2955 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2956 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
2957 
2958 	/* Make sure the hypervisor sees the complete request */
2959 	mb();
2960 
2961 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2962 				cpu_to_be64(u64_crq[0]),
2963 				cpu_to_be64(u64_crq[1]));
2964 
2965 	if (rc) {
2966 		if (rc == H_CLOSED) {
2967 			dev_warn(dev, "CRQ Queue closed\n");
2968 			if (adapter->resetting)
2969 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
2970 		}
2971 
2972 		dev_warn(dev, "Send error (rc=%d)\n", rc);
2973 	}
2974 
2975 	return rc;
2976 }
2977 
2978 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2979 {
2980 	union ibmvnic_crq crq;
2981 
2982 	memset(&crq, 0, sizeof(crq));
2983 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2984 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
2985 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2986 
2987 	return ibmvnic_send_crq(adapter, &crq);
2988 }
2989 
2990 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2991 {
2992 	union ibmvnic_crq crq;
2993 
2994 	memset(&crq, 0, sizeof(crq));
2995 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2996 	crq.version_exchange.cmd = VERSION_EXCHANGE;
2997 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2998 
2999 	return ibmvnic_send_crq(adapter, &crq);
3000 }
3001 
3002 struct vnic_login_client_data {
3003 	u8	type;
3004 	__be16	len;
3005 	char	name;
3006 } __packed;
3007 
3008 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3009 {
3010 	int len;
3011 
3012 	/* Calculate the amount of buffer space needed for the
3013 	 * vnic client data in the login buffer. There are four entries,
3014 	 * OS name, LPAR name, device name, and a null last entry.
3015 	 */
3016 	len = 4 * sizeof(struct vnic_login_client_data);
3017 	len += 6; /* "Linux" plus NULL */
3018 	len += strlen(utsname()->nodename) + 1;
3019 	len += strlen(adapter->netdev->name) + 1;
3020 
3021 	return len;
3022 }
3023 
3024 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3025 				 struct vnic_login_client_data *vlcd)
3026 {
3027 	const char *os_name = "Linux";
3028 	int len;
3029 
3030 	/* Type 1 - LPAR OS */
3031 	vlcd->type = 1;
3032 	len = strlen(os_name) + 1;
3033 	vlcd->len = cpu_to_be16(len);
3034 	strncpy(&vlcd->name, os_name, len);
3035 	vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
3036 
3037 	/* Type 2 - LPAR name */
3038 	vlcd->type = 2;
3039 	len = strlen(utsname()->nodename) + 1;
3040 	vlcd->len = cpu_to_be16(len);
3041 	strncpy(&vlcd->name, utsname()->nodename, len);
3042 	vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
3043 
3044 	/* Type 3 - device name */
3045 	vlcd->type = 3;
3046 	len = strlen(adapter->netdev->name) + 1;
3047 	vlcd->len = cpu_to_be16(len);
3048 	strncpy(&vlcd->name, adapter->netdev->name, len);
3049 }
3050 
3051 static void send_login(struct ibmvnic_adapter *adapter)
3052 {
3053 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3054 	struct ibmvnic_login_buffer *login_buffer;
3055 	struct device *dev = &adapter->vdev->dev;
3056 	dma_addr_t rsp_buffer_token;
3057 	dma_addr_t buffer_token;
3058 	size_t rsp_buffer_size;
3059 	union ibmvnic_crq crq;
3060 	size_t buffer_size;
3061 	__be64 *tx_list_p;
3062 	__be64 *rx_list_p;
3063 	int client_data_len;
3064 	struct vnic_login_client_data *vlcd;
3065 	int i;
3066 
3067 	release_login_rsp_buffer(adapter);
3068 	client_data_len = vnic_client_data_len(adapter);
3069 
3070 	buffer_size =
3071 	    sizeof(struct ibmvnic_login_buffer) +
3072 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3073 	    client_data_len;
3074 
3075 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3076 	if (!login_buffer)
3077 		goto buf_alloc_failed;
3078 
3079 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3080 				      DMA_TO_DEVICE);
3081 	if (dma_mapping_error(dev, buffer_token)) {
3082 		dev_err(dev, "Couldn't map login buffer\n");
3083 		goto buf_map_failed;
3084 	}
3085 
3086 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3087 			  sizeof(u64) * adapter->req_tx_queues +
3088 			  sizeof(u64) * adapter->req_rx_queues +
3089 			  sizeof(u64) * adapter->req_rx_queues +
3090 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3091 
3092 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3093 	if (!login_rsp_buffer)
3094 		goto buf_rsp_alloc_failed;
3095 
3096 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3097 					  rsp_buffer_size, DMA_FROM_DEVICE);
3098 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3099 		dev_err(dev, "Couldn't map login rsp buffer\n");
3100 		goto buf_rsp_map_failed;
3101 	}
3102 
3103 	adapter->login_buf = login_buffer;
3104 	adapter->login_buf_token = buffer_token;
3105 	adapter->login_buf_sz = buffer_size;
3106 	adapter->login_rsp_buf = login_rsp_buffer;
3107 	adapter->login_rsp_buf_token = rsp_buffer_token;
3108 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3109 
3110 	login_buffer->len = cpu_to_be32(buffer_size);
3111 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3112 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3113 	login_buffer->off_txcomp_subcrqs =
3114 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3115 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3116 	login_buffer->off_rxcomp_subcrqs =
3117 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3118 			sizeof(u64) * adapter->req_tx_queues);
3119 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3120 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3121 
3122 	tx_list_p = (__be64 *)((char *)login_buffer +
3123 				      sizeof(struct ibmvnic_login_buffer));
3124 	rx_list_p = (__be64 *)((char *)login_buffer +
3125 				      sizeof(struct ibmvnic_login_buffer) +
3126 				      sizeof(u64) * adapter->req_tx_queues);
3127 
3128 	for (i = 0; i < adapter->req_tx_queues; i++) {
3129 		if (adapter->tx_scrq[i]) {
3130 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3131 						   crq_num);
3132 		}
3133 	}
3134 
3135 	for (i = 0; i < adapter->req_rx_queues; i++) {
3136 		if (adapter->rx_scrq[i]) {
3137 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3138 						   crq_num);
3139 		}
3140 	}
3141 
3142 	/* Insert vNIC login client data */
3143 	vlcd = (struct vnic_login_client_data *)
3144 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3145 	login_buffer->client_data_offset =
3146 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3147 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3148 
3149 	vnic_add_client_data(adapter, vlcd);
3150 
3151 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3152 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3153 		netdev_dbg(adapter->netdev, "%016lx\n",
3154 			   ((unsigned long int *)(adapter->login_buf))[i]);
3155 	}
3156 
3157 	memset(&crq, 0, sizeof(crq));
3158 	crq.login.first = IBMVNIC_CRQ_CMD;
3159 	crq.login.cmd = LOGIN;
3160 	crq.login.ioba = cpu_to_be32(buffer_token);
3161 	crq.login.len = cpu_to_be32(buffer_size);
3162 	ibmvnic_send_crq(adapter, &crq);
3163 
3164 	return;
3165 
3166 buf_rsp_map_failed:
3167 	kfree(login_rsp_buffer);
3168 buf_rsp_alloc_failed:
3169 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3170 buf_map_failed:
3171 	kfree(login_buffer);
3172 buf_alloc_failed:
3173 	return;
3174 }
3175 
3176 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3177 			     u32 len, u8 map_id)
3178 {
3179 	union ibmvnic_crq crq;
3180 
3181 	memset(&crq, 0, sizeof(crq));
3182 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3183 	crq.request_map.cmd = REQUEST_MAP;
3184 	crq.request_map.map_id = map_id;
3185 	crq.request_map.ioba = cpu_to_be32(addr);
3186 	crq.request_map.len = cpu_to_be32(len);
3187 	ibmvnic_send_crq(adapter, &crq);
3188 }
3189 
3190 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3191 {
3192 	union ibmvnic_crq crq;
3193 
3194 	memset(&crq, 0, sizeof(crq));
3195 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3196 	crq.request_unmap.cmd = REQUEST_UNMAP;
3197 	crq.request_unmap.map_id = map_id;
3198 	ibmvnic_send_crq(adapter, &crq);
3199 }
3200 
3201 static void send_map_query(struct ibmvnic_adapter *adapter)
3202 {
3203 	union ibmvnic_crq crq;
3204 
3205 	memset(&crq, 0, sizeof(crq));
3206 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3207 	crq.query_map.cmd = QUERY_MAP;
3208 	ibmvnic_send_crq(adapter, &crq);
3209 }
3210 
3211 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3212 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3213 {
3214 	union ibmvnic_crq crq;
3215 
3216 	atomic_set(&adapter->running_cap_crqs, 0);
3217 	memset(&crq, 0, sizeof(crq));
3218 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3219 	crq.query_capability.cmd = QUERY_CAPABILITY;
3220 
3221 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3222 	atomic_inc(&adapter->running_cap_crqs);
3223 	ibmvnic_send_crq(adapter, &crq);
3224 
3225 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3226 	atomic_inc(&adapter->running_cap_crqs);
3227 	ibmvnic_send_crq(adapter, &crq);
3228 
3229 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3230 	atomic_inc(&adapter->running_cap_crqs);
3231 	ibmvnic_send_crq(adapter, &crq);
3232 
3233 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3234 	atomic_inc(&adapter->running_cap_crqs);
3235 	ibmvnic_send_crq(adapter, &crq);
3236 
3237 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3238 	atomic_inc(&adapter->running_cap_crqs);
3239 	ibmvnic_send_crq(adapter, &crq);
3240 
3241 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3242 	atomic_inc(&adapter->running_cap_crqs);
3243 	ibmvnic_send_crq(adapter, &crq);
3244 
3245 	crq.query_capability.capability =
3246 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3247 	atomic_inc(&adapter->running_cap_crqs);
3248 	ibmvnic_send_crq(adapter, &crq);
3249 
3250 	crq.query_capability.capability =
3251 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3252 	atomic_inc(&adapter->running_cap_crqs);
3253 	ibmvnic_send_crq(adapter, &crq);
3254 
3255 	crq.query_capability.capability =
3256 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3257 	atomic_inc(&adapter->running_cap_crqs);
3258 	ibmvnic_send_crq(adapter, &crq);
3259 
3260 	crq.query_capability.capability =
3261 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3262 	atomic_inc(&adapter->running_cap_crqs);
3263 	ibmvnic_send_crq(adapter, &crq);
3264 
3265 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3266 	atomic_inc(&adapter->running_cap_crqs);
3267 	ibmvnic_send_crq(adapter, &crq);
3268 
3269 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3270 	atomic_inc(&adapter->running_cap_crqs);
3271 	ibmvnic_send_crq(adapter, &crq);
3272 
3273 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3274 	atomic_inc(&adapter->running_cap_crqs);
3275 	ibmvnic_send_crq(adapter, &crq);
3276 
3277 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3278 	atomic_inc(&adapter->running_cap_crqs);
3279 	ibmvnic_send_crq(adapter, &crq);
3280 
3281 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3282 	atomic_inc(&adapter->running_cap_crqs);
3283 	ibmvnic_send_crq(adapter, &crq);
3284 
3285 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3286 	atomic_inc(&adapter->running_cap_crqs);
3287 	ibmvnic_send_crq(adapter, &crq);
3288 
3289 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3290 	atomic_inc(&adapter->running_cap_crqs);
3291 	ibmvnic_send_crq(adapter, &crq);
3292 
3293 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3294 	atomic_inc(&adapter->running_cap_crqs);
3295 	ibmvnic_send_crq(adapter, &crq);
3296 
3297 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3298 	atomic_inc(&adapter->running_cap_crqs);
3299 	ibmvnic_send_crq(adapter, &crq);
3300 
3301 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3302 	atomic_inc(&adapter->running_cap_crqs);
3303 	ibmvnic_send_crq(adapter, &crq);
3304 
3305 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3306 	atomic_inc(&adapter->running_cap_crqs);
3307 	ibmvnic_send_crq(adapter, &crq);
3308 
3309 	crq.query_capability.capability =
3310 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3311 	atomic_inc(&adapter->running_cap_crqs);
3312 	ibmvnic_send_crq(adapter, &crq);
3313 
3314 	crq.query_capability.capability =
3315 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3316 	atomic_inc(&adapter->running_cap_crqs);
3317 	ibmvnic_send_crq(adapter, &crq);
3318 
3319 	crq.query_capability.capability =
3320 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3321 	atomic_inc(&adapter->running_cap_crqs);
3322 	ibmvnic_send_crq(adapter, &crq);
3323 
3324 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3325 	atomic_inc(&adapter->running_cap_crqs);
3326 	ibmvnic_send_crq(adapter, &crq);
3327 }
3328 
3329 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3330 				struct ibmvnic_adapter *adapter)
3331 {
3332 	struct device *dev = &adapter->vdev->dev;
3333 
3334 	if (crq->get_vpd_size_rsp.rc.code) {
3335 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3336 			crq->get_vpd_size_rsp.rc.code);
3337 		complete(&adapter->fw_done);
3338 		return;
3339 	}
3340 
3341 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3342 	complete(&adapter->fw_done);
3343 }
3344 
3345 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3346 			   struct ibmvnic_adapter *adapter)
3347 {
3348 	struct device *dev = &adapter->vdev->dev;
3349 	unsigned char *substr = NULL;
3350 	u8 fw_level_len = 0;
3351 
3352 	memset(adapter->fw_version, 0, 32);
3353 
3354 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3355 			 DMA_FROM_DEVICE);
3356 
3357 	if (crq->get_vpd_rsp.rc.code) {
3358 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3359 			crq->get_vpd_rsp.rc.code);
3360 		goto complete;
3361 	}
3362 
3363 	/* get the position of the firmware version info
3364 	 * located after the ASCII 'RM' substring in the buffer
3365 	 */
3366 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3367 	if (!substr) {
3368 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3369 		goto complete;
3370 	}
3371 
3372 	/* get length of firmware level ASCII substring */
3373 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3374 		fw_level_len = *(substr + 2);
3375 	} else {
3376 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3377 		goto complete;
3378 	}
3379 
3380 	/* copy firmware version string from vpd into adapter */
3381 	if ((substr + 3 + fw_level_len) <
3382 	    (adapter->vpd->buff + adapter->vpd->len)) {
3383 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3384 	} else {
3385 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3386 	}
3387 
3388 complete:
3389 	if (adapter->fw_version[0] == '\0')
3390 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3391 	complete(&adapter->fw_done);
3392 }
3393 
3394 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3395 {
3396 	struct device *dev = &adapter->vdev->dev;
3397 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3398 	union ibmvnic_crq crq;
3399 	int i;
3400 
3401 	dma_unmap_single(dev, adapter->ip_offload_tok,
3402 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3403 
3404 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3405 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3406 		netdev_dbg(adapter->netdev, "%016lx\n",
3407 			   ((unsigned long int *)(buf))[i]);
3408 
3409 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3410 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3411 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3412 		   buf->tcp_ipv4_chksum);
3413 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3414 		   buf->tcp_ipv6_chksum);
3415 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3416 		   buf->udp_ipv4_chksum);
3417 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3418 		   buf->udp_ipv6_chksum);
3419 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3420 		   buf->large_tx_ipv4);
3421 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3422 		   buf->large_tx_ipv6);
3423 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3424 		   buf->large_rx_ipv4);
3425 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3426 		   buf->large_rx_ipv6);
3427 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3428 		   buf->max_ipv4_header_size);
3429 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3430 		   buf->max_ipv6_header_size);
3431 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3432 		   buf->max_tcp_header_size);
3433 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3434 		   buf->max_udp_header_size);
3435 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3436 		   buf->max_large_tx_size);
3437 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3438 		   buf->max_large_rx_size);
3439 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3440 		   buf->ipv6_extension_header);
3441 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3442 		   buf->tcp_pseudosum_req);
3443 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3444 		   buf->num_ipv6_ext_headers);
3445 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3446 		   buf->off_ipv6_ext_headers);
3447 
3448 	adapter->ip_offload_ctrl_tok =
3449 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
3450 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3451 
3452 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3453 		dev_err(dev, "Couldn't map ip offload control buffer\n");
3454 		return;
3455 	}
3456 
3457 	adapter->ip_offload_ctrl.len =
3458 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3459 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3460 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3461 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3462 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3463 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3464 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3465 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3466 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3467 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3468 
3469 	/* large_rx disabled for now, additional features needed */
3470 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3471 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3472 
3473 	adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3474 
3475 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3476 		adapter->netdev->features |= NETIF_F_IP_CSUM;
3477 
3478 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3479 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3480 
3481 	if ((adapter->netdev->features &
3482 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3483 		adapter->netdev->features |= NETIF_F_RXCSUM;
3484 
3485 	if (buf->large_tx_ipv4)
3486 		adapter->netdev->features |= NETIF_F_TSO;
3487 	if (buf->large_tx_ipv6)
3488 		adapter->netdev->features |= NETIF_F_TSO6;
3489 
3490 	adapter->netdev->hw_features |= adapter->netdev->features;
3491 
3492 	memset(&crq, 0, sizeof(crq));
3493 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3494 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3495 	crq.control_ip_offload.len =
3496 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3497 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3498 	ibmvnic_send_crq(adapter, &crq);
3499 }
3500 
3501 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3502 				  struct ibmvnic_adapter *adapter)
3503 {
3504 	struct device *dev = &adapter->vdev->dev;
3505 	struct ibmvnic_error_buff *error_buff, *tmp;
3506 	unsigned long flags;
3507 	bool found = false;
3508 	int i;
3509 
3510 	if (!crq->request_error_rsp.rc.code) {
3511 		dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3512 			 crq->request_error_rsp.rc.code);
3513 		return;
3514 	}
3515 
3516 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3517 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3518 		if (error_buff->error_id == crq->request_error_rsp.error_id) {
3519 			found = true;
3520 			list_del(&error_buff->list);
3521 			break;
3522 		}
3523 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3524 
3525 	if (!found) {
3526 		dev_err(dev, "Couldn't find error id %x\n",
3527 			be32_to_cpu(crq->request_error_rsp.error_id));
3528 		return;
3529 	}
3530 
3531 	dev_err(dev, "Detailed info for error id %x:",
3532 		be32_to_cpu(crq->request_error_rsp.error_id));
3533 
3534 	for (i = 0; i < error_buff->len; i++) {
3535 		pr_cont("%02x", (int)error_buff->buff[i]);
3536 		if (i % 8 == 7)
3537 			pr_cont(" ");
3538 	}
3539 	pr_cont("\n");
3540 
3541 	dma_unmap_single(dev, error_buff->dma, error_buff->len,
3542 			 DMA_FROM_DEVICE);
3543 	kfree(error_buff->buff);
3544 	kfree(error_buff);
3545 }
3546 
3547 static void request_error_information(struct ibmvnic_adapter *adapter,
3548 				      union ibmvnic_crq *err_crq)
3549 {
3550 	struct device *dev = &adapter->vdev->dev;
3551 	struct net_device *netdev = adapter->netdev;
3552 	struct ibmvnic_error_buff *error_buff;
3553 	unsigned long timeout = msecs_to_jiffies(30000);
3554 	union ibmvnic_crq crq;
3555 	unsigned long flags;
3556 	int rc, detail_len;
3557 
3558 	error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3559 	if (!error_buff)
3560 		return;
3561 
3562 	detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3563 	error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3564 	if (!error_buff->buff) {
3565 		kfree(error_buff);
3566 		return;
3567 	}
3568 
3569 	error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3570 					 DMA_FROM_DEVICE);
3571 	if (dma_mapping_error(dev, error_buff->dma)) {
3572 		netdev_err(netdev, "Couldn't map error buffer\n");
3573 		kfree(error_buff->buff);
3574 		kfree(error_buff);
3575 		return;
3576 	}
3577 
3578 	error_buff->len = detail_len;
3579 	error_buff->error_id = err_crq->error_indication.error_id;
3580 
3581 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3582 	list_add_tail(&error_buff->list, &adapter->errors);
3583 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3584 
3585 	memset(&crq, 0, sizeof(crq));
3586 	crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3587 	crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3588 	crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3589 	crq.request_error_info.len = cpu_to_be32(detail_len);
3590 	crq.request_error_info.error_id = err_crq->error_indication.error_id;
3591 
3592 	rc = ibmvnic_send_crq(adapter, &crq);
3593 	if (rc) {
3594 		netdev_err(netdev, "failed to request error information\n");
3595 		goto err_info_fail;
3596 	}
3597 
3598 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3599 		netdev_err(netdev, "timeout waiting for error information\n");
3600 		goto err_info_fail;
3601 	}
3602 
3603 	return;
3604 
3605 err_info_fail:
3606 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3607 	list_del(&error_buff->list);
3608 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3609 
3610 	kfree(error_buff->buff);
3611 	kfree(error_buff);
3612 }
3613 
3614 static void handle_error_indication(union ibmvnic_crq *crq,
3615 				    struct ibmvnic_adapter *adapter)
3616 {
3617 	struct device *dev = &adapter->vdev->dev;
3618 
3619 	dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3620 		crq->error_indication.flags
3621 			& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3622 		be32_to_cpu(crq->error_indication.error_id),
3623 		be16_to_cpu(crq->error_indication.error_cause));
3624 
3625 	if (be32_to_cpu(crq->error_indication.error_id))
3626 		request_error_information(adapter, crq);
3627 
3628 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3629 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3630 	else
3631 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3632 }
3633 
3634 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3635 				 struct ibmvnic_adapter *adapter)
3636 {
3637 	struct net_device *netdev = adapter->netdev;
3638 	struct device *dev = &adapter->vdev->dev;
3639 	long rc;
3640 
3641 	rc = crq->change_mac_addr_rsp.rc.code;
3642 	if (rc) {
3643 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3644 		goto out;
3645 	}
3646 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3647 	       ETH_ALEN);
3648 out:
3649 	complete(&adapter->fw_done);
3650 	return rc;
3651 }
3652 
3653 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3654 				   struct ibmvnic_adapter *adapter)
3655 {
3656 	struct device *dev = &adapter->vdev->dev;
3657 	u64 *req_value;
3658 	char *name;
3659 
3660 	atomic_dec(&adapter->running_cap_crqs);
3661 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3662 	case REQ_TX_QUEUES:
3663 		req_value = &adapter->req_tx_queues;
3664 		name = "tx";
3665 		break;
3666 	case REQ_RX_QUEUES:
3667 		req_value = &adapter->req_rx_queues;
3668 		name = "rx";
3669 		break;
3670 	case REQ_RX_ADD_QUEUES:
3671 		req_value = &adapter->req_rx_add_queues;
3672 		name = "rx_add";
3673 		break;
3674 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3675 		req_value = &adapter->req_tx_entries_per_subcrq;
3676 		name = "tx_entries_per_subcrq";
3677 		break;
3678 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3679 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3680 		name = "rx_add_entries_per_subcrq";
3681 		break;
3682 	case REQ_MTU:
3683 		req_value = &adapter->req_mtu;
3684 		name = "mtu";
3685 		break;
3686 	case PROMISC_REQUESTED:
3687 		req_value = &adapter->promisc;
3688 		name = "promisc";
3689 		break;
3690 	default:
3691 		dev_err(dev, "Got invalid cap request rsp %d\n",
3692 			crq->request_capability.capability);
3693 		return;
3694 	}
3695 
3696 	switch (crq->request_capability_rsp.rc.code) {
3697 	case SUCCESS:
3698 		break;
3699 	case PARTIALSUCCESS:
3700 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3701 			 *req_value,
3702 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3703 					       number), name);
3704 
3705 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3706 		    REQ_MTU) {
3707 			pr_err("mtu of %llu is not supported. Reverting.\n",
3708 			       *req_value);
3709 			*req_value = adapter->fallback.mtu;
3710 		} else {
3711 			*req_value =
3712 				be64_to_cpu(crq->request_capability_rsp.number);
3713 		}
3714 
3715 		ibmvnic_send_req_caps(adapter, 1);
3716 		return;
3717 	default:
3718 		dev_err(dev, "Error %d in request cap rsp\n",
3719 			crq->request_capability_rsp.rc.code);
3720 		return;
3721 	}
3722 
3723 	/* Done receiving requested capabilities, query IP offload support */
3724 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3725 		union ibmvnic_crq newcrq;
3726 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3727 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3728 		    &adapter->ip_offload_buf;
3729 
3730 		adapter->wait_capability = false;
3731 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3732 							 buf_sz,
3733 							 DMA_FROM_DEVICE);
3734 
3735 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3736 			if (!firmware_has_feature(FW_FEATURE_CMO))
3737 				dev_err(dev, "Couldn't map offload buffer\n");
3738 			return;
3739 		}
3740 
3741 		memset(&newcrq, 0, sizeof(newcrq));
3742 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3743 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3744 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3745 		newcrq.query_ip_offload.ioba =
3746 		    cpu_to_be32(adapter->ip_offload_tok);
3747 
3748 		ibmvnic_send_crq(adapter, &newcrq);
3749 	}
3750 }
3751 
3752 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3753 			    struct ibmvnic_adapter *adapter)
3754 {
3755 	struct device *dev = &adapter->vdev->dev;
3756 	struct net_device *netdev = adapter->netdev;
3757 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3758 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3759 	int i;
3760 
3761 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3762 			 DMA_BIDIRECTIONAL);
3763 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3764 			 adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3765 
3766 	/* If the number of queues requested can't be allocated by the
3767 	 * server, the login response will return with code 1. We will need
3768 	 * to resend the login buffer with fewer queues requested.
3769 	 */
3770 	if (login_rsp_crq->generic.rc.code) {
3771 		adapter->renegotiate = true;
3772 		complete(&adapter->init_done);
3773 		return 0;
3774 	}
3775 
3776 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
3777 
3778 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3779 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3780 		netdev_dbg(adapter->netdev, "%016lx\n",
3781 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3782 	}
3783 
3784 	/* Sanity checks */
3785 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3786 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
3787 	     adapter->req_rx_add_queues !=
3788 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3789 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3790 		ibmvnic_remove(adapter->vdev);
3791 		return -EIO;
3792 	}
3793 	release_login_buffer(adapter);
3794 	complete(&adapter->init_done);
3795 
3796 	return 0;
3797 }
3798 
3799 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3800 				     struct ibmvnic_adapter *adapter)
3801 {
3802 	struct device *dev = &adapter->vdev->dev;
3803 	long rc;
3804 
3805 	rc = crq->request_unmap_rsp.rc.code;
3806 	if (rc)
3807 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3808 }
3809 
3810 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3811 				 struct ibmvnic_adapter *adapter)
3812 {
3813 	struct net_device *netdev = adapter->netdev;
3814 	struct device *dev = &adapter->vdev->dev;
3815 	long rc;
3816 
3817 	rc = crq->query_map_rsp.rc.code;
3818 	if (rc) {
3819 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3820 		return;
3821 	}
3822 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3823 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3824 		   crq->query_map_rsp.free_pages);
3825 }
3826 
3827 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3828 				 struct ibmvnic_adapter *adapter)
3829 {
3830 	struct net_device *netdev = adapter->netdev;
3831 	struct device *dev = &adapter->vdev->dev;
3832 	long rc;
3833 
3834 	atomic_dec(&adapter->running_cap_crqs);
3835 	netdev_dbg(netdev, "Outstanding queries: %d\n",
3836 		   atomic_read(&adapter->running_cap_crqs));
3837 	rc = crq->query_capability.rc.code;
3838 	if (rc) {
3839 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3840 		goto out;
3841 	}
3842 
3843 	switch (be16_to_cpu(crq->query_capability.capability)) {
3844 	case MIN_TX_QUEUES:
3845 		adapter->min_tx_queues =
3846 		    be64_to_cpu(crq->query_capability.number);
3847 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
3848 			   adapter->min_tx_queues);
3849 		break;
3850 	case MIN_RX_QUEUES:
3851 		adapter->min_rx_queues =
3852 		    be64_to_cpu(crq->query_capability.number);
3853 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
3854 			   adapter->min_rx_queues);
3855 		break;
3856 	case MIN_RX_ADD_QUEUES:
3857 		adapter->min_rx_add_queues =
3858 		    be64_to_cpu(crq->query_capability.number);
3859 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3860 			   adapter->min_rx_add_queues);
3861 		break;
3862 	case MAX_TX_QUEUES:
3863 		adapter->max_tx_queues =
3864 		    be64_to_cpu(crq->query_capability.number);
3865 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
3866 			   adapter->max_tx_queues);
3867 		break;
3868 	case MAX_RX_QUEUES:
3869 		adapter->max_rx_queues =
3870 		    be64_to_cpu(crq->query_capability.number);
3871 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
3872 			   adapter->max_rx_queues);
3873 		break;
3874 	case MAX_RX_ADD_QUEUES:
3875 		adapter->max_rx_add_queues =
3876 		    be64_to_cpu(crq->query_capability.number);
3877 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3878 			   adapter->max_rx_add_queues);
3879 		break;
3880 	case MIN_TX_ENTRIES_PER_SUBCRQ:
3881 		adapter->min_tx_entries_per_subcrq =
3882 		    be64_to_cpu(crq->query_capability.number);
3883 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3884 			   adapter->min_tx_entries_per_subcrq);
3885 		break;
3886 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3887 		adapter->min_rx_add_entries_per_subcrq =
3888 		    be64_to_cpu(crq->query_capability.number);
3889 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3890 			   adapter->min_rx_add_entries_per_subcrq);
3891 		break;
3892 	case MAX_TX_ENTRIES_PER_SUBCRQ:
3893 		adapter->max_tx_entries_per_subcrq =
3894 		    be64_to_cpu(crq->query_capability.number);
3895 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3896 			   adapter->max_tx_entries_per_subcrq);
3897 		break;
3898 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3899 		adapter->max_rx_add_entries_per_subcrq =
3900 		    be64_to_cpu(crq->query_capability.number);
3901 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3902 			   adapter->max_rx_add_entries_per_subcrq);
3903 		break;
3904 	case TCP_IP_OFFLOAD:
3905 		adapter->tcp_ip_offload =
3906 		    be64_to_cpu(crq->query_capability.number);
3907 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3908 			   adapter->tcp_ip_offload);
3909 		break;
3910 	case PROMISC_SUPPORTED:
3911 		adapter->promisc_supported =
3912 		    be64_to_cpu(crq->query_capability.number);
3913 		netdev_dbg(netdev, "promisc_supported = %lld\n",
3914 			   adapter->promisc_supported);
3915 		break;
3916 	case MIN_MTU:
3917 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3918 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3919 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3920 		break;
3921 	case MAX_MTU:
3922 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3923 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3924 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3925 		break;
3926 	case MAX_MULTICAST_FILTERS:
3927 		adapter->max_multicast_filters =
3928 		    be64_to_cpu(crq->query_capability.number);
3929 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3930 			   adapter->max_multicast_filters);
3931 		break;
3932 	case VLAN_HEADER_INSERTION:
3933 		adapter->vlan_header_insertion =
3934 		    be64_to_cpu(crq->query_capability.number);
3935 		if (adapter->vlan_header_insertion)
3936 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3937 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3938 			   adapter->vlan_header_insertion);
3939 		break;
3940 	case RX_VLAN_HEADER_INSERTION:
3941 		adapter->rx_vlan_header_insertion =
3942 		    be64_to_cpu(crq->query_capability.number);
3943 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3944 			   adapter->rx_vlan_header_insertion);
3945 		break;
3946 	case MAX_TX_SG_ENTRIES:
3947 		adapter->max_tx_sg_entries =
3948 		    be64_to_cpu(crq->query_capability.number);
3949 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3950 			   adapter->max_tx_sg_entries);
3951 		break;
3952 	case RX_SG_SUPPORTED:
3953 		adapter->rx_sg_supported =
3954 		    be64_to_cpu(crq->query_capability.number);
3955 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3956 			   adapter->rx_sg_supported);
3957 		break;
3958 	case OPT_TX_COMP_SUB_QUEUES:
3959 		adapter->opt_tx_comp_sub_queues =
3960 		    be64_to_cpu(crq->query_capability.number);
3961 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3962 			   adapter->opt_tx_comp_sub_queues);
3963 		break;
3964 	case OPT_RX_COMP_QUEUES:
3965 		adapter->opt_rx_comp_queues =
3966 		    be64_to_cpu(crq->query_capability.number);
3967 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3968 			   adapter->opt_rx_comp_queues);
3969 		break;
3970 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3971 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
3972 		    be64_to_cpu(crq->query_capability.number);
3973 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3974 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
3975 		break;
3976 	case OPT_TX_ENTRIES_PER_SUBCRQ:
3977 		adapter->opt_tx_entries_per_subcrq =
3978 		    be64_to_cpu(crq->query_capability.number);
3979 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3980 			   adapter->opt_tx_entries_per_subcrq);
3981 		break;
3982 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3983 		adapter->opt_rxba_entries_per_subcrq =
3984 		    be64_to_cpu(crq->query_capability.number);
3985 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3986 			   adapter->opt_rxba_entries_per_subcrq);
3987 		break;
3988 	case TX_RX_DESC_REQ:
3989 		adapter->tx_rx_desc_req = crq->query_capability.number;
3990 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3991 			   adapter->tx_rx_desc_req);
3992 		break;
3993 
3994 	default:
3995 		netdev_err(netdev, "Got invalid cap rsp %d\n",
3996 			   crq->query_capability.capability);
3997 	}
3998 
3999 out:
4000 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4001 		adapter->wait_capability = false;
4002 		ibmvnic_send_req_caps(adapter, 0);
4003 	}
4004 }
4005 
4006 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4007 			       struct ibmvnic_adapter *adapter)
4008 {
4009 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4010 	struct net_device *netdev = adapter->netdev;
4011 	struct device *dev = &adapter->vdev->dev;
4012 	u64 *u64_crq = (u64 *)crq;
4013 	long rc;
4014 
4015 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4016 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4017 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4018 	switch (gen_crq->first) {
4019 	case IBMVNIC_CRQ_INIT_RSP:
4020 		switch (gen_crq->cmd) {
4021 		case IBMVNIC_CRQ_INIT:
4022 			dev_info(dev, "Partner initialized\n");
4023 			adapter->from_passive_init = true;
4024 			complete(&adapter->init_done);
4025 			break;
4026 		case IBMVNIC_CRQ_INIT_COMPLETE:
4027 			dev_info(dev, "Partner initialization complete\n");
4028 			send_version_xchg(adapter);
4029 			break;
4030 		default:
4031 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4032 		}
4033 		return;
4034 	case IBMVNIC_CRQ_XPORT_EVENT:
4035 		netif_carrier_off(netdev);
4036 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4037 			dev_info(dev, "Migrated, re-enabling adapter\n");
4038 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4039 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4040 			dev_info(dev, "Backing device failover detected\n");
4041 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4042 		} else {
4043 			/* The adapter lost the connection */
4044 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4045 				gen_crq->cmd);
4046 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4047 		}
4048 		return;
4049 	case IBMVNIC_CRQ_CMD_RSP:
4050 		break;
4051 	default:
4052 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4053 			gen_crq->first);
4054 		return;
4055 	}
4056 
4057 	switch (gen_crq->cmd) {
4058 	case VERSION_EXCHANGE_RSP:
4059 		rc = crq->version_exchange_rsp.rc.code;
4060 		if (rc) {
4061 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4062 			break;
4063 		}
4064 		dev_info(dev, "Partner protocol version is %d\n",
4065 			 crq->version_exchange_rsp.version);
4066 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4067 		    ibmvnic_version)
4068 			ibmvnic_version =
4069 			    be16_to_cpu(crq->version_exchange_rsp.version);
4070 		send_cap_queries(adapter);
4071 		break;
4072 	case QUERY_CAPABILITY_RSP:
4073 		handle_query_cap_rsp(crq, adapter);
4074 		break;
4075 	case QUERY_MAP_RSP:
4076 		handle_query_map_rsp(crq, adapter);
4077 		break;
4078 	case REQUEST_MAP_RSP:
4079 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4080 		complete(&adapter->fw_done);
4081 		break;
4082 	case REQUEST_UNMAP_RSP:
4083 		handle_request_unmap_rsp(crq, adapter);
4084 		break;
4085 	case REQUEST_CAPABILITY_RSP:
4086 		handle_request_cap_rsp(crq, adapter);
4087 		break;
4088 	case LOGIN_RSP:
4089 		netdev_dbg(netdev, "Got Login Response\n");
4090 		handle_login_rsp(crq, adapter);
4091 		break;
4092 	case LOGICAL_LINK_STATE_RSP:
4093 		netdev_dbg(netdev,
4094 			   "Got Logical Link State Response, state: %d rc: %d\n",
4095 			   crq->logical_link_state_rsp.link_state,
4096 			   crq->logical_link_state_rsp.rc.code);
4097 		adapter->logical_link_state =
4098 		    crq->logical_link_state_rsp.link_state;
4099 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4100 		complete(&adapter->init_done);
4101 		break;
4102 	case LINK_STATE_INDICATION:
4103 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4104 		adapter->phys_link_state =
4105 		    crq->link_state_indication.phys_link_state;
4106 		adapter->logical_link_state =
4107 		    crq->link_state_indication.logical_link_state;
4108 		break;
4109 	case CHANGE_MAC_ADDR_RSP:
4110 		netdev_dbg(netdev, "Got MAC address change Response\n");
4111 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4112 		break;
4113 	case ERROR_INDICATION:
4114 		netdev_dbg(netdev, "Got Error Indication\n");
4115 		handle_error_indication(crq, adapter);
4116 		break;
4117 	case REQUEST_ERROR_RSP:
4118 		netdev_dbg(netdev, "Got Error Detail Response\n");
4119 		handle_error_info_rsp(crq, adapter);
4120 		break;
4121 	case REQUEST_STATISTICS_RSP:
4122 		netdev_dbg(netdev, "Got Statistics Response\n");
4123 		complete(&adapter->stats_done);
4124 		break;
4125 	case QUERY_IP_OFFLOAD_RSP:
4126 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4127 		handle_query_ip_offload_rsp(adapter);
4128 		break;
4129 	case MULTICAST_CTRL_RSP:
4130 		netdev_dbg(netdev, "Got multicast control Response\n");
4131 		break;
4132 	case CONTROL_IP_OFFLOAD_RSP:
4133 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4134 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4135 				 sizeof(adapter->ip_offload_ctrl),
4136 				 DMA_TO_DEVICE);
4137 		complete(&adapter->init_done);
4138 		break;
4139 	case COLLECT_FW_TRACE_RSP:
4140 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4141 		complete(&adapter->fw_done);
4142 		break;
4143 	case GET_VPD_SIZE_RSP:
4144 		handle_vpd_size_rsp(crq, adapter);
4145 		break;
4146 	case GET_VPD_RSP:
4147 		handle_vpd_rsp(crq, adapter);
4148 		break;
4149 	default:
4150 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4151 			   gen_crq->cmd);
4152 	}
4153 }
4154 
4155 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4156 {
4157 	struct ibmvnic_adapter *adapter = instance;
4158 
4159 	tasklet_schedule(&adapter->tasklet);
4160 	return IRQ_HANDLED;
4161 }
4162 
4163 static void ibmvnic_tasklet(void *data)
4164 {
4165 	struct ibmvnic_adapter *adapter = data;
4166 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4167 	union ibmvnic_crq *crq;
4168 	unsigned long flags;
4169 	bool done = false;
4170 
4171 	spin_lock_irqsave(&queue->lock, flags);
4172 	while (!done) {
4173 		/* Pull all the valid messages off the CRQ */
4174 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4175 			ibmvnic_handle_crq(crq, adapter);
4176 			crq->generic.first = 0;
4177 		}
4178 
4179 		/* remain in tasklet until all
4180 		 * capabilities responses are received
4181 		 */
4182 		if (!adapter->wait_capability)
4183 			done = true;
4184 	}
4185 	/* if capabilities CRQ's were sent in this tasklet, the following
4186 	 * tasklet must wait until all responses are received
4187 	 */
4188 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4189 		adapter->wait_capability = true;
4190 	spin_unlock_irqrestore(&queue->lock, flags);
4191 }
4192 
4193 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4194 {
4195 	struct vio_dev *vdev = adapter->vdev;
4196 	int rc;
4197 
4198 	do {
4199 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4200 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4201 
4202 	if (rc)
4203 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4204 
4205 	return rc;
4206 }
4207 
4208 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4209 {
4210 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4211 	struct device *dev = &adapter->vdev->dev;
4212 	struct vio_dev *vdev = adapter->vdev;
4213 	int rc;
4214 
4215 	/* Close the CRQ */
4216 	do {
4217 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4218 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4219 
4220 	/* Clean out the queue */
4221 	memset(crq->msgs, 0, PAGE_SIZE);
4222 	crq->cur = 0;
4223 
4224 	/* And re-open it again */
4225 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4226 				crq->msg_token, PAGE_SIZE);
4227 
4228 	if (rc == H_CLOSED)
4229 		/* Adapter is good, but other end is not ready */
4230 		dev_warn(dev, "Partner adapter not ready\n");
4231 	else if (rc != 0)
4232 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4233 
4234 	return rc;
4235 }
4236 
4237 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4238 {
4239 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4240 	struct vio_dev *vdev = adapter->vdev;
4241 	long rc;
4242 
4243 	if (!crq->msgs)
4244 		return;
4245 
4246 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4247 	free_irq(vdev->irq, adapter);
4248 	tasklet_kill(&adapter->tasklet);
4249 	do {
4250 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4251 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4252 
4253 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4254 			 DMA_BIDIRECTIONAL);
4255 	free_page((unsigned long)crq->msgs);
4256 	crq->msgs = NULL;
4257 }
4258 
4259 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4260 {
4261 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4262 	struct device *dev = &adapter->vdev->dev;
4263 	struct vio_dev *vdev = adapter->vdev;
4264 	int rc, retrc = -ENOMEM;
4265 
4266 	if (crq->msgs)
4267 		return 0;
4268 
4269 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4270 	/* Should we allocate more than one page? */
4271 
4272 	if (!crq->msgs)
4273 		return -ENOMEM;
4274 
4275 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4276 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4277 					DMA_BIDIRECTIONAL);
4278 	if (dma_mapping_error(dev, crq->msg_token))
4279 		goto map_failed;
4280 
4281 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4282 				crq->msg_token, PAGE_SIZE);
4283 
4284 	if (rc == H_RESOURCE)
4285 		/* maybe kexecing and resource is busy. try a reset */
4286 		rc = ibmvnic_reset_crq(adapter);
4287 	retrc = rc;
4288 
4289 	if (rc == H_CLOSED) {
4290 		dev_warn(dev, "Partner adapter not ready\n");
4291 	} else if (rc) {
4292 		dev_warn(dev, "Error %d opening adapter\n", rc);
4293 		goto reg_crq_failed;
4294 	}
4295 
4296 	retrc = 0;
4297 
4298 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4299 		     (unsigned long)adapter);
4300 
4301 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4302 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4303 			 adapter);
4304 	if (rc) {
4305 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4306 			vdev->irq, rc);
4307 		goto req_irq_failed;
4308 	}
4309 
4310 	rc = vio_enable_interrupts(vdev);
4311 	if (rc) {
4312 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4313 		goto req_irq_failed;
4314 	}
4315 
4316 	crq->cur = 0;
4317 	spin_lock_init(&crq->lock);
4318 
4319 	return retrc;
4320 
4321 req_irq_failed:
4322 	tasklet_kill(&adapter->tasklet);
4323 	do {
4324 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4325 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4326 reg_crq_failed:
4327 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4328 map_failed:
4329 	free_page((unsigned long)crq->msgs);
4330 	crq->msgs = NULL;
4331 	return retrc;
4332 }
4333 
4334 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4335 {
4336 	struct device *dev = &adapter->vdev->dev;
4337 	unsigned long timeout = msecs_to_jiffies(30000);
4338 	int rc;
4339 
4340 	if (adapter->resetting && !adapter->wait_for_reset) {
4341 		rc = ibmvnic_reset_crq(adapter);
4342 		if (!rc)
4343 			rc = vio_enable_interrupts(adapter->vdev);
4344 	} else {
4345 		rc = init_crq_queue(adapter);
4346 	}
4347 
4348 	if (rc) {
4349 		dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
4350 		return rc;
4351 	}
4352 
4353 	adapter->from_passive_init = false;
4354 
4355 	init_completion(&adapter->init_done);
4356 	adapter->init_done_rc = 0;
4357 	ibmvnic_send_crq_init(adapter);
4358 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4359 		dev_err(dev, "Initialization sequence timed out\n");
4360 		return -1;
4361 	}
4362 
4363 	if (adapter->init_done_rc) {
4364 		release_crq_queue(adapter);
4365 		return adapter->init_done_rc;
4366 	}
4367 
4368 	if (adapter->from_passive_init) {
4369 		adapter->state = VNIC_OPEN;
4370 		adapter->from_passive_init = false;
4371 		return -1;
4372 	}
4373 
4374 	if (adapter->resetting && !adapter->wait_for_reset)
4375 		rc = reset_sub_crq_queues(adapter);
4376 	else
4377 		rc = init_sub_crqs(adapter);
4378 	if (rc) {
4379 		dev_err(dev, "Initialization of sub crqs failed\n");
4380 		release_crq_queue(adapter);
4381 		return rc;
4382 	}
4383 
4384 	rc = init_sub_crq_irqs(adapter);
4385 	if (rc) {
4386 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4387 		release_crq_queue(adapter);
4388 	}
4389 
4390 	return rc;
4391 }
4392 
4393 static struct device_attribute dev_attr_failover;
4394 
4395 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4396 {
4397 	struct ibmvnic_adapter *adapter;
4398 	struct net_device *netdev;
4399 	unsigned char *mac_addr_p;
4400 	int rc;
4401 
4402 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4403 		dev->unit_address);
4404 
4405 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4406 							VETH_MAC_ADDR, NULL);
4407 	if (!mac_addr_p) {
4408 		dev_err(&dev->dev,
4409 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4410 			__FILE__, __LINE__);
4411 		return 0;
4412 	}
4413 
4414 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4415 				   IBMVNIC_MAX_QUEUES);
4416 	if (!netdev)
4417 		return -ENOMEM;
4418 
4419 	adapter = netdev_priv(netdev);
4420 	adapter->state = VNIC_PROBING;
4421 	dev_set_drvdata(&dev->dev, netdev);
4422 	adapter->vdev = dev;
4423 	adapter->netdev = netdev;
4424 
4425 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
4426 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4427 	netdev->irq = dev->irq;
4428 	netdev->netdev_ops = &ibmvnic_netdev_ops;
4429 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4430 	SET_NETDEV_DEV(netdev, &dev->dev);
4431 
4432 	spin_lock_init(&adapter->stats_lock);
4433 
4434 	INIT_LIST_HEAD(&adapter->errors);
4435 	spin_lock_init(&adapter->error_list_lock);
4436 
4437 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4438 	INIT_LIST_HEAD(&adapter->rwi_list);
4439 	mutex_init(&adapter->reset_lock);
4440 	mutex_init(&adapter->rwi_lock);
4441 	adapter->resetting = false;
4442 
4443 	adapter->mac_change_pending = false;
4444 
4445 	do {
4446 		rc = ibmvnic_init(adapter);
4447 		if (rc && rc != EAGAIN)
4448 			goto ibmvnic_init_fail;
4449 	} while (rc == EAGAIN);
4450 
4451 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4452 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4453 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4454 
4455 	rc = device_create_file(&dev->dev, &dev_attr_failover);
4456 	if (rc)
4457 		goto ibmvnic_init_fail;
4458 
4459 	netif_carrier_off(netdev);
4460 	rc = register_netdev(netdev);
4461 	if (rc) {
4462 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4463 		goto ibmvnic_register_fail;
4464 	}
4465 	dev_info(&dev->dev, "ibmvnic registered\n");
4466 
4467 	adapter->state = VNIC_PROBED;
4468 
4469 	adapter->wait_for_reset = false;
4470 
4471 	return 0;
4472 
4473 ibmvnic_register_fail:
4474 	device_remove_file(&dev->dev, &dev_attr_failover);
4475 
4476 ibmvnic_init_fail:
4477 	release_sub_crqs(adapter);
4478 	release_crq_queue(adapter);
4479 	free_netdev(netdev);
4480 
4481 	return rc;
4482 }
4483 
4484 static int ibmvnic_remove(struct vio_dev *dev)
4485 {
4486 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
4487 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4488 
4489 	adapter->state = VNIC_REMOVING;
4490 	unregister_netdev(netdev);
4491 	mutex_lock(&adapter->reset_lock);
4492 
4493 	release_resources(adapter);
4494 	release_sub_crqs(adapter);
4495 	release_crq_queue(adapter);
4496 
4497 	adapter->state = VNIC_REMOVED;
4498 
4499 	mutex_unlock(&adapter->reset_lock);
4500 	device_remove_file(&dev->dev, &dev_attr_failover);
4501 	free_netdev(netdev);
4502 	dev_set_drvdata(&dev->dev, NULL);
4503 
4504 	return 0;
4505 }
4506 
4507 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4508 			      const char *buf, size_t count)
4509 {
4510 	struct net_device *netdev = dev_get_drvdata(dev);
4511 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4512 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4513 	__be64 session_token;
4514 	long rc;
4515 
4516 	if (!sysfs_streq(buf, "1"))
4517 		return -EINVAL;
4518 
4519 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4520 			 H_GET_SESSION_TOKEN, 0, 0, 0);
4521 	if (rc) {
4522 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4523 			   rc);
4524 		return -EINVAL;
4525 	}
4526 
4527 	session_token = (__be64)retbuf[0];
4528 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4529 		   be64_to_cpu(session_token));
4530 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4531 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
4532 	if (rc) {
4533 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4534 			   rc);
4535 		return -EINVAL;
4536 	}
4537 
4538 	return count;
4539 }
4540 
4541 static DEVICE_ATTR_WO(failover);
4542 
4543 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4544 {
4545 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4546 	struct ibmvnic_adapter *adapter;
4547 	struct iommu_table *tbl;
4548 	unsigned long ret = 0;
4549 	int i;
4550 
4551 	tbl = get_iommu_table_base(&vdev->dev);
4552 
4553 	/* netdev inits at probe time along with the structures we need below*/
4554 	if (!netdev)
4555 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4556 
4557 	adapter = netdev_priv(netdev);
4558 
4559 	ret += PAGE_SIZE; /* the crq message queue */
4560 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4561 
4562 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4563 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4564 
4565 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4566 	     i++)
4567 		ret += adapter->rx_pool[i].size *
4568 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4569 
4570 	return ret;
4571 }
4572 
4573 static int ibmvnic_resume(struct device *dev)
4574 {
4575 	struct net_device *netdev = dev_get_drvdata(dev);
4576 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4577 
4578 	if (adapter->state != VNIC_OPEN)
4579 		return 0;
4580 
4581 	tasklet_schedule(&adapter->tasklet);
4582 
4583 	return 0;
4584 }
4585 
4586 static const struct vio_device_id ibmvnic_device_table[] = {
4587 	{"network", "IBM,vnic"},
4588 	{"", "" }
4589 };
4590 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4591 
4592 static const struct dev_pm_ops ibmvnic_pm_ops = {
4593 	.resume = ibmvnic_resume
4594 };
4595 
4596 static struct vio_driver ibmvnic_driver = {
4597 	.id_table       = ibmvnic_device_table,
4598 	.probe          = ibmvnic_probe,
4599 	.remove         = ibmvnic_remove,
4600 	.get_desired_dma = ibmvnic_get_desired_dma,
4601 	.name		= ibmvnic_driver_name,
4602 	.pm		= &ibmvnic_pm_ops,
4603 };
4604 
4605 /* module functions */
4606 static int __init ibmvnic_module_init(void)
4607 {
4608 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4609 		IBMVNIC_DRIVER_VERSION);
4610 
4611 	return vio_register_driver(&ibmvnic_driver);
4612 }
4613 
4614 static void __exit ibmvnic_module_exit(void)
4615 {
4616 	vio_unregister_driver(&ibmvnic_driver);
4617 }
4618 
4619 module_init(ibmvnic_module_init);
4620 module_exit(ibmvnic_module_exit);
4621