xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision da2ef666)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/if_arp.h>
63 #include <linux/in.h>
64 #include <linux/ip.h>
65 #include <linux/ipv6.h>
66 #include <linux/irq.h>
67 #include <linux/kthread.h>
68 #include <linux/seq_file.h>
69 #include <linux/interrupt.h>
70 #include <net/net_namespace.h>
71 #include <asm/hvcall.h>
72 #include <linux/atomic.h>
73 #include <asm/vio.h>
74 #include <asm/iommu.h>
75 #include <linux/uaccess.h>
76 #include <asm/firmware.h>
77 #include <linux/workqueue.h>
78 #include <linux/if_vlan.h>
79 #include <linux/utsname.h>
80 
81 #include "ibmvnic.h"
82 
83 static const char ibmvnic_driver_name[] = "ibmvnic";
84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
85 
86 MODULE_AUTHOR("Santiago Leon");
87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
88 MODULE_LICENSE("GPL");
89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
90 
91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
92 static int ibmvnic_remove(struct vio_dev *);
93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
99 		       union sub_crq *sub_crq);
100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
102 static int enable_scrq_irq(struct ibmvnic_adapter *,
103 			   struct ibmvnic_sub_crq_queue *);
104 static int disable_scrq_irq(struct ibmvnic_adapter *,
105 			    struct ibmvnic_sub_crq_queue *);
106 static int pending_scrq(struct ibmvnic_adapter *,
107 			struct ibmvnic_sub_crq_queue *);
108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
109 					struct ibmvnic_sub_crq_queue *);
110 static int ibmvnic_poll(struct napi_struct *napi, int data);
111 static void send_map_query(struct ibmvnic_adapter *adapter);
112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
114 static int send_login(struct ibmvnic_adapter *adapter);
115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
116 static int init_sub_crqs(struct ibmvnic_adapter *);
117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
118 static int ibmvnic_init(struct ibmvnic_adapter *);
119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
120 static void release_crq_queue(struct ibmvnic_adapter *);
121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
123 
124 struct ibmvnic_stat {
125 	char name[ETH_GSTRING_LEN];
126 	int offset;
127 };
128 
129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
130 			     offsetof(struct ibmvnic_statistics, stat))
131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
132 
133 static const struct ibmvnic_stat ibmvnic_stats[] = {
134 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
135 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
136 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
137 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
138 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
139 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
140 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
141 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
142 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
143 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
144 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
145 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
146 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
147 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
148 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
149 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
150 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
151 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
152 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
153 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
154 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
155 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
156 };
157 
158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
159 			  unsigned long length, unsigned long *number,
160 			  unsigned long *irq)
161 {
162 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
163 	long rc;
164 
165 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
166 	*number = retbuf[0];
167 	*irq = retbuf[1];
168 
169 	return rc;
170 }
171 
172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
173 				struct ibmvnic_long_term_buff *ltb, int size)
174 {
175 	struct device *dev = &adapter->vdev->dev;
176 	int rc;
177 
178 	ltb->size = size;
179 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
180 				       GFP_KERNEL);
181 
182 	if (!ltb->buff) {
183 		dev_err(dev, "Couldn't alloc long term buffer\n");
184 		return -ENOMEM;
185 	}
186 	ltb->map_id = adapter->map_id;
187 	adapter->map_id++;
188 
189 	init_completion(&adapter->fw_done);
190 	rc = send_request_map(adapter, ltb->addr,
191 			      ltb->size, ltb->map_id);
192 	if (rc) {
193 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
194 		return rc;
195 	}
196 	wait_for_completion(&adapter->fw_done);
197 
198 	if (adapter->fw_done_rc) {
199 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
200 			adapter->fw_done_rc);
201 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
202 		return -1;
203 	}
204 	return 0;
205 }
206 
207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
208 				struct ibmvnic_long_term_buff *ltb)
209 {
210 	struct device *dev = &adapter->vdev->dev;
211 
212 	if (!ltb->buff)
213 		return;
214 
215 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
216 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
217 		send_request_unmap(adapter, ltb->map_id);
218 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
219 }
220 
221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
222 				struct ibmvnic_long_term_buff *ltb)
223 {
224 	int rc;
225 
226 	memset(ltb->buff, 0, ltb->size);
227 
228 	init_completion(&adapter->fw_done);
229 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
230 	if (rc)
231 		return rc;
232 	wait_for_completion(&adapter->fw_done);
233 
234 	if (adapter->fw_done_rc) {
235 		dev_info(&adapter->vdev->dev,
236 			 "Reset failed, attempting to free and reallocate buffer\n");
237 		free_long_term_buff(adapter, ltb);
238 		return alloc_long_term_buff(adapter, ltb, ltb->size);
239 	}
240 	return 0;
241 }
242 
243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
244 {
245 	int i;
246 
247 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
248 	     i++)
249 		adapter->rx_pool[i].active = 0;
250 }
251 
252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
253 			      struct ibmvnic_rx_pool *pool)
254 {
255 	int count = pool->size - atomic_read(&pool->available);
256 	struct device *dev = &adapter->vdev->dev;
257 	int buffers_added = 0;
258 	unsigned long lpar_rc;
259 	union sub_crq sub_crq;
260 	struct sk_buff *skb;
261 	unsigned int offset;
262 	dma_addr_t dma_addr;
263 	unsigned char *dst;
264 	u64 *handle_array;
265 	int shift = 0;
266 	int index;
267 	int i;
268 
269 	if (!pool->active)
270 		return;
271 
272 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
273 				      be32_to_cpu(adapter->login_rsp_buf->
274 				      off_rxadd_subcrqs));
275 
276 	for (i = 0; i < count; ++i) {
277 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
278 		if (!skb) {
279 			dev_err(dev, "Couldn't replenish rx buff\n");
280 			adapter->replenish_no_mem++;
281 			break;
282 		}
283 
284 		index = pool->free_map[pool->next_free];
285 
286 		if (pool->rx_buff[index].skb)
287 			dev_err(dev, "Inconsistent free_map!\n");
288 
289 		/* Copy the skb to the long term mapped DMA buffer */
290 		offset = index * pool->buff_size;
291 		dst = pool->long_term_buff.buff + offset;
292 		memset(dst, 0, pool->buff_size);
293 		dma_addr = pool->long_term_buff.addr + offset;
294 		pool->rx_buff[index].data = dst;
295 
296 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
297 		pool->rx_buff[index].dma = dma_addr;
298 		pool->rx_buff[index].skb = skb;
299 		pool->rx_buff[index].pool_index = pool->index;
300 		pool->rx_buff[index].size = pool->buff_size;
301 
302 		memset(&sub_crq, 0, sizeof(sub_crq));
303 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
304 		sub_crq.rx_add.correlator =
305 		    cpu_to_be64((u64)&pool->rx_buff[index]);
306 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
307 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
308 
309 		/* The length field of the sCRQ is defined to be 24 bits so the
310 		 * buffer size needs to be left shifted by a byte before it is
311 		 * converted to big endian to prevent the last byte from being
312 		 * truncated.
313 		 */
314 #ifdef __LITTLE_ENDIAN__
315 		shift = 8;
316 #endif
317 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
318 
319 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
320 				      &sub_crq);
321 		if (lpar_rc != H_SUCCESS)
322 			goto failure;
323 
324 		buffers_added++;
325 		adapter->replenish_add_buff_success++;
326 		pool->next_free = (pool->next_free + 1) % pool->size;
327 	}
328 	atomic_add(buffers_added, &pool->available);
329 	return;
330 
331 failure:
332 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
333 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
334 	pool->free_map[pool->next_free] = index;
335 	pool->rx_buff[index].skb = NULL;
336 
337 	dev_kfree_skb_any(skb);
338 	adapter->replenish_add_buff_failure++;
339 	atomic_add(buffers_added, &pool->available);
340 
341 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
342 		/* Disable buffer pool replenishment and report carrier off if
343 		 * queue is closed or pending failover.
344 		 * Firmware guarantees that a signal will be sent to the
345 		 * driver, triggering a reset.
346 		 */
347 		deactivate_rx_pools(adapter);
348 		netif_carrier_off(adapter->netdev);
349 	}
350 }
351 
352 static void replenish_pools(struct ibmvnic_adapter *adapter)
353 {
354 	int i;
355 
356 	adapter->replenish_task_cycles++;
357 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
358 	     i++) {
359 		if (adapter->rx_pool[i].active)
360 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
361 	}
362 }
363 
364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
365 {
366 	kfree(adapter->tx_stats_buffers);
367 	kfree(adapter->rx_stats_buffers);
368 	adapter->tx_stats_buffers = NULL;
369 	adapter->rx_stats_buffers = NULL;
370 }
371 
372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
373 {
374 	adapter->tx_stats_buffers =
375 				kcalloc(IBMVNIC_MAX_QUEUES,
376 					sizeof(struct ibmvnic_tx_queue_stats),
377 					GFP_KERNEL);
378 	if (!adapter->tx_stats_buffers)
379 		return -ENOMEM;
380 
381 	adapter->rx_stats_buffers =
382 				kcalloc(IBMVNIC_MAX_QUEUES,
383 					sizeof(struct ibmvnic_rx_queue_stats),
384 					GFP_KERNEL);
385 	if (!adapter->rx_stats_buffers)
386 		return -ENOMEM;
387 
388 	return 0;
389 }
390 
391 static void release_stats_token(struct ibmvnic_adapter *adapter)
392 {
393 	struct device *dev = &adapter->vdev->dev;
394 
395 	if (!adapter->stats_token)
396 		return;
397 
398 	dma_unmap_single(dev, adapter->stats_token,
399 			 sizeof(struct ibmvnic_statistics),
400 			 DMA_FROM_DEVICE);
401 	adapter->stats_token = 0;
402 }
403 
404 static int init_stats_token(struct ibmvnic_adapter *adapter)
405 {
406 	struct device *dev = &adapter->vdev->dev;
407 	dma_addr_t stok;
408 
409 	stok = dma_map_single(dev, &adapter->stats,
410 			      sizeof(struct ibmvnic_statistics),
411 			      DMA_FROM_DEVICE);
412 	if (dma_mapping_error(dev, stok)) {
413 		dev_err(dev, "Couldn't map stats buffer\n");
414 		return -1;
415 	}
416 
417 	adapter->stats_token = stok;
418 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
419 	return 0;
420 }
421 
422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
423 {
424 	struct ibmvnic_rx_pool *rx_pool;
425 	int rx_scrqs;
426 	int i, j, rc;
427 	u64 *size_array;
428 
429 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
430 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
431 
432 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
433 	for (i = 0; i < rx_scrqs; i++) {
434 		rx_pool = &adapter->rx_pool[i];
435 
436 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
437 
438 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
439 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
440 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
441 			alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
442 					     rx_pool->size *
443 					     rx_pool->buff_size);
444 		} else {
445 			rc = reset_long_term_buff(adapter,
446 						  &rx_pool->long_term_buff);
447 		}
448 
449 		if (rc)
450 			return rc;
451 
452 		for (j = 0; j < rx_pool->size; j++)
453 			rx_pool->free_map[j] = j;
454 
455 		memset(rx_pool->rx_buff, 0,
456 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
457 
458 		atomic_set(&rx_pool->available, 0);
459 		rx_pool->next_alloc = 0;
460 		rx_pool->next_free = 0;
461 		rx_pool->active = 1;
462 	}
463 
464 	return 0;
465 }
466 
467 static void release_rx_pools(struct ibmvnic_adapter *adapter)
468 {
469 	struct ibmvnic_rx_pool *rx_pool;
470 	int i, j;
471 
472 	if (!adapter->rx_pool)
473 		return;
474 
475 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
476 		rx_pool = &adapter->rx_pool[i];
477 
478 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
479 
480 		kfree(rx_pool->free_map);
481 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
482 
483 		if (!rx_pool->rx_buff)
484 			continue;
485 
486 		for (j = 0; j < rx_pool->size; j++) {
487 			if (rx_pool->rx_buff[j].skb) {
488 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
489 				rx_pool->rx_buff[i].skb = NULL;
490 			}
491 		}
492 
493 		kfree(rx_pool->rx_buff);
494 	}
495 
496 	kfree(adapter->rx_pool);
497 	adapter->rx_pool = NULL;
498 	adapter->num_active_rx_pools = 0;
499 }
500 
501 static int init_rx_pools(struct net_device *netdev)
502 {
503 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
504 	struct device *dev = &adapter->vdev->dev;
505 	struct ibmvnic_rx_pool *rx_pool;
506 	int rxadd_subcrqs;
507 	u64 *size_array;
508 	int i, j;
509 
510 	rxadd_subcrqs =
511 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
512 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
513 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
514 
515 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
516 				   sizeof(struct ibmvnic_rx_pool),
517 				   GFP_KERNEL);
518 	if (!adapter->rx_pool) {
519 		dev_err(dev, "Failed to allocate rx pools\n");
520 		return -1;
521 	}
522 
523 	adapter->num_active_rx_pools = rxadd_subcrqs;
524 
525 	for (i = 0; i < rxadd_subcrqs; i++) {
526 		rx_pool = &adapter->rx_pool[i];
527 
528 		netdev_dbg(adapter->netdev,
529 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
530 			   i, adapter->req_rx_add_entries_per_subcrq,
531 			   be64_to_cpu(size_array[i]));
532 
533 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
534 		rx_pool->index = i;
535 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
536 		rx_pool->active = 1;
537 
538 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
539 					    GFP_KERNEL);
540 		if (!rx_pool->free_map) {
541 			release_rx_pools(adapter);
542 			return -1;
543 		}
544 
545 		rx_pool->rx_buff = kcalloc(rx_pool->size,
546 					   sizeof(struct ibmvnic_rx_buff),
547 					   GFP_KERNEL);
548 		if (!rx_pool->rx_buff) {
549 			dev_err(dev, "Couldn't alloc rx buffers\n");
550 			release_rx_pools(adapter);
551 			return -1;
552 		}
553 
554 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
555 					 rx_pool->size * rx_pool->buff_size)) {
556 			release_rx_pools(adapter);
557 			return -1;
558 		}
559 
560 		for (j = 0; j < rx_pool->size; ++j)
561 			rx_pool->free_map[j] = j;
562 
563 		atomic_set(&rx_pool->available, 0);
564 		rx_pool->next_alloc = 0;
565 		rx_pool->next_free = 0;
566 	}
567 
568 	return 0;
569 }
570 
571 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
572 			     struct ibmvnic_tx_pool *tx_pool)
573 {
574 	int rc, i;
575 
576 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
577 	if (rc)
578 		return rc;
579 
580 	memset(tx_pool->tx_buff, 0,
581 	       tx_pool->num_buffers *
582 	       sizeof(struct ibmvnic_tx_buff));
583 
584 	for (i = 0; i < tx_pool->num_buffers; i++)
585 		tx_pool->free_map[i] = i;
586 
587 	tx_pool->consumer_index = 0;
588 	tx_pool->producer_index = 0;
589 
590 	return 0;
591 }
592 
593 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
594 {
595 	int tx_scrqs;
596 	int i, rc;
597 
598 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
599 	for (i = 0; i < tx_scrqs; i++) {
600 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
601 		if (rc)
602 			return rc;
603 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
604 		if (rc)
605 			return rc;
606 	}
607 
608 	return 0;
609 }
610 
611 static void release_vpd_data(struct ibmvnic_adapter *adapter)
612 {
613 	if (!adapter->vpd)
614 		return;
615 
616 	kfree(adapter->vpd->buff);
617 	kfree(adapter->vpd);
618 
619 	adapter->vpd = NULL;
620 }
621 
622 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
623 				struct ibmvnic_tx_pool *tx_pool)
624 {
625 	kfree(tx_pool->tx_buff);
626 	kfree(tx_pool->free_map);
627 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
628 }
629 
630 static void release_tx_pools(struct ibmvnic_adapter *adapter)
631 {
632 	int i;
633 
634 	if (!adapter->tx_pool)
635 		return;
636 
637 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
638 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
639 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
640 	}
641 
642 	kfree(adapter->tx_pool);
643 	adapter->tx_pool = NULL;
644 	kfree(adapter->tso_pool);
645 	adapter->tso_pool = NULL;
646 	adapter->num_active_tx_pools = 0;
647 }
648 
649 static int init_one_tx_pool(struct net_device *netdev,
650 			    struct ibmvnic_tx_pool *tx_pool,
651 			    int num_entries, int buf_size)
652 {
653 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
654 	int i;
655 
656 	tx_pool->tx_buff = kcalloc(num_entries,
657 				   sizeof(struct ibmvnic_tx_buff),
658 				   GFP_KERNEL);
659 	if (!tx_pool->tx_buff)
660 		return -1;
661 
662 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
663 				 num_entries * buf_size))
664 		return -1;
665 
666 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
667 	if (!tx_pool->free_map)
668 		return -1;
669 
670 	for (i = 0; i < num_entries; i++)
671 		tx_pool->free_map[i] = i;
672 
673 	tx_pool->consumer_index = 0;
674 	tx_pool->producer_index = 0;
675 	tx_pool->num_buffers = num_entries;
676 	tx_pool->buf_size = buf_size;
677 
678 	return 0;
679 }
680 
681 static int init_tx_pools(struct net_device *netdev)
682 {
683 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
684 	int tx_subcrqs;
685 	int i, rc;
686 
687 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
688 	adapter->tx_pool = kcalloc(tx_subcrqs,
689 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
690 	if (!adapter->tx_pool)
691 		return -1;
692 
693 	adapter->tso_pool = kcalloc(tx_subcrqs,
694 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
695 	if (!adapter->tso_pool)
696 		return -1;
697 
698 	adapter->num_active_tx_pools = tx_subcrqs;
699 
700 	for (i = 0; i < tx_subcrqs; i++) {
701 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
702 				      adapter->req_tx_entries_per_subcrq,
703 				      adapter->req_mtu + VLAN_HLEN);
704 		if (rc) {
705 			release_tx_pools(adapter);
706 			return rc;
707 		}
708 
709 		init_one_tx_pool(netdev, &adapter->tso_pool[i],
710 				 IBMVNIC_TSO_BUFS,
711 				 IBMVNIC_TSO_BUF_SZ);
712 		if (rc) {
713 			release_tx_pools(adapter);
714 			return rc;
715 		}
716 	}
717 
718 	return 0;
719 }
720 
721 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
722 {
723 	int i;
724 
725 	if (adapter->napi_enabled)
726 		return;
727 
728 	for (i = 0; i < adapter->req_rx_queues; i++)
729 		napi_enable(&adapter->napi[i]);
730 
731 	adapter->napi_enabled = true;
732 }
733 
734 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
735 {
736 	int i;
737 
738 	if (!adapter->napi_enabled)
739 		return;
740 
741 	for (i = 0; i < adapter->req_rx_queues; i++) {
742 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
743 		napi_disable(&adapter->napi[i]);
744 	}
745 
746 	adapter->napi_enabled = false;
747 }
748 
749 static int init_napi(struct ibmvnic_adapter *adapter)
750 {
751 	int i;
752 
753 	adapter->napi = kcalloc(adapter->req_rx_queues,
754 				sizeof(struct napi_struct), GFP_KERNEL);
755 	if (!adapter->napi)
756 		return -ENOMEM;
757 
758 	for (i = 0; i < adapter->req_rx_queues; i++) {
759 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
760 		netif_napi_add(adapter->netdev, &adapter->napi[i],
761 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
762 	}
763 
764 	adapter->num_active_rx_napi = adapter->req_rx_queues;
765 	return 0;
766 }
767 
768 static void release_napi(struct ibmvnic_adapter *adapter)
769 {
770 	int i;
771 
772 	if (!adapter->napi)
773 		return;
774 
775 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
776 		if (&adapter->napi[i]) {
777 			netdev_dbg(adapter->netdev,
778 				   "Releasing napi[%d]\n", i);
779 			netif_napi_del(&adapter->napi[i]);
780 		}
781 	}
782 
783 	kfree(adapter->napi);
784 	adapter->napi = NULL;
785 	adapter->num_active_rx_napi = 0;
786 	adapter->napi_enabled = false;
787 }
788 
789 static int ibmvnic_login(struct net_device *netdev)
790 {
791 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
792 	unsigned long timeout = msecs_to_jiffies(30000);
793 	int retry_count = 0;
794 	bool retry;
795 	int rc;
796 
797 	do {
798 		retry = false;
799 		if (retry_count > IBMVNIC_MAX_QUEUES) {
800 			netdev_warn(netdev, "Login attempts exceeded\n");
801 			return -1;
802 		}
803 
804 		adapter->init_done_rc = 0;
805 		reinit_completion(&adapter->init_done);
806 		rc = send_login(adapter);
807 		if (rc) {
808 			netdev_warn(netdev, "Unable to login\n");
809 			return rc;
810 		}
811 
812 		if (!wait_for_completion_timeout(&adapter->init_done,
813 						 timeout)) {
814 			netdev_warn(netdev, "Login timed out\n");
815 			return -1;
816 		}
817 
818 		if (adapter->init_done_rc == PARTIALSUCCESS) {
819 			retry_count++;
820 			release_sub_crqs(adapter, 1);
821 
822 			retry = true;
823 			netdev_dbg(netdev,
824 				   "Received partial success, retrying...\n");
825 			adapter->init_done_rc = 0;
826 			reinit_completion(&adapter->init_done);
827 			send_cap_queries(adapter);
828 			if (!wait_for_completion_timeout(&adapter->init_done,
829 							 timeout)) {
830 				netdev_warn(netdev,
831 					    "Capabilities query timed out\n");
832 				return -1;
833 			}
834 
835 			rc = init_sub_crqs(adapter);
836 			if (rc) {
837 				netdev_warn(netdev,
838 					    "SCRQ initialization failed\n");
839 				return -1;
840 			}
841 
842 			rc = init_sub_crq_irqs(adapter);
843 			if (rc) {
844 				netdev_warn(netdev,
845 					    "SCRQ irq initialization failed\n");
846 				return -1;
847 			}
848 		} else if (adapter->init_done_rc) {
849 			netdev_warn(netdev, "Adapter login failed\n");
850 			return -1;
851 		}
852 	} while (retry);
853 
854 	/* handle pending MAC address changes after successful login */
855 	if (adapter->mac_change_pending) {
856 		__ibmvnic_set_mac(netdev, &adapter->desired.mac);
857 		adapter->mac_change_pending = false;
858 	}
859 
860 	return 0;
861 }
862 
863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
864 {
865 	kfree(adapter->login_buf);
866 	adapter->login_buf = NULL;
867 }
868 
869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
870 {
871 	kfree(adapter->login_rsp_buf);
872 	adapter->login_rsp_buf = NULL;
873 }
874 
875 static void release_resources(struct ibmvnic_adapter *adapter)
876 {
877 	release_vpd_data(adapter);
878 
879 	release_tx_pools(adapter);
880 	release_rx_pools(adapter);
881 
882 	release_napi(adapter);
883 	release_login_rsp_buffer(adapter);
884 }
885 
886 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
887 {
888 	struct net_device *netdev = adapter->netdev;
889 	unsigned long timeout = msecs_to_jiffies(30000);
890 	union ibmvnic_crq crq;
891 	bool resend;
892 	int rc;
893 
894 	netdev_dbg(netdev, "setting link state %d\n", link_state);
895 
896 	memset(&crq, 0, sizeof(crq));
897 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
898 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
899 	crq.logical_link_state.link_state = link_state;
900 
901 	do {
902 		resend = false;
903 
904 		reinit_completion(&adapter->init_done);
905 		rc = ibmvnic_send_crq(adapter, &crq);
906 		if (rc) {
907 			netdev_err(netdev, "Failed to set link state\n");
908 			return rc;
909 		}
910 
911 		if (!wait_for_completion_timeout(&adapter->init_done,
912 						 timeout)) {
913 			netdev_err(netdev, "timeout setting link state\n");
914 			return -1;
915 		}
916 
917 		if (adapter->init_done_rc == 1) {
918 			/* Partuial success, delay and re-send */
919 			mdelay(1000);
920 			resend = true;
921 		} else if (adapter->init_done_rc) {
922 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
923 				    adapter->init_done_rc);
924 			return adapter->init_done_rc;
925 		}
926 	} while (resend);
927 
928 	return 0;
929 }
930 
931 static int set_real_num_queues(struct net_device *netdev)
932 {
933 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
934 	int rc;
935 
936 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
937 		   adapter->req_tx_queues, adapter->req_rx_queues);
938 
939 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
940 	if (rc) {
941 		netdev_err(netdev, "failed to set the number of tx queues\n");
942 		return rc;
943 	}
944 
945 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
946 	if (rc)
947 		netdev_err(netdev, "failed to set the number of rx queues\n");
948 
949 	return rc;
950 }
951 
952 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
953 {
954 	struct device *dev = &adapter->vdev->dev;
955 	union ibmvnic_crq crq;
956 	int len = 0;
957 	int rc;
958 
959 	if (adapter->vpd->buff)
960 		len = adapter->vpd->len;
961 
962 	init_completion(&adapter->fw_done);
963 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
964 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
965 	rc = ibmvnic_send_crq(adapter, &crq);
966 	if (rc)
967 		return rc;
968 	wait_for_completion(&adapter->fw_done);
969 
970 	if (!adapter->vpd->len)
971 		return -ENODATA;
972 
973 	if (!adapter->vpd->buff)
974 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
975 	else if (adapter->vpd->len != len)
976 		adapter->vpd->buff =
977 			krealloc(adapter->vpd->buff,
978 				 adapter->vpd->len, GFP_KERNEL);
979 
980 	if (!adapter->vpd->buff) {
981 		dev_err(dev, "Could allocate VPD buffer\n");
982 		return -ENOMEM;
983 	}
984 
985 	adapter->vpd->dma_addr =
986 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
987 			       DMA_FROM_DEVICE);
988 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
989 		dev_err(dev, "Could not map VPD buffer\n");
990 		kfree(adapter->vpd->buff);
991 		adapter->vpd->buff = NULL;
992 		return -ENOMEM;
993 	}
994 
995 	reinit_completion(&adapter->fw_done);
996 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
997 	crq.get_vpd.cmd = GET_VPD;
998 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
999 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1000 	rc = ibmvnic_send_crq(adapter, &crq);
1001 	if (rc) {
1002 		kfree(adapter->vpd->buff);
1003 		adapter->vpd->buff = NULL;
1004 		return rc;
1005 	}
1006 	wait_for_completion(&adapter->fw_done);
1007 
1008 	return 0;
1009 }
1010 
1011 static int init_resources(struct ibmvnic_adapter *adapter)
1012 {
1013 	struct net_device *netdev = adapter->netdev;
1014 	int rc;
1015 
1016 	rc = set_real_num_queues(netdev);
1017 	if (rc)
1018 		return rc;
1019 
1020 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1021 	if (!adapter->vpd)
1022 		return -ENOMEM;
1023 
1024 	/* Vital Product Data (VPD) */
1025 	rc = ibmvnic_get_vpd(adapter);
1026 	if (rc) {
1027 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1028 		return rc;
1029 	}
1030 
1031 	adapter->map_id = 1;
1032 
1033 	rc = init_napi(adapter);
1034 	if (rc)
1035 		return rc;
1036 
1037 	send_map_query(adapter);
1038 
1039 	rc = init_rx_pools(netdev);
1040 	if (rc)
1041 		return rc;
1042 
1043 	rc = init_tx_pools(netdev);
1044 	return rc;
1045 }
1046 
1047 static int __ibmvnic_open(struct net_device *netdev)
1048 {
1049 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1050 	enum vnic_state prev_state = adapter->state;
1051 	int i, rc;
1052 
1053 	adapter->state = VNIC_OPENING;
1054 	replenish_pools(adapter);
1055 	ibmvnic_napi_enable(adapter);
1056 
1057 	/* We're ready to receive frames, enable the sub-crq interrupts and
1058 	 * set the logical link state to up
1059 	 */
1060 	for (i = 0; i < adapter->req_rx_queues; i++) {
1061 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1062 		if (prev_state == VNIC_CLOSED)
1063 			enable_irq(adapter->rx_scrq[i]->irq);
1064 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1065 	}
1066 
1067 	for (i = 0; i < adapter->req_tx_queues; i++) {
1068 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1069 		if (prev_state == VNIC_CLOSED)
1070 			enable_irq(adapter->tx_scrq[i]->irq);
1071 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1072 	}
1073 
1074 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1075 	if (rc) {
1076 		for (i = 0; i < adapter->req_rx_queues; i++)
1077 			napi_disable(&adapter->napi[i]);
1078 		release_resources(adapter);
1079 		return rc;
1080 	}
1081 
1082 	netif_tx_start_all_queues(netdev);
1083 
1084 	if (prev_state == VNIC_CLOSED) {
1085 		for (i = 0; i < adapter->req_rx_queues; i++)
1086 			napi_schedule(&adapter->napi[i]);
1087 	}
1088 
1089 	adapter->state = VNIC_OPEN;
1090 	return rc;
1091 }
1092 
1093 static int ibmvnic_open(struct net_device *netdev)
1094 {
1095 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1096 	int rc;
1097 
1098 	/* If device failover is pending, just set device state and return.
1099 	 * Device operation will be handled by reset routine.
1100 	 */
1101 	if (adapter->failover_pending) {
1102 		adapter->state = VNIC_OPEN;
1103 		return 0;
1104 	}
1105 
1106 	mutex_lock(&adapter->reset_lock);
1107 
1108 	if (adapter->state != VNIC_CLOSED) {
1109 		rc = ibmvnic_login(netdev);
1110 		if (rc) {
1111 			mutex_unlock(&adapter->reset_lock);
1112 			return rc;
1113 		}
1114 
1115 		rc = init_resources(adapter);
1116 		if (rc) {
1117 			netdev_err(netdev, "failed to initialize resources\n");
1118 			release_resources(adapter);
1119 			mutex_unlock(&adapter->reset_lock);
1120 			return rc;
1121 		}
1122 	}
1123 
1124 	rc = __ibmvnic_open(netdev);
1125 	netif_carrier_on(netdev);
1126 
1127 	mutex_unlock(&adapter->reset_lock);
1128 
1129 	return rc;
1130 }
1131 
1132 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1133 {
1134 	struct ibmvnic_rx_pool *rx_pool;
1135 	struct ibmvnic_rx_buff *rx_buff;
1136 	u64 rx_entries;
1137 	int rx_scrqs;
1138 	int i, j;
1139 
1140 	if (!adapter->rx_pool)
1141 		return;
1142 
1143 	rx_scrqs = adapter->num_active_rx_pools;
1144 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1145 
1146 	/* Free any remaining skbs in the rx buffer pools */
1147 	for (i = 0; i < rx_scrqs; i++) {
1148 		rx_pool = &adapter->rx_pool[i];
1149 		if (!rx_pool || !rx_pool->rx_buff)
1150 			continue;
1151 
1152 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1153 		for (j = 0; j < rx_entries; j++) {
1154 			rx_buff = &rx_pool->rx_buff[j];
1155 			if (rx_buff && rx_buff->skb) {
1156 				dev_kfree_skb_any(rx_buff->skb);
1157 				rx_buff->skb = NULL;
1158 			}
1159 		}
1160 	}
1161 }
1162 
1163 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1164 			      struct ibmvnic_tx_pool *tx_pool)
1165 {
1166 	struct ibmvnic_tx_buff *tx_buff;
1167 	u64 tx_entries;
1168 	int i;
1169 
1170 	if (!tx_pool || !tx_pool->tx_buff)
1171 		return;
1172 
1173 	tx_entries = tx_pool->num_buffers;
1174 
1175 	for (i = 0; i < tx_entries; i++) {
1176 		tx_buff = &tx_pool->tx_buff[i];
1177 		if (tx_buff && tx_buff->skb) {
1178 			dev_kfree_skb_any(tx_buff->skb);
1179 			tx_buff->skb = NULL;
1180 		}
1181 	}
1182 }
1183 
1184 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1185 {
1186 	int tx_scrqs;
1187 	int i;
1188 
1189 	if (!adapter->tx_pool || !adapter->tso_pool)
1190 		return;
1191 
1192 	tx_scrqs = adapter->num_active_tx_pools;
1193 
1194 	/* Free any remaining skbs in the tx buffer pools */
1195 	for (i = 0; i < tx_scrqs; i++) {
1196 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1197 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1198 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1199 	}
1200 }
1201 
1202 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1203 {
1204 	struct net_device *netdev = adapter->netdev;
1205 	int i;
1206 
1207 	if (adapter->tx_scrq) {
1208 		for (i = 0; i < adapter->req_tx_queues; i++)
1209 			if (adapter->tx_scrq[i]->irq) {
1210 				netdev_dbg(netdev,
1211 					   "Disabling tx_scrq[%d] irq\n", i);
1212 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1213 				disable_irq(adapter->tx_scrq[i]->irq);
1214 			}
1215 	}
1216 
1217 	if (adapter->rx_scrq) {
1218 		for (i = 0; i < adapter->req_rx_queues; i++) {
1219 			if (adapter->rx_scrq[i]->irq) {
1220 				netdev_dbg(netdev,
1221 					   "Disabling rx_scrq[%d] irq\n", i);
1222 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1223 				disable_irq(adapter->rx_scrq[i]->irq);
1224 			}
1225 		}
1226 	}
1227 }
1228 
1229 static void ibmvnic_cleanup(struct net_device *netdev)
1230 {
1231 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1232 
1233 	/* ensure that transmissions are stopped if called by do_reset */
1234 	if (adapter->resetting)
1235 		netif_tx_disable(netdev);
1236 	else
1237 		netif_tx_stop_all_queues(netdev);
1238 
1239 	ibmvnic_napi_disable(adapter);
1240 	ibmvnic_disable_irqs(adapter);
1241 
1242 	clean_rx_pools(adapter);
1243 	clean_tx_pools(adapter);
1244 }
1245 
1246 static int __ibmvnic_close(struct net_device *netdev)
1247 {
1248 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1249 	int rc = 0;
1250 
1251 	adapter->state = VNIC_CLOSING;
1252 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1253 	if (rc)
1254 		return rc;
1255 	adapter->state = VNIC_CLOSED;
1256 	return 0;
1257 }
1258 
1259 static int ibmvnic_close(struct net_device *netdev)
1260 {
1261 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1262 	int rc;
1263 
1264 	/* If device failover is pending, just set device state and return.
1265 	 * Device operation will be handled by reset routine.
1266 	 */
1267 	if (adapter->failover_pending) {
1268 		adapter->state = VNIC_CLOSED;
1269 		return 0;
1270 	}
1271 
1272 	mutex_lock(&adapter->reset_lock);
1273 	rc = __ibmvnic_close(netdev);
1274 	ibmvnic_cleanup(netdev);
1275 	mutex_unlock(&adapter->reset_lock);
1276 
1277 	return rc;
1278 }
1279 
1280 /**
1281  * build_hdr_data - creates L2/L3/L4 header data buffer
1282  * @hdr_field - bitfield determining needed headers
1283  * @skb - socket buffer
1284  * @hdr_len - array of header lengths
1285  * @tot_len - total length of data
1286  *
1287  * Reads hdr_field to determine which headers are needed by firmware.
1288  * Builds a buffer containing these headers.  Saves individual header
1289  * lengths and total buffer length to be used to build descriptors.
1290  */
1291 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1292 			  int *hdr_len, u8 *hdr_data)
1293 {
1294 	int len = 0;
1295 	u8 *hdr;
1296 
1297 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1298 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1299 	else
1300 		hdr_len[0] = sizeof(struct ethhdr);
1301 
1302 	if (skb->protocol == htons(ETH_P_IP)) {
1303 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1304 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1305 			hdr_len[2] = tcp_hdrlen(skb);
1306 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1307 			hdr_len[2] = sizeof(struct udphdr);
1308 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1309 		hdr_len[1] = sizeof(struct ipv6hdr);
1310 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1311 			hdr_len[2] = tcp_hdrlen(skb);
1312 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1313 			hdr_len[2] = sizeof(struct udphdr);
1314 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1315 		hdr_len[1] = arp_hdr_len(skb->dev);
1316 		hdr_len[2] = 0;
1317 	}
1318 
1319 	memset(hdr_data, 0, 120);
1320 	if ((hdr_field >> 6) & 1) {
1321 		hdr = skb_mac_header(skb);
1322 		memcpy(hdr_data, hdr, hdr_len[0]);
1323 		len += hdr_len[0];
1324 	}
1325 
1326 	if ((hdr_field >> 5) & 1) {
1327 		hdr = skb_network_header(skb);
1328 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1329 		len += hdr_len[1];
1330 	}
1331 
1332 	if ((hdr_field >> 4) & 1) {
1333 		hdr = skb_transport_header(skb);
1334 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1335 		len += hdr_len[2];
1336 	}
1337 	return len;
1338 }
1339 
1340 /**
1341  * create_hdr_descs - create header and header extension descriptors
1342  * @hdr_field - bitfield determining needed headers
1343  * @data - buffer containing header data
1344  * @len - length of data buffer
1345  * @hdr_len - array of individual header lengths
1346  * @scrq_arr - descriptor array
1347  *
1348  * Creates header and, if needed, header extension descriptors and
1349  * places them in a descriptor array, scrq_arr
1350  */
1351 
1352 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1353 			    union sub_crq *scrq_arr)
1354 {
1355 	union sub_crq hdr_desc;
1356 	int tmp_len = len;
1357 	int num_descs = 0;
1358 	u8 *data, *cur;
1359 	int tmp;
1360 
1361 	while (tmp_len > 0) {
1362 		cur = hdr_data + len - tmp_len;
1363 
1364 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1365 		if (cur != hdr_data) {
1366 			data = hdr_desc.hdr_ext.data;
1367 			tmp = tmp_len > 29 ? 29 : tmp_len;
1368 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1369 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1370 			hdr_desc.hdr_ext.len = tmp;
1371 		} else {
1372 			data = hdr_desc.hdr.data;
1373 			tmp = tmp_len > 24 ? 24 : tmp_len;
1374 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1375 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1376 			hdr_desc.hdr.len = tmp;
1377 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1378 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1379 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1380 			hdr_desc.hdr.flag = hdr_field << 1;
1381 		}
1382 		memcpy(data, cur, tmp);
1383 		tmp_len -= tmp;
1384 		*scrq_arr = hdr_desc;
1385 		scrq_arr++;
1386 		num_descs++;
1387 	}
1388 
1389 	return num_descs;
1390 }
1391 
1392 /**
1393  * build_hdr_descs_arr - build a header descriptor array
1394  * @skb - socket buffer
1395  * @num_entries - number of descriptors to be sent
1396  * @subcrq - first TX descriptor
1397  * @hdr_field - bit field determining which headers will be sent
1398  *
1399  * This function will build a TX descriptor array with applicable
1400  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1401  */
1402 
1403 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1404 				int *num_entries, u8 hdr_field)
1405 {
1406 	int hdr_len[3] = {0, 0, 0};
1407 	int tot_len;
1408 	u8 *hdr_data = txbuff->hdr_data;
1409 
1410 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1411 				 txbuff->hdr_data);
1412 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1413 			 txbuff->indir_arr + 1);
1414 }
1415 
1416 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1417 				    struct net_device *netdev)
1418 {
1419 	/* For some backing devices, mishandling of small packets
1420 	 * can result in a loss of connection or TX stall. Device
1421 	 * architects recommend that no packet should be smaller
1422 	 * than the minimum MTU value provided to the driver, so
1423 	 * pad any packets to that length
1424 	 */
1425 	if (skb->len < netdev->min_mtu)
1426 		return skb_put_padto(skb, netdev->min_mtu);
1427 
1428 	return 0;
1429 }
1430 
1431 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1432 {
1433 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1434 	int queue_num = skb_get_queue_mapping(skb);
1435 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1436 	struct device *dev = &adapter->vdev->dev;
1437 	struct ibmvnic_tx_buff *tx_buff = NULL;
1438 	struct ibmvnic_sub_crq_queue *tx_scrq;
1439 	struct ibmvnic_tx_pool *tx_pool;
1440 	unsigned int tx_send_failed = 0;
1441 	unsigned int tx_map_failed = 0;
1442 	unsigned int tx_dropped = 0;
1443 	unsigned int tx_packets = 0;
1444 	unsigned int tx_bytes = 0;
1445 	dma_addr_t data_dma_addr;
1446 	struct netdev_queue *txq;
1447 	unsigned long lpar_rc;
1448 	union sub_crq tx_crq;
1449 	unsigned int offset;
1450 	int num_entries = 1;
1451 	unsigned char *dst;
1452 	u64 *handle_array;
1453 	int index = 0;
1454 	u8 proto = 0;
1455 	int ret = 0;
1456 
1457 	if (adapter->resetting) {
1458 		if (!netif_subqueue_stopped(netdev, skb))
1459 			netif_stop_subqueue(netdev, queue_num);
1460 		dev_kfree_skb_any(skb);
1461 
1462 		tx_send_failed++;
1463 		tx_dropped++;
1464 		ret = NETDEV_TX_OK;
1465 		goto out;
1466 	}
1467 
1468 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1469 		tx_dropped++;
1470 		tx_send_failed++;
1471 		ret = NETDEV_TX_OK;
1472 		goto out;
1473 	}
1474 	if (skb_is_gso(skb))
1475 		tx_pool = &adapter->tso_pool[queue_num];
1476 	else
1477 		tx_pool = &adapter->tx_pool[queue_num];
1478 
1479 	tx_scrq = adapter->tx_scrq[queue_num];
1480 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1481 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1482 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1483 
1484 	index = tx_pool->free_map[tx_pool->consumer_index];
1485 
1486 	if (index == IBMVNIC_INVALID_MAP) {
1487 		dev_kfree_skb_any(skb);
1488 		tx_send_failed++;
1489 		tx_dropped++;
1490 		ret = NETDEV_TX_OK;
1491 		goto out;
1492 	}
1493 
1494 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1495 
1496 	offset = index * tx_pool->buf_size;
1497 	dst = tx_pool->long_term_buff.buff + offset;
1498 	memset(dst, 0, tx_pool->buf_size);
1499 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1500 
1501 	if (skb_shinfo(skb)->nr_frags) {
1502 		int cur, i;
1503 
1504 		/* Copy the head */
1505 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1506 		cur = skb_headlen(skb);
1507 
1508 		/* Copy the frags */
1509 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1510 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1511 
1512 			memcpy(dst + cur,
1513 			       page_address(skb_frag_page(frag)) +
1514 			       frag->page_offset, skb_frag_size(frag));
1515 			cur += skb_frag_size(frag);
1516 		}
1517 	} else {
1518 		skb_copy_from_linear_data(skb, dst, skb->len);
1519 	}
1520 
1521 	tx_pool->consumer_index =
1522 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1523 
1524 	tx_buff = &tx_pool->tx_buff[index];
1525 	tx_buff->skb = skb;
1526 	tx_buff->data_dma[0] = data_dma_addr;
1527 	tx_buff->data_len[0] = skb->len;
1528 	tx_buff->index = index;
1529 	tx_buff->pool_index = queue_num;
1530 	tx_buff->last_frag = true;
1531 
1532 	memset(&tx_crq, 0, sizeof(tx_crq));
1533 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1534 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1535 	tx_crq.v1.n_crq_elem = 1;
1536 	tx_crq.v1.n_sge = 1;
1537 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1538 
1539 	if (skb_is_gso(skb))
1540 		tx_crq.v1.correlator =
1541 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1542 	else
1543 		tx_crq.v1.correlator = cpu_to_be32(index);
1544 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1545 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1546 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1547 
1548 	if (adapter->vlan_header_insertion) {
1549 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1550 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1551 	}
1552 
1553 	if (skb->protocol == htons(ETH_P_IP)) {
1554 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1555 		proto = ip_hdr(skb)->protocol;
1556 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1557 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1558 		proto = ipv6_hdr(skb)->nexthdr;
1559 	}
1560 
1561 	if (proto == IPPROTO_TCP)
1562 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1563 	else if (proto == IPPROTO_UDP)
1564 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1565 
1566 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1567 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1568 		hdrs += 2;
1569 	}
1570 	if (skb_is_gso(skb)) {
1571 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1572 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1573 		hdrs += 2;
1574 	}
1575 	/* determine if l2/3/4 headers are sent to firmware */
1576 	if ((*hdrs >> 7) & 1) {
1577 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1578 		tx_crq.v1.n_crq_elem = num_entries;
1579 		tx_buff->num_entries = num_entries;
1580 		tx_buff->indir_arr[0] = tx_crq;
1581 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1582 						    sizeof(tx_buff->indir_arr),
1583 						    DMA_TO_DEVICE);
1584 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1585 			dev_kfree_skb_any(skb);
1586 			tx_buff->skb = NULL;
1587 			if (!firmware_has_feature(FW_FEATURE_CMO))
1588 				dev_err(dev, "tx: unable to map descriptor array\n");
1589 			tx_map_failed++;
1590 			tx_dropped++;
1591 			ret = NETDEV_TX_OK;
1592 			goto tx_err_out;
1593 		}
1594 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1595 					       (u64)tx_buff->indir_dma,
1596 					       (u64)num_entries);
1597 	} else {
1598 		tx_buff->num_entries = num_entries;
1599 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1600 				      &tx_crq);
1601 	}
1602 	if (lpar_rc != H_SUCCESS) {
1603 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1604 			dev_err_ratelimited(dev, "tx: send failed\n");
1605 		dev_kfree_skb_any(skb);
1606 		tx_buff->skb = NULL;
1607 
1608 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1609 			/* Disable TX and report carrier off if queue is closed
1610 			 * or pending failover.
1611 			 * Firmware guarantees that a signal will be sent to the
1612 			 * driver, triggering a reset or some other action.
1613 			 */
1614 			netif_tx_stop_all_queues(netdev);
1615 			netif_carrier_off(netdev);
1616 		}
1617 
1618 		tx_send_failed++;
1619 		tx_dropped++;
1620 		ret = NETDEV_TX_OK;
1621 		goto tx_err_out;
1622 	}
1623 
1624 	if (atomic_add_return(num_entries, &tx_scrq->used)
1625 					>= adapter->req_tx_entries_per_subcrq) {
1626 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1627 		netif_stop_subqueue(netdev, queue_num);
1628 	}
1629 
1630 	tx_packets++;
1631 	tx_bytes += skb->len;
1632 	txq->trans_start = jiffies;
1633 	ret = NETDEV_TX_OK;
1634 	goto out;
1635 
1636 tx_err_out:
1637 	/* roll back consumer index and map array*/
1638 	if (tx_pool->consumer_index == 0)
1639 		tx_pool->consumer_index =
1640 			tx_pool->num_buffers - 1;
1641 	else
1642 		tx_pool->consumer_index--;
1643 	tx_pool->free_map[tx_pool->consumer_index] = index;
1644 out:
1645 	netdev->stats.tx_dropped += tx_dropped;
1646 	netdev->stats.tx_bytes += tx_bytes;
1647 	netdev->stats.tx_packets += tx_packets;
1648 	adapter->tx_send_failed += tx_send_failed;
1649 	adapter->tx_map_failed += tx_map_failed;
1650 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1651 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1652 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1653 
1654 	return ret;
1655 }
1656 
1657 static void ibmvnic_set_multi(struct net_device *netdev)
1658 {
1659 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1660 	struct netdev_hw_addr *ha;
1661 	union ibmvnic_crq crq;
1662 
1663 	memset(&crq, 0, sizeof(crq));
1664 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1665 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1666 
1667 	if (netdev->flags & IFF_PROMISC) {
1668 		if (!adapter->promisc_supported)
1669 			return;
1670 	} else {
1671 		if (netdev->flags & IFF_ALLMULTI) {
1672 			/* Accept all multicast */
1673 			memset(&crq, 0, sizeof(crq));
1674 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1675 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1676 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1677 			ibmvnic_send_crq(adapter, &crq);
1678 		} else if (netdev_mc_empty(netdev)) {
1679 			/* Reject all multicast */
1680 			memset(&crq, 0, sizeof(crq));
1681 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1682 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1683 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1684 			ibmvnic_send_crq(adapter, &crq);
1685 		} else {
1686 			/* Accept one or more multicast(s) */
1687 			netdev_for_each_mc_addr(ha, netdev) {
1688 				memset(&crq, 0, sizeof(crq));
1689 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1690 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1691 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1692 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1693 						ha->addr);
1694 				ibmvnic_send_crq(adapter, &crq);
1695 			}
1696 		}
1697 	}
1698 }
1699 
1700 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1701 {
1702 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1703 	struct sockaddr *addr = p;
1704 	union ibmvnic_crq crq;
1705 	int rc;
1706 
1707 	if (!is_valid_ether_addr(addr->sa_data))
1708 		return -EADDRNOTAVAIL;
1709 
1710 	memset(&crq, 0, sizeof(crq));
1711 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1712 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1713 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1714 
1715 	init_completion(&adapter->fw_done);
1716 	rc = ibmvnic_send_crq(adapter, &crq);
1717 	if (rc)
1718 		return rc;
1719 	wait_for_completion(&adapter->fw_done);
1720 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1721 	return adapter->fw_done_rc ? -EIO : 0;
1722 }
1723 
1724 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1725 {
1726 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1727 	struct sockaddr *addr = p;
1728 	int rc;
1729 
1730 	if (adapter->state == VNIC_PROBED) {
1731 		memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1732 		adapter->mac_change_pending = true;
1733 		return 0;
1734 	}
1735 
1736 	rc = __ibmvnic_set_mac(netdev, addr);
1737 
1738 	return rc;
1739 }
1740 
1741 /**
1742  * do_reset returns zero if we are able to keep processing reset events, or
1743  * non-zero if we hit a fatal error and must halt.
1744  */
1745 static int do_reset(struct ibmvnic_adapter *adapter,
1746 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1747 {
1748 	u64 old_num_rx_queues, old_num_tx_queues;
1749 	struct net_device *netdev = adapter->netdev;
1750 	int i, rc;
1751 
1752 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1753 		   rwi->reset_reason);
1754 
1755 	netif_carrier_off(netdev);
1756 	adapter->reset_reason = rwi->reset_reason;
1757 
1758 	old_num_rx_queues = adapter->req_rx_queues;
1759 	old_num_tx_queues = adapter->req_tx_queues;
1760 
1761 	ibmvnic_cleanup(netdev);
1762 
1763 	if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1764 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1765 		rc = __ibmvnic_close(netdev);
1766 		if (rc)
1767 			return rc;
1768 	}
1769 
1770 	if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1771 	    adapter->wait_for_reset) {
1772 		release_resources(adapter);
1773 		release_sub_crqs(adapter, 1);
1774 		release_crq_queue(adapter);
1775 	}
1776 
1777 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1778 		/* remove the closed state so when we call open it appears
1779 		 * we are coming from the probed state.
1780 		 */
1781 		adapter->state = VNIC_PROBED;
1782 
1783 		if (adapter->wait_for_reset) {
1784 			rc = init_crq_queue(adapter);
1785 		} else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1786 			rc = ibmvnic_reenable_crq_queue(adapter);
1787 			release_sub_crqs(adapter, 1);
1788 		} else {
1789 			rc = ibmvnic_reset_crq(adapter);
1790 			if (!rc)
1791 				rc = vio_enable_interrupts(adapter->vdev);
1792 		}
1793 
1794 		if (rc) {
1795 			netdev_err(adapter->netdev,
1796 				   "Couldn't initialize crq. rc=%d\n", rc);
1797 			return rc;
1798 		}
1799 
1800 		rc = ibmvnic_reset_init(adapter);
1801 		if (rc)
1802 			return IBMVNIC_INIT_FAILED;
1803 
1804 		/* If the adapter was in PROBE state prior to the reset,
1805 		 * exit here.
1806 		 */
1807 		if (reset_state == VNIC_PROBED)
1808 			return 0;
1809 
1810 		rc = ibmvnic_login(netdev);
1811 		if (rc) {
1812 			adapter->state = reset_state;
1813 			return rc;
1814 		}
1815 
1816 		if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1817 		    adapter->wait_for_reset) {
1818 			rc = init_resources(adapter);
1819 			if (rc)
1820 				return rc;
1821 		} else if (adapter->req_rx_queues != old_num_rx_queues ||
1822 			   adapter->req_tx_queues != old_num_tx_queues) {
1823 			adapter->map_id = 1;
1824 			release_rx_pools(adapter);
1825 			release_tx_pools(adapter);
1826 			init_rx_pools(netdev);
1827 			init_tx_pools(netdev);
1828 
1829 			release_napi(adapter);
1830 			init_napi(adapter);
1831 		} else {
1832 			rc = reset_tx_pools(adapter);
1833 			if (rc)
1834 				return rc;
1835 
1836 			rc = reset_rx_pools(adapter);
1837 			if (rc)
1838 				return rc;
1839 		}
1840 		ibmvnic_disable_irqs(adapter);
1841 	}
1842 	adapter->state = VNIC_CLOSED;
1843 
1844 	if (reset_state == VNIC_CLOSED)
1845 		return 0;
1846 
1847 	rc = __ibmvnic_open(netdev);
1848 	if (rc) {
1849 		if (list_empty(&adapter->rwi_list))
1850 			adapter->state = VNIC_CLOSED;
1851 		else
1852 			adapter->state = reset_state;
1853 
1854 		return 0;
1855 	}
1856 
1857 	/* kick napi */
1858 	for (i = 0; i < adapter->req_rx_queues; i++)
1859 		napi_schedule(&adapter->napi[i]);
1860 
1861 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1862 	    adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1863 		netdev_notify_peers(netdev);
1864 
1865 	netif_carrier_on(netdev);
1866 
1867 	return 0;
1868 }
1869 
1870 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1871 			 struct ibmvnic_rwi *rwi, u32 reset_state)
1872 {
1873 	struct net_device *netdev = adapter->netdev;
1874 	int rc;
1875 
1876 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1877 		   rwi->reset_reason);
1878 
1879 	netif_carrier_off(netdev);
1880 	adapter->reset_reason = rwi->reset_reason;
1881 
1882 	ibmvnic_cleanup(netdev);
1883 	release_resources(adapter);
1884 	release_sub_crqs(adapter, 0);
1885 	release_crq_queue(adapter);
1886 
1887 	/* remove the closed state so when we call open it appears
1888 	 * we are coming from the probed state.
1889 	 */
1890 	adapter->state = VNIC_PROBED;
1891 
1892 	rc = init_crq_queue(adapter);
1893 	if (rc) {
1894 		netdev_err(adapter->netdev,
1895 			   "Couldn't initialize crq. rc=%d\n", rc);
1896 		return rc;
1897 	}
1898 
1899 	rc = ibmvnic_init(adapter);
1900 	if (rc)
1901 		return rc;
1902 
1903 	/* If the adapter was in PROBE state prior to the reset,
1904 	 * exit here.
1905 	 */
1906 	if (reset_state == VNIC_PROBED)
1907 		return 0;
1908 
1909 	rc = ibmvnic_login(netdev);
1910 	if (rc) {
1911 		adapter->state = VNIC_PROBED;
1912 		return 0;
1913 	}
1914 	/* netif_set_real_num_xx_queues needs to take rtnl lock here
1915 	 * unless wait_for_reset is set, in which case the rtnl lock
1916 	 * has already been taken before initializing the reset
1917 	 */
1918 	if (!adapter->wait_for_reset) {
1919 		rtnl_lock();
1920 		rc = init_resources(adapter);
1921 		rtnl_unlock();
1922 	} else {
1923 		rc = init_resources(adapter);
1924 	}
1925 	if (rc)
1926 		return rc;
1927 
1928 	ibmvnic_disable_irqs(adapter);
1929 	adapter->state = VNIC_CLOSED;
1930 
1931 	if (reset_state == VNIC_CLOSED)
1932 		return 0;
1933 
1934 	rc = __ibmvnic_open(netdev);
1935 	if (rc) {
1936 		if (list_empty(&adapter->rwi_list))
1937 			adapter->state = VNIC_CLOSED;
1938 		else
1939 			adapter->state = reset_state;
1940 
1941 		return 0;
1942 	}
1943 
1944 	netif_carrier_on(netdev);
1945 
1946 	return 0;
1947 }
1948 
1949 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1950 {
1951 	struct ibmvnic_rwi *rwi;
1952 
1953 	mutex_lock(&adapter->rwi_lock);
1954 
1955 	if (!list_empty(&adapter->rwi_list)) {
1956 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1957 				       list);
1958 		list_del(&rwi->list);
1959 	} else {
1960 		rwi = NULL;
1961 	}
1962 
1963 	mutex_unlock(&adapter->rwi_lock);
1964 	return rwi;
1965 }
1966 
1967 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1968 {
1969 	struct ibmvnic_rwi *rwi;
1970 
1971 	rwi = get_next_rwi(adapter);
1972 	while (rwi) {
1973 		kfree(rwi);
1974 		rwi = get_next_rwi(adapter);
1975 	}
1976 }
1977 
1978 static void __ibmvnic_reset(struct work_struct *work)
1979 {
1980 	struct ibmvnic_rwi *rwi;
1981 	struct ibmvnic_adapter *adapter;
1982 	struct net_device *netdev;
1983 	u32 reset_state;
1984 	int rc = 0;
1985 
1986 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1987 	netdev = adapter->netdev;
1988 
1989 	mutex_lock(&adapter->reset_lock);
1990 	reset_state = adapter->state;
1991 
1992 	rwi = get_next_rwi(adapter);
1993 	while (rwi) {
1994 		if (adapter->force_reset_recovery) {
1995 			adapter->force_reset_recovery = false;
1996 			rc = do_hard_reset(adapter, rwi, reset_state);
1997 		} else {
1998 			rc = do_reset(adapter, rwi, reset_state);
1999 		}
2000 		kfree(rwi);
2001 		if (rc && rc != IBMVNIC_INIT_FAILED &&
2002 		    !adapter->force_reset_recovery)
2003 			break;
2004 
2005 		rwi = get_next_rwi(adapter);
2006 	}
2007 
2008 	if (adapter->wait_for_reset) {
2009 		adapter->wait_for_reset = false;
2010 		adapter->reset_done_rc = rc;
2011 		complete(&adapter->reset_done);
2012 	}
2013 
2014 	if (rc) {
2015 		netdev_dbg(adapter->netdev, "Reset failed\n");
2016 		free_all_rwi(adapter);
2017 		mutex_unlock(&adapter->reset_lock);
2018 		return;
2019 	}
2020 
2021 	adapter->resetting = false;
2022 	mutex_unlock(&adapter->reset_lock);
2023 }
2024 
2025 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2026 			 enum ibmvnic_reset_reason reason)
2027 {
2028 	struct list_head *entry, *tmp_entry;
2029 	struct ibmvnic_rwi *rwi, *tmp;
2030 	struct net_device *netdev = adapter->netdev;
2031 	int ret;
2032 
2033 	if (adapter->state == VNIC_REMOVING ||
2034 	    adapter->state == VNIC_REMOVED ||
2035 	    adapter->failover_pending) {
2036 		ret = EBUSY;
2037 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2038 		goto err;
2039 	}
2040 
2041 	if (adapter->state == VNIC_PROBING) {
2042 		netdev_warn(netdev, "Adapter reset during probe\n");
2043 		ret = adapter->init_done_rc = EAGAIN;
2044 		goto err;
2045 	}
2046 
2047 	mutex_lock(&adapter->rwi_lock);
2048 
2049 	list_for_each(entry, &adapter->rwi_list) {
2050 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2051 		if (tmp->reset_reason == reason) {
2052 			netdev_dbg(netdev, "Skipping matching reset\n");
2053 			mutex_unlock(&adapter->rwi_lock);
2054 			ret = EBUSY;
2055 			goto err;
2056 		}
2057 	}
2058 
2059 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
2060 	if (!rwi) {
2061 		mutex_unlock(&adapter->rwi_lock);
2062 		ibmvnic_close(netdev);
2063 		ret = ENOMEM;
2064 		goto err;
2065 	}
2066 	/* if we just received a transport event,
2067 	 * flush reset queue and process this reset
2068 	 */
2069 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2070 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2071 			list_del(entry);
2072 	}
2073 	rwi->reset_reason = reason;
2074 	list_add_tail(&rwi->list, &adapter->rwi_list);
2075 	mutex_unlock(&adapter->rwi_lock);
2076 	adapter->resetting = true;
2077 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2078 	schedule_work(&adapter->ibmvnic_reset);
2079 
2080 	return 0;
2081 err:
2082 	if (adapter->wait_for_reset)
2083 		adapter->wait_for_reset = false;
2084 	return -ret;
2085 }
2086 
2087 static void ibmvnic_tx_timeout(struct net_device *dev)
2088 {
2089 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2090 
2091 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2092 }
2093 
2094 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2095 				  struct ibmvnic_rx_buff *rx_buff)
2096 {
2097 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2098 
2099 	rx_buff->skb = NULL;
2100 
2101 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2102 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2103 
2104 	atomic_dec(&pool->available);
2105 }
2106 
2107 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2108 {
2109 	struct net_device *netdev = napi->dev;
2110 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2111 	int scrq_num = (int)(napi - adapter->napi);
2112 	int frames_processed = 0;
2113 
2114 restart_poll:
2115 	while (frames_processed < budget) {
2116 		struct sk_buff *skb;
2117 		struct ibmvnic_rx_buff *rx_buff;
2118 		union sub_crq *next;
2119 		u32 length;
2120 		u16 offset;
2121 		u8 flags = 0;
2122 
2123 		if (unlikely(adapter->resetting &&
2124 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2125 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2126 			napi_complete_done(napi, frames_processed);
2127 			return frames_processed;
2128 		}
2129 
2130 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2131 			break;
2132 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2133 		rx_buff =
2134 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2135 							  rx_comp.correlator);
2136 		/* do error checking */
2137 		if (next->rx_comp.rc) {
2138 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2139 				   be16_to_cpu(next->rx_comp.rc));
2140 			/* free the entry */
2141 			next->rx_comp.first = 0;
2142 			dev_kfree_skb_any(rx_buff->skb);
2143 			remove_buff_from_pool(adapter, rx_buff);
2144 			continue;
2145 		} else if (!rx_buff->skb) {
2146 			/* free the entry */
2147 			next->rx_comp.first = 0;
2148 			remove_buff_from_pool(adapter, rx_buff);
2149 			continue;
2150 		}
2151 
2152 		length = be32_to_cpu(next->rx_comp.len);
2153 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2154 		flags = next->rx_comp.flags;
2155 		skb = rx_buff->skb;
2156 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2157 					length);
2158 
2159 		/* VLAN Header has been stripped by the system firmware and
2160 		 * needs to be inserted by the driver
2161 		 */
2162 		if (adapter->rx_vlan_header_insertion &&
2163 		    (flags & IBMVNIC_VLAN_STRIPPED))
2164 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2165 					       ntohs(next->rx_comp.vlan_tci));
2166 
2167 		/* free the entry */
2168 		next->rx_comp.first = 0;
2169 		remove_buff_from_pool(adapter, rx_buff);
2170 
2171 		skb_put(skb, length);
2172 		skb->protocol = eth_type_trans(skb, netdev);
2173 		skb_record_rx_queue(skb, scrq_num);
2174 
2175 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2176 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2177 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2178 		}
2179 
2180 		length = skb->len;
2181 		napi_gro_receive(napi, skb); /* send it up */
2182 		netdev->stats.rx_packets++;
2183 		netdev->stats.rx_bytes += length;
2184 		adapter->rx_stats_buffers[scrq_num].packets++;
2185 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2186 		frames_processed++;
2187 	}
2188 
2189 	if (adapter->state != VNIC_CLOSING)
2190 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2191 
2192 	if (frames_processed < budget) {
2193 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2194 		napi_complete_done(napi, frames_processed);
2195 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2196 		    napi_reschedule(napi)) {
2197 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2198 			goto restart_poll;
2199 		}
2200 	}
2201 	return frames_processed;
2202 }
2203 
2204 #ifdef CONFIG_NET_POLL_CONTROLLER
2205 static void ibmvnic_netpoll_controller(struct net_device *dev)
2206 {
2207 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2208 	int i;
2209 
2210 	replenish_pools(netdev_priv(dev));
2211 	for (i = 0; i < adapter->req_rx_queues; i++)
2212 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
2213 				     adapter->rx_scrq[i]);
2214 }
2215 #endif
2216 
2217 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2218 {
2219 	int rc, ret;
2220 
2221 	adapter->fallback.mtu = adapter->req_mtu;
2222 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2223 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2224 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2225 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2226 
2227 	init_completion(&adapter->reset_done);
2228 	adapter->wait_for_reset = true;
2229 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2230 	if (rc)
2231 		return rc;
2232 	wait_for_completion(&adapter->reset_done);
2233 
2234 	ret = 0;
2235 	if (adapter->reset_done_rc) {
2236 		ret = -EIO;
2237 		adapter->desired.mtu = adapter->fallback.mtu;
2238 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2239 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2240 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2241 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2242 
2243 		init_completion(&adapter->reset_done);
2244 		adapter->wait_for_reset = true;
2245 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2246 		if (rc)
2247 			return ret;
2248 		wait_for_completion(&adapter->reset_done);
2249 	}
2250 	adapter->wait_for_reset = false;
2251 
2252 	return ret;
2253 }
2254 
2255 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2256 {
2257 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2258 
2259 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2260 
2261 	return wait_for_reset(adapter);
2262 }
2263 
2264 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2265 						struct net_device *dev,
2266 						netdev_features_t features)
2267 {
2268 	/* Some backing hardware adapters can not
2269 	 * handle packets with a MSS less than 224
2270 	 * or with only one segment.
2271 	 */
2272 	if (skb_is_gso(skb)) {
2273 		if (skb_shinfo(skb)->gso_size < 224 ||
2274 		    skb_shinfo(skb)->gso_segs == 1)
2275 			features &= ~NETIF_F_GSO_MASK;
2276 	}
2277 
2278 	return features;
2279 }
2280 
2281 static const struct net_device_ops ibmvnic_netdev_ops = {
2282 	.ndo_open		= ibmvnic_open,
2283 	.ndo_stop		= ibmvnic_close,
2284 	.ndo_start_xmit		= ibmvnic_xmit,
2285 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2286 	.ndo_set_mac_address	= ibmvnic_set_mac,
2287 	.ndo_validate_addr	= eth_validate_addr,
2288 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2289 #ifdef CONFIG_NET_POLL_CONTROLLER
2290 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
2291 #endif
2292 	.ndo_change_mtu		= ibmvnic_change_mtu,
2293 	.ndo_features_check     = ibmvnic_features_check,
2294 };
2295 
2296 /* ethtool functions */
2297 
2298 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2299 				      struct ethtool_link_ksettings *cmd)
2300 {
2301 	u32 supported, advertising;
2302 
2303 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2304 			  SUPPORTED_FIBRE);
2305 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2306 			    ADVERTISED_FIBRE);
2307 	cmd->base.speed = SPEED_1000;
2308 	cmd->base.duplex = DUPLEX_FULL;
2309 	cmd->base.port = PORT_FIBRE;
2310 	cmd->base.phy_address = 0;
2311 	cmd->base.autoneg = AUTONEG_ENABLE;
2312 
2313 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2314 						supported);
2315 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2316 						advertising);
2317 
2318 	return 0;
2319 }
2320 
2321 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2322 				struct ethtool_drvinfo *info)
2323 {
2324 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2325 
2326 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2327 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2328 	strlcpy(info->fw_version, adapter->fw_version,
2329 		sizeof(info->fw_version));
2330 }
2331 
2332 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2333 {
2334 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2335 
2336 	return adapter->msg_enable;
2337 }
2338 
2339 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2340 {
2341 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2342 
2343 	adapter->msg_enable = data;
2344 }
2345 
2346 static u32 ibmvnic_get_link(struct net_device *netdev)
2347 {
2348 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2349 
2350 	/* Don't need to send a query because we request a logical link up at
2351 	 * init and then we wait for link state indications
2352 	 */
2353 	return adapter->logical_link_state;
2354 }
2355 
2356 static void ibmvnic_get_ringparam(struct net_device *netdev,
2357 				  struct ethtool_ringparam *ring)
2358 {
2359 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2360 
2361 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2362 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2363 	ring->rx_mini_max_pending = 0;
2364 	ring->rx_jumbo_max_pending = 0;
2365 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2366 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2367 	ring->rx_mini_pending = 0;
2368 	ring->rx_jumbo_pending = 0;
2369 }
2370 
2371 static int ibmvnic_set_ringparam(struct net_device *netdev,
2372 				 struct ethtool_ringparam *ring)
2373 {
2374 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2375 
2376 	if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2377 	    ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2378 		netdev_err(netdev, "Invalid request.\n");
2379 		netdev_err(netdev, "Max tx buffers = %llu\n",
2380 			   adapter->max_rx_add_entries_per_subcrq);
2381 		netdev_err(netdev, "Max rx buffers = %llu\n",
2382 			   adapter->max_tx_entries_per_subcrq);
2383 		return -EINVAL;
2384 	}
2385 
2386 	adapter->desired.rx_entries = ring->rx_pending;
2387 	adapter->desired.tx_entries = ring->tx_pending;
2388 
2389 	return wait_for_reset(adapter);
2390 }
2391 
2392 static void ibmvnic_get_channels(struct net_device *netdev,
2393 				 struct ethtool_channels *channels)
2394 {
2395 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2396 
2397 	channels->max_rx = adapter->max_rx_queues;
2398 	channels->max_tx = adapter->max_tx_queues;
2399 	channels->max_other = 0;
2400 	channels->max_combined = 0;
2401 	channels->rx_count = adapter->req_rx_queues;
2402 	channels->tx_count = adapter->req_tx_queues;
2403 	channels->other_count = 0;
2404 	channels->combined_count = 0;
2405 }
2406 
2407 static int ibmvnic_set_channels(struct net_device *netdev,
2408 				struct ethtool_channels *channels)
2409 {
2410 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2411 
2412 	adapter->desired.rx_queues = channels->rx_count;
2413 	adapter->desired.tx_queues = channels->tx_count;
2414 
2415 	return wait_for_reset(adapter);
2416 }
2417 
2418 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2419 {
2420 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2421 	int i;
2422 
2423 	if (stringset != ETH_SS_STATS)
2424 		return;
2425 
2426 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2427 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2428 
2429 	for (i = 0; i < adapter->req_tx_queues; i++) {
2430 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2431 		data += ETH_GSTRING_LEN;
2432 
2433 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2434 		data += ETH_GSTRING_LEN;
2435 
2436 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2437 		data += ETH_GSTRING_LEN;
2438 	}
2439 
2440 	for (i = 0; i < adapter->req_rx_queues; i++) {
2441 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2442 		data += ETH_GSTRING_LEN;
2443 
2444 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2445 		data += ETH_GSTRING_LEN;
2446 
2447 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2448 		data += ETH_GSTRING_LEN;
2449 	}
2450 }
2451 
2452 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2453 {
2454 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2455 
2456 	switch (sset) {
2457 	case ETH_SS_STATS:
2458 		return ARRAY_SIZE(ibmvnic_stats) +
2459 		       adapter->req_tx_queues * NUM_TX_STATS +
2460 		       adapter->req_rx_queues * NUM_RX_STATS;
2461 	default:
2462 		return -EOPNOTSUPP;
2463 	}
2464 }
2465 
2466 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2467 				      struct ethtool_stats *stats, u64 *data)
2468 {
2469 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2470 	union ibmvnic_crq crq;
2471 	int i, j;
2472 	int rc;
2473 
2474 	memset(&crq, 0, sizeof(crq));
2475 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2476 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2477 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2478 	crq.request_statistics.len =
2479 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2480 
2481 	/* Wait for data to be written */
2482 	init_completion(&adapter->stats_done);
2483 	rc = ibmvnic_send_crq(adapter, &crq);
2484 	if (rc)
2485 		return;
2486 	wait_for_completion(&adapter->stats_done);
2487 
2488 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2489 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2490 						ibmvnic_stats[i].offset));
2491 
2492 	for (j = 0; j < adapter->req_tx_queues; j++) {
2493 		data[i] = adapter->tx_stats_buffers[j].packets;
2494 		i++;
2495 		data[i] = adapter->tx_stats_buffers[j].bytes;
2496 		i++;
2497 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2498 		i++;
2499 	}
2500 
2501 	for (j = 0; j < adapter->req_rx_queues; j++) {
2502 		data[i] = adapter->rx_stats_buffers[j].packets;
2503 		i++;
2504 		data[i] = adapter->rx_stats_buffers[j].bytes;
2505 		i++;
2506 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2507 		i++;
2508 	}
2509 }
2510 
2511 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2512 	.get_drvinfo		= ibmvnic_get_drvinfo,
2513 	.get_msglevel		= ibmvnic_get_msglevel,
2514 	.set_msglevel		= ibmvnic_set_msglevel,
2515 	.get_link		= ibmvnic_get_link,
2516 	.get_ringparam		= ibmvnic_get_ringparam,
2517 	.set_ringparam		= ibmvnic_set_ringparam,
2518 	.get_channels		= ibmvnic_get_channels,
2519 	.set_channels		= ibmvnic_set_channels,
2520 	.get_strings            = ibmvnic_get_strings,
2521 	.get_sset_count         = ibmvnic_get_sset_count,
2522 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2523 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2524 };
2525 
2526 /* Routines for managing CRQs/sCRQs  */
2527 
2528 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2529 				   struct ibmvnic_sub_crq_queue *scrq)
2530 {
2531 	int rc;
2532 
2533 	if (scrq->irq) {
2534 		free_irq(scrq->irq, scrq);
2535 		irq_dispose_mapping(scrq->irq);
2536 		scrq->irq = 0;
2537 	}
2538 
2539 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2540 	atomic_set(&scrq->used, 0);
2541 	scrq->cur = 0;
2542 
2543 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2544 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2545 	return rc;
2546 }
2547 
2548 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2549 {
2550 	int i, rc;
2551 
2552 	for (i = 0; i < adapter->req_tx_queues; i++) {
2553 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2554 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2555 		if (rc)
2556 			return rc;
2557 	}
2558 
2559 	for (i = 0; i < adapter->req_rx_queues; i++) {
2560 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2561 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2562 		if (rc)
2563 			return rc;
2564 	}
2565 
2566 	return rc;
2567 }
2568 
2569 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2570 				  struct ibmvnic_sub_crq_queue *scrq,
2571 				  bool do_h_free)
2572 {
2573 	struct device *dev = &adapter->vdev->dev;
2574 	long rc;
2575 
2576 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2577 
2578 	if (do_h_free) {
2579 		/* Close the sub-crqs */
2580 		do {
2581 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2582 						adapter->vdev->unit_address,
2583 						scrq->crq_num);
2584 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2585 
2586 		if (rc) {
2587 			netdev_err(adapter->netdev,
2588 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2589 				   scrq->crq_num, rc);
2590 		}
2591 	}
2592 
2593 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2594 			 DMA_BIDIRECTIONAL);
2595 	free_pages((unsigned long)scrq->msgs, 2);
2596 	kfree(scrq);
2597 }
2598 
2599 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2600 							*adapter)
2601 {
2602 	struct device *dev = &adapter->vdev->dev;
2603 	struct ibmvnic_sub_crq_queue *scrq;
2604 	int rc;
2605 
2606 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2607 	if (!scrq)
2608 		return NULL;
2609 
2610 	scrq->msgs =
2611 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2612 	if (!scrq->msgs) {
2613 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2614 		goto zero_page_failed;
2615 	}
2616 
2617 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2618 					 DMA_BIDIRECTIONAL);
2619 	if (dma_mapping_error(dev, scrq->msg_token)) {
2620 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2621 		goto map_failed;
2622 	}
2623 
2624 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2625 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2626 
2627 	if (rc == H_RESOURCE)
2628 		rc = ibmvnic_reset_crq(adapter);
2629 
2630 	if (rc == H_CLOSED) {
2631 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2632 	} else if (rc) {
2633 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2634 		goto reg_failed;
2635 	}
2636 
2637 	scrq->adapter = adapter;
2638 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2639 	spin_lock_init(&scrq->lock);
2640 
2641 	netdev_dbg(adapter->netdev,
2642 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2643 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2644 
2645 	return scrq;
2646 
2647 reg_failed:
2648 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2649 			 DMA_BIDIRECTIONAL);
2650 map_failed:
2651 	free_pages((unsigned long)scrq->msgs, 2);
2652 zero_page_failed:
2653 	kfree(scrq);
2654 
2655 	return NULL;
2656 }
2657 
2658 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2659 {
2660 	int i;
2661 
2662 	if (adapter->tx_scrq) {
2663 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2664 			if (!adapter->tx_scrq[i])
2665 				continue;
2666 
2667 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2668 				   i);
2669 			if (adapter->tx_scrq[i]->irq) {
2670 				free_irq(adapter->tx_scrq[i]->irq,
2671 					 adapter->tx_scrq[i]);
2672 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2673 				adapter->tx_scrq[i]->irq = 0;
2674 			}
2675 
2676 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2677 					      do_h_free);
2678 		}
2679 
2680 		kfree(adapter->tx_scrq);
2681 		adapter->tx_scrq = NULL;
2682 		adapter->num_active_tx_scrqs = 0;
2683 	}
2684 
2685 	if (adapter->rx_scrq) {
2686 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2687 			if (!adapter->rx_scrq[i])
2688 				continue;
2689 
2690 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2691 				   i);
2692 			if (adapter->rx_scrq[i]->irq) {
2693 				free_irq(adapter->rx_scrq[i]->irq,
2694 					 adapter->rx_scrq[i]);
2695 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2696 				adapter->rx_scrq[i]->irq = 0;
2697 			}
2698 
2699 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2700 					      do_h_free);
2701 		}
2702 
2703 		kfree(adapter->rx_scrq);
2704 		adapter->rx_scrq = NULL;
2705 		adapter->num_active_rx_scrqs = 0;
2706 	}
2707 }
2708 
2709 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2710 			    struct ibmvnic_sub_crq_queue *scrq)
2711 {
2712 	struct device *dev = &adapter->vdev->dev;
2713 	unsigned long rc;
2714 
2715 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2716 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2717 	if (rc)
2718 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2719 			scrq->hw_irq, rc);
2720 	return rc;
2721 }
2722 
2723 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2724 			   struct ibmvnic_sub_crq_queue *scrq)
2725 {
2726 	struct device *dev = &adapter->vdev->dev;
2727 	unsigned long rc;
2728 
2729 	if (scrq->hw_irq > 0x100000000ULL) {
2730 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2731 		return 1;
2732 	}
2733 
2734 	if (adapter->resetting &&
2735 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
2736 		u64 val = (0xff000000) | scrq->hw_irq;
2737 
2738 		rc = plpar_hcall_norets(H_EOI, val);
2739 		if (rc)
2740 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2741 				val, rc);
2742 	}
2743 
2744 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2745 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2746 	if (rc)
2747 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2748 			scrq->hw_irq, rc);
2749 	return rc;
2750 }
2751 
2752 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2753 			       struct ibmvnic_sub_crq_queue *scrq)
2754 {
2755 	struct device *dev = &adapter->vdev->dev;
2756 	struct ibmvnic_tx_pool *tx_pool;
2757 	struct ibmvnic_tx_buff *txbuff;
2758 	union sub_crq *next;
2759 	int index;
2760 	int i, j;
2761 	u8 *first;
2762 
2763 restart_loop:
2764 	while (pending_scrq(adapter, scrq)) {
2765 		unsigned int pool = scrq->pool_index;
2766 		int num_entries = 0;
2767 
2768 		next = ibmvnic_next_scrq(adapter, scrq);
2769 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2770 			if (next->tx_comp.rcs[i]) {
2771 				dev_err(dev, "tx error %x\n",
2772 					next->tx_comp.rcs[i]);
2773 				continue;
2774 			}
2775 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2776 			if (index & IBMVNIC_TSO_POOL_MASK) {
2777 				tx_pool = &adapter->tso_pool[pool];
2778 				index &= ~IBMVNIC_TSO_POOL_MASK;
2779 			} else {
2780 				tx_pool = &adapter->tx_pool[pool];
2781 			}
2782 
2783 			txbuff = &tx_pool->tx_buff[index];
2784 
2785 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2786 				if (!txbuff->data_dma[j])
2787 					continue;
2788 
2789 				txbuff->data_dma[j] = 0;
2790 			}
2791 			/* if sub_crq was sent indirectly */
2792 			first = &txbuff->indir_arr[0].generic.first;
2793 			if (*first == IBMVNIC_CRQ_CMD) {
2794 				dma_unmap_single(dev, txbuff->indir_dma,
2795 						 sizeof(txbuff->indir_arr),
2796 						 DMA_TO_DEVICE);
2797 				*first = 0;
2798 			}
2799 
2800 			if (txbuff->last_frag) {
2801 				dev_kfree_skb_any(txbuff->skb);
2802 				txbuff->skb = NULL;
2803 			}
2804 
2805 			num_entries += txbuff->num_entries;
2806 
2807 			tx_pool->free_map[tx_pool->producer_index] = index;
2808 			tx_pool->producer_index =
2809 				(tx_pool->producer_index + 1) %
2810 					tx_pool->num_buffers;
2811 		}
2812 		/* remove tx_comp scrq*/
2813 		next->tx_comp.first = 0;
2814 
2815 		if (atomic_sub_return(num_entries, &scrq->used) <=
2816 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2817 		    __netif_subqueue_stopped(adapter->netdev,
2818 					     scrq->pool_index)) {
2819 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2820 			netdev_dbg(adapter->netdev, "Started queue %d\n",
2821 				   scrq->pool_index);
2822 		}
2823 	}
2824 
2825 	enable_scrq_irq(adapter, scrq);
2826 
2827 	if (pending_scrq(adapter, scrq)) {
2828 		disable_scrq_irq(adapter, scrq);
2829 		goto restart_loop;
2830 	}
2831 
2832 	return 0;
2833 }
2834 
2835 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2836 {
2837 	struct ibmvnic_sub_crq_queue *scrq = instance;
2838 	struct ibmvnic_adapter *adapter = scrq->adapter;
2839 
2840 	disable_scrq_irq(adapter, scrq);
2841 	ibmvnic_complete_tx(adapter, scrq);
2842 
2843 	return IRQ_HANDLED;
2844 }
2845 
2846 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2847 {
2848 	struct ibmvnic_sub_crq_queue *scrq = instance;
2849 	struct ibmvnic_adapter *adapter = scrq->adapter;
2850 
2851 	/* When booting a kdump kernel we can hit pending interrupts
2852 	 * prior to completing driver initialization.
2853 	 */
2854 	if (unlikely(adapter->state != VNIC_OPEN))
2855 		return IRQ_NONE;
2856 
2857 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2858 
2859 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2860 		disable_scrq_irq(adapter, scrq);
2861 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2862 	}
2863 
2864 	return IRQ_HANDLED;
2865 }
2866 
2867 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2868 {
2869 	struct device *dev = &adapter->vdev->dev;
2870 	struct ibmvnic_sub_crq_queue *scrq;
2871 	int i = 0, j = 0;
2872 	int rc = 0;
2873 
2874 	for (i = 0; i < adapter->req_tx_queues; i++) {
2875 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2876 			   i);
2877 		scrq = adapter->tx_scrq[i];
2878 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2879 
2880 		if (!scrq->irq) {
2881 			rc = -EINVAL;
2882 			dev_err(dev, "Error mapping irq\n");
2883 			goto req_tx_irq_failed;
2884 		}
2885 
2886 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2887 				 0, "ibmvnic_tx", scrq);
2888 
2889 		if (rc) {
2890 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2891 				scrq->irq, rc);
2892 			irq_dispose_mapping(scrq->irq);
2893 			goto req_tx_irq_failed;
2894 		}
2895 	}
2896 
2897 	for (i = 0; i < adapter->req_rx_queues; i++) {
2898 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2899 			   i);
2900 		scrq = adapter->rx_scrq[i];
2901 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2902 		if (!scrq->irq) {
2903 			rc = -EINVAL;
2904 			dev_err(dev, "Error mapping irq\n");
2905 			goto req_rx_irq_failed;
2906 		}
2907 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2908 				 0, "ibmvnic_rx", scrq);
2909 		if (rc) {
2910 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2911 				scrq->irq, rc);
2912 			irq_dispose_mapping(scrq->irq);
2913 			goto req_rx_irq_failed;
2914 		}
2915 	}
2916 	return rc;
2917 
2918 req_rx_irq_failed:
2919 	for (j = 0; j < i; j++) {
2920 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2921 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2922 	}
2923 	i = adapter->req_tx_queues;
2924 req_tx_irq_failed:
2925 	for (j = 0; j < i; j++) {
2926 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2927 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2928 	}
2929 	release_sub_crqs(adapter, 1);
2930 	return rc;
2931 }
2932 
2933 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2934 {
2935 	struct device *dev = &adapter->vdev->dev;
2936 	struct ibmvnic_sub_crq_queue **allqueues;
2937 	int registered_queues = 0;
2938 	int total_queues;
2939 	int more = 0;
2940 	int i;
2941 
2942 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2943 
2944 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2945 	if (!allqueues)
2946 		return -1;
2947 
2948 	for (i = 0; i < total_queues; i++) {
2949 		allqueues[i] = init_sub_crq_queue(adapter);
2950 		if (!allqueues[i]) {
2951 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2952 			break;
2953 		}
2954 		registered_queues++;
2955 	}
2956 
2957 	/* Make sure we were able to register the minimum number of queues */
2958 	if (registered_queues <
2959 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2960 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2961 		goto tx_failed;
2962 	}
2963 
2964 	/* Distribute the failed allocated queues*/
2965 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2966 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2967 		switch (i % 3) {
2968 		case 0:
2969 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2970 				adapter->req_rx_queues--;
2971 			else
2972 				more++;
2973 			break;
2974 		case 1:
2975 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2976 				adapter->req_tx_queues--;
2977 			else
2978 				more++;
2979 			break;
2980 		}
2981 	}
2982 
2983 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2984 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2985 	if (!adapter->tx_scrq)
2986 		goto tx_failed;
2987 
2988 	for (i = 0; i < adapter->req_tx_queues; i++) {
2989 		adapter->tx_scrq[i] = allqueues[i];
2990 		adapter->tx_scrq[i]->pool_index = i;
2991 		adapter->num_active_tx_scrqs++;
2992 	}
2993 
2994 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2995 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2996 	if (!adapter->rx_scrq)
2997 		goto rx_failed;
2998 
2999 	for (i = 0; i < adapter->req_rx_queues; i++) {
3000 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3001 		adapter->rx_scrq[i]->scrq_num = i;
3002 		adapter->num_active_rx_scrqs++;
3003 	}
3004 
3005 	kfree(allqueues);
3006 	return 0;
3007 
3008 rx_failed:
3009 	kfree(adapter->tx_scrq);
3010 	adapter->tx_scrq = NULL;
3011 tx_failed:
3012 	for (i = 0; i < registered_queues; i++)
3013 		release_sub_crq_queue(adapter, allqueues[i], 1);
3014 	kfree(allqueues);
3015 	return -1;
3016 }
3017 
3018 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3019 {
3020 	struct device *dev = &adapter->vdev->dev;
3021 	union ibmvnic_crq crq;
3022 	int max_entries;
3023 
3024 	if (!retry) {
3025 		/* Sub-CRQ entries are 32 byte long */
3026 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3027 
3028 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3029 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3030 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3031 			return;
3032 		}
3033 
3034 		if (adapter->desired.mtu)
3035 			adapter->req_mtu = adapter->desired.mtu;
3036 		else
3037 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3038 
3039 		if (!adapter->desired.tx_entries)
3040 			adapter->desired.tx_entries =
3041 					adapter->max_tx_entries_per_subcrq;
3042 		if (!adapter->desired.rx_entries)
3043 			adapter->desired.rx_entries =
3044 					adapter->max_rx_add_entries_per_subcrq;
3045 
3046 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3047 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3048 
3049 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3050 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3051 			adapter->desired.tx_entries = max_entries;
3052 		}
3053 
3054 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3055 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3056 			adapter->desired.rx_entries = max_entries;
3057 		}
3058 
3059 		if (adapter->desired.tx_entries)
3060 			adapter->req_tx_entries_per_subcrq =
3061 					adapter->desired.tx_entries;
3062 		else
3063 			adapter->req_tx_entries_per_subcrq =
3064 					adapter->max_tx_entries_per_subcrq;
3065 
3066 		if (adapter->desired.rx_entries)
3067 			adapter->req_rx_add_entries_per_subcrq =
3068 					adapter->desired.rx_entries;
3069 		else
3070 			adapter->req_rx_add_entries_per_subcrq =
3071 					adapter->max_rx_add_entries_per_subcrq;
3072 
3073 		if (adapter->desired.tx_queues)
3074 			adapter->req_tx_queues =
3075 					adapter->desired.tx_queues;
3076 		else
3077 			adapter->req_tx_queues =
3078 					adapter->opt_tx_comp_sub_queues;
3079 
3080 		if (adapter->desired.rx_queues)
3081 			adapter->req_rx_queues =
3082 					adapter->desired.rx_queues;
3083 		else
3084 			adapter->req_rx_queues =
3085 					adapter->opt_rx_comp_queues;
3086 
3087 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3088 	}
3089 
3090 	memset(&crq, 0, sizeof(crq));
3091 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3092 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3093 
3094 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3095 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3096 	atomic_inc(&adapter->running_cap_crqs);
3097 	ibmvnic_send_crq(adapter, &crq);
3098 
3099 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3100 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3101 	atomic_inc(&adapter->running_cap_crqs);
3102 	ibmvnic_send_crq(adapter, &crq);
3103 
3104 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3105 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3106 	atomic_inc(&adapter->running_cap_crqs);
3107 	ibmvnic_send_crq(adapter, &crq);
3108 
3109 	crq.request_capability.capability =
3110 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3111 	crq.request_capability.number =
3112 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3113 	atomic_inc(&adapter->running_cap_crqs);
3114 	ibmvnic_send_crq(adapter, &crq);
3115 
3116 	crq.request_capability.capability =
3117 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3118 	crq.request_capability.number =
3119 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3120 	atomic_inc(&adapter->running_cap_crqs);
3121 	ibmvnic_send_crq(adapter, &crq);
3122 
3123 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3124 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3125 	atomic_inc(&adapter->running_cap_crqs);
3126 	ibmvnic_send_crq(adapter, &crq);
3127 
3128 	if (adapter->netdev->flags & IFF_PROMISC) {
3129 		if (adapter->promisc_supported) {
3130 			crq.request_capability.capability =
3131 			    cpu_to_be16(PROMISC_REQUESTED);
3132 			crq.request_capability.number = cpu_to_be64(1);
3133 			atomic_inc(&adapter->running_cap_crqs);
3134 			ibmvnic_send_crq(adapter, &crq);
3135 		}
3136 	} else {
3137 		crq.request_capability.capability =
3138 		    cpu_to_be16(PROMISC_REQUESTED);
3139 		crq.request_capability.number = cpu_to_be64(0);
3140 		atomic_inc(&adapter->running_cap_crqs);
3141 		ibmvnic_send_crq(adapter, &crq);
3142 	}
3143 }
3144 
3145 static int pending_scrq(struct ibmvnic_adapter *adapter,
3146 			struct ibmvnic_sub_crq_queue *scrq)
3147 {
3148 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3149 
3150 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3151 		return 1;
3152 	else
3153 		return 0;
3154 }
3155 
3156 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3157 					struct ibmvnic_sub_crq_queue *scrq)
3158 {
3159 	union sub_crq *entry;
3160 	unsigned long flags;
3161 
3162 	spin_lock_irqsave(&scrq->lock, flags);
3163 	entry = &scrq->msgs[scrq->cur];
3164 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3165 		if (++scrq->cur == scrq->size)
3166 			scrq->cur = 0;
3167 	} else {
3168 		entry = NULL;
3169 	}
3170 	spin_unlock_irqrestore(&scrq->lock, flags);
3171 
3172 	return entry;
3173 }
3174 
3175 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3176 {
3177 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3178 	union ibmvnic_crq *crq;
3179 
3180 	crq = &queue->msgs[queue->cur];
3181 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3182 		if (++queue->cur == queue->size)
3183 			queue->cur = 0;
3184 	} else {
3185 		crq = NULL;
3186 	}
3187 
3188 	return crq;
3189 }
3190 
3191 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3192 {
3193 	switch (rc) {
3194 	case H_PARAMETER:
3195 		dev_warn_ratelimited(dev,
3196 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3197 				     func, rc);
3198 		break;
3199 	case H_CLOSED:
3200 		dev_warn_ratelimited(dev,
3201 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3202 				     func, rc);
3203 		break;
3204 	default:
3205 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3206 		break;
3207 	}
3208 }
3209 
3210 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3211 		       union sub_crq *sub_crq)
3212 {
3213 	unsigned int ua = adapter->vdev->unit_address;
3214 	struct device *dev = &adapter->vdev->dev;
3215 	u64 *u64_crq = (u64 *)sub_crq;
3216 	int rc;
3217 
3218 	netdev_dbg(adapter->netdev,
3219 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3220 		   (unsigned long int)cpu_to_be64(remote_handle),
3221 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3222 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3223 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3224 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3225 
3226 	/* Make sure the hypervisor sees the complete request */
3227 	mb();
3228 
3229 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3230 				cpu_to_be64(remote_handle),
3231 				cpu_to_be64(u64_crq[0]),
3232 				cpu_to_be64(u64_crq[1]),
3233 				cpu_to_be64(u64_crq[2]),
3234 				cpu_to_be64(u64_crq[3]));
3235 
3236 	if (rc)
3237 		print_subcrq_error(dev, rc, __func__);
3238 
3239 	return rc;
3240 }
3241 
3242 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3243 				u64 remote_handle, u64 ioba, u64 num_entries)
3244 {
3245 	unsigned int ua = adapter->vdev->unit_address;
3246 	struct device *dev = &adapter->vdev->dev;
3247 	int rc;
3248 
3249 	/* Make sure the hypervisor sees the complete request */
3250 	mb();
3251 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3252 				cpu_to_be64(remote_handle),
3253 				ioba, num_entries);
3254 
3255 	if (rc)
3256 		print_subcrq_error(dev, rc, __func__);
3257 
3258 	return rc;
3259 }
3260 
3261 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3262 			    union ibmvnic_crq *crq)
3263 {
3264 	unsigned int ua = adapter->vdev->unit_address;
3265 	struct device *dev = &adapter->vdev->dev;
3266 	u64 *u64_crq = (u64 *)crq;
3267 	int rc;
3268 
3269 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3270 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3271 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3272 
3273 	if (!adapter->crq.active &&
3274 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3275 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3276 		return -EINVAL;
3277 	}
3278 
3279 	/* Make sure the hypervisor sees the complete request */
3280 	mb();
3281 
3282 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3283 				cpu_to_be64(u64_crq[0]),
3284 				cpu_to_be64(u64_crq[1]));
3285 
3286 	if (rc) {
3287 		if (rc == H_CLOSED) {
3288 			dev_warn(dev, "CRQ Queue closed\n");
3289 			if (adapter->resetting)
3290 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3291 		}
3292 
3293 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3294 	}
3295 
3296 	return rc;
3297 }
3298 
3299 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3300 {
3301 	union ibmvnic_crq crq;
3302 
3303 	memset(&crq, 0, sizeof(crq));
3304 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3305 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3306 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3307 
3308 	return ibmvnic_send_crq(adapter, &crq);
3309 }
3310 
3311 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3312 {
3313 	union ibmvnic_crq crq;
3314 
3315 	memset(&crq, 0, sizeof(crq));
3316 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3317 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3318 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3319 
3320 	return ibmvnic_send_crq(adapter, &crq);
3321 }
3322 
3323 struct vnic_login_client_data {
3324 	u8	type;
3325 	__be16	len;
3326 	char	name[];
3327 } __packed;
3328 
3329 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3330 {
3331 	int len;
3332 
3333 	/* Calculate the amount of buffer space needed for the
3334 	 * vnic client data in the login buffer. There are four entries,
3335 	 * OS name, LPAR name, device name, and a null last entry.
3336 	 */
3337 	len = 4 * sizeof(struct vnic_login_client_data);
3338 	len += 6; /* "Linux" plus NULL */
3339 	len += strlen(utsname()->nodename) + 1;
3340 	len += strlen(adapter->netdev->name) + 1;
3341 
3342 	return len;
3343 }
3344 
3345 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3346 				 struct vnic_login_client_data *vlcd)
3347 {
3348 	const char *os_name = "Linux";
3349 	int len;
3350 
3351 	/* Type 1 - LPAR OS */
3352 	vlcd->type = 1;
3353 	len = strlen(os_name) + 1;
3354 	vlcd->len = cpu_to_be16(len);
3355 	strncpy(vlcd->name, os_name, len);
3356 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3357 
3358 	/* Type 2 - LPAR name */
3359 	vlcd->type = 2;
3360 	len = strlen(utsname()->nodename) + 1;
3361 	vlcd->len = cpu_to_be16(len);
3362 	strncpy(vlcd->name, utsname()->nodename, len);
3363 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3364 
3365 	/* Type 3 - device name */
3366 	vlcd->type = 3;
3367 	len = strlen(adapter->netdev->name) + 1;
3368 	vlcd->len = cpu_to_be16(len);
3369 	strncpy(vlcd->name, adapter->netdev->name, len);
3370 }
3371 
3372 static int send_login(struct ibmvnic_adapter *adapter)
3373 {
3374 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3375 	struct ibmvnic_login_buffer *login_buffer;
3376 	struct device *dev = &adapter->vdev->dev;
3377 	dma_addr_t rsp_buffer_token;
3378 	dma_addr_t buffer_token;
3379 	size_t rsp_buffer_size;
3380 	union ibmvnic_crq crq;
3381 	size_t buffer_size;
3382 	__be64 *tx_list_p;
3383 	__be64 *rx_list_p;
3384 	int client_data_len;
3385 	struct vnic_login_client_data *vlcd;
3386 	int i;
3387 
3388 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3389 		netdev_err(adapter->netdev,
3390 			   "RX or TX queues are not allocated, device login failed\n");
3391 		return -1;
3392 	}
3393 
3394 	release_login_rsp_buffer(adapter);
3395 	client_data_len = vnic_client_data_len(adapter);
3396 
3397 	buffer_size =
3398 	    sizeof(struct ibmvnic_login_buffer) +
3399 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3400 	    client_data_len;
3401 
3402 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3403 	if (!login_buffer)
3404 		goto buf_alloc_failed;
3405 
3406 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3407 				      DMA_TO_DEVICE);
3408 	if (dma_mapping_error(dev, buffer_token)) {
3409 		dev_err(dev, "Couldn't map login buffer\n");
3410 		goto buf_map_failed;
3411 	}
3412 
3413 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3414 			  sizeof(u64) * adapter->req_tx_queues +
3415 			  sizeof(u64) * adapter->req_rx_queues +
3416 			  sizeof(u64) * adapter->req_rx_queues +
3417 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3418 
3419 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3420 	if (!login_rsp_buffer)
3421 		goto buf_rsp_alloc_failed;
3422 
3423 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3424 					  rsp_buffer_size, DMA_FROM_DEVICE);
3425 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3426 		dev_err(dev, "Couldn't map login rsp buffer\n");
3427 		goto buf_rsp_map_failed;
3428 	}
3429 
3430 	adapter->login_buf = login_buffer;
3431 	adapter->login_buf_token = buffer_token;
3432 	adapter->login_buf_sz = buffer_size;
3433 	adapter->login_rsp_buf = login_rsp_buffer;
3434 	adapter->login_rsp_buf_token = rsp_buffer_token;
3435 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3436 
3437 	login_buffer->len = cpu_to_be32(buffer_size);
3438 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3439 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3440 	login_buffer->off_txcomp_subcrqs =
3441 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3442 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3443 	login_buffer->off_rxcomp_subcrqs =
3444 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3445 			sizeof(u64) * adapter->req_tx_queues);
3446 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3447 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3448 
3449 	tx_list_p = (__be64 *)((char *)login_buffer +
3450 				      sizeof(struct ibmvnic_login_buffer));
3451 	rx_list_p = (__be64 *)((char *)login_buffer +
3452 				      sizeof(struct ibmvnic_login_buffer) +
3453 				      sizeof(u64) * adapter->req_tx_queues);
3454 
3455 	for (i = 0; i < adapter->req_tx_queues; i++) {
3456 		if (adapter->tx_scrq[i]) {
3457 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3458 						   crq_num);
3459 		}
3460 	}
3461 
3462 	for (i = 0; i < adapter->req_rx_queues; i++) {
3463 		if (adapter->rx_scrq[i]) {
3464 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3465 						   crq_num);
3466 		}
3467 	}
3468 
3469 	/* Insert vNIC login client data */
3470 	vlcd = (struct vnic_login_client_data *)
3471 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3472 	login_buffer->client_data_offset =
3473 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3474 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3475 
3476 	vnic_add_client_data(adapter, vlcd);
3477 
3478 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3479 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3480 		netdev_dbg(adapter->netdev, "%016lx\n",
3481 			   ((unsigned long int *)(adapter->login_buf))[i]);
3482 	}
3483 
3484 	memset(&crq, 0, sizeof(crq));
3485 	crq.login.first = IBMVNIC_CRQ_CMD;
3486 	crq.login.cmd = LOGIN;
3487 	crq.login.ioba = cpu_to_be32(buffer_token);
3488 	crq.login.len = cpu_to_be32(buffer_size);
3489 	ibmvnic_send_crq(adapter, &crq);
3490 
3491 	return 0;
3492 
3493 buf_rsp_map_failed:
3494 	kfree(login_rsp_buffer);
3495 buf_rsp_alloc_failed:
3496 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3497 buf_map_failed:
3498 	kfree(login_buffer);
3499 buf_alloc_failed:
3500 	return -1;
3501 }
3502 
3503 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3504 			    u32 len, u8 map_id)
3505 {
3506 	union ibmvnic_crq crq;
3507 
3508 	memset(&crq, 0, sizeof(crq));
3509 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3510 	crq.request_map.cmd = REQUEST_MAP;
3511 	crq.request_map.map_id = map_id;
3512 	crq.request_map.ioba = cpu_to_be32(addr);
3513 	crq.request_map.len = cpu_to_be32(len);
3514 	return ibmvnic_send_crq(adapter, &crq);
3515 }
3516 
3517 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3518 {
3519 	union ibmvnic_crq crq;
3520 
3521 	memset(&crq, 0, sizeof(crq));
3522 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3523 	crq.request_unmap.cmd = REQUEST_UNMAP;
3524 	crq.request_unmap.map_id = map_id;
3525 	return ibmvnic_send_crq(adapter, &crq);
3526 }
3527 
3528 static void send_map_query(struct ibmvnic_adapter *adapter)
3529 {
3530 	union ibmvnic_crq crq;
3531 
3532 	memset(&crq, 0, sizeof(crq));
3533 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3534 	crq.query_map.cmd = QUERY_MAP;
3535 	ibmvnic_send_crq(adapter, &crq);
3536 }
3537 
3538 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3539 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3540 {
3541 	union ibmvnic_crq crq;
3542 
3543 	atomic_set(&adapter->running_cap_crqs, 0);
3544 	memset(&crq, 0, sizeof(crq));
3545 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3546 	crq.query_capability.cmd = QUERY_CAPABILITY;
3547 
3548 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3549 	atomic_inc(&adapter->running_cap_crqs);
3550 	ibmvnic_send_crq(adapter, &crq);
3551 
3552 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3553 	atomic_inc(&adapter->running_cap_crqs);
3554 	ibmvnic_send_crq(adapter, &crq);
3555 
3556 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3557 	atomic_inc(&adapter->running_cap_crqs);
3558 	ibmvnic_send_crq(adapter, &crq);
3559 
3560 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3561 	atomic_inc(&adapter->running_cap_crqs);
3562 	ibmvnic_send_crq(adapter, &crq);
3563 
3564 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3565 	atomic_inc(&adapter->running_cap_crqs);
3566 	ibmvnic_send_crq(adapter, &crq);
3567 
3568 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3569 	atomic_inc(&adapter->running_cap_crqs);
3570 	ibmvnic_send_crq(adapter, &crq);
3571 
3572 	crq.query_capability.capability =
3573 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3574 	atomic_inc(&adapter->running_cap_crqs);
3575 	ibmvnic_send_crq(adapter, &crq);
3576 
3577 	crq.query_capability.capability =
3578 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3579 	atomic_inc(&adapter->running_cap_crqs);
3580 	ibmvnic_send_crq(adapter, &crq);
3581 
3582 	crq.query_capability.capability =
3583 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3584 	atomic_inc(&adapter->running_cap_crqs);
3585 	ibmvnic_send_crq(adapter, &crq);
3586 
3587 	crq.query_capability.capability =
3588 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3589 	atomic_inc(&adapter->running_cap_crqs);
3590 	ibmvnic_send_crq(adapter, &crq);
3591 
3592 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3593 	atomic_inc(&adapter->running_cap_crqs);
3594 	ibmvnic_send_crq(adapter, &crq);
3595 
3596 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3597 	atomic_inc(&adapter->running_cap_crqs);
3598 	ibmvnic_send_crq(adapter, &crq);
3599 
3600 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3601 	atomic_inc(&adapter->running_cap_crqs);
3602 	ibmvnic_send_crq(adapter, &crq);
3603 
3604 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3605 	atomic_inc(&adapter->running_cap_crqs);
3606 	ibmvnic_send_crq(adapter, &crq);
3607 
3608 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3609 	atomic_inc(&adapter->running_cap_crqs);
3610 	ibmvnic_send_crq(adapter, &crq);
3611 
3612 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3613 	atomic_inc(&adapter->running_cap_crqs);
3614 	ibmvnic_send_crq(adapter, &crq);
3615 
3616 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3617 	atomic_inc(&adapter->running_cap_crqs);
3618 	ibmvnic_send_crq(adapter, &crq);
3619 
3620 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3621 	atomic_inc(&adapter->running_cap_crqs);
3622 	ibmvnic_send_crq(adapter, &crq);
3623 
3624 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3625 	atomic_inc(&adapter->running_cap_crqs);
3626 	ibmvnic_send_crq(adapter, &crq);
3627 
3628 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3629 	atomic_inc(&adapter->running_cap_crqs);
3630 	ibmvnic_send_crq(adapter, &crq);
3631 
3632 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3633 	atomic_inc(&adapter->running_cap_crqs);
3634 	ibmvnic_send_crq(adapter, &crq);
3635 
3636 	crq.query_capability.capability =
3637 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3638 	atomic_inc(&adapter->running_cap_crqs);
3639 	ibmvnic_send_crq(adapter, &crq);
3640 
3641 	crq.query_capability.capability =
3642 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3643 	atomic_inc(&adapter->running_cap_crqs);
3644 	ibmvnic_send_crq(adapter, &crq);
3645 
3646 	crq.query_capability.capability =
3647 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3648 	atomic_inc(&adapter->running_cap_crqs);
3649 	ibmvnic_send_crq(adapter, &crq);
3650 
3651 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3652 	atomic_inc(&adapter->running_cap_crqs);
3653 	ibmvnic_send_crq(adapter, &crq);
3654 }
3655 
3656 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3657 				struct ibmvnic_adapter *adapter)
3658 {
3659 	struct device *dev = &adapter->vdev->dev;
3660 
3661 	if (crq->get_vpd_size_rsp.rc.code) {
3662 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3663 			crq->get_vpd_size_rsp.rc.code);
3664 		complete(&adapter->fw_done);
3665 		return;
3666 	}
3667 
3668 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3669 	complete(&adapter->fw_done);
3670 }
3671 
3672 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3673 			   struct ibmvnic_adapter *adapter)
3674 {
3675 	struct device *dev = &adapter->vdev->dev;
3676 	unsigned char *substr = NULL;
3677 	u8 fw_level_len = 0;
3678 
3679 	memset(adapter->fw_version, 0, 32);
3680 
3681 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3682 			 DMA_FROM_DEVICE);
3683 
3684 	if (crq->get_vpd_rsp.rc.code) {
3685 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3686 			crq->get_vpd_rsp.rc.code);
3687 		goto complete;
3688 	}
3689 
3690 	/* get the position of the firmware version info
3691 	 * located after the ASCII 'RM' substring in the buffer
3692 	 */
3693 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3694 	if (!substr) {
3695 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3696 		goto complete;
3697 	}
3698 
3699 	/* get length of firmware level ASCII substring */
3700 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3701 		fw_level_len = *(substr + 2);
3702 	} else {
3703 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3704 		goto complete;
3705 	}
3706 
3707 	/* copy firmware version string from vpd into adapter */
3708 	if ((substr + 3 + fw_level_len) <
3709 	    (adapter->vpd->buff + adapter->vpd->len)) {
3710 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3711 	} else {
3712 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3713 	}
3714 
3715 complete:
3716 	if (adapter->fw_version[0] == '\0')
3717 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3718 	complete(&adapter->fw_done);
3719 }
3720 
3721 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3722 {
3723 	struct device *dev = &adapter->vdev->dev;
3724 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3725 	union ibmvnic_crq crq;
3726 	int i;
3727 
3728 	dma_unmap_single(dev, adapter->ip_offload_tok,
3729 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3730 
3731 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3732 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3733 		netdev_dbg(adapter->netdev, "%016lx\n",
3734 			   ((unsigned long int *)(buf))[i]);
3735 
3736 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3737 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3738 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3739 		   buf->tcp_ipv4_chksum);
3740 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3741 		   buf->tcp_ipv6_chksum);
3742 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3743 		   buf->udp_ipv4_chksum);
3744 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3745 		   buf->udp_ipv6_chksum);
3746 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3747 		   buf->large_tx_ipv4);
3748 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3749 		   buf->large_tx_ipv6);
3750 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3751 		   buf->large_rx_ipv4);
3752 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3753 		   buf->large_rx_ipv6);
3754 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3755 		   buf->max_ipv4_header_size);
3756 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3757 		   buf->max_ipv6_header_size);
3758 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3759 		   buf->max_tcp_header_size);
3760 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3761 		   buf->max_udp_header_size);
3762 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3763 		   buf->max_large_tx_size);
3764 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3765 		   buf->max_large_rx_size);
3766 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3767 		   buf->ipv6_extension_header);
3768 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3769 		   buf->tcp_pseudosum_req);
3770 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3771 		   buf->num_ipv6_ext_headers);
3772 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3773 		   buf->off_ipv6_ext_headers);
3774 
3775 	adapter->ip_offload_ctrl_tok =
3776 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
3777 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3778 
3779 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3780 		dev_err(dev, "Couldn't map ip offload control buffer\n");
3781 		return;
3782 	}
3783 
3784 	adapter->ip_offload_ctrl.len =
3785 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3786 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3787 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3788 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3789 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3790 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3791 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3792 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3793 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3794 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3795 
3796 	/* large_rx disabled for now, additional features needed */
3797 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3798 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3799 
3800 	adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3801 
3802 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3803 		adapter->netdev->features |= NETIF_F_IP_CSUM;
3804 
3805 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3806 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3807 
3808 	if ((adapter->netdev->features &
3809 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3810 		adapter->netdev->features |= NETIF_F_RXCSUM;
3811 
3812 	if (buf->large_tx_ipv4)
3813 		adapter->netdev->features |= NETIF_F_TSO;
3814 	if (buf->large_tx_ipv6)
3815 		adapter->netdev->features |= NETIF_F_TSO6;
3816 
3817 	adapter->netdev->hw_features |= adapter->netdev->features;
3818 
3819 	memset(&crq, 0, sizeof(crq));
3820 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3821 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3822 	crq.control_ip_offload.len =
3823 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3824 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3825 	ibmvnic_send_crq(adapter, &crq);
3826 }
3827 
3828 static const char *ibmvnic_fw_err_cause(u16 cause)
3829 {
3830 	switch (cause) {
3831 	case ADAPTER_PROBLEM:
3832 		return "adapter problem";
3833 	case BUS_PROBLEM:
3834 		return "bus problem";
3835 	case FW_PROBLEM:
3836 		return "firmware problem";
3837 	case DD_PROBLEM:
3838 		return "device driver problem";
3839 	case EEH_RECOVERY:
3840 		return "EEH recovery";
3841 	case FW_UPDATED:
3842 		return "firmware updated";
3843 	case LOW_MEMORY:
3844 		return "low Memory";
3845 	default:
3846 		return "unknown";
3847 	}
3848 }
3849 
3850 static void handle_error_indication(union ibmvnic_crq *crq,
3851 				    struct ibmvnic_adapter *adapter)
3852 {
3853 	struct device *dev = &adapter->vdev->dev;
3854 	u16 cause;
3855 
3856 	cause = be16_to_cpu(crq->error_indication.error_cause);
3857 
3858 	dev_warn_ratelimited(dev,
3859 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
3860 			     crq->error_indication.flags
3861 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3862 			     ibmvnic_fw_err_cause(cause));
3863 
3864 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3865 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3866 	else
3867 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3868 }
3869 
3870 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3871 				 struct ibmvnic_adapter *adapter)
3872 {
3873 	struct net_device *netdev = adapter->netdev;
3874 	struct device *dev = &adapter->vdev->dev;
3875 	long rc;
3876 
3877 	rc = crq->change_mac_addr_rsp.rc.code;
3878 	if (rc) {
3879 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3880 		goto out;
3881 	}
3882 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3883 	       ETH_ALEN);
3884 out:
3885 	complete(&adapter->fw_done);
3886 	return rc;
3887 }
3888 
3889 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3890 				   struct ibmvnic_adapter *adapter)
3891 {
3892 	struct device *dev = &adapter->vdev->dev;
3893 	u64 *req_value;
3894 	char *name;
3895 
3896 	atomic_dec(&adapter->running_cap_crqs);
3897 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3898 	case REQ_TX_QUEUES:
3899 		req_value = &adapter->req_tx_queues;
3900 		name = "tx";
3901 		break;
3902 	case REQ_RX_QUEUES:
3903 		req_value = &adapter->req_rx_queues;
3904 		name = "rx";
3905 		break;
3906 	case REQ_RX_ADD_QUEUES:
3907 		req_value = &adapter->req_rx_add_queues;
3908 		name = "rx_add";
3909 		break;
3910 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3911 		req_value = &adapter->req_tx_entries_per_subcrq;
3912 		name = "tx_entries_per_subcrq";
3913 		break;
3914 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3915 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3916 		name = "rx_add_entries_per_subcrq";
3917 		break;
3918 	case REQ_MTU:
3919 		req_value = &adapter->req_mtu;
3920 		name = "mtu";
3921 		break;
3922 	case PROMISC_REQUESTED:
3923 		req_value = &adapter->promisc;
3924 		name = "promisc";
3925 		break;
3926 	default:
3927 		dev_err(dev, "Got invalid cap request rsp %d\n",
3928 			crq->request_capability.capability);
3929 		return;
3930 	}
3931 
3932 	switch (crq->request_capability_rsp.rc.code) {
3933 	case SUCCESS:
3934 		break;
3935 	case PARTIALSUCCESS:
3936 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3937 			 *req_value,
3938 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3939 					       number), name);
3940 
3941 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3942 		    REQ_MTU) {
3943 			pr_err("mtu of %llu is not supported. Reverting.\n",
3944 			       *req_value);
3945 			*req_value = adapter->fallback.mtu;
3946 		} else {
3947 			*req_value =
3948 				be64_to_cpu(crq->request_capability_rsp.number);
3949 		}
3950 
3951 		ibmvnic_send_req_caps(adapter, 1);
3952 		return;
3953 	default:
3954 		dev_err(dev, "Error %d in request cap rsp\n",
3955 			crq->request_capability_rsp.rc.code);
3956 		return;
3957 	}
3958 
3959 	/* Done receiving requested capabilities, query IP offload support */
3960 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3961 		union ibmvnic_crq newcrq;
3962 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3963 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3964 		    &adapter->ip_offload_buf;
3965 
3966 		adapter->wait_capability = false;
3967 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3968 							 buf_sz,
3969 							 DMA_FROM_DEVICE);
3970 
3971 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3972 			if (!firmware_has_feature(FW_FEATURE_CMO))
3973 				dev_err(dev, "Couldn't map offload buffer\n");
3974 			return;
3975 		}
3976 
3977 		memset(&newcrq, 0, sizeof(newcrq));
3978 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3979 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3980 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3981 		newcrq.query_ip_offload.ioba =
3982 		    cpu_to_be32(adapter->ip_offload_tok);
3983 
3984 		ibmvnic_send_crq(adapter, &newcrq);
3985 	}
3986 }
3987 
3988 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3989 			    struct ibmvnic_adapter *adapter)
3990 {
3991 	struct device *dev = &adapter->vdev->dev;
3992 	struct net_device *netdev = adapter->netdev;
3993 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3994 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3995 	int i;
3996 
3997 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3998 			 DMA_TO_DEVICE);
3999 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4000 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4001 
4002 	/* If the number of queues requested can't be allocated by the
4003 	 * server, the login response will return with code 1. We will need
4004 	 * to resend the login buffer with fewer queues requested.
4005 	 */
4006 	if (login_rsp_crq->generic.rc.code) {
4007 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4008 		complete(&adapter->init_done);
4009 		return 0;
4010 	}
4011 
4012 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4013 
4014 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4015 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4016 		netdev_dbg(adapter->netdev, "%016lx\n",
4017 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4018 	}
4019 
4020 	/* Sanity checks */
4021 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4022 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4023 	     adapter->req_rx_add_queues !=
4024 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4025 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4026 		ibmvnic_remove(adapter->vdev);
4027 		return -EIO;
4028 	}
4029 	release_login_buffer(adapter);
4030 	complete(&adapter->init_done);
4031 
4032 	return 0;
4033 }
4034 
4035 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4036 				     struct ibmvnic_adapter *adapter)
4037 {
4038 	struct device *dev = &adapter->vdev->dev;
4039 	long rc;
4040 
4041 	rc = crq->request_unmap_rsp.rc.code;
4042 	if (rc)
4043 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4044 }
4045 
4046 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4047 				 struct ibmvnic_adapter *adapter)
4048 {
4049 	struct net_device *netdev = adapter->netdev;
4050 	struct device *dev = &adapter->vdev->dev;
4051 	long rc;
4052 
4053 	rc = crq->query_map_rsp.rc.code;
4054 	if (rc) {
4055 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4056 		return;
4057 	}
4058 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4059 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4060 		   crq->query_map_rsp.free_pages);
4061 }
4062 
4063 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4064 				 struct ibmvnic_adapter *adapter)
4065 {
4066 	struct net_device *netdev = adapter->netdev;
4067 	struct device *dev = &adapter->vdev->dev;
4068 	long rc;
4069 
4070 	atomic_dec(&adapter->running_cap_crqs);
4071 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4072 		   atomic_read(&adapter->running_cap_crqs));
4073 	rc = crq->query_capability.rc.code;
4074 	if (rc) {
4075 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4076 		goto out;
4077 	}
4078 
4079 	switch (be16_to_cpu(crq->query_capability.capability)) {
4080 	case MIN_TX_QUEUES:
4081 		adapter->min_tx_queues =
4082 		    be64_to_cpu(crq->query_capability.number);
4083 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4084 			   adapter->min_tx_queues);
4085 		break;
4086 	case MIN_RX_QUEUES:
4087 		adapter->min_rx_queues =
4088 		    be64_to_cpu(crq->query_capability.number);
4089 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4090 			   adapter->min_rx_queues);
4091 		break;
4092 	case MIN_RX_ADD_QUEUES:
4093 		adapter->min_rx_add_queues =
4094 		    be64_to_cpu(crq->query_capability.number);
4095 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4096 			   adapter->min_rx_add_queues);
4097 		break;
4098 	case MAX_TX_QUEUES:
4099 		adapter->max_tx_queues =
4100 		    be64_to_cpu(crq->query_capability.number);
4101 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4102 			   adapter->max_tx_queues);
4103 		break;
4104 	case MAX_RX_QUEUES:
4105 		adapter->max_rx_queues =
4106 		    be64_to_cpu(crq->query_capability.number);
4107 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4108 			   adapter->max_rx_queues);
4109 		break;
4110 	case MAX_RX_ADD_QUEUES:
4111 		adapter->max_rx_add_queues =
4112 		    be64_to_cpu(crq->query_capability.number);
4113 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4114 			   adapter->max_rx_add_queues);
4115 		break;
4116 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4117 		adapter->min_tx_entries_per_subcrq =
4118 		    be64_to_cpu(crq->query_capability.number);
4119 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4120 			   adapter->min_tx_entries_per_subcrq);
4121 		break;
4122 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4123 		adapter->min_rx_add_entries_per_subcrq =
4124 		    be64_to_cpu(crq->query_capability.number);
4125 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4126 			   adapter->min_rx_add_entries_per_subcrq);
4127 		break;
4128 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4129 		adapter->max_tx_entries_per_subcrq =
4130 		    be64_to_cpu(crq->query_capability.number);
4131 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4132 			   adapter->max_tx_entries_per_subcrq);
4133 		break;
4134 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4135 		adapter->max_rx_add_entries_per_subcrq =
4136 		    be64_to_cpu(crq->query_capability.number);
4137 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4138 			   adapter->max_rx_add_entries_per_subcrq);
4139 		break;
4140 	case TCP_IP_OFFLOAD:
4141 		adapter->tcp_ip_offload =
4142 		    be64_to_cpu(crq->query_capability.number);
4143 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4144 			   adapter->tcp_ip_offload);
4145 		break;
4146 	case PROMISC_SUPPORTED:
4147 		adapter->promisc_supported =
4148 		    be64_to_cpu(crq->query_capability.number);
4149 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4150 			   adapter->promisc_supported);
4151 		break;
4152 	case MIN_MTU:
4153 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4154 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4155 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4156 		break;
4157 	case MAX_MTU:
4158 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4159 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4160 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4161 		break;
4162 	case MAX_MULTICAST_FILTERS:
4163 		adapter->max_multicast_filters =
4164 		    be64_to_cpu(crq->query_capability.number);
4165 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4166 			   adapter->max_multicast_filters);
4167 		break;
4168 	case VLAN_HEADER_INSERTION:
4169 		adapter->vlan_header_insertion =
4170 		    be64_to_cpu(crq->query_capability.number);
4171 		if (adapter->vlan_header_insertion)
4172 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4173 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4174 			   adapter->vlan_header_insertion);
4175 		break;
4176 	case RX_VLAN_HEADER_INSERTION:
4177 		adapter->rx_vlan_header_insertion =
4178 		    be64_to_cpu(crq->query_capability.number);
4179 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4180 			   adapter->rx_vlan_header_insertion);
4181 		break;
4182 	case MAX_TX_SG_ENTRIES:
4183 		adapter->max_tx_sg_entries =
4184 		    be64_to_cpu(crq->query_capability.number);
4185 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4186 			   adapter->max_tx_sg_entries);
4187 		break;
4188 	case RX_SG_SUPPORTED:
4189 		adapter->rx_sg_supported =
4190 		    be64_to_cpu(crq->query_capability.number);
4191 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4192 			   adapter->rx_sg_supported);
4193 		break;
4194 	case OPT_TX_COMP_SUB_QUEUES:
4195 		adapter->opt_tx_comp_sub_queues =
4196 		    be64_to_cpu(crq->query_capability.number);
4197 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4198 			   adapter->opt_tx_comp_sub_queues);
4199 		break;
4200 	case OPT_RX_COMP_QUEUES:
4201 		adapter->opt_rx_comp_queues =
4202 		    be64_to_cpu(crq->query_capability.number);
4203 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4204 			   adapter->opt_rx_comp_queues);
4205 		break;
4206 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4207 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4208 		    be64_to_cpu(crq->query_capability.number);
4209 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4210 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4211 		break;
4212 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4213 		adapter->opt_tx_entries_per_subcrq =
4214 		    be64_to_cpu(crq->query_capability.number);
4215 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4216 			   adapter->opt_tx_entries_per_subcrq);
4217 		break;
4218 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4219 		adapter->opt_rxba_entries_per_subcrq =
4220 		    be64_to_cpu(crq->query_capability.number);
4221 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4222 			   adapter->opt_rxba_entries_per_subcrq);
4223 		break;
4224 	case TX_RX_DESC_REQ:
4225 		adapter->tx_rx_desc_req = crq->query_capability.number;
4226 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4227 			   adapter->tx_rx_desc_req);
4228 		break;
4229 
4230 	default:
4231 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4232 			   crq->query_capability.capability);
4233 	}
4234 
4235 out:
4236 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4237 		adapter->wait_capability = false;
4238 		ibmvnic_send_req_caps(adapter, 0);
4239 	}
4240 }
4241 
4242 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4243 			       struct ibmvnic_adapter *adapter)
4244 {
4245 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4246 	struct net_device *netdev = adapter->netdev;
4247 	struct device *dev = &adapter->vdev->dev;
4248 	u64 *u64_crq = (u64 *)crq;
4249 	long rc;
4250 
4251 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4252 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4253 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4254 	switch (gen_crq->first) {
4255 	case IBMVNIC_CRQ_INIT_RSP:
4256 		switch (gen_crq->cmd) {
4257 		case IBMVNIC_CRQ_INIT:
4258 			dev_info(dev, "Partner initialized\n");
4259 			adapter->from_passive_init = true;
4260 			adapter->failover_pending = false;
4261 			if (!completion_done(&adapter->init_done)) {
4262 				complete(&adapter->init_done);
4263 				adapter->init_done_rc = -EIO;
4264 			}
4265 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4266 			break;
4267 		case IBMVNIC_CRQ_INIT_COMPLETE:
4268 			dev_info(dev, "Partner initialization complete\n");
4269 			adapter->crq.active = true;
4270 			send_version_xchg(adapter);
4271 			break;
4272 		default:
4273 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4274 		}
4275 		return;
4276 	case IBMVNIC_CRQ_XPORT_EVENT:
4277 		netif_carrier_off(netdev);
4278 		adapter->crq.active = false;
4279 		if (adapter->resetting)
4280 			adapter->force_reset_recovery = true;
4281 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4282 			dev_info(dev, "Migrated, re-enabling adapter\n");
4283 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4284 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4285 			dev_info(dev, "Backing device failover detected\n");
4286 			adapter->failover_pending = true;
4287 		} else {
4288 			/* The adapter lost the connection */
4289 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4290 				gen_crq->cmd);
4291 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4292 		}
4293 		return;
4294 	case IBMVNIC_CRQ_CMD_RSP:
4295 		break;
4296 	default:
4297 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4298 			gen_crq->first);
4299 		return;
4300 	}
4301 
4302 	switch (gen_crq->cmd) {
4303 	case VERSION_EXCHANGE_RSP:
4304 		rc = crq->version_exchange_rsp.rc.code;
4305 		if (rc) {
4306 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4307 			break;
4308 		}
4309 		dev_info(dev, "Partner protocol version is %d\n",
4310 			 crq->version_exchange_rsp.version);
4311 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4312 		    ibmvnic_version)
4313 			ibmvnic_version =
4314 			    be16_to_cpu(crq->version_exchange_rsp.version);
4315 		send_cap_queries(adapter);
4316 		break;
4317 	case QUERY_CAPABILITY_RSP:
4318 		handle_query_cap_rsp(crq, adapter);
4319 		break;
4320 	case QUERY_MAP_RSP:
4321 		handle_query_map_rsp(crq, adapter);
4322 		break;
4323 	case REQUEST_MAP_RSP:
4324 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4325 		complete(&adapter->fw_done);
4326 		break;
4327 	case REQUEST_UNMAP_RSP:
4328 		handle_request_unmap_rsp(crq, adapter);
4329 		break;
4330 	case REQUEST_CAPABILITY_RSP:
4331 		handle_request_cap_rsp(crq, adapter);
4332 		break;
4333 	case LOGIN_RSP:
4334 		netdev_dbg(netdev, "Got Login Response\n");
4335 		handle_login_rsp(crq, adapter);
4336 		break;
4337 	case LOGICAL_LINK_STATE_RSP:
4338 		netdev_dbg(netdev,
4339 			   "Got Logical Link State Response, state: %d rc: %d\n",
4340 			   crq->logical_link_state_rsp.link_state,
4341 			   crq->logical_link_state_rsp.rc.code);
4342 		adapter->logical_link_state =
4343 		    crq->logical_link_state_rsp.link_state;
4344 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4345 		complete(&adapter->init_done);
4346 		break;
4347 	case LINK_STATE_INDICATION:
4348 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4349 		adapter->phys_link_state =
4350 		    crq->link_state_indication.phys_link_state;
4351 		adapter->logical_link_state =
4352 		    crq->link_state_indication.logical_link_state;
4353 		break;
4354 	case CHANGE_MAC_ADDR_RSP:
4355 		netdev_dbg(netdev, "Got MAC address change Response\n");
4356 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4357 		break;
4358 	case ERROR_INDICATION:
4359 		netdev_dbg(netdev, "Got Error Indication\n");
4360 		handle_error_indication(crq, adapter);
4361 		break;
4362 	case REQUEST_STATISTICS_RSP:
4363 		netdev_dbg(netdev, "Got Statistics Response\n");
4364 		complete(&adapter->stats_done);
4365 		break;
4366 	case QUERY_IP_OFFLOAD_RSP:
4367 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4368 		handle_query_ip_offload_rsp(adapter);
4369 		break;
4370 	case MULTICAST_CTRL_RSP:
4371 		netdev_dbg(netdev, "Got multicast control Response\n");
4372 		break;
4373 	case CONTROL_IP_OFFLOAD_RSP:
4374 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4375 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4376 				 sizeof(adapter->ip_offload_ctrl),
4377 				 DMA_TO_DEVICE);
4378 		complete(&adapter->init_done);
4379 		break;
4380 	case COLLECT_FW_TRACE_RSP:
4381 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4382 		complete(&adapter->fw_done);
4383 		break;
4384 	case GET_VPD_SIZE_RSP:
4385 		handle_vpd_size_rsp(crq, adapter);
4386 		break;
4387 	case GET_VPD_RSP:
4388 		handle_vpd_rsp(crq, adapter);
4389 		break;
4390 	default:
4391 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4392 			   gen_crq->cmd);
4393 	}
4394 }
4395 
4396 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4397 {
4398 	struct ibmvnic_adapter *adapter = instance;
4399 
4400 	tasklet_schedule(&adapter->tasklet);
4401 	return IRQ_HANDLED;
4402 }
4403 
4404 static void ibmvnic_tasklet(void *data)
4405 {
4406 	struct ibmvnic_adapter *adapter = data;
4407 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4408 	union ibmvnic_crq *crq;
4409 	unsigned long flags;
4410 	bool done = false;
4411 
4412 	spin_lock_irqsave(&queue->lock, flags);
4413 	while (!done) {
4414 		/* Pull all the valid messages off the CRQ */
4415 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4416 			ibmvnic_handle_crq(crq, adapter);
4417 			crq->generic.first = 0;
4418 		}
4419 
4420 		/* remain in tasklet until all
4421 		 * capabilities responses are received
4422 		 */
4423 		if (!adapter->wait_capability)
4424 			done = true;
4425 	}
4426 	/* if capabilities CRQ's were sent in this tasklet, the following
4427 	 * tasklet must wait until all responses are received
4428 	 */
4429 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4430 		adapter->wait_capability = true;
4431 	spin_unlock_irqrestore(&queue->lock, flags);
4432 }
4433 
4434 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4435 {
4436 	struct vio_dev *vdev = adapter->vdev;
4437 	int rc;
4438 
4439 	do {
4440 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4441 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4442 
4443 	if (rc)
4444 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4445 
4446 	return rc;
4447 }
4448 
4449 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4450 {
4451 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4452 	struct device *dev = &adapter->vdev->dev;
4453 	struct vio_dev *vdev = adapter->vdev;
4454 	int rc;
4455 
4456 	/* Close the CRQ */
4457 	do {
4458 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4459 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4460 
4461 	/* Clean out the queue */
4462 	memset(crq->msgs, 0, PAGE_SIZE);
4463 	crq->cur = 0;
4464 	crq->active = false;
4465 
4466 	/* And re-open it again */
4467 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4468 				crq->msg_token, PAGE_SIZE);
4469 
4470 	if (rc == H_CLOSED)
4471 		/* Adapter is good, but other end is not ready */
4472 		dev_warn(dev, "Partner adapter not ready\n");
4473 	else if (rc != 0)
4474 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4475 
4476 	return rc;
4477 }
4478 
4479 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4480 {
4481 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4482 	struct vio_dev *vdev = adapter->vdev;
4483 	long rc;
4484 
4485 	if (!crq->msgs)
4486 		return;
4487 
4488 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4489 	free_irq(vdev->irq, adapter);
4490 	tasklet_kill(&adapter->tasklet);
4491 	do {
4492 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4493 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4494 
4495 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4496 			 DMA_BIDIRECTIONAL);
4497 	free_page((unsigned long)crq->msgs);
4498 	crq->msgs = NULL;
4499 	crq->active = false;
4500 }
4501 
4502 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4503 {
4504 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4505 	struct device *dev = &adapter->vdev->dev;
4506 	struct vio_dev *vdev = adapter->vdev;
4507 	int rc, retrc = -ENOMEM;
4508 
4509 	if (crq->msgs)
4510 		return 0;
4511 
4512 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4513 	/* Should we allocate more than one page? */
4514 
4515 	if (!crq->msgs)
4516 		return -ENOMEM;
4517 
4518 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4519 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4520 					DMA_BIDIRECTIONAL);
4521 	if (dma_mapping_error(dev, crq->msg_token))
4522 		goto map_failed;
4523 
4524 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4525 				crq->msg_token, PAGE_SIZE);
4526 
4527 	if (rc == H_RESOURCE)
4528 		/* maybe kexecing and resource is busy. try a reset */
4529 		rc = ibmvnic_reset_crq(adapter);
4530 	retrc = rc;
4531 
4532 	if (rc == H_CLOSED) {
4533 		dev_warn(dev, "Partner adapter not ready\n");
4534 	} else if (rc) {
4535 		dev_warn(dev, "Error %d opening adapter\n", rc);
4536 		goto reg_crq_failed;
4537 	}
4538 
4539 	retrc = 0;
4540 
4541 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4542 		     (unsigned long)adapter);
4543 
4544 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4545 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4546 			 adapter);
4547 	if (rc) {
4548 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4549 			vdev->irq, rc);
4550 		goto req_irq_failed;
4551 	}
4552 
4553 	rc = vio_enable_interrupts(vdev);
4554 	if (rc) {
4555 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4556 		goto req_irq_failed;
4557 	}
4558 
4559 	crq->cur = 0;
4560 	spin_lock_init(&crq->lock);
4561 
4562 	return retrc;
4563 
4564 req_irq_failed:
4565 	tasklet_kill(&adapter->tasklet);
4566 	do {
4567 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4568 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4569 reg_crq_failed:
4570 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4571 map_failed:
4572 	free_page((unsigned long)crq->msgs);
4573 	crq->msgs = NULL;
4574 	return retrc;
4575 }
4576 
4577 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4578 {
4579 	struct device *dev = &adapter->vdev->dev;
4580 	unsigned long timeout = msecs_to_jiffies(30000);
4581 	u64 old_num_rx_queues, old_num_tx_queues;
4582 	int rc;
4583 
4584 	adapter->from_passive_init = false;
4585 
4586 	old_num_rx_queues = adapter->req_rx_queues;
4587 	old_num_tx_queues = adapter->req_tx_queues;
4588 
4589 	init_completion(&adapter->init_done);
4590 	adapter->init_done_rc = 0;
4591 	ibmvnic_send_crq_init(adapter);
4592 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4593 		dev_err(dev, "Initialization sequence timed out\n");
4594 		return -1;
4595 	}
4596 
4597 	if (adapter->init_done_rc) {
4598 		release_crq_queue(adapter);
4599 		return adapter->init_done_rc;
4600 	}
4601 
4602 	if (adapter->from_passive_init) {
4603 		adapter->state = VNIC_OPEN;
4604 		adapter->from_passive_init = false;
4605 		return -1;
4606 	}
4607 
4608 	if (adapter->resetting && !adapter->wait_for_reset &&
4609 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
4610 		if (adapter->req_rx_queues != old_num_rx_queues ||
4611 		    adapter->req_tx_queues != old_num_tx_queues) {
4612 			release_sub_crqs(adapter, 0);
4613 			rc = init_sub_crqs(adapter);
4614 		} else {
4615 			rc = reset_sub_crq_queues(adapter);
4616 		}
4617 	} else {
4618 		rc = init_sub_crqs(adapter);
4619 	}
4620 
4621 	if (rc) {
4622 		dev_err(dev, "Initialization of sub crqs failed\n");
4623 		release_crq_queue(adapter);
4624 		return rc;
4625 	}
4626 
4627 	rc = init_sub_crq_irqs(adapter);
4628 	if (rc) {
4629 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4630 		release_crq_queue(adapter);
4631 	}
4632 
4633 	return rc;
4634 }
4635 
4636 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4637 {
4638 	struct device *dev = &adapter->vdev->dev;
4639 	unsigned long timeout = msecs_to_jiffies(30000);
4640 	int rc;
4641 
4642 	adapter->from_passive_init = false;
4643 
4644 	init_completion(&adapter->init_done);
4645 	adapter->init_done_rc = 0;
4646 	ibmvnic_send_crq_init(adapter);
4647 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4648 		dev_err(dev, "Initialization sequence timed out\n");
4649 		return -1;
4650 	}
4651 
4652 	if (adapter->init_done_rc) {
4653 		release_crq_queue(adapter);
4654 		return adapter->init_done_rc;
4655 	}
4656 
4657 	if (adapter->from_passive_init) {
4658 		adapter->state = VNIC_OPEN;
4659 		adapter->from_passive_init = false;
4660 		return -1;
4661 	}
4662 
4663 	rc = init_sub_crqs(adapter);
4664 	if (rc) {
4665 		dev_err(dev, "Initialization of sub crqs failed\n");
4666 		release_crq_queue(adapter);
4667 		return rc;
4668 	}
4669 
4670 	rc = init_sub_crq_irqs(adapter);
4671 	if (rc) {
4672 		dev_err(dev, "Failed to initialize sub crq irqs\n");
4673 		release_crq_queue(adapter);
4674 	}
4675 
4676 	return rc;
4677 }
4678 
4679 static struct device_attribute dev_attr_failover;
4680 
4681 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4682 {
4683 	struct ibmvnic_adapter *adapter;
4684 	struct net_device *netdev;
4685 	unsigned char *mac_addr_p;
4686 	int rc;
4687 
4688 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4689 		dev->unit_address);
4690 
4691 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4692 							VETH_MAC_ADDR, NULL);
4693 	if (!mac_addr_p) {
4694 		dev_err(&dev->dev,
4695 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4696 			__FILE__, __LINE__);
4697 		return 0;
4698 	}
4699 
4700 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4701 				   IBMVNIC_MAX_QUEUES);
4702 	if (!netdev)
4703 		return -ENOMEM;
4704 
4705 	adapter = netdev_priv(netdev);
4706 	adapter->state = VNIC_PROBING;
4707 	dev_set_drvdata(&dev->dev, netdev);
4708 	adapter->vdev = dev;
4709 	adapter->netdev = netdev;
4710 
4711 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
4712 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4713 	netdev->irq = dev->irq;
4714 	netdev->netdev_ops = &ibmvnic_netdev_ops;
4715 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4716 	SET_NETDEV_DEV(netdev, &dev->dev);
4717 
4718 	spin_lock_init(&adapter->stats_lock);
4719 
4720 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4721 	INIT_LIST_HEAD(&adapter->rwi_list);
4722 	mutex_init(&adapter->reset_lock);
4723 	mutex_init(&adapter->rwi_lock);
4724 	adapter->resetting = false;
4725 
4726 	adapter->mac_change_pending = false;
4727 
4728 	do {
4729 		rc = init_crq_queue(adapter);
4730 		if (rc) {
4731 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4732 				rc);
4733 			goto ibmvnic_init_fail;
4734 		}
4735 
4736 		rc = ibmvnic_init(adapter);
4737 		if (rc && rc != EAGAIN)
4738 			goto ibmvnic_init_fail;
4739 	} while (rc == EAGAIN);
4740 
4741 	rc = init_stats_buffers(adapter);
4742 	if (rc)
4743 		goto ibmvnic_init_fail;
4744 
4745 	rc = init_stats_token(adapter);
4746 	if (rc)
4747 		goto ibmvnic_stats_fail;
4748 
4749 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4750 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4751 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4752 
4753 	rc = device_create_file(&dev->dev, &dev_attr_failover);
4754 	if (rc)
4755 		goto ibmvnic_dev_file_err;
4756 
4757 	netif_carrier_off(netdev);
4758 	rc = register_netdev(netdev);
4759 	if (rc) {
4760 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4761 		goto ibmvnic_register_fail;
4762 	}
4763 	dev_info(&dev->dev, "ibmvnic registered\n");
4764 
4765 	adapter->state = VNIC_PROBED;
4766 
4767 	adapter->wait_for_reset = false;
4768 
4769 	return 0;
4770 
4771 ibmvnic_register_fail:
4772 	device_remove_file(&dev->dev, &dev_attr_failover);
4773 
4774 ibmvnic_dev_file_err:
4775 	release_stats_token(adapter);
4776 
4777 ibmvnic_stats_fail:
4778 	release_stats_buffers(adapter);
4779 
4780 ibmvnic_init_fail:
4781 	release_sub_crqs(adapter, 1);
4782 	release_crq_queue(adapter);
4783 	free_netdev(netdev);
4784 
4785 	return rc;
4786 }
4787 
4788 static int ibmvnic_remove(struct vio_dev *dev)
4789 {
4790 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
4791 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4792 
4793 	adapter->state = VNIC_REMOVING;
4794 	unregister_netdev(netdev);
4795 	mutex_lock(&adapter->reset_lock);
4796 
4797 	release_resources(adapter);
4798 	release_sub_crqs(adapter, 1);
4799 	release_crq_queue(adapter);
4800 
4801 	release_stats_token(adapter);
4802 	release_stats_buffers(adapter);
4803 
4804 	adapter->state = VNIC_REMOVED;
4805 
4806 	mutex_unlock(&adapter->reset_lock);
4807 	device_remove_file(&dev->dev, &dev_attr_failover);
4808 	free_netdev(netdev);
4809 	dev_set_drvdata(&dev->dev, NULL);
4810 
4811 	return 0;
4812 }
4813 
4814 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4815 			      const char *buf, size_t count)
4816 {
4817 	struct net_device *netdev = dev_get_drvdata(dev);
4818 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4819 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4820 	__be64 session_token;
4821 	long rc;
4822 
4823 	if (!sysfs_streq(buf, "1"))
4824 		return -EINVAL;
4825 
4826 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4827 			 H_GET_SESSION_TOKEN, 0, 0, 0);
4828 	if (rc) {
4829 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4830 			   rc);
4831 		return -EINVAL;
4832 	}
4833 
4834 	session_token = (__be64)retbuf[0];
4835 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4836 		   be64_to_cpu(session_token));
4837 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4838 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
4839 	if (rc) {
4840 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4841 			   rc);
4842 		return -EINVAL;
4843 	}
4844 
4845 	return count;
4846 }
4847 
4848 static DEVICE_ATTR_WO(failover);
4849 
4850 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4851 {
4852 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4853 	struct ibmvnic_adapter *adapter;
4854 	struct iommu_table *tbl;
4855 	unsigned long ret = 0;
4856 	int i;
4857 
4858 	tbl = get_iommu_table_base(&vdev->dev);
4859 
4860 	/* netdev inits at probe time along with the structures we need below*/
4861 	if (!netdev)
4862 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4863 
4864 	adapter = netdev_priv(netdev);
4865 
4866 	ret += PAGE_SIZE; /* the crq message queue */
4867 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4868 
4869 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4870 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4871 
4872 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4873 	     i++)
4874 		ret += adapter->rx_pool[i].size *
4875 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4876 
4877 	return ret;
4878 }
4879 
4880 static int ibmvnic_resume(struct device *dev)
4881 {
4882 	struct net_device *netdev = dev_get_drvdata(dev);
4883 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4884 
4885 	if (adapter->state != VNIC_OPEN)
4886 		return 0;
4887 
4888 	tasklet_schedule(&adapter->tasklet);
4889 
4890 	return 0;
4891 }
4892 
4893 static const struct vio_device_id ibmvnic_device_table[] = {
4894 	{"network", "IBM,vnic"},
4895 	{"", "" }
4896 };
4897 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4898 
4899 static const struct dev_pm_ops ibmvnic_pm_ops = {
4900 	.resume = ibmvnic_resume
4901 };
4902 
4903 static struct vio_driver ibmvnic_driver = {
4904 	.id_table       = ibmvnic_device_table,
4905 	.probe          = ibmvnic_probe,
4906 	.remove         = ibmvnic_remove,
4907 	.get_desired_dma = ibmvnic_get_desired_dma,
4908 	.name		= ibmvnic_driver_name,
4909 	.pm		= &ibmvnic_pm_ops,
4910 };
4911 
4912 /* module functions */
4913 static int __init ibmvnic_module_init(void)
4914 {
4915 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4916 		IBMVNIC_DRIVER_VERSION);
4917 
4918 	return vio_register_driver(&ibmvnic_driver);
4919 }
4920 
4921 static void __exit ibmvnic_module_exit(void)
4922 {
4923 	vio_unregister_driver(&ibmvnic_driver);
4924 }
4925 
4926 module_init(ibmvnic_module_init);
4927 module_exit(ibmvnic_module_exit);
4928