xref: /linux/drivers/net/ethernet/ibm/ibmvnic.c (revision e3b9626f09d429788d929c9b9000a069fcfc056e)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /*                                                                        */
4 /*  IBM System i and System p Virtual NIC Device Driver                   */
5 /*  Copyright (C) 2014 IBM Corp.                                          */
6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
9 /*                                                                        */
10 /*                                                                        */
11 /* This module contains the implementation of a virtual ethernet device   */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
15 /*									   */
16 /* Messages are passed between the VNIC driver and the VNIC server using  */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
20 /* are used by the driver to notify the server that a packet is           */
21 /* ready for transmission or that a buffer has been added to receive a    */
22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
23 /* driver that a packet transmission has been completed or that a packet  */
24 /* has been received and placed in a waiting buffer.                      */
25 /*                                                                        */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
28 /* or receive has been completed, the VNIC driver is required to use      */
29 /* "long term mapping". This entails that large, continuous DMA mapped    */
30 /* buffers are allocated on driver initialization and these buffers are   */
31 /* then continuously reused to pass skbs to and from the VNIC server.     */
32 /*                                                                        */
33 /**************************************************************************/
34 
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69 
70 #include "ibmvnic.h"
71 
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74 
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79 
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static int ibmvnic_remove(struct vio_dev *);
82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
88 		       union sub_crq *sub_crq);
89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
91 static int enable_scrq_irq(struct ibmvnic_adapter *,
92 			   struct ibmvnic_sub_crq_queue *);
93 static int disable_scrq_irq(struct ibmvnic_adapter *,
94 			    struct ibmvnic_sub_crq_queue *);
95 static int pending_scrq(struct ibmvnic_adapter *,
96 			struct ibmvnic_sub_crq_queue *);
97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
98 					struct ibmvnic_sub_crq_queue *);
99 static int ibmvnic_poll(struct napi_struct *napi, int data);
100 static void send_map_query(struct ibmvnic_adapter *adapter);
101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
103 static int send_login(struct ibmvnic_adapter *adapter);
104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
105 static int init_sub_crqs(struct ibmvnic_adapter *);
106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
107 static int ibmvnic_reset_init(struct ibmvnic_adapter *, bool reset);
108 static void release_crq_queue(struct ibmvnic_adapter *);
109 static int __ibmvnic_set_mac(struct net_device *, u8 *);
110 static int init_crq_queue(struct ibmvnic_adapter *adapter);
111 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
112 
113 struct ibmvnic_stat {
114 	char name[ETH_GSTRING_LEN];
115 	int offset;
116 };
117 
118 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
119 			     offsetof(struct ibmvnic_statistics, stat))
120 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
121 
122 static const struct ibmvnic_stat ibmvnic_stats[] = {
123 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
124 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
125 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
126 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
127 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
128 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
129 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
130 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
131 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
132 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
133 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
134 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
135 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
136 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
137 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
138 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
139 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
140 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
141 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
142 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
143 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
144 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
145 };
146 
147 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
148 			  unsigned long length, unsigned long *number,
149 			  unsigned long *irq)
150 {
151 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
152 	long rc;
153 
154 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
155 	*number = retbuf[0];
156 	*irq = retbuf[1];
157 
158 	return rc;
159 }
160 
161 /**
162  * ibmvnic_wait_for_completion - Check device state and wait for completion
163  * @adapter: private device data
164  * @comp_done: completion structure to wait for
165  * @timeout: time to wait in milliseconds
166  *
167  * Wait for a completion signal or until the timeout limit is reached
168  * while checking that the device is still active.
169  */
170 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
171 				       struct completion *comp_done,
172 				       unsigned long timeout)
173 {
174 	struct net_device *netdev;
175 	unsigned long div_timeout;
176 	u8 retry;
177 
178 	netdev = adapter->netdev;
179 	retry = 5;
180 	div_timeout = msecs_to_jiffies(timeout / retry);
181 	while (true) {
182 		if (!adapter->crq.active) {
183 			netdev_err(netdev, "Device down!\n");
184 			return -ENODEV;
185 		}
186 		if (!retry--)
187 			break;
188 		if (wait_for_completion_timeout(comp_done, div_timeout))
189 			return 0;
190 	}
191 	netdev_err(netdev, "Operation timed out.\n");
192 	return -ETIMEDOUT;
193 }
194 
195 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
196 				struct ibmvnic_long_term_buff *ltb, int size)
197 {
198 	struct device *dev = &adapter->vdev->dev;
199 	int rc;
200 
201 	ltb->size = size;
202 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
203 				       GFP_KERNEL);
204 
205 	if (!ltb->buff) {
206 		dev_err(dev, "Couldn't alloc long term buffer\n");
207 		return -ENOMEM;
208 	}
209 	ltb->map_id = adapter->map_id;
210 	adapter->map_id++;
211 
212 	mutex_lock(&adapter->fw_lock);
213 	adapter->fw_done_rc = 0;
214 	reinit_completion(&adapter->fw_done);
215 	rc = send_request_map(adapter, ltb->addr,
216 			      ltb->size, ltb->map_id);
217 	if (rc) {
218 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
219 		mutex_unlock(&adapter->fw_lock);
220 		return rc;
221 	}
222 
223 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
224 	if (rc) {
225 		dev_err(dev,
226 			"Long term map request aborted or timed out,rc = %d\n",
227 			rc);
228 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
229 		mutex_unlock(&adapter->fw_lock);
230 		return rc;
231 	}
232 
233 	if (adapter->fw_done_rc) {
234 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
235 			adapter->fw_done_rc);
236 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
237 		mutex_unlock(&adapter->fw_lock);
238 		return -1;
239 	}
240 	mutex_unlock(&adapter->fw_lock);
241 	return 0;
242 }
243 
244 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
245 				struct ibmvnic_long_term_buff *ltb)
246 {
247 	struct device *dev = &adapter->vdev->dev;
248 
249 	if (!ltb->buff)
250 		return;
251 
252 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
253 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
254 		send_request_unmap(adapter, ltb->map_id);
255 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
256 }
257 
258 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
259 				struct ibmvnic_long_term_buff *ltb)
260 {
261 	struct device *dev = &adapter->vdev->dev;
262 	int rc;
263 
264 	memset(ltb->buff, 0, ltb->size);
265 
266 	mutex_lock(&adapter->fw_lock);
267 	adapter->fw_done_rc = 0;
268 
269 	reinit_completion(&adapter->fw_done);
270 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
271 	if (rc) {
272 		mutex_unlock(&adapter->fw_lock);
273 		return rc;
274 	}
275 
276 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
277 	if (rc) {
278 		dev_info(dev,
279 			 "Reset failed, long term map request timed out or aborted\n");
280 		mutex_unlock(&adapter->fw_lock);
281 		return rc;
282 	}
283 
284 	if (adapter->fw_done_rc) {
285 		dev_info(dev,
286 			 "Reset failed, attempting to free and reallocate buffer\n");
287 		free_long_term_buff(adapter, ltb);
288 		mutex_unlock(&adapter->fw_lock);
289 		return alloc_long_term_buff(adapter, ltb, ltb->size);
290 	}
291 	mutex_unlock(&adapter->fw_lock);
292 	return 0;
293 }
294 
295 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
296 {
297 	int i;
298 
299 	for (i = 0; i < adapter->num_active_rx_pools; i++)
300 		adapter->rx_pool[i].active = 0;
301 }
302 
303 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
304 			      struct ibmvnic_rx_pool *pool)
305 {
306 	int count = pool->size - atomic_read(&pool->available);
307 	u64 handle = adapter->rx_scrq[pool->index]->handle;
308 	struct device *dev = &adapter->vdev->dev;
309 	int buffers_added = 0;
310 	unsigned long lpar_rc;
311 	union sub_crq sub_crq;
312 	struct sk_buff *skb;
313 	unsigned int offset;
314 	dma_addr_t dma_addr;
315 	unsigned char *dst;
316 	int shift = 0;
317 	int index;
318 	int i;
319 
320 	if (!pool->active)
321 		return;
322 
323 	for (i = 0; i < count; ++i) {
324 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
325 		if (!skb) {
326 			dev_err(dev, "Couldn't replenish rx buff\n");
327 			adapter->replenish_no_mem++;
328 			break;
329 		}
330 
331 		index = pool->free_map[pool->next_free];
332 
333 		if (pool->rx_buff[index].skb)
334 			dev_err(dev, "Inconsistent free_map!\n");
335 
336 		/* Copy the skb to the long term mapped DMA buffer */
337 		offset = index * pool->buff_size;
338 		dst = pool->long_term_buff.buff + offset;
339 		memset(dst, 0, pool->buff_size);
340 		dma_addr = pool->long_term_buff.addr + offset;
341 		pool->rx_buff[index].data = dst;
342 
343 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
344 		pool->rx_buff[index].dma = dma_addr;
345 		pool->rx_buff[index].skb = skb;
346 		pool->rx_buff[index].pool_index = pool->index;
347 		pool->rx_buff[index].size = pool->buff_size;
348 
349 		memset(&sub_crq, 0, sizeof(sub_crq));
350 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
351 		sub_crq.rx_add.correlator =
352 		    cpu_to_be64((u64)&pool->rx_buff[index]);
353 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
354 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
355 
356 		/* The length field of the sCRQ is defined to be 24 bits so the
357 		 * buffer size needs to be left shifted by a byte before it is
358 		 * converted to big endian to prevent the last byte from being
359 		 * truncated.
360 		 */
361 #ifdef __LITTLE_ENDIAN__
362 		shift = 8;
363 #endif
364 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
365 
366 		lpar_rc = send_subcrq(adapter, handle, &sub_crq);
367 		if (lpar_rc != H_SUCCESS)
368 			goto failure;
369 
370 		buffers_added++;
371 		adapter->replenish_add_buff_success++;
372 		pool->next_free = (pool->next_free + 1) % pool->size;
373 	}
374 	atomic_add(buffers_added, &pool->available);
375 	return;
376 
377 failure:
378 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
379 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
380 	pool->free_map[pool->next_free] = index;
381 	pool->rx_buff[index].skb = NULL;
382 
383 	dev_kfree_skb_any(skb);
384 	adapter->replenish_add_buff_failure++;
385 	atomic_add(buffers_added, &pool->available);
386 
387 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
388 		/* Disable buffer pool replenishment and report carrier off if
389 		 * queue is closed or pending failover.
390 		 * Firmware guarantees that a signal will be sent to the
391 		 * driver, triggering a reset.
392 		 */
393 		deactivate_rx_pools(adapter);
394 		netif_carrier_off(adapter->netdev);
395 	}
396 }
397 
398 static void replenish_pools(struct ibmvnic_adapter *adapter)
399 {
400 	int i;
401 
402 	adapter->replenish_task_cycles++;
403 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
404 		if (adapter->rx_pool[i].active)
405 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
406 	}
407 }
408 
409 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
410 {
411 	kfree(adapter->tx_stats_buffers);
412 	kfree(adapter->rx_stats_buffers);
413 	adapter->tx_stats_buffers = NULL;
414 	adapter->rx_stats_buffers = NULL;
415 }
416 
417 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
418 {
419 	adapter->tx_stats_buffers =
420 				kcalloc(IBMVNIC_MAX_QUEUES,
421 					sizeof(struct ibmvnic_tx_queue_stats),
422 					GFP_KERNEL);
423 	if (!adapter->tx_stats_buffers)
424 		return -ENOMEM;
425 
426 	adapter->rx_stats_buffers =
427 				kcalloc(IBMVNIC_MAX_QUEUES,
428 					sizeof(struct ibmvnic_rx_queue_stats),
429 					GFP_KERNEL);
430 	if (!adapter->rx_stats_buffers)
431 		return -ENOMEM;
432 
433 	return 0;
434 }
435 
436 static void release_stats_token(struct ibmvnic_adapter *adapter)
437 {
438 	struct device *dev = &adapter->vdev->dev;
439 
440 	if (!adapter->stats_token)
441 		return;
442 
443 	dma_unmap_single(dev, adapter->stats_token,
444 			 sizeof(struct ibmvnic_statistics),
445 			 DMA_FROM_DEVICE);
446 	adapter->stats_token = 0;
447 }
448 
449 static int init_stats_token(struct ibmvnic_adapter *adapter)
450 {
451 	struct device *dev = &adapter->vdev->dev;
452 	dma_addr_t stok;
453 
454 	stok = dma_map_single(dev, &adapter->stats,
455 			      sizeof(struct ibmvnic_statistics),
456 			      DMA_FROM_DEVICE);
457 	if (dma_mapping_error(dev, stok)) {
458 		dev_err(dev, "Couldn't map stats buffer\n");
459 		return -1;
460 	}
461 
462 	adapter->stats_token = stok;
463 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
464 	return 0;
465 }
466 
467 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
468 {
469 	struct ibmvnic_rx_pool *rx_pool;
470 	u64 buff_size;
471 	int rx_scrqs;
472 	int i, j, rc;
473 
474 	buff_size = adapter->cur_rx_buf_sz;
475 	rx_scrqs = adapter->num_active_rx_pools;
476 	for (i = 0; i < rx_scrqs; i++) {
477 		rx_pool = &adapter->rx_pool[i];
478 
479 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
480 
481 		if (rx_pool->buff_size != buff_size) {
482 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
483 			rx_pool->buff_size = buff_size;
484 			rc = alloc_long_term_buff(adapter,
485 						  &rx_pool->long_term_buff,
486 						  rx_pool->size *
487 						  rx_pool->buff_size);
488 		} else {
489 			rc = reset_long_term_buff(adapter,
490 						  &rx_pool->long_term_buff);
491 		}
492 
493 		if (rc)
494 			return rc;
495 
496 		for (j = 0; j < rx_pool->size; j++)
497 			rx_pool->free_map[j] = j;
498 
499 		memset(rx_pool->rx_buff, 0,
500 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
501 
502 		atomic_set(&rx_pool->available, 0);
503 		rx_pool->next_alloc = 0;
504 		rx_pool->next_free = 0;
505 		rx_pool->active = 1;
506 	}
507 
508 	return 0;
509 }
510 
511 static void release_rx_pools(struct ibmvnic_adapter *adapter)
512 {
513 	struct ibmvnic_rx_pool *rx_pool;
514 	int i, j;
515 
516 	if (!adapter->rx_pool)
517 		return;
518 
519 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
520 		rx_pool = &adapter->rx_pool[i];
521 
522 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
523 
524 		kfree(rx_pool->free_map);
525 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
526 
527 		if (!rx_pool->rx_buff)
528 			continue;
529 
530 		for (j = 0; j < rx_pool->size; j++) {
531 			if (rx_pool->rx_buff[j].skb) {
532 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
533 				rx_pool->rx_buff[j].skb = NULL;
534 			}
535 		}
536 
537 		kfree(rx_pool->rx_buff);
538 	}
539 
540 	kfree(adapter->rx_pool);
541 	adapter->rx_pool = NULL;
542 	adapter->num_active_rx_pools = 0;
543 }
544 
545 static int init_rx_pools(struct net_device *netdev)
546 {
547 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
548 	struct device *dev = &adapter->vdev->dev;
549 	struct ibmvnic_rx_pool *rx_pool;
550 	int rxadd_subcrqs;
551 	u64 buff_size;
552 	int i, j;
553 
554 	rxadd_subcrqs = adapter->num_active_rx_scrqs;
555 	buff_size = adapter->cur_rx_buf_sz;
556 
557 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
558 				   sizeof(struct ibmvnic_rx_pool),
559 				   GFP_KERNEL);
560 	if (!adapter->rx_pool) {
561 		dev_err(dev, "Failed to allocate rx pools\n");
562 		return -1;
563 	}
564 
565 	adapter->num_active_rx_pools = rxadd_subcrqs;
566 
567 	for (i = 0; i < rxadd_subcrqs; i++) {
568 		rx_pool = &adapter->rx_pool[i];
569 
570 		netdev_dbg(adapter->netdev,
571 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
572 			   i, adapter->req_rx_add_entries_per_subcrq,
573 			   buff_size);
574 
575 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
576 		rx_pool->index = i;
577 		rx_pool->buff_size = buff_size;
578 		rx_pool->active = 1;
579 
580 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
581 					    GFP_KERNEL);
582 		if (!rx_pool->free_map) {
583 			release_rx_pools(adapter);
584 			return -1;
585 		}
586 
587 		rx_pool->rx_buff = kcalloc(rx_pool->size,
588 					   sizeof(struct ibmvnic_rx_buff),
589 					   GFP_KERNEL);
590 		if (!rx_pool->rx_buff) {
591 			dev_err(dev, "Couldn't alloc rx buffers\n");
592 			release_rx_pools(adapter);
593 			return -1;
594 		}
595 
596 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
597 					 rx_pool->size * rx_pool->buff_size)) {
598 			release_rx_pools(adapter);
599 			return -1;
600 		}
601 
602 		for (j = 0; j < rx_pool->size; ++j)
603 			rx_pool->free_map[j] = j;
604 
605 		atomic_set(&rx_pool->available, 0);
606 		rx_pool->next_alloc = 0;
607 		rx_pool->next_free = 0;
608 	}
609 
610 	return 0;
611 }
612 
613 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
614 			     struct ibmvnic_tx_pool *tx_pool)
615 {
616 	int rc, i;
617 
618 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
619 	if (rc)
620 		return rc;
621 
622 	memset(tx_pool->tx_buff, 0,
623 	       tx_pool->num_buffers *
624 	       sizeof(struct ibmvnic_tx_buff));
625 
626 	for (i = 0; i < tx_pool->num_buffers; i++)
627 		tx_pool->free_map[i] = i;
628 
629 	tx_pool->consumer_index = 0;
630 	tx_pool->producer_index = 0;
631 
632 	return 0;
633 }
634 
635 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
636 {
637 	int tx_scrqs;
638 	int i, rc;
639 
640 	tx_scrqs = adapter->num_active_tx_pools;
641 	for (i = 0; i < tx_scrqs; i++) {
642 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
643 		if (rc)
644 			return rc;
645 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
646 		if (rc)
647 			return rc;
648 	}
649 
650 	return 0;
651 }
652 
653 static void release_vpd_data(struct ibmvnic_adapter *adapter)
654 {
655 	if (!adapter->vpd)
656 		return;
657 
658 	kfree(adapter->vpd->buff);
659 	kfree(adapter->vpd);
660 
661 	adapter->vpd = NULL;
662 }
663 
664 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
665 				struct ibmvnic_tx_pool *tx_pool)
666 {
667 	kfree(tx_pool->tx_buff);
668 	kfree(tx_pool->free_map);
669 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
670 }
671 
672 static void release_tx_pools(struct ibmvnic_adapter *adapter)
673 {
674 	int i;
675 
676 	if (!adapter->tx_pool)
677 		return;
678 
679 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
680 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
681 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
682 	}
683 
684 	kfree(adapter->tx_pool);
685 	adapter->tx_pool = NULL;
686 	kfree(adapter->tso_pool);
687 	adapter->tso_pool = NULL;
688 	adapter->num_active_tx_pools = 0;
689 }
690 
691 static int init_one_tx_pool(struct net_device *netdev,
692 			    struct ibmvnic_tx_pool *tx_pool,
693 			    int num_entries, int buf_size)
694 {
695 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
696 	int i;
697 
698 	tx_pool->tx_buff = kcalloc(num_entries,
699 				   sizeof(struct ibmvnic_tx_buff),
700 				   GFP_KERNEL);
701 	if (!tx_pool->tx_buff)
702 		return -1;
703 
704 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
705 				 num_entries * buf_size))
706 		return -1;
707 
708 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
709 	if (!tx_pool->free_map)
710 		return -1;
711 
712 	for (i = 0; i < num_entries; i++)
713 		tx_pool->free_map[i] = i;
714 
715 	tx_pool->consumer_index = 0;
716 	tx_pool->producer_index = 0;
717 	tx_pool->num_buffers = num_entries;
718 	tx_pool->buf_size = buf_size;
719 
720 	return 0;
721 }
722 
723 static int init_tx_pools(struct net_device *netdev)
724 {
725 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
726 	int tx_subcrqs;
727 	int i, rc;
728 
729 	tx_subcrqs = adapter->num_active_tx_scrqs;
730 	adapter->tx_pool = kcalloc(tx_subcrqs,
731 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
732 	if (!adapter->tx_pool)
733 		return -1;
734 
735 	adapter->tso_pool = kcalloc(tx_subcrqs,
736 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
737 	if (!adapter->tso_pool)
738 		return -1;
739 
740 	adapter->num_active_tx_pools = tx_subcrqs;
741 
742 	for (i = 0; i < tx_subcrqs; i++) {
743 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
744 				      adapter->req_tx_entries_per_subcrq,
745 				      adapter->req_mtu + VLAN_HLEN);
746 		if (rc) {
747 			release_tx_pools(adapter);
748 			return rc;
749 		}
750 
751 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
752 				      IBMVNIC_TSO_BUFS,
753 				      IBMVNIC_TSO_BUF_SZ);
754 		if (rc) {
755 			release_tx_pools(adapter);
756 			return rc;
757 		}
758 	}
759 
760 	return 0;
761 }
762 
763 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
764 {
765 	int i;
766 
767 	if (adapter->napi_enabled)
768 		return;
769 
770 	for (i = 0; i < adapter->req_rx_queues; i++)
771 		napi_enable(&adapter->napi[i]);
772 
773 	adapter->napi_enabled = true;
774 }
775 
776 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
777 {
778 	int i;
779 
780 	if (!adapter->napi_enabled)
781 		return;
782 
783 	for (i = 0; i < adapter->req_rx_queues; i++) {
784 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
785 		napi_disable(&adapter->napi[i]);
786 	}
787 
788 	adapter->napi_enabled = false;
789 }
790 
791 static int init_napi(struct ibmvnic_adapter *adapter)
792 {
793 	int i;
794 
795 	adapter->napi = kcalloc(adapter->req_rx_queues,
796 				sizeof(struct napi_struct), GFP_KERNEL);
797 	if (!adapter->napi)
798 		return -ENOMEM;
799 
800 	for (i = 0; i < adapter->req_rx_queues; i++) {
801 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
802 		netif_napi_add(adapter->netdev, &adapter->napi[i],
803 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
804 	}
805 
806 	adapter->num_active_rx_napi = adapter->req_rx_queues;
807 	return 0;
808 }
809 
810 static void release_napi(struct ibmvnic_adapter *adapter)
811 {
812 	int i;
813 
814 	if (!adapter->napi)
815 		return;
816 
817 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
818 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
819 		netif_napi_del(&adapter->napi[i]);
820 	}
821 
822 	kfree(adapter->napi);
823 	adapter->napi = NULL;
824 	adapter->num_active_rx_napi = 0;
825 	adapter->napi_enabled = false;
826 }
827 
828 static int ibmvnic_login(struct net_device *netdev)
829 {
830 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
831 	unsigned long timeout = msecs_to_jiffies(30000);
832 	int retry_count = 0;
833 	int retries = 10;
834 	bool retry;
835 	int rc;
836 
837 	do {
838 		retry = false;
839 		if (retry_count > retries) {
840 			netdev_warn(netdev, "Login attempts exceeded\n");
841 			return -1;
842 		}
843 
844 		adapter->init_done_rc = 0;
845 		reinit_completion(&adapter->init_done);
846 		rc = send_login(adapter);
847 		if (rc) {
848 			netdev_warn(netdev, "Unable to login\n");
849 			return rc;
850 		}
851 
852 		if (!wait_for_completion_timeout(&adapter->init_done,
853 						 timeout)) {
854 			netdev_warn(netdev, "Login timed out, retrying...\n");
855 			retry = true;
856 			adapter->init_done_rc = 0;
857 			retry_count++;
858 			continue;
859 		}
860 
861 		if (adapter->init_done_rc == ABORTED) {
862 			netdev_warn(netdev, "Login aborted, retrying...\n");
863 			retry = true;
864 			adapter->init_done_rc = 0;
865 			retry_count++;
866 			/* FW or device may be busy, so
867 			 * wait a bit before retrying login
868 			 */
869 			msleep(500);
870 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
871 			retry_count++;
872 			release_sub_crqs(adapter, 1);
873 
874 			retry = true;
875 			netdev_dbg(netdev,
876 				   "Received partial success, retrying...\n");
877 			adapter->init_done_rc = 0;
878 			reinit_completion(&adapter->init_done);
879 			send_cap_queries(adapter);
880 			if (!wait_for_completion_timeout(&adapter->init_done,
881 							 timeout)) {
882 				netdev_warn(netdev,
883 					    "Capabilities query timed out\n");
884 				return -1;
885 			}
886 
887 			rc = init_sub_crqs(adapter);
888 			if (rc) {
889 				netdev_warn(netdev,
890 					    "SCRQ initialization failed\n");
891 				return -1;
892 			}
893 
894 			rc = init_sub_crq_irqs(adapter);
895 			if (rc) {
896 				netdev_warn(netdev,
897 					    "SCRQ irq initialization failed\n");
898 				return -1;
899 			}
900 		} else if (adapter->init_done_rc) {
901 			netdev_warn(netdev, "Adapter login failed\n");
902 			return -1;
903 		}
904 	} while (retry);
905 
906 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
907 
908 	return 0;
909 }
910 
911 static void release_login_buffer(struct ibmvnic_adapter *adapter)
912 {
913 	kfree(adapter->login_buf);
914 	adapter->login_buf = NULL;
915 }
916 
917 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
918 {
919 	kfree(adapter->login_rsp_buf);
920 	adapter->login_rsp_buf = NULL;
921 }
922 
923 static void release_resources(struct ibmvnic_adapter *adapter)
924 {
925 	release_vpd_data(adapter);
926 
927 	release_tx_pools(adapter);
928 	release_rx_pools(adapter);
929 
930 	release_napi(adapter);
931 	release_login_rsp_buffer(adapter);
932 }
933 
934 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
935 {
936 	struct net_device *netdev = adapter->netdev;
937 	unsigned long timeout = msecs_to_jiffies(30000);
938 	union ibmvnic_crq crq;
939 	bool resend;
940 	int rc;
941 
942 	netdev_dbg(netdev, "setting link state %d\n", link_state);
943 
944 	memset(&crq, 0, sizeof(crq));
945 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
946 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
947 	crq.logical_link_state.link_state = link_state;
948 
949 	do {
950 		resend = false;
951 
952 		reinit_completion(&adapter->init_done);
953 		rc = ibmvnic_send_crq(adapter, &crq);
954 		if (rc) {
955 			netdev_err(netdev, "Failed to set link state\n");
956 			return rc;
957 		}
958 
959 		if (!wait_for_completion_timeout(&adapter->init_done,
960 						 timeout)) {
961 			netdev_err(netdev, "timeout setting link state\n");
962 			return -1;
963 		}
964 
965 		if (adapter->init_done_rc == PARTIALSUCCESS) {
966 			/* Partuial success, delay and re-send */
967 			mdelay(1000);
968 			resend = true;
969 		} else if (adapter->init_done_rc) {
970 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
971 				    adapter->init_done_rc);
972 			return adapter->init_done_rc;
973 		}
974 	} while (resend);
975 
976 	return 0;
977 }
978 
979 static int set_real_num_queues(struct net_device *netdev)
980 {
981 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
982 	int rc;
983 
984 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
985 		   adapter->req_tx_queues, adapter->req_rx_queues);
986 
987 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
988 	if (rc) {
989 		netdev_err(netdev, "failed to set the number of tx queues\n");
990 		return rc;
991 	}
992 
993 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
994 	if (rc)
995 		netdev_err(netdev, "failed to set the number of rx queues\n");
996 
997 	return rc;
998 }
999 
1000 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1001 {
1002 	struct device *dev = &adapter->vdev->dev;
1003 	union ibmvnic_crq crq;
1004 	int len = 0;
1005 	int rc;
1006 
1007 	if (adapter->vpd->buff)
1008 		len = adapter->vpd->len;
1009 
1010 	mutex_lock(&adapter->fw_lock);
1011 	adapter->fw_done_rc = 0;
1012 	reinit_completion(&adapter->fw_done);
1013 
1014 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1015 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1016 	rc = ibmvnic_send_crq(adapter, &crq);
1017 	if (rc) {
1018 		mutex_unlock(&adapter->fw_lock);
1019 		return rc;
1020 	}
1021 
1022 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1023 	if (rc) {
1024 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1025 		mutex_unlock(&adapter->fw_lock);
1026 		return rc;
1027 	}
1028 	mutex_unlock(&adapter->fw_lock);
1029 
1030 	if (!adapter->vpd->len)
1031 		return -ENODATA;
1032 
1033 	if (!adapter->vpd->buff)
1034 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1035 	else if (adapter->vpd->len != len)
1036 		adapter->vpd->buff =
1037 			krealloc(adapter->vpd->buff,
1038 				 adapter->vpd->len, GFP_KERNEL);
1039 
1040 	if (!adapter->vpd->buff) {
1041 		dev_err(dev, "Could allocate VPD buffer\n");
1042 		return -ENOMEM;
1043 	}
1044 
1045 	adapter->vpd->dma_addr =
1046 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1047 			       DMA_FROM_DEVICE);
1048 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1049 		dev_err(dev, "Could not map VPD buffer\n");
1050 		kfree(adapter->vpd->buff);
1051 		adapter->vpd->buff = NULL;
1052 		return -ENOMEM;
1053 	}
1054 
1055 	mutex_lock(&adapter->fw_lock);
1056 	adapter->fw_done_rc = 0;
1057 	reinit_completion(&adapter->fw_done);
1058 
1059 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1060 	crq.get_vpd.cmd = GET_VPD;
1061 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1062 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1063 	rc = ibmvnic_send_crq(adapter, &crq);
1064 	if (rc) {
1065 		kfree(adapter->vpd->buff);
1066 		adapter->vpd->buff = NULL;
1067 		mutex_unlock(&adapter->fw_lock);
1068 		return rc;
1069 	}
1070 
1071 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1072 	if (rc) {
1073 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1074 		kfree(adapter->vpd->buff);
1075 		adapter->vpd->buff = NULL;
1076 		mutex_unlock(&adapter->fw_lock);
1077 		return rc;
1078 	}
1079 
1080 	mutex_unlock(&adapter->fw_lock);
1081 	return 0;
1082 }
1083 
1084 static int init_resources(struct ibmvnic_adapter *adapter)
1085 {
1086 	struct net_device *netdev = adapter->netdev;
1087 	int rc;
1088 
1089 	rc = set_real_num_queues(netdev);
1090 	if (rc)
1091 		return rc;
1092 
1093 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1094 	if (!adapter->vpd)
1095 		return -ENOMEM;
1096 
1097 	/* Vital Product Data (VPD) */
1098 	rc = ibmvnic_get_vpd(adapter);
1099 	if (rc) {
1100 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1101 		return rc;
1102 	}
1103 
1104 	adapter->map_id = 1;
1105 
1106 	rc = init_napi(adapter);
1107 	if (rc)
1108 		return rc;
1109 
1110 	send_map_query(adapter);
1111 
1112 	rc = init_rx_pools(netdev);
1113 	if (rc)
1114 		return rc;
1115 
1116 	rc = init_tx_pools(netdev);
1117 	return rc;
1118 }
1119 
1120 static int __ibmvnic_open(struct net_device *netdev)
1121 {
1122 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1123 	enum vnic_state prev_state = adapter->state;
1124 	int i, rc;
1125 
1126 	adapter->state = VNIC_OPENING;
1127 	replenish_pools(adapter);
1128 	ibmvnic_napi_enable(adapter);
1129 
1130 	/* We're ready to receive frames, enable the sub-crq interrupts and
1131 	 * set the logical link state to up
1132 	 */
1133 	for (i = 0; i < adapter->req_rx_queues; i++) {
1134 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1135 		if (prev_state == VNIC_CLOSED)
1136 			enable_irq(adapter->rx_scrq[i]->irq);
1137 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1138 	}
1139 
1140 	for (i = 0; i < adapter->req_tx_queues; i++) {
1141 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1142 		if (prev_state == VNIC_CLOSED)
1143 			enable_irq(adapter->tx_scrq[i]->irq);
1144 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1145 	}
1146 
1147 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1148 	if (rc) {
1149 		for (i = 0; i < adapter->req_rx_queues; i++)
1150 			napi_disable(&adapter->napi[i]);
1151 		release_resources(adapter);
1152 		return rc;
1153 	}
1154 
1155 	netif_tx_start_all_queues(netdev);
1156 
1157 	if (prev_state == VNIC_CLOSED) {
1158 		for (i = 0; i < adapter->req_rx_queues; i++)
1159 			napi_schedule(&adapter->napi[i]);
1160 	}
1161 
1162 	adapter->state = VNIC_OPEN;
1163 	return rc;
1164 }
1165 
1166 static int ibmvnic_open(struct net_device *netdev)
1167 {
1168 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1169 	int rc;
1170 
1171 	/* If device failover is pending, just set device state and return.
1172 	 * Device operation will be handled by reset routine.
1173 	 */
1174 	if (adapter->failover_pending) {
1175 		adapter->state = VNIC_OPEN;
1176 		return 0;
1177 	}
1178 
1179 	if (adapter->state != VNIC_CLOSED) {
1180 		rc = ibmvnic_login(netdev);
1181 		if (rc)
1182 			return rc;
1183 
1184 		rc = init_resources(adapter);
1185 		if (rc) {
1186 			netdev_err(netdev, "failed to initialize resources\n");
1187 			release_resources(adapter);
1188 			return rc;
1189 		}
1190 	}
1191 
1192 	rc = __ibmvnic_open(netdev);
1193 
1194 	return rc;
1195 }
1196 
1197 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1198 {
1199 	struct ibmvnic_rx_pool *rx_pool;
1200 	struct ibmvnic_rx_buff *rx_buff;
1201 	u64 rx_entries;
1202 	int rx_scrqs;
1203 	int i, j;
1204 
1205 	if (!adapter->rx_pool)
1206 		return;
1207 
1208 	rx_scrqs = adapter->num_active_rx_pools;
1209 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1210 
1211 	/* Free any remaining skbs in the rx buffer pools */
1212 	for (i = 0; i < rx_scrqs; i++) {
1213 		rx_pool = &adapter->rx_pool[i];
1214 		if (!rx_pool || !rx_pool->rx_buff)
1215 			continue;
1216 
1217 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1218 		for (j = 0; j < rx_entries; j++) {
1219 			rx_buff = &rx_pool->rx_buff[j];
1220 			if (rx_buff && rx_buff->skb) {
1221 				dev_kfree_skb_any(rx_buff->skb);
1222 				rx_buff->skb = NULL;
1223 			}
1224 		}
1225 	}
1226 }
1227 
1228 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1229 			      struct ibmvnic_tx_pool *tx_pool)
1230 {
1231 	struct ibmvnic_tx_buff *tx_buff;
1232 	u64 tx_entries;
1233 	int i;
1234 
1235 	if (!tx_pool || !tx_pool->tx_buff)
1236 		return;
1237 
1238 	tx_entries = tx_pool->num_buffers;
1239 
1240 	for (i = 0; i < tx_entries; i++) {
1241 		tx_buff = &tx_pool->tx_buff[i];
1242 		if (tx_buff && tx_buff->skb) {
1243 			dev_kfree_skb_any(tx_buff->skb);
1244 			tx_buff->skb = NULL;
1245 		}
1246 	}
1247 }
1248 
1249 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1250 {
1251 	int tx_scrqs;
1252 	int i;
1253 
1254 	if (!adapter->tx_pool || !adapter->tso_pool)
1255 		return;
1256 
1257 	tx_scrqs = adapter->num_active_tx_pools;
1258 
1259 	/* Free any remaining skbs in the tx buffer pools */
1260 	for (i = 0; i < tx_scrqs; i++) {
1261 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1262 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1263 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1264 	}
1265 }
1266 
1267 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1268 {
1269 	struct net_device *netdev = adapter->netdev;
1270 	int i;
1271 
1272 	if (adapter->tx_scrq) {
1273 		for (i = 0; i < adapter->req_tx_queues; i++)
1274 			if (adapter->tx_scrq[i]->irq) {
1275 				netdev_dbg(netdev,
1276 					   "Disabling tx_scrq[%d] irq\n", i);
1277 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1278 				disable_irq(adapter->tx_scrq[i]->irq);
1279 			}
1280 	}
1281 
1282 	if (adapter->rx_scrq) {
1283 		for (i = 0; i < adapter->req_rx_queues; i++) {
1284 			if (adapter->rx_scrq[i]->irq) {
1285 				netdev_dbg(netdev,
1286 					   "Disabling rx_scrq[%d] irq\n", i);
1287 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1288 				disable_irq(adapter->rx_scrq[i]->irq);
1289 			}
1290 		}
1291 	}
1292 }
1293 
1294 static void ibmvnic_cleanup(struct net_device *netdev)
1295 {
1296 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1297 
1298 	/* ensure that transmissions are stopped if called by do_reset */
1299 	if (test_bit(0, &adapter->resetting))
1300 		netif_tx_disable(netdev);
1301 	else
1302 		netif_tx_stop_all_queues(netdev);
1303 
1304 	ibmvnic_napi_disable(adapter);
1305 	ibmvnic_disable_irqs(adapter);
1306 
1307 	clean_rx_pools(adapter);
1308 	clean_tx_pools(adapter);
1309 }
1310 
1311 static int __ibmvnic_close(struct net_device *netdev)
1312 {
1313 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1314 	int rc = 0;
1315 
1316 	adapter->state = VNIC_CLOSING;
1317 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1318 	if (rc)
1319 		return rc;
1320 	adapter->state = VNIC_CLOSED;
1321 	return 0;
1322 }
1323 
1324 static int ibmvnic_close(struct net_device *netdev)
1325 {
1326 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1327 	int rc;
1328 
1329 	/* If device failover is pending, just set device state and return.
1330 	 * Device operation will be handled by reset routine.
1331 	 */
1332 	if (adapter->failover_pending) {
1333 		adapter->state = VNIC_CLOSED;
1334 		return 0;
1335 	}
1336 
1337 	rc = __ibmvnic_close(netdev);
1338 	ibmvnic_cleanup(netdev);
1339 
1340 	return rc;
1341 }
1342 
1343 /**
1344  * build_hdr_data - creates L2/L3/L4 header data buffer
1345  * @hdr_field - bitfield determining needed headers
1346  * @skb - socket buffer
1347  * @hdr_len - array of header lengths
1348  * @tot_len - total length of data
1349  *
1350  * Reads hdr_field to determine which headers are needed by firmware.
1351  * Builds a buffer containing these headers.  Saves individual header
1352  * lengths and total buffer length to be used to build descriptors.
1353  */
1354 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1355 			  int *hdr_len, u8 *hdr_data)
1356 {
1357 	int len = 0;
1358 	u8 *hdr;
1359 
1360 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1361 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1362 	else
1363 		hdr_len[0] = sizeof(struct ethhdr);
1364 
1365 	if (skb->protocol == htons(ETH_P_IP)) {
1366 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1367 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1368 			hdr_len[2] = tcp_hdrlen(skb);
1369 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1370 			hdr_len[2] = sizeof(struct udphdr);
1371 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1372 		hdr_len[1] = sizeof(struct ipv6hdr);
1373 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1374 			hdr_len[2] = tcp_hdrlen(skb);
1375 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1376 			hdr_len[2] = sizeof(struct udphdr);
1377 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1378 		hdr_len[1] = arp_hdr_len(skb->dev);
1379 		hdr_len[2] = 0;
1380 	}
1381 
1382 	memset(hdr_data, 0, 120);
1383 	if ((hdr_field >> 6) & 1) {
1384 		hdr = skb_mac_header(skb);
1385 		memcpy(hdr_data, hdr, hdr_len[0]);
1386 		len += hdr_len[0];
1387 	}
1388 
1389 	if ((hdr_field >> 5) & 1) {
1390 		hdr = skb_network_header(skb);
1391 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1392 		len += hdr_len[1];
1393 	}
1394 
1395 	if ((hdr_field >> 4) & 1) {
1396 		hdr = skb_transport_header(skb);
1397 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1398 		len += hdr_len[2];
1399 	}
1400 	return len;
1401 }
1402 
1403 /**
1404  * create_hdr_descs - create header and header extension descriptors
1405  * @hdr_field - bitfield determining needed headers
1406  * @data - buffer containing header data
1407  * @len - length of data buffer
1408  * @hdr_len - array of individual header lengths
1409  * @scrq_arr - descriptor array
1410  *
1411  * Creates header and, if needed, header extension descriptors and
1412  * places them in a descriptor array, scrq_arr
1413  */
1414 
1415 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1416 			    union sub_crq *scrq_arr)
1417 {
1418 	union sub_crq hdr_desc;
1419 	int tmp_len = len;
1420 	int num_descs = 0;
1421 	u8 *data, *cur;
1422 	int tmp;
1423 
1424 	while (tmp_len > 0) {
1425 		cur = hdr_data + len - tmp_len;
1426 
1427 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1428 		if (cur != hdr_data) {
1429 			data = hdr_desc.hdr_ext.data;
1430 			tmp = tmp_len > 29 ? 29 : tmp_len;
1431 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1432 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1433 			hdr_desc.hdr_ext.len = tmp;
1434 		} else {
1435 			data = hdr_desc.hdr.data;
1436 			tmp = tmp_len > 24 ? 24 : tmp_len;
1437 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1438 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1439 			hdr_desc.hdr.len = tmp;
1440 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1441 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1442 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1443 			hdr_desc.hdr.flag = hdr_field << 1;
1444 		}
1445 		memcpy(data, cur, tmp);
1446 		tmp_len -= tmp;
1447 		*scrq_arr = hdr_desc;
1448 		scrq_arr++;
1449 		num_descs++;
1450 	}
1451 
1452 	return num_descs;
1453 }
1454 
1455 /**
1456  * build_hdr_descs_arr - build a header descriptor array
1457  * @skb - socket buffer
1458  * @num_entries - number of descriptors to be sent
1459  * @subcrq - first TX descriptor
1460  * @hdr_field - bit field determining which headers will be sent
1461  *
1462  * This function will build a TX descriptor array with applicable
1463  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1464  */
1465 
1466 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1467 				int *num_entries, u8 hdr_field)
1468 {
1469 	int hdr_len[3] = {0, 0, 0};
1470 	int tot_len;
1471 	u8 *hdr_data = txbuff->hdr_data;
1472 
1473 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1474 				 txbuff->hdr_data);
1475 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1476 			 txbuff->indir_arr + 1);
1477 }
1478 
1479 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1480 				    struct net_device *netdev)
1481 {
1482 	/* For some backing devices, mishandling of small packets
1483 	 * can result in a loss of connection or TX stall. Device
1484 	 * architects recommend that no packet should be smaller
1485 	 * than the minimum MTU value provided to the driver, so
1486 	 * pad any packets to that length
1487 	 */
1488 	if (skb->len < netdev->min_mtu)
1489 		return skb_put_padto(skb, netdev->min_mtu);
1490 
1491 	return 0;
1492 }
1493 
1494 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1495 {
1496 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1497 	int queue_num = skb_get_queue_mapping(skb);
1498 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1499 	struct device *dev = &adapter->vdev->dev;
1500 	struct ibmvnic_tx_buff *tx_buff = NULL;
1501 	struct ibmvnic_sub_crq_queue *tx_scrq;
1502 	struct ibmvnic_tx_pool *tx_pool;
1503 	unsigned int tx_send_failed = 0;
1504 	unsigned int tx_map_failed = 0;
1505 	unsigned int tx_dropped = 0;
1506 	unsigned int tx_packets = 0;
1507 	unsigned int tx_bytes = 0;
1508 	dma_addr_t data_dma_addr;
1509 	struct netdev_queue *txq;
1510 	unsigned long lpar_rc;
1511 	union sub_crq tx_crq;
1512 	unsigned int offset;
1513 	int num_entries = 1;
1514 	unsigned char *dst;
1515 	int index = 0;
1516 	u8 proto = 0;
1517 	u64 handle;
1518 	netdev_tx_t ret = NETDEV_TX_OK;
1519 
1520 	if (test_bit(0, &adapter->resetting)) {
1521 		if (!netif_subqueue_stopped(netdev, skb))
1522 			netif_stop_subqueue(netdev, queue_num);
1523 		dev_kfree_skb_any(skb);
1524 
1525 		tx_send_failed++;
1526 		tx_dropped++;
1527 		ret = NETDEV_TX_OK;
1528 		goto out;
1529 	}
1530 
1531 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1532 		tx_dropped++;
1533 		tx_send_failed++;
1534 		ret = NETDEV_TX_OK;
1535 		goto out;
1536 	}
1537 	if (skb_is_gso(skb))
1538 		tx_pool = &adapter->tso_pool[queue_num];
1539 	else
1540 		tx_pool = &adapter->tx_pool[queue_num];
1541 
1542 	tx_scrq = adapter->tx_scrq[queue_num];
1543 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1544 	handle = tx_scrq->handle;
1545 
1546 	index = tx_pool->free_map[tx_pool->consumer_index];
1547 
1548 	if (index == IBMVNIC_INVALID_MAP) {
1549 		dev_kfree_skb_any(skb);
1550 		tx_send_failed++;
1551 		tx_dropped++;
1552 		ret = NETDEV_TX_OK;
1553 		goto out;
1554 	}
1555 
1556 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1557 
1558 	offset = index * tx_pool->buf_size;
1559 	dst = tx_pool->long_term_buff.buff + offset;
1560 	memset(dst, 0, tx_pool->buf_size);
1561 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1562 
1563 	if (skb_shinfo(skb)->nr_frags) {
1564 		int cur, i;
1565 
1566 		/* Copy the head */
1567 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1568 		cur = skb_headlen(skb);
1569 
1570 		/* Copy the frags */
1571 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1572 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1573 
1574 			memcpy(dst + cur,
1575 			       page_address(skb_frag_page(frag)) +
1576 			       skb_frag_off(frag), skb_frag_size(frag));
1577 			cur += skb_frag_size(frag);
1578 		}
1579 	} else {
1580 		skb_copy_from_linear_data(skb, dst, skb->len);
1581 	}
1582 
1583 	tx_pool->consumer_index =
1584 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1585 
1586 	tx_buff = &tx_pool->tx_buff[index];
1587 	tx_buff->skb = skb;
1588 	tx_buff->data_dma[0] = data_dma_addr;
1589 	tx_buff->data_len[0] = skb->len;
1590 	tx_buff->index = index;
1591 	tx_buff->pool_index = queue_num;
1592 	tx_buff->last_frag = true;
1593 
1594 	memset(&tx_crq, 0, sizeof(tx_crq));
1595 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1596 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1597 	tx_crq.v1.n_crq_elem = 1;
1598 	tx_crq.v1.n_sge = 1;
1599 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1600 
1601 	if (skb_is_gso(skb))
1602 		tx_crq.v1.correlator =
1603 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1604 	else
1605 		tx_crq.v1.correlator = cpu_to_be32(index);
1606 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1607 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1608 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1609 
1610 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1611 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1612 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1613 	}
1614 
1615 	if (skb->protocol == htons(ETH_P_IP)) {
1616 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1617 		proto = ip_hdr(skb)->protocol;
1618 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1619 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1620 		proto = ipv6_hdr(skb)->nexthdr;
1621 	}
1622 
1623 	if (proto == IPPROTO_TCP)
1624 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1625 	else if (proto == IPPROTO_UDP)
1626 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1627 
1628 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1629 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1630 		hdrs += 2;
1631 	}
1632 	if (skb_is_gso(skb)) {
1633 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1634 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1635 		hdrs += 2;
1636 	}
1637 	/* determine if l2/3/4 headers are sent to firmware */
1638 	if ((*hdrs >> 7) & 1) {
1639 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1640 		tx_crq.v1.n_crq_elem = num_entries;
1641 		tx_buff->num_entries = num_entries;
1642 		tx_buff->indir_arr[0] = tx_crq;
1643 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1644 						    sizeof(tx_buff->indir_arr),
1645 						    DMA_TO_DEVICE);
1646 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1647 			dev_kfree_skb_any(skb);
1648 			tx_buff->skb = NULL;
1649 			if (!firmware_has_feature(FW_FEATURE_CMO))
1650 				dev_err(dev, "tx: unable to map descriptor array\n");
1651 			tx_map_failed++;
1652 			tx_dropped++;
1653 			ret = NETDEV_TX_OK;
1654 			goto tx_err_out;
1655 		}
1656 		lpar_rc = send_subcrq_indirect(adapter, handle,
1657 					       (u64)tx_buff->indir_dma,
1658 					       (u64)num_entries);
1659 		dma_unmap_single(dev, tx_buff->indir_dma,
1660 				 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1661 	} else {
1662 		tx_buff->num_entries = num_entries;
1663 		lpar_rc = send_subcrq(adapter, handle,
1664 				      &tx_crq);
1665 	}
1666 	if (lpar_rc != H_SUCCESS) {
1667 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1668 			dev_err_ratelimited(dev, "tx: send failed\n");
1669 		dev_kfree_skb_any(skb);
1670 		tx_buff->skb = NULL;
1671 
1672 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1673 			/* Disable TX and report carrier off if queue is closed
1674 			 * or pending failover.
1675 			 * Firmware guarantees that a signal will be sent to the
1676 			 * driver, triggering a reset or some other action.
1677 			 */
1678 			netif_tx_stop_all_queues(netdev);
1679 			netif_carrier_off(netdev);
1680 		}
1681 
1682 		tx_send_failed++;
1683 		tx_dropped++;
1684 		ret = NETDEV_TX_OK;
1685 		goto tx_err_out;
1686 	}
1687 
1688 	if (atomic_add_return(num_entries, &tx_scrq->used)
1689 					>= adapter->req_tx_entries_per_subcrq) {
1690 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1691 		netif_stop_subqueue(netdev, queue_num);
1692 	}
1693 
1694 	tx_packets++;
1695 	tx_bytes += skb->len;
1696 	txq->trans_start = jiffies;
1697 	ret = NETDEV_TX_OK;
1698 	goto out;
1699 
1700 tx_err_out:
1701 	/* roll back consumer index and map array*/
1702 	if (tx_pool->consumer_index == 0)
1703 		tx_pool->consumer_index =
1704 			tx_pool->num_buffers - 1;
1705 	else
1706 		tx_pool->consumer_index--;
1707 	tx_pool->free_map[tx_pool->consumer_index] = index;
1708 out:
1709 	netdev->stats.tx_dropped += tx_dropped;
1710 	netdev->stats.tx_bytes += tx_bytes;
1711 	netdev->stats.tx_packets += tx_packets;
1712 	adapter->tx_send_failed += tx_send_failed;
1713 	adapter->tx_map_failed += tx_map_failed;
1714 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1715 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1716 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1717 
1718 	return ret;
1719 }
1720 
1721 static void ibmvnic_set_multi(struct net_device *netdev)
1722 {
1723 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1724 	struct netdev_hw_addr *ha;
1725 	union ibmvnic_crq crq;
1726 
1727 	memset(&crq, 0, sizeof(crq));
1728 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1729 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1730 
1731 	if (netdev->flags & IFF_PROMISC) {
1732 		if (!adapter->promisc_supported)
1733 			return;
1734 	} else {
1735 		if (netdev->flags & IFF_ALLMULTI) {
1736 			/* Accept all multicast */
1737 			memset(&crq, 0, sizeof(crq));
1738 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1739 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1740 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1741 			ibmvnic_send_crq(adapter, &crq);
1742 		} else if (netdev_mc_empty(netdev)) {
1743 			/* Reject all multicast */
1744 			memset(&crq, 0, sizeof(crq));
1745 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1746 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1747 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1748 			ibmvnic_send_crq(adapter, &crq);
1749 		} else {
1750 			/* Accept one or more multicast(s) */
1751 			netdev_for_each_mc_addr(ha, netdev) {
1752 				memset(&crq, 0, sizeof(crq));
1753 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1754 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1755 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1756 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1757 						ha->addr);
1758 				ibmvnic_send_crq(adapter, &crq);
1759 			}
1760 		}
1761 	}
1762 }
1763 
1764 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1765 {
1766 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1767 	union ibmvnic_crq crq;
1768 	int rc;
1769 
1770 	if (!is_valid_ether_addr(dev_addr)) {
1771 		rc = -EADDRNOTAVAIL;
1772 		goto err;
1773 	}
1774 
1775 	memset(&crq, 0, sizeof(crq));
1776 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1777 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1778 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1779 
1780 	mutex_lock(&adapter->fw_lock);
1781 	adapter->fw_done_rc = 0;
1782 	reinit_completion(&adapter->fw_done);
1783 
1784 	rc = ibmvnic_send_crq(adapter, &crq);
1785 	if (rc) {
1786 		rc = -EIO;
1787 		mutex_unlock(&adapter->fw_lock);
1788 		goto err;
1789 	}
1790 
1791 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1792 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1793 	if (rc || adapter->fw_done_rc) {
1794 		rc = -EIO;
1795 		mutex_unlock(&adapter->fw_lock);
1796 		goto err;
1797 	}
1798 	mutex_unlock(&adapter->fw_lock);
1799 	return 0;
1800 err:
1801 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1802 	return rc;
1803 }
1804 
1805 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1806 {
1807 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1808 	struct sockaddr *addr = p;
1809 	int rc;
1810 
1811 	rc = 0;
1812 	ether_addr_copy(adapter->mac_addr, addr->sa_data);
1813 	if (adapter->state != VNIC_PROBED)
1814 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1815 
1816 	return rc;
1817 }
1818 
1819 /**
1820  * do_change_param_reset returns zero if we are able to keep processing reset
1821  * events, or non-zero if we hit a fatal error and must halt.
1822  */
1823 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1824 				 struct ibmvnic_rwi *rwi,
1825 				 u32 reset_state)
1826 {
1827 	struct net_device *netdev = adapter->netdev;
1828 	int i, rc;
1829 
1830 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1831 		   rwi->reset_reason);
1832 
1833 	netif_carrier_off(netdev);
1834 	adapter->reset_reason = rwi->reset_reason;
1835 
1836 	ibmvnic_cleanup(netdev);
1837 
1838 	if (reset_state == VNIC_OPEN) {
1839 		rc = __ibmvnic_close(netdev);
1840 		if (rc)
1841 			return rc;
1842 	}
1843 
1844 	release_resources(adapter);
1845 	release_sub_crqs(adapter, 1);
1846 	release_crq_queue(adapter);
1847 
1848 	adapter->state = VNIC_PROBED;
1849 
1850 	rc = init_crq_queue(adapter);
1851 
1852 	if (rc) {
1853 		netdev_err(adapter->netdev,
1854 			   "Couldn't initialize crq. rc=%d\n", rc);
1855 		return rc;
1856 	}
1857 
1858 	rc = ibmvnic_reset_init(adapter, true);
1859 	if (rc)
1860 		return IBMVNIC_INIT_FAILED;
1861 
1862 	/* If the adapter was in PROBE state prior to the reset,
1863 	 * exit here.
1864 	 */
1865 	if (reset_state == VNIC_PROBED)
1866 		return 0;
1867 
1868 	rc = ibmvnic_login(netdev);
1869 	if (rc) {
1870 		adapter->state = reset_state;
1871 		return rc;
1872 	}
1873 
1874 	rc = init_resources(adapter);
1875 	if (rc)
1876 		return rc;
1877 
1878 	ibmvnic_disable_irqs(adapter);
1879 
1880 	adapter->state = VNIC_CLOSED;
1881 
1882 	if (reset_state == VNIC_CLOSED)
1883 		return 0;
1884 
1885 	rc = __ibmvnic_open(netdev);
1886 	if (rc)
1887 		return IBMVNIC_OPEN_FAILED;
1888 
1889 	/* refresh device's multicast list */
1890 	ibmvnic_set_multi(netdev);
1891 
1892 	/* kick napi */
1893 	for (i = 0; i < adapter->req_rx_queues; i++)
1894 		napi_schedule(&adapter->napi[i]);
1895 
1896 	return 0;
1897 }
1898 
1899 /**
1900  * do_reset returns zero if we are able to keep processing reset events, or
1901  * non-zero if we hit a fatal error and must halt.
1902  */
1903 static int do_reset(struct ibmvnic_adapter *adapter,
1904 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1905 {
1906 	u64 old_num_rx_queues, old_num_tx_queues;
1907 	u64 old_num_rx_slots, old_num_tx_slots;
1908 	struct net_device *netdev = adapter->netdev;
1909 	int i, rc;
1910 
1911 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1912 		   rwi->reset_reason);
1913 
1914 	rtnl_lock();
1915 
1916 	netif_carrier_off(netdev);
1917 	adapter->reset_reason = rwi->reset_reason;
1918 
1919 	old_num_rx_queues = adapter->req_rx_queues;
1920 	old_num_tx_queues = adapter->req_tx_queues;
1921 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1922 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1923 
1924 	ibmvnic_cleanup(netdev);
1925 
1926 	if (reset_state == VNIC_OPEN &&
1927 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
1928 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1929 		adapter->state = VNIC_CLOSING;
1930 
1931 		/* Release the RTNL lock before link state change and
1932 		 * re-acquire after the link state change to allow
1933 		 * linkwatch_event to grab the RTNL lock and run during
1934 		 * a reset.
1935 		 */
1936 		rtnl_unlock();
1937 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1938 		rtnl_lock();
1939 		if (rc)
1940 			goto out;
1941 
1942 		if (adapter->state != VNIC_CLOSING) {
1943 			rc = -1;
1944 			goto out;
1945 		}
1946 
1947 		adapter->state = VNIC_CLOSED;
1948 	}
1949 
1950 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1951 		/* remove the closed state so when we call open it appears
1952 		 * we are coming from the probed state.
1953 		 */
1954 		adapter->state = VNIC_PROBED;
1955 
1956 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1957 			rc = ibmvnic_reenable_crq_queue(adapter);
1958 			release_sub_crqs(adapter, 1);
1959 		} else {
1960 			rc = ibmvnic_reset_crq(adapter);
1961 			if (rc == H_CLOSED || rc == H_SUCCESS) {
1962 				rc = vio_enable_interrupts(adapter->vdev);
1963 				if (rc)
1964 					netdev_err(adapter->netdev,
1965 						   "Reset failed to enable interrupts. rc=%d\n",
1966 						   rc);
1967 			}
1968 		}
1969 
1970 		if (rc) {
1971 			netdev_err(adapter->netdev,
1972 				   "Reset couldn't initialize crq. rc=%d\n", rc);
1973 			goto out;
1974 		}
1975 
1976 		rc = ibmvnic_reset_init(adapter, true);
1977 		if (rc) {
1978 			rc = IBMVNIC_INIT_FAILED;
1979 			goto out;
1980 		}
1981 
1982 		/* If the adapter was in PROBE state prior to the reset,
1983 		 * exit here.
1984 		 */
1985 		if (reset_state == VNIC_PROBED) {
1986 			rc = 0;
1987 			goto out;
1988 		}
1989 
1990 		rc = ibmvnic_login(netdev);
1991 		if (rc) {
1992 			adapter->state = reset_state;
1993 			goto out;
1994 		}
1995 
1996 		if (adapter->req_rx_queues != old_num_rx_queues ||
1997 		    adapter->req_tx_queues != old_num_tx_queues ||
1998 		    adapter->req_rx_add_entries_per_subcrq !=
1999 		    old_num_rx_slots ||
2000 		    adapter->req_tx_entries_per_subcrq !=
2001 		    old_num_tx_slots) {
2002 			release_rx_pools(adapter);
2003 			release_tx_pools(adapter);
2004 			release_napi(adapter);
2005 			release_vpd_data(adapter);
2006 
2007 			rc = init_resources(adapter);
2008 			if (rc)
2009 				goto out;
2010 
2011 		} else {
2012 			rc = reset_tx_pools(adapter);
2013 			if (rc)
2014 				goto out;
2015 
2016 			rc = reset_rx_pools(adapter);
2017 			if (rc)
2018 				goto out;
2019 		}
2020 		ibmvnic_disable_irqs(adapter);
2021 	}
2022 	adapter->state = VNIC_CLOSED;
2023 
2024 	if (reset_state == VNIC_CLOSED) {
2025 		rc = 0;
2026 		goto out;
2027 	}
2028 
2029 	rc = __ibmvnic_open(netdev);
2030 	if (rc) {
2031 		rc = IBMVNIC_OPEN_FAILED;
2032 		goto out;
2033 	}
2034 
2035 	/* refresh device's multicast list */
2036 	ibmvnic_set_multi(netdev);
2037 
2038 	/* kick napi */
2039 	for (i = 0; i < adapter->req_rx_queues; i++)
2040 		napi_schedule(&adapter->napi[i]);
2041 
2042 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2043 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2044 
2045 	rc = 0;
2046 
2047 out:
2048 	rtnl_unlock();
2049 
2050 	return rc;
2051 }
2052 
2053 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2054 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2055 {
2056 	struct net_device *netdev = adapter->netdev;
2057 	int rc;
2058 
2059 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2060 		   rwi->reset_reason);
2061 
2062 	netif_carrier_off(netdev);
2063 	adapter->reset_reason = rwi->reset_reason;
2064 
2065 	ibmvnic_cleanup(netdev);
2066 	release_resources(adapter);
2067 	release_sub_crqs(adapter, 0);
2068 	release_crq_queue(adapter);
2069 
2070 	/* remove the closed state so when we call open it appears
2071 	 * we are coming from the probed state.
2072 	 */
2073 	adapter->state = VNIC_PROBED;
2074 
2075 	reinit_completion(&adapter->init_done);
2076 	rc = init_crq_queue(adapter);
2077 	if (rc) {
2078 		netdev_err(adapter->netdev,
2079 			   "Couldn't initialize crq. rc=%d\n", rc);
2080 		return rc;
2081 	}
2082 
2083 	rc = ibmvnic_reset_init(adapter, false);
2084 	if (rc)
2085 		return rc;
2086 
2087 	/* If the adapter was in PROBE state prior to the reset,
2088 	 * exit here.
2089 	 */
2090 	if (reset_state == VNIC_PROBED)
2091 		return 0;
2092 
2093 	rc = ibmvnic_login(netdev);
2094 	if (rc) {
2095 		adapter->state = VNIC_PROBED;
2096 		return 0;
2097 	}
2098 
2099 	rc = init_resources(adapter);
2100 	if (rc)
2101 		return rc;
2102 
2103 	ibmvnic_disable_irqs(adapter);
2104 	adapter->state = VNIC_CLOSED;
2105 
2106 	if (reset_state == VNIC_CLOSED)
2107 		return 0;
2108 
2109 	rc = __ibmvnic_open(netdev);
2110 	if (rc)
2111 		return IBMVNIC_OPEN_FAILED;
2112 
2113 	return 0;
2114 }
2115 
2116 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2117 {
2118 	struct ibmvnic_rwi *rwi;
2119 	unsigned long flags;
2120 
2121 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2122 
2123 	if (!list_empty(&adapter->rwi_list)) {
2124 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2125 				       list);
2126 		list_del(&rwi->list);
2127 	} else {
2128 		rwi = NULL;
2129 	}
2130 
2131 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2132 	return rwi;
2133 }
2134 
2135 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2136 {
2137 	struct ibmvnic_rwi *rwi;
2138 
2139 	rwi = get_next_rwi(adapter);
2140 	while (rwi) {
2141 		kfree(rwi);
2142 		rwi = get_next_rwi(adapter);
2143 	}
2144 }
2145 
2146 static void __ibmvnic_reset(struct work_struct *work)
2147 {
2148 	struct ibmvnic_rwi *rwi;
2149 	struct ibmvnic_adapter *adapter;
2150 	bool saved_state = false;
2151 	unsigned long flags;
2152 	u32 reset_state;
2153 	int rc = 0;
2154 
2155 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2156 
2157 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2158 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2159 				      IBMVNIC_RESET_DELAY);
2160 		return;
2161 	}
2162 
2163 	rwi = get_next_rwi(adapter);
2164 	while (rwi) {
2165 		spin_lock_irqsave(&adapter->state_lock, flags);
2166 
2167 		if (adapter->state == VNIC_REMOVING ||
2168 		    adapter->state == VNIC_REMOVED) {
2169 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2170 			kfree(rwi);
2171 			rc = EBUSY;
2172 			break;
2173 		}
2174 
2175 		if (!saved_state) {
2176 			reset_state = adapter->state;
2177 			adapter->state = VNIC_RESETTING;
2178 			saved_state = true;
2179 		}
2180 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2181 
2182 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2183 			/* CHANGE_PARAM requestor holds rtnl_lock */
2184 			rc = do_change_param_reset(adapter, rwi, reset_state);
2185 		} else if (adapter->force_reset_recovery) {
2186 			/* Transport event occurred during previous reset */
2187 			if (adapter->wait_for_reset) {
2188 				/* Previous was CHANGE_PARAM; caller locked */
2189 				adapter->force_reset_recovery = false;
2190 				rc = do_hard_reset(adapter, rwi, reset_state);
2191 			} else {
2192 				rtnl_lock();
2193 				adapter->force_reset_recovery = false;
2194 				rc = do_hard_reset(adapter, rwi, reset_state);
2195 				rtnl_unlock();
2196 			}
2197 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2198 				adapter->from_passive_init)) {
2199 			rc = do_reset(adapter, rwi, reset_state);
2200 		}
2201 		kfree(rwi);
2202 		if (rc == IBMVNIC_OPEN_FAILED) {
2203 			if (list_empty(&adapter->rwi_list))
2204 				adapter->state = VNIC_CLOSED;
2205 			else
2206 				adapter->state = reset_state;
2207 			rc = 0;
2208 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2209 		    !adapter->force_reset_recovery)
2210 			break;
2211 
2212 		rwi = get_next_rwi(adapter);
2213 
2214 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2215 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2216 			adapter->force_reset_recovery = true;
2217 	}
2218 
2219 	if (adapter->wait_for_reset) {
2220 		adapter->reset_done_rc = rc;
2221 		complete(&adapter->reset_done);
2222 	}
2223 
2224 	if (rc) {
2225 		netdev_dbg(adapter->netdev, "Reset failed\n");
2226 		free_all_rwi(adapter);
2227 	}
2228 
2229 	clear_bit_unlock(0, &adapter->resetting);
2230 }
2231 
2232 static void __ibmvnic_delayed_reset(struct work_struct *work)
2233 {
2234 	struct ibmvnic_adapter *adapter;
2235 
2236 	adapter = container_of(work, struct ibmvnic_adapter,
2237 			       ibmvnic_delayed_reset.work);
2238 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2239 }
2240 
2241 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2242 			 enum ibmvnic_reset_reason reason)
2243 {
2244 	struct list_head *entry, *tmp_entry;
2245 	struct ibmvnic_rwi *rwi, *tmp;
2246 	struct net_device *netdev = adapter->netdev;
2247 	unsigned long flags;
2248 	int ret;
2249 
2250 	if (adapter->state == VNIC_REMOVING ||
2251 	    adapter->state == VNIC_REMOVED ||
2252 	    adapter->failover_pending) {
2253 		ret = EBUSY;
2254 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2255 		goto err;
2256 	}
2257 
2258 	if (adapter->state == VNIC_PROBING) {
2259 		netdev_warn(netdev, "Adapter reset during probe\n");
2260 		ret = adapter->init_done_rc = EAGAIN;
2261 		goto err;
2262 	}
2263 
2264 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2265 
2266 	list_for_each(entry, &adapter->rwi_list) {
2267 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2268 		if (tmp->reset_reason == reason) {
2269 			netdev_dbg(netdev, "Skipping matching reset\n");
2270 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2271 			ret = EBUSY;
2272 			goto err;
2273 		}
2274 	}
2275 
2276 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2277 	if (!rwi) {
2278 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2279 		ibmvnic_close(netdev);
2280 		ret = ENOMEM;
2281 		goto err;
2282 	}
2283 	/* if we just received a transport event,
2284 	 * flush reset queue and process this reset
2285 	 */
2286 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2287 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2288 			list_del(entry);
2289 	}
2290 	rwi->reset_reason = reason;
2291 	list_add_tail(&rwi->list, &adapter->rwi_list);
2292 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2293 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2294 	schedule_work(&adapter->ibmvnic_reset);
2295 
2296 	return 0;
2297 err:
2298 	return -ret;
2299 }
2300 
2301 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2302 {
2303 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2304 
2305 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2306 }
2307 
2308 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2309 				  struct ibmvnic_rx_buff *rx_buff)
2310 {
2311 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2312 
2313 	rx_buff->skb = NULL;
2314 
2315 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2316 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2317 
2318 	atomic_dec(&pool->available);
2319 }
2320 
2321 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2322 {
2323 	struct net_device *netdev = napi->dev;
2324 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2325 	int scrq_num = (int)(napi - adapter->napi);
2326 	int frames_processed = 0;
2327 
2328 restart_poll:
2329 	while (frames_processed < budget) {
2330 		struct sk_buff *skb;
2331 		struct ibmvnic_rx_buff *rx_buff;
2332 		union sub_crq *next;
2333 		u32 length;
2334 		u16 offset;
2335 		u8 flags = 0;
2336 
2337 		if (unlikely(test_bit(0, &adapter->resetting) &&
2338 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2339 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2340 			napi_complete_done(napi, frames_processed);
2341 			return frames_processed;
2342 		}
2343 
2344 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2345 			break;
2346 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2347 		rx_buff =
2348 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2349 							  rx_comp.correlator);
2350 		/* do error checking */
2351 		if (next->rx_comp.rc) {
2352 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2353 				   be16_to_cpu(next->rx_comp.rc));
2354 			/* free the entry */
2355 			next->rx_comp.first = 0;
2356 			dev_kfree_skb_any(rx_buff->skb);
2357 			remove_buff_from_pool(adapter, rx_buff);
2358 			continue;
2359 		} else if (!rx_buff->skb) {
2360 			/* free the entry */
2361 			next->rx_comp.first = 0;
2362 			remove_buff_from_pool(adapter, rx_buff);
2363 			continue;
2364 		}
2365 
2366 		length = be32_to_cpu(next->rx_comp.len);
2367 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2368 		flags = next->rx_comp.flags;
2369 		skb = rx_buff->skb;
2370 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2371 					length);
2372 
2373 		/* VLAN Header has been stripped by the system firmware and
2374 		 * needs to be inserted by the driver
2375 		 */
2376 		if (adapter->rx_vlan_header_insertion &&
2377 		    (flags & IBMVNIC_VLAN_STRIPPED))
2378 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2379 					       ntohs(next->rx_comp.vlan_tci));
2380 
2381 		/* free the entry */
2382 		next->rx_comp.first = 0;
2383 		remove_buff_from_pool(adapter, rx_buff);
2384 
2385 		skb_put(skb, length);
2386 		skb->protocol = eth_type_trans(skb, netdev);
2387 		skb_record_rx_queue(skb, scrq_num);
2388 
2389 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2390 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2391 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2392 		}
2393 
2394 		length = skb->len;
2395 		napi_gro_receive(napi, skb); /* send it up */
2396 		netdev->stats.rx_packets++;
2397 		netdev->stats.rx_bytes += length;
2398 		adapter->rx_stats_buffers[scrq_num].packets++;
2399 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2400 		frames_processed++;
2401 	}
2402 
2403 	if (adapter->state != VNIC_CLOSING)
2404 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2405 
2406 	if (frames_processed < budget) {
2407 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2408 		napi_complete_done(napi, frames_processed);
2409 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2410 		    napi_reschedule(napi)) {
2411 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2412 			goto restart_poll;
2413 		}
2414 	}
2415 	return frames_processed;
2416 }
2417 
2418 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2419 {
2420 	int rc, ret;
2421 
2422 	adapter->fallback.mtu = adapter->req_mtu;
2423 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2424 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2425 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2426 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2427 
2428 	reinit_completion(&adapter->reset_done);
2429 	adapter->wait_for_reset = true;
2430 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2431 
2432 	if (rc) {
2433 		ret = rc;
2434 		goto out;
2435 	}
2436 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2437 	if (rc) {
2438 		ret = -ENODEV;
2439 		goto out;
2440 	}
2441 
2442 	ret = 0;
2443 	if (adapter->reset_done_rc) {
2444 		ret = -EIO;
2445 		adapter->desired.mtu = adapter->fallback.mtu;
2446 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2447 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2448 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2449 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2450 
2451 		reinit_completion(&adapter->reset_done);
2452 		adapter->wait_for_reset = true;
2453 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2454 		if (rc) {
2455 			ret = rc;
2456 			goto out;
2457 		}
2458 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2459 						 60000);
2460 		if (rc) {
2461 			ret = -ENODEV;
2462 			goto out;
2463 		}
2464 	}
2465 out:
2466 	adapter->wait_for_reset = false;
2467 
2468 	return ret;
2469 }
2470 
2471 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2472 {
2473 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2474 
2475 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2476 
2477 	return wait_for_reset(adapter);
2478 }
2479 
2480 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2481 						struct net_device *dev,
2482 						netdev_features_t features)
2483 {
2484 	/* Some backing hardware adapters can not
2485 	 * handle packets with a MSS less than 224
2486 	 * or with only one segment.
2487 	 */
2488 	if (skb_is_gso(skb)) {
2489 		if (skb_shinfo(skb)->gso_size < 224 ||
2490 		    skb_shinfo(skb)->gso_segs == 1)
2491 			features &= ~NETIF_F_GSO_MASK;
2492 	}
2493 
2494 	return features;
2495 }
2496 
2497 static const struct net_device_ops ibmvnic_netdev_ops = {
2498 	.ndo_open		= ibmvnic_open,
2499 	.ndo_stop		= ibmvnic_close,
2500 	.ndo_start_xmit		= ibmvnic_xmit,
2501 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2502 	.ndo_set_mac_address	= ibmvnic_set_mac,
2503 	.ndo_validate_addr	= eth_validate_addr,
2504 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2505 	.ndo_change_mtu		= ibmvnic_change_mtu,
2506 	.ndo_features_check     = ibmvnic_features_check,
2507 };
2508 
2509 /* ethtool functions */
2510 
2511 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2512 				      struct ethtool_link_ksettings *cmd)
2513 {
2514 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2515 	int rc;
2516 
2517 	rc = send_query_phys_parms(adapter);
2518 	if (rc) {
2519 		adapter->speed = SPEED_UNKNOWN;
2520 		adapter->duplex = DUPLEX_UNKNOWN;
2521 	}
2522 	cmd->base.speed = adapter->speed;
2523 	cmd->base.duplex = adapter->duplex;
2524 	cmd->base.port = PORT_FIBRE;
2525 	cmd->base.phy_address = 0;
2526 	cmd->base.autoneg = AUTONEG_ENABLE;
2527 
2528 	return 0;
2529 }
2530 
2531 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2532 				struct ethtool_drvinfo *info)
2533 {
2534 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2535 
2536 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2537 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2538 	strlcpy(info->fw_version, adapter->fw_version,
2539 		sizeof(info->fw_version));
2540 }
2541 
2542 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2543 {
2544 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2545 
2546 	return adapter->msg_enable;
2547 }
2548 
2549 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2550 {
2551 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2552 
2553 	adapter->msg_enable = data;
2554 }
2555 
2556 static u32 ibmvnic_get_link(struct net_device *netdev)
2557 {
2558 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2559 
2560 	/* Don't need to send a query because we request a logical link up at
2561 	 * init and then we wait for link state indications
2562 	 */
2563 	return adapter->logical_link_state;
2564 }
2565 
2566 static void ibmvnic_get_ringparam(struct net_device *netdev,
2567 				  struct ethtool_ringparam *ring)
2568 {
2569 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2570 
2571 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2572 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2573 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2574 	} else {
2575 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2576 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2577 	}
2578 	ring->rx_mini_max_pending = 0;
2579 	ring->rx_jumbo_max_pending = 0;
2580 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2581 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2582 	ring->rx_mini_pending = 0;
2583 	ring->rx_jumbo_pending = 0;
2584 }
2585 
2586 static int ibmvnic_set_ringparam(struct net_device *netdev,
2587 				 struct ethtool_ringparam *ring)
2588 {
2589 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2590 	int ret;
2591 
2592 	ret = 0;
2593 	adapter->desired.rx_entries = ring->rx_pending;
2594 	adapter->desired.tx_entries = ring->tx_pending;
2595 
2596 	ret = wait_for_reset(adapter);
2597 
2598 	if (!ret &&
2599 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2600 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2601 		netdev_info(netdev,
2602 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2603 			    ring->rx_pending, ring->tx_pending,
2604 			    adapter->req_rx_add_entries_per_subcrq,
2605 			    adapter->req_tx_entries_per_subcrq);
2606 	return ret;
2607 }
2608 
2609 static void ibmvnic_get_channels(struct net_device *netdev,
2610 				 struct ethtool_channels *channels)
2611 {
2612 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2613 
2614 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2615 		channels->max_rx = adapter->max_rx_queues;
2616 		channels->max_tx = adapter->max_tx_queues;
2617 	} else {
2618 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2619 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2620 	}
2621 
2622 	channels->max_other = 0;
2623 	channels->max_combined = 0;
2624 	channels->rx_count = adapter->req_rx_queues;
2625 	channels->tx_count = adapter->req_tx_queues;
2626 	channels->other_count = 0;
2627 	channels->combined_count = 0;
2628 }
2629 
2630 static int ibmvnic_set_channels(struct net_device *netdev,
2631 				struct ethtool_channels *channels)
2632 {
2633 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2634 	int ret;
2635 
2636 	ret = 0;
2637 	adapter->desired.rx_queues = channels->rx_count;
2638 	adapter->desired.tx_queues = channels->tx_count;
2639 
2640 	ret = wait_for_reset(adapter);
2641 
2642 	if (!ret &&
2643 	    (adapter->req_rx_queues != channels->rx_count ||
2644 	     adapter->req_tx_queues != channels->tx_count))
2645 		netdev_info(netdev,
2646 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2647 			    channels->rx_count, channels->tx_count,
2648 			    adapter->req_rx_queues, adapter->req_tx_queues);
2649 	return ret;
2650 
2651 }
2652 
2653 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2654 {
2655 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2656 	int i;
2657 
2658 	switch (stringset) {
2659 	case ETH_SS_STATS:
2660 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2661 				i++, data += ETH_GSTRING_LEN)
2662 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2663 
2664 		for (i = 0; i < adapter->req_tx_queues; i++) {
2665 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2666 			data += ETH_GSTRING_LEN;
2667 
2668 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2669 			data += ETH_GSTRING_LEN;
2670 
2671 			snprintf(data, ETH_GSTRING_LEN,
2672 				 "tx%d_dropped_packets", i);
2673 			data += ETH_GSTRING_LEN;
2674 		}
2675 
2676 		for (i = 0; i < adapter->req_rx_queues; i++) {
2677 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2678 			data += ETH_GSTRING_LEN;
2679 
2680 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2681 			data += ETH_GSTRING_LEN;
2682 
2683 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2684 			data += ETH_GSTRING_LEN;
2685 		}
2686 		break;
2687 
2688 	case ETH_SS_PRIV_FLAGS:
2689 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2690 			strcpy(data + i * ETH_GSTRING_LEN,
2691 			       ibmvnic_priv_flags[i]);
2692 		break;
2693 	default:
2694 		return;
2695 	}
2696 }
2697 
2698 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2699 {
2700 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2701 
2702 	switch (sset) {
2703 	case ETH_SS_STATS:
2704 		return ARRAY_SIZE(ibmvnic_stats) +
2705 		       adapter->req_tx_queues * NUM_TX_STATS +
2706 		       adapter->req_rx_queues * NUM_RX_STATS;
2707 	case ETH_SS_PRIV_FLAGS:
2708 		return ARRAY_SIZE(ibmvnic_priv_flags);
2709 	default:
2710 		return -EOPNOTSUPP;
2711 	}
2712 }
2713 
2714 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2715 				      struct ethtool_stats *stats, u64 *data)
2716 {
2717 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2718 	union ibmvnic_crq crq;
2719 	int i, j;
2720 	int rc;
2721 
2722 	memset(&crq, 0, sizeof(crq));
2723 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2724 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2725 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2726 	crq.request_statistics.len =
2727 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2728 
2729 	/* Wait for data to be written */
2730 	reinit_completion(&adapter->stats_done);
2731 	rc = ibmvnic_send_crq(adapter, &crq);
2732 	if (rc)
2733 		return;
2734 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2735 	if (rc)
2736 		return;
2737 
2738 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2739 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2740 						ibmvnic_stats[i].offset));
2741 
2742 	for (j = 0; j < adapter->req_tx_queues; j++) {
2743 		data[i] = adapter->tx_stats_buffers[j].packets;
2744 		i++;
2745 		data[i] = adapter->tx_stats_buffers[j].bytes;
2746 		i++;
2747 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2748 		i++;
2749 	}
2750 
2751 	for (j = 0; j < adapter->req_rx_queues; j++) {
2752 		data[i] = adapter->rx_stats_buffers[j].packets;
2753 		i++;
2754 		data[i] = adapter->rx_stats_buffers[j].bytes;
2755 		i++;
2756 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2757 		i++;
2758 	}
2759 }
2760 
2761 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2762 {
2763 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2764 
2765 	return adapter->priv_flags;
2766 }
2767 
2768 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2769 {
2770 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2771 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2772 
2773 	if (which_maxes)
2774 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2775 	else
2776 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2777 
2778 	return 0;
2779 }
2780 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2781 	.get_drvinfo		= ibmvnic_get_drvinfo,
2782 	.get_msglevel		= ibmvnic_get_msglevel,
2783 	.set_msglevel		= ibmvnic_set_msglevel,
2784 	.get_link		= ibmvnic_get_link,
2785 	.get_ringparam		= ibmvnic_get_ringparam,
2786 	.set_ringparam		= ibmvnic_set_ringparam,
2787 	.get_channels		= ibmvnic_get_channels,
2788 	.set_channels		= ibmvnic_set_channels,
2789 	.get_strings            = ibmvnic_get_strings,
2790 	.get_sset_count         = ibmvnic_get_sset_count,
2791 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2792 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2793 	.get_priv_flags		= ibmvnic_get_priv_flags,
2794 	.set_priv_flags		= ibmvnic_set_priv_flags,
2795 };
2796 
2797 /* Routines for managing CRQs/sCRQs  */
2798 
2799 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2800 				   struct ibmvnic_sub_crq_queue *scrq)
2801 {
2802 	int rc;
2803 
2804 	if (scrq->irq) {
2805 		free_irq(scrq->irq, scrq);
2806 		irq_dispose_mapping(scrq->irq);
2807 		scrq->irq = 0;
2808 	}
2809 
2810 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2811 	atomic_set(&scrq->used, 0);
2812 	scrq->cur = 0;
2813 
2814 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2815 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2816 	return rc;
2817 }
2818 
2819 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2820 {
2821 	int i, rc;
2822 
2823 	for (i = 0; i < adapter->req_tx_queues; i++) {
2824 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2825 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2826 		if (rc)
2827 			return rc;
2828 	}
2829 
2830 	for (i = 0; i < adapter->req_rx_queues; i++) {
2831 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2832 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2833 		if (rc)
2834 			return rc;
2835 	}
2836 
2837 	return rc;
2838 }
2839 
2840 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2841 				  struct ibmvnic_sub_crq_queue *scrq,
2842 				  bool do_h_free)
2843 {
2844 	struct device *dev = &adapter->vdev->dev;
2845 	long rc;
2846 
2847 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2848 
2849 	if (do_h_free) {
2850 		/* Close the sub-crqs */
2851 		do {
2852 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2853 						adapter->vdev->unit_address,
2854 						scrq->crq_num);
2855 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2856 
2857 		if (rc) {
2858 			netdev_err(adapter->netdev,
2859 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2860 				   scrq->crq_num, rc);
2861 		}
2862 	}
2863 
2864 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2865 			 DMA_BIDIRECTIONAL);
2866 	free_pages((unsigned long)scrq->msgs, 2);
2867 	kfree(scrq);
2868 }
2869 
2870 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2871 							*adapter)
2872 {
2873 	struct device *dev = &adapter->vdev->dev;
2874 	struct ibmvnic_sub_crq_queue *scrq;
2875 	int rc;
2876 
2877 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2878 	if (!scrq)
2879 		return NULL;
2880 
2881 	scrq->msgs =
2882 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2883 	if (!scrq->msgs) {
2884 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2885 		goto zero_page_failed;
2886 	}
2887 
2888 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2889 					 DMA_BIDIRECTIONAL);
2890 	if (dma_mapping_error(dev, scrq->msg_token)) {
2891 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2892 		goto map_failed;
2893 	}
2894 
2895 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2896 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2897 
2898 	if (rc == H_RESOURCE)
2899 		rc = ibmvnic_reset_crq(adapter);
2900 
2901 	if (rc == H_CLOSED) {
2902 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2903 	} else if (rc) {
2904 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2905 		goto reg_failed;
2906 	}
2907 
2908 	scrq->adapter = adapter;
2909 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2910 	spin_lock_init(&scrq->lock);
2911 
2912 	netdev_dbg(adapter->netdev,
2913 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2914 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2915 
2916 	return scrq;
2917 
2918 reg_failed:
2919 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2920 			 DMA_BIDIRECTIONAL);
2921 map_failed:
2922 	free_pages((unsigned long)scrq->msgs, 2);
2923 zero_page_failed:
2924 	kfree(scrq);
2925 
2926 	return NULL;
2927 }
2928 
2929 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2930 {
2931 	int i;
2932 
2933 	if (adapter->tx_scrq) {
2934 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2935 			if (!adapter->tx_scrq[i])
2936 				continue;
2937 
2938 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2939 				   i);
2940 			if (adapter->tx_scrq[i]->irq) {
2941 				free_irq(adapter->tx_scrq[i]->irq,
2942 					 adapter->tx_scrq[i]);
2943 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2944 				adapter->tx_scrq[i]->irq = 0;
2945 			}
2946 
2947 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2948 					      do_h_free);
2949 		}
2950 
2951 		kfree(adapter->tx_scrq);
2952 		adapter->tx_scrq = NULL;
2953 		adapter->num_active_tx_scrqs = 0;
2954 	}
2955 
2956 	if (adapter->rx_scrq) {
2957 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2958 			if (!adapter->rx_scrq[i])
2959 				continue;
2960 
2961 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2962 				   i);
2963 			if (adapter->rx_scrq[i]->irq) {
2964 				free_irq(adapter->rx_scrq[i]->irq,
2965 					 adapter->rx_scrq[i]);
2966 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2967 				adapter->rx_scrq[i]->irq = 0;
2968 			}
2969 
2970 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2971 					      do_h_free);
2972 		}
2973 
2974 		kfree(adapter->rx_scrq);
2975 		adapter->rx_scrq = NULL;
2976 		adapter->num_active_rx_scrqs = 0;
2977 	}
2978 }
2979 
2980 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2981 			    struct ibmvnic_sub_crq_queue *scrq)
2982 {
2983 	struct device *dev = &adapter->vdev->dev;
2984 	unsigned long rc;
2985 
2986 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2987 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2988 	if (rc)
2989 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2990 			scrq->hw_irq, rc);
2991 	return rc;
2992 }
2993 
2994 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2995 			   struct ibmvnic_sub_crq_queue *scrq)
2996 {
2997 	struct device *dev = &adapter->vdev->dev;
2998 	unsigned long rc;
2999 
3000 	if (scrq->hw_irq > 0x100000000ULL) {
3001 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3002 		return 1;
3003 	}
3004 
3005 	if (test_bit(0, &adapter->resetting) &&
3006 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3007 		u64 val = (0xff000000) | scrq->hw_irq;
3008 
3009 		rc = plpar_hcall_norets(H_EOI, val);
3010 		/* H_EOI would fail with rc = H_FUNCTION when running
3011 		 * in XIVE mode which is expected, but not an error.
3012 		 */
3013 		if (rc && (rc != H_FUNCTION))
3014 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3015 				val, rc);
3016 	}
3017 
3018 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3019 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3020 	if (rc)
3021 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3022 			scrq->hw_irq, rc);
3023 	return rc;
3024 }
3025 
3026 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3027 			       struct ibmvnic_sub_crq_queue *scrq)
3028 {
3029 	struct device *dev = &adapter->vdev->dev;
3030 	struct ibmvnic_tx_pool *tx_pool;
3031 	struct ibmvnic_tx_buff *txbuff;
3032 	union sub_crq *next;
3033 	int index;
3034 	int i, j;
3035 
3036 restart_loop:
3037 	while (pending_scrq(adapter, scrq)) {
3038 		unsigned int pool = scrq->pool_index;
3039 		int num_entries = 0;
3040 
3041 		next = ibmvnic_next_scrq(adapter, scrq);
3042 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3043 			if (next->tx_comp.rcs[i]) {
3044 				dev_err(dev, "tx error %x\n",
3045 					next->tx_comp.rcs[i]);
3046 				continue;
3047 			}
3048 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3049 			if (index & IBMVNIC_TSO_POOL_MASK) {
3050 				tx_pool = &adapter->tso_pool[pool];
3051 				index &= ~IBMVNIC_TSO_POOL_MASK;
3052 			} else {
3053 				tx_pool = &adapter->tx_pool[pool];
3054 			}
3055 
3056 			txbuff = &tx_pool->tx_buff[index];
3057 
3058 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3059 				if (!txbuff->data_dma[j])
3060 					continue;
3061 
3062 				txbuff->data_dma[j] = 0;
3063 			}
3064 
3065 			if (txbuff->last_frag) {
3066 				dev_kfree_skb_any(txbuff->skb);
3067 				txbuff->skb = NULL;
3068 			}
3069 
3070 			num_entries += txbuff->num_entries;
3071 
3072 			tx_pool->free_map[tx_pool->producer_index] = index;
3073 			tx_pool->producer_index =
3074 				(tx_pool->producer_index + 1) %
3075 					tx_pool->num_buffers;
3076 		}
3077 		/* remove tx_comp scrq*/
3078 		next->tx_comp.first = 0;
3079 
3080 		if (atomic_sub_return(num_entries, &scrq->used) <=
3081 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3082 		    __netif_subqueue_stopped(adapter->netdev,
3083 					     scrq->pool_index)) {
3084 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3085 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3086 				   scrq->pool_index);
3087 		}
3088 	}
3089 
3090 	enable_scrq_irq(adapter, scrq);
3091 
3092 	if (pending_scrq(adapter, scrq)) {
3093 		disable_scrq_irq(adapter, scrq);
3094 		goto restart_loop;
3095 	}
3096 
3097 	return 0;
3098 }
3099 
3100 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3101 {
3102 	struct ibmvnic_sub_crq_queue *scrq = instance;
3103 	struct ibmvnic_adapter *adapter = scrq->adapter;
3104 
3105 	disable_scrq_irq(adapter, scrq);
3106 	ibmvnic_complete_tx(adapter, scrq);
3107 
3108 	return IRQ_HANDLED;
3109 }
3110 
3111 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3112 {
3113 	struct ibmvnic_sub_crq_queue *scrq = instance;
3114 	struct ibmvnic_adapter *adapter = scrq->adapter;
3115 
3116 	/* When booting a kdump kernel we can hit pending interrupts
3117 	 * prior to completing driver initialization.
3118 	 */
3119 	if (unlikely(adapter->state != VNIC_OPEN))
3120 		return IRQ_NONE;
3121 
3122 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3123 
3124 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3125 		disable_scrq_irq(adapter, scrq);
3126 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3127 	}
3128 
3129 	return IRQ_HANDLED;
3130 }
3131 
3132 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3133 {
3134 	struct device *dev = &adapter->vdev->dev;
3135 	struct ibmvnic_sub_crq_queue *scrq;
3136 	int i = 0, j = 0;
3137 	int rc = 0;
3138 
3139 	for (i = 0; i < adapter->req_tx_queues; i++) {
3140 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3141 			   i);
3142 		scrq = adapter->tx_scrq[i];
3143 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3144 
3145 		if (!scrq->irq) {
3146 			rc = -EINVAL;
3147 			dev_err(dev, "Error mapping irq\n");
3148 			goto req_tx_irq_failed;
3149 		}
3150 
3151 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3152 			 adapter->vdev->unit_address, i);
3153 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3154 				 0, scrq->name, scrq);
3155 
3156 		if (rc) {
3157 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3158 				scrq->irq, rc);
3159 			irq_dispose_mapping(scrq->irq);
3160 			goto req_tx_irq_failed;
3161 		}
3162 	}
3163 
3164 	for (i = 0; i < adapter->req_rx_queues; i++) {
3165 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3166 			   i);
3167 		scrq = adapter->rx_scrq[i];
3168 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3169 		if (!scrq->irq) {
3170 			rc = -EINVAL;
3171 			dev_err(dev, "Error mapping irq\n");
3172 			goto req_rx_irq_failed;
3173 		}
3174 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3175 			 adapter->vdev->unit_address, i);
3176 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3177 				 0, scrq->name, scrq);
3178 		if (rc) {
3179 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3180 				scrq->irq, rc);
3181 			irq_dispose_mapping(scrq->irq);
3182 			goto req_rx_irq_failed;
3183 		}
3184 	}
3185 	return rc;
3186 
3187 req_rx_irq_failed:
3188 	for (j = 0; j < i; j++) {
3189 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3190 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3191 	}
3192 	i = adapter->req_tx_queues;
3193 req_tx_irq_failed:
3194 	for (j = 0; j < i; j++) {
3195 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3196 		irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3197 	}
3198 	release_sub_crqs(adapter, 1);
3199 	return rc;
3200 }
3201 
3202 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3203 {
3204 	struct device *dev = &adapter->vdev->dev;
3205 	struct ibmvnic_sub_crq_queue **allqueues;
3206 	int registered_queues = 0;
3207 	int total_queues;
3208 	int more = 0;
3209 	int i;
3210 
3211 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3212 
3213 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3214 	if (!allqueues)
3215 		return -1;
3216 
3217 	for (i = 0; i < total_queues; i++) {
3218 		allqueues[i] = init_sub_crq_queue(adapter);
3219 		if (!allqueues[i]) {
3220 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3221 			break;
3222 		}
3223 		registered_queues++;
3224 	}
3225 
3226 	/* Make sure we were able to register the minimum number of queues */
3227 	if (registered_queues <
3228 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3229 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3230 		goto tx_failed;
3231 	}
3232 
3233 	/* Distribute the failed allocated queues*/
3234 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3235 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3236 		switch (i % 3) {
3237 		case 0:
3238 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3239 				adapter->req_rx_queues--;
3240 			else
3241 				more++;
3242 			break;
3243 		case 1:
3244 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3245 				adapter->req_tx_queues--;
3246 			else
3247 				more++;
3248 			break;
3249 		}
3250 	}
3251 
3252 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3253 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3254 	if (!adapter->tx_scrq)
3255 		goto tx_failed;
3256 
3257 	for (i = 0; i < adapter->req_tx_queues; i++) {
3258 		adapter->tx_scrq[i] = allqueues[i];
3259 		adapter->tx_scrq[i]->pool_index = i;
3260 		adapter->num_active_tx_scrqs++;
3261 	}
3262 
3263 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3264 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3265 	if (!adapter->rx_scrq)
3266 		goto rx_failed;
3267 
3268 	for (i = 0; i < adapter->req_rx_queues; i++) {
3269 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3270 		adapter->rx_scrq[i]->scrq_num = i;
3271 		adapter->num_active_rx_scrqs++;
3272 	}
3273 
3274 	kfree(allqueues);
3275 	return 0;
3276 
3277 rx_failed:
3278 	kfree(adapter->tx_scrq);
3279 	adapter->tx_scrq = NULL;
3280 tx_failed:
3281 	for (i = 0; i < registered_queues; i++)
3282 		release_sub_crq_queue(adapter, allqueues[i], 1);
3283 	kfree(allqueues);
3284 	return -1;
3285 }
3286 
3287 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3288 {
3289 	struct device *dev = &adapter->vdev->dev;
3290 	union ibmvnic_crq crq;
3291 	int max_entries;
3292 
3293 	if (!retry) {
3294 		/* Sub-CRQ entries are 32 byte long */
3295 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3296 
3297 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3298 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3299 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3300 			return;
3301 		}
3302 
3303 		if (adapter->desired.mtu)
3304 			adapter->req_mtu = adapter->desired.mtu;
3305 		else
3306 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3307 
3308 		if (!adapter->desired.tx_entries)
3309 			adapter->desired.tx_entries =
3310 					adapter->max_tx_entries_per_subcrq;
3311 		if (!adapter->desired.rx_entries)
3312 			adapter->desired.rx_entries =
3313 					adapter->max_rx_add_entries_per_subcrq;
3314 
3315 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3316 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3317 
3318 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3319 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3320 			adapter->desired.tx_entries = max_entries;
3321 		}
3322 
3323 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3324 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3325 			adapter->desired.rx_entries = max_entries;
3326 		}
3327 
3328 		if (adapter->desired.tx_entries)
3329 			adapter->req_tx_entries_per_subcrq =
3330 					adapter->desired.tx_entries;
3331 		else
3332 			adapter->req_tx_entries_per_subcrq =
3333 					adapter->max_tx_entries_per_subcrq;
3334 
3335 		if (adapter->desired.rx_entries)
3336 			adapter->req_rx_add_entries_per_subcrq =
3337 					adapter->desired.rx_entries;
3338 		else
3339 			adapter->req_rx_add_entries_per_subcrq =
3340 					adapter->max_rx_add_entries_per_subcrq;
3341 
3342 		if (adapter->desired.tx_queues)
3343 			adapter->req_tx_queues =
3344 					adapter->desired.tx_queues;
3345 		else
3346 			adapter->req_tx_queues =
3347 					adapter->opt_tx_comp_sub_queues;
3348 
3349 		if (adapter->desired.rx_queues)
3350 			adapter->req_rx_queues =
3351 					adapter->desired.rx_queues;
3352 		else
3353 			adapter->req_rx_queues =
3354 					adapter->opt_rx_comp_queues;
3355 
3356 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3357 	}
3358 
3359 	memset(&crq, 0, sizeof(crq));
3360 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3361 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3362 
3363 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3364 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3365 	atomic_inc(&adapter->running_cap_crqs);
3366 	ibmvnic_send_crq(adapter, &crq);
3367 
3368 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3369 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3370 	atomic_inc(&adapter->running_cap_crqs);
3371 	ibmvnic_send_crq(adapter, &crq);
3372 
3373 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3374 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3375 	atomic_inc(&adapter->running_cap_crqs);
3376 	ibmvnic_send_crq(adapter, &crq);
3377 
3378 	crq.request_capability.capability =
3379 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3380 	crq.request_capability.number =
3381 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3382 	atomic_inc(&adapter->running_cap_crqs);
3383 	ibmvnic_send_crq(adapter, &crq);
3384 
3385 	crq.request_capability.capability =
3386 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3387 	crq.request_capability.number =
3388 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3389 	atomic_inc(&adapter->running_cap_crqs);
3390 	ibmvnic_send_crq(adapter, &crq);
3391 
3392 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3393 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3394 	atomic_inc(&adapter->running_cap_crqs);
3395 	ibmvnic_send_crq(adapter, &crq);
3396 
3397 	if (adapter->netdev->flags & IFF_PROMISC) {
3398 		if (adapter->promisc_supported) {
3399 			crq.request_capability.capability =
3400 			    cpu_to_be16(PROMISC_REQUESTED);
3401 			crq.request_capability.number = cpu_to_be64(1);
3402 			atomic_inc(&adapter->running_cap_crqs);
3403 			ibmvnic_send_crq(adapter, &crq);
3404 		}
3405 	} else {
3406 		crq.request_capability.capability =
3407 		    cpu_to_be16(PROMISC_REQUESTED);
3408 		crq.request_capability.number = cpu_to_be64(0);
3409 		atomic_inc(&adapter->running_cap_crqs);
3410 		ibmvnic_send_crq(adapter, &crq);
3411 	}
3412 }
3413 
3414 static int pending_scrq(struct ibmvnic_adapter *adapter,
3415 			struct ibmvnic_sub_crq_queue *scrq)
3416 {
3417 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3418 
3419 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3420 		return 1;
3421 	else
3422 		return 0;
3423 }
3424 
3425 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3426 					struct ibmvnic_sub_crq_queue *scrq)
3427 {
3428 	union sub_crq *entry;
3429 	unsigned long flags;
3430 
3431 	spin_lock_irqsave(&scrq->lock, flags);
3432 	entry = &scrq->msgs[scrq->cur];
3433 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3434 		if (++scrq->cur == scrq->size)
3435 			scrq->cur = 0;
3436 	} else {
3437 		entry = NULL;
3438 	}
3439 	spin_unlock_irqrestore(&scrq->lock, flags);
3440 
3441 	return entry;
3442 }
3443 
3444 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3445 {
3446 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3447 	union ibmvnic_crq *crq;
3448 
3449 	crq = &queue->msgs[queue->cur];
3450 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3451 		if (++queue->cur == queue->size)
3452 			queue->cur = 0;
3453 	} else {
3454 		crq = NULL;
3455 	}
3456 
3457 	return crq;
3458 }
3459 
3460 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3461 {
3462 	switch (rc) {
3463 	case H_PARAMETER:
3464 		dev_warn_ratelimited(dev,
3465 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3466 				     func, rc);
3467 		break;
3468 	case H_CLOSED:
3469 		dev_warn_ratelimited(dev,
3470 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3471 				     func, rc);
3472 		break;
3473 	default:
3474 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3475 		break;
3476 	}
3477 }
3478 
3479 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3480 		       union sub_crq *sub_crq)
3481 {
3482 	unsigned int ua = adapter->vdev->unit_address;
3483 	struct device *dev = &adapter->vdev->dev;
3484 	u64 *u64_crq = (u64 *)sub_crq;
3485 	int rc;
3486 
3487 	netdev_dbg(adapter->netdev,
3488 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3489 		   (unsigned long int)cpu_to_be64(remote_handle),
3490 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3491 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3492 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3493 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3494 
3495 	/* Make sure the hypervisor sees the complete request */
3496 	mb();
3497 
3498 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3499 				cpu_to_be64(remote_handle),
3500 				cpu_to_be64(u64_crq[0]),
3501 				cpu_to_be64(u64_crq[1]),
3502 				cpu_to_be64(u64_crq[2]),
3503 				cpu_to_be64(u64_crq[3]));
3504 
3505 	if (rc)
3506 		print_subcrq_error(dev, rc, __func__);
3507 
3508 	return rc;
3509 }
3510 
3511 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3512 				u64 remote_handle, u64 ioba, u64 num_entries)
3513 {
3514 	unsigned int ua = adapter->vdev->unit_address;
3515 	struct device *dev = &adapter->vdev->dev;
3516 	int rc;
3517 
3518 	/* Make sure the hypervisor sees the complete request */
3519 	mb();
3520 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3521 				cpu_to_be64(remote_handle),
3522 				ioba, num_entries);
3523 
3524 	if (rc)
3525 		print_subcrq_error(dev, rc, __func__);
3526 
3527 	return rc;
3528 }
3529 
3530 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3531 			    union ibmvnic_crq *crq)
3532 {
3533 	unsigned int ua = adapter->vdev->unit_address;
3534 	struct device *dev = &adapter->vdev->dev;
3535 	u64 *u64_crq = (u64 *)crq;
3536 	int rc;
3537 
3538 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3539 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3540 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3541 
3542 	if (!adapter->crq.active &&
3543 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3544 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3545 		return -EINVAL;
3546 	}
3547 
3548 	/* Make sure the hypervisor sees the complete request */
3549 	mb();
3550 
3551 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3552 				cpu_to_be64(u64_crq[0]),
3553 				cpu_to_be64(u64_crq[1]));
3554 
3555 	if (rc) {
3556 		if (rc == H_CLOSED) {
3557 			dev_warn(dev, "CRQ Queue closed\n");
3558 			/* do not reset, report the fail, wait for passive init from server */
3559 		}
3560 
3561 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3562 	}
3563 
3564 	return rc;
3565 }
3566 
3567 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3568 {
3569 	struct device *dev = &adapter->vdev->dev;
3570 	union ibmvnic_crq crq;
3571 	int retries = 100;
3572 	int rc;
3573 
3574 	memset(&crq, 0, sizeof(crq));
3575 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3576 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3577 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3578 
3579 	do {
3580 		rc = ibmvnic_send_crq(adapter, &crq);
3581 		if (rc != H_CLOSED)
3582 			break;
3583 		retries--;
3584 		msleep(50);
3585 
3586 	} while (retries > 0);
3587 
3588 	if (rc) {
3589 		dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3590 		return rc;
3591 	}
3592 
3593 	return 0;
3594 }
3595 
3596 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3597 {
3598 	union ibmvnic_crq crq;
3599 
3600 	memset(&crq, 0, sizeof(crq));
3601 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3602 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3603 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3604 
3605 	return ibmvnic_send_crq(adapter, &crq);
3606 }
3607 
3608 struct vnic_login_client_data {
3609 	u8	type;
3610 	__be16	len;
3611 	char	name[];
3612 } __packed;
3613 
3614 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3615 {
3616 	int len;
3617 
3618 	/* Calculate the amount of buffer space needed for the
3619 	 * vnic client data in the login buffer. There are four entries,
3620 	 * OS name, LPAR name, device name, and a null last entry.
3621 	 */
3622 	len = 4 * sizeof(struct vnic_login_client_data);
3623 	len += 6; /* "Linux" plus NULL */
3624 	len += strlen(utsname()->nodename) + 1;
3625 	len += strlen(adapter->netdev->name) + 1;
3626 
3627 	return len;
3628 }
3629 
3630 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3631 				 struct vnic_login_client_data *vlcd)
3632 {
3633 	const char *os_name = "Linux";
3634 	int len;
3635 
3636 	/* Type 1 - LPAR OS */
3637 	vlcd->type = 1;
3638 	len = strlen(os_name) + 1;
3639 	vlcd->len = cpu_to_be16(len);
3640 	strncpy(vlcd->name, os_name, len);
3641 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3642 
3643 	/* Type 2 - LPAR name */
3644 	vlcd->type = 2;
3645 	len = strlen(utsname()->nodename) + 1;
3646 	vlcd->len = cpu_to_be16(len);
3647 	strncpy(vlcd->name, utsname()->nodename, len);
3648 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3649 
3650 	/* Type 3 - device name */
3651 	vlcd->type = 3;
3652 	len = strlen(adapter->netdev->name) + 1;
3653 	vlcd->len = cpu_to_be16(len);
3654 	strncpy(vlcd->name, adapter->netdev->name, len);
3655 }
3656 
3657 static int send_login(struct ibmvnic_adapter *adapter)
3658 {
3659 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3660 	struct ibmvnic_login_buffer *login_buffer;
3661 	struct device *dev = &adapter->vdev->dev;
3662 	dma_addr_t rsp_buffer_token;
3663 	dma_addr_t buffer_token;
3664 	size_t rsp_buffer_size;
3665 	union ibmvnic_crq crq;
3666 	size_t buffer_size;
3667 	__be64 *tx_list_p;
3668 	__be64 *rx_list_p;
3669 	int client_data_len;
3670 	struct vnic_login_client_data *vlcd;
3671 	int i;
3672 
3673 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3674 		netdev_err(adapter->netdev,
3675 			   "RX or TX queues are not allocated, device login failed\n");
3676 		return -1;
3677 	}
3678 
3679 	release_login_rsp_buffer(adapter);
3680 	client_data_len = vnic_client_data_len(adapter);
3681 
3682 	buffer_size =
3683 	    sizeof(struct ibmvnic_login_buffer) +
3684 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3685 	    client_data_len;
3686 
3687 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3688 	if (!login_buffer)
3689 		goto buf_alloc_failed;
3690 
3691 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3692 				      DMA_TO_DEVICE);
3693 	if (dma_mapping_error(dev, buffer_token)) {
3694 		dev_err(dev, "Couldn't map login buffer\n");
3695 		goto buf_map_failed;
3696 	}
3697 
3698 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3699 			  sizeof(u64) * adapter->req_tx_queues +
3700 			  sizeof(u64) * adapter->req_rx_queues +
3701 			  sizeof(u64) * adapter->req_rx_queues +
3702 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3703 
3704 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3705 	if (!login_rsp_buffer)
3706 		goto buf_rsp_alloc_failed;
3707 
3708 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3709 					  rsp_buffer_size, DMA_FROM_DEVICE);
3710 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3711 		dev_err(dev, "Couldn't map login rsp buffer\n");
3712 		goto buf_rsp_map_failed;
3713 	}
3714 
3715 	adapter->login_buf = login_buffer;
3716 	adapter->login_buf_token = buffer_token;
3717 	adapter->login_buf_sz = buffer_size;
3718 	adapter->login_rsp_buf = login_rsp_buffer;
3719 	adapter->login_rsp_buf_token = rsp_buffer_token;
3720 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3721 
3722 	login_buffer->len = cpu_to_be32(buffer_size);
3723 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3724 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3725 	login_buffer->off_txcomp_subcrqs =
3726 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3727 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3728 	login_buffer->off_rxcomp_subcrqs =
3729 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3730 			sizeof(u64) * adapter->req_tx_queues);
3731 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3732 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3733 
3734 	tx_list_p = (__be64 *)((char *)login_buffer +
3735 				      sizeof(struct ibmvnic_login_buffer));
3736 	rx_list_p = (__be64 *)((char *)login_buffer +
3737 				      sizeof(struct ibmvnic_login_buffer) +
3738 				      sizeof(u64) * adapter->req_tx_queues);
3739 
3740 	for (i = 0; i < adapter->req_tx_queues; i++) {
3741 		if (adapter->tx_scrq[i]) {
3742 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3743 						   crq_num);
3744 		}
3745 	}
3746 
3747 	for (i = 0; i < adapter->req_rx_queues; i++) {
3748 		if (adapter->rx_scrq[i]) {
3749 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3750 						   crq_num);
3751 		}
3752 	}
3753 
3754 	/* Insert vNIC login client data */
3755 	vlcd = (struct vnic_login_client_data *)
3756 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3757 	login_buffer->client_data_offset =
3758 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3759 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3760 
3761 	vnic_add_client_data(adapter, vlcd);
3762 
3763 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3764 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3765 		netdev_dbg(adapter->netdev, "%016lx\n",
3766 			   ((unsigned long int *)(adapter->login_buf))[i]);
3767 	}
3768 
3769 	memset(&crq, 0, sizeof(crq));
3770 	crq.login.first = IBMVNIC_CRQ_CMD;
3771 	crq.login.cmd = LOGIN;
3772 	crq.login.ioba = cpu_to_be32(buffer_token);
3773 	crq.login.len = cpu_to_be32(buffer_size);
3774 	ibmvnic_send_crq(adapter, &crq);
3775 
3776 	return 0;
3777 
3778 buf_rsp_map_failed:
3779 	kfree(login_rsp_buffer);
3780 buf_rsp_alloc_failed:
3781 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3782 buf_map_failed:
3783 	kfree(login_buffer);
3784 buf_alloc_failed:
3785 	return -1;
3786 }
3787 
3788 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3789 			    u32 len, u8 map_id)
3790 {
3791 	union ibmvnic_crq crq;
3792 
3793 	memset(&crq, 0, sizeof(crq));
3794 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3795 	crq.request_map.cmd = REQUEST_MAP;
3796 	crq.request_map.map_id = map_id;
3797 	crq.request_map.ioba = cpu_to_be32(addr);
3798 	crq.request_map.len = cpu_to_be32(len);
3799 	return ibmvnic_send_crq(adapter, &crq);
3800 }
3801 
3802 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3803 {
3804 	union ibmvnic_crq crq;
3805 
3806 	memset(&crq, 0, sizeof(crq));
3807 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3808 	crq.request_unmap.cmd = REQUEST_UNMAP;
3809 	crq.request_unmap.map_id = map_id;
3810 	return ibmvnic_send_crq(adapter, &crq);
3811 }
3812 
3813 static void send_map_query(struct ibmvnic_adapter *adapter)
3814 {
3815 	union ibmvnic_crq crq;
3816 
3817 	memset(&crq, 0, sizeof(crq));
3818 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3819 	crq.query_map.cmd = QUERY_MAP;
3820 	ibmvnic_send_crq(adapter, &crq);
3821 }
3822 
3823 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3824 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3825 {
3826 	union ibmvnic_crq crq;
3827 
3828 	atomic_set(&adapter->running_cap_crqs, 0);
3829 	memset(&crq, 0, sizeof(crq));
3830 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3831 	crq.query_capability.cmd = QUERY_CAPABILITY;
3832 
3833 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3834 	atomic_inc(&adapter->running_cap_crqs);
3835 	ibmvnic_send_crq(adapter, &crq);
3836 
3837 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3838 	atomic_inc(&adapter->running_cap_crqs);
3839 	ibmvnic_send_crq(adapter, &crq);
3840 
3841 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3842 	atomic_inc(&adapter->running_cap_crqs);
3843 	ibmvnic_send_crq(adapter, &crq);
3844 
3845 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3846 	atomic_inc(&adapter->running_cap_crqs);
3847 	ibmvnic_send_crq(adapter, &crq);
3848 
3849 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3850 	atomic_inc(&adapter->running_cap_crqs);
3851 	ibmvnic_send_crq(adapter, &crq);
3852 
3853 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3854 	atomic_inc(&adapter->running_cap_crqs);
3855 	ibmvnic_send_crq(adapter, &crq);
3856 
3857 	crq.query_capability.capability =
3858 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3859 	atomic_inc(&adapter->running_cap_crqs);
3860 	ibmvnic_send_crq(adapter, &crq);
3861 
3862 	crq.query_capability.capability =
3863 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3864 	atomic_inc(&adapter->running_cap_crqs);
3865 	ibmvnic_send_crq(adapter, &crq);
3866 
3867 	crq.query_capability.capability =
3868 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3869 	atomic_inc(&adapter->running_cap_crqs);
3870 	ibmvnic_send_crq(adapter, &crq);
3871 
3872 	crq.query_capability.capability =
3873 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3874 	atomic_inc(&adapter->running_cap_crqs);
3875 	ibmvnic_send_crq(adapter, &crq);
3876 
3877 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3878 	atomic_inc(&adapter->running_cap_crqs);
3879 	ibmvnic_send_crq(adapter, &crq);
3880 
3881 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3882 	atomic_inc(&adapter->running_cap_crqs);
3883 	ibmvnic_send_crq(adapter, &crq);
3884 
3885 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3886 	atomic_inc(&adapter->running_cap_crqs);
3887 	ibmvnic_send_crq(adapter, &crq);
3888 
3889 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3890 	atomic_inc(&adapter->running_cap_crqs);
3891 	ibmvnic_send_crq(adapter, &crq);
3892 
3893 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3894 	atomic_inc(&adapter->running_cap_crqs);
3895 	ibmvnic_send_crq(adapter, &crq);
3896 
3897 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3898 	atomic_inc(&adapter->running_cap_crqs);
3899 	ibmvnic_send_crq(adapter, &crq);
3900 
3901 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3902 	atomic_inc(&adapter->running_cap_crqs);
3903 	ibmvnic_send_crq(adapter, &crq);
3904 
3905 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3906 	atomic_inc(&adapter->running_cap_crqs);
3907 	ibmvnic_send_crq(adapter, &crq);
3908 
3909 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3910 	atomic_inc(&adapter->running_cap_crqs);
3911 	ibmvnic_send_crq(adapter, &crq);
3912 
3913 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3914 	atomic_inc(&adapter->running_cap_crqs);
3915 	ibmvnic_send_crq(adapter, &crq);
3916 
3917 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3918 	atomic_inc(&adapter->running_cap_crqs);
3919 	ibmvnic_send_crq(adapter, &crq);
3920 
3921 	crq.query_capability.capability =
3922 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3923 	atomic_inc(&adapter->running_cap_crqs);
3924 	ibmvnic_send_crq(adapter, &crq);
3925 
3926 	crq.query_capability.capability =
3927 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3928 	atomic_inc(&adapter->running_cap_crqs);
3929 	ibmvnic_send_crq(adapter, &crq);
3930 
3931 	crq.query_capability.capability =
3932 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3933 	atomic_inc(&adapter->running_cap_crqs);
3934 	ibmvnic_send_crq(adapter, &crq);
3935 
3936 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3937 	atomic_inc(&adapter->running_cap_crqs);
3938 	ibmvnic_send_crq(adapter, &crq);
3939 }
3940 
3941 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3942 				struct ibmvnic_adapter *adapter)
3943 {
3944 	struct device *dev = &adapter->vdev->dev;
3945 
3946 	if (crq->get_vpd_size_rsp.rc.code) {
3947 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3948 			crq->get_vpd_size_rsp.rc.code);
3949 		complete(&adapter->fw_done);
3950 		return;
3951 	}
3952 
3953 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3954 	complete(&adapter->fw_done);
3955 }
3956 
3957 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3958 			   struct ibmvnic_adapter *adapter)
3959 {
3960 	struct device *dev = &adapter->vdev->dev;
3961 	unsigned char *substr = NULL;
3962 	u8 fw_level_len = 0;
3963 
3964 	memset(adapter->fw_version, 0, 32);
3965 
3966 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3967 			 DMA_FROM_DEVICE);
3968 
3969 	if (crq->get_vpd_rsp.rc.code) {
3970 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3971 			crq->get_vpd_rsp.rc.code);
3972 		goto complete;
3973 	}
3974 
3975 	/* get the position of the firmware version info
3976 	 * located after the ASCII 'RM' substring in the buffer
3977 	 */
3978 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3979 	if (!substr) {
3980 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3981 		goto complete;
3982 	}
3983 
3984 	/* get length of firmware level ASCII substring */
3985 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3986 		fw_level_len = *(substr + 2);
3987 	} else {
3988 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3989 		goto complete;
3990 	}
3991 
3992 	/* copy firmware version string from vpd into adapter */
3993 	if ((substr + 3 + fw_level_len) <
3994 	    (adapter->vpd->buff + adapter->vpd->len)) {
3995 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3996 	} else {
3997 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3998 	}
3999 
4000 complete:
4001 	if (adapter->fw_version[0] == '\0')
4002 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4003 	complete(&adapter->fw_done);
4004 }
4005 
4006 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4007 {
4008 	struct device *dev = &adapter->vdev->dev;
4009 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4010 	netdev_features_t old_hw_features = 0;
4011 	union ibmvnic_crq crq;
4012 	int i;
4013 
4014 	dma_unmap_single(dev, adapter->ip_offload_tok,
4015 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4016 
4017 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4018 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4019 		netdev_dbg(adapter->netdev, "%016lx\n",
4020 			   ((unsigned long int *)(buf))[i]);
4021 
4022 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4023 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4024 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4025 		   buf->tcp_ipv4_chksum);
4026 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4027 		   buf->tcp_ipv6_chksum);
4028 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4029 		   buf->udp_ipv4_chksum);
4030 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4031 		   buf->udp_ipv6_chksum);
4032 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4033 		   buf->large_tx_ipv4);
4034 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4035 		   buf->large_tx_ipv6);
4036 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4037 		   buf->large_rx_ipv4);
4038 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4039 		   buf->large_rx_ipv6);
4040 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4041 		   buf->max_ipv4_header_size);
4042 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4043 		   buf->max_ipv6_header_size);
4044 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4045 		   buf->max_tcp_header_size);
4046 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4047 		   buf->max_udp_header_size);
4048 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4049 		   buf->max_large_tx_size);
4050 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4051 		   buf->max_large_rx_size);
4052 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4053 		   buf->ipv6_extension_header);
4054 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4055 		   buf->tcp_pseudosum_req);
4056 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4057 		   buf->num_ipv6_ext_headers);
4058 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4059 		   buf->off_ipv6_ext_headers);
4060 
4061 	adapter->ip_offload_ctrl_tok =
4062 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
4063 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
4064 
4065 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4066 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4067 		return;
4068 	}
4069 
4070 	adapter->ip_offload_ctrl.len =
4071 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4072 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
4073 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
4074 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
4075 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4076 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
4077 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4078 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
4079 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
4080 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
4081 
4082 	/* large_rx disabled for now, additional features needed */
4083 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
4084 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
4085 
4086 	if (adapter->state != VNIC_PROBING) {
4087 		old_hw_features = adapter->netdev->hw_features;
4088 		adapter->netdev->hw_features = 0;
4089 	}
4090 
4091 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4092 
4093 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4094 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4095 
4096 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4097 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4098 
4099 	if ((adapter->netdev->features &
4100 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4101 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4102 
4103 	if (buf->large_tx_ipv4)
4104 		adapter->netdev->hw_features |= NETIF_F_TSO;
4105 	if (buf->large_tx_ipv6)
4106 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4107 
4108 	if (adapter->state == VNIC_PROBING) {
4109 		adapter->netdev->features |= adapter->netdev->hw_features;
4110 	} else if (old_hw_features != adapter->netdev->hw_features) {
4111 		netdev_features_t tmp = 0;
4112 
4113 		/* disable features no longer supported */
4114 		adapter->netdev->features &= adapter->netdev->hw_features;
4115 		/* turn on features now supported if previously enabled */
4116 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4117 			adapter->netdev->hw_features;
4118 		adapter->netdev->features |=
4119 				tmp & adapter->netdev->wanted_features;
4120 	}
4121 
4122 	memset(&crq, 0, sizeof(crq));
4123 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4124 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4125 	crq.control_ip_offload.len =
4126 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4127 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4128 	ibmvnic_send_crq(adapter, &crq);
4129 }
4130 
4131 static const char *ibmvnic_fw_err_cause(u16 cause)
4132 {
4133 	switch (cause) {
4134 	case ADAPTER_PROBLEM:
4135 		return "adapter problem";
4136 	case BUS_PROBLEM:
4137 		return "bus problem";
4138 	case FW_PROBLEM:
4139 		return "firmware problem";
4140 	case DD_PROBLEM:
4141 		return "device driver problem";
4142 	case EEH_RECOVERY:
4143 		return "EEH recovery";
4144 	case FW_UPDATED:
4145 		return "firmware updated";
4146 	case LOW_MEMORY:
4147 		return "low Memory";
4148 	default:
4149 		return "unknown";
4150 	}
4151 }
4152 
4153 static void handle_error_indication(union ibmvnic_crq *crq,
4154 				    struct ibmvnic_adapter *adapter)
4155 {
4156 	struct device *dev = &adapter->vdev->dev;
4157 	u16 cause;
4158 
4159 	cause = be16_to_cpu(crq->error_indication.error_cause);
4160 
4161 	dev_warn_ratelimited(dev,
4162 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4163 			     crq->error_indication.flags
4164 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4165 			     ibmvnic_fw_err_cause(cause));
4166 
4167 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4168 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4169 	else
4170 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4171 }
4172 
4173 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4174 				 struct ibmvnic_adapter *adapter)
4175 {
4176 	struct net_device *netdev = adapter->netdev;
4177 	struct device *dev = &adapter->vdev->dev;
4178 	long rc;
4179 
4180 	rc = crq->change_mac_addr_rsp.rc.code;
4181 	if (rc) {
4182 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4183 		goto out;
4184 	}
4185 	ether_addr_copy(netdev->dev_addr,
4186 			&crq->change_mac_addr_rsp.mac_addr[0]);
4187 out:
4188 	complete(&adapter->fw_done);
4189 	return rc;
4190 }
4191 
4192 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4193 				   struct ibmvnic_adapter *adapter)
4194 {
4195 	struct device *dev = &adapter->vdev->dev;
4196 	u64 *req_value;
4197 	char *name;
4198 
4199 	atomic_dec(&adapter->running_cap_crqs);
4200 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4201 	case REQ_TX_QUEUES:
4202 		req_value = &adapter->req_tx_queues;
4203 		name = "tx";
4204 		break;
4205 	case REQ_RX_QUEUES:
4206 		req_value = &adapter->req_rx_queues;
4207 		name = "rx";
4208 		break;
4209 	case REQ_RX_ADD_QUEUES:
4210 		req_value = &adapter->req_rx_add_queues;
4211 		name = "rx_add";
4212 		break;
4213 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4214 		req_value = &adapter->req_tx_entries_per_subcrq;
4215 		name = "tx_entries_per_subcrq";
4216 		break;
4217 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4218 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4219 		name = "rx_add_entries_per_subcrq";
4220 		break;
4221 	case REQ_MTU:
4222 		req_value = &adapter->req_mtu;
4223 		name = "mtu";
4224 		break;
4225 	case PROMISC_REQUESTED:
4226 		req_value = &adapter->promisc;
4227 		name = "promisc";
4228 		break;
4229 	default:
4230 		dev_err(dev, "Got invalid cap request rsp %d\n",
4231 			crq->request_capability.capability);
4232 		return;
4233 	}
4234 
4235 	switch (crq->request_capability_rsp.rc.code) {
4236 	case SUCCESS:
4237 		break;
4238 	case PARTIALSUCCESS:
4239 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4240 			 *req_value,
4241 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4242 					       number), name);
4243 
4244 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4245 		    REQ_MTU) {
4246 			pr_err("mtu of %llu is not supported. Reverting.\n",
4247 			       *req_value);
4248 			*req_value = adapter->fallback.mtu;
4249 		} else {
4250 			*req_value =
4251 				be64_to_cpu(crq->request_capability_rsp.number);
4252 		}
4253 
4254 		ibmvnic_send_req_caps(adapter, 1);
4255 		return;
4256 	default:
4257 		dev_err(dev, "Error %d in request cap rsp\n",
4258 			crq->request_capability_rsp.rc.code);
4259 		return;
4260 	}
4261 
4262 	/* Done receiving requested capabilities, query IP offload support */
4263 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4264 		union ibmvnic_crq newcrq;
4265 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4266 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4267 		    &adapter->ip_offload_buf;
4268 
4269 		adapter->wait_capability = false;
4270 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4271 							 buf_sz,
4272 							 DMA_FROM_DEVICE);
4273 
4274 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4275 			if (!firmware_has_feature(FW_FEATURE_CMO))
4276 				dev_err(dev, "Couldn't map offload buffer\n");
4277 			return;
4278 		}
4279 
4280 		memset(&newcrq, 0, sizeof(newcrq));
4281 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4282 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4283 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4284 		newcrq.query_ip_offload.ioba =
4285 		    cpu_to_be32(adapter->ip_offload_tok);
4286 
4287 		ibmvnic_send_crq(adapter, &newcrq);
4288 	}
4289 }
4290 
4291 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4292 			    struct ibmvnic_adapter *adapter)
4293 {
4294 	struct device *dev = &adapter->vdev->dev;
4295 	struct net_device *netdev = adapter->netdev;
4296 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4297 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4298 	u64 *tx_handle_array;
4299 	u64 *rx_handle_array;
4300 	int num_tx_pools;
4301 	int num_rx_pools;
4302 	u64 *size_array;
4303 	int i;
4304 
4305 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4306 			 DMA_TO_DEVICE);
4307 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4308 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4309 
4310 	/* If the number of queues requested can't be allocated by the
4311 	 * server, the login response will return with code 1. We will need
4312 	 * to resend the login buffer with fewer queues requested.
4313 	 */
4314 	if (login_rsp_crq->generic.rc.code) {
4315 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4316 		complete(&adapter->init_done);
4317 		return 0;
4318 	}
4319 
4320 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4321 
4322 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4323 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4324 		netdev_dbg(adapter->netdev, "%016lx\n",
4325 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4326 	}
4327 
4328 	/* Sanity checks */
4329 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4330 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4331 	     adapter->req_rx_add_queues !=
4332 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4333 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4334 		ibmvnic_remove(adapter->vdev);
4335 		return -EIO;
4336 	}
4337 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4338 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4339 	/* variable buffer sizes are not supported, so just read the
4340 	 * first entry.
4341 	 */
4342 	adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4343 
4344 	num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4345 	num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4346 
4347 	tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4348 				  be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4349 	rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4350 				  be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4351 
4352 	for (i = 0; i < num_tx_pools; i++)
4353 		adapter->tx_scrq[i]->handle = tx_handle_array[i];
4354 
4355 	for (i = 0; i < num_rx_pools; i++)
4356 		adapter->rx_scrq[i]->handle = rx_handle_array[i];
4357 
4358 	adapter->num_active_tx_scrqs = num_tx_pools;
4359 	adapter->num_active_rx_scrqs = num_rx_pools;
4360 	release_login_rsp_buffer(adapter);
4361 	release_login_buffer(adapter);
4362 	complete(&adapter->init_done);
4363 
4364 	return 0;
4365 }
4366 
4367 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4368 				     struct ibmvnic_adapter *adapter)
4369 {
4370 	struct device *dev = &adapter->vdev->dev;
4371 	long rc;
4372 
4373 	rc = crq->request_unmap_rsp.rc.code;
4374 	if (rc)
4375 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4376 }
4377 
4378 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4379 				 struct ibmvnic_adapter *adapter)
4380 {
4381 	struct net_device *netdev = adapter->netdev;
4382 	struct device *dev = &adapter->vdev->dev;
4383 	long rc;
4384 
4385 	rc = crq->query_map_rsp.rc.code;
4386 	if (rc) {
4387 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4388 		return;
4389 	}
4390 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4391 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4392 		   crq->query_map_rsp.free_pages);
4393 }
4394 
4395 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4396 				 struct ibmvnic_adapter *adapter)
4397 {
4398 	struct net_device *netdev = adapter->netdev;
4399 	struct device *dev = &adapter->vdev->dev;
4400 	long rc;
4401 
4402 	atomic_dec(&adapter->running_cap_crqs);
4403 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4404 		   atomic_read(&adapter->running_cap_crqs));
4405 	rc = crq->query_capability.rc.code;
4406 	if (rc) {
4407 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4408 		goto out;
4409 	}
4410 
4411 	switch (be16_to_cpu(crq->query_capability.capability)) {
4412 	case MIN_TX_QUEUES:
4413 		adapter->min_tx_queues =
4414 		    be64_to_cpu(crq->query_capability.number);
4415 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4416 			   adapter->min_tx_queues);
4417 		break;
4418 	case MIN_RX_QUEUES:
4419 		adapter->min_rx_queues =
4420 		    be64_to_cpu(crq->query_capability.number);
4421 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4422 			   adapter->min_rx_queues);
4423 		break;
4424 	case MIN_RX_ADD_QUEUES:
4425 		adapter->min_rx_add_queues =
4426 		    be64_to_cpu(crq->query_capability.number);
4427 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4428 			   adapter->min_rx_add_queues);
4429 		break;
4430 	case MAX_TX_QUEUES:
4431 		adapter->max_tx_queues =
4432 		    be64_to_cpu(crq->query_capability.number);
4433 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4434 			   adapter->max_tx_queues);
4435 		break;
4436 	case MAX_RX_QUEUES:
4437 		adapter->max_rx_queues =
4438 		    be64_to_cpu(crq->query_capability.number);
4439 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4440 			   adapter->max_rx_queues);
4441 		break;
4442 	case MAX_RX_ADD_QUEUES:
4443 		adapter->max_rx_add_queues =
4444 		    be64_to_cpu(crq->query_capability.number);
4445 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4446 			   adapter->max_rx_add_queues);
4447 		break;
4448 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4449 		adapter->min_tx_entries_per_subcrq =
4450 		    be64_to_cpu(crq->query_capability.number);
4451 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4452 			   adapter->min_tx_entries_per_subcrq);
4453 		break;
4454 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4455 		adapter->min_rx_add_entries_per_subcrq =
4456 		    be64_to_cpu(crq->query_capability.number);
4457 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4458 			   adapter->min_rx_add_entries_per_subcrq);
4459 		break;
4460 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4461 		adapter->max_tx_entries_per_subcrq =
4462 		    be64_to_cpu(crq->query_capability.number);
4463 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4464 			   adapter->max_tx_entries_per_subcrq);
4465 		break;
4466 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4467 		adapter->max_rx_add_entries_per_subcrq =
4468 		    be64_to_cpu(crq->query_capability.number);
4469 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4470 			   adapter->max_rx_add_entries_per_subcrq);
4471 		break;
4472 	case TCP_IP_OFFLOAD:
4473 		adapter->tcp_ip_offload =
4474 		    be64_to_cpu(crq->query_capability.number);
4475 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4476 			   adapter->tcp_ip_offload);
4477 		break;
4478 	case PROMISC_SUPPORTED:
4479 		adapter->promisc_supported =
4480 		    be64_to_cpu(crq->query_capability.number);
4481 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4482 			   adapter->promisc_supported);
4483 		break;
4484 	case MIN_MTU:
4485 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4486 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4487 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4488 		break;
4489 	case MAX_MTU:
4490 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4491 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4492 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4493 		break;
4494 	case MAX_MULTICAST_FILTERS:
4495 		adapter->max_multicast_filters =
4496 		    be64_to_cpu(crq->query_capability.number);
4497 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4498 			   adapter->max_multicast_filters);
4499 		break;
4500 	case VLAN_HEADER_INSERTION:
4501 		adapter->vlan_header_insertion =
4502 		    be64_to_cpu(crq->query_capability.number);
4503 		if (adapter->vlan_header_insertion)
4504 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4505 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4506 			   adapter->vlan_header_insertion);
4507 		break;
4508 	case RX_VLAN_HEADER_INSERTION:
4509 		adapter->rx_vlan_header_insertion =
4510 		    be64_to_cpu(crq->query_capability.number);
4511 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4512 			   adapter->rx_vlan_header_insertion);
4513 		break;
4514 	case MAX_TX_SG_ENTRIES:
4515 		adapter->max_tx_sg_entries =
4516 		    be64_to_cpu(crq->query_capability.number);
4517 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4518 			   adapter->max_tx_sg_entries);
4519 		break;
4520 	case RX_SG_SUPPORTED:
4521 		adapter->rx_sg_supported =
4522 		    be64_to_cpu(crq->query_capability.number);
4523 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4524 			   adapter->rx_sg_supported);
4525 		break;
4526 	case OPT_TX_COMP_SUB_QUEUES:
4527 		adapter->opt_tx_comp_sub_queues =
4528 		    be64_to_cpu(crq->query_capability.number);
4529 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4530 			   adapter->opt_tx_comp_sub_queues);
4531 		break;
4532 	case OPT_RX_COMP_QUEUES:
4533 		adapter->opt_rx_comp_queues =
4534 		    be64_to_cpu(crq->query_capability.number);
4535 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4536 			   adapter->opt_rx_comp_queues);
4537 		break;
4538 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4539 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4540 		    be64_to_cpu(crq->query_capability.number);
4541 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4542 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4543 		break;
4544 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4545 		adapter->opt_tx_entries_per_subcrq =
4546 		    be64_to_cpu(crq->query_capability.number);
4547 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4548 			   adapter->opt_tx_entries_per_subcrq);
4549 		break;
4550 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4551 		adapter->opt_rxba_entries_per_subcrq =
4552 		    be64_to_cpu(crq->query_capability.number);
4553 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4554 			   adapter->opt_rxba_entries_per_subcrq);
4555 		break;
4556 	case TX_RX_DESC_REQ:
4557 		adapter->tx_rx_desc_req = crq->query_capability.number;
4558 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4559 			   adapter->tx_rx_desc_req);
4560 		break;
4561 
4562 	default:
4563 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4564 			   crq->query_capability.capability);
4565 	}
4566 
4567 out:
4568 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4569 		adapter->wait_capability = false;
4570 		ibmvnic_send_req_caps(adapter, 0);
4571 	}
4572 }
4573 
4574 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4575 {
4576 	union ibmvnic_crq crq;
4577 	int rc;
4578 
4579 	memset(&crq, 0, sizeof(crq));
4580 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4581 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4582 
4583 	mutex_lock(&adapter->fw_lock);
4584 	adapter->fw_done_rc = 0;
4585 	reinit_completion(&adapter->fw_done);
4586 
4587 	rc = ibmvnic_send_crq(adapter, &crq);
4588 	if (rc) {
4589 		mutex_unlock(&adapter->fw_lock);
4590 		return rc;
4591 	}
4592 
4593 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4594 	if (rc) {
4595 		mutex_unlock(&adapter->fw_lock);
4596 		return rc;
4597 	}
4598 
4599 	mutex_unlock(&adapter->fw_lock);
4600 	return adapter->fw_done_rc ? -EIO : 0;
4601 }
4602 
4603 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4604 				       struct ibmvnic_adapter *adapter)
4605 {
4606 	struct net_device *netdev = adapter->netdev;
4607 	int rc;
4608 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4609 
4610 	rc = crq->query_phys_parms_rsp.rc.code;
4611 	if (rc) {
4612 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4613 		return rc;
4614 	}
4615 	switch (rspeed) {
4616 	case IBMVNIC_10MBPS:
4617 		adapter->speed = SPEED_10;
4618 		break;
4619 	case IBMVNIC_100MBPS:
4620 		adapter->speed = SPEED_100;
4621 		break;
4622 	case IBMVNIC_1GBPS:
4623 		adapter->speed = SPEED_1000;
4624 		break;
4625 	case IBMVNIC_10GBP:
4626 		adapter->speed = SPEED_10000;
4627 		break;
4628 	case IBMVNIC_25GBPS:
4629 		adapter->speed = SPEED_25000;
4630 		break;
4631 	case IBMVNIC_40GBPS:
4632 		adapter->speed = SPEED_40000;
4633 		break;
4634 	case IBMVNIC_50GBPS:
4635 		adapter->speed = SPEED_50000;
4636 		break;
4637 	case IBMVNIC_100GBPS:
4638 		adapter->speed = SPEED_100000;
4639 		break;
4640 	default:
4641 		if (netif_carrier_ok(netdev))
4642 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4643 		adapter->speed = SPEED_UNKNOWN;
4644 	}
4645 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4646 		adapter->duplex = DUPLEX_FULL;
4647 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4648 		adapter->duplex = DUPLEX_HALF;
4649 	else
4650 		adapter->duplex = DUPLEX_UNKNOWN;
4651 
4652 	return rc;
4653 }
4654 
4655 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4656 			       struct ibmvnic_adapter *adapter)
4657 {
4658 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4659 	struct net_device *netdev = adapter->netdev;
4660 	struct device *dev = &adapter->vdev->dev;
4661 	u64 *u64_crq = (u64 *)crq;
4662 	long rc;
4663 
4664 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4665 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4666 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4667 	switch (gen_crq->first) {
4668 	case IBMVNIC_CRQ_INIT_RSP:
4669 		switch (gen_crq->cmd) {
4670 		case IBMVNIC_CRQ_INIT:
4671 			dev_info(dev, "Partner initialized\n");
4672 			adapter->from_passive_init = true;
4673 			adapter->failover_pending = false;
4674 			if (!completion_done(&adapter->init_done)) {
4675 				complete(&adapter->init_done);
4676 				adapter->init_done_rc = -EIO;
4677 			}
4678 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4679 			break;
4680 		case IBMVNIC_CRQ_INIT_COMPLETE:
4681 			dev_info(dev, "Partner initialization complete\n");
4682 			adapter->crq.active = true;
4683 			send_version_xchg(adapter);
4684 			break;
4685 		default:
4686 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4687 		}
4688 		return;
4689 	case IBMVNIC_CRQ_XPORT_EVENT:
4690 		netif_carrier_off(netdev);
4691 		adapter->crq.active = false;
4692 		/* terminate any thread waiting for a response
4693 		 * from the device
4694 		 */
4695 		if (!completion_done(&adapter->fw_done)) {
4696 			adapter->fw_done_rc = -EIO;
4697 			complete(&adapter->fw_done);
4698 		}
4699 		if (!completion_done(&adapter->stats_done))
4700 			complete(&adapter->stats_done);
4701 		if (test_bit(0, &adapter->resetting))
4702 			adapter->force_reset_recovery = true;
4703 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4704 			dev_info(dev, "Migrated, re-enabling adapter\n");
4705 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4706 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4707 			dev_info(dev, "Backing device failover detected\n");
4708 			adapter->failover_pending = true;
4709 		} else {
4710 			/* The adapter lost the connection */
4711 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4712 				gen_crq->cmd);
4713 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4714 		}
4715 		return;
4716 	case IBMVNIC_CRQ_CMD_RSP:
4717 		break;
4718 	default:
4719 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4720 			gen_crq->first);
4721 		return;
4722 	}
4723 
4724 	switch (gen_crq->cmd) {
4725 	case VERSION_EXCHANGE_RSP:
4726 		rc = crq->version_exchange_rsp.rc.code;
4727 		if (rc) {
4728 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4729 			break;
4730 		}
4731 		ibmvnic_version =
4732 			    be16_to_cpu(crq->version_exchange_rsp.version);
4733 		dev_info(dev, "Partner protocol version is %d\n",
4734 			 ibmvnic_version);
4735 		send_cap_queries(adapter);
4736 		break;
4737 	case QUERY_CAPABILITY_RSP:
4738 		handle_query_cap_rsp(crq, adapter);
4739 		break;
4740 	case QUERY_MAP_RSP:
4741 		handle_query_map_rsp(crq, adapter);
4742 		break;
4743 	case REQUEST_MAP_RSP:
4744 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4745 		complete(&adapter->fw_done);
4746 		break;
4747 	case REQUEST_UNMAP_RSP:
4748 		handle_request_unmap_rsp(crq, adapter);
4749 		break;
4750 	case REQUEST_CAPABILITY_RSP:
4751 		handle_request_cap_rsp(crq, adapter);
4752 		break;
4753 	case LOGIN_RSP:
4754 		netdev_dbg(netdev, "Got Login Response\n");
4755 		handle_login_rsp(crq, adapter);
4756 		break;
4757 	case LOGICAL_LINK_STATE_RSP:
4758 		netdev_dbg(netdev,
4759 			   "Got Logical Link State Response, state: %d rc: %d\n",
4760 			   crq->logical_link_state_rsp.link_state,
4761 			   crq->logical_link_state_rsp.rc.code);
4762 		adapter->logical_link_state =
4763 		    crq->logical_link_state_rsp.link_state;
4764 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4765 		complete(&adapter->init_done);
4766 		break;
4767 	case LINK_STATE_INDICATION:
4768 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4769 		adapter->phys_link_state =
4770 		    crq->link_state_indication.phys_link_state;
4771 		adapter->logical_link_state =
4772 		    crq->link_state_indication.logical_link_state;
4773 		if (adapter->phys_link_state && adapter->logical_link_state)
4774 			netif_carrier_on(netdev);
4775 		else
4776 			netif_carrier_off(netdev);
4777 		break;
4778 	case CHANGE_MAC_ADDR_RSP:
4779 		netdev_dbg(netdev, "Got MAC address change Response\n");
4780 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4781 		break;
4782 	case ERROR_INDICATION:
4783 		netdev_dbg(netdev, "Got Error Indication\n");
4784 		handle_error_indication(crq, adapter);
4785 		break;
4786 	case REQUEST_STATISTICS_RSP:
4787 		netdev_dbg(netdev, "Got Statistics Response\n");
4788 		complete(&adapter->stats_done);
4789 		break;
4790 	case QUERY_IP_OFFLOAD_RSP:
4791 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4792 		handle_query_ip_offload_rsp(adapter);
4793 		break;
4794 	case MULTICAST_CTRL_RSP:
4795 		netdev_dbg(netdev, "Got multicast control Response\n");
4796 		break;
4797 	case CONTROL_IP_OFFLOAD_RSP:
4798 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4799 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4800 				 sizeof(adapter->ip_offload_ctrl),
4801 				 DMA_TO_DEVICE);
4802 		complete(&adapter->init_done);
4803 		break;
4804 	case COLLECT_FW_TRACE_RSP:
4805 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4806 		complete(&adapter->fw_done);
4807 		break;
4808 	case GET_VPD_SIZE_RSP:
4809 		handle_vpd_size_rsp(crq, adapter);
4810 		break;
4811 	case GET_VPD_RSP:
4812 		handle_vpd_rsp(crq, adapter);
4813 		break;
4814 	case QUERY_PHYS_PARMS_RSP:
4815 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4816 		complete(&adapter->fw_done);
4817 		break;
4818 	default:
4819 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4820 			   gen_crq->cmd);
4821 	}
4822 }
4823 
4824 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4825 {
4826 	struct ibmvnic_adapter *adapter = instance;
4827 
4828 	tasklet_schedule(&adapter->tasklet);
4829 	return IRQ_HANDLED;
4830 }
4831 
4832 static void ibmvnic_tasklet(void *data)
4833 {
4834 	struct ibmvnic_adapter *adapter = data;
4835 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4836 	union ibmvnic_crq *crq;
4837 	unsigned long flags;
4838 	bool done = false;
4839 
4840 	spin_lock_irqsave(&queue->lock, flags);
4841 	while (!done) {
4842 		/* Pull all the valid messages off the CRQ */
4843 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4844 			ibmvnic_handle_crq(crq, adapter);
4845 			crq->generic.first = 0;
4846 		}
4847 
4848 		/* remain in tasklet until all
4849 		 * capabilities responses are received
4850 		 */
4851 		if (!adapter->wait_capability)
4852 			done = true;
4853 	}
4854 	/* if capabilities CRQ's were sent in this tasklet, the following
4855 	 * tasklet must wait until all responses are received
4856 	 */
4857 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4858 		adapter->wait_capability = true;
4859 	spin_unlock_irqrestore(&queue->lock, flags);
4860 }
4861 
4862 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4863 {
4864 	struct vio_dev *vdev = adapter->vdev;
4865 	int rc;
4866 
4867 	do {
4868 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4869 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4870 
4871 	if (rc)
4872 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4873 
4874 	return rc;
4875 }
4876 
4877 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4878 {
4879 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4880 	struct device *dev = &adapter->vdev->dev;
4881 	struct vio_dev *vdev = adapter->vdev;
4882 	int rc;
4883 
4884 	/* Close the CRQ */
4885 	do {
4886 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4887 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4888 
4889 	/* Clean out the queue */
4890 	memset(crq->msgs, 0, PAGE_SIZE);
4891 	crq->cur = 0;
4892 	crq->active = false;
4893 
4894 	/* And re-open it again */
4895 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4896 				crq->msg_token, PAGE_SIZE);
4897 
4898 	if (rc == H_CLOSED)
4899 		/* Adapter is good, but other end is not ready */
4900 		dev_warn(dev, "Partner adapter not ready\n");
4901 	else if (rc != 0)
4902 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4903 
4904 	return rc;
4905 }
4906 
4907 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4908 {
4909 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4910 	struct vio_dev *vdev = adapter->vdev;
4911 	long rc;
4912 
4913 	if (!crq->msgs)
4914 		return;
4915 
4916 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4917 	free_irq(vdev->irq, adapter);
4918 	tasklet_kill(&adapter->tasklet);
4919 	do {
4920 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4921 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4922 
4923 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4924 			 DMA_BIDIRECTIONAL);
4925 	free_page((unsigned long)crq->msgs);
4926 	crq->msgs = NULL;
4927 	crq->active = false;
4928 }
4929 
4930 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4931 {
4932 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4933 	struct device *dev = &adapter->vdev->dev;
4934 	struct vio_dev *vdev = adapter->vdev;
4935 	int rc, retrc = -ENOMEM;
4936 
4937 	if (crq->msgs)
4938 		return 0;
4939 
4940 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4941 	/* Should we allocate more than one page? */
4942 
4943 	if (!crq->msgs)
4944 		return -ENOMEM;
4945 
4946 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4947 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4948 					DMA_BIDIRECTIONAL);
4949 	if (dma_mapping_error(dev, crq->msg_token))
4950 		goto map_failed;
4951 
4952 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4953 				crq->msg_token, PAGE_SIZE);
4954 
4955 	if (rc == H_RESOURCE)
4956 		/* maybe kexecing and resource is busy. try a reset */
4957 		rc = ibmvnic_reset_crq(adapter);
4958 	retrc = rc;
4959 
4960 	if (rc == H_CLOSED) {
4961 		dev_warn(dev, "Partner adapter not ready\n");
4962 	} else if (rc) {
4963 		dev_warn(dev, "Error %d opening adapter\n", rc);
4964 		goto reg_crq_failed;
4965 	}
4966 
4967 	retrc = 0;
4968 
4969 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4970 		     (unsigned long)adapter);
4971 
4972 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4973 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4974 		 adapter->vdev->unit_address);
4975 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4976 	if (rc) {
4977 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4978 			vdev->irq, rc);
4979 		goto req_irq_failed;
4980 	}
4981 
4982 	rc = vio_enable_interrupts(vdev);
4983 	if (rc) {
4984 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4985 		goto req_irq_failed;
4986 	}
4987 
4988 	crq->cur = 0;
4989 	spin_lock_init(&crq->lock);
4990 
4991 	return retrc;
4992 
4993 req_irq_failed:
4994 	tasklet_kill(&adapter->tasklet);
4995 	do {
4996 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4997 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4998 reg_crq_failed:
4999 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5000 map_failed:
5001 	free_page((unsigned long)crq->msgs);
5002 	crq->msgs = NULL;
5003 	return retrc;
5004 }
5005 
5006 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5007 {
5008 	struct device *dev = &adapter->vdev->dev;
5009 	unsigned long timeout = msecs_to_jiffies(30000);
5010 	u64 old_num_rx_queues, old_num_tx_queues;
5011 	int rc;
5012 
5013 	adapter->from_passive_init = false;
5014 
5015 	if (reset) {
5016 		old_num_rx_queues = adapter->req_rx_queues;
5017 		old_num_tx_queues = adapter->req_tx_queues;
5018 		reinit_completion(&adapter->init_done);
5019 	}
5020 
5021 	adapter->init_done_rc = 0;
5022 	rc = ibmvnic_send_crq_init(adapter);
5023 	if (rc) {
5024 		dev_err(dev, "Send crq init failed with error %d\n", rc);
5025 		return rc;
5026 	}
5027 
5028 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5029 		dev_err(dev, "Initialization sequence timed out\n");
5030 		return -1;
5031 	}
5032 
5033 	if (adapter->init_done_rc) {
5034 		release_crq_queue(adapter);
5035 		return adapter->init_done_rc;
5036 	}
5037 
5038 	if (reset &&
5039 	    test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5040 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5041 		if (adapter->req_rx_queues != old_num_rx_queues ||
5042 		    adapter->req_tx_queues != old_num_tx_queues) {
5043 			release_sub_crqs(adapter, 0);
5044 			rc = init_sub_crqs(adapter);
5045 		} else {
5046 			rc = reset_sub_crq_queues(adapter);
5047 		}
5048 	} else {
5049 		rc = init_sub_crqs(adapter);
5050 	}
5051 
5052 	if (rc) {
5053 		dev_err(dev, "Initialization of sub crqs failed\n");
5054 		release_crq_queue(adapter);
5055 		return rc;
5056 	}
5057 
5058 	rc = init_sub_crq_irqs(adapter);
5059 	if (rc) {
5060 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5061 		release_crq_queue(adapter);
5062 	}
5063 
5064 	return rc;
5065 }
5066 
5067 static struct device_attribute dev_attr_failover;
5068 
5069 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5070 {
5071 	struct ibmvnic_adapter *adapter;
5072 	struct net_device *netdev;
5073 	unsigned char *mac_addr_p;
5074 	int rc;
5075 
5076 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5077 		dev->unit_address);
5078 
5079 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5080 							VETH_MAC_ADDR, NULL);
5081 	if (!mac_addr_p) {
5082 		dev_err(&dev->dev,
5083 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5084 			__FILE__, __LINE__);
5085 		return 0;
5086 	}
5087 
5088 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5089 				   IBMVNIC_MAX_QUEUES);
5090 	if (!netdev)
5091 		return -ENOMEM;
5092 
5093 	adapter = netdev_priv(netdev);
5094 	adapter->state = VNIC_PROBING;
5095 	dev_set_drvdata(&dev->dev, netdev);
5096 	adapter->vdev = dev;
5097 	adapter->netdev = netdev;
5098 
5099 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5100 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5101 	netdev->irq = dev->irq;
5102 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5103 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5104 	SET_NETDEV_DEV(netdev, &dev->dev);
5105 
5106 	spin_lock_init(&adapter->stats_lock);
5107 
5108 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5109 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5110 			  __ibmvnic_delayed_reset);
5111 	INIT_LIST_HEAD(&adapter->rwi_list);
5112 	spin_lock_init(&adapter->rwi_lock);
5113 	spin_lock_init(&adapter->state_lock);
5114 	mutex_init(&adapter->fw_lock);
5115 	init_completion(&adapter->init_done);
5116 	init_completion(&adapter->fw_done);
5117 	init_completion(&adapter->reset_done);
5118 	init_completion(&adapter->stats_done);
5119 	clear_bit(0, &adapter->resetting);
5120 
5121 	do {
5122 		rc = init_crq_queue(adapter);
5123 		if (rc) {
5124 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5125 				rc);
5126 			goto ibmvnic_init_fail;
5127 		}
5128 
5129 		rc = ibmvnic_reset_init(adapter, false);
5130 		if (rc && rc != EAGAIN)
5131 			goto ibmvnic_init_fail;
5132 	} while (rc == EAGAIN);
5133 
5134 	rc = init_stats_buffers(adapter);
5135 	if (rc)
5136 		goto ibmvnic_init_fail;
5137 
5138 	rc = init_stats_token(adapter);
5139 	if (rc)
5140 		goto ibmvnic_stats_fail;
5141 
5142 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5143 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5144 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5145 
5146 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5147 	if (rc)
5148 		goto ibmvnic_dev_file_err;
5149 
5150 	netif_carrier_off(netdev);
5151 	rc = register_netdev(netdev);
5152 	if (rc) {
5153 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5154 		goto ibmvnic_register_fail;
5155 	}
5156 	dev_info(&dev->dev, "ibmvnic registered\n");
5157 
5158 	adapter->state = VNIC_PROBED;
5159 
5160 	adapter->wait_for_reset = false;
5161 
5162 	return 0;
5163 
5164 ibmvnic_register_fail:
5165 	device_remove_file(&dev->dev, &dev_attr_failover);
5166 
5167 ibmvnic_dev_file_err:
5168 	release_stats_token(adapter);
5169 
5170 ibmvnic_stats_fail:
5171 	release_stats_buffers(adapter);
5172 
5173 ibmvnic_init_fail:
5174 	release_sub_crqs(adapter, 1);
5175 	release_crq_queue(adapter);
5176 	mutex_destroy(&adapter->fw_lock);
5177 	free_netdev(netdev);
5178 
5179 	return rc;
5180 }
5181 
5182 static int ibmvnic_remove(struct vio_dev *dev)
5183 {
5184 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5185 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5186 	unsigned long flags;
5187 
5188 	spin_lock_irqsave(&adapter->state_lock, flags);
5189 	if (adapter->state == VNIC_RESETTING) {
5190 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5191 		return -EBUSY;
5192 	}
5193 
5194 	adapter->state = VNIC_REMOVING;
5195 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5196 
5197 	flush_work(&adapter->ibmvnic_reset);
5198 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5199 
5200 	rtnl_lock();
5201 	unregister_netdevice(netdev);
5202 
5203 	release_resources(adapter);
5204 	release_sub_crqs(adapter, 1);
5205 	release_crq_queue(adapter);
5206 
5207 	release_stats_token(adapter);
5208 	release_stats_buffers(adapter);
5209 
5210 	adapter->state = VNIC_REMOVED;
5211 
5212 	rtnl_unlock();
5213 	mutex_destroy(&adapter->fw_lock);
5214 	device_remove_file(&dev->dev, &dev_attr_failover);
5215 	free_netdev(netdev);
5216 	dev_set_drvdata(&dev->dev, NULL);
5217 
5218 	return 0;
5219 }
5220 
5221 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5222 			      const char *buf, size_t count)
5223 {
5224 	struct net_device *netdev = dev_get_drvdata(dev);
5225 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5226 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5227 	__be64 session_token;
5228 	long rc;
5229 
5230 	if (!sysfs_streq(buf, "1"))
5231 		return -EINVAL;
5232 
5233 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5234 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5235 	if (rc) {
5236 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5237 			   rc);
5238 		return -EINVAL;
5239 	}
5240 
5241 	session_token = (__be64)retbuf[0];
5242 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5243 		   be64_to_cpu(session_token));
5244 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5245 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5246 	if (rc) {
5247 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5248 			   rc);
5249 		return -EINVAL;
5250 	}
5251 
5252 	return count;
5253 }
5254 
5255 static DEVICE_ATTR_WO(failover);
5256 
5257 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5258 {
5259 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5260 	struct ibmvnic_adapter *adapter;
5261 	struct iommu_table *tbl;
5262 	unsigned long ret = 0;
5263 	int i;
5264 
5265 	tbl = get_iommu_table_base(&vdev->dev);
5266 
5267 	/* netdev inits at probe time along with the structures we need below*/
5268 	if (!netdev)
5269 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5270 
5271 	adapter = netdev_priv(netdev);
5272 
5273 	ret += PAGE_SIZE; /* the crq message queue */
5274 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5275 
5276 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5277 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5278 
5279 	for (i = 0; i < adapter->num_active_rx_pools; i++)
5280 		ret += adapter->rx_pool[i].size *
5281 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5282 
5283 	return ret;
5284 }
5285 
5286 static int ibmvnic_resume(struct device *dev)
5287 {
5288 	struct net_device *netdev = dev_get_drvdata(dev);
5289 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5290 
5291 	if (adapter->state != VNIC_OPEN)
5292 		return 0;
5293 
5294 	tasklet_schedule(&adapter->tasklet);
5295 
5296 	return 0;
5297 }
5298 
5299 static const struct vio_device_id ibmvnic_device_table[] = {
5300 	{"network", "IBM,vnic"},
5301 	{"", "" }
5302 };
5303 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5304 
5305 static const struct dev_pm_ops ibmvnic_pm_ops = {
5306 	.resume = ibmvnic_resume
5307 };
5308 
5309 static struct vio_driver ibmvnic_driver = {
5310 	.id_table       = ibmvnic_device_table,
5311 	.probe          = ibmvnic_probe,
5312 	.remove         = ibmvnic_remove,
5313 	.get_desired_dma = ibmvnic_get_desired_dma,
5314 	.name		= ibmvnic_driver_name,
5315 	.pm		= &ibmvnic_pm_ops,
5316 };
5317 
5318 /* module functions */
5319 static int __init ibmvnic_module_init(void)
5320 {
5321 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5322 		IBMVNIC_DRIVER_VERSION);
5323 
5324 	return vio_register_driver(&ibmvnic_driver);
5325 }
5326 
5327 static void __exit ibmvnic_module_exit(void)
5328 {
5329 	vio_unregister_driver(&ibmvnic_driver);
5330 }
5331 
5332 module_init(ibmvnic_module_init);
5333 module_exit(ibmvnic_module_exit);
5334