xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision d96fc832bcb6269d96e33d506f33033d7ed08598)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/if_arp.h>
63 #include <linux/in.h>
64 #include <linux/ip.h>
65 #include <linux/ipv6.h>
66 #include <linux/irq.h>
67 #include <linux/kthread.h>
68 #include <linux/seq_file.h>
69 #include <linux/interrupt.h>
70 #include <net/net_namespace.h>
71 #include <asm/hvcall.h>
72 #include <linux/atomic.h>
73 #include <asm/vio.h>
74 #include <asm/iommu.h>
75 #include <linux/uaccess.h>
76 #include <asm/firmware.h>
77 #include <linux/workqueue.h>
78 #include <linux/if_vlan.h>
79 #include <linux/utsname.h>
80 
81 #include "ibmvnic.h"
82 
83 static const char ibmvnic_driver_name[] = "ibmvnic";
84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
85 
86 MODULE_AUTHOR("Santiago Leon");
87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
88 MODULE_LICENSE("GPL");
89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
90 
91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
92 static int ibmvnic_remove(struct vio_dev *);
93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
99 		       union sub_crq *sub_crq);
100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
102 static int enable_scrq_irq(struct ibmvnic_adapter *,
103 			   struct ibmvnic_sub_crq_queue *);
104 static int disable_scrq_irq(struct ibmvnic_adapter *,
105 			    struct ibmvnic_sub_crq_queue *);
106 static int pending_scrq(struct ibmvnic_adapter *,
107 			struct ibmvnic_sub_crq_queue *);
108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
109 					struct ibmvnic_sub_crq_queue *);
110 static int ibmvnic_poll(struct napi_struct *napi, int data);
111 static void send_map_query(struct ibmvnic_adapter *adapter);
112 static 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(IBMVNIC_MAX_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(IBMVNIC_MAX_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 = rxadd_subcrqs;
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 	return 0;
558 }
559 
560 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
561 {
562 	struct ibmvnic_tx_pool *tx_pool;
563 	int tx_scrqs;
564 	int i, j, rc;
565 
566 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
567 	for (i = 0; i < tx_scrqs; i++) {
568 		netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
569 
570 		tx_pool = &adapter->tx_pool[i];
571 
572 		rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
573 		if (rc)
574 			return rc;
575 
576 		rc = reset_long_term_buff(adapter, &tx_pool->tso_ltb);
577 		if (rc)
578 			return rc;
579 
580 		memset(tx_pool->tx_buff, 0,
581 		       adapter->req_tx_entries_per_subcrq *
582 		       sizeof(struct ibmvnic_tx_buff));
583 
584 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
585 			tx_pool->free_map[j] = j;
586 
587 		tx_pool->consumer_index = 0;
588 		tx_pool->producer_index = 0;
589 		tx_pool->tso_index = 0;
590 	}
591 
592 	return 0;
593 }
594 
595 static void release_vpd_data(struct ibmvnic_adapter *adapter)
596 {
597 	if (!adapter->vpd)
598 		return;
599 
600 	kfree(adapter->vpd->buff);
601 	kfree(adapter->vpd);
602 
603 	adapter->vpd = NULL;
604 }
605 
606 static void release_tx_pools(struct ibmvnic_adapter *adapter)
607 {
608 	struct ibmvnic_tx_pool *tx_pool;
609 	int i;
610 
611 	if (!adapter->tx_pool)
612 		return;
613 
614 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
615 		netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
616 		tx_pool = &adapter->tx_pool[i];
617 		kfree(tx_pool->tx_buff);
618 		free_long_term_buff(adapter, &tx_pool->long_term_buff);
619 		free_long_term_buff(adapter, &tx_pool->tso_ltb);
620 		kfree(tx_pool->free_map);
621 	}
622 
623 	kfree(adapter->tx_pool);
624 	adapter->tx_pool = NULL;
625 	adapter->num_active_tx_pools = 0;
626 }
627 
628 static int init_tx_pools(struct net_device *netdev)
629 {
630 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
631 	struct device *dev = &adapter->vdev->dev;
632 	struct ibmvnic_tx_pool *tx_pool;
633 	int tx_subcrqs;
634 	int i, j;
635 
636 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
637 	adapter->tx_pool = kcalloc(tx_subcrqs,
638 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
639 	if (!adapter->tx_pool)
640 		return -1;
641 
642 	adapter->num_active_tx_pools = tx_subcrqs;
643 
644 	for (i = 0; i < tx_subcrqs; i++) {
645 		tx_pool = &adapter->tx_pool[i];
646 
647 		netdev_dbg(adapter->netdev,
648 			   "Initializing tx_pool[%d], %lld buffs\n",
649 			   i, adapter->req_tx_entries_per_subcrq);
650 
651 		tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
652 					   sizeof(struct ibmvnic_tx_buff),
653 					   GFP_KERNEL);
654 		if (!tx_pool->tx_buff) {
655 			dev_err(dev, "tx pool buffer allocation failed\n");
656 			release_tx_pools(adapter);
657 			return -1;
658 		}
659 
660 		if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
661 					 adapter->req_tx_entries_per_subcrq *
662 					 adapter->req_mtu)) {
663 			release_tx_pools(adapter);
664 			return -1;
665 		}
666 
667 		/* alloc TSO ltb */
668 		if (alloc_long_term_buff(adapter, &tx_pool->tso_ltb,
669 					 IBMVNIC_TSO_BUFS *
670 					 IBMVNIC_TSO_BUF_SZ)) {
671 			release_tx_pools(adapter);
672 			return -1;
673 		}
674 
675 		tx_pool->tso_index = 0;
676 
677 		tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
678 					    sizeof(int), GFP_KERNEL);
679 		if (!tx_pool->free_map) {
680 			release_tx_pools(adapter);
681 			return -1;
682 		}
683 
684 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
685 			tx_pool->free_map[j] = j;
686 
687 		tx_pool->consumer_index = 0;
688 		tx_pool->producer_index = 0;
689 	}
690 
691 	return 0;
692 }
693 
694 static void release_error_buffers(struct ibmvnic_adapter *adapter)
695 {
696 	struct device *dev = &adapter->vdev->dev;
697 	struct ibmvnic_error_buff *error_buff, *tmp;
698 	unsigned long flags;
699 
700 	spin_lock_irqsave(&adapter->error_list_lock, flags);
701 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
702 		list_del(&error_buff->list);
703 		dma_unmap_single(dev, error_buff->dma, error_buff->len,
704 				 DMA_FROM_DEVICE);
705 		kfree(error_buff->buff);
706 		kfree(error_buff);
707 	}
708 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
709 }
710 
711 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
712 {
713 	int i;
714 
715 	if (adapter->napi_enabled)
716 		return;
717 
718 	for (i = 0; i < adapter->req_rx_queues; i++)
719 		napi_enable(&adapter->napi[i]);
720 
721 	adapter->napi_enabled = true;
722 }
723 
724 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
725 {
726 	int i;
727 
728 	if (!adapter->napi_enabled)
729 		return;
730 
731 	for (i = 0; i < adapter->req_rx_queues; i++) {
732 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
733 		napi_disable(&adapter->napi[i]);
734 	}
735 
736 	adapter->napi_enabled = false;
737 }
738 
739 static int init_napi(struct ibmvnic_adapter *adapter)
740 {
741 	int i;
742 
743 	adapter->napi = kcalloc(adapter->req_rx_queues,
744 				sizeof(struct napi_struct), GFP_KERNEL);
745 	if (!adapter->napi)
746 		return -ENOMEM;
747 
748 	for (i = 0; i < adapter->req_rx_queues; i++) {
749 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
750 		netif_napi_add(adapter->netdev, &adapter->napi[i],
751 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
752 	}
753 
754 	adapter->num_active_rx_napi = adapter->req_rx_queues;
755 	return 0;
756 }
757 
758 static void release_napi(struct ibmvnic_adapter *adapter)
759 {
760 	int i;
761 
762 	if (!adapter->napi)
763 		return;
764 
765 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
766 		if (&adapter->napi[i]) {
767 			netdev_dbg(adapter->netdev,
768 				   "Releasing napi[%d]\n", i);
769 			netif_napi_del(&adapter->napi[i]);
770 		}
771 	}
772 
773 	kfree(adapter->napi);
774 	adapter->napi = NULL;
775 	adapter->num_active_rx_napi = 0;
776 }
777 
778 static int ibmvnic_login(struct net_device *netdev)
779 {
780 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
781 	unsigned long timeout = msecs_to_jiffies(30000);
782 	struct device *dev = &adapter->vdev->dev;
783 	int rc;
784 
785 	do {
786 		if (adapter->renegotiate) {
787 			adapter->renegotiate = false;
788 			release_sub_crqs(adapter, 1);
789 
790 			reinit_completion(&adapter->init_done);
791 			send_cap_queries(adapter);
792 			if (!wait_for_completion_timeout(&adapter->init_done,
793 							 timeout)) {
794 				dev_err(dev, "Capabilities query timeout\n");
795 				return -1;
796 			}
797 			rc = init_sub_crqs(adapter);
798 			if (rc) {
799 				dev_err(dev,
800 					"Initialization of SCRQ's failed\n");
801 				return -1;
802 			}
803 			rc = init_sub_crq_irqs(adapter);
804 			if (rc) {
805 				dev_err(dev,
806 					"Initialization of SCRQ's irqs failed\n");
807 				return -1;
808 			}
809 		}
810 
811 		reinit_completion(&adapter->init_done);
812 		send_login(adapter);
813 		if (!wait_for_completion_timeout(&adapter->init_done,
814 						 timeout)) {
815 			dev_err(dev, "Login timeout\n");
816 			return -1;
817 		}
818 	} while (adapter->renegotiate);
819 
820 	/* handle pending MAC address changes after successful login */
821 	if (adapter->mac_change_pending) {
822 		__ibmvnic_set_mac(netdev, &adapter->desired.mac);
823 		adapter->mac_change_pending = false;
824 	}
825 
826 	return 0;
827 }
828 
829 static void release_login_buffer(struct ibmvnic_adapter *adapter)
830 {
831 	kfree(adapter->login_buf);
832 	adapter->login_buf = NULL;
833 }
834 
835 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
836 {
837 	kfree(adapter->login_rsp_buf);
838 	adapter->login_rsp_buf = NULL;
839 }
840 
841 static void release_resources(struct ibmvnic_adapter *adapter)
842 {
843 	release_vpd_data(adapter);
844 
845 	release_tx_pools(adapter);
846 	release_rx_pools(adapter);
847 
848 	release_stats_token(adapter);
849 	release_stats_buffers(adapter);
850 	release_error_buffers(adapter);
851 	release_napi(adapter);
852 	release_login_rsp_buffer(adapter);
853 }
854 
855 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
856 {
857 	struct net_device *netdev = adapter->netdev;
858 	unsigned long timeout = msecs_to_jiffies(30000);
859 	union ibmvnic_crq crq;
860 	bool resend;
861 	int rc;
862 
863 	netdev_dbg(netdev, "setting link state %d\n", link_state);
864 
865 	memset(&crq, 0, sizeof(crq));
866 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
867 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
868 	crq.logical_link_state.link_state = link_state;
869 
870 	do {
871 		resend = false;
872 
873 		reinit_completion(&adapter->init_done);
874 		rc = ibmvnic_send_crq(adapter, &crq);
875 		if (rc) {
876 			netdev_err(netdev, "Failed to set link state\n");
877 			return rc;
878 		}
879 
880 		if (!wait_for_completion_timeout(&adapter->init_done,
881 						 timeout)) {
882 			netdev_err(netdev, "timeout setting link state\n");
883 			return -1;
884 		}
885 
886 		if (adapter->init_done_rc == 1) {
887 			/* Partuial success, delay and re-send */
888 			mdelay(1000);
889 			resend = true;
890 		}
891 	} while (resend);
892 
893 	return 0;
894 }
895 
896 static int set_real_num_queues(struct net_device *netdev)
897 {
898 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
899 	int rc;
900 
901 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
902 		   adapter->req_tx_queues, adapter->req_rx_queues);
903 
904 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
905 	if (rc) {
906 		netdev_err(netdev, "failed to set the number of tx queues\n");
907 		return rc;
908 	}
909 
910 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
911 	if (rc)
912 		netdev_err(netdev, "failed to set the number of rx queues\n");
913 
914 	return rc;
915 }
916 
917 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
918 {
919 	struct device *dev = &adapter->vdev->dev;
920 	union ibmvnic_crq crq;
921 	int len = 0;
922 
923 	if (adapter->vpd->buff)
924 		len = adapter->vpd->len;
925 
926 	init_completion(&adapter->fw_done);
927 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
928 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
929 	ibmvnic_send_crq(adapter, &crq);
930 	wait_for_completion(&adapter->fw_done);
931 
932 	if (!adapter->vpd->len)
933 		return -ENODATA;
934 
935 	if (!adapter->vpd->buff)
936 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
937 	else if (adapter->vpd->len != len)
938 		adapter->vpd->buff =
939 			krealloc(adapter->vpd->buff,
940 				 adapter->vpd->len, GFP_KERNEL);
941 
942 	if (!adapter->vpd->buff) {
943 		dev_err(dev, "Could allocate VPD buffer\n");
944 		return -ENOMEM;
945 	}
946 
947 	adapter->vpd->dma_addr =
948 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
949 			       DMA_FROM_DEVICE);
950 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
951 		dev_err(dev, "Could not map VPD buffer\n");
952 		kfree(adapter->vpd->buff);
953 		adapter->vpd->buff = NULL;
954 		return -ENOMEM;
955 	}
956 
957 	reinit_completion(&adapter->fw_done);
958 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
959 	crq.get_vpd.cmd = GET_VPD;
960 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
961 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
962 	ibmvnic_send_crq(adapter, &crq);
963 	wait_for_completion(&adapter->fw_done);
964 
965 	return 0;
966 }
967 
968 static int init_resources(struct ibmvnic_adapter *adapter)
969 {
970 	struct net_device *netdev = adapter->netdev;
971 	int rc;
972 
973 	rc = set_real_num_queues(netdev);
974 	if (rc)
975 		return rc;
976 
977 	rc = init_stats_buffers(adapter);
978 	if (rc)
979 		return rc;
980 
981 	rc = init_stats_token(adapter);
982 	if (rc)
983 		return rc;
984 
985 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
986 	if (!adapter->vpd)
987 		return -ENOMEM;
988 
989 	/* Vital Product Data (VPD) */
990 	rc = ibmvnic_get_vpd(adapter);
991 	if (rc) {
992 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
993 		return rc;
994 	}
995 
996 	adapter->map_id = 1;
997 
998 	rc = init_napi(adapter);
999 	if (rc)
1000 		return rc;
1001 
1002 	send_map_query(adapter);
1003 
1004 	rc = init_rx_pools(netdev);
1005 	if (rc)
1006 		return rc;
1007 
1008 	rc = init_tx_pools(netdev);
1009 	return rc;
1010 }
1011 
1012 static int __ibmvnic_open(struct net_device *netdev)
1013 {
1014 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1015 	enum vnic_state prev_state = adapter->state;
1016 	int i, rc;
1017 
1018 	adapter->state = VNIC_OPENING;
1019 	replenish_pools(adapter);
1020 	ibmvnic_napi_enable(adapter);
1021 
1022 	/* We're ready to receive frames, enable the sub-crq interrupts and
1023 	 * set the logical link state to up
1024 	 */
1025 	for (i = 0; i < adapter->req_rx_queues; i++) {
1026 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1027 		if (prev_state == VNIC_CLOSED)
1028 			enable_irq(adapter->rx_scrq[i]->irq);
1029 		else
1030 			enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1031 	}
1032 
1033 	for (i = 0; i < adapter->req_tx_queues; i++) {
1034 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1035 		if (prev_state == VNIC_CLOSED)
1036 			enable_irq(adapter->tx_scrq[i]->irq);
1037 		else
1038 			enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1039 	}
1040 
1041 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1042 	if (rc) {
1043 		for (i = 0; i < adapter->req_rx_queues; i++)
1044 			napi_disable(&adapter->napi[i]);
1045 		release_resources(adapter);
1046 		return rc;
1047 	}
1048 
1049 	netif_tx_start_all_queues(netdev);
1050 
1051 	if (prev_state == VNIC_CLOSED) {
1052 		for (i = 0; i < adapter->req_rx_queues; i++)
1053 			napi_schedule(&adapter->napi[i]);
1054 	}
1055 
1056 	adapter->state = VNIC_OPEN;
1057 	return rc;
1058 }
1059 
1060 static int ibmvnic_open(struct net_device *netdev)
1061 {
1062 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1063 	int rc;
1064 
1065 	mutex_lock(&adapter->reset_lock);
1066 
1067 	if (adapter->state != VNIC_CLOSED) {
1068 		rc = ibmvnic_login(netdev);
1069 		if (rc) {
1070 			mutex_unlock(&adapter->reset_lock);
1071 			return rc;
1072 		}
1073 
1074 		rc = init_resources(adapter);
1075 		if (rc) {
1076 			netdev_err(netdev, "failed to initialize resources\n");
1077 			release_resources(adapter);
1078 			mutex_unlock(&adapter->reset_lock);
1079 			return rc;
1080 		}
1081 	}
1082 
1083 	rc = __ibmvnic_open(netdev);
1084 	netif_carrier_on(netdev);
1085 
1086 	mutex_unlock(&adapter->reset_lock);
1087 
1088 	return rc;
1089 }
1090 
1091 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1092 {
1093 	struct ibmvnic_rx_pool *rx_pool;
1094 	u64 rx_entries;
1095 	int rx_scrqs;
1096 	int i, j;
1097 
1098 	if (!adapter->rx_pool)
1099 		return;
1100 
1101 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
1102 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1103 
1104 	/* Free any remaining skbs in the rx buffer pools */
1105 	for (i = 0; i < rx_scrqs; i++) {
1106 		rx_pool = &adapter->rx_pool[i];
1107 		if (!rx_pool)
1108 			continue;
1109 
1110 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1111 		for (j = 0; j < rx_entries; j++) {
1112 			if (rx_pool->rx_buff[j].skb) {
1113 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
1114 				rx_pool->rx_buff[j].skb = NULL;
1115 			}
1116 		}
1117 	}
1118 }
1119 
1120 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1121 {
1122 	struct ibmvnic_tx_pool *tx_pool;
1123 	u64 tx_entries;
1124 	int tx_scrqs;
1125 	int i, j;
1126 
1127 	if (!adapter->tx_pool)
1128 		return;
1129 
1130 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
1131 	tx_entries = adapter->req_tx_entries_per_subcrq;
1132 
1133 	/* Free any remaining skbs in the tx buffer pools */
1134 	for (i = 0; i < tx_scrqs; i++) {
1135 		tx_pool = &adapter->tx_pool[i];
1136 		if (!tx_pool)
1137 			continue;
1138 
1139 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1140 		for (j = 0; j < tx_entries; j++) {
1141 			if (tx_pool->tx_buff[j].skb) {
1142 				dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
1143 				tx_pool->tx_buff[j].skb = NULL;
1144 			}
1145 		}
1146 	}
1147 }
1148 
1149 static int __ibmvnic_close(struct net_device *netdev)
1150 {
1151 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1152 	int rc = 0;
1153 	int i;
1154 
1155 	adapter->state = VNIC_CLOSING;
1156 
1157 	/* ensure that transmissions are stopped if called by do_reset */
1158 	if (adapter->resetting)
1159 		netif_tx_disable(netdev);
1160 	else
1161 		netif_tx_stop_all_queues(netdev);
1162 
1163 	ibmvnic_napi_disable(adapter);
1164 
1165 	if (adapter->tx_scrq) {
1166 		for (i = 0; i < adapter->req_tx_queues; i++)
1167 			if (adapter->tx_scrq[i]->irq) {
1168 				netdev_dbg(adapter->netdev,
1169 					   "Disabling tx_scrq[%d] irq\n", i);
1170 				disable_irq(adapter->tx_scrq[i]->irq);
1171 			}
1172 	}
1173 
1174 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1175 	if (rc)
1176 		return rc;
1177 
1178 	if (adapter->rx_scrq) {
1179 		for (i = 0; i < adapter->req_rx_queues; i++) {
1180 			int retries = 10;
1181 
1182 			while (pending_scrq(adapter, adapter->rx_scrq[i])) {
1183 				retries--;
1184 				mdelay(100);
1185 
1186 				if (retries == 0)
1187 					break;
1188 			}
1189 
1190 			if (adapter->rx_scrq[i]->irq) {
1191 				netdev_dbg(adapter->netdev,
1192 					   "Disabling rx_scrq[%d] irq\n", i);
1193 				disable_irq(adapter->rx_scrq[i]->irq);
1194 			}
1195 		}
1196 	}
1197 	clean_rx_pools(adapter);
1198 	clean_tx_pools(adapter);
1199 	adapter->state = VNIC_CLOSED;
1200 	return rc;
1201 }
1202 
1203 static int ibmvnic_close(struct net_device *netdev)
1204 {
1205 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1206 	int rc;
1207 
1208 	mutex_lock(&adapter->reset_lock);
1209 	rc = __ibmvnic_close(netdev);
1210 	mutex_unlock(&adapter->reset_lock);
1211 
1212 	return rc;
1213 }
1214 
1215 /**
1216  * build_hdr_data - creates L2/L3/L4 header data buffer
1217  * @hdr_field - bitfield determining needed headers
1218  * @skb - socket buffer
1219  * @hdr_len - array of header lengths
1220  * @tot_len - total length of data
1221  *
1222  * Reads hdr_field to determine which headers are needed by firmware.
1223  * Builds a buffer containing these headers.  Saves individual header
1224  * lengths and total buffer length to be used to build descriptors.
1225  */
1226 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1227 			  int *hdr_len, u8 *hdr_data)
1228 {
1229 	int len = 0;
1230 	u8 *hdr;
1231 
1232 	hdr_len[0] = sizeof(struct ethhdr);
1233 
1234 	if (skb->protocol == htons(ETH_P_IP)) {
1235 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1236 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1237 			hdr_len[2] = tcp_hdrlen(skb);
1238 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1239 			hdr_len[2] = sizeof(struct udphdr);
1240 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1241 		hdr_len[1] = sizeof(struct ipv6hdr);
1242 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1243 			hdr_len[2] = tcp_hdrlen(skb);
1244 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1245 			hdr_len[2] = sizeof(struct udphdr);
1246 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1247 		hdr_len[1] = arp_hdr_len(skb->dev);
1248 		hdr_len[2] = 0;
1249 	}
1250 
1251 	memset(hdr_data, 0, 120);
1252 	if ((hdr_field >> 6) & 1) {
1253 		hdr = skb_mac_header(skb);
1254 		memcpy(hdr_data, hdr, hdr_len[0]);
1255 		len += hdr_len[0];
1256 	}
1257 
1258 	if ((hdr_field >> 5) & 1) {
1259 		hdr = skb_network_header(skb);
1260 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1261 		len += hdr_len[1];
1262 	}
1263 
1264 	if ((hdr_field >> 4) & 1) {
1265 		hdr = skb_transport_header(skb);
1266 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1267 		len += hdr_len[2];
1268 	}
1269 	return len;
1270 }
1271 
1272 /**
1273  * create_hdr_descs - create header and header extension descriptors
1274  * @hdr_field - bitfield determining needed headers
1275  * @data - buffer containing header data
1276  * @len - length of data buffer
1277  * @hdr_len - array of individual header lengths
1278  * @scrq_arr - descriptor array
1279  *
1280  * Creates header and, if needed, header extension descriptors and
1281  * places them in a descriptor array, scrq_arr
1282  */
1283 
1284 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1285 			    union sub_crq *scrq_arr)
1286 {
1287 	union sub_crq hdr_desc;
1288 	int tmp_len = len;
1289 	int num_descs = 0;
1290 	u8 *data, *cur;
1291 	int tmp;
1292 
1293 	while (tmp_len > 0) {
1294 		cur = hdr_data + len - tmp_len;
1295 
1296 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1297 		if (cur != hdr_data) {
1298 			data = hdr_desc.hdr_ext.data;
1299 			tmp = tmp_len > 29 ? 29 : tmp_len;
1300 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1301 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1302 			hdr_desc.hdr_ext.len = tmp;
1303 		} else {
1304 			data = hdr_desc.hdr.data;
1305 			tmp = tmp_len > 24 ? 24 : tmp_len;
1306 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1307 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1308 			hdr_desc.hdr.len = tmp;
1309 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1310 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1311 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1312 			hdr_desc.hdr.flag = hdr_field << 1;
1313 		}
1314 		memcpy(data, cur, tmp);
1315 		tmp_len -= tmp;
1316 		*scrq_arr = hdr_desc;
1317 		scrq_arr++;
1318 		num_descs++;
1319 	}
1320 
1321 	return num_descs;
1322 }
1323 
1324 /**
1325  * build_hdr_descs_arr - build a header descriptor array
1326  * @skb - socket buffer
1327  * @num_entries - number of descriptors to be sent
1328  * @subcrq - first TX descriptor
1329  * @hdr_field - bit field determining which headers will be sent
1330  *
1331  * This function will build a TX descriptor array with applicable
1332  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1333  */
1334 
1335 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1336 				int *num_entries, u8 hdr_field)
1337 {
1338 	int hdr_len[3] = {0, 0, 0};
1339 	int tot_len;
1340 	u8 *hdr_data = txbuff->hdr_data;
1341 
1342 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1343 				 txbuff->hdr_data);
1344 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1345 			 txbuff->indir_arr + 1);
1346 }
1347 
1348 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1349 {
1350 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1351 	int queue_num = skb_get_queue_mapping(skb);
1352 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1353 	struct device *dev = &adapter->vdev->dev;
1354 	struct ibmvnic_tx_buff *tx_buff = NULL;
1355 	struct ibmvnic_sub_crq_queue *tx_scrq;
1356 	struct ibmvnic_tx_pool *tx_pool;
1357 	unsigned int tx_send_failed = 0;
1358 	unsigned int tx_map_failed = 0;
1359 	unsigned int tx_dropped = 0;
1360 	unsigned int tx_packets = 0;
1361 	unsigned int tx_bytes = 0;
1362 	dma_addr_t data_dma_addr;
1363 	struct netdev_queue *txq;
1364 	unsigned long lpar_rc;
1365 	union sub_crq tx_crq;
1366 	unsigned int offset;
1367 	int num_entries = 1;
1368 	unsigned char *dst;
1369 	u64 *handle_array;
1370 	int index = 0;
1371 	u8 proto = 0;
1372 	int ret = 0;
1373 
1374 	if (adapter->resetting) {
1375 		if (!netif_subqueue_stopped(netdev, skb))
1376 			netif_stop_subqueue(netdev, queue_num);
1377 		dev_kfree_skb_any(skb);
1378 
1379 		tx_send_failed++;
1380 		tx_dropped++;
1381 		ret = NETDEV_TX_OK;
1382 		goto out;
1383 	}
1384 
1385 	tx_pool = &adapter->tx_pool[queue_num];
1386 	tx_scrq = adapter->tx_scrq[queue_num];
1387 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1388 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1389 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1390 
1391 	index = tx_pool->free_map[tx_pool->consumer_index];
1392 
1393 	if (skb_is_gso(skb)) {
1394 		offset = tx_pool->tso_index * IBMVNIC_TSO_BUF_SZ;
1395 		dst = tx_pool->tso_ltb.buff + offset;
1396 		memset(dst, 0, IBMVNIC_TSO_BUF_SZ);
1397 		data_dma_addr = tx_pool->tso_ltb.addr + offset;
1398 		tx_pool->tso_index++;
1399 		if (tx_pool->tso_index == IBMVNIC_TSO_BUFS)
1400 			tx_pool->tso_index = 0;
1401 	} else {
1402 		offset = index * adapter->req_mtu;
1403 		dst = tx_pool->long_term_buff.buff + offset;
1404 		memset(dst, 0, adapter->req_mtu);
1405 		data_dma_addr = tx_pool->long_term_buff.addr + offset;
1406 	}
1407 
1408 	if (skb_shinfo(skb)->nr_frags) {
1409 		int cur, i;
1410 
1411 		/* Copy the head */
1412 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1413 		cur = skb_headlen(skb);
1414 
1415 		/* Copy the frags */
1416 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1417 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1418 
1419 			memcpy(dst + cur,
1420 			       page_address(skb_frag_page(frag)) +
1421 			       frag->page_offset, skb_frag_size(frag));
1422 			cur += skb_frag_size(frag);
1423 		}
1424 	} else {
1425 		skb_copy_from_linear_data(skb, dst, skb->len);
1426 	}
1427 
1428 	tx_pool->consumer_index =
1429 	    (tx_pool->consumer_index + 1) %
1430 		adapter->req_tx_entries_per_subcrq;
1431 
1432 	tx_buff = &tx_pool->tx_buff[index];
1433 	tx_buff->skb = skb;
1434 	tx_buff->data_dma[0] = data_dma_addr;
1435 	tx_buff->data_len[0] = skb->len;
1436 	tx_buff->index = index;
1437 	tx_buff->pool_index = queue_num;
1438 	tx_buff->last_frag = true;
1439 
1440 	memset(&tx_crq, 0, sizeof(tx_crq));
1441 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1442 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1443 	tx_crq.v1.n_crq_elem = 1;
1444 	tx_crq.v1.n_sge = 1;
1445 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1446 	tx_crq.v1.correlator = cpu_to_be32(index);
1447 	if (skb_is_gso(skb))
1448 		tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->tso_ltb.map_id);
1449 	else
1450 		tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1451 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1452 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1453 
1454 	if (adapter->vlan_header_insertion) {
1455 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1456 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1457 	}
1458 
1459 	if (skb->protocol == htons(ETH_P_IP)) {
1460 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1461 		proto = ip_hdr(skb)->protocol;
1462 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1463 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1464 		proto = ipv6_hdr(skb)->nexthdr;
1465 	}
1466 
1467 	if (proto == IPPROTO_TCP)
1468 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1469 	else if (proto == IPPROTO_UDP)
1470 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1471 
1472 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1473 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1474 		hdrs += 2;
1475 	}
1476 	if (skb_is_gso(skb)) {
1477 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1478 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1479 		hdrs += 2;
1480 	}
1481 	/* determine if l2/3/4 headers are sent to firmware */
1482 	if ((*hdrs >> 7) & 1) {
1483 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1484 		tx_crq.v1.n_crq_elem = num_entries;
1485 		tx_buff->indir_arr[0] = tx_crq;
1486 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1487 						    sizeof(tx_buff->indir_arr),
1488 						    DMA_TO_DEVICE);
1489 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1490 			dev_kfree_skb_any(skb);
1491 			tx_buff->skb = NULL;
1492 			if (!firmware_has_feature(FW_FEATURE_CMO))
1493 				dev_err(dev, "tx: unable to map descriptor array\n");
1494 			tx_map_failed++;
1495 			tx_dropped++;
1496 			ret = NETDEV_TX_OK;
1497 			goto out;
1498 		}
1499 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1500 					       (u64)tx_buff->indir_dma,
1501 					       (u64)num_entries);
1502 	} else {
1503 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1504 				      &tx_crq);
1505 	}
1506 	if (lpar_rc != H_SUCCESS) {
1507 		dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1508 
1509 		if (tx_pool->consumer_index == 0)
1510 			tx_pool->consumer_index =
1511 				adapter->req_tx_entries_per_subcrq - 1;
1512 		else
1513 			tx_pool->consumer_index--;
1514 
1515 		dev_kfree_skb_any(skb);
1516 		tx_buff->skb = NULL;
1517 
1518 		if (lpar_rc == H_CLOSED) {
1519 			/* Disable TX and report carrier off if queue is closed.
1520 			 * Firmware guarantees that a signal will be sent to the
1521 			 * driver, triggering a reset or some other action.
1522 			 */
1523 			netif_tx_stop_all_queues(netdev);
1524 			netif_carrier_off(netdev);
1525 		}
1526 
1527 		tx_send_failed++;
1528 		tx_dropped++;
1529 		ret = NETDEV_TX_OK;
1530 		goto out;
1531 	}
1532 
1533 	if (atomic_add_return(num_entries, &tx_scrq->used)
1534 					>= adapter->req_tx_entries_per_subcrq) {
1535 		netdev_info(netdev, "Stopping queue %d\n", queue_num);
1536 		netif_stop_subqueue(netdev, queue_num);
1537 	}
1538 
1539 	tx_buff->num_entries = num_entries;
1540 	tx_packets++;
1541 	tx_bytes += skb->len;
1542 	txq->trans_start = jiffies;
1543 	ret = NETDEV_TX_OK;
1544 
1545 out:
1546 	netdev->stats.tx_dropped += tx_dropped;
1547 	netdev->stats.tx_bytes += tx_bytes;
1548 	netdev->stats.tx_packets += tx_packets;
1549 	adapter->tx_send_failed += tx_send_failed;
1550 	adapter->tx_map_failed += tx_map_failed;
1551 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1552 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1553 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1554 
1555 	return ret;
1556 }
1557 
1558 static void ibmvnic_set_multi(struct net_device *netdev)
1559 {
1560 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1561 	struct netdev_hw_addr *ha;
1562 	union ibmvnic_crq crq;
1563 
1564 	memset(&crq, 0, sizeof(crq));
1565 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1566 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1567 
1568 	if (netdev->flags & IFF_PROMISC) {
1569 		if (!adapter->promisc_supported)
1570 			return;
1571 	} else {
1572 		if (netdev->flags & IFF_ALLMULTI) {
1573 			/* Accept all multicast */
1574 			memset(&crq, 0, sizeof(crq));
1575 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1576 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1577 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1578 			ibmvnic_send_crq(adapter, &crq);
1579 		} else if (netdev_mc_empty(netdev)) {
1580 			/* Reject all multicast */
1581 			memset(&crq, 0, sizeof(crq));
1582 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1583 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1584 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1585 			ibmvnic_send_crq(adapter, &crq);
1586 		} else {
1587 			/* Accept one or more multicast(s) */
1588 			netdev_for_each_mc_addr(ha, netdev) {
1589 				memset(&crq, 0, sizeof(crq));
1590 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1591 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1592 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1593 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1594 						ha->addr);
1595 				ibmvnic_send_crq(adapter, &crq);
1596 			}
1597 		}
1598 	}
1599 }
1600 
1601 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1602 {
1603 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1604 	struct sockaddr *addr = p;
1605 	union ibmvnic_crq crq;
1606 
1607 	if (!is_valid_ether_addr(addr->sa_data))
1608 		return -EADDRNOTAVAIL;
1609 
1610 	memset(&crq, 0, sizeof(crq));
1611 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1612 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1613 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1614 
1615 	init_completion(&adapter->fw_done);
1616 	ibmvnic_send_crq(adapter, &crq);
1617 	wait_for_completion(&adapter->fw_done);
1618 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1619 	return adapter->fw_done_rc ? -EIO : 0;
1620 }
1621 
1622 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1623 {
1624 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1625 	struct sockaddr *addr = p;
1626 	int rc;
1627 
1628 	if (adapter->state == VNIC_PROBED) {
1629 		memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1630 		adapter->mac_change_pending = true;
1631 		return 0;
1632 	}
1633 
1634 	rc = __ibmvnic_set_mac(netdev, addr);
1635 
1636 	return rc;
1637 }
1638 
1639 /**
1640  * do_reset returns zero if we are able to keep processing reset events, or
1641  * non-zero if we hit a fatal error and must halt.
1642  */
1643 static int do_reset(struct ibmvnic_adapter *adapter,
1644 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1645 {
1646 	u64 old_num_rx_queues, old_num_tx_queues;
1647 	struct net_device *netdev = adapter->netdev;
1648 	int i, rc;
1649 
1650 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1651 		   rwi->reset_reason);
1652 
1653 	netif_carrier_off(netdev);
1654 	adapter->reset_reason = rwi->reset_reason;
1655 
1656 	old_num_rx_queues = adapter->req_rx_queues;
1657 	old_num_tx_queues = adapter->req_tx_queues;
1658 
1659 	if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1660 		rc = ibmvnic_reenable_crq_queue(adapter);
1661 		if (rc)
1662 			return 0;
1663 	}
1664 
1665 	rc = __ibmvnic_close(netdev);
1666 	if (rc)
1667 		return rc;
1668 
1669 	if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1670 	    adapter->wait_for_reset) {
1671 		release_resources(adapter);
1672 		release_sub_crqs(adapter, 1);
1673 		release_crq_queue(adapter);
1674 	}
1675 
1676 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1677 		/* remove the closed state so when we call open it appears
1678 		 * we are coming from the probed state.
1679 		 */
1680 		adapter->state = VNIC_PROBED;
1681 
1682 		rc = ibmvnic_init(adapter);
1683 		if (rc)
1684 			return IBMVNIC_INIT_FAILED;
1685 
1686 		/* If the adapter was in PROBE state prior to the reset,
1687 		 * exit here.
1688 		 */
1689 		if (reset_state == VNIC_PROBED)
1690 			return 0;
1691 
1692 		rc = ibmvnic_login(netdev);
1693 		if (rc) {
1694 			adapter->state = VNIC_PROBED;
1695 			return 0;
1696 		}
1697 
1698 		if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1699 		    adapter->wait_for_reset) {
1700 			rc = init_resources(adapter);
1701 			if (rc)
1702 				return rc;
1703 		} else if (adapter->req_rx_queues != old_num_rx_queues ||
1704 			   adapter->req_tx_queues != old_num_tx_queues) {
1705 			adapter->map_id = 1;
1706 			release_rx_pools(adapter);
1707 			release_tx_pools(adapter);
1708 			init_rx_pools(netdev);
1709 			init_tx_pools(netdev);
1710 
1711 			release_napi(adapter);
1712 			init_napi(adapter);
1713 		} else {
1714 			rc = reset_tx_pools(adapter);
1715 			if (rc)
1716 				return rc;
1717 
1718 			rc = reset_rx_pools(adapter);
1719 			if (rc)
1720 				return rc;
1721 
1722 			if (reset_state == VNIC_CLOSED)
1723 				return 0;
1724 		}
1725 	}
1726 
1727 	rc = __ibmvnic_open(netdev);
1728 	if (rc) {
1729 		if (list_empty(&adapter->rwi_list))
1730 			adapter->state = VNIC_CLOSED;
1731 		else
1732 			adapter->state = reset_state;
1733 
1734 		return 0;
1735 	}
1736 
1737 	/* kick napi */
1738 	for (i = 0; i < adapter->req_rx_queues; i++)
1739 		napi_schedule(&adapter->napi[i]);
1740 
1741 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1742 		netdev_notify_peers(netdev);
1743 
1744 	netif_carrier_on(netdev);
1745 
1746 	return 0;
1747 }
1748 
1749 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1750 {
1751 	struct ibmvnic_rwi *rwi;
1752 
1753 	mutex_lock(&adapter->rwi_lock);
1754 
1755 	if (!list_empty(&adapter->rwi_list)) {
1756 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1757 				       list);
1758 		list_del(&rwi->list);
1759 	} else {
1760 		rwi = NULL;
1761 	}
1762 
1763 	mutex_unlock(&adapter->rwi_lock);
1764 	return rwi;
1765 }
1766 
1767 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1768 {
1769 	struct ibmvnic_rwi *rwi;
1770 
1771 	rwi = get_next_rwi(adapter);
1772 	while (rwi) {
1773 		kfree(rwi);
1774 		rwi = get_next_rwi(adapter);
1775 	}
1776 }
1777 
1778 static void __ibmvnic_reset(struct work_struct *work)
1779 {
1780 	struct ibmvnic_rwi *rwi;
1781 	struct ibmvnic_adapter *adapter;
1782 	struct net_device *netdev;
1783 	u32 reset_state;
1784 	int rc = 0;
1785 
1786 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1787 	netdev = adapter->netdev;
1788 
1789 	mutex_lock(&adapter->reset_lock);
1790 	adapter->resetting = true;
1791 	reset_state = adapter->state;
1792 
1793 	rwi = get_next_rwi(adapter);
1794 	while (rwi) {
1795 		rc = do_reset(adapter, rwi, reset_state);
1796 		kfree(rwi);
1797 		if (rc && rc != IBMVNIC_INIT_FAILED)
1798 			break;
1799 
1800 		rwi = get_next_rwi(adapter);
1801 	}
1802 
1803 	if (adapter->wait_for_reset) {
1804 		adapter->wait_for_reset = false;
1805 		adapter->reset_done_rc = rc;
1806 		complete(&adapter->reset_done);
1807 	}
1808 
1809 	if (rc) {
1810 		netdev_dbg(adapter->netdev, "Reset failed\n");
1811 		free_all_rwi(adapter);
1812 		mutex_unlock(&adapter->reset_lock);
1813 		return;
1814 	}
1815 
1816 	adapter->resetting = false;
1817 	mutex_unlock(&adapter->reset_lock);
1818 }
1819 
1820 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1821 			  enum ibmvnic_reset_reason reason)
1822 {
1823 	struct ibmvnic_rwi *rwi, *tmp;
1824 	struct net_device *netdev = adapter->netdev;
1825 	struct list_head *entry;
1826 
1827 	if (adapter->state == VNIC_REMOVING ||
1828 	    adapter->state == VNIC_REMOVED) {
1829 		netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1830 		return;
1831 	}
1832 
1833 	if (adapter->state == VNIC_PROBING) {
1834 		netdev_warn(netdev, "Adapter reset during probe\n");
1835 		adapter->init_done_rc = EAGAIN;
1836 		return;
1837 	}
1838 
1839 	mutex_lock(&adapter->rwi_lock);
1840 
1841 	list_for_each(entry, &adapter->rwi_list) {
1842 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
1843 		if (tmp->reset_reason == reason) {
1844 			netdev_dbg(netdev, "Skipping matching reset\n");
1845 			mutex_unlock(&adapter->rwi_lock);
1846 			return;
1847 		}
1848 	}
1849 
1850 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1851 	if (!rwi) {
1852 		mutex_unlock(&adapter->rwi_lock);
1853 		ibmvnic_close(netdev);
1854 		return;
1855 	}
1856 
1857 	rwi->reset_reason = reason;
1858 	list_add_tail(&rwi->list, &adapter->rwi_list);
1859 	mutex_unlock(&adapter->rwi_lock);
1860 
1861 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1862 	schedule_work(&adapter->ibmvnic_reset);
1863 }
1864 
1865 static void ibmvnic_tx_timeout(struct net_device *dev)
1866 {
1867 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1868 
1869 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1870 }
1871 
1872 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1873 				  struct ibmvnic_rx_buff *rx_buff)
1874 {
1875 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1876 
1877 	rx_buff->skb = NULL;
1878 
1879 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1880 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1881 
1882 	atomic_dec(&pool->available);
1883 }
1884 
1885 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1886 {
1887 	struct net_device *netdev = napi->dev;
1888 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1889 	int scrq_num = (int)(napi - adapter->napi);
1890 	int frames_processed = 0;
1891 
1892 restart_poll:
1893 	while (frames_processed < budget) {
1894 		struct sk_buff *skb;
1895 		struct ibmvnic_rx_buff *rx_buff;
1896 		union sub_crq *next;
1897 		u32 length;
1898 		u16 offset;
1899 		u8 flags = 0;
1900 
1901 		if (unlikely(adapter->resetting &&
1902 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
1903 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1904 			napi_complete_done(napi, frames_processed);
1905 			return frames_processed;
1906 		}
1907 
1908 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1909 			break;
1910 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1911 		rx_buff =
1912 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1913 							  rx_comp.correlator);
1914 		/* do error checking */
1915 		if (next->rx_comp.rc) {
1916 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1917 				   be16_to_cpu(next->rx_comp.rc));
1918 			/* free the entry */
1919 			next->rx_comp.first = 0;
1920 			dev_kfree_skb_any(rx_buff->skb);
1921 			remove_buff_from_pool(adapter, rx_buff);
1922 			continue;
1923 		} else if (!rx_buff->skb) {
1924 			/* free the entry */
1925 			next->rx_comp.first = 0;
1926 			remove_buff_from_pool(adapter, rx_buff);
1927 			continue;
1928 		}
1929 
1930 		length = be32_to_cpu(next->rx_comp.len);
1931 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
1932 		flags = next->rx_comp.flags;
1933 		skb = rx_buff->skb;
1934 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
1935 					length);
1936 
1937 		/* VLAN Header has been stripped by the system firmware and
1938 		 * needs to be inserted by the driver
1939 		 */
1940 		if (adapter->rx_vlan_header_insertion &&
1941 		    (flags & IBMVNIC_VLAN_STRIPPED))
1942 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1943 					       ntohs(next->rx_comp.vlan_tci));
1944 
1945 		/* free the entry */
1946 		next->rx_comp.first = 0;
1947 		remove_buff_from_pool(adapter, rx_buff);
1948 
1949 		skb_put(skb, length);
1950 		skb->protocol = eth_type_trans(skb, netdev);
1951 		skb_record_rx_queue(skb, scrq_num);
1952 
1953 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1954 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1955 			skb->ip_summed = CHECKSUM_UNNECESSARY;
1956 		}
1957 
1958 		length = skb->len;
1959 		napi_gro_receive(napi, skb); /* send it up */
1960 		netdev->stats.rx_packets++;
1961 		netdev->stats.rx_bytes += length;
1962 		adapter->rx_stats_buffers[scrq_num].packets++;
1963 		adapter->rx_stats_buffers[scrq_num].bytes += length;
1964 		frames_processed++;
1965 	}
1966 
1967 	if (adapter->state != VNIC_CLOSING)
1968 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1969 
1970 	if (frames_processed < budget) {
1971 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1972 		napi_complete_done(napi, frames_processed);
1973 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1974 		    napi_reschedule(napi)) {
1975 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1976 			goto restart_poll;
1977 		}
1978 	}
1979 	return frames_processed;
1980 }
1981 
1982 #ifdef CONFIG_NET_POLL_CONTROLLER
1983 static void ibmvnic_netpoll_controller(struct net_device *dev)
1984 {
1985 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1986 	int i;
1987 
1988 	replenish_pools(netdev_priv(dev));
1989 	for (i = 0; i < adapter->req_rx_queues; i++)
1990 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1991 				     adapter->rx_scrq[i]);
1992 }
1993 #endif
1994 
1995 static int wait_for_reset(struct ibmvnic_adapter *adapter)
1996 {
1997 	adapter->fallback.mtu = adapter->req_mtu;
1998 	adapter->fallback.rx_queues = adapter->req_rx_queues;
1999 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2000 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2001 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2002 
2003 	init_completion(&adapter->reset_done);
2004 	ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2005 	adapter->wait_for_reset = true;
2006 	wait_for_completion(&adapter->reset_done);
2007 
2008 	if (adapter->reset_done_rc) {
2009 		adapter->desired.mtu = adapter->fallback.mtu;
2010 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2011 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2012 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2013 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2014 
2015 		init_completion(&adapter->reset_done);
2016 		ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2017 		wait_for_completion(&adapter->reset_done);
2018 	}
2019 	adapter->wait_for_reset = false;
2020 
2021 	return adapter->reset_done_rc;
2022 }
2023 
2024 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2025 {
2026 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2027 
2028 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2029 
2030 	return wait_for_reset(adapter);
2031 }
2032 
2033 static const struct net_device_ops ibmvnic_netdev_ops = {
2034 	.ndo_open		= ibmvnic_open,
2035 	.ndo_stop		= ibmvnic_close,
2036 	.ndo_start_xmit		= ibmvnic_xmit,
2037 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2038 	.ndo_set_mac_address	= ibmvnic_set_mac,
2039 	.ndo_validate_addr	= eth_validate_addr,
2040 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2041 #ifdef CONFIG_NET_POLL_CONTROLLER
2042 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
2043 #endif
2044 	.ndo_change_mtu		= ibmvnic_change_mtu,
2045 };
2046 
2047 /* ethtool functions */
2048 
2049 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2050 				      struct ethtool_link_ksettings *cmd)
2051 {
2052 	u32 supported, advertising;
2053 
2054 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2055 			  SUPPORTED_FIBRE);
2056 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2057 			    ADVERTISED_FIBRE);
2058 	cmd->base.speed = SPEED_1000;
2059 	cmd->base.duplex = DUPLEX_FULL;
2060 	cmd->base.port = PORT_FIBRE;
2061 	cmd->base.phy_address = 0;
2062 	cmd->base.autoneg = AUTONEG_ENABLE;
2063 
2064 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2065 						supported);
2066 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2067 						advertising);
2068 
2069 	return 0;
2070 }
2071 
2072 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2073 				struct ethtool_drvinfo *info)
2074 {
2075 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2076 
2077 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2078 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2079 	strlcpy(info->fw_version, adapter->fw_version,
2080 		sizeof(info->fw_version));
2081 }
2082 
2083 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2084 {
2085 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2086 
2087 	return adapter->msg_enable;
2088 }
2089 
2090 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2091 {
2092 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2093 
2094 	adapter->msg_enable = data;
2095 }
2096 
2097 static u32 ibmvnic_get_link(struct net_device *netdev)
2098 {
2099 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2100 
2101 	/* Don't need to send a query because we request a logical link up at
2102 	 * init and then we wait for link state indications
2103 	 */
2104 	return adapter->logical_link_state;
2105 }
2106 
2107 static void ibmvnic_get_ringparam(struct net_device *netdev,
2108 				  struct ethtool_ringparam *ring)
2109 {
2110 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2111 
2112 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2113 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2114 	ring->rx_mini_max_pending = 0;
2115 	ring->rx_jumbo_max_pending = 0;
2116 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2117 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2118 	ring->rx_mini_pending = 0;
2119 	ring->rx_jumbo_pending = 0;
2120 }
2121 
2122 static int ibmvnic_set_ringparam(struct net_device *netdev,
2123 				 struct ethtool_ringparam *ring)
2124 {
2125 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2126 
2127 	if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2128 	    ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2129 		netdev_err(netdev, "Invalid request.\n");
2130 		netdev_err(netdev, "Max tx buffers = %llu\n",
2131 			   adapter->max_rx_add_entries_per_subcrq);
2132 		netdev_err(netdev, "Max rx buffers = %llu\n",
2133 			   adapter->max_tx_entries_per_subcrq);
2134 		return -EINVAL;
2135 	}
2136 
2137 	adapter->desired.rx_entries = ring->rx_pending;
2138 	adapter->desired.tx_entries = ring->tx_pending;
2139 
2140 	return wait_for_reset(adapter);
2141 }
2142 
2143 static void ibmvnic_get_channels(struct net_device *netdev,
2144 				 struct ethtool_channels *channels)
2145 {
2146 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2147 
2148 	channels->max_rx = adapter->max_rx_queues;
2149 	channels->max_tx = adapter->max_tx_queues;
2150 	channels->max_other = 0;
2151 	channels->max_combined = 0;
2152 	channels->rx_count = adapter->req_rx_queues;
2153 	channels->tx_count = adapter->req_tx_queues;
2154 	channels->other_count = 0;
2155 	channels->combined_count = 0;
2156 }
2157 
2158 static int ibmvnic_set_channels(struct net_device *netdev,
2159 				struct ethtool_channels *channels)
2160 {
2161 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2162 
2163 	adapter->desired.rx_queues = channels->rx_count;
2164 	adapter->desired.tx_queues = channels->tx_count;
2165 
2166 	return wait_for_reset(adapter);
2167 }
2168 
2169 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2170 {
2171 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2172 	int i;
2173 
2174 	if (stringset != ETH_SS_STATS)
2175 		return;
2176 
2177 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2178 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2179 
2180 	for (i = 0; i < adapter->req_tx_queues; i++) {
2181 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2182 		data += ETH_GSTRING_LEN;
2183 
2184 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2185 		data += ETH_GSTRING_LEN;
2186 
2187 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2188 		data += ETH_GSTRING_LEN;
2189 	}
2190 
2191 	for (i = 0; i < adapter->req_rx_queues; i++) {
2192 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2193 		data += ETH_GSTRING_LEN;
2194 
2195 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2196 		data += ETH_GSTRING_LEN;
2197 
2198 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2199 		data += ETH_GSTRING_LEN;
2200 	}
2201 }
2202 
2203 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2204 {
2205 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2206 
2207 	switch (sset) {
2208 	case ETH_SS_STATS:
2209 		return ARRAY_SIZE(ibmvnic_stats) +
2210 		       adapter->req_tx_queues * NUM_TX_STATS +
2211 		       adapter->req_rx_queues * NUM_RX_STATS;
2212 	default:
2213 		return -EOPNOTSUPP;
2214 	}
2215 }
2216 
2217 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2218 				      struct ethtool_stats *stats, u64 *data)
2219 {
2220 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2221 	union ibmvnic_crq crq;
2222 	int i, j;
2223 
2224 	memset(&crq, 0, sizeof(crq));
2225 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2226 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2227 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2228 	crq.request_statistics.len =
2229 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2230 
2231 	/* Wait for data to be written */
2232 	init_completion(&adapter->stats_done);
2233 	ibmvnic_send_crq(adapter, &crq);
2234 	wait_for_completion(&adapter->stats_done);
2235 
2236 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2237 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2238 						ibmvnic_stats[i].offset));
2239 
2240 	for (j = 0; j < adapter->req_tx_queues; j++) {
2241 		data[i] = adapter->tx_stats_buffers[j].packets;
2242 		i++;
2243 		data[i] = adapter->tx_stats_buffers[j].bytes;
2244 		i++;
2245 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2246 		i++;
2247 	}
2248 
2249 	for (j = 0; j < adapter->req_rx_queues; j++) {
2250 		data[i] = adapter->rx_stats_buffers[j].packets;
2251 		i++;
2252 		data[i] = adapter->rx_stats_buffers[j].bytes;
2253 		i++;
2254 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2255 		i++;
2256 	}
2257 }
2258 
2259 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2260 	.get_drvinfo		= ibmvnic_get_drvinfo,
2261 	.get_msglevel		= ibmvnic_get_msglevel,
2262 	.set_msglevel		= ibmvnic_set_msglevel,
2263 	.get_link		= ibmvnic_get_link,
2264 	.get_ringparam		= ibmvnic_get_ringparam,
2265 	.set_ringparam		= ibmvnic_set_ringparam,
2266 	.get_channels		= ibmvnic_get_channels,
2267 	.set_channels		= ibmvnic_set_channels,
2268 	.get_strings            = ibmvnic_get_strings,
2269 	.get_sset_count         = ibmvnic_get_sset_count,
2270 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2271 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2272 };
2273 
2274 /* Routines for managing CRQs/sCRQs  */
2275 
2276 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2277 				   struct ibmvnic_sub_crq_queue *scrq)
2278 {
2279 	int rc;
2280 
2281 	if (scrq->irq) {
2282 		free_irq(scrq->irq, scrq);
2283 		irq_dispose_mapping(scrq->irq);
2284 		scrq->irq = 0;
2285 	}
2286 
2287 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2288 	scrq->cur = 0;
2289 
2290 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2291 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2292 	return rc;
2293 }
2294 
2295 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2296 {
2297 	int i, rc;
2298 
2299 	for (i = 0; i < adapter->req_tx_queues; i++) {
2300 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2301 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2302 		if (rc)
2303 			return rc;
2304 	}
2305 
2306 	for (i = 0; i < adapter->req_rx_queues; i++) {
2307 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2308 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2309 		if (rc)
2310 			return rc;
2311 	}
2312 
2313 	return rc;
2314 }
2315 
2316 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2317 				  struct ibmvnic_sub_crq_queue *scrq,
2318 				  bool do_h_free)
2319 {
2320 	struct device *dev = &adapter->vdev->dev;
2321 	long rc;
2322 
2323 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2324 
2325 	if (do_h_free) {
2326 		/* Close the sub-crqs */
2327 		do {
2328 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2329 						adapter->vdev->unit_address,
2330 						scrq->crq_num);
2331 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2332 
2333 		if (rc) {
2334 			netdev_err(adapter->netdev,
2335 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2336 				   scrq->crq_num, rc);
2337 		}
2338 	}
2339 
2340 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2341 			 DMA_BIDIRECTIONAL);
2342 	free_pages((unsigned long)scrq->msgs, 2);
2343 	kfree(scrq);
2344 }
2345 
2346 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2347 							*adapter)
2348 {
2349 	struct device *dev = &adapter->vdev->dev;
2350 	struct ibmvnic_sub_crq_queue *scrq;
2351 	int rc;
2352 
2353 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2354 	if (!scrq)
2355 		return NULL;
2356 
2357 	scrq->msgs =
2358 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2359 	if (!scrq->msgs) {
2360 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2361 		goto zero_page_failed;
2362 	}
2363 
2364 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2365 					 DMA_BIDIRECTIONAL);
2366 	if (dma_mapping_error(dev, scrq->msg_token)) {
2367 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2368 		goto map_failed;
2369 	}
2370 
2371 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2372 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2373 
2374 	if (rc == H_RESOURCE)
2375 		rc = ibmvnic_reset_crq(adapter);
2376 
2377 	if (rc == H_CLOSED) {
2378 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2379 	} else if (rc) {
2380 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2381 		goto reg_failed;
2382 	}
2383 
2384 	scrq->adapter = adapter;
2385 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2386 	spin_lock_init(&scrq->lock);
2387 
2388 	netdev_dbg(adapter->netdev,
2389 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2390 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2391 
2392 	return scrq;
2393 
2394 reg_failed:
2395 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2396 			 DMA_BIDIRECTIONAL);
2397 map_failed:
2398 	free_pages((unsigned long)scrq->msgs, 2);
2399 zero_page_failed:
2400 	kfree(scrq);
2401 
2402 	return NULL;
2403 }
2404 
2405 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2406 {
2407 	int i;
2408 
2409 	if (adapter->tx_scrq) {
2410 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2411 			if (!adapter->tx_scrq[i])
2412 				continue;
2413 
2414 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2415 				   i);
2416 			if (adapter->tx_scrq[i]->irq) {
2417 				free_irq(adapter->tx_scrq[i]->irq,
2418 					 adapter->tx_scrq[i]);
2419 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2420 				adapter->tx_scrq[i]->irq = 0;
2421 			}
2422 
2423 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2424 					      do_h_free);
2425 		}
2426 
2427 		kfree(adapter->tx_scrq);
2428 		adapter->tx_scrq = NULL;
2429 		adapter->num_active_tx_scrqs = 0;
2430 	}
2431 
2432 	if (adapter->rx_scrq) {
2433 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2434 			if (!adapter->rx_scrq[i])
2435 				continue;
2436 
2437 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2438 				   i);
2439 			if (adapter->rx_scrq[i]->irq) {
2440 				free_irq(adapter->rx_scrq[i]->irq,
2441 					 adapter->rx_scrq[i]);
2442 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2443 				adapter->rx_scrq[i]->irq = 0;
2444 			}
2445 
2446 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2447 					      do_h_free);
2448 		}
2449 
2450 		kfree(adapter->rx_scrq);
2451 		adapter->rx_scrq = NULL;
2452 		adapter->num_active_rx_scrqs = 0;
2453 	}
2454 }
2455 
2456 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2457 			    struct ibmvnic_sub_crq_queue *scrq)
2458 {
2459 	struct device *dev = &adapter->vdev->dev;
2460 	unsigned long rc;
2461 
2462 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2463 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2464 	if (rc)
2465 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2466 			scrq->hw_irq, rc);
2467 	return rc;
2468 }
2469 
2470 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2471 			   struct ibmvnic_sub_crq_queue *scrq)
2472 {
2473 	struct device *dev = &adapter->vdev->dev;
2474 	unsigned long rc;
2475 
2476 	if (scrq->hw_irq > 0x100000000ULL) {
2477 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2478 		return 1;
2479 	}
2480 
2481 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2482 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2483 	if (rc)
2484 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2485 			scrq->hw_irq, rc);
2486 	return rc;
2487 }
2488 
2489 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2490 			       struct ibmvnic_sub_crq_queue *scrq)
2491 {
2492 	struct device *dev = &adapter->vdev->dev;
2493 	struct ibmvnic_tx_buff *txbuff;
2494 	union sub_crq *next;
2495 	int index;
2496 	int i, j;
2497 	u8 first;
2498 
2499 restart_loop:
2500 	while (pending_scrq(adapter, scrq)) {
2501 		unsigned int pool = scrq->pool_index;
2502 		int num_entries = 0;
2503 
2504 		next = ibmvnic_next_scrq(adapter, scrq);
2505 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2506 			if (next->tx_comp.rcs[i]) {
2507 				dev_err(dev, "tx error %x\n",
2508 					next->tx_comp.rcs[i]);
2509 				continue;
2510 			}
2511 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2512 			txbuff = &adapter->tx_pool[pool].tx_buff[index];
2513 
2514 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2515 				if (!txbuff->data_dma[j])
2516 					continue;
2517 
2518 				txbuff->data_dma[j] = 0;
2519 			}
2520 			/* if sub_crq was sent indirectly */
2521 			first = txbuff->indir_arr[0].generic.first;
2522 			if (first == IBMVNIC_CRQ_CMD) {
2523 				dma_unmap_single(dev, txbuff->indir_dma,
2524 						 sizeof(txbuff->indir_arr),
2525 						 DMA_TO_DEVICE);
2526 			}
2527 
2528 			if (txbuff->last_frag) {
2529 				dev_kfree_skb_any(txbuff->skb);
2530 				txbuff->skb = NULL;
2531 			}
2532 
2533 			num_entries += txbuff->num_entries;
2534 
2535 			adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2536 						     producer_index] = index;
2537 			adapter->tx_pool[pool].producer_index =
2538 			    (adapter->tx_pool[pool].producer_index + 1) %
2539 			    adapter->req_tx_entries_per_subcrq;
2540 		}
2541 		/* remove tx_comp scrq*/
2542 		next->tx_comp.first = 0;
2543 
2544 		if (atomic_sub_return(num_entries, &scrq->used) <=
2545 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2546 		    __netif_subqueue_stopped(adapter->netdev,
2547 					     scrq->pool_index)) {
2548 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2549 			netdev_info(adapter->netdev, "Started queue %d\n",
2550 				    scrq->pool_index);
2551 		}
2552 	}
2553 
2554 	enable_scrq_irq(adapter, scrq);
2555 
2556 	if (pending_scrq(adapter, scrq)) {
2557 		disable_scrq_irq(adapter, scrq);
2558 		goto restart_loop;
2559 	}
2560 
2561 	return 0;
2562 }
2563 
2564 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2565 {
2566 	struct ibmvnic_sub_crq_queue *scrq = instance;
2567 	struct ibmvnic_adapter *adapter = scrq->adapter;
2568 
2569 	disable_scrq_irq(adapter, scrq);
2570 	ibmvnic_complete_tx(adapter, scrq);
2571 
2572 	return IRQ_HANDLED;
2573 }
2574 
2575 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2576 {
2577 	struct ibmvnic_sub_crq_queue *scrq = instance;
2578 	struct ibmvnic_adapter *adapter = scrq->adapter;
2579 
2580 	/* When booting a kdump kernel we can hit pending interrupts
2581 	 * prior to completing driver initialization.
2582 	 */
2583 	if (unlikely(adapter->state != VNIC_OPEN))
2584 		return IRQ_NONE;
2585 
2586 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2587 
2588 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2589 		disable_scrq_irq(adapter, scrq);
2590 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2591 	}
2592 
2593 	return IRQ_HANDLED;
2594 }
2595 
2596 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2597 {
2598 	struct device *dev = &adapter->vdev->dev;
2599 	struct ibmvnic_sub_crq_queue *scrq;
2600 	int i = 0, j = 0;
2601 	int rc = 0;
2602 
2603 	for (i = 0; i < adapter->req_tx_queues; i++) {
2604 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2605 			   i);
2606 		scrq = adapter->tx_scrq[i];
2607 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2608 
2609 		if (!scrq->irq) {
2610 			rc = -EINVAL;
2611 			dev_err(dev, "Error mapping irq\n");
2612 			goto req_tx_irq_failed;
2613 		}
2614 
2615 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2616 				 0, "ibmvnic_tx", scrq);
2617 
2618 		if (rc) {
2619 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2620 				scrq->irq, rc);
2621 			irq_dispose_mapping(scrq->irq);
2622 			goto req_tx_irq_failed;
2623 		}
2624 	}
2625 
2626 	for (i = 0; i < adapter->req_rx_queues; i++) {
2627 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2628 			   i);
2629 		scrq = adapter->rx_scrq[i];
2630 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2631 		if (!scrq->irq) {
2632 			rc = -EINVAL;
2633 			dev_err(dev, "Error mapping irq\n");
2634 			goto req_rx_irq_failed;
2635 		}
2636 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2637 				 0, "ibmvnic_rx", scrq);
2638 		if (rc) {
2639 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2640 				scrq->irq, rc);
2641 			irq_dispose_mapping(scrq->irq);
2642 			goto req_rx_irq_failed;
2643 		}
2644 	}
2645 	return rc;
2646 
2647 req_rx_irq_failed:
2648 	for (j = 0; j < i; j++) {
2649 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2650 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2651 	}
2652 	i = adapter->req_tx_queues;
2653 req_tx_irq_failed:
2654 	for (j = 0; j < i; j++) {
2655 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2656 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2657 	}
2658 	release_sub_crqs(adapter, 1);
2659 	return rc;
2660 }
2661 
2662 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2663 {
2664 	struct device *dev = &adapter->vdev->dev;
2665 	struct ibmvnic_sub_crq_queue **allqueues;
2666 	int registered_queues = 0;
2667 	int total_queues;
2668 	int more = 0;
2669 	int i;
2670 
2671 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2672 
2673 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2674 	if (!allqueues)
2675 		return -1;
2676 
2677 	for (i = 0; i < total_queues; i++) {
2678 		allqueues[i] = init_sub_crq_queue(adapter);
2679 		if (!allqueues[i]) {
2680 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2681 			break;
2682 		}
2683 		registered_queues++;
2684 	}
2685 
2686 	/* Make sure we were able to register the minimum number of queues */
2687 	if (registered_queues <
2688 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2689 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2690 		goto tx_failed;
2691 	}
2692 
2693 	/* Distribute the failed allocated queues*/
2694 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2695 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2696 		switch (i % 3) {
2697 		case 0:
2698 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2699 				adapter->req_rx_queues--;
2700 			else
2701 				more++;
2702 			break;
2703 		case 1:
2704 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2705 				adapter->req_tx_queues--;
2706 			else
2707 				more++;
2708 			break;
2709 		}
2710 	}
2711 
2712 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2713 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2714 	if (!adapter->tx_scrq)
2715 		goto tx_failed;
2716 
2717 	for (i = 0; i < adapter->req_tx_queues; i++) {
2718 		adapter->tx_scrq[i] = allqueues[i];
2719 		adapter->tx_scrq[i]->pool_index = i;
2720 		adapter->num_active_tx_scrqs++;
2721 	}
2722 
2723 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2724 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2725 	if (!adapter->rx_scrq)
2726 		goto rx_failed;
2727 
2728 	for (i = 0; i < adapter->req_rx_queues; i++) {
2729 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2730 		adapter->rx_scrq[i]->scrq_num = i;
2731 		adapter->num_active_rx_scrqs++;
2732 	}
2733 
2734 	kfree(allqueues);
2735 	return 0;
2736 
2737 rx_failed:
2738 	kfree(adapter->tx_scrq);
2739 	adapter->tx_scrq = NULL;
2740 tx_failed:
2741 	for (i = 0; i < registered_queues; i++)
2742 		release_sub_crq_queue(adapter, allqueues[i], 1);
2743 	kfree(allqueues);
2744 	return -1;
2745 }
2746 
2747 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2748 {
2749 	struct device *dev = &adapter->vdev->dev;
2750 	union ibmvnic_crq crq;
2751 	int max_entries;
2752 
2753 	if (!retry) {
2754 		/* Sub-CRQ entries are 32 byte long */
2755 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2756 
2757 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
2758 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
2759 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2760 			return;
2761 		}
2762 
2763 		if (adapter->desired.mtu)
2764 			adapter->req_mtu = adapter->desired.mtu;
2765 		else
2766 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2767 
2768 		if (!adapter->desired.tx_entries)
2769 			adapter->desired.tx_entries =
2770 					adapter->max_tx_entries_per_subcrq;
2771 		if (!adapter->desired.rx_entries)
2772 			adapter->desired.rx_entries =
2773 					adapter->max_rx_add_entries_per_subcrq;
2774 
2775 		max_entries = IBMVNIC_MAX_LTB_SIZE /
2776 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
2777 
2778 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2779 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
2780 			adapter->desired.tx_entries = max_entries;
2781 		}
2782 
2783 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
2784 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
2785 			adapter->desired.rx_entries = max_entries;
2786 		}
2787 
2788 		if (adapter->desired.tx_entries)
2789 			adapter->req_tx_entries_per_subcrq =
2790 					adapter->desired.tx_entries;
2791 		else
2792 			adapter->req_tx_entries_per_subcrq =
2793 					adapter->max_tx_entries_per_subcrq;
2794 
2795 		if (adapter->desired.rx_entries)
2796 			adapter->req_rx_add_entries_per_subcrq =
2797 					adapter->desired.rx_entries;
2798 		else
2799 			adapter->req_rx_add_entries_per_subcrq =
2800 					adapter->max_rx_add_entries_per_subcrq;
2801 
2802 		if (adapter->desired.tx_queues)
2803 			adapter->req_tx_queues =
2804 					adapter->desired.tx_queues;
2805 		else
2806 			adapter->req_tx_queues =
2807 					adapter->opt_tx_comp_sub_queues;
2808 
2809 		if (adapter->desired.rx_queues)
2810 			adapter->req_rx_queues =
2811 					adapter->desired.rx_queues;
2812 		else
2813 			adapter->req_rx_queues =
2814 					adapter->opt_rx_comp_queues;
2815 
2816 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2817 	}
2818 
2819 	memset(&crq, 0, sizeof(crq));
2820 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
2821 	crq.request_capability.cmd = REQUEST_CAPABILITY;
2822 
2823 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2824 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2825 	atomic_inc(&adapter->running_cap_crqs);
2826 	ibmvnic_send_crq(adapter, &crq);
2827 
2828 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2829 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2830 	atomic_inc(&adapter->running_cap_crqs);
2831 	ibmvnic_send_crq(adapter, &crq);
2832 
2833 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2834 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2835 	atomic_inc(&adapter->running_cap_crqs);
2836 	ibmvnic_send_crq(adapter, &crq);
2837 
2838 	crq.request_capability.capability =
2839 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2840 	crq.request_capability.number =
2841 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2842 	atomic_inc(&adapter->running_cap_crqs);
2843 	ibmvnic_send_crq(adapter, &crq);
2844 
2845 	crq.request_capability.capability =
2846 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2847 	crq.request_capability.number =
2848 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2849 	atomic_inc(&adapter->running_cap_crqs);
2850 	ibmvnic_send_crq(adapter, &crq);
2851 
2852 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2853 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2854 	atomic_inc(&adapter->running_cap_crqs);
2855 	ibmvnic_send_crq(adapter, &crq);
2856 
2857 	if (adapter->netdev->flags & IFF_PROMISC) {
2858 		if (adapter->promisc_supported) {
2859 			crq.request_capability.capability =
2860 			    cpu_to_be16(PROMISC_REQUESTED);
2861 			crq.request_capability.number = cpu_to_be64(1);
2862 			atomic_inc(&adapter->running_cap_crqs);
2863 			ibmvnic_send_crq(adapter, &crq);
2864 		}
2865 	} else {
2866 		crq.request_capability.capability =
2867 		    cpu_to_be16(PROMISC_REQUESTED);
2868 		crq.request_capability.number = cpu_to_be64(0);
2869 		atomic_inc(&adapter->running_cap_crqs);
2870 		ibmvnic_send_crq(adapter, &crq);
2871 	}
2872 }
2873 
2874 static int pending_scrq(struct ibmvnic_adapter *adapter,
2875 			struct ibmvnic_sub_crq_queue *scrq)
2876 {
2877 	union sub_crq *entry = &scrq->msgs[scrq->cur];
2878 
2879 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2880 		return 1;
2881 	else
2882 		return 0;
2883 }
2884 
2885 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2886 					struct ibmvnic_sub_crq_queue *scrq)
2887 {
2888 	union sub_crq *entry;
2889 	unsigned long flags;
2890 
2891 	spin_lock_irqsave(&scrq->lock, flags);
2892 	entry = &scrq->msgs[scrq->cur];
2893 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2894 		if (++scrq->cur == scrq->size)
2895 			scrq->cur = 0;
2896 	} else {
2897 		entry = NULL;
2898 	}
2899 	spin_unlock_irqrestore(&scrq->lock, flags);
2900 
2901 	return entry;
2902 }
2903 
2904 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2905 {
2906 	struct ibmvnic_crq_queue *queue = &adapter->crq;
2907 	union ibmvnic_crq *crq;
2908 
2909 	crq = &queue->msgs[queue->cur];
2910 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2911 		if (++queue->cur == queue->size)
2912 			queue->cur = 0;
2913 	} else {
2914 		crq = NULL;
2915 	}
2916 
2917 	return crq;
2918 }
2919 
2920 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2921 		       union sub_crq *sub_crq)
2922 {
2923 	unsigned int ua = adapter->vdev->unit_address;
2924 	struct device *dev = &adapter->vdev->dev;
2925 	u64 *u64_crq = (u64 *)sub_crq;
2926 	int rc;
2927 
2928 	netdev_dbg(adapter->netdev,
2929 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2930 		   (unsigned long int)cpu_to_be64(remote_handle),
2931 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2932 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
2933 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
2934 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
2935 
2936 	/* Make sure the hypervisor sees the complete request */
2937 	mb();
2938 
2939 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2940 				cpu_to_be64(remote_handle),
2941 				cpu_to_be64(u64_crq[0]),
2942 				cpu_to_be64(u64_crq[1]),
2943 				cpu_to_be64(u64_crq[2]),
2944 				cpu_to_be64(u64_crq[3]));
2945 
2946 	if (rc) {
2947 		if (rc == H_CLOSED)
2948 			dev_warn(dev, "CRQ Queue closed\n");
2949 		dev_err(dev, "Send error (rc=%d)\n", rc);
2950 	}
2951 
2952 	return rc;
2953 }
2954 
2955 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2956 				u64 remote_handle, u64 ioba, u64 num_entries)
2957 {
2958 	unsigned int ua = adapter->vdev->unit_address;
2959 	struct device *dev = &adapter->vdev->dev;
2960 	int rc;
2961 
2962 	/* Make sure the hypervisor sees the complete request */
2963 	mb();
2964 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2965 				cpu_to_be64(remote_handle),
2966 				ioba, num_entries);
2967 
2968 	if (rc) {
2969 		if (rc == H_CLOSED)
2970 			dev_warn(dev, "CRQ Queue closed\n");
2971 		dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2972 	}
2973 
2974 	return rc;
2975 }
2976 
2977 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2978 			    union ibmvnic_crq *crq)
2979 {
2980 	unsigned int ua = adapter->vdev->unit_address;
2981 	struct device *dev = &adapter->vdev->dev;
2982 	u64 *u64_crq = (u64 *)crq;
2983 	int rc;
2984 
2985 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2986 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2987 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
2988 
2989 	/* Make sure the hypervisor sees the complete request */
2990 	mb();
2991 
2992 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2993 				cpu_to_be64(u64_crq[0]),
2994 				cpu_to_be64(u64_crq[1]));
2995 
2996 	if (rc) {
2997 		if (rc == H_CLOSED) {
2998 			dev_warn(dev, "CRQ Queue closed\n");
2999 			if (adapter->resetting)
3000 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3001 		}
3002 
3003 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3004 	}
3005 
3006 	return rc;
3007 }
3008 
3009 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3010 {
3011 	union ibmvnic_crq crq;
3012 
3013 	memset(&crq, 0, sizeof(crq));
3014 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3015 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3016 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3017 
3018 	return ibmvnic_send_crq(adapter, &crq);
3019 }
3020 
3021 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3022 {
3023 	union ibmvnic_crq crq;
3024 
3025 	memset(&crq, 0, sizeof(crq));
3026 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3027 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3028 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3029 
3030 	return ibmvnic_send_crq(adapter, &crq);
3031 }
3032 
3033 struct vnic_login_client_data {
3034 	u8	type;
3035 	__be16	len;
3036 	char	name;
3037 } __packed;
3038 
3039 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3040 {
3041 	int len;
3042 
3043 	/* Calculate the amount of buffer space needed for the
3044 	 * vnic client data in the login buffer. There are four entries,
3045 	 * OS name, LPAR name, device name, and a null last entry.
3046 	 */
3047 	len = 4 * sizeof(struct vnic_login_client_data);
3048 	len += 6; /* "Linux" plus NULL */
3049 	len += strlen(utsname()->nodename) + 1;
3050 	len += strlen(adapter->netdev->name) + 1;
3051 
3052 	return len;
3053 }
3054 
3055 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3056 				 struct vnic_login_client_data *vlcd)
3057 {
3058 	const char *os_name = "Linux";
3059 	int len;
3060 
3061 	/* Type 1 - LPAR OS */
3062 	vlcd->type = 1;
3063 	len = strlen(os_name) + 1;
3064 	vlcd->len = cpu_to_be16(len);
3065 	strncpy(&vlcd->name, os_name, len);
3066 	vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
3067 
3068 	/* Type 2 - LPAR name */
3069 	vlcd->type = 2;
3070 	len = strlen(utsname()->nodename) + 1;
3071 	vlcd->len = cpu_to_be16(len);
3072 	strncpy(&vlcd->name, utsname()->nodename, len);
3073 	vlcd = (struct vnic_login_client_data *)((char *)&vlcd->name + len);
3074 
3075 	/* Type 3 - device name */
3076 	vlcd->type = 3;
3077 	len = strlen(adapter->netdev->name) + 1;
3078 	vlcd->len = cpu_to_be16(len);
3079 	strncpy(&vlcd->name, adapter->netdev->name, len);
3080 }
3081 
3082 static void send_login(struct ibmvnic_adapter *adapter)
3083 {
3084 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3085 	struct ibmvnic_login_buffer *login_buffer;
3086 	struct device *dev = &adapter->vdev->dev;
3087 	dma_addr_t rsp_buffer_token;
3088 	dma_addr_t buffer_token;
3089 	size_t rsp_buffer_size;
3090 	union ibmvnic_crq crq;
3091 	size_t buffer_size;
3092 	__be64 *tx_list_p;
3093 	__be64 *rx_list_p;
3094 	int client_data_len;
3095 	struct vnic_login_client_data *vlcd;
3096 	int i;
3097 
3098 	release_login_rsp_buffer(adapter);
3099 	client_data_len = vnic_client_data_len(adapter);
3100 
3101 	buffer_size =
3102 	    sizeof(struct ibmvnic_login_buffer) +
3103 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3104 	    client_data_len;
3105 
3106 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3107 	if (!login_buffer)
3108 		goto buf_alloc_failed;
3109 
3110 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3111 				      DMA_TO_DEVICE);
3112 	if (dma_mapping_error(dev, buffer_token)) {
3113 		dev_err(dev, "Couldn't map login buffer\n");
3114 		goto buf_map_failed;
3115 	}
3116 
3117 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3118 			  sizeof(u64) * adapter->req_tx_queues +
3119 			  sizeof(u64) * adapter->req_rx_queues +
3120 			  sizeof(u64) * adapter->req_rx_queues +
3121 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3122 
3123 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3124 	if (!login_rsp_buffer)
3125 		goto buf_rsp_alloc_failed;
3126 
3127 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3128 					  rsp_buffer_size, DMA_FROM_DEVICE);
3129 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3130 		dev_err(dev, "Couldn't map login rsp buffer\n");
3131 		goto buf_rsp_map_failed;
3132 	}
3133 
3134 	adapter->login_buf = login_buffer;
3135 	adapter->login_buf_token = buffer_token;
3136 	adapter->login_buf_sz = buffer_size;
3137 	adapter->login_rsp_buf = login_rsp_buffer;
3138 	adapter->login_rsp_buf_token = rsp_buffer_token;
3139 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3140 
3141 	login_buffer->len = cpu_to_be32(buffer_size);
3142 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3143 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3144 	login_buffer->off_txcomp_subcrqs =
3145 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3146 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3147 	login_buffer->off_rxcomp_subcrqs =
3148 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3149 			sizeof(u64) * adapter->req_tx_queues);
3150 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3151 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3152 
3153 	tx_list_p = (__be64 *)((char *)login_buffer +
3154 				      sizeof(struct ibmvnic_login_buffer));
3155 	rx_list_p = (__be64 *)((char *)login_buffer +
3156 				      sizeof(struct ibmvnic_login_buffer) +
3157 				      sizeof(u64) * adapter->req_tx_queues);
3158 
3159 	for (i = 0; i < adapter->req_tx_queues; i++) {
3160 		if (adapter->tx_scrq[i]) {
3161 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3162 						   crq_num);
3163 		}
3164 	}
3165 
3166 	for (i = 0; i < adapter->req_rx_queues; i++) {
3167 		if (adapter->rx_scrq[i]) {
3168 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3169 						   crq_num);
3170 		}
3171 	}
3172 
3173 	/* Insert vNIC login client data */
3174 	vlcd = (struct vnic_login_client_data *)
3175 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3176 	login_buffer->client_data_offset =
3177 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3178 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3179 
3180 	vnic_add_client_data(adapter, vlcd);
3181 
3182 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3183 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3184 		netdev_dbg(adapter->netdev, "%016lx\n",
3185 			   ((unsigned long int *)(adapter->login_buf))[i]);
3186 	}
3187 
3188 	memset(&crq, 0, sizeof(crq));
3189 	crq.login.first = IBMVNIC_CRQ_CMD;
3190 	crq.login.cmd = LOGIN;
3191 	crq.login.ioba = cpu_to_be32(buffer_token);
3192 	crq.login.len = cpu_to_be32(buffer_size);
3193 	ibmvnic_send_crq(adapter, &crq);
3194 
3195 	return;
3196 
3197 buf_rsp_map_failed:
3198 	kfree(login_rsp_buffer);
3199 buf_rsp_alloc_failed:
3200 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3201 buf_map_failed:
3202 	kfree(login_buffer);
3203 buf_alloc_failed:
3204 	return;
3205 }
3206 
3207 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3208 			     u32 len, u8 map_id)
3209 {
3210 	union ibmvnic_crq crq;
3211 
3212 	memset(&crq, 0, sizeof(crq));
3213 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3214 	crq.request_map.cmd = REQUEST_MAP;
3215 	crq.request_map.map_id = map_id;
3216 	crq.request_map.ioba = cpu_to_be32(addr);
3217 	crq.request_map.len = cpu_to_be32(len);
3218 	ibmvnic_send_crq(adapter, &crq);
3219 }
3220 
3221 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3222 {
3223 	union ibmvnic_crq crq;
3224 
3225 	memset(&crq, 0, sizeof(crq));
3226 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3227 	crq.request_unmap.cmd = REQUEST_UNMAP;
3228 	crq.request_unmap.map_id = map_id;
3229 	ibmvnic_send_crq(adapter, &crq);
3230 }
3231 
3232 static void send_map_query(struct ibmvnic_adapter *adapter)
3233 {
3234 	union ibmvnic_crq crq;
3235 
3236 	memset(&crq, 0, sizeof(crq));
3237 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3238 	crq.query_map.cmd = QUERY_MAP;
3239 	ibmvnic_send_crq(adapter, &crq);
3240 }
3241 
3242 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3243 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3244 {
3245 	union ibmvnic_crq crq;
3246 
3247 	atomic_set(&adapter->running_cap_crqs, 0);
3248 	memset(&crq, 0, sizeof(crq));
3249 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3250 	crq.query_capability.cmd = QUERY_CAPABILITY;
3251 
3252 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3253 	atomic_inc(&adapter->running_cap_crqs);
3254 	ibmvnic_send_crq(adapter, &crq);
3255 
3256 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3257 	atomic_inc(&adapter->running_cap_crqs);
3258 	ibmvnic_send_crq(adapter, &crq);
3259 
3260 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3261 	atomic_inc(&adapter->running_cap_crqs);
3262 	ibmvnic_send_crq(adapter, &crq);
3263 
3264 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3265 	atomic_inc(&adapter->running_cap_crqs);
3266 	ibmvnic_send_crq(adapter, &crq);
3267 
3268 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3269 	atomic_inc(&adapter->running_cap_crqs);
3270 	ibmvnic_send_crq(adapter, &crq);
3271 
3272 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3273 	atomic_inc(&adapter->running_cap_crqs);
3274 	ibmvnic_send_crq(adapter, &crq);
3275 
3276 	crq.query_capability.capability =
3277 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3278 	atomic_inc(&adapter->running_cap_crqs);
3279 	ibmvnic_send_crq(adapter, &crq);
3280 
3281 	crq.query_capability.capability =
3282 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3283 	atomic_inc(&adapter->running_cap_crqs);
3284 	ibmvnic_send_crq(adapter, &crq);
3285 
3286 	crq.query_capability.capability =
3287 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3288 	atomic_inc(&adapter->running_cap_crqs);
3289 	ibmvnic_send_crq(adapter, &crq);
3290 
3291 	crq.query_capability.capability =
3292 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3293 	atomic_inc(&adapter->running_cap_crqs);
3294 	ibmvnic_send_crq(adapter, &crq);
3295 
3296 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3297 	atomic_inc(&adapter->running_cap_crqs);
3298 	ibmvnic_send_crq(adapter, &crq);
3299 
3300 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3301 	atomic_inc(&adapter->running_cap_crqs);
3302 	ibmvnic_send_crq(adapter, &crq);
3303 
3304 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3305 	atomic_inc(&adapter->running_cap_crqs);
3306 	ibmvnic_send_crq(adapter, &crq);
3307 
3308 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3309 	atomic_inc(&adapter->running_cap_crqs);
3310 	ibmvnic_send_crq(adapter, &crq);
3311 
3312 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3313 	atomic_inc(&adapter->running_cap_crqs);
3314 	ibmvnic_send_crq(adapter, &crq);
3315 
3316 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3317 	atomic_inc(&adapter->running_cap_crqs);
3318 	ibmvnic_send_crq(adapter, &crq);
3319 
3320 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3321 	atomic_inc(&adapter->running_cap_crqs);
3322 	ibmvnic_send_crq(adapter, &crq);
3323 
3324 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3325 	atomic_inc(&adapter->running_cap_crqs);
3326 	ibmvnic_send_crq(adapter, &crq);
3327 
3328 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3329 	atomic_inc(&adapter->running_cap_crqs);
3330 	ibmvnic_send_crq(adapter, &crq);
3331 
3332 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3333 	atomic_inc(&adapter->running_cap_crqs);
3334 	ibmvnic_send_crq(adapter, &crq);
3335 
3336 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3337 	atomic_inc(&adapter->running_cap_crqs);
3338 	ibmvnic_send_crq(adapter, &crq);
3339 
3340 	crq.query_capability.capability =
3341 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3342 	atomic_inc(&adapter->running_cap_crqs);
3343 	ibmvnic_send_crq(adapter, &crq);
3344 
3345 	crq.query_capability.capability =
3346 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3347 	atomic_inc(&adapter->running_cap_crqs);
3348 	ibmvnic_send_crq(adapter, &crq);
3349 
3350 	crq.query_capability.capability =
3351 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3352 	atomic_inc(&adapter->running_cap_crqs);
3353 	ibmvnic_send_crq(adapter, &crq);
3354 
3355 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3356 	atomic_inc(&adapter->running_cap_crqs);
3357 	ibmvnic_send_crq(adapter, &crq);
3358 }
3359 
3360 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3361 				struct ibmvnic_adapter *adapter)
3362 {
3363 	struct device *dev = &adapter->vdev->dev;
3364 
3365 	if (crq->get_vpd_size_rsp.rc.code) {
3366 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3367 			crq->get_vpd_size_rsp.rc.code);
3368 		complete(&adapter->fw_done);
3369 		return;
3370 	}
3371 
3372 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3373 	complete(&adapter->fw_done);
3374 }
3375 
3376 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3377 			   struct ibmvnic_adapter *adapter)
3378 {
3379 	struct device *dev = &adapter->vdev->dev;
3380 	unsigned char *substr = NULL;
3381 	u8 fw_level_len = 0;
3382 
3383 	memset(adapter->fw_version, 0, 32);
3384 
3385 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3386 			 DMA_FROM_DEVICE);
3387 
3388 	if (crq->get_vpd_rsp.rc.code) {
3389 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3390 			crq->get_vpd_rsp.rc.code);
3391 		goto complete;
3392 	}
3393 
3394 	/* get the position of the firmware version info
3395 	 * located after the ASCII 'RM' substring in the buffer
3396 	 */
3397 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3398 	if (!substr) {
3399 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3400 		goto complete;
3401 	}
3402 
3403 	/* get length of firmware level ASCII substring */
3404 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3405 		fw_level_len = *(substr + 2);
3406 	} else {
3407 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3408 		goto complete;
3409 	}
3410 
3411 	/* copy firmware version string from vpd into adapter */
3412 	if ((substr + 3 + fw_level_len) <
3413 	    (adapter->vpd->buff + adapter->vpd->len)) {
3414 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3415 	} else {
3416 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3417 	}
3418 
3419 complete:
3420 	if (adapter->fw_version[0] == '\0')
3421 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3422 	complete(&adapter->fw_done);
3423 }
3424 
3425 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3426 {
3427 	struct device *dev = &adapter->vdev->dev;
3428 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3429 	union ibmvnic_crq crq;
3430 	int i;
3431 
3432 	dma_unmap_single(dev, adapter->ip_offload_tok,
3433 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3434 
3435 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3436 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3437 		netdev_dbg(adapter->netdev, "%016lx\n",
3438 			   ((unsigned long int *)(buf))[i]);
3439 
3440 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3441 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3442 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3443 		   buf->tcp_ipv4_chksum);
3444 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3445 		   buf->tcp_ipv6_chksum);
3446 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3447 		   buf->udp_ipv4_chksum);
3448 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3449 		   buf->udp_ipv6_chksum);
3450 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3451 		   buf->large_tx_ipv4);
3452 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3453 		   buf->large_tx_ipv6);
3454 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3455 		   buf->large_rx_ipv4);
3456 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3457 		   buf->large_rx_ipv6);
3458 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3459 		   buf->max_ipv4_header_size);
3460 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3461 		   buf->max_ipv6_header_size);
3462 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3463 		   buf->max_tcp_header_size);
3464 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3465 		   buf->max_udp_header_size);
3466 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3467 		   buf->max_large_tx_size);
3468 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3469 		   buf->max_large_rx_size);
3470 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3471 		   buf->ipv6_extension_header);
3472 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3473 		   buf->tcp_pseudosum_req);
3474 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3475 		   buf->num_ipv6_ext_headers);
3476 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3477 		   buf->off_ipv6_ext_headers);
3478 
3479 	adapter->ip_offload_ctrl_tok =
3480 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
3481 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3482 
3483 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3484 		dev_err(dev, "Couldn't map ip offload control buffer\n");
3485 		return;
3486 	}
3487 
3488 	adapter->ip_offload_ctrl.len =
3489 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3490 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3491 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3492 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3493 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3494 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3495 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3496 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3497 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3498 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3499 
3500 	/* large_rx disabled for now, additional features needed */
3501 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3502 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3503 
3504 	adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3505 
3506 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3507 		adapter->netdev->features |= NETIF_F_IP_CSUM;
3508 
3509 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3510 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3511 
3512 	if ((adapter->netdev->features &
3513 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3514 		adapter->netdev->features |= NETIF_F_RXCSUM;
3515 
3516 	if (buf->large_tx_ipv4)
3517 		adapter->netdev->features |= NETIF_F_TSO;
3518 	if (buf->large_tx_ipv6)
3519 		adapter->netdev->features |= NETIF_F_TSO6;
3520 
3521 	adapter->netdev->hw_features |= adapter->netdev->features;
3522 
3523 	memset(&crq, 0, sizeof(crq));
3524 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3525 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3526 	crq.control_ip_offload.len =
3527 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3528 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3529 	ibmvnic_send_crq(adapter, &crq);
3530 }
3531 
3532 static void handle_error_info_rsp(union ibmvnic_crq *crq,
3533 				  struct ibmvnic_adapter *adapter)
3534 {
3535 	struct device *dev = &adapter->vdev->dev;
3536 	struct ibmvnic_error_buff *error_buff, *tmp;
3537 	unsigned long flags;
3538 	bool found = false;
3539 	int i;
3540 
3541 	if (!crq->request_error_rsp.rc.code) {
3542 		dev_info(dev, "Request Error Rsp returned with rc=%x\n",
3543 			 crq->request_error_rsp.rc.code);
3544 		return;
3545 	}
3546 
3547 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3548 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
3549 		if (error_buff->error_id == crq->request_error_rsp.error_id) {
3550 			found = true;
3551 			list_del(&error_buff->list);
3552 			break;
3553 		}
3554 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3555 
3556 	if (!found) {
3557 		dev_err(dev, "Couldn't find error id %x\n",
3558 			be32_to_cpu(crq->request_error_rsp.error_id));
3559 		return;
3560 	}
3561 
3562 	dev_err(dev, "Detailed info for error id %x:",
3563 		be32_to_cpu(crq->request_error_rsp.error_id));
3564 
3565 	for (i = 0; i < error_buff->len; i++) {
3566 		pr_cont("%02x", (int)error_buff->buff[i]);
3567 		if (i % 8 == 7)
3568 			pr_cont(" ");
3569 	}
3570 	pr_cont("\n");
3571 
3572 	dma_unmap_single(dev, error_buff->dma, error_buff->len,
3573 			 DMA_FROM_DEVICE);
3574 	kfree(error_buff->buff);
3575 	kfree(error_buff);
3576 }
3577 
3578 static void request_error_information(struct ibmvnic_adapter *adapter,
3579 				      union ibmvnic_crq *err_crq)
3580 {
3581 	struct device *dev = &adapter->vdev->dev;
3582 	struct net_device *netdev = adapter->netdev;
3583 	struct ibmvnic_error_buff *error_buff;
3584 	unsigned long timeout = msecs_to_jiffies(30000);
3585 	union ibmvnic_crq crq;
3586 	unsigned long flags;
3587 	int rc, detail_len;
3588 
3589 	error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3590 	if (!error_buff)
3591 		return;
3592 
3593 	detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3594 	error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3595 	if (!error_buff->buff) {
3596 		kfree(error_buff);
3597 		return;
3598 	}
3599 
3600 	error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3601 					 DMA_FROM_DEVICE);
3602 	if (dma_mapping_error(dev, error_buff->dma)) {
3603 		netdev_err(netdev, "Couldn't map error buffer\n");
3604 		kfree(error_buff->buff);
3605 		kfree(error_buff);
3606 		return;
3607 	}
3608 
3609 	error_buff->len = detail_len;
3610 	error_buff->error_id = err_crq->error_indication.error_id;
3611 
3612 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3613 	list_add_tail(&error_buff->list, &adapter->errors);
3614 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3615 
3616 	memset(&crq, 0, sizeof(crq));
3617 	crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3618 	crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3619 	crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3620 	crq.request_error_info.len = cpu_to_be32(detail_len);
3621 	crq.request_error_info.error_id = err_crq->error_indication.error_id;
3622 
3623 	rc = ibmvnic_send_crq(adapter, &crq);
3624 	if (rc) {
3625 		netdev_err(netdev, "failed to request error information\n");
3626 		goto err_info_fail;
3627 	}
3628 
3629 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3630 		netdev_err(netdev, "timeout waiting for error information\n");
3631 		goto err_info_fail;
3632 	}
3633 
3634 	return;
3635 
3636 err_info_fail:
3637 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3638 	list_del(&error_buff->list);
3639 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3640 
3641 	kfree(error_buff->buff);
3642 	kfree(error_buff);
3643 }
3644 
3645 static void handle_error_indication(union ibmvnic_crq *crq,
3646 				    struct ibmvnic_adapter *adapter)
3647 {
3648 	struct device *dev = &adapter->vdev->dev;
3649 
3650 	dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3651 		crq->error_indication.flags
3652 			& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3653 		be32_to_cpu(crq->error_indication.error_id),
3654 		be16_to_cpu(crq->error_indication.error_cause));
3655 
3656 	if (be32_to_cpu(crq->error_indication.error_id))
3657 		request_error_information(adapter, crq);
3658 
3659 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3660 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3661 	else
3662 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3663 }
3664 
3665 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3666 				 struct ibmvnic_adapter *adapter)
3667 {
3668 	struct net_device *netdev = adapter->netdev;
3669 	struct device *dev = &adapter->vdev->dev;
3670 	long rc;
3671 
3672 	rc = crq->change_mac_addr_rsp.rc.code;
3673 	if (rc) {
3674 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3675 		goto out;
3676 	}
3677 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3678 	       ETH_ALEN);
3679 out:
3680 	complete(&adapter->fw_done);
3681 	return rc;
3682 }
3683 
3684 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3685 				   struct ibmvnic_adapter *adapter)
3686 {
3687 	struct device *dev = &adapter->vdev->dev;
3688 	u64 *req_value;
3689 	char *name;
3690 
3691 	atomic_dec(&adapter->running_cap_crqs);
3692 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3693 	case REQ_TX_QUEUES:
3694 		req_value = &adapter->req_tx_queues;
3695 		name = "tx";
3696 		break;
3697 	case REQ_RX_QUEUES:
3698 		req_value = &adapter->req_rx_queues;
3699 		name = "rx";
3700 		break;
3701 	case REQ_RX_ADD_QUEUES:
3702 		req_value = &adapter->req_rx_add_queues;
3703 		name = "rx_add";
3704 		break;
3705 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3706 		req_value = &adapter->req_tx_entries_per_subcrq;
3707 		name = "tx_entries_per_subcrq";
3708 		break;
3709 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3710 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3711 		name = "rx_add_entries_per_subcrq";
3712 		break;
3713 	case REQ_MTU:
3714 		req_value = &adapter->req_mtu;
3715 		name = "mtu";
3716 		break;
3717 	case PROMISC_REQUESTED:
3718 		req_value = &adapter->promisc;
3719 		name = "promisc";
3720 		break;
3721 	default:
3722 		dev_err(dev, "Got invalid cap request rsp %d\n",
3723 			crq->request_capability.capability);
3724 		return;
3725 	}
3726 
3727 	switch (crq->request_capability_rsp.rc.code) {
3728 	case SUCCESS:
3729 		break;
3730 	case PARTIALSUCCESS:
3731 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3732 			 *req_value,
3733 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3734 					       number), name);
3735 
3736 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3737 		    REQ_MTU) {
3738 			pr_err("mtu of %llu is not supported. Reverting.\n",
3739 			       *req_value);
3740 			*req_value = adapter->fallback.mtu;
3741 		} else {
3742 			*req_value =
3743 				be64_to_cpu(crq->request_capability_rsp.number);
3744 		}
3745 
3746 		ibmvnic_send_req_caps(adapter, 1);
3747 		return;
3748 	default:
3749 		dev_err(dev, "Error %d in request cap rsp\n",
3750 			crq->request_capability_rsp.rc.code);
3751 		return;
3752 	}
3753 
3754 	/* Done receiving requested capabilities, query IP offload support */
3755 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3756 		union ibmvnic_crq newcrq;
3757 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3758 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3759 		    &adapter->ip_offload_buf;
3760 
3761 		adapter->wait_capability = false;
3762 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3763 							 buf_sz,
3764 							 DMA_FROM_DEVICE);
3765 
3766 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3767 			if (!firmware_has_feature(FW_FEATURE_CMO))
3768 				dev_err(dev, "Couldn't map offload buffer\n");
3769 			return;
3770 		}
3771 
3772 		memset(&newcrq, 0, sizeof(newcrq));
3773 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3774 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3775 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3776 		newcrq.query_ip_offload.ioba =
3777 		    cpu_to_be32(adapter->ip_offload_tok);
3778 
3779 		ibmvnic_send_crq(adapter, &newcrq);
3780 	}
3781 }
3782 
3783 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3784 			    struct ibmvnic_adapter *adapter)
3785 {
3786 	struct device *dev = &adapter->vdev->dev;
3787 	struct net_device *netdev = adapter->netdev;
3788 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3789 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3790 	int i;
3791 
3792 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3793 			 DMA_BIDIRECTIONAL);
3794 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3795 			 adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3796 
3797 	/* If the number of queues requested can't be allocated by the
3798 	 * server, the login response will return with code 1. We will need
3799 	 * to resend the login buffer with fewer queues requested.
3800 	 */
3801 	if (login_rsp_crq->generic.rc.code) {
3802 		adapter->renegotiate = true;
3803 		complete(&adapter->init_done);
3804 		return 0;
3805 	}
3806 
3807 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
3808 
3809 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3810 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3811 		netdev_dbg(adapter->netdev, "%016lx\n",
3812 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3813 	}
3814 
3815 	/* Sanity checks */
3816 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3817 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
3818 	     adapter->req_rx_add_queues !=
3819 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3820 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3821 		ibmvnic_remove(adapter->vdev);
3822 		return -EIO;
3823 	}
3824 	release_login_buffer(adapter);
3825 	complete(&adapter->init_done);
3826 
3827 	return 0;
3828 }
3829 
3830 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3831 				     struct ibmvnic_adapter *adapter)
3832 {
3833 	struct device *dev = &adapter->vdev->dev;
3834 	long rc;
3835 
3836 	rc = crq->request_unmap_rsp.rc.code;
3837 	if (rc)
3838 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3839 }
3840 
3841 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3842 				 struct ibmvnic_adapter *adapter)
3843 {
3844 	struct net_device *netdev = adapter->netdev;
3845 	struct device *dev = &adapter->vdev->dev;
3846 	long rc;
3847 
3848 	rc = crq->query_map_rsp.rc.code;
3849 	if (rc) {
3850 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3851 		return;
3852 	}
3853 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3854 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3855 		   crq->query_map_rsp.free_pages);
3856 }
3857 
3858 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3859 				 struct ibmvnic_adapter *adapter)
3860 {
3861 	struct net_device *netdev = adapter->netdev;
3862 	struct device *dev = &adapter->vdev->dev;
3863 	long rc;
3864 
3865 	atomic_dec(&adapter->running_cap_crqs);
3866 	netdev_dbg(netdev, "Outstanding queries: %d\n",
3867 		   atomic_read(&adapter->running_cap_crqs));
3868 	rc = crq->query_capability.rc.code;
3869 	if (rc) {
3870 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3871 		goto out;
3872 	}
3873 
3874 	switch (be16_to_cpu(crq->query_capability.capability)) {
3875 	case MIN_TX_QUEUES:
3876 		adapter->min_tx_queues =
3877 		    be64_to_cpu(crq->query_capability.number);
3878 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
3879 			   adapter->min_tx_queues);
3880 		break;
3881 	case MIN_RX_QUEUES:
3882 		adapter->min_rx_queues =
3883 		    be64_to_cpu(crq->query_capability.number);
3884 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
3885 			   adapter->min_rx_queues);
3886 		break;
3887 	case MIN_RX_ADD_QUEUES:
3888 		adapter->min_rx_add_queues =
3889 		    be64_to_cpu(crq->query_capability.number);
3890 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3891 			   adapter->min_rx_add_queues);
3892 		break;
3893 	case MAX_TX_QUEUES:
3894 		adapter->max_tx_queues =
3895 		    be64_to_cpu(crq->query_capability.number);
3896 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
3897 			   adapter->max_tx_queues);
3898 		break;
3899 	case MAX_RX_QUEUES:
3900 		adapter->max_rx_queues =
3901 		    be64_to_cpu(crq->query_capability.number);
3902 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
3903 			   adapter->max_rx_queues);
3904 		break;
3905 	case MAX_RX_ADD_QUEUES:
3906 		adapter->max_rx_add_queues =
3907 		    be64_to_cpu(crq->query_capability.number);
3908 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3909 			   adapter->max_rx_add_queues);
3910 		break;
3911 	case MIN_TX_ENTRIES_PER_SUBCRQ:
3912 		adapter->min_tx_entries_per_subcrq =
3913 		    be64_to_cpu(crq->query_capability.number);
3914 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3915 			   adapter->min_tx_entries_per_subcrq);
3916 		break;
3917 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3918 		adapter->min_rx_add_entries_per_subcrq =
3919 		    be64_to_cpu(crq->query_capability.number);
3920 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3921 			   adapter->min_rx_add_entries_per_subcrq);
3922 		break;
3923 	case MAX_TX_ENTRIES_PER_SUBCRQ:
3924 		adapter->max_tx_entries_per_subcrq =
3925 		    be64_to_cpu(crq->query_capability.number);
3926 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3927 			   adapter->max_tx_entries_per_subcrq);
3928 		break;
3929 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3930 		adapter->max_rx_add_entries_per_subcrq =
3931 		    be64_to_cpu(crq->query_capability.number);
3932 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3933 			   adapter->max_rx_add_entries_per_subcrq);
3934 		break;
3935 	case TCP_IP_OFFLOAD:
3936 		adapter->tcp_ip_offload =
3937 		    be64_to_cpu(crq->query_capability.number);
3938 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3939 			   adapter->tcp_ip_offload);
3940 		break;
3941 	case PROMISC_SUPPORTED:
3942 		adapter->promisc_supported =
3943 		    be64_to_cpu(crq->query_capability.number);
3944 		netdev_dbg(netdev, "promisc_supported = %lld\n",
3945 			   adapter->promisc_supported);
3946 		break;
3947 	case MIN_MTU:
3948 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3949 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3950 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3951 		break;
3952 	case MAX_MTU:
3953 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3954 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3955 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3956 		break;
3957 	case MAX_MULTICAST_FILTERS:
3958 		adapter->max_multicast_filters =
3959 		    be64_to_cpu(crq->query_capability.number);
3960 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3961 			   adapter->max_multicast_filters);
3962 		break;
3963 	case VLAN_HEADER_INSERTION:
3964 		adapter->vlan_header_insertion =
3965 		    be64_to_cpu(crq->query_capability.number);
3966 		if (adapter->vlan_header_insertion)
3967 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3968 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3969 			   adapter->vlan_header_insertion);
3970 		break;
3971 	case RX_VLAN_HEADER_INSERTION:
3972 		adapter->rx_vlan_header_insertion =
3973 		    be64_to_cpu(crq->query_capability.number);
3974 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3975 			   adapter->rx_vlan_header_insertion);
3976 		break;
3977 	case MAX_TX_SG_ENTRIES:
3978 		adapter->max_tx_sg_entries =
3979 		    be64_to_cpu(crq->query_capability.number);
3980 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3981 			   adapter->max_tx_sg_entries);
3982 		break;
3983 	case RX_SG_SUPPORTED:
3984 		adapter->rx_sg_supported =
3985 		    be64_to_cpu(crq->query_capability.number);
3986 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3987 			   adapter->rx_sg_supported);
3988 		break;
3989 	case OPT_TX_COMP_SUB_QUEUES:
3990 		adapter->opt_tx_comp_sub_queues =
3991 		    be64_to_cpu(crq->query_capability.number);
3992 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3993 			   adapter->opt_tx_comp_sub_queues);
3994 		break;
3995 	case OPT_RX_COMP_QUEUES:
3996 		adapter->opt_rx_comp_queues =
3997 		    be64_to_cpu(crq->query_capability.number);
3998 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3999 			   adapter->opt_rx_comp_queues);
4000 		break;
4001 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4002 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4003 		    be64_to_cpu(crq->query_capability.number);
4004 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4005 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4006 		break;
4007 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4008 		adapter->opt_tx_entries_per_subcrq =
4009 		    be64_to_cpu(crq->query_capability.number);
4010 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4011 			   adapter->opt_tx_entries_per_subcrq);
4012 		break;
4013 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4014 		adapter->opt_rxba_entries_per_subcrq =
4015 		    be64_to_cpu(crq->query_capability.number);
4016 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4017 			   adapter->opt_rxba_entries_per_subcrq);
4018 		break;
4019 	case TX_RX_DESC_REQ:
4020 		adapter->tx_rx_desc_req = crq->query_capability.number;
4021 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4022 			   adapter->tx_rx_desc_req);
4023 		break;
4024 
4025 	default:
4026 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4027 			   crq->query_capability.capability);
4028 	}
4029 
4030 out:
4031 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4032 		adapter->wait_capability = false;
4033 		ibmvnic_send_req_caps(adapter, 0);
4034 	}
4035 }
4036 
4037 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4038 			       struct ibmvnic_adapter *adapter)
4039 {
4040 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4041 	struct net_device *netdev = adapter->netdev;
4042 	struct device *dev = &adapter->vdev->dev;
4043 	u64 *u64_crq = (u64 *)crq;
4044 	long rc;
4045 
4046 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4047 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4048 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4049 	switch (gen_crq->first) {
4050 	case IBMVNIC_CRQ_INIT_RSP:
4051 		switch (gen_crq->cmd) {
4052 		case IBMVNIC_CRQ_INIT:
4053 			dev_info(dev, "Partner initialized\n");
4054 			adapter->from_passive_init = true;
4055 			complete(&adapter->init_done);
4056 			break;
4057 		case IBMVNIC_CRQ_INIT_COMPLETE:
4058 			dev_info(dev, "Partner initialization complete\n");
4059 			send_version_xchg(adapter);
4060 			break;
4061 		default:
4062 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4063 		}
4064 		return;
4065 	case IBMVNIC_CRQ_XPORT_EVENT:
4066 		netif_carrier_off(netdev);
4067 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4068 			dev_info(dev, "Migrated, re-enabling adapter\n");
4069 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4070 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4071 			dev_info(dev, "Backing device failover detected\n");
4072 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4073 		} else {
4074 			/* The adapter lost the connection */
4075 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4076 				gen_crq->cmd);
4077 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4078 		}
4079 		return;
4080 	case IBMVNIC_CRQ_CMD_RSP:
4081 		break;
4082 	default:
4083 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4084 			gen_crq->first);
4085 		return;
4086 	}
4087 
4088 	switch (gen_crq->cmd) {
4089 	case VERSION_EXCHANGE_RSP:
4090 		rc = crq->version_exchange_rsp.rc.code;
4091 		if (rc) {
4092 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4093 			break;
4094 		}
4095 		dev_info(dev, "Partner protocol version is %d\n",
4096 			 crq->version_exchange_rsp.version);
4097 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4098 		    ibmvnic_version)
4099 			ibmvnic_version =
4100 			    be16_to_cpu(crq->version_exchange_rsp.version);
4101 		send_cap_queries(adapter);
4102 		break;
4103 	case QUERY_CAPABILITY_RSP:
4104 		handle_query_cap_rsp(crq, adapter);
4105 		break;
4106 	case QUERY_MAP_RSP:
4107 		handle_query_map_rsp(crq, adapter);
4108 		break;
4109 	case REQUEST_MAP_RSP:
4110 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4111 		complete(&adapter->fw_done);
4112 		break;
4113 	case REQUEST_UNMAP_RSP:
4114 		handle_request_unmap_rsp(crq, adapter);
4115 		break;
4116 	case REQUEST_CAPABILITY_RSP:
4117 		handle_request_cap_rsp(crq, adapter);
4118 		break;
4119 	case LOGIN_RSP:
4120 		netdev_dbg(netdev, "Got Login Response\n");
4121 		handle_login_rsp(crq, adapter);
4122 		break;
4123 	case LOGICAL_LINK_STATE_RSP:
4124 		netdev_dbg(netdev,
4125 			   "Got Logical Link State Response, state: %d rc: %d\n",
4126 			   crq->logical_link_state_rsp.link_state,
4127 			   crq->logical_link_state_rsp.rc.code);
4128 		adapter->logical_link_state =
4129 		    crq->logical_link_state_rsp.link_state;
4130 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4131 		complete(&adapter->init_done);
4132 		break;
4133 	case LINK_STATE_INDICATION:
4134 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4135 		adapter->phys_link_state =
4136 		    crq->link_state_indication.phys_link_state;
4137 		adapter->logical_link_state =
4138 		    crq->link_state_indication.logical_link_state;
4139 		break;
4140 	case CHANGE_MAC_ADDR_RSP:
4141 		netdev_dbg(netdev, "Got MAC address change Response\n");
4142 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4143 		break;
4144 	case ERROR_INDICATION:
4145 		netdev_dbg(netdev, "Got Error Indication\n");
4146 		handle_error_indication(crq, adapter);
4147 		break;
4148 	case REQUEST_ERROR_RSP:
4149 		netdev_dbg(netdev, "Got Error Detail Response\n");
4150 		handle_error_info_rsp(crq, adapter);
4151 		break;
4152 	case REQUEST_STATISTICS_RSP:
4153 		netdev_dbg(netdev, "Got Statistics Response\n");
4154 		complete(&adapter->stats_done);
4155 		break;
4156 	case QUERY_IP_OFFLOAD_RSP:
4157 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4158 		handle_query_ip_offload_rsp(adapter);
4159 		break;
4160 	case MULTICAST_CTRL_RSP:
4161 		netdev_dbg(netdev, "Got multicast control Response\n");
4162 		break;
4163 	case CONTROL_IP_OFFLOAD_RSP:
4164 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4165 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4166 				 sizeof(adapter->ip_offload_ctrl),
4167 				 DMA_TO_DEVICE);
4168 		complete(&adapter->init_done);
4169 		break;
4170 	case COLLECT_FW_TRACE_RSP:
4171 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4172 		complete(&adapter->fw_done);
4173 		break;
4174 	case GET_VPD_SIZE_RSP:
4175 		handle_vpd_size_rsp(crq, adapter);
4176 		break;
4177 	case GET_VPD_RSP:
4178 		handle_vpd_rsp(crq, adapter);
4179 		break;
4180 	default:
4181 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4182 			   gen_crq->cmd);
4183 	}
4184 }
4185 
4186 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4187 {
4188 	struct ibmvnic_adapter *adapter = instance;
4189 
4190 	tasklet_schedule(&adapter->tasklet);
4191 	return IRQ_HANDLED;
4192 }
4193 
4194 static void ibmvnic_tasklet(void *data)
4195 {
4196 	struct ibmvnic_adapter *adapter = data;
4197 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4198 	union ibmvnic_crq *crq;
4199 	unsigned long flags;
4200 	bool done = false;
4201 
4202 	spin_lock_irqsave(&queue->lock, flags);
4203 	while (!done) {
4204 		/* Pull all the valid messages off the CRQ */
4205 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4206 			ibmvnic_handle_crq(crq, adapter);
4207 			crq->generic.first = 0;
4208 		}
4209 
4210 		/* remain in tasklet until all
4211 		 * capabilities responses are received
4212 		 */
4213 		if (!adapter->wait_capability)
4214 			done = true;
4215 	}
4216 	/* if capabilities CRQ's were sent in this tasklet, the following
4217 	 * tasklet must wait until all responses are received
4218 	 */
4219 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4220 		adapter->wait_capability = true;
4221 	spin_unlock_irqrestore(&queue->lock, flags);
4222 }
4223 
4224 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4225 {
4226 	struct vio_dev *vdev = adapter->vdev;
4227 	int rc;
4228 
4229 	do {
4230 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4231 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4232 
4233 	if (rc)
4234 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4235 
4236 	return rc;
4237 }
4238 
4239 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4240 {
4241 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4242 	struct device *dev = &adapter->vdev->dev;
4243 	struct vio_dev *vdev = adapter->vdev;
4244 	int rc;
4245 
4246 	/* Close the CRQ */
4247 	do {
4248 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4249 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4250 
4251 	/* Clean out the queue */
4252 	memset(crq->msgs, 0, PAGE_SIZE);
4253 	crq->cur = 0;
4254 
4255 	/* And re-open it again */
4256 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4257 				crq->msg_token, PAGE_SIZE);
4258 
4259 	if (rc == H_CLOSED)
4260 		/* Adapter is good, but other end is not ready */
4261 		dev_warn(dev, "Partner adapter not ready\n");
4262 	else if (rc != 0)
4263 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4264 
4265 	return rc;
4266 }
4267 
4268 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4269 {
4270 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4271 	struct vio_dev *vdev = adapter->vdev;
4272 	long rc;
4273 
4274 	if (!crq->msgs)
4275 		return;
4276 
4277 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4278 	free_irq(vdev->irq, adapter);
4279 	tasklet_kill(&adapter->tasklet);
4280 	do {
4281 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4282 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4283 
4284 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4285 			 DMA_BIDIRECTIONAL);
4286 	free_page((unsigned long)crq->msgs);
4287 	crq->msgs = NULL;
4288 }
4289 
4290 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4291 {
4292 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4293 	struct device *dev = &adapter->vdev->dev;
4294 	struct vio_dev *vdev = adapter->vdev;
4295 	int rc, retrc = -ENOMEM;
4296 
4297 	if (crq->msgs)
4298 		return 0;
4299 
4300 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4301 	/* Should we allocate more than one page? */
4302 
4303 	if (!crq->msgs)
4304 		return -ENOMEM;
4305 
4306 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4307 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4308 					DMA_BIDIRECTIONAL);
4309 	if (dma_mapping_error(dev, crq->msg_token))
4310 		goto map_failed;
4311 
4312 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4313 				crq->msg_token, PAGE_SIZE);
4314 
4315 	if (rc == H_RESOURCE)
4316 		/* maybe kexecing and resource is busy. try a reset */
4317 		rc = ibmvnic_reset_crq(adapter);
4318 	retrc = rc;
4319 
4320 	if (rc == H_CLOSED) {
4321 		dev_warn(dev, "Partner adapter not ready\n");
4322 	} else if (rc) {
4323 		dev_warn(dev, "Error %d opening adapter\n", rc);
4324 		goto reg_crq_failed;
4325 	}
4326 
4327 	retrc = 0;
4328 
4329 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4330 		     (unsigned long)adapter);
4331 
4332 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4333 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4334 			 adapter);
4335 	if (rc) {
4336 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4337 			vdev->irq, rc);
4338 		goto req_irq_failed;
4339 	}
4340 
4341 	rc = vio_enable_interrupts(vdev);
4342 	if (rc) {
4343 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4344 		goto req_irq_failed;
4345 	}
4346 
4347 	crq->cur = 0;
4348 	spin_lock_init(&crq->lock);
4349 
4350 	return retrc;
4351 
4352 req_irq_failed:
4353 	tasklet_kill(&adapter->tasklet);
4354 	do {
4355 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4356 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4357 reg_crq_failed:
4358 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4359 map_failed:
4360 	free_page((unsigned long)crq->msgs);
4361 	crq->msgs = NULL;
4362 	return retrc;
4363 }
4364 
4365 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4366 {
4367 	struct device *dev = &adapter->vdev->dev;
4368 	unsigned long timeout = msecs_to_jiffies(30000);
4369 	u64 old_num_rx_queues, old_num_tx_queues;
4370 	int rc;
4371 
4372 	if (adapter->resetting && !adapter->wait_for_reset) {
4373 		rc = ibmvnic_reset_crq(adapter);
4374 		if (!rc)
4375 			rc = vio_enable_interrupts(adapter->vdev);
4376 	} else {
4377 		rc = init_crq_queue(adapter);
4378 	}
4379 
4380 	if (rc) {
4381 		dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
4382 		return rc;
4383 	}
4384 
4385 	adapter->from_passive_init = false;
4386 
4387 	old_num_rx_queues = adapter->req_rx_queues;
4388 	old_num_tx_queues = adapter->req_tx_queues;
4389 
4390 	init_completion(&adapter->init_done);
4391 	adapter->init_done_rc = 0;
4392 	ibmvnic_send_crq_init(adapter);
4393 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4394 		dev_err(dev, "Initialization sequence timed out\n");
4395 		return -1;
4396 	}
4397 
4398 	if (adapter->init_done_rc) {
4399 		release_crq_queue(adapter);
4400 		return adapter->init_done_rc;
4401 	}
4402 
4403 	if (adapter->from_passive_init) {
4404 		adapter->state = VNIC_OPEN;
4405 		adapter->from_passive_init = false;
4406 		return -1;
4407 	}
4408 
4409 	if (adapter->resetting && !adapter->wait_for_reset) {
4410 		if (adapter->req_rx_queues != old_num_rx_queues ||
4411 		    adapter->req_tx_queues != old_num_tx_queues) {
4412 			release_sub_crqs(adapter, 0);
4413 			rc = init_sub_crqs(adapter);
4414 		} else {
4415 			rc = reset_sub_crq_queues(adapter);
4416 		}
4417 	} else {
4418 		rc = init_sub_crqs(adapter);
4419 	}
4420 
4421 	if (rc) {
4422 		dev_err(dev, "Initialization of sub crqs failed\n");
4423 		release_crq_queue(adapter);
4424 		return rc;
4425 	}
4426 
4427 	rc = init_sub_crq_irqs(adapter);
4428 	if (rc) {
4429 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4430 		release_crq_queue(adapter);
4431 	}
4432 
4433 	return rc;
4434 }
4435 
4436 static struct device_attribute dev_attr_failover;
4437 
4438 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4439 {
4440 	struct ibmvnic_adapter *adapter;
4441 	struct net_device *netdev;
4442 	unsigned char *mac_addr_p;
4443 	int rc;
4444 
4445 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4446 		dev->unit_address);
4447 
4448 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4449 							VETH_MAC_ADDR, NULL);
4450 	if (!mac_addr_p) {
4451 		dev_err(&dev->dev,
4452 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4453 			__FILE__, __LINE__);
4454 		return 0;
4455 	}
4456 
4457 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4458 				   IBMVNIC_MAX_QUEUES);
4459 	if (!netdev)
4460 		return -ENOMEM;
4461 
4462 	adapter = netdev_priv(netdev);
4463 	adapter->state = VNIC_PROBING;
4464 	dev_set_drvdata(&dev->dev, netdev);
4465 	adapter->vdev = dev;
4466 	adapter->netdev = netdev;
4467 
4468 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
4469 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4470 	netdev->irq = dev->irq;
4471 	netdev->netdev_ops = &ibmvnic_netdev_ops;
4472 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4473 	SET_NETDEV_DEV(netdev, &dev->dev);
4474 
4475 	spin_lock_init(&adapter->stats_lock);
4476 
4477 	INIT_LIST_HEAD(&adapter->errors);
4478 	spin_lock_init(&adapter->error_list_lock);
4479 
4480 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4481 	INIT_LIST_HEAD(&adapter->rwi_list);
4482 	mutex_init(&adapter->reset_lock);
4483 	mutex_init(&adapter->rwi_lock);
4484 	adapter->resetting = false;
4485 
4486 	adapter->mac_change_pending = false;
4487 
4488 	do {
4489 		rc = ibmvnic_init(adapter);
4490 		if (rc && rc != EAGAIN)
4491 			goto ibmvnic_init_fail;
4492 	} while (rc == EAGAIN);
4493 
4494 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4495 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4496 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4497 
4498 	rc = device_create_file(&dev->dev, &dev_attr_failover);
4499 	if (rc)
4500 		goto ibmvnic_init_fail;
4501 
4502 	netif_carrier_off(netdev);
4503 	rc = register_netdev(netdev);
4504 	if (rc) {
4505 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4506 		goto ibmvnic_register_fail;
4507 	}
4508 	dev_info(&dev->dev, "ibmvnic registered\n");
4509 
4510 	adapter->state = VNIC_PROBED;
4511 
4512 	adapter->wait_for_reset = false;
4513 
4514 	return 0;
4515 
4516 ibmvnic_register_fail:
4517 	device_remove_file(&dev->dev, &dev_attr_failover);
4518 
4519 ibmvnic_init_fail:
4520 	release_sub_crqs(adapter, 1);
4521 	release_crq_queue(adapter);
4522 	free_netdev(netdev);
4523 
4524 	return rc;
4525 }
4526 
4527 static int ibmvnic_remove(struct vio_dev *dev)
4528 {
4529 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
4530 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4531 
4532 	adapter->state = VNIC_REMOVING;
4533 	unregister_netdev(netdev);
4534 	mutex_lock(&adapter->reset_lock);
4535 
4536 	release_resources(adapter);
4537 	release_sub_crqs(adapter, 1);
4538 	release_crq_queue(adapter);
4539 
4540 	adapter->state = VNIC_REMOVED;
4541 
4542 	mutex_unlock(&adapter->reset_lock);
4543 	device_remove_file(&dev->dev, &dev_attr_failover);
4544 	free_netdev(netdev);
4545 	dev_set_drvdata(&dev->dev, NULL);
4546 
4547 	return 0;
4548 }
4549 
4550 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4551 			      const char *buf, size_t count)
4552 {
4553 	struct net_device *netdev = dev_get_drvdata(dev);
4554 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4555 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4556 	__be64 session_token;
4557 	long rc;
4558 
4559 	if (!sysfs_streq(buf, "1"))
4560 		return -EINVAL;
4561 
4562 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4563 			 H_GET_SESSION_TOKEN, 0, 0, 0);
4564 	if (rc) {
4565 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4566 			   rc);
4567 		return -EINVAL;
4568 	}
4569 
4570 	session_token = (__be64)retbuf[0];
4571 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4572 		   be64_to_cpu(session_token));
4573 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4574 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
4575 	if (rc) {
4576 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4577 			   rc);
4578 		return -EINVAL;
4579 	}
4580 
4581 	return count;
4582 }
4583 
4584 static DEVICE_ATTR_WO(failover);
4585 
4586 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4587 {
4588 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4589 	struct ibmvnic_adapter *adapter;
4590 	struct iommu_table *tbl;
4591 	unsigned long ret = 0;
4592 	int i;
4593 
4594 	tbl = get_iommu_table_base(&vdev->dev);
4595 
4596 	/* netdev inits at probe time along with the structures we need below*/
4597 	if (!netdev)
4598 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4599 
4600 	adapter = netdev_priv(netdev);
4601 
4602 	ret += PAGE_SIZE; /* the crq message queue */
4603 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4604 
4605 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4606 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4607 
4608 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4609 	     i++)
4610 		ret += adapter->rx_pool[i].size *
4611 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4612 
4613 	return ret;
4614 }
4615 
4616 static int ibmvnic_resume(struct device *dev)
4617 {
4618 	struct net_device *netdev = dev_get_drvdata(dev);
4619 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4620 
4621 	if (adapter->state != VNIC_OPEN)
4622 		return 0;
4623 
4624 	tasklet_schedule(&adapter->tasklet);
4625 
4626 	return 0;
4627 }
4628 
4629 static const struct vio_device_id ibmvnic_device_table[] = {
4630 	{"network", "IBM,vnic"},
4631 	{"", "" }
4632 };
4633 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4634 
4635 static const struct dev_pm_ops ibmvnic_pm_ops = {
4636 	.resume = ibmvnic_resume
4637 };
4638 
4639 static struct vio_driver ibmvnic_driver = {
4640 	.id_table       = ibmvnic_device_table,
4641 	.probe          = ibmvnic_probe,
4642 	.remove         = ibmvnic_remove,
4643 	.get_desired_dma = ibmvnic_get_desired_dma,
4644 	.name		= ibmvnic_driver_name,
4645 	.pm		= &ibmvnic_pm_ops,
4646 };
4647 
4648 /* module functions */
4649 static int __init ibmvnic_module_init(void)
4650 {
4651 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4652 		IBMVNIC_DRIVER_VERSION);
4653 
4654 	return vio_register_driver(&ibmvnic_driver);
4655 }
4656 
4657 static void __exit ibmvnic_module_exit(void)
4658 {
4659 	vio_unregister_driver(&ibmvnic_driver);
4660 }
4661 
4662 module_init(ibmvnic_module_init);
4663 module_exit(ibmvnic_module_exit);
4664