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