xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision c32e64e852f3f5c0fd709f84bc94736840088375)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/if_arp.h>
63 #include <linux/in.h>
64 #include <linux/ip.h>
65 #include <linux/ipv6.h>
66 #include <linux/irq.h>
67 #include <linux/kthread.h>
68 #include <linux/seq_file.h>
69 #include <linux/interrupt.h>
70 #include <net/net_namespace.h>
71 #include <asm/hvcall.h>
72 #include <linux/atomic.h>
73 #include <asm/vio.h>
74 #include <asm/iommu.h>
75 #include <linux/uaccess.h>
76 #include <asm/firmware.h>
77 #include <linux/workqueue.h>
78 #include <linux/if_vlan.h>
79 #include <linux/utsname.h>
80 
81 #include "ibmvnic.h"
82 
83 static const char ibmvnic_driver_name[] = "ibmvnic";
84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
85 
86 MODULE_AUTHOR("Santiago Leon");
87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
88 MODULE_LICENSE("GPL");
89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
90 
91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
92 static int ibmvnic_remove(struct vio_dev *);
93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
99 		       union sub_crq *sub_crq);
100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
102 static int enable_scrq_irq(struct ibmvnic_adapter *,
103 			   struct ibmvnic_sub_crq_queue *);
104 static int disable_scrq_irq(struct ibmvnic_adapter *,
105 			    struct ibmvnic_sub_crq_queue *);
106 static int pending_scrq(struct ibmvnic_adapter *,
107 			struct ibmvnic_sub_crq_queue *);
108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
109 					struct ibmvnic_sub_crq_queue *);
110 static int ibmvnic_poll(struct napi_struct *napi, int data);
111 static void send_map_query(struct ibmvnic_adapter *adapter);
112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
114 static int send_login(struct ibmvnic_adapter *adapter);
115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
116 static int init_sub_crqs(struct ibmvnic_adapter *);
117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
118 static int ibmvnic_init(struct ibmvnic_adapter *);
119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
120 static void release_crq_queue(struct ibmvnic_adapter *);
121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
123 
124 struct ibmvnic_stat {
125 	char name[ETH_GSTRING_LEN];
126 	int offset;
127 };
128 
129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
130 			     offsetof(struct ibmvnic_statistics, stat))
131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
132 
133 static const struct ibmvnic_stat ibmvnic_stats[] = {
134 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
135 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
136 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
137 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
138 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
139 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
140 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
141 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
142 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
143 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
144 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
145 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
146 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
147 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
148 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
149 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
150 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
151 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
152 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
153 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
154 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
155 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
156 };
157 
158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
159 			  unsigned long length, unsigned long *number,
160 			  unsigned long *irq)
161 {
162 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
163 	long rc;
164 
165 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
166 	*number = retbuf[0];
167 	*irq = retbuf[1];
168 
169 	return rc;
170 }
171 
172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
173 				struct ibmvnic_long_term_buff *ltb, int size)
174 {
175 	struct device *dev = &adapter->vdev->dev;
176 	int rc;
177 
178 	ltb->size = size;
179 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
180 				       GFP_KERNEL);
181 
182 	if (!ltb->buff) {
183 		dev_err(dev, "Couldn't alloc long term buffer\n");
184 		return -ENOMEM;
185 	}
186 	ltb->map_id = adapter->map_id;
187 	adapter->map_id++;
188 
189 	init_completion(&adapter->fw_done);
190 	rc = send_request_map(adapter, ltb->addr,
191 			      ltb->size, ltb->map_id);
192 	if (rc) {
193 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
194 		return rc;
195 	}
196 	wait_for_completion(&adapter->fw_done);
197 
198 	if (adapter->fw_done_rc) {
199 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
200 			adapter->fw_done_rc);
201 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
202 		return -1;
203 	}
204 	return 0;
205 }
206 
207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
208 				struct ibmvnic_long_term_buff *ltb)
209 {
210 	struct device *dev = &adapter->vdev->dev;
211 
212 	if (!ltb->buff)
213 		return;
214 
215 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
216 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
217 		send_request_unmap(adapter, ltb->map_id);
218 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
219 }
220 
221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
222 				struct ibmvnic_long_term_buff *ltb)
223 {
224 	int rc;
225 
226 	memset(ltb->buff, 0, ltb->size);
227 
228 	init_completion(&adapter->fw_done);
229 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
230 	if (rc)
231 		return rc;
232 	wait_for_completion(&adapter->fw_done);
233 
234 	if (adapter->fw_done_rc) {
235 		dev_info(&adapter->vdev->dev,
236 			 "Reset failed, attempting to free and reallocate buffer\n");
237 		free_long_term_buff(adapter, ltb);
238 		return alloc_long_term_buff(adapter, ltb, ltb->size);
239 	}
240 	return 0;
241 }
242 
243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
244 {
245 	int i;
246 
247 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
248 	     i++)
249 		adapter->rx_pool[i].active = 0;
250 }
251 
252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
253 			      struct ibmvnic_rx_pool *pool)
254 {
255 	int count = pool->size - atomic_read(&pool->available);
256 	struct device *dev = &adapter->vdev->dev;
257 	int buffers_added = 0;
258 	unsigned long lpar_rc;
259 	union sub_crq sub_crq;
260 	struct sk_buff *skb;
261 	unsigned int offset;
262 	dma_addr_t dma_addr;
263 	unsigned char *dst;
264 	u64 *handle_array;
265 	int shift = 0;
266 	int index;
267 	int i;
268 
269 	if (!pool->active)
270 		return;
271 
272 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
273 				      be32_to_cpu(adapter->login_rsp_buf->
274 				      off_rxadd_subcrqs));
275 
276 	for (i = 0; i < count; ++i) {
277 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
278 		if (!skb) {
279 			dev_err(dev, "Couldn't replenish rx buff\n");
280 			adapter->replenish_no_mem++;
281 			break;
282 		}
283 
284 		index = pool->free_map[pool->next_free];
285 
286 		if (pool->rx_buff[index].skb)
287 			dev_err(dev, "Inconsistent free_map!\n");
288 
289 		/* Copy the skb to the long term mapped DMA buffer */
290 		offset = index * pool->buff_size;
291 		dst = pool->long_term_buff.buff + offset;
292 		memset(dst, 0, pool->buff_size);
293 		dma_addr = pool->long_term_buff.addr + offset;
294 		pool->rx_buff[index].data = dst;
295 
296 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
297 		pool->rx_buff[index].dma = dma_addr;
298 		pool->rx_buff[index].skb = skb;
299 		pool->rx_buff[index].pool_index = pool->index;
300 		pool->rx_buff[index].size = pool->buff_size;
301 
302 		memset(&sub_crq, 0, sizeof(sub_crq));
303 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
304 		sub_crq.rx_add.correlator =
305 		    cpu_to_be64((u64)&pool->rx_buff[index]);
306 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
307 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
308 
309 		/* The length field of the sCRQ is defined to be 24 bits so the
310 		 * buffer size needs to be left shifted by a byte before it is
311 		 * converted to big endian to prevent the last byte from being
312 		 * truncated.
313 		 */
314 #ifdef __LITTLE_ENDIAN__
315 		shift = 8;
316 #endif
317 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
318 
319 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
320 				      &sub_crq);
321 		if (lpar_rc != H_SUCCESS)
322 			goto failure;
323 
324 		buffers_added++;
325 		adapter->replenish_add_buff_success++;
326 		pool->next_free = (pool->next_free + 1) % pool->size;
327 	}
328 	atomic_add(buffers_added, &pool->available);
329 	return;
330 
331 failure:
332 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
333 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
334 	pool->free_map[pool->next_free] = index;
335 	pool->rx_buff[index].skb = NULL;
336 
337 	dev_kfree_skb_any(skb);
338 	adapter->replenish_add_buff_failure++;
339 	atomic_add(buffers_added, &pool->available);
340 
341 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
342 		/* Disable buffer pool replenishment and report carrier off if
343 		 * queue is closed or pending failover.
344 		 * Firmware guarantees that a signal will be sent to the
345 		 * driver, triggering a reset.
346 		 */
347 		deactivate_rx_pools(adapter);
348 		netif_carrier_off(adapter->netdev);
349 	}
350 }
351 
352 static void replenish_pools(struct ibmvnic_adapter *adapter)
353 {
354 	int i;
355 
356 	adapter->replenish_task_cycles++;
357 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
358 	     i++) {
359 		if (adapter->rx_pool[i].active)
360 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
361 	}
362 }
363 
364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
365 {
366 	kfree(adapter->tx_stats_buffers);
367 	kfree(adapter->rx_stats_buffers);
368 	adapter->tx_stats_buffers = NULL;
369 	adapter->rx_stats_buffers = NULL;
370 }
371 
372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
373 {
374 	adapter->tx_stats_buffers =
375 				kcalloc(IBMVNIC_MAX_QUEUES,
376 					sizeof(struct ibmvnic_tx_queue_stats),
377 					GFP_KERNEL);
378 	if (!adapter->tx_stats_buffers)
379 		return -ENOMEM;
380 
381 	adapter->rx_stats_buffers =
382 				kcalloc(IBMVNIC_MAX_QUEUES,
383 					sizeof(struct ibmvnic_rx_queue_stats),
384 					GFP_KERNEL);
385 	if (!adapter->rx_stats_buffers)
386 		return -ENOMEM;
387 
388 	return 0;
389 }
390 
391 static void release_stats_token(struct ibmvnic_adapter *adapter)
392 {
393 	struct device *dev = &adapter->vdev->dev;
394 
395 	if (!adapter->stats_token)
396 		return;
397 
398 	dma_unmap_single(dev, adapter->stats_token,
399 			 sizeof(struct ibmvnic_statistics),
400 			 DMA_FROM_DEVICE);
401 	adapter->stats_token = 0;
402 }
403 
404 static int init_stats_token(struct ibmvnic_adapter *adapter)
405 {
406 	struct device *dev = &adapter->vdev->dev;
407 	dma_addr_t stok;
408 
409 	stok = dma_map_single(dev, &adapter->stats,
410 			      sizeof(struct ibmvnic_statistics),
411 			      DMA_FROM_DEVICE);
412 	if (dma_mapping_error(dev, stok)) {
413 		dev_err(dev, "Couldn't map stats buffer\n");
414 		return -1;
415 	}
416 
417 	adapter->stats_token = stok;
418 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
419 	return 0;
420 }
421 
422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
423 {
424 	struct ibmvnic_rx_pool *rx_pool;
425 	int rx_scrqs;
426 	int i, j, rc;
427 	u64 *size_array;
428 
429 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
430 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
431 
432 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
433 	for (i = 0; i < rx_scrqs; i++) {
434 		rx_pool = &adapter->rx_pool[i];
435 
436 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
437 
438 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
439 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
440 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
441 			alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
442 					     rx_pool->size *
443 					     rx_pool->buff_size);
444 		} else {
445 			rc = reset_long_term_buff(adapter,
446 						  &rx_pool->long_term_buff);
447 		}
448 
449 		if (rc)
450 			return rc;
451 
452 		for (j = 0; j < rx_pool->size; j++)
453 			rx_pool->free_map[j] = j;
454 
455 		memset(rx_pool->rx_buff, 0,
456 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
457 
458 		atomic_set(&rx_pool->available, 0);
459 		rx_pool->next_alloc = 0;
460 		rx_pool->next_free = 0;
461 		rx_pool->active = 1;
462 	}
463 
464 	return 0;
465 }
466 
467 static void release_rx_pools(struct ibmvnic_adapter *adapter)
468 {
469 	struct ibmvnic_rx_pool *rx_pool;
470 	int i, j;
471 
472 	if (!adapter->rx_pool)
473 		return;
474 
475 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
476 		rx_pool = &adapter->rx_pool[i];
477 
478 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
479 
480 		kfree(rx_pool->free_map);
481 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
482 
483 		if (!rx_pool->rx_buff)
484 			continue;
485 
486 		for (j = 0; j < rx_pool->size; j++) {
487 			if (rx_pool->rx_buff[j].skb) {
488 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
489 				rx_pool->rx_buff[i].skb = NULL;
490 			}
491 		}
492 
493 		kfree(rx_pool->rx_buff);
494 	}
495 
496 	kfree(adapter->rx_pool);
497 	adapter->rx_pool = NULL;
498 	adapter->num_active_rx_pools = 0;
499 }
500 
501 static int init_rx_pools(struct net_device *netdev)
502 {
503 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
504 	struct device *dev = &adapter->vdev->dev;
505 	struct ibmvnic_rx_pool *rx_pool;
506 	int rxadd_subcrqs;
507 	u64 *size_array;
508 	int i, j;
509 
510 	rxadd_subcrqs =
511 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
512 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
513 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
514 
515 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
516 				   sizeof(struct ibmvnic_rx_pool),
517 				   GFP_KERNEL);
518 	if (!adapter->rx_pool) {
519 		dev_err(dev, "Failed to allocate rx pools\n");
520 		return -1;
521 	}
522 
523 	adapter->num_active_rx_pools = rxadd_subcrqs;
524 
525 	for (i = 0; i < rxadd_subcrqs; i++) {
526 		rx_pool = &adapter->rx_pool[i];
527 
528 		netdev_dbg(adapter->netdev,
529 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
530 			   i, adapter->req_rx_add_entries_per_subcrq,
531 			   be64_to_cpu(size_array[i]));
532 
533 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
534 		rx_pool->index = i;
535 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
536 		rx_pool->active = 1;
537 
538 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
539 					    GFP_KERNEL);
540 		if (!rx_pool->free_map) {
541 			release_rx_pools(adapter);
542 			return -1;
543 		}
544 
545 		rx_pool->rx_buff = kcalloc(rx_pool->size,
546 					   sizeof(struct ibmvnic_rx_buff),
547 					   GFP_KERNEL);
548 		if (!rx_pool->rx_buff) {
549 			dev_err(dev, "Couldn't alloc rx buffers\n");
550 			release_rx_pools(adapter);
551 			return -1;
552 		}
553 
554 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
555 					 rx_pool->size * rx_pool->buff_size)) {
556 			release_rx_pools(adapter);
557 			return -1;
558 		}
559 
560 		for (j = 0; j < rx_pool->size; ++j)
561 			rx_pool->free_map[j] = j;
562 
563 		atomic_set(&rx_pool->available, 0);
564 		rx_pool->next_alloc = 0;
565 		rx_pool->next_free = 0;
566 	}
567 
568 	return 0;
569 }
570 
571 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
572 			     struct ibmvnic_tx_pool *tx_pool)
573 {
574 	int rc, i;
575 
576 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
577 	if (rc)
578 		return rc;
579 
580 	memset(tx_pool->tx_buff, 0,
581 	       tx_pool->num_buffers *
582 	       sizeof(struct ibmvnic_tx_buff));
583 
584 	for (i = 0; i < tx_pool->num_buffers; i++)
585 		tx_pool->free_map[i] = i;
586 
587 	tx_pool->consumer_index = 0;
588 	tx_pool->producer_index = 0;
589 
590 	return 0;
591 }
592 
593 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
594 {
595 	int tx_scrqs;
596 	int i, rc;
597 
598 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
599 	for (i = 0; i < tx_scrqs; i++) {
600 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
601 		if (rc)
602 			return rc;
603 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
604 		if (rc)
605 			return rc;
606 	}
607 
608 	return 0;
609 }
610 
611 static void release_vpd_data(struct ibmvnic_adapter *adapter)
612 {
613 	if (!adapter->vpd)
614 		return;
615 
616 	kfree(adapter->vpd->buff);
617 	kfree(adapter->vpd);
618 
619 	adapter->vpd = NULL;
620 }
621 
622 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
623 				struct ibmvnic_tx_pool *tx_pool)
624 {
625 	kfree(tx_pool->tx_buff);
626 	kfree(tx_pool->free_map);
627 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
628 }
629 
630 static void release_tx_pools(struct ibmvnic_adapter *adapter)
631 {
632 	int i;
633 
634 	if (!adapter->tx_pool)
635 		return;
636 
637 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
638 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
639 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
640 	}
641 
642 	kfree(adapter->tx_pool);
643 	adapter->tx_pool = NULL;
644 	kfree(adapter->tso_pool);
645 	adapter->tso_pool = NULL;
646 	adapter->num_active_tx_pools = 0;
647 }
648 
649 static int init_one_tx_pool(struct net_device *netdev,
650 			    struct ibmvnic_tx_pool *tx_pool,
651 			    int num_entries, int buf_size)
652 {
653 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
654 	int i;
655 
656 	tx_pool->tx_buff = kcalloc(num_entries,
657 				   sizeof(struct ibmvnic_tx_buff),
658 				   GFP_KERNEL);
659 	if (!tx_pool->tx_buff)
660 		return -1;
661 
662 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
663 				 num_entries * buf_size))
664 		return -1;
665 
666 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
667 	if (!tx_pool->free_map)
668 		return -1;
669 
670 	for (i = 0; i < num_entries; i++)
671 		tx_pool->free_map[i] = i;
672 
673 	tx_pool->consumer_index = 0;
674 	tx_pool->producer_index = 0;
675 	tx_pool->num_buffers = num_entries;
676 	tx_pool->buf_size = buf_size;
677 
678 	return 0;
679 }
680 
681 static int init_tx_pools(struct net_device *netdev)
682 {
683 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
684 	int tx_subcrqs;
685 	int i, rc;
686 
687 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
688 	adapter->tx_pool = kcalloc(tx_subcrqs,
689 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
690 	if (!adapter->tx_pool)
691 		return -1;
692 
693 	adapter->tso_pool = kcalloc(tx_subcrqs,
694 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
695 	if (!adapter->tso_pool)
696 		return -1;
697 
698 	adapter->num_active_tx_pools = tx_subcrqs;
699 
700 	for (i = 0; i < tx_subcrqs; i++) {
701 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
702 				      adapter->req_tx_entries_per_subcrq,
703 				      adapter->req_mtu + VLAN_HLEN);
704 		if (rc) {
705 			release_tx_pools(adapter);
706 			return rc;
707 		}
708 
709 		init_one_tx_pool(netdev, &adapter->tso_pool[i],
710 				 IBMVNIC_TSO_BUFS,
711 				 IBMVNIC_TSO_BUF_SZ);
712 		if (rc) {
713 			release_tx_pools(adapter);
714 			return rc;
715 		}
716 	}
717 
718 	return 0;
719 }
720 
721 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
722 {
723 	int i;
724 
725 	if (adapter->napi_enabled)
726 		return;
727 
728 	for (i = 0; i < adapter->req_rx_queues; i++)
729 		napi_enable(&adapter->napi[i]);
730 
731 	adapter->napi_enabled = true;
732 }
733 
734 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
735 {
736 	int i;
737 
738 	if (!adapter->napi_enabled)
739 		return;
740 
741 	for (i = 0; i < adapter->req_rx_queues; i++) {
742 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
743 		napi_disable(&adapter->napi[i]);
744 	}
745 
746 	adapter->napi_enabled = false;
747 }
748 
749 static int init_napi(struct ibmvnic_adapter *adapter)
750 {
751 	int i;
752 
753 	adapter->napi = kcalloc(adapter->req_rx_queues,
754 				sizeof(struct napi_struct), GFP_KERNEL);
755 	if (!adapter->napi)
756 		return -ENOMEM;
757 
758 	for (i = 0; i < adapter->req_rx_queues; i++) {
759 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
760 		netif_napi_add(adapter->netdev, &adapter->napi[i],
761 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
762 	}
763 
764 	adapter->num_active_rx_napi = adapter->req_rx_queues;
765 	return 0;
766 }
767 
768 static void release_napi(struct ibmvnic_adapter *adapter)
769 {
770 	int i;
771 
772 	if (!adapter->napi)
773 		return;
774 
775 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
776 		if (&adapter->napi[i]) {
777 			netdev_dbg(adapter->netdev,
778 				   "Releasing napi[%d]\n", i);
779 			netif_napi_del(&adapter->napi[i]);
780 		}
781 	}
782 
783 	kfree(adapter->napi);
784 	adapter->napi = NULL;
785 	adapter->num_active_rx_napi = 0;
786 	adapter->napi_enabled = false;
787 }
788 
789 static int ibmvnic_login(struct net_device *netdev)
790 {
791 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
792 	unsigned long timeout = msecs_to_jiffies(30000);
793 	int retry_count = 0;
794 	bool retry;
795 	int rc;
796 
797 	do {
798 		retry = false;
799 		if (retry_count > IBMVNIC_MAX_QUEUES) {
800 			netdev_warn(netdev, "Login attempts exceeded\n");
801 			return -1;
802 		}
803 
804 		adapter->init_done_rc = 0;
805 		reinit_completion(&adapter->init_done);
806 		rc = send_login(adapter);
807 		if (rc) {
808 			netdev_warn(netdev, "Unable to login\n");
809 			return rc;
810 		}
811 
812 		if (!wait_for_completion_timeout(&adapter->init_done,
813 						 timeout)) {
814 			netdev_warn(netdev, "Login timed out\n");
815 			return -1;
816 		}
817 
818 		if (adapter->init_done_rc == PARTIALSUCCESS) {
819 			retry_count++;
820 			release_sub_crqs(adapter, 1);
821 
822 			retry = true;
823 			netdev_dbg(netdev,
824 				   "Received partial success, retrying...\n");
825 			adapter->init_done_rc = 0;
826 			reinit_completion(&adapter->init_done);
827 			send_cap_queries(adapter);
828 			if (!wait_for_completion_timeout(&adapter->init_done,
829 							 timeout)) {
830 				netdev_warn(netdev,
831 					    "Capabilities query timed out\n");
832 				return -1;
833 			}
834 
835 			rc = init_sub_crqs(adapter);
836 			if (rc) {
837 				netdev_warn(netdev,
838 					    "SCRQ initialization failed\n");
839 				return -1;
840 			}
841 
842 			rc = init_sub_crq_irqs(adapter);
843 			if (rc) {
844 				netdev_warn(netdev,
845 					    "SCRQ irq initialization failed\n");
846 				return -1;
847 			}
848 		} else if (adapter->init_done_rc) {
849 			netdev_warn(netdev, "Adapter login failed\n");
850 			return -1;
851 		}
852 	} while (retry);
853 
854 	/* handle pending MAC address changes after successful login */
855 	if (adapter->mac_change_pending) {
856 		__ibmvnic_set_mac(netdev, &adapter->desired.mac);
857 		adapter->mac_change_pending = false;
858 	}
859 
860 	return 0;
861 }
862 
863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
864 {
865 	kfree(adapter->login_buf);
866 	adapter->login_buf = NULL;
867 }
868 
869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
870 {
871 	kfree(adapter->login_rsp_buf);
872 	adapter->login_rsp_buf = NULL;
873 }
874 
875 static void release_resources(struct ibmvnic_adapter *adapter)
876 {
877 	release_vpd_data(adapter);
878 
879 	release_tx_pools(adapter);
880 	release_rx_pools(adapter);
881 
882 	release_napi(adapter);
883 	release_login_rsp_buffer(adapter);
884 }
885 
886 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
887 {
888 	struct net_device *netdev = adapter->netdev;
889 	unsigned long timeout = msecs_to_jiffies(30000);
890 	union ibmvnic_crq crq;
891 	bool resend;
892 	int rc;
893 
894 	netdev_dbg(netdev, "setting link state %d\n", link_state);
895 
896 	memset(&crq, 0, sizeof(crq));
897 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
898 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
899 	crq.logical_link_state.link_state = link_state;
900 
901 	do {
902 		resend = false;
903 
904 		reinit_completion(&adapter->init_done);
905 		rc = ibmvnic_send_crq(adapter, &crq);
906 		if (rc) {
907 			netdev_err(netdev, "Failed to set link state\n");
908 			return rc;
909 		}
910 
911 		if (!wait_for_completion_timeout(&adapter->init_done,
912 						 timeout)) {
913 			netdev_err(netdev, "timeout setting link state\n");
914 			return -1;
915 		}
916 
917 		if (adapter->init_done_rc == 1) {
918 			/* Partuial success, delay and re-send */
919 			mdelay(1000);
920 			resend = true;
921 		} else if (adapter->init_done_rc) {
922 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
923 				    adapter->init_done_rc);
924 			return adapter->init_done_rc;
925 		}
926 	} while (resend);
927 
928 	return 0;
929 }
930 
931 static int set_real_num_queues(struct net_device *netdev)
932 {
933 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
934 	int rc;
935 
936 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
937 		   adapter->req_tx_queues, adapter->req_rx_queues);
938 
939 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
940 	if (rc) {
941 		netdev_err(netdev, "failed to set the number of tx queues\n");
942 		return rc;
943 	}
944 
945 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
946 	if (rc)
947 		netdev_err(netdev, "failed to set the number of rx queues\n");
948 
949 	return rc;
950 }
951 
952 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
953 {
954 	struct device *dev = &adapter->vdev->dev;
955 	union ibmvnic_crq crq;
956 	int len = 0;
957 	int rc;
958 
959 	if (adapter->vpd->buff)
960 		len = adapter->vpd->len;
961 
962 	init_completion(&adapter->fw_done);
963 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
964 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
965 	rc = ibmvnic_send_crq(adapter, &crq);
966 	if (rc)
967 		return rc;
968 	wait_for_completion(&adapter->fw_done);
969 
970 	if (!adapter->vpd->len)
971 		return -ENODATA;
972 
973 	if (!adapter->vpd->buff)
974 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
975 	else if (adapter->vpd->len != len)
976 		adapter->vpd->buff =
977 			krealloc(adapter->vpd->buff,
978 				 adapter->vpd->len, GFP_KERNEL);
979 
980 	if (!adapter->vpd->buff) {
981 		dev_err(dev, "Could allocate VPD buffer\n");
982 		return -ENOMEM;
983 	}
984 
985 	adapter->vpd->dma_addr =
986 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
987 			       DMA_FROM_DEVICE);
988 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
989 		dev_err(dev, "Could not map VPD buffer\n");
990 		kfree(adapter->vpd->buff);
991 		adapter->vpd->buff = NULL;
992 		return -ENOMEM;
993 	}
994 
995 	reinit_completion(&adapter->fw_done);
996 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
997 	crq.get_vpd.cmd = GET_VPD;
998 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
999 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1000 	rc = ibmvnic_send_crq(adapter, &crq);
1001 	if (rc) {
1002 		kfree(adapter->vpd->buff);
1003 		adapter->vpd->buff = NULL;
1004 		return rc;
1005 	}
1006 	wait_for_completion(&adapter->fw_done);
1007 
1008 	return 0;
1009 }
1010 
1011 static int init_resources(struct ibmvnic_adapter *adapter)
1012 {
1013 	struct net_device *netdev = adapter->netdev;
1014 	int rc;
1015 
1016 	rc = set_real_num_queues(netdev);
1017 	if (rc)
1018 		return rc;
1019 
1020 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1021 	if (!adapter->vpd)
1022 		return -ENOMEM;
1023 
1024 	/* Vital Product Data (VPD) */
1025 	rc = ibmvnic_get_vpd(adapter);
1026 	if (rc) {
1027 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1028 		return rc;
1029 	}
1030 
1031 	adapter->map_id = 1;
1032 
1033 	rc = init_napi(adapter);
1034 	if (rc)
1035 		return rc;
1036 
1037 	send_map_query(adapter);
1038 
1039 	rc = init_rx_pools(netdev);
1040 	if (rc)
1041 		return rc;
1042 
1043 	rc = init_tx_pools(netdev);
1044 	return rc;
1045 }
1046 
1047 static int __ibmvnic_open(struct net_device *netdev)
1048 {
1049 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1050 	enum vnic_state prev_state = adapter->state;
1051 	int i, rc;
1052 
1053 	adapter->state = VNIC_OPENING;
1054 	replenish_pools(adapter);
1055 	ibmvnic_napi_enable(adapter);
1056 
1057 	/* We're ready to receive frames, enable the sub-crq interrupts and
1058 	 * set the logical link state to up
1059 	 */
1060 	for (i = 0; i < adapter->req_rx_queues; i++) {
1061 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1062 		if (prev_state == VNIC_CLOSED)
1063 			enable_irq(adapter->rx_scrq[i]->irq);
1064 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1065 	}
1066 
1067 	for (i = 0; i < adapter->req_tx_queues; i++) {
1068 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1069 		if (prev_state == VNIC_CLOSED)
1070 			enable_irq(adapter->tx_scrq[i]->irq);
1071 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1072 	}
1073 
1074 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1075 	if (rc) {
1076 		for (i = 0; i < adapter->req_rx_queues; i++)
1077 			napi_disable(&adapter->napi[i]);
1078 		release_resources(adapter);
1079 		return rc;
1080 	}
1081 
1082 	netif_tx_start_all_queues(netdev);
1083 
1084 	if (prev_state == VNIC_CLOSED) {
1085 		for (i = 0; i < adapter->req_rx_queues; i++)
1086 			napi_schedule(&adapter->napi[i]);
1087 	}
1088 
1089 	adapter->state = VNIC_OPEN;
1090 	return rc;
1091 }
1092 
1093 static int ibmvnic_open(struct net_device *netdev)
1094 {
1095 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1096 	int rc;
1097 
1098 	/* If device failover is pending, just set device state and return.
1099 	 * Device operation will be handled by reset routine.
1100 	 */
1101 	if (adapter->failover_pending) {
1102 		adapter->state = VNIC_OPEN;
1103 		return 0;
1104 	}
1105 
1106 	mutex_lock(&adapter->reset_lock);
1107 
1108 	if (adapter->state != VNIC_CLOSED) {
1109 		rc = ibmvnic_login(netdev);
1110 		if (rc) {
1111 			mutex_unlock(&adapter->reset_lock);
1112 			return rc;
1113 		}
1114 
1115 		rc = init_resources(adapter);
1116 		if (rc) {
1117 			netdev_err(netdev, "failed to initialize resources\n");
1118 			release_resources(adapter);
1119 			mutex_unlock(&adapter->reset_lock);
1120 			return rc;
1121 		}
1122 	}
1123 
1124 	rc = __ibmvnic_open(netdev);
1125 	netif_carrier_on(netdev);
1126 
1127 	mutex_unlock(&adapter->reset_lock);
1128 
1129 	return rc;
1130 }
1131 
1132 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1133 {
1134 	struct ibmvnic_rx_pool *rx_pool;
1135 	struct ibmvnic_rx_buff *rx_buff;
1136 	u64 rx_entries;
1137 	int rx_scrqs;
1138 	int i, j;
1139 
1140 	if (!adapter->rx_pool)
1141 		return;
1142 
1143 	rx_scrqs = adapter->num_active_rx_pools;
1144 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1145 
1146 	/* Free any remaining skbs in the rx buffer pools */
1147 	for (i = 0; i < rx_scrqs; i++) {
1148 		rx_pool = &adapter->rx_pool[i];
1149 		if (!rx_pool || !rx_pool->rx_buff)
1150 			continue;
1151 
1152 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1153 		for (j = 0; j < rx_entries; j++) {
1154 			rx_buff = &rx_pool->rx_buff[j];
1155 			if (rx_buff && rx_buff->skb) {
1156 				dev_kfree_skb_any(rx_buff->skb);
1157 				rx_buff->skb = NULL;
1158 			}
1159 		}
1160 	}
1161 }
1162 
1163 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1164 			      struct ibmvnic_tx_pool *tx_pool)
1165 {
1166 	struct ibmvnic_tx_buff *tx_buff;
1167 	u64 tx_entries;
1168 	int i;
1169 
1170 	if (!tx_pool || !tx_pool->tx_buff)
1171 		return;
1172 
1173 	tx_entries = tx_pool->num_buffers;
1174 
1175 	for (i = 0; i < tx_entries; i++) {
1176 		tx_buff = &tx_pool->tx_buff[i];
1177 		if (tx_buff && tx_buff->skb) {
1178 			dev_kfree_skb_any(tx_buff->skb);
1179 			tx_buff->skb = NULL;
1180 		}
1181 	}
1182 }
1183 
1184 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1185 {
1186 	int tx_scrqs;
1187 	int i;
1188 
1189 	if (!adapter->tx_pool || !adapter->tso_pool)
1190 		return;
1191 
1192 	tx_scrqs = adapter->num_active_tx_pools;
1193 
1194 	/* Free any remaining skbs in the tx buffer pools */
1195 	for (i = 0; i < tx_scrqs; i++) {
1196 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1197 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1198 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1199 	}
1200 }
1201 
1202 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1203 {
1204 	struct net_device *netdev = adapter->netdev;
1205 	int i;
1206 
1207 	if (adapter->tx_scrq) {
1208 		for (i = 0; i < adapter->req_tx_queues; i++)
1209 			if (adapter->tx_scrq[i]->irq) {
1210 				netdev_dbg(netdev,
1211 					   "Disabling tx_scrq[%d] irq\n", i);
1212 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1213 				disable_irq(adapter->tx_scrq[i]->irq);
1214 			}
1215 	}
1216 
1217 	if (adapter->rx_scrq) {
1218 		for (i = 0; i < adapter->req_rx_queues; i++) {
1219 			if (adapter->rx_scrq[i]->irq) {
1220 				netdev_dbg(netdev,
1221 					   "Disabling rx_scrq[%d] irq\n", i);
1222 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1223 				disable_irq(adapter->rx_scrq[i]->irq);
1224 			}
1225 		}
1226 	}
1227 }
1228 
1229 static void ibmvnic_cleanup(struct net_device *netdev)
1230 {
1231 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1232 
1233 	/* ensure that transmissions are stopped if called by do_reset */
1234 	if (adapter->resetting)
1235 		netif_tx_disable(netdev);
1236 	else
1237 		netif_tx_stop_all_queues(netdev);
1238 
1239 	ibmvnic_napi_disable(adapter);
1240 	ibmvnic_disable_irqs(adapter);
1241 
1242 	clean_rx_pools(adapter);
1243 	clean_tx_pools(adapter);
1244 }
1245 
1246 static int __ibmvnic_close(struct net_device *netdev)
1247 {
1248 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1249 	int rc = 0;
1250 
1251 	adapter->state = VNIC_CLOSING;
1252 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1253 	if (rc)
1254 		return rc;
1255 	adapter->state = VNIC_CLOSED;
1256 	return 0;
1257 }
1258 
1259 static int ibmvnic_close(struct net_device *netdev)
1260 {
1261 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1262 	int rc;
1263 
1264 	/* If device failover is pending, just set device state and return.
1265 	 * Device operation will be handled by reset routine.
1266 	 */
1267 	if (adapter->failover_pending) {
1268 		adapter->state = VNIC_CLOSED;
1269 		return 0;
1270 	}
1271 
1272 	mutex_lock(&adapter->reset_lock);
1273 	rc = __ibmvnic_close(netdev);
1274 	ibmvnic_cleanup(netdev);
1275 	mutex_unlock(&adapter->reset_lock);
1276 
1277 	return rc;
1278 }
1279 
1280 /**
1281  * build_hdr_data - creates L2/L3/L4 header data buffer
1282  * @hdr_field - bitfield determining needed headers
1283  * @skb - socket buffer
1284  * @hdr_len - array of header lengths
1285  * @tot_len - total length of data
1286  *
1287  * Reads hdr_field to determine which headers are needed by firmware.
1288  * Builds a buffer containing these headers.  Saves individual header
1289  * lengths and total buffer length to be used to build descriptors.
1290  */
1291 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1292 			  int *hdr_len, u8 *hdr_data)
1293 {
1294 	int len = 0;
1295 	u8 *hdr;
1296 
1297 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1298 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1299 	else
1300 		hdr_len[0] = sizeof(struct ethhdr);
1301 
1302 	if (skb->protocol == htons(ETH_P_IP)) {
1303 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1304 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1305 			hdr_len[2] = tcp_hdrlen(skb);
1306 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1307 			hdr_len[2] = sizeof(struct udphdr);
1308 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1309 		hdr_len[1] = sizeof(struct ipv6hdr);
1310 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1311 			hdr_len[2] = tcp_hdrlen(skb);
1312 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1313 			hdr_len[2] = sizeof(struct udphdr);
1314 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1315 		hdr_len[1] = arp_hdr_len(skb->dev);
1316 		hdr_len[2] = 0;
1317 	}
1318 
1319 	memset(hdr_data, 0, 120);
1320 	if ((hdr_field >> 6) & 1) {
1321 		hdr = skb_mac_header(skb);
1322 		memcpy(hdr_data, hdr, hdr_len[0]);
1323 		len += hdr_len[0];
1324 	}
1325 
1326 	if ((hdr_field >> 5) & 1) {
1327 		hdr = skb_network_header(skb);
1328 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1329 		len += hdr_len[1];
1330 	}
1331 
1332 	if ((hdr_field >> 4) & 1) {
1333 		hdr = skb_transport_header(skb);
1334 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1335 		len += hdr_len[2];
1336 	}
1337 	return len;
1338 }
1339 
1340 /**
1341  * create_hdr_descs - create header and header extension descriptors
1342  * @hdr_field - bitfield determining needed headers
1343  * @data - buffer containing header data
1344  * @len - length of data buffer
1345  * @hdr_len - array of individual header lengths
1346  * @scrq_arr - descriptor array
1347  *
1348  * Creates header and, if needed, header extension descriptors and
1349  * places them in a descriptor array, scrq_arr
1350  */
1351 
1352 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1353 			    union sub_crq *scrq_arr)
1354 {
1355 	union sub_crq hdr_desc;
1356 	int tmp_len = len;
1357 	int num_descs = 0;
1358 	u8 *data, *cur;
1359 	int tmp;
1360 
1361 	while (tmp_len > 0) {
1362 		cur = hdr_data + len - tmp_len;
1363 
1364 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1365 		if (cur != hdr_data) {
1366 			data = hdr_desc.hdr_ext.data;
1367 			tmp = tmp_len > 29 ? 29 : tmp_len;
1368 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1369 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1370 			hdr_desc.hdr_ext.len = tmp;
1371 		} else {
1372 			data = hdr_desc.hdr.data;
1373 			tmp = tmp_len > 24 ? 24 : tmp_len;
1374 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1375 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1376 			hdr_desc.hdr.len = tmp;
1377 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1378 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1379 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1380 			hdr_desc.hdr.flag = hdr_field << 1;
1381 		}
1382 		memcpy(data, cur, tmp);
1383 		tmp_len -= tmp;
1384 		*scrq_arr = hdr_desc;
1385 		scrq_arr++;
1386 		num_descs++;
1387 	}
1388 
1389 	return num_descs;
1390 }
1391 
1392 /**
1393  * build_hdr_descs_arr - build a header descriptor array
1394  * @skb - socket buffer
1395  * @num_entries - number of descriptors to be sent
1396  * @subcrq - first TX descriptor
1397  * @hdr_field - bit field determining which headers will be sent
1398  *
1399  * This function will build a TX descriptor array with applicable
1400  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1401  */
1402 
1403 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1404 				int *num_entries, u8 hdr_field)
1405 {
1406 	int hdr_len[3] = {0, 0, 0};
1407 	int tot_len;
1408 	u8 *hdr_data = txbuff->hdr_data;
1409 
1410 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1411 				 txbuff->hdr_data);
1412 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1413 			 txbuff->indir_arr + 1);
1414 }
1415 
1416 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1417 				    struct net_device *netdev)
1418 {
1419 	/* For some backing devices, mishandling of small packets
1420 	 * can result in a loss of connection or TX stall. Device
1421 	 * architects recommend that no packet should be smaller
1422 	 * than the minimum MTU value provided to the driver, so
1423 	 * pad any packets to that length
1424 	 */
1425 	if (skb->len < netdev->min_mtu)
1426 		return skb_put_padto(skb, netdev->min_mtu);
1427 
1428 	return 0;
1429 }
1430 
1431 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1432 {
1433 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1434 	int queue_num = skb_get_queue_mapping(skb);
1435 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1436 	struct device *dev = &adapter->vdev->dev;
1437 	struct ibmvnic_tx_buff *tx_buff = NULL;
1438 	struct ibmvnic_sub_crq_queue *tx_scrq;
1439 	struct ibmvnic_tx_pool *tx_pool;
1440 	unsigned int tx_send_failed = 0;
1441 	unsigned int tx_map_failed = 0;
1442 	unsigned int tx_dropped = 0;
1443 	unsigned int tx_packets = 0;
1444 	unsigned int tx_bytes = 0;
1445 	dma_addr_t data_dma_addr;
1446 	struct netdev_queue *txq;
1447 	unsigned long lpar_rc;
1448 	union sub_crq tx_crq;
1449 	unsigned int offset;
1450 	int num_entries = 1;
1451 	unsigned char *dst;
1452 	u64 *handle_array;
1453 	int index = 0;
1454 	u8 proto = 0;
1455 	int ret = 0;
1456 
1457 	if (adapter->resetting) {
1458 		if (!netif_subqueue_stopped(netdev, skb))
1459 			netif_stop_subqueue(netdev, queue_num);
1460 		dev_kfree_skb_any(skb);
1461 
1462 		tx_send_failed++;
1463 		tx_dropped++;
1464 		ret = NETDEV_TX_OK;
1465 		goto out;
1466 	}
1467 
1468 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1469 		tx_dropped++;
1470 		tx_send_failed++;
1471 		ret = NETDEV_TX_OK;
1472 		goto out;
1473 	}
1474 	if (skb_is_gso(skb))
1475 		tx_pool = &adapter->tso_pool[queue_num];
1476 	else
1477 		tx_pool = &adapter->tx_pool[queue_num];
1478 
1479 	tx_scrq = adapter->tx_scrq[queue_num];
1480 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1481 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1482 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1483 
1484 	index = tx_pool->free_map[tx_pool->consumer_index];
1485 
1486 	if (index == IBMVNIC_INVALID_MAP) {
1487 		dev_kfree_skb_any(skb);
1488 		tx_send_failed++;
1489 		tx_dropped++;
1490 		ret = NETDEV_TX_OK;
1491 		goto out;
1492 	}
1493 
1494 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1495 
1496 	offset = index * tx_pool->buf_size;
1497 	dst = tx_pool->long_term_buff.buff + offset;
1498 	memset(dst, 0, tx_pool->buf_size);
1499 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1500 
1501 	if (skb_shinfo(skb)->nr_frags) {
1502 		int cur, i;
1503 
1504 		/* Copy the head */
1505 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1506 		cur = skb_headlen(skb);
1507 
1508 		/* Copy the frags */
1509 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1510 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1511 
1512 			memcpy(dst + cur,
1513 			       page_address(skb_frag_page(frag)) +
1514 			       frag->page_offset, skb_frag_size(frag));
1515 			cur += skb_frag_size(frag);
1516 		}
1517 	} else {
1518 		skb_copy_from_linear_data(skb, dst, skb->len);
1519 	}
1520 
1521 	tx_pool->consumer_index =
1522 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1523 
1524 	tx_buff = &tx_pool->tx_buff[index];
1525 	tx_buff->skb = skb;
1526 	tx_buff->data_dma[0] = data_dma_addr;
1527 	tx_buff->data_len[0] = skb->len;
1528 	tx_buff->index = index;
1529 	tx_buff->pool_index = queue_num;
1530 	tx_buff->last_frag = true;
1531 
1532 	memset(&tx_crq, 0, sizeof(tx_crq));
1533 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1534 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1535 	tx_crq.v1.n_crq_elem = 1;
1536 	tx_crq.v1.n_sge = 1;
1537 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1538 
1539 	if (skb_is_gso(skb))
1540 		tx_crq.v1.correlator =
1541 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1542 	else
1543 		tx_crq.v1.correlator = cpu_to_be32(index);
1544 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1545 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1546 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1547 
1548 	if (adapter->vlan_header_insertion) {
1549 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1550 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1551 	}
1552 
1553 	if (skb->protocol == htons(ETH_P_IP)) {
1554 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1555 		proto = ip_hdr(skb)->protocol;
1556 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1557 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1558 		proto = ipv6_hdr(skb)->nexthdr;
1559 	}
1560 
1561 	if (proto == IPPROTO_TCP)
1562 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1563 	else if (proto == IPPROTO_UDP)
1564 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1565 
1566 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1567 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1568 		hdrs += 2;
1569 	}
1570 	if (skb_is_gso(skb)) {
1571 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1572 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1573 		hdrs += 2;
1574 	}
1575 	/* determine if l2/3/4 headers are sent to firmware */
1576 	if ((*hdrs >> 7) & 1) {
1577 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1578 		tx_crq.v1.n_crq_elem = num_entries;
1579 		tx_buff->num_entries = num_entries;
1580 		tx_buff->indir_arr[0] = tx_crq;
1581 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1582 						    sizeof(tx_buff->indir_arr),
1583 						    DMA_TO_DEVICE);
1584 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1585 			dev_kfree_skb_any(skb);
1586 			tx_buff->skb = NULL;
1587 			if (!firmware_has_feature(FW_FEATURE_CMO))
1588 				dev_err(dev, "tx: unable to map descriptor array\n");
1589 			tx_map_failed++;
1590 			tx_dropped++;
1591 			ret = NETDEV_TX_OK;
1592 			goto tx_err_out;
1593 		}
1594 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1595 					       (u64)tx_buff->indir_dma,
1596 					       (u64)num_entries);
1597 	} else {
1598 		tx_buff->num_entries = num_entries;
1599 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1600 				      &tx_crq);
1601 	}
1602 	if (lpar_rc != H_SUCCESS) {
1603 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1604 			dev_err_ratelimited(dev, "tx: send failed\n");
1605 		dev_kfree_skb_any(skb);
1606 		tx_buff->skb = NULL;
1607 
1608 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1609 			/* Disable TX and report carrier off if queue is closed
1610 			 * or pending failover.
1611 			 * Firmware guarantees that a signal will be sent to the
1612 			 * driver, triggering a reset or some other action.
1613 			 */
1614 			netif_tx_stop_all_queues(netdev);
1615 			netif_carrier_off(netdev);
1616 		}
1617 
1618 		tx_send_failed++;
1619 		tx_dropped++;
1620 		ret = NETDEV_TX_OK;
1621 		goto tx_err_out;
1622 	}
1623 
1624 	if (atomic_add_return(num_entries, &tx_scrq->used)
1625 					>= adapter->req_tx_entries_per_subcrq) {
1626 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1627 		netif_stop_subqueue(netdev, queue_num);
1628 	}
1629 
1630 	tx_packets++;
1631 	tx_bytes += skb->len;
1632 	txq->trans_start = jiffies;
1633 	ret = NETDEV_TX_OK;
1634 	goto out;
1635 
1636 tx_err_out:
1637 	/* roll back consumer index and map array*/
1638 	if (tx_pool->consumer_index == 0)
1639 		tx_pool->consumer_index =
1640 			tx_pool->num_buffers - 1;
1641 	else
1642 		tx_pool->consumer_index--;
1643 	tx_pool->free_map[tx_pool->consumer_index] = index;
1644 out:
1645 	netdev->stats.tx_dropped += tx_dropped;
1646 	netdev->stats.tx_bytes += tx_bytes;
1647 	netdev->stats.tx_packets += tx_packets;
1648 	adapter->tx_send_failed += tx_send_failed;
1649 	adapter->tx_map_failed += tx_map_failed;
1650 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1651 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1652 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1653 
1654 	return ret;
1655 }
1656 
1657 static void ibmvnic_set_multi(struct net_device *netdev)
1658 {
1659 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1660 	struct netdev_hw_addr *ha;
1661 	union ibmvnic_crq crq;
1662 
1663 	memset(&crq, 0, sizeof(crq));
1664 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1665 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1666 
1667 	if (netdev->flags & IFF_PROMISC) {
1668 		if (!adapter->promisc_supported)
1669 			return;
1670 	} else {
1671 		if (netdev->flags & IFF_ALLMULTI) {
1672 			/* Accept all multicast */
1673 			memset(&crq, 0, sizeof(crq));
1674 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1675 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1676 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1677 			ibmvnic_send_crq(adapter, &crq);
1678 		} else if (netdev_mc_empty(netdev)) {
1679 			/* Reject all multicast */
1680 			memset(&crq, 0, sizeof(crq));
1681 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1682 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1683 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1684 			ibmvnic_send_crq(adapter, &crq);
1685 		} else {
1686 			/* Accept one or more multicast(s) */
1687 			netdev_for_each_mc_addr(ha, netdev) {
1688 				memset(&crq, 0, sizeof(crq));
1689 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1690 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1691 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1692 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1693 						ha->addr);
1694 				ibmvnic_send_crq(adapter, &crq);
1695 			}
1696 		}
1697 	}
1698 }
1699 
1700 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1701 {
1702 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1703 	struct sockaddr *addr = p;
1704 	union ibmvnic_crq crq;
1705 	int rc;
1706 
1707 	if (!is_valid_ether_addr(addr->sa_data))
1708 		return -EADDRNOTAVAIL;
1709 
1710 	memset(&crq, 0, sizeof(crq));
1711 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1712 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1713 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1714 
1715 	init_completion(&adapter->fw_done);
1716 	rc = ibmvnic_send_crq(adapter, &crq);
1717 	if (rc)
1718 		return rc;
1719 	wait_for_completion(&adapter->fw_done);
1720 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1721 	return adapter->fw_done_rc ? -EIO : 0;
1722 }
1723 
1724 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1725 {
1726 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1727 	struct sockaddr *addr = p;
1728 	int rc;
1729 
1730 	if (adapter->state == VNIC_PROBED) {
1731 		memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1732 		adapter->mac_change_pending = true;
1733 		return 0;
1734 	}
1735 
1736 	rc = __ibmvnic_set_mac(netdev, addr);
1737 
1738 	return rc;
1739 }
1740 
1741 /**
1742  * do_reset returns zero if we are able to keep processing reset events, or
1743  * non-zero if we hit a fatal error and must halt.
1744  */
1745 static int do_reset(struct ibmvnic_adapter *adapter,
1746 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1747 {
1748 	u64 old_num_rx_queues, old_num_tx_queues;
1749 	struct net_device *netdev = adapter->netdev;
1750 	int i, rc;
1751 
1752 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1753 		   rwi->reset_reason);
1754 
1755 	netif_carrier_off(netdev);
1756 	adapter->reset_reason = rwi->reset_reason;
1757 
1758 	old_num_rx_queues = adapter->req_rx_queues;
1759 	old_num_tx_queues = adapter->req_tx_queues;
1760 
1761 	ibmvnic_cleanup(netdev);
1762 
1763 	if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1764 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1765 		rc = __ibmvnic_close(netdev);
1766 		if (rc)
1767 			return rc;
1768 	}
1769 
1770 	if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1771 	    adapter->wait_for_reset) {
1772 		release_resources(adapter);
1773 		release_sub_crqs(adapter, 1);
1774 		release_crq_queue(adapter);
1775 	}
1776 
1777 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1778 		/* remove the closed state so when we call open it appears
1779 		 * we are coming from the probed state.
1780 		 */
1781 		adapter->state = VNIC_PROBED;
1782 
1783 		if (adapter->wait_for_reset) {
1784 			rc = init_crq_queue(adapter);
1785 		} else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1786 			rc = ibmvnic_reenable_crq_queue(adapter);
1787 			release_sub_crqs(adapter, 1);
1788 		} else {
1789 			rc = ibmvnic_reset_crq(adapter);
1790 			if (!rc)
1791 				rc = vio_enable_interrupts(adapter->vdev);
1792 		}
1793 
1794 		if (rc) {
1795 			netdev_err(adapter->netdev,
1796 				   "Couldn't initialize crq. rc=%d\n", rc);
1797 			return rc;
1798 		}
1799 
1800 		rc = ibmvnic_reset_init(adapter);
1801 		if (rc)
1802 			return IBMVNIC_INIT_FAILED;
1803 
1804 		/* If the adapter was in PROBE state prior to the reset,
1805 		 * exit here.
1806 		 */
1807 		if (reset_state == VNIC_PROBED)
1808 			return 0;
1809 
1810 		rc = ibmvnic_login(netdev);
1811 		if (rc) {
1812 			adapter->state = reset_state;
1813 			return rc;
1814 		}
1815 
1816 		if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1817 		    adapter->wait_for_reset) {
1818 			rc = init_resources(adapter);
1819 			if (rc)
1820 				return rc;
1821 		} else if (adapter->req_rx_queues != old_num_rx_queues ||
1822 			   adapter->req_tx_queues != old_num_tx_queues) {
1823 			adapter->map_id = 1;
1824 			release_rx_pools(adapter);
1825 			release_tx_pools(adapter);
1826 			rc = init_rx_pools(netdev);
1827 			if (rc)
1828 				return rc;
1829 			rc = init_tx_pools(netdev);
1830 			if (rc)
1831 				return rc;
1832 
1833 			release_napi(adapter);
1834 			rc = init_napi(adapter);
1835 			if (rc)
1836 				return rc;
1837 		} else {
1838 			rc = reset_tx_pools(adapter);
1839 			if (rc)
1840 				return rc;
1841 
1842 			rc = reset_rx_pools(adapter);
1843 			if (rc)
1844 				return rc;
1845 		}
1846 		ibmvnic_disable_irqs(adapter);
1847 	}
1848 	adapter->state = VNIC_CLOSED;
1849 
1850 	if (reset_state == VNIC_CLOSED)
1851 		return 0;
1852 
1853 	rc = __ibmvnic_open(netdev);
1854 	if (rc) {
1855 		if (list_empty(&adapter->rwi_list))
1856 			adapter->state = VNIC_CLOSED;
1857 		else
1858 			adapter->state = reset_state;
1859 
1860 		return 0;
1861 	}
1862 
1863 	/* kick napi */
1864 	for (i = 0; i < adapter->req_rx_queues; i++)
1865 		napi_schedule(&adapter->napi[i]);
1866 
1867 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1868 	    adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1869 		netdev_notify_peers(netdev);
1870 
1871 	netif_carrier_on(netdev);
1872 
1873 	return 0;
1874 }
1875 
1876 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1877 			 struct ibmvnic_rwi *rwi, u32 reset_state)
1878 {
1879 	struct net_device *netdev = adapter->netdev;
1880 	int rc;
1881 
1882 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1883 		   rwi->reset_reason);
1884 
1885 	netif_carrier_off(netdev);
1886 	adapter->reset_reason = rwi->reset_reason;
1887 
1888 	ibmvnic_cleanup(netdev);
1889 	release_resources(adapter);
1890 	release_sub_crqs(adapter, 0);
1891 	release_crq_queue(adapter);
1892 
1893 	/* remove the closed state so when we call open it appears
1894 	 * we are coming from the probed state.
1895 	 */
1896 	adapter->state = VNIC_PROBED;
1897 
1898 	rc = init_crq_queue(adapter);
1899 	if (rc) {
1900 		netdev_err(adapter->netdev,
1901 			   "Couldn't initialize crq. rc=%d\n", rc);
1902 		return rc;
1903 	}
1904 
1905 	rc = ibmvnic_init(adapter);
1906 	if (rc)
1907 		return rc;
1908 
1909 	/* If the adapter was in PROBE state prior to the reset,
1910 	 * exit here.
1911 	 */
1912 	if (reset_state == VNIC_PROBED)
1913 		return 0;
1914 
1915 	rc = ibmvnic_login(netdev);
1916 	if (rc) {
1917 		adapter->state = VNIC_PROBED;
1918 		return 0;
1919 	}
1920 	/* netif_set_real_num_xx_queues needs to take rtnl lock here
1921 	 * unless wait_for_reset is set, in which case the rtnl lock
1922 	 * has already been taken before initializing the reset
1923 	 */
1924 	if (!adapter->wait_for_reset) {
1925 		rtnl_lock();
1926 		rc = init_resources(adapter);
1927 		rtnl_unlock();
1928 	} else {
1929 		rc = init_resources(adapter);
1930 	}
1931 	if (rc)
1932 		return rc;
1933 
1934 	ibmvnic_disable_irqs(adapter);
1935 	adapter->state = VNIC_CLOSED;
1936 
1937 	if (reset_state == VNIC_CLOSED)
1938 		return 0;
1939 
1940 	rc = __ibmvnic_open(netdev);
1941 	if (rc) {
1942 		if (list_empty(&adapter->rwi_list))
1943 			adapter->state = VNIC_CLOSED;
1944 		else
1945 			adapter->state = reset_state;
1946 
1947 		return 0;
1948 	}
1949 
1950 	netif_carrier_on(netdev);
1951 
1952 	return 0;
1953 }
1954 
1955 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1956 {
1957 	struct ibmvnic_rwi *rwi;
1958 
1959 	mutex_lock(&adapter->rwi_lock);
1960 
1961 	if (!list_empty(&adapter->rwi_list)) {
1962 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1963 				       list);
1964 		list_del(&rwi->list);
1965 	} else {
1966 		rwi = NULL;
1967 	}
1968 
1969 	mutex_unlock(&adapter->rwi_lock);
1970 	return rwi;
1971 }
1972 
1973 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1974 {
1975 	struct ibmvnic_rwi *rwi;
1976 
1977 	rwi = get_next_rwi(adapter);
1978 	while (rwi) {
1979 		kfree(rwi);
1980 		rwi = get_next_rwi(adapter);
1981 	}
1982 }
1983 
1984 static void __ibmvnic_reset(struct work_struct *work)
1985 {
1986 	struct ibmvnic_rwi *rwi;
1987 	struct ibmvnic_adapter *adapter;
1988 	struct net_device *netdev;
1989 	u32 reset_state;
1990 	int rc = 0;
1991 
1992 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1993 	netdev = adapter->netdev;
1994 
1995 	mutex_lock(&adapter->reset_lock);
1996 	reset_state = adapter->state;
1997 
1998 	rwi = get_next_rwi(adapter);
1999 	while (rwi) {
2000 		if (adapter->force_reset_recovery) {
2001 			adapter->force_reset_recovery = false;
2002 			rc = do_hard_reset(adapter, rwi, reset_state);
2003 		} else {
2004 			rc = do_reset(adapter, rwi, reset_state);
2005 		}
2006 		kfree(rwi);
2007 		if (rc && rc != IBMVNIC_INIT_FAILED &&
2008 		    !adapter->force_reset_recovery)
2009 			break;
2010 
2011 		rwi = get_next_rwi(adapter);
2012 	}
2013 
2014 	if (adapter->wait_for_reset) {
2015 		adapter->wait_for_reset = false;
2016 		adapter->reset_done_rc = rc;
2017 		complete(&adapter->reset_done);
2018 	}
2019 
2020 	if (rc) {
2021 		netdev_dbg(adapter->netdev, "Reset failed\n");
2022 		free_all_rwi(adapter);
2023 		mutex_unlock(&adapter->reset_lock);
2024 		return;
2025 	}
2026 
2027 	adapter->resetting = false;
2028 	mutex_unlock(&adapter->reset_lock);
2029 }
2030 
2031 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2032 			 enum ibmvnic_reset_reason reason)
2033 {
2034 	struct list_head *entry, *tmp_entry;
2035 	struct ibmvnic_rwi *rwi, *tmp;
2036 	struct net_device *netdev = adapter->netdev;
2037 	int ret;
2038 
2039 	if (adapter->state == VNIC_REMOVING ||
2040 	    adapter->state == VNIC_REMOVED ||
2041 	    adapter->failover_pending) {
2042 		ret = EBUSY;
2043 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2044 		goto err;
2045 	}
2046 
2047 	if (adapter->state == VNIC_PROBING) {
2048 		netdev_warn(netdev, "Adapter reset during probe\n");
2049 		ret = adapter->init_done_rc = EAGAIN;
2050 		goto err;
2051 	}
2052 
2053 	mutex_lock(&adapter->rwi_lock);
2054 
2055 	list_for_each(entry, &adapter->rwi_list) {
2056 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2057 		if (tmp->reset_reason == reason) {
2058 			netdev_dbg(netdev, "Skipping matching reset\n");
2059 			mutex_unlock(&adapter->rwi_lock);
2060 			ret = EBUSY;
2061 			goto err;
2062 		}
2063 	}
2064 
2065 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
2066 	if (!rwi) {
2067 		mutex_unlock(&adapter->rwi_lock);
2068 		ibmvnic_close(netdev);
2069 		ret = ENOMEM;
2070 		goto err;
2071 	}
2072 	/* if we just received a transport event,
2073 	 * flush reset queue and process this reset
2074 	 */
2075 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2076 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2077 			list_del(entry);
2078 	}
2079 	rwi->reset_reason = reason;
2080 	list_add_tail(&rwi->list, &adapter->rwi_list);
2081 	mutex_unlock(&adapter->rwi_lock);
2082 	adapter->resetting = true;
2083 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2084 	schedule_work(&adapter->ibmvnic_reset);
2085 
2086 	return 0;
2087 err:
2088 	if (adapter->wait_for_reset)
2089 		adapter->wait_for_reset = false;
2090 	return -ret;
2091 }
2092 
2093 static void ibmvnic_tx_timeout(struct net_device *dev)
2094 {
2095 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2096 
2097 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2098 }
2099 
2100 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2101 				  struct ibmvnic_rx_buff *rx_buff)
2102 {
2103 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2104 
2105 	rx_buff->skb = NULL;
2106 
2107 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2108 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2109 
2110 	atomic_dec(&pool->available);
2111 }
2112 
2113 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2114 {
2115 	struct net_device *netdev = napi->dev;
2116 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2117 	int scrq_num = (int)(napi - adapter->napi);
2118 	int frames_processed = 0;
2119 
2120 restart_poll:
2121 	while (frames_processed < budget) {
2122 		struct sk_buff *skb;
2123 		struct ibmvnic_rx_buff *rx_buff;
2124 		union sub_crq *next;
2125 		u32 length;
2126 		u16 offset;
2127 		u8 flags = 0;
2128 
2129 		if (unlikely(adapter->resetting &&
2130 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2131 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2132 			napi_complete_done(napi, frames_processed);
2133 			return frames_processed;
2134 		}
2135 
2136 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2137 			break;
2138 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2139 		rx_buff =
2140 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2141 							  rx_comp.correlator);
2142 		/* do error checking */
2143 		if (next->rx_comp.rc) {
2144 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2145 				   be16_to_cpu(next->rx_comp.rc));
2146 			/* free the entry */
2147 			next->rx_comp.first = 0;
2148 			dev_kfree_skb_any(rx_buff->skb);
2149 			remove_buff_from_pool(adapter, rx_buff);
2150 			continue;
2151 		} else if (!rx_buff->skb) {
2152 			/* free the entry */
2153 			next->rx_comp.first = 0;
2154 			remove_buff_from_pool(adapter, rx_buff);
2155 			continue;
2156 		}
2157 
2158 		length = be32_to_cpu(next->rx_comp.len);
2159 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2160 		flags = next->rx_comp.flags;
2161 		skb = rx_buff->skb;
2162 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2163 					length);
2164 
2165 		/* VLAN Header has been stripped by the system firmware and
2166 		 * needs to be inserted by the driver
2167 		 */
2168 		if (adapter->rx_vlan_header_insertion &&
2169 		    (flags & IBMVNIC_VLAN_STRIPPED))
2170 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2171 					       ntohs(next->rx_comp.vlan_tci));
2172 
2173 		/* free the entry */
2174 		next->rx_comp.first = 0;
2175 		remove_buff_from_pool(adapter, rx_buff);
2176 
2177 		skb_put(skb, length);
2178 		skb->protocol = eth_type_trans(skb, netdev);
2179 		skb_record_rx_queue(skb, scrq_num);
2180 
2181 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2182 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2183 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2184 		}
2185 
2186 		length = skb->len;
2187 		napi_gro_receive(napi, skb); /* send it up */
2188 		netdev->stats.rx_packets++;
2189 		netdev->stats.rx_bytes += length;
2190 		adapter->rx_stats_buffers[scrq_num].packets++;
2191 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2192 		frames_processed++;
2193 	}
2194 
2195 	if (adapter->state != VNIC_CLOSING)
2196 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2197 
2198 	if (frames_processed < budget) {
2199 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2200 		napi_complete_done(napi, frames_processed);
2201 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2202 		    napi_reschedule(napi)) {
2203 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2204 			goto restart_poll;
2205 		}
2206 	}
2207 	return frames_processed;
2208 }
2209 
2210 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2211 {
2212 	int rc, ret;
2213 
2214 	adapter->fallback.mtu = adapter->req_mtu;
2215 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2216 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2217 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2218 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2219 
2220 	init_completion(&adapter->reset_done);
2221 	adapter->wait_for_reset = true;
2222 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2223 	if (rc)
2224 		return rc;
2225 	wait_for_completion(&adapter->reset_done);
2226 
2227 	ret = 0;
2228 	if (adapter->reset_done_rc) {
2229 		ret = -EIO;
2230 		adapter->desired.mtu = adapter->fallback.mtu;
2231 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2232 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2233 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2234 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2235 
2236 		init_completion(&adapter->reset_done);
2237 		adapter->wait_for_reset = true;
2238 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2239 		if (rc)
2240 			return ret;
2241 		wait_for_completion(&adapter->reset_done);
2242 	}
2243 	adapter->wait_for_reset = false;
2244 
2245 	return ret;
2246 }
2247 
2248 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2249 {
2250 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2251 
2252 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2253 
2254 	return wait_for_reset(adapter);
2255 }
2256 
2257 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2258 						struct net_device *dev,
2259 						netdev_features_t features)
2260 {
2261 	/* Some backing hardware adapters can not
2262 	 * handle packets with a MSS less than 224
2263 	 * or with only one segment.
2264 	 */
2265 	if (skb_is_gso(skb)) {
2266 		if (skb_shinfo(skb)->gso_size < 224 ||
2267 		    skb_shinfo(skb)->gso_segs == 1)
2268 			features &= ~NETIF_F_GSO_MASK;
2269 	}
2270 
2271 	return features;
2272 }
2273 
2274 static const struct net_device_ops ibmvnic_netdev_ops = {
2275 	.ndo_open		= ibmvnic_open,
2276 	.ndo_stop		= ibmvnic_close,
2277 	.ndo_start_xmit		= ibmvnic_xmit,
2278 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2279 	.ndo_set_mac_address	= ibmvnic_set_mac,
2280 	.ndo_validate_addr	= eth_validate_addr,
2281 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2282 	.ndo_change_mtu		= ibmvnic_change_mtu,
2283 	.ndo_features_check     = ibmvnic_features_check,
2284 };
2285 
2286 /* ethtool functions */
2287 
2288 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2289 				      struct ethtool_link_ksettings *cmd)
2290 {
2291 	u32 supported, advertising;
2292 
2293 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2294 			  SUPPORTED_FIBRE);
2295 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2296 			    ADVERTISED_FIBRE);
2297 	cmd->base.speed = SPEED_1000;
2298 	cmd->base.duplex = DUPLEX_FULL;
2299 	cmd->base.port = PORT_FIBRE;
2300 	cmd->base.phy_address = 0;
2301 	cmd->base.autoneg = AUTONEG_ENABLE;
2302 
2303 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2304 						supported);
2305 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2306 						advertising);
2307 
2308 	return 0;
2309 }
2310 
2311 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2312 				struct ethtool_drvinfo *info)
2313 {
2314 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2315 
2316 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2317 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2318 	strlcpy(info->fw_version, adapter->fw_version,
2319 		sizeof(info->fw_version));
2320 }
2321 
2322 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2323 {
2324 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2325 
2326 	return adapter->msg_enable;
2327 }
2328 
2329 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2330 {
2331 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2332 
2333 	adapter->msg_enable = data;
2334 }
2335 
2336 static u32 ibmvnic_get_link(struct net_device *netdev)
2337 {
2338 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2339 
2340 	/* Don't need to send a query because we request a logical link up at
2341 	 * init and then we wait for link state indications
2342 	 */
2343 	return adapter->logical_link_state;
2344 }
2345 
2346 static void ibmvnic_get_ringparam(struct net_device *netdev,
2347 				  struct ethtool_ringparam *ring)
2348 {
2349 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2350 
2351 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2352 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2353 	ring->rx_mini_max_pending = 0;
2354 	ring->rx_jumbo_max_pending = 0;
2355 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2356 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2357 	ring->rx_mini_pending = 0;
2358 	ring->rx_jumbo_pending = 0;
2359 }
2360 
2361 static int ibmvnic_set_ringparam(struct net_device *netdev,
2362 				 struct ethtool_ringparam *ring)
2363 {
2364 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2365 
2366 	if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2367 	    ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2368 		netdev_err(netdev, "Invalid request.\n");
2369 		netdev_err(netdev, "Max tx buffers = %llu\n",
2370 			   adapter->max_rx_add_entries_per_subcrq);
2371 		netdev_err(netdev, "Max rx buffers = %llu\n",
2372 			   adapter->max_tx_entries_per_subcrq);
2373 		return -EINVAL;
2374 	}
2375 
2376 	adapter->desired.rx_entries = ring->rx_pending;
2377 	adapter->desired.tx_entries = ring->tx_pending;
2378 
2379 	return wait_for_reset(adapter);
2380 }
2381 
2382 static void ibmvnic_get_channels(struct net_device *netdev,
2383 				 struct ethtool_channels *channels)
2384 {
2385 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2386 
2387 	channels->max_rx = adapter->max_rx_queues;
2388 	channels->max_tx = adapter->max_tx_queues;
2389 	channels->max_other = 0;
2390 	channels->max_combined = 0;
2391 	channels->rx_count = adapter->req_rx_queues;
2392 	channels->tx_count = adapter->req_tx_queues;
2393 	channels->other_count = 0;
2394 	channels->combined_count = 0;
2395 }
2396 
2397 static int ibmvnic_set_channels(struct net_device *netdev,
2398 				struct ethtool_channels *channels)
2399 {
2400 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2401 
2402 	adapter->desired.rx_queues = channels->rx_count;
2403 	adapter->desired.tx_queues = channels->tx_count;
2404 
2405 	return wait_for_reset(adapter);
2406 }
2407 
2408 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2409 {
2410 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2411 	int i;
2412 
2413 	if (stringset != ETH_SS_STATS)
2414 		return;
2415 
2416 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2417 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2418 
2419 	for (i = 0; i < adapter->req_tx_queues; i++) {
2420 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2421 		data += ETH_GSTRING_LEN;
2422 
2423 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2424 		data += ETH_GSTRING_LEN;
2425 
2426 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2427 		data += ETH_GSTRING_LEN;
2428 	}
2429 
2430 	for (i = 0; i < adapter->req_rx_queues; i++) {
2431 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2432 		data += ETH_GSTRING_LEN;
2433 
2434 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2435 		data += ETH_GSTRING_LEN;
2436 
2437 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2438 		data += ETH_GSTRING_LEN;
2439 	}
2440 }
2441 
2442 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2443 {
2444 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2445 
2446 	switch (sset) {
2447 	case ETH_SS_STATS:
2448 		return ARRAY_SIZE(ibmvnic_stats) +
2449 		       adapter->req_tx_queues * NUM_TX_STATS +
2450 		       adapter->req_rx_queues * NUM_RX_STATS;
2451 	default:
2452 		return -EOPNOTSUPP;
2453 	}
2454 }
2455 
2456 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2457 				      struct ethtool_stats *stats, u64 *data)
2458 {
2459 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2460 	union ibmvnic_crq crq;
2461 	int i, j;
2462 	int rc;
2463 
2464 	memset(&crq, 0, sizeof(crq));
2465 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2466 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2467 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2468 	crq.request_statistics.len =
2469 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2470 
2471 	/* Wait for data to be written */
2472 	init_completion(&adapter->stats_done);
2473 	rc = ibmvnic_send_crq(adapter, &crq);
2474 	if (rc)
2475 		return;
2476 	wait_for_completion(&adapter->stats_done);
2477 
2478 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2479 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2480 						ibmvnic_stats[i].offset));
2481 
2482 	for (j = 0; j < adapter->req_tx_queues; j++) {
2483 		data[i] = adapter->tx_stats_buffers[j].packets;
2484 		i++;
2485 		data[i] = adapter->tx_stats_buffers[j].bytes;
2486 		i++;
2487 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2488 		i++;
2489 	}
2490 
2491 	for (j = 0; j < adapter->req_rx_queues; j++) {
2492 		data[i] = adapter->rx_stats_buffers[j].packets;
2493 		i++;
2494 		data[i] = adapter->rx_stats_buffers[j].bytes;
2495 		i++;
2496 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2497 		i++;
2498 	}
2499 }
2500 
2501 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2502 	.get_drvinfo		= ibmvnic_get_drvinfo,
2503 	.get_msglevel		= ibmvnic_get_msglevel,
2504 	.set_msglevel		= ibmvnic_set_msglevel,
2505 	.get_link		= ibmvnic_get_link,
2506 	.get_ringparam		= ibmvnic_get_ringparam,
2507 	.set_ringparam		= ibmvnic_set_ringparam,
2508 	.get_channels		= ibmvnic_get_channels,
2509 	.set_channels		= ibmvnic_set_channels,
2510 	.get_strings            = ibmvnic_get_strings,
2511 	.get_sset_count         = ibmvnic_get_sset_count,
2512 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2513 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2514 };
2515 
2516 /* Routines for managing CRQs/sCRQs  */
2517 
2518 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2519 				   struct ibmvnic_sub_crq_queue *scrq)
2520 {
2521 	int rc;
2522 
2523 	if (scrq->irq) {
2524 		free_irq(scrq->irq, scrq);
2525 		irq_dispose_mapping(scrq->irq);
2526 		scrq->irq = 0;
2527 	}
2528 
2529 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2530 	atomic_set(&scrq->used, 0);
2531 	scrq->cur = 0;
2532 
2533 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2534 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2535 	return rc;
2536 }
2537 
2538 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2539 {
2540 	int i, rc;
2541 
2542 	for (i = 0; i < adapter->req_tx_queues; i++) {
2543 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2544 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2545 		if (rc)
2546 			return rc;
2547 	}
2548 
2549 	for (i = 0; i < adapter->req_rx_queues; i++) {
2550 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2551 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2552 		if (rc)
2553 			return rc;
2554 	}
2555 
2556 	return rc;
2557 }
2558 
2559 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2560 				  struct ibmvnic_sub_crq_queue *scrq,
2561 				  bool do_h_free)
2562 {
2563 	struct device *dev = &adapter->vdev->dev;
2564 	long rc;
2565 
2566 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2567 
2568 	if (do_h_free) {
2569 		/* Close the sub-crqs */
2570 		do {
2571 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2572 						adapter->vdev->unit_address,
2573 						scrq->crq_num);
2574 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2575 
2576 		if (rc) {
2577 			netdev_err(adapter->netdev,
2578 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2579 				   scrq->crq_num, rc);
2580 		}
2581 	}
2582 
2583 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2584 			 DMA_BIDIRECTIONAL);
2585 	free_pages((unsigned long)scrq->msgs, 2);
2586 	kfree(scrq);
2587 }
2588 
2589 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2590 							*adapter)
2591 {
2592 	struct device *dev = &adapter->vdev->dev;
2593 	struct ibmvnic_sub_crq_queue *scrq;
2594 	int rc;
2595 
2596 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2597 	if (!scrq)
2598 		return NULL;
2599 
2600 	scrq->msgs =
2601 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2602 	if (!scrq->msgs) {
2603 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2604 		goto zero_page_failed;
2605 	}
2606 
2607 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2608 					 DMA_BIDIRECTIONAL);
2609 	if (dma_mapping_error(dev, scrq->msg_token)) {
2610 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2611 		goto map_failed;
2612 	}
2613 
2614 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2615 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2616 
2617 	if (rc == H_RESOURCE)
2618 		rc = ibmvnic_reset_crq(adapter);
2619 
2620 	if (rc == H_CLOSED) {
2621 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2622 	} else if (rc) {
2623 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2624 		goto reg_failed;
2625 	}
2626 
2627 	scrq->adapter = adapter;
2628 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2629 	spin_lock_init(&scrq->lock);
2630 
2631 	netdev_dbg(adapter->netdev,
2632 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2633 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2634 
2635 	return scrq;
2636 
2637 reg_failed:
2638 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2639 			 DMA_BIDIRECTIONAL);
2640 map_failed:
2641 	free_pages((unsigned long)scrq->msgs, 2);
2642 zero_page_failed:
2643 	kfree(scrq);
2644 
2645 	return NULL;
2646 }
2647 
2648 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2649 {
2650 	int i;
2651 
2652 	if (adapter->tx_scrq) {
2653 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2654 			if (!adapter->tx_scrq[i])
2655 				continue;
2656 
2657 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2658 				   i);
2659 			if (adapter->tx_scrq[i]->irq) {
2660 				free_irq(adapter->tx_scrq[i]->irq,
2661 					 adapter->tx_scrq[i]);
2662 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2663 				adapter->tx_scrq[i]->irq = 0;
2664 			}
2665 
2666 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2667 					      do_h_free);
2668 		}
2669 
2670 		kfree(adapter->tx_scrq);
2671 		adapter->tx_scrq = NULL;
2672 		adapter->num_active_tx_scrqs = 0;
2673 	}
2674 
2675 	if (adapter->rx_scrq) {
2676 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2677 			if (!adapter->rx_scrq[i])
2678 				continue;
2679 
2680 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2681 				   i);
2682 			if (adapter->rx_scrq[i]->irq) {
2683 				free_irq(adapter->rx_scrq[i]->irq,
2684 					 adapter->rx_scrq[i]);
2685 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2686 				adapter->rx_scrq[i]->irq = 0;
2687 			}
2688 
2689 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2690 					      do_h_free);
2691 		}
2692 
2693 		kfree(adapter->rx_scrq);
2694 		adapter->rx_scrq = NULL;
2695 		adapter->num_active_rx_scrqs = 0;
2696 	}
2697 }
2698 
2699 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2700 			    struct ibmvnic_sub_crq_queue *scrq)
2701 {
2702 	struct device *dev = &adapter->vdev->dev;
2703 	unsigned long rc;
2704 
2705 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2706 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2707 	if (rc)
2708 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2709 			scrq->hw_irq, rc);
2710 	return rc;
2711 }
2712 
2713 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2714 			   struct ibmvnic_sub_crq_queue *scrq)
2715 {
2716 	struct device *dev = &adapter->vdev->dev;
2717 	unsigned long rc;
2718 
2719 	if (scrq->hw_irq > 0x100000000ULL) {
2720 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2721 		return 1;
2722 	}
2723 
2724 	if (adapter->resetting &&
2725 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
2726 		u64 val = (0xff000000) | scrq->hw_irq;
2727 
2728 		rc = plpar_hcall_norets(H_EOI, val);
2729 		if (rc)
2730 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2731 				val, rc);
2732 	}
2733 
2734 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2735 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2736 	if (rc)
2737 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2738 			scrq->hw_irq, rc);
2739 	return rc;
2740 }
2741 
2742 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2743 			       struct ibmvnic_sub_crq_queue *scrq)
2744 {
2745 	struct device *dev = &adapter->vdev->dev;
2746 	struct ibmvnic_tx_pool *tx_pool;
2747 	struct ibmvnic_tx_buff *txbuff;
2748 	union sub_crq *next;
2749 	int index;
2750 	int i, j;
2751 	u8 *first;
2752 
2753 restart_loop:
2754 	while (pending_scrq(adapter, scrq)) {
2755 		unsigned int pool = scrq->pool_index;
2756 		int num_entries = 0;
2757 
2758 		next = ibmvnic_next_scrq(adapter, scrq);
2759 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2760 			if (next->tx_comp.rcs[i]) {
2761 				dev_err(dev, "tx error %x\n",
2762 					next->tx_comp.rcs[i]);
2763 				continue;
2764 			}
2765 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2766 			if (index & IBMVNIC_TSO_POOL_MASK) {
2767 				tx_pool = &adapter->tso_pool[pool];
2768 				index &= ~IBMVNIC_TSO_POOL_MASK;
2769 			} else {
2770 				tx_pool = &adapter->tx_pool[pool];
2771 			}
2772 
2773 			txbuff = &tx_pool->tx_buff[index];
2774 
2775 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2776 				if (!txbuff->data_dma[j])
2777 					continue;
2778 
2779 				txbuff->data_dma[j] = 0;
2780 			}
2781 			/* if sub_crq was sent indirectly */
2782 			first = &txbuff->indir_arr[0].generic.first;
2783 			if (*first == IBMVNIC_CRQ_CMD) {
2784 				dma_unmap_single(dev, txbuff->indir_dma,
2785 						 sizeof(txbuff->indir_arr),
2786 						 DMA_TO_DEVICE);
2787 				*first = 0;
2788 			}
2789 
2790 			if (txbuff->last_frag) {
2791 				dev_kfree_skb_any(txbuff->skb);
2792 				txbuff->skb = NULL;
2793 			}
2794 
2795 			num_entries += txbuff->num_entries;
2796 
2797 			tx_pool->free_map[tx_pool->producer_index] = index;
2798 			tx_pool->producer_index =
2799 				(tx_pool->producer_index + 1) %
2800 					tx_pool->num_buffers;
2801 		}
2802 		/* remove tx_comp scrq*/
2803 		next->tx_comp.first = 0;
2804 
2805 		if (atomic_sub_return(num_entries, &scrq->used) <=
2806 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2807 		    __netif_subqueue_stopped(adapter->netdev,
2808 					     scrq->pool_index)) {
2809 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2810 			netdev_dbg(adapter->netdev, "Started queue %d\n",
2811 				   scrq->pool_index);
2812 		}
2813 	}
2814 
2815 	enable_scrq_irq(adapter, scrq);
2816 
2817 	if (pending_scrq(adapter, scrq)) {
2818 		disable_scrq_irq(adapter, scrq);
2819 		goto restart_loop;
2820 	}
2821 
2822 	return 0;
2823 }
2824 
2825 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2826 {
2827 	struct ibmvnic_sub_crq_queue *scrq = instance;
2828 	struct ibmvnic_adapter *adapter = scrq->adapter;
2829 
2830 	disable_scrq_irq(adapter, scrq);
2831 	ibmvnic_complete_tx(adapter, scrq);
2832 
2833 	return IRQ_HANDLED;
2834 }
2835 
2836 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2837 {
2838 	struct ibmvnic_sub_crq_queue *scrq = instance;
2839 	struct ibmvnic_adapter *adapter = scrq->adapter;
2840 
2841 	/* When booting a kdump kernel we can hit pending interrupts
2842 	 * prior to completing driver initialization.
2843 	 */
2844 	if (unlikely(adapter->state != VNIC_OPEN))
2845 		return IRQ_NONE;
2846 
2847 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2848 
2849 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2850 		disable_scrq_irq(adapter, scrq);
2851 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2852 	}
2853 
2854 	return IRQ_HANDLED;
2855 }
2856 
2857 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2858 {
2859 	struct device *dev = &adapter->vdev->dev;
2860 	struct ibmvnic_sub_crq_queue *scrq;
2861 	int i = 0, j = 0;
2862 	int rc = 0;
2863 
2864 	for (i = 0; i < adapter->req_tx_queues; i++) {
2865 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2866 			   i);
2867 		scrq = adapter->tx_scrq[i];
2868 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2869 
2870 		if (!scrq->irq) {
2871 			rc = -EINVAL;
2872 			dev_err(dev, "Error mapping irq\n");
2873 			goto req_tx_irq_failed;
2874 		}
2875 
2876 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2877 				 0, "ibmvnic_tx", scrq);
2878 
2879 		if (rc) {
2880 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2881 				scrq->irq, rc);
2882 			irq_dispose_mapping(scrq->irq);
2883 			goto req_tx_irq_failed;
2884 		}
2885 	}
2886 
2887 	for (i = 0; i < adapter->req_rx_queues; i++) {
2888 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2889 			   i);
2890 		scrq = adapter->rx_scrq[i];
2891 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2892 		if (!scrq->irq) {
2893 			rc = -EINVAL;
2894 			dev_err(dev, "Error mapping irq\n");
2895 			goto req_rx_irq_failed;
2896 		}
2897 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2898 				 0, "ibmvnic_rx", scrq);
2899 		if (rc) {
2900 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2901 				scrq->irq, rc);
2902 			irq_dispose_mapping(scrq->irq);
2903 			goto req_rx_irq_failed;
2904 		}
2905 	}
2906 	return rc;
2907 
2908 req_rx_irq_failed:
2909 	for (j = 0; j < i; j++) {
2910 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2911 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2912 	}
2913 	i = adapter->req_tx_queues;
2914 req_tx_irq_failed:
2915 	for (j = 0; j < i; j++) {
2916 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2917 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2918 	}
2919 	release_sub_crqs(adapter, 1);
2920 	return rc;
2921 }
2922 
2923 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2924 {
2925 	struct device *dev = &adapter->vdev->dev;
2926 	struct ibmvnic_sub_crq_queue **allqueues;
2927 	int registered_queues = 0;
2928 	int total_queues;
2929 	int more = 0;
2930 	int i;
2931 
2932 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2933 
2934 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2935 	if (!allqueues)
2936 		return -1;
2937 
2938 	for (i = 0; i < total_queues; i++) {
2939 		allqueues[i] = init_sub_crq_queue(adapter);
2940 		if (!allqueues[i]) {
2941 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2942 			break;
2943 		}
2944 		registered_queues++;
2945 	}
2946 
2947 	/* Make sure we were able to register the minimum number of queues */
2948 	if (registered_queues <
2949 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2950 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2951 		goto tx_failed;
2952 	}
2953 
2954 	/* Distribute the failed allocated queues*/
2955 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2956 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2957 		switch (i % 3) {
2958 		case 0:
2959 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2960 				adapter->req_rx_queues--;
2961 			else
2962 				more++;
2963 			break;
2964 		case 1:
2965 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2966 				adapter->req_tx_queues--;
2967 			else
2968 				more++;
2969 			break;
2970 		}
2971 	}
2972 
2973 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2974 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2975 	if (!adapter->tx_scrq)
2976 		goto tx_failed;
2977 
2978 	for (i = 0; i < adapter->req_tx_queues; i++) {
2979 		adapter->tx_scrq[i] = allqueues[i];
2980 		adapter->tx_scrq[i]->pool_index = i;
2981 		adapter->num_active_tx_scrqs++;
2982 	}
2983 
2984 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2985 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2986 	if (!adapter->rx_scrq)
2987 		goto rx_failed;
2988 
2989 	for (i = 0; i < adapter->req_rx_queues; i++) {
2990 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2991 		adapter->rx_scrq[i]->scrq_num = i;
2992 		adapter->num_active_rx_scrqs++;
2993 	}
2994 
2995 	kfree(allqueues);
2996 	return 0;
2997 
2998 rx_failed:
2999 	kfree(adapter->tx_scrq);
3000 	adapter->tx_scrq = NULL;
3001 tx_failed:
3002 	for (i = 0; i < registered_queues; i++)
3003 		release_sub_crq_queue(adapter, allqueues[i], 1);
3004 	kfree(allqueues);
3005 	return -1;
3006 }
3007 
3008 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3009 {
3010 	struct device *dev = &adapter->vdev->dev;
3011 	union ibmvnic_crq crq;
3012 	int max_entries;
3013 
3014 	if (!retry) {
3015 		/* Sub-CRQ entries are 32 byte long */
3016 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3017 
3018 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3019 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3020 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3021 			return;
3022 		}
3023 
3024 		if (adapter->desired.mtu)
3025 			adapter->req_mtu = adapter->desired.mtu;
3026 		else
3027 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3028 
3029 		if (!adapter->desired.tx_entries)
3030 			adapter->desired.tx_entries =
3031 					adapter->max_tx_entries_per_subcrq;
3032 		if (!adapter->desired.rx_entries)
3033 			adapter->desired.rx_entries =
3034 					adapter->max_rx_add_entries_per_subcrq;
3035 
3036 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3037 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3038 
3039 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3040 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3041 			adapter->desired.tx_entries = max_entries;
3042 		}
3043 
3044 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3045 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3046 			adapter->desired.rx_entries = max_entries;
3047 		}
3048 
3049 		if (adapter->desired.tx_entries)
3050 			adapter->req_tx_entries_per_subcrq =
3051 					adapter->desired.tx_entries;
3052 		else
3053 			adapter->req_tx_entries_per_subcrq =
3054 					adapter->max_tx_entries_per_subcrq;
3055 
3056 		if (adapter->desired.rx_entries)
3057 			adapter->req_rx_add_entries_per_subcrq =
3058 					adapter->desired.rx_entries;
3059 		else
3060 			adapter->req_rx_add_entries_per_subcrq =
3061 					adapter->max_rx_add_entries_per_subcrq;
3062 
3063 		if (adapter->desired.tx_queues)
3064 			adapter->req_tx_queues =
3065 					adapter->desired.tx_queues;
3066 		else
3067 			adapter->req_tx_queues =
3068 					adapter->opt_tx_comp_sub_queues;
3069 
3070 		if (adapter->desired.rx_queues)
3071 			adapter->req_rx_queues =
3072 					adapter->desired.rx_queues;
3073 		else
3074 			adapter->req_rx_queues =
3075 					adapter->opt_rx_comp_queues;
3076 
3077 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3078 	}
3079 
3080 	memset(&crq, 0, sizeof(crq));
3081 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3082 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3083 
3084 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3085 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3086 	atomic_inc(&adapter->running_cap_crqs);
3087 	ibmvnic_send_crq(adapter, &crq);
3088 
3089 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3090 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3091 	atomic_inc(&adapter->running_cap_crqs);
3092 	ibmvnic_send_crq(adapter, &crq);
3093 
3094 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3095 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3096 	atomic_inc(&adapter->running_cap_crqs);
3097 	ibmvnic_send_crq(adapter, &crq);
3098 
3099 	crq.request_capability.capability =
3100 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3101 	crq.request_capability.number =
3102 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3103 	atomic_inc(&adapter->running_cap_crqs);
3104 	ibmvnic_send_crq(adapter, &crq);
3105 
3106 	crq.request_capability.capability =
3107 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3108 	crq.request_capability.number =
3109 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3110 	atomic_inc(&adapter->running_cap_crqs);
3111 	ibmvnic_send_crq(adapter, &crq);
3112 
3113 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3114 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3115 	atomic_inc(&adapter->running_cap_crqs);
3116 	ibmvnic_send_crq(adapter, &crq);
3117 
3118 	if (adapter->netdev->flags & IFF_PROMISC) {
3119 		if (adapter->promisc_supported) {
3120 			crq.request_capability.capability =
3121 			    cpu_to_be16(PROMISC_REQUESTED);
3122 			crq.request_capability.number = cpu_to_be64(1);
3123 			atomic_inc(&adapter->running_cap_crqs);
3124 			ibmvnic_send_crq(adapter, &crq);
3125 		}
3126 	} else {
3127 		crq.request_capability.capability =
3128 		    cpu_to_be16(PROMISC_REQUESTED);
3129 		crq.request_capability.number = cpu_to_be64(0);
3130 		atomic_inc(&adapter->running_cap_crqs);
3131 		ibmvnic_send_crq(adapter, &crq);
3132 	}
3133 }
3134 
3135 static int pending_scrq(struct ibmvnic_adapter *adapter,
3136 			struct ibmvnic_sub_crq_queue *scrq)
3137 {
3138 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3139 
3140 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3141 		return 1;
3142 	else
3143 		return 0;
3144 }
3145 
3146 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3147 					struct ibmvnic_sub_crq_queue *scrq)
3148 {
3149 	union sub_crq *entry;
3150 	unsigned long flags;
3151 
3152 	spin_lock_irqsave(&scrq->lock, flags);
3153 	entry = &scrq->msgs[scrq->cur];
3154 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3155 		if (++scrq->cur == scrq->size)
3156 			scrq->cur = 0;
3157 	} else {
3158 		entry = NULL;
3159 	}
3160 	spin_unlock_irqrestore(&scrq->lock, flags);
3161 
3162 	return entry;
3163 }
3164 
3165 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3166 {
3167 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3168 	union ibmvnic_crq *crq;
3169 
3170 	crq = &queue->msgs[queue->cur];
3171 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3172 		if (++queue->cur == queue->size)
3173 			queue->cur = 0;
3174 	} else {
3175 		crq = NULL;
3176 	}
3177 
3178 	return crq;
3179 }
3180 
3181 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3182 {
3183 	switch (rc) {
3184 	case H_PARAMETER:
3185 		dev_warn_ratelimited(dev,
3186 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3187 				     func, rc);
3188 		break;
3189 	case H_CLOSED:
3190 		dev_warn_ratelimited(dev,
3191 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3192 				     func, rc);
3193 		break;
3194 	default:
3195 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3196 		break;
3197 	}
3198 }
3199 
3200 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3201 		       union sub_crq *sub_crq)
3202 {
3203 	unsigned int ua = adapter->vdev->unit_address;
3204 	struct device *dev = &adapter->vdev->dev;
3205 	u64 *u64_crq = (u64 *)sub_crq;
3206 	int rc;
3207 
3208 	netdev_dbg(adapter->netdev,
3209 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3210 		   (unsigned long int)cpu_to_be64(remote_handle),
3211 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3212 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3213 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3214 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3215 
3216 	/* Make sure the hypervisor sees the complete request */
3217 	mb();
3218 
3219 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3220 				cpu_to_be64(remote_handle),
3221 				cpu_to_be64(u64_crq[0]),
3222 				cpu_to_be64(u64_crq[1]),
3223 				cpu_to_be64(u64_crq[2]),
3224 				cpu_to_be64(u64_crq[3]));
3225 
3226 	if (rc)
3227 		print_subcrq_error(dev, rc, __func__);
3228 
3229 	return rc;
3230 }
3231 
3232 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3233 				u64 remote_handle, u64 ioba, u64 num_entries)
3234 {
3235 	unsigned int ua = adapter->vdev->unit_address;
3236 	struct device *dev = &adapter->vdev->dev;
3237 	int rc;
3238 
3239 	/* Make sure the hypervisor sees the complete request */
3240 	mb();
3241 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3242 				cpu_to_be64(remote_handle),
3243 				ioba, num_entries);
3244 
3245 	if (rc)
3246 		print_subcrq_error(dev, rc, __func__);
3247 
3248 	return rc;
3249 }
3250 
3251 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3252 			    union ibmvnic_crq *crq)
3253 {
3254 	unsigned int ua = adapter->vdev->unit_address;
3255 	struct device *dev = &adapter->vdev->dev;
3256 	u64 *u64_crq = (u64 *)crq;
3257 	int rc;
3258 
3259 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3260 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3261 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3262 
3263 	if (!adapter->crq.active &&
3264 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3265 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3266 		return -EINVAL;
3267 	}
3268 
3269 	/* Make sure the hypervisor sees the complete request */
3270 	mb();
3271 
3272 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3273 				cpu_to_be64(u64_crq[0]),
3274 				cpu_to_be64(u64_crq[1]));
3275 
3276 	if (rc) {
3277 		if (rc == H_CLOSED) {
3278 			dev_warn(dev, "CRQ Queue closed\n");
3279 			if (adapter->resetting)
3280 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3281 		}
3282 
3283 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3284 	}
3285 
3286 	return rc;
3287 }
3288 
3289 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3290 {
3291 	union ibmvnic_crq crq;
3292 
3293 	memset(&crq, 0, sizeof(crq));
3294 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3295 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3296 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3297 
3298 	return ibmvnic_send_crq(adapter, &crq);
3299 }
3300 
3301 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3302 {
3303 	union ibmvnic_crq crq;
3304 
3305 	memset(&crq, 0, sizeof(crq));
3306 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3307 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3308 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3309 
3310 	return ibmvnic_send_crq(adapter, &crq);
3311 }
3312 
3313 struct vnic_login_client_data {
3314 	u8	type;
3315 	__be16	len;
3316 	char	name[];
3317 } __packed;
3318 
3319 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3320 {
3321 	int len;
3322 
3323 	/* Calculate the amount of buffer space needed for the
3324 	 * vnic client data in the login buffer. There are four entries,
3325 	 * OS name, LPAR name, device name, and a null last entry.
3326 	 */
3327 	len = 4 * sizeof(struct vnic_login_client_data);
3328 	len += 6; /* "Linux" plus NULL */
3329 	len += strlen(utsname()->nodename) + 1;
3330 	len += strlen(adapter->netdev->name) + 1;
3331 
3332 	return len;
3333 }
3334 
3335 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3336 				 struct vnic_login_client_data *vlcd)
3337 {
3338 	const char *os_name = "Linux";
3339 	int len;
3340 
3341 	/* Type 1 - LPAR OS */
3342 	vlcd->type = 1;
3343 	len = strlen(os_name) + 1;
3344 	vlcd->len = cpu_to_be16(len);
3345 	strncpy(vlcd->name, os_name, len);
3346 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3347 
3348 	/* Type 2 - LPAR name */
3349 	vlcd->type = 2;
3350 	len = strlen(utsname()->nodename) + 1;
3351 	vlcd->len = cpu_to_be16(len);
3352 	strncpy(vlcd->name, utsname()->nodename, len);
3353 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3354 
3355 	/* Type 3 - device name */
3356 	vlcd->type = 3;
3357 	len = strlen(adapter->netdev->name) + 1;
3358 	vlcd->len = cpu_to_be16(len);
3359 	strncpy(vlcd->name, adapter->netdev->name, len);
3360 }
3361 
3362 static int send_login(struct ibmvnic_adapter *adapter)
3363 {
3364 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3365 	struct ibmvnic_login_buffer *login_buffer;
3366 	struct device *dev = &adapter->vdev->dev;
3367 	dma_addr_t rsp_buffer_token;
3368 	dma_addr_t buffer_token;
3369 	size_t rsp_buffer_size;
3370 	union ibmvnic_crq crq;
3371 	size_t buffer_size;
3372 	__be64 *tx_list_p;
3373 	__be64 *rx_list_p;
3374 	int client_data_len;
3375 	struct vnic_login_client_data *vlcd;
3376 	int i;
3377 
3378 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3379 		netdev_err(adapter->netdev,
3380 			   "RX or TX queues are not allocated, device login failed\n");
3381 		return -1;
3382 	}
3383 
3384 	release_login_rsp_buffer(adapter);
3385 	client_data_len = vnic_client_data_len(adapter);
3386 
3387 	buffer_size =
3388 	    sizeof(struct ibmvnic_login_buffer) +
3389 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3390 	    client_data_len;
3391 
3392 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3393 	if (!login_buffer)
3394 		goto buf_alloc_failed;
3395 
3396 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3397 				      DMA_TO_DEVICE);
3398 	if (dma_mapping_error(dev, buffer_token)) {
3399 		dev_err(dev, "Couldn't map login buffer\n");
3400 		goto buf_map_failed;
3401 	}
3402 
3403 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3404 			  sizeof(u64) * adapter->req_tx_queues +
3405 			  sizeof(u64) * adapter->req_rx_queues +
3406 			  sizeof(u64) * adapter->req_rx_queues +
3407 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3408 
3409 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3410 	if (!login_rsp_buffer)
3411 		goto buf_rsp_alloc_failed;
3412 
3413 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3414 					  rsp_buffer_size, DMA_FROM_DEVICE);
3415 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3416 		dev_err(dev, "Couldn't map login rsp buffer\n");
3417 		goto buf_rsp_map_failed;
3418 	}
3419 
3420 	adapter->login_buf = login_buffer;
3421 	adapter->login_buf_token = buffer_token;
3422 	adapter->login_buf_sz = buffer_size;
3423 	adapter->login_rsp_buf = login_rsp_buffer;
3424 	adapter->login_rsp_buf_token = rsp_buffer_token;
3425 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3426 
3427 	login_buffer->len = cpu_to_be32(buffer_size);
3428 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3429 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3430 	login_buffer->off_txcomp_subcrqs =
3431 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3432 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3433 	login_buffer->off_rxcomp_subcrqs =
3434 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3435 			sizeof(u64) * adapter->req_tx_queues);
3436 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3437 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3438 
3439 	tx_list_p = (__be64 *)((char *)login_buffer +
3440 				      sizeof(struct ibmvnic_login_buffer));
3441 	rx_list_p = (__be64 *)((char *)login_buffer +
3442 				      sizeof(struct ibmvnic_login_buffer) +
3443 				      sizeof(u64) * adapter->req_tx_queues);
3444 
3445 	for (i = 0; i < adapter->req_tx_queues; i++) {
3446 		if (adapter->tx_scrq[i]) {
3447 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3448 						   crq_num);
3449 		}
3450 	}
3451 
3452 	for (i = 0; i < adapter->req_rx_queues; i++) {
3453 		if (adapter->rx_scrq[i]) {
3454 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3455 						   crq_num);
3456 		}
3457 	}
3458 
3459 	/* Insert vNIC login client data */
3460 	vlcd = (struct vnic_login_client_data *)
3461 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3462 	login_buffer->client_data_offset =
3463 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3464 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3465 
3466 	vnic_add_client_data(adapter, vlcd);
3467 
3468 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3469 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3470 		netdev_dbg(adapter->netdev, "%016lx\n",
3471 			   ((unsigned long int *)(adapter->login_buf))[i]);
3472 	}
3473 
3474 	memset(&crq, 0, sizeof(crq));
3475 	crq.login.first = IBMVNIC_CRQ_CMD;
3476 	crq.login.cmd = LOGIN;
3477 	crq.login.ioba = cpu_to_be32(buffer_token);
3478 	crq.login.len = cpu_to_be32(buffer_size);
3479 	ibmvnic_send_crq(adapter, &crq);
3480 
3481 	return 0;
3482 
3483 buf_rsp_map_failed:
3484 	kfree(login_rsp_buffer);
3485 buf_rsp_alloc_failed:
3486 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3487 buf_map_failed:
3488 	kfree(login_buffer);
3489 buf_alloc_failed:
3490 	return -1;
3491 }
3492 
3493 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3494 			    u32 len, u8 map_id)
3495 {
3496 	union ibmvnic_crq crq;
3497 
3498 	memset(&crq, 0, sizeof(crq));
3499 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3500 	crq.request_map.cmd = REQUEST_MAP;
3501 	crq.request_map.map_id = map_id;
3502 	crq.request_map.ioba = cpu_to_be32(addr);
3503 	crq.request_map.len = cpu_to_be32(len);
3504 	return ibmvnic_send_crq(adapter, &crq);
3505 }
3506 
3507 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3508 {
3509 	union ibmvnic_crq crq;
3510 
3511 	memset(&crq, 0, sizeof(crq));
3512 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3513 	crq.request_unmap.cmd = REQUEST_UNMAP;
3514 	crq.request_unmap.map_id = map_id;
3515 	return ibmvnic_send_crq(adapter, &crq);
3516 }
3517 
3518 static void send_map_query(struct ibmvnic_adapter *adapter)
3519 {
3520 	union ibmvnic_crq crq;
3521 
3522 	memset(&crq, 0, sizeof(crq));
3523 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3524 	crq.query_map.cmd = QUERY_MAP;
3525 	ibmvnic_send_crq(adapter, &crq);
3526 }
3527 
3528 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3529 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3530 {
3531 	union ibmvnic_crq crq;
3532 
3533 	atomic_set(&adapter->running_cap_crqs, 0);
3534 	memset(&crq, 0, sizeof(crq));
3535 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3536 	crq.query_capability.cmd = QUERY_CAPABILITY;
3537 
3538 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3539 	atomic_inc(&adapter->running_cap_crqs);
3540 	ibmvnic_send_crq(adapter, &crq);
3541 
3542 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3543 	atomic_inc(&adapter->running_cap_crqs);
3544 	ibmvnic_send_crq(adapter, &crq);
3545 
3546 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3547 	atomic_inc(&adapter->running_cap_crqs);
3548 	ibmvnic_send_crq(adapter, &crq);
3549 
3550 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3551 	atomic_inc(&adapter->running_cap_crqs);
3552 	ibmvnic_send_crq(adapter, &crq);
3553 
3554 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3555 	atomic_inc(&adapter->running_cap_crqs);
3556 	ibmvnic_send_crq(adapter, &crq);
3557 
3558 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3559 	atomic_inc(&adapter->running_cap_crqs);
3560 	ibmvnic_send_crq(adapter, &crq);
3561 
3562 	crq.query_capability.capability =
3563 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3564 	atomic_inc(&adapter->running_cap_crqs);
3565 	ibmvnic_send_crq(adapter, &crq);
3566 
3567 	crq.query_capability.capability =
3568 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3569 	atomic_inc(&adapter->running_cap_crqs);
3570 	ibmvnic_send_crq(adapter, &crq);
3571 
3572 	crq.query_capability.capability =
3573 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3574 	atomic_inc(&adapter->running_cap_crqs);
3575 	ibmvnic_send_crq(adapter, &crq);
3576 
3577 	crq.query_capability.capability =
3578 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3579 	atomic_inc(&adapter->running_cap_crqs);
3580 	ibmvnic_send_crq(adapter, &crq);
3581 
3582 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3583 	atomic_inc(&adapter->running_cap_crqs);
3584 	ibmvnic_send_crq(adapter, &crq);
3585 
3586 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3587 	atomic_inc(&adapter->running_cap_crqs);
3588 	ibmvnic_send_crq(adapter, &crq);
3589 
3590 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3591 	atomic_inc(&adapter->running_cap_crqs);
3592 	ibmvnic_send_crq(adapter, &crq);
3593 
3594 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3595 	atomic_inc(&adapter->running_cap_crqs);
3596 	ibmvnic_send_crq(adapter, &crq);
3597 
3598 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3599 	atomic_inc(&adapter->running_cap_crqs);
3600 	ibmvnic_send_crq(adapter, &crq);
3601 
3602 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3603 	atomic_inc(&adapter->running_cap_crqs);
3604 	ibmvnic_send_crq(adapter, &crq);
3605 
3606 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3607 	atomic_inc(&adapter->running_cap_crqs);
3608 	ibmvnic_send_crq(adapter, &crq);
3609 
3610 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3611 	atomic_inc(&adapter->running_cap_crqs);
3612 	ibmvnic_send_crq(adapter, &crq);
3613 
3614 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3615 	atomic_inc(&adapter->running_cap_crqs);
3616 	ibmvnic_send_crq(adapter, &crq);
3617 
3618 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3619 	atomic_inc(&adapter->running_cap_crqs);
3620 	ibmvnic_send_crq(adapter, &crq);
3621 
3622 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3623 	atomic_inc(&adapter->running_cap_crqs);
3624 	ibmvnic_send_crq(adapter, &crq);
3625 
3626 	crq.query_capability.capability =
3627 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3628 	atomic_inc(&adapter->running_cap_crqs);
3629 	ibmvnic_send_crq(adapter, &crq);
3630 
3631 	crq.query_capability.capability =
3632 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3633 	atomic_inc(&adapter->running_cap_crqs);
3634 	ibmvnic_send_crq(adapter, &crq);
3635 
3636 	crq.query_capability.capability =
3637 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3638 	atomic_inc(&adapter->running_cap_crqs);
3639 	ibmvnic_send_crq(adapter, &crq);
3640 
3641 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3642 	atomic_inc(&adapter->running_cap_crqs);
3643 	ibmvnic_send_crq(adapter, &crq);
3644 }
3645 
3646 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3647 				struct ibmvnic_adapter *adapter)
3648 {
3649 	struct device *dev = &adapter->vdev->dev;
3650 
3651 	if (crq->get_vpd_size_rsp.rc.code) {
3652 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3653 			crq->get_vpd_size_rsp.rc.code);
3654 		complete(&adapter->fw_done);
3655 		return;
3656 	}
3657 
3658 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3659 	complete(&adapter->fw_done);
3660 }
3661 
3662 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3663 			   struct ibmvnic_adapter *adapter)
3664 {
3665 	struct device *dev = &adapter->vdev->dev;
3666 	unsigned char *substr = NULL;
3667 	u8 fw_level_len = 0;
3668 
3669 	memset(adapter->fw_version, 0, 32);
3670 
3671 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3672 			 DMA_FROM_DEVICE);
3673 
3674 	if (crq->get_vpd_rsp.rc.code) {
3675 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3676 			crq->get_vpd_rsp.rc.code);
3677 		goto complete;
3678 	}
3679 
3680 	/* get the position of the firmware version info
3681 	 * located after the ASCII 'RM' substring in the buffer
3682 	 */
3683 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3684 	if (!substr) {
3685 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3686 		goto complete;
3687 	}
3688 
3689 	/* get length of firmware level ASCII substring */
3690 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3691 		fw_level_len = *(substr + 2);
3692 	} else {
3693 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3694 		goto complete;
3695 	}
3696 
3697 	/* copy firmware version string from vpd into adapter */
3698 	if ((substr + 3 + fw_level_len) <
3699 	    (adapter->vpd->buff + adapter->vpd->len)) {
3700 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3701 	} else {
3702 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3703 	}
3704 
3705 complete:
3706 	if (adapter->fw_version[0] == '\0')
3707 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3708 	complete(&adapter->fw_done);
3709 }
3710 
3711 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3712 {
3713 	struct device *dev = &adapter->vdev->dev;
3714 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3715 	union ibmvnic_crq crq;
3716 	int i;
3717 
3718 	dma_unmap_single(dev, adapter->ip_offload_tok,
3719 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3720 
3721 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3722 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3723 		netdev_dbg(adapter->netdev, "%016lx\n",
3724 			   ((unsigned long int *)(buf))[i]);
3725 
3726 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3727 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3728 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3729 		   buf->tcp_ipv4_chksum);
3730 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3731 		   buf->tcp_ipv6_chksum);
3732 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3733 		   buf->udp_ipv4_chksum);
3734 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3735 		   buf->udp_ipv6_chksum);
3736 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3737 		   buf->large_tx_ipv4);
3738 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3739 		   buf->large_tx_ipv6);
3740 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3741 		   buf->large_rx_ipv4);
3742 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3743 		   buf->large_rx_ipv6);
3744 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3745 		   buf->max_ipv4_header_size);
3746 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3747 		   buf->max_ipv6_header_size);
3748 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3749 		   buf->max_tcp_header_size);
3750 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3751 		   buf->max_udp_header_size);
3752 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3753 		   buf->max_large_tx_size);
3754 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3755 		   buf->max_large_rx_size);
3756 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3757 		   buf->ipv6_extension_header);
3758 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3759 		   buf->tcp_pseudosum_req);
3760 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3761 		   buf->num_ipv6_ext_headers);
3762 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3763 		   buf->off_ipv6_ext_headers);
3764 
3765 	adapter->ip_offload_ctrl_tok =
3766 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
3767 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3768 
3769 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3770 		dev_err(dev, "Couldn't map ip offload control buffer\n");
3771 		return;
3772 	}
3773 
3774 	adapter->ip_offload_ctrl.len =
3775 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3776 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3777 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3778 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3779 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3780 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3781 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3782 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3783 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3784 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3785 
3786 	/* large_rx disabled for now, additional features needed */
3787 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3788 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3789 
3790 	adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3791 
3792 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3793 		adapter->netdev->features |= NETIF_F_IP_CSUM;
3794 
3795 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3796 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3797 
3798 	if ((adapter->netdev->features &
3799 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3800 		adapter->netdev->features |= NETIF_F_RXCSUM;
3801 
3802 	if (buf->large_tx_ipv4)
3803 		adapter->netdev->features |= NETIF_F_TSO;
3804 	if (buf->large_tx_ipv6)
3805 		adapter->netdev->features |= NETIF_F_TSO6;
3806 
3807 	adapter->netdev->hw_features |= adapter->netdev->features;
3808 
3809 	memset(&crq, 0, sizeof(crq));
3810 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3811 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3812 	crq.control_ip_offload.len =
3813 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3814 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3815 	ibmvnic_send_crq(adapter, &crq);
3816 }
3817 
3818 static const char *ibmvnic_fw_err_cause(u16 cause)
3819 {
3820 	switch (cause) {
3821 	case ADAPTER_PROBLEM:
3822 		return "adapter problem";
3823 	case BUS_PROBLEM:
3824 		return "bus problem";
3825 	case FW_PROBLEM:
3826 		return "firmware problem";
3827 	case DD_PROBLEM:
3828 		return "device driver problem";
3829 	case EEH_RECOVERY:
3830 		return "EEH recovery";
3831 	case FW_UPDATED:
3832 		return "firmware updated";
3833 	case LOW_MEMORY:
3834 		return "low Memory";
3835 	default:
3836 		return "unknown";
3837 	}
3838 }
3839 
3840 static void handle_error_indication(union ibmvnic_crq *crq,
3841 				    struct ibmvnic_adapter *adapter)
3842 {
3843 	struct device *dev = &adapter->vdev->dev;
3844 	u16 cause;
3845 
3846 	cause = be16_to_cpu(crq->error_indication.error_cause);
3847 
3848 	dev_warn_ratelimited(dev,
3849 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
3850 			     crq->error_indication.flags
3851 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3852 			     ibmvnic_fw_err_cause(cause));
3853 
3854 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3855 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3856 	else
3857 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3858 }
3859 
3860 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3861 				 struct ibmvnic_adapter *adapter)
3862 {
3863 	struct net_device *netdev = adapter->netdev;
3864 	struct device *dev = &adapter->vdev->dev;
3865 	long rc;
3866 
3867 	rc = crq->change_mac_addr_rsp.rc.code;
3868 	if (rc) {
3869 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3870 		goto out;
3871 	}
3872 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3873 	       ETH_ALEN);
3874 out:
3875 	complete(&adapter->fw_done);
3876 	return rc;
3877 }
3878 
3879 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3880 				   struct ibmvnic_adapter *adapter)
3881 {
3882 	struct device *dev = &adapter->vdev->dev;
3883 	u64 *req_value;
3884 	char *name;
3885 
3886 	atomic_dec(&adapter->running_cap_crqs);
3887 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3888 	case REQ_TX_QUEUES:
3889 		req_value = &adapter->req_tx_queues;
3890 		name = "tx";
3891 		break;
3892 	case REQ_RX_QUEUES:
3893 		req_value = &adapter->req_rx_queues;
3894 		name = "rx";
3895 		break;
3896 	case REQ_RX_ADD_QUEUES:
3897 		req_value = &adapter->req_rx_add_queues;
3898 		name = "rx_add";
3899 		break;
3900 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3901 		req_value = &adapter->req_tx_entries_per_subcrq;
3902 		name = "tx_entries_per_subcrq";
3903 		break;
3904 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3905 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3906 		name = "rx_add_entries_per_subcrq";
3907 		break;
3908 	case REQ_MTU:
3909 		req_value = &adapter->req_mtu;
3910 		name = "mtu";
3911 		break;
3912 	case PROMISC_REQUESTED:
3913 		req_value = &adapter->promisc;
3914 		name = "promisc";
3915 		break;
3916 	default:
3917 		dev_err(dev, "Got invalid cap request rsp %d\n",
3918 			crq->request_capability.capability);
3919 		return;
3920 	}
3921 
3922 	switch (crq->request_capability_rsp.rc.code) {
3923 	case SUCCESS:
3924 		break;
3925 	case PARTIALSUCCESS:
3926 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3927 			 *req_value,
3928 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3929 					       number), name);
3930 
3931 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3932 		    REQ_MTU) {
3933 			pr_err("mtu of %llu is not supported. Reverting.\n",
3934 			       *req_value);
3935 			*req_value = adapter->fallback.mtu;
3936 		} else {
3937 			*req_value =
3938 				be64_to_cpu(crq->request_capability_rsp.number);
3939 		}
3940 
3941 		ibmvnic_send_req_caps(adapter, 1);
3942 		return;
3943 	default:
3944 		dev_err(dev, "Error %d in request cap rsp\n",
3945 			crq->request_capability_rsp.rc.code);
3946 		return;
3947 	}
3948 
3949 	/* Done receiving requested capabilities, query IP offload support */
3950 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3951 		union ibmvnic_crq newcrq;
3952 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3953 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3954 		    &adapter->ip_offload_buf;
3955 
3956 		adapter->wait_capability = false;
3957 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3958 							 buf_sz,
3959 							 DMA_FROM_DEVICE);
3960 
3961 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3962 			if (!firmware_has_feature(FW_FEATURE_CMO))
3963 				dev_err(dev, "Couldn't map offload buffer\n");
3964 			return;
3965 		}
3966 
3967 		memset(&newcrq, 0, sizeof(newcrq));
3968 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3969 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3970 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3971 		newcrq.query_ip_offload.ioba =
3972 		    cpu_to_be32(adapter->ip_offload_tok);
3973 
3974 		ibmvnic_send_crq(adapter, &newcrq);
3975 	}
3976 }
3977 
3978 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3979 			    struct ibmvnic_adapter *adapter)
3980 {
3981 	struct device *dev = &adapter->vdev->dev;
3982 	struct net_device *netdev = adapter->netdev;
3983 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3984 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3985 	int i;
3986 
3987 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3988 			 DMA_TO_DEVICE);
3989 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3990 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
3991 
3992 	/* If the number of queues requested can't be allocated by the
3993 	 * server, the login response will return with code 1. We will need
3994 	 * to resend the login buffer with fewer queues requested.
3995 	 */
3996 	if (login_rsp_crq->generic.rc.code) {
3997 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
3998 		complete(&adapter->init_done);
3999 		return 0;
4000 	}
4001 
4002 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4003 
4004 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4005 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4006 		netdev_dbg(adapter->netdev, "%016lx\n",
4007 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4008 	}
4009 
4010 	/* Sanity checks */
4011 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4012 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4013 	     adapter->req_rx_add_queues !=
4014 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4015 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4016 		ibmvnic_remove(adapter->vdev);
4017 		return -EIO;
4018 	}
4019 	release_login_buffer(adapter);
4020 	complete(&adapter->init_done);
4021 
4022 	return 0;
4023 }
4024 
4025 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4026 				     struct ibmvnic_adapter *adapter)
4027 {
4028 	struct device *dev = &adapter->vdev->dev;
4029 	long rc;
4030 
4031 	rc = crq->request_unmap_rsp.rc.code;
4032 	if (rc)
4033 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4034 }
4035 
4036 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4037 				 struct ibmvnic_adapter *adapter)
4038 {
4039 	struct net_device *netdev = adapter->netdev;
4040 	struct device *dev = &adapter->vdev->dev;
4041 	long rc;
4042 
4043 	rc = crq->query_map_rsp.rc.code;
4044 	if (rc) {
4045 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4046 		return;
4047 	}
4048 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4049 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4050 		   crq->query_map_rsp.free_pages);
4051 }
4052 
4053 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4054 				 struct ibmvnic_adapter *adapter)
4055 {
4056 	struct net_device *netdev = adapter->netdev;
4057 	struct device *dev = &adapter->vdev->dev;
4058 	long rc;
4059 
4060 	atomic_dec(&adapter->running_cap_crqs);
4061 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4062 		   atomic_read(&adapter->running_cap_crqs));
4063 	rc = crq->query_capability.rc.code;
4064 	if (rc) {
4065 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4066 		goto out;
4067 	}
4068 
4069 	switch (be16_to_cpu(crq->query_capability.capability)) {
4070 	case MIN_TX_QUEUES:
4071 		adapter->min_tx_queues =
4072 		    be64_to_cpu(crq->query_capability.number);
4073 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4074 			   adapter->min_tx_queues);
4075 		break;
4076 	case MIN_RX_QUEUES:
4077 		adapter->min_rx_queues =
4078 		    be64_to_cpu(crq->query_capability.number);
4079 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4080 			   adapter->min_rx_queues);
4081 		break;
4082 	case MIN_RX_ADD_QUEUES:
4083 		adapter->min_rx_add_queues =
4084 		    be64_to_cpu(crq->query_capability.number);
4085 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4086 			   adapter->min_rx_add_queues);
4087 		break;
4088 	case MAX_TX_QUEUES:
4089 		adapter->max_tx_queues =
4090 		    be64_to_cpu(crq->query_capability.number);
4091 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4092 			   adapter->max_tx_queues);
4093 		break;
4094 	case MAX_RX_QUEUES:
4095 		adapter->max_rx_queues =
4096 		    be64_to_cpu(crq->query_capability.number);
4097 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4098 			   adapter->max_rx_queues);
4099 		break;
4100 	case MAX_RX_ADD_QUEUES:
4101 		adapter->max_rx_add_queues =
4102 		    be64_to_cpu(crq->query_capability.number);
4103 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4104 			   adapter->max_rx_add_queues);
4105 		break;
4106 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4107 		adapter->min_tx_entries_per_subcrq =
4108 		    be64_to_cpu(crq->query_capability.number);
4109 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4110 			   adapter->min_tx_entries_per_subcrq);
4111 		break;
4112 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4113 		adapter->min_rx_add_entries_per_subcrq =
4114 		    be64_to_cpu(crq->query_capability.number);
4115 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4116 			   adapter->min_rx_add_entries_per_subcrq);
4117 		break;
4118 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4119 		adapter->max_tx_entries_per_subcrq =
4120 		    be64_to_cpu(crq->query_capability.number);
4121 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4122 			   adapter->max_tx_entries_per_subcrq);
4123 		break;
4124 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4125 		adapter->max_rx_add_entries_per_subcrq =
4126 		    be64_to_cpu(crq->query_capability.number);
4127 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4128 			   adapter->max_rx_add_entries_per_subcrq);
4129 		break;
4130 	case TCP_IP_OFFLOAD:
4131 		adapter->tcp_ip_offload =
4132 		    be64_to_cpu(crq->query_capability.number);
4133 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4134 			   adapter->tcp_ip_offload);
4135 		break;
4136 	case PROMISC_SUPPORTED:
4137 		adapter->promisc_supported =
4138 		    be64_to_cpu(crq->query_capability.number);
4139 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4140 			   adapter->promisc_supported);
4141 		break;
4142 	case MIN_MTU:
4143 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4144 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4145 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4146 		break;
4147 	case MAX_MTU:
4148 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4149 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4150 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4151 		break;
4152 	case MAX_MULTICAST_FILTERS:
4153 		adapter->max_multicast_filters =
4154 		    be64_to_cpu(crq->query_capability.number);
4155 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4156 			   adapter->max_multicast_filters);
4157 		break;
4158 	case VLAN_HEADER_INSERTION:
4159 		adapter->vlan_header_insertion =
4160 		    be64_to_cpu(crq->query_capability.number);
4161 		if (adapter->vlan_header_insertion)
4162 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4163 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4164 			   adapter->vlan_header_insertion);
4165 		break;
4166 	case RX_VLAN_HEADER_INSERTION:
4167 		adapter->rx_vlan_header_insertion =
4168 		    be64_to_cpu(crq->query_capability.number);
4169 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4170 			   adapter->rx_vlan_header_insertion);
4171 		break;
4172 	case MAX_TX_SG_ENTRIES:
4173 		adapter->max_tx_sg_entries =
4174 		    be64_to_cpu(crq->query_capability.number);
4175 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4176 			   adapter->max_tx_sg_entries);
4177 		break;
4178 	case RX_SG_SUPPORTED:
4179 		adapter->rx_sg_supported =
4180 		    be64_to_cpu(crq->query_capability.number);
4181 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4182 			   adapter->rx_sg_supported);
4183 		break;
4184 	case OPT_TX_COMP_SUB_QUEUES:
4185 		adapter->opt_tx_comp_sub_queues =
4186 		    be64_to_cpu(crq->query_capability.number);
4187 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4188 			   adapter->opt_tx_comp_sub_queues);
4189 		break;
4190 	case OPT_RX_COMP_QUEUES:
4191 		adapter->opt_rx_comp_queues =
4192 		    be64_to_cpu(crq->query_capability.number);
4193 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4194 			   adapter->opt_rx_comp_queues);
4195 		break;
4196 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4197 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4198 		    be64_to_cpu(crq->query_capability.number);
4199 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4200 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4201 		break;
4202 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4203 		adapter->opt_tx_entries_per_subcrq =
4204 		    be64_to_cpu(crq->query_capability.number);
4205 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4206 			   adapter->opt_tx_entries_per_subcrq);
4207 		break;
4208 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4209 		adapter->opt_rxba_entries_per_subcrq =
4210 		    be64_to_cpu(crq->query_capability.number);
4211 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4212 			   adapter->opt_rxba_entries_per_subcrq);
4213 		break;
4214 	case TX_RX_DESC_REQ:
4215 		adapter->tx_rx_desc_req = crq->query_capability.number;
4216 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4217 			   adapter->tx_rx_desc_req);
4218 		break;
4219 
4220 	default:
4221 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4222 			   crq->query_capability.capability);
4223 	}
4224 
4225 out:
4226 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4227 		adapter->wait_capability = false;
4228 		ibmvnic_send_req_caps(adapter, 0);
4229 	}
4230 }
4231 
4232 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4233 			       struct ibmvnic_adapter *adapter)
4234 {
4235 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4236 	struct net_device *netdev = adapter->netdev;
4237 	struct device *dev = &adapter->vdev->dev;
4238 	u64 *u64_crq = (u64 *)crq;
4239 	long rc;
4240 
4241 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4242 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4243 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4244 	switch (gen_crq->first) {
4245 	case IBMVNIC_CRQ_INIT_RSP:
4246 		switch (gen_crq->cmd) {
4247 		case IBMVNIC_CRQ_INIT:
4248 			dev_info(dev, "Partner initialized\n");
4249 			adapter->from_passive_init = true;
4250 			adapter->failover_pending = false;
4251 			if (!completion_done(&adapter->init_done)) {
4252 				complete(&adapter->init_done);
4253 				adapter->init_done_rc = -EIO;
4254 			}
4255 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4256 			break;
4257 		case IBMVNIC_CRQ_INIT_COMPLETE:
4258 			dev_info(dev, "Partner initialization complete\n");
4259 			adapter->crq.active = true;
4260 			send_version_xchg(adapter);
4261 			break;
4262 		default:
4263 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4264 		}
4265 		return;
4266 	case IBMVNIC_CRQ_XPORT_EVENT:
4267 		netif_carrier_off(netdev);
4268 		adapter->crq.active = false;
4269 		if (adapter->resetting)
4270 			adapter->force_reset_recovery = true;
4271 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4272 			dev_info(dev, "Migrated, re-enabling adapter\n");
4273 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4274 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4275 			dev_info(dev, "Backing device failover detected\n");
4276 			adapter->failover_pending = true;
4277 		} else {
4278 			/* The adapter lost the connection */
4279 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4280 				gen_crq->cmd);
4281 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4282 		}
4283 		return;
4284 	case IBMVNIC_CRQ_CMD_RSP:
4285 		break;
4286 	default:
4287 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4288 			gen_crq->first);
4289 		return;
4290 	}
4291 
4292 	switch (gen_crq->cmd) {
4293 	case VERSION_EXCHANGE_RSP:
4294 		rc = crq->version_exchange_rsp.rc.code;
4295 		if (rc) {
4296 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4297 			break;
4298 		}
4299 		dev_info(dev, "Partner protocol version is %d\n",
4300 			 crq->version_exchange_rsp.version);
4301 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4302 		    ibmvnic_version)
4303 			ibmvnic_version =
4304 			    be16_to_cpu(crq->version_exchange_rsp.version);
4305 		send_cap_queries(adapter);
4306 		break;
4307 	case QUERY_CAPABILITY_RSP:
4308 		handle_query_cap_rsp(crq, adapter);
4309 		break;
4310 	case QUERY_MAP_RSP:
4311 		handle_query_map_rsp(crq, adapter);
4312 		break;
4313 	case REQUEST_MAP_RSP:
4314 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4315 		complete(&adapter->fw_done);
4316 		break;
4317 	case REQUEST_UNMAP_RSP:
4318 		handle_request_unmap_rsp(crq, adapter);
4319 		break;
4320 	case REQUEST_CAPABILITY_RSP:
4321 		handle_request_cap_rsp(crq, adapter);
4322 		break;
4323 	case LOGIN_RSP:
4324 		netdev_dbg(netdev, "Got Login Response\n");
4325 		handle_login_rsp(crq, adapter);
4326 		break;
4327 	case LOGICAL_LINK_STATE_RSP:
4328 		netdev_dbg(netdev,
4329 			   "Got Logical Link State Response, state: %d rc: %d\n",
4330 			   crq->logical_link_state_rsp.link_state,
4331 			   crq->logical_link_state_rsp.rc.code);
4332 		adapter->logical_link_state =
4333 		    crq->logical_link_state_rsp.link_state;
4334 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4335 		complete(&adapter->init_done);
4336 		break;
4337 	case LINK_STATE_INDICATION:
4338 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4339 		adapter->phys_link_state =
4340 		    crq->link_state_indication.phys_link_state;
4341 		adapter->logical_link_state =
4342 		    crq->link_state_indication.logical_link_state;
4343 		break;
4344 	case CHANGE_MAC_ADDR_RSP:
4345 		netdev_dbg(netdev, "Got MAC address change Response\n");
4346 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4347 		break;
4348 	case ERROR_INDICATION:
4349 		netdev_dbg(netdev, "Got Error Indication\n");
4350 		handle_error_indication(crq, adapter);
4351 		break;
4352 	case REQUEST_STATISTICS_RSP:
4353 		netdev_dbg(netdev, "Got Statistics Response\n");
4354 		complete(&adapter->stats_done);
4355 		break;
4356 	case QUERY_IP_OFFLOAD_RSP:
4357 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4358 		handle_query_ip_offload_rsp(adapter);
4359 		break;
4360 	case MULTICAST_CTRL_RSP:
4361 		netdev_dbg(netdev, "Got multicast control Response\n");
4362 		break;
4363 	case CONTROL_IP_OFFLOAD_RSP:
4364 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4365 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4366 				 sizeof(adapter->ip_offload_ctrl),
4367 				 DMA_TO_DEVICE);
4368 		complete(&adapter->init_done);
4369 		break;
4370 	case COLLECT_FW_TRACE_RSP:
4371 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4372 		complete(&adapter->fw_done);
4373 		break;
4374 	case GET_VPD_SIZE_RSP:
4375 		handle_vpd_size_rsp(crq, adapter);
4376 		break;
4377 	case GET_VPD_RSP:
4378 		handle_vpd_rsp(crq, adapter);
4379 		break;
4380 	default:
4381 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4382 			   gen_crq->cmd);
4383 	}
4384 }
4385 
4386 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4387 {
4388 	struct ibmvnic_adapter *adapter = instance;
4389 
4390 	tasklet_schedule(&adapter->tasklet);
4391 	return IRQ_HANDLED;
4392 }
4393 
4394 static void ibmvnic_tasklet(void *data)
4395 {
4396 	struct ibmvnic_adapter *adapter = data;
4397 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4398 	union ibmvnic_crq *crq;
4399 	unsigned long flags;
4400 	bool done = false;
4401 
4402 	spin_lock_irqsave(&queue->lock, flags);
4403 	while (!done) {
4404 		/* Pull all the valid messages off the CRQ */
4405 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4406 			ibmvnic_handle_crq(crq, adapter);
4407 			crq->generic.first = 0;
4408 		}
4409 
4410 		/* remain in tasklet until all
4411 		 * capabilities responses are received
4412 		 */
4413 		if (!adapter->wait_capability)
4414 			done = true;
4415 	}
4416 	/* if capabilities CRQ's were sent in this tasklet, the following
4417 	 * tasklet must wait until all responses are received
4418 	 */
4419 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4420 		adapter->wait_capability = true;
4421 	spin_unlock_irqrestore(&queue->lock, flags);
4422 }
4423 
4424 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4425 {
4426 	struct vio_dev *vdev = adapter->vdev;
4427 	int rc;
4428 
4429 	do {
4430 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4431 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4432 
4433 	if (rc)
4434 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4435 
4436 	return rc;
4437 }
4438 
4439 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4440 {
4441 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4442 	struct device *dev = &adapter->vdev->dev;
4443 	struct vio_dev *vdev = adapter->vdev;
4444 	int rc;
4445 
4446 	/* Close the CRQ */
4447 	do {
4448 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4449 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4450 
4451 	/* Clean out the queue */
4452 	memset(crq->msgs, 0, PAGE_SIZE);
4453 	crq->cur = 0;
4454 	crq->active = false;
4455 
4456 	/* And re-open it again */
4457 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4458 				crq->msg_token, PAGE_SIZE);
4459 
4460 	if (rc == H_CLOSED)
4461 		/* Adapter is good, but other end is not ready */
4462 		dev_warn(dev, "Partner adapter not ready\n");
4463 	else if (rc != 0)
4464 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4465 
4466 	return rc;
4467 }
4468 
4469 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4470 {
4471 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4472 	struct vio_dev *vdev = adapter->vdev;
4473 	long rc;
4474 
4475 	if (!crq->msgs)
4476 		return;
4477 
4478 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4479 	free_irq(vdev->irq, adapter);
4480 	tasklet_kill(&adapter->tasklet);
4481 	do {
4482 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4483 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4484 
4485 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4486 			 DMA_BIDIRECTIONAL);
4487 	free_page((unsigned long)crq->msgs);
4488 	crq->msgs = NULL;
4489 	crq->active = false;
4490 }
4491 
4492 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4493 {
4494 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4495 	struct device *dev = &adapter->vdev->dev;
4496 	struct vio_dev *vdev = adapter->vdev;
4497 	int rc, retrc = -ENOMEM;
4498 
4499 	if (crq->msgs)
4500 		return 0;
4501 
4502 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4503 	/* Should we allocate more than one page? */
4504 
4505 	if (!crq->msgs)
4506 		return -ENOMEM;
4507 
4508 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4509 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4510 					DMA_BIDIRECTIONAL);
4511 	if (dma_mapping_error(dev, crq->msg_token))
4512 		goto map_failed;
4513 
4514 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4515 				crq->msg_token, PAGE_SIZE);
4516 
4517 	if (rc == H_RESOURCE)
4518 		/* maybe kexecing and resource is busy. try a reset */
4519 		rc = ibmvnic_reset_crq(adapter);
4520 	retrc = rc;
4521 
4522 	if (rc == H_CLOSED) {
4523 		dev_warn(dev, "Partner adapter not ready\n");
4524 	} else if (rc) {
4525 		dev_warn(dev, "Error %d opening adapter\n", rc);
4526 		goto reg_crq_failed;
4527 	}
4528 
4529 	retrc = 0;
4530 
4531 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4532 		     (unsigned long)adapter);
4533 
4534 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4535 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4536 			 adapter);
4537 	if (rc) {
4538 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4539 			vdev->irq, rc);
4540 		goto req_irq_failed;
4541 	}
4542 
4543 	rc = vio_enable_interrupts(vdev);
4544 	if (rc) {
4545 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4546 		goto req_irq_failed;
4547 	}
4548 
4549 	crq->cur = 0;
4550 	spin_lock_init(&crq->lock);
4551 
4552 	return retrc;
4553 
4554 req_irq_failed:
4555 	tasklet_kill(&adapter->tasklet);
4556 	do {
4557 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4558 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4559 reg_crq_failed:
4560 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4561 map_failed:
4562 	free_page((unsigned long)crq->msgs);
4563 	crq->msgs = NULL;
4564 	return retrc;
4565 }
4566 
4567 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4568 {
4569 	struct device *dev = &adapter->vdev->dev;
4570 	unsigned long timeout = msecs_to_jiffies(30000);
4571 	u64 old_num_rx_queues, old_num_tx_queues;
4572 	int rc;
4573 
4574 	adapter->from_passive_init = false;
4575 
4576 	old_num_rx_queues = adapter->req_rx_queues;
4577 	old_num_tx_queues = adapter->req_tx_queues;
4578 
4579 	init_completion(&adapter->init_done);
4580 	adapter->init_done_rc = 0;
4581 	ibmvnic_send_crq_init(adapter);
4582 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4583 		dev_err(dev, "Initialization sequence timed out\n");
4584 		return -1;
4585 	}
4586 
4587 	if (adapter->init_done_rc) {
4588 		release_crq_queue(adapter);
4589 		return adapter->init_done_rc;
4590 	}
4591 
4592 	if (adapter->from_passive_init) {
4593 		adapter->state = VNIC_OPEN;
4594 		adapter->from_passive_init = false;
4595 		return -1;
4596 	}
4597 
4598 	if (adapter->resetting && !adapter->wait_for_reset &&
4599 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
4600 		if (adapter->req_rx_queues != old_num_rx_queues ||
4601 		    adapter->req_tx_queues != old_num_tx_queues) {
4602 			release_sub_crqs(adapter, 0);
4603 			rc = init_sub_crqs(adapter);
4604 		} else {
4605 			rc = reset_sub_crq_queues(adapter);
4606 		}
4607 	} else {
4608 		rc = init_sub_crqs(adapter);
4609 	}
4610 
4611 	if (rc) {
4612 		dev_err(dev, "Initialization of sub crqs failed\n");
4613 		release_crq_queue(adapter);
4614 		return rc;
4615 	}
4616 
4617 	rc = init_sub_crq_irqs(adapter);
4618 	if (rc) {
4619 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4620 		release_crq_queue(adapter);
4621 	}
4622 
4623 	return rc;
4624 }
4625 
4626 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4627 {
4628 	struct device *dev = &adapter->vdev->dev;
4629 	unsigned long timeout = msecs_to_jiffies(30000);
4630 	int rc;
4631 
4632 	adapter->from_passive_init = false;
4633 
4634 	init_completion(&adapter->init_done);
4635 	adapter->init_done_rc = 0;
4636 	ibmvnic_send_crq_init(adapter);
4637 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4638 		dev_err(dev, "Initialization sequence timed out\n");
4639 		return -1;
4640 	}
4641 
4642 	if (adapter->init_done_rc) {
4643 		release_crq_queue(adapter);
4644 		return adapter->init_done_rc;
4645 	}
4646 
4647 	if (adapter->from_passive_init) {
4648 		adapter->state = VNIC_OPEN;
4649 		adapter->from_passive_init = false;
4650 		return -1;
4651 	}
4652 
4653 	rc = init_sub_crqs(adapter);
4654 	if (rc) {
4655 		dev_err(dev, "Initialization of sub crqs failed\n");
4656 		release_crq_queue(adapter);
4657 		return rc;
4658 	}
4659 
4660 	rc = init_sub_crq_irqs(adapter);
4661 	if (rc) {
4662 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4663 		release_crq_queue(adapter);
4664 	}
4665 
4666 	return rc;
4667 }
4668 
4669 static struct device_attribute dev_attr_failover;
4670 
4671 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4672 {
4673 	struct ibmvnic_adapter *adapter;
4674 	struct net_device *netdev;
4675 	unsigned char *mac_addr_p;
4676 	int rc;
4677 
4678 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4679 		dev->unit_address);
4680 
4681 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4682 							VETH_MAC_ADDR, NULL);
4683 	if (!mac_addr_p) {
4684 		dev_err(&dev->dev,
4685 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4686 			__FILE__, __LINE__);
4687 		return 0;
4688 	}
4689 
4690 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4691 				   IBMVNIC_MAX_QUEUES);
4692 	if (!netdev)
4693 		return -ENOMEM;
4694 
4695 	adapter = netdev_priv(netdev);
4696 	adapter->state = VNIC_PROBING;
4697 	dev_set_drvdata(&dev->dev, netdev);
4698 	adapter->vdev = dev;
4699 	adapter->netdev = netdev;
4700 
4701 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
4702 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4703 	netdev->irq = dev->irq;
4704 	netdev->netdev_ops = &ibmvnic_netdev_ops;
4705 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4706 	SET_NETDEV_DEV(netdev, &dev->dev);
4707 
4708 	spin_lock_init(&adapter->stats_lock);
4709 
4710 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4711 	INIT_LIST_HEAD(&adapter->rwi_list);
4712 	mutex_init(&adapter->reset_lock);
4713 	mutex_init(&adapter->rwi_lock);
4714 	adapter->resetting = false;
4715 
4716 	adapter->mac_change_pending = false;
4717 
4718 	do {
4719 		rc = init_crq_queue(adapter);
4720 		if (rc) {
4721 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4722 				rc);
4723 			goto ibmvnic_init_fail;
4724 		}
4725 
4726 		rc = ibmvnic_init(adapter);
4727 		if (rc && rc != EAGAIN)
4728 			goto ibmvnic_init_fail;
4729 	} while (rc == EAGAIN);
4730 
4731 	rc = init_stats_buffers(adapter);
4732 	if (rc)
4733 		goto ibmvnic_init_fail;
4734 
4735 	rc = init_stats_token(adapter);
4736 	if (rc)
4737 		goto ibmvnic_stats_fail;
4738 
4739 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4740 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4741 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4742 
4743 	rc = device_create_file(&dev->dev, &dev_attr_failover);
4744 	if (rc)
4745 		goto ibmvnic_dev_file_err;
4746 
4747 	netif_carrier_off(netdev);
4748 	rc = register_netdev(netdev);
4749 	if (rc) {
4750 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4751 		goto ibmvnic_register_fail;
4752 	}
4753 	dev_info(&dev->dev, "ibmvnic registered\n");
4754 
4755 	adapter->state = VNIC_PROBED;
4756 
4757 	adapter->wait_for_reset = false;
4758 
4759 	return 0;
4760 
4761 ibmvnic_register_fail:
4762 	device_remove_file(&dev->dev, &dev_attr_failover);
4763 
4764 ibmvnic_dev_file_err:
4765 	release_stats_token(adapter);
4766 
4767 ibmvnic_stats_fail:
4768 	release_stats_buffers(adapter);
4769 
4770 ibmvnic_init_fail:
4771 	release_sub_crqs(adapter, 1);
4772 	release_crq_queue(adapter);
4773 	free_netdev(netdev);
4774 
4775 	return rc;
4776 }
4777 
4778 static int ibmvnic_remove(struct vio_dev *dev)
4779 {
4780 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
4781 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4782 
4783 	adapter->state = VNIC_REMOVING;
4784 	unregister_netdev(netdev);
4785 	mutex_lock(&adapter->reset_lock);
4786 
4787 	release_resources(adapter);
4788 	release_sub_crqs(adapter, 1);
4789 	release_crq_queue(adapter);
4790 
4791 	release_stats_token(adapter);
4792 	release_stats_buffers(adapter);
4793 
4794 	adapter->state = VNIC_REMOVED;
4795 
4796 	mutex_unlock(&adapter->reset_lock);
4797 	device_remove_file(&dev->dev, &dev_attr_failover);
4798 	free_netdev(netdev);
4799 	dev_set_drvdata(&dev->dev, NULL);
4800 
4801 	return 0;
4802 }
4803 
4804 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4805 			      const char *buf, size_t count)
4806 {
4807 	struct net_device *netdev = dev_get_drvdata(dev);
4808 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4809 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4810 	__be64 session_token;
4811 	long rc;
4812 
4813 	if (!sysfs_streq(buf, "1"))
4814 		return -EINVAL;
4815 
4816 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4817 			 H_GET_SESSION_TOKEN, 0, 0, 0);
4818 	if (rc) {
4819 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4820 			   rc);
4821 		return -EINVAL;
4822 	}
4823 
4824 	session_token = (__be64)retbuf[0];
4825 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4826 		   be64_to_cpu(session_token));
4827 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4828 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
4829 	if (rc) {
4830 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4831 			   rc);
4832 		return -EINVAL;
4833 	}
4834 
4835 	return count;
4836 }
4837 
4838 static DEVICE_ATTR_WO(failover);
4839 
4840 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4841 {
4842 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4843 	struct ibmvnic_adapter *adapter;
4844 	struct iommu_table *tbl;
4845 	unsigned long ret = 0;
4846 	int i;
4847 
4848 	tbl = get_iommu_table_base(&vdev->dev);
4849 
4850 	/* netdev inits at probe time along with the structures we need below*/
4851 	if (!netdev)
4852 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4853 
4854 	adapter = netdev_priv(netdev);
4855 
4856 	ret += PAGE_SIZE; /* the crq message queue */
4857 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4858 
4859 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4860 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4861 
4862 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4863 	     i++)
4864 		ret += adapter->rx_pool[i].size *
4865 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4866 
4867 	return ret;
4868 }
4869 
4870 static int ibmvnic_resume(struct device *dev)
4871 {
4872 	struct net_device *netdev = dev_get_drvdata(dev);
4873 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4874 
4875 	if (adapter->state != VNIC_OPEN)
4876 		return 0;
4877 
4878 	tasklet_schedule(&adapter->tasklet);
4879 
4880 	return 0;
4881 }
4882 
4883 static const struct vio_device_id ibmvnic_device_table[] = {
4884 	{"network", "IBM,vnic"},
4885 	{"", "" }
4886 };
4887 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4888 
4889 static const struct dev_pm_ops ibmvnic_pm_ops = {
4890 	.resume = ibmvnic_resume
4891 };
4892 
4893 static struct vio_driver ibmvnic_driver = {
4894 	.id_table       = ibmvnic_device_table,
4895 	.probe          = ibmvnic_probe,
4896 	.remove         = ibmvnic_remove,
4897 	.get_desired_dma = ibmvnic_get_desired_dma,
4898 	.name		= ibmvnic_driver_name,
4899 	.pm		= &ibmvnic_pm_ops,
4900 };
4901 
4902 /* module functions */
4903 static int __init ibmvnic_module_init(void)
4904 {
4905 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4906 		IBMVNIC_DRIVER_VERSION);
4907 
4908 	return vio_register_driver(&ibmvnic_driver);
4909 }
4910 
4911 static void __exit ibmvnic_module_exit(void)
4912 {
4913 	vio_unregister_driver(&ibmvnic_driver);
4914 }
4915 
4916 module_init(ibmvnic_module_init);
4917 module_exit(ibmvnic_module_exit);
4918