xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 911b8eac)
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_query_map(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_query_cap(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 	if (!adapter->rx_pool)
475 		return -1;
476 
477 	buff_size = adapter->cur_rx_buf_sz;
478 	rx_scrqs = adapter->num_active_rx_pools;
479 	for (i = 0; i < rx_scrqs; i++) {
480 		rx_pool = &adapter->rx_pool[i];
481 
482 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
483 
484 		if (rx_pool->buff_size != buff_size) {
485 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
486 			rx_pool->buff_size = buff_size;
487 			rc = alloc_long_term_buff(adapter,
488 						  &rx_pool->long_term_buff,
489 						  rx_pool->size *
490 						  rx_pool->buff_size);
491 		} else {
492 			rc = reset_long_term_buff(adapter,
493 						  &rx_pool->long_term_buff);
494 		}
495 
496 		if (rc)
497 			return rc;
498 
499 		for (j = 0; j < rx_pool->size; j++)
500 			rx_pool->free_map[j] = j;
501 
502 		memset(rx_pool->rx_buff, 0,
503 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
504 
505 		atomic_set(&rx_pool->available, 0);
506 		rx_pool->next_alloc = 0;
507 		rx_pool->next_free = 0;
508 		rx_pool->active = 1;
509 	}
510 
511 	return 0;
512 }
513 
514 static void release_rx_pools(struct ibmvnic_adapter *adapter)
515 {
516 	struct ibmvnic_rx_pool *rx_pool;
517 	int i, j;
518 
519 	if (!adapter->rx_pool)
520 		return;
521 
522 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
523 		rx_pool = &adapter->rx_pool[i];
524 
525 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
526 
527 		kfree(rx_pool->free_map);
528 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
529 
530 		if (!rx_pool->rx_buff)
531 			continue;
532 
533 		for (j = 0; j < rx_pool->size; j++) {
534 			if (rx_pool->rx_buff[j].skb) {
535 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
536 				rx_pool->rx_buff[j].skb = NULL;
537 			}
538 		}
539 
540 		kfree(rx_pool->rx_buff);
541 	}
542 
543 	kfree(adapter->rx_pool);
544 	adapter->rx_pool = NULL;
545 	adapter->num_active_rx_pools = 0;
546 }
547 
548 static int init_rx_pools(struct net_device *netdev)
549 {
550 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
551 	struct device *dev = &adapter->vdev->dev;
552 	struct ibmvnic_rx_pool *rx_pool;
553 	int rxadd_subcrqs;
554 	u64 buff_size;
555 	int i, j;
556 
557 	rxadd_subcrqs = adapter->num_active_rx_scrqs;
558 	buff_size = adapter->cur_rx_buf_sz;
559 
560 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
561 				   sizeof(struct ibmvnic_rx_pool),
562 				   GFP_KERNEL);
563 	if (!adapter->rx_pool) {
564 		dev_err(dev, "Failed to allocate rx pools\n");
565 		return -1;
566 	}
567 
568 	adapter->num_active_rx_pools = rxadd_subcrqs;
569 
570 	for (i = 0; i < rxadd_subcrqs; i++) {
571 		rx_pool = &adapter->rx_pool[i];
572 
573 		netdev_dbg(adapter->netdev,
574 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
575 			   i, adapter->req_rx_add_entries_per_subcrq,
576 			   buff_size);
577 
578 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
579 		rx_pool->index = i;
580 		rx_pool->buff_size = buff_size;
581 		rx_pool->active = 1;
582 
583 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
584 					    GFP_KERNEL);
585 		if (!rx_pool->free_map) {
586 			release_rx_pools(adapter);
587 			return -1;
588 		}
589 
590 		rx_pool->rx_buff = kcalloc(rx_pool->size,
591 					   sizeof(struct ibmvnic_rx_buff),
592 					   GFP_KERNEL);
593 		if (!rx_pool->rx_buff) {
594 			dev_err(dev, "Couldn't alloc rx buffers\n");
595 			release_rx_pools(adapter);
596 			return -1;
597 		}
598 
599 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
600 					 rx_pool->size * rx_pool->buff_size)) {
601 			release_rx_pools(adapter);
602 			return -1;
603 		}
604 
605 		for (j = 0; j < rx_pool->size; ++j)
606 			rx_pool->free_map[j] = j;
607 
608 		atomic_set(&rx_pool->available, 0);
609 		rx_pool->next_alloc = 0;
610 		rx_pool->next_free = 0;
611 	}
612 
613 	return 0;
614 }
615 
616 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
617 			     struct ibmvnic_tx_pool *tx_pool)
618 {
619 	int rc, i;
620 
621 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
622 	if (rc)
623 		return rc;
624 
625 	memset(tx_pool->tx_buff, 0,
626 	       tx_pool->num_buffers *
627 	       sizeof(struct ibmvnic_tx_buff));
628 
629 	for (i = 0; i < tx_pool->num_buffers; i++)
630 		tx_pool->free_map[i] = i;
631 
632 	tx_pool->consumer_index = 0;
633 	tx_pool->producer_index = 0;
634 
635 	return 0;
636 }
637 
638 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
639 {
640 	int tx_scrqs;
641 	int i, rc;
642 
643 	if (!adapter->tx_pool)
644 		return -1;
645 
646 	tx_scrqs = adapter->num_active_tx_pools;
647 	for (i = 0; i < tx_scrqs; i++) {
648 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
649 		if (rc)
650 			return rc;
651 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
652 		if (rc)
653 			return rc;
654 	}
655 
656 	return 0;
657 }
658 
659 static void release_vpd_data(struct ibmvnic_adapter *adapter)
660 {
661 	if (!adapter->vpd)
662 		return;
663 
664 	kfree(adapter->vpd->buff);
665 	kfree(adapter->vpd);
666 
667 	adapter->vpd = NULL;
668 }
669 
670 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
671 				struct ibmvnic_tx_pool *tx_pool)
672 {
673 	kfree(tx_pool->tx_buff);
674 	kfree(tx_pool->free_map);
675 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
676 }
677 
678 static void release_tx_pools(struct ibmvnic_adapter *adapter)
679 {
680 	int i;
681 
682 	if (!adapter->tx_pool)
683 		return;
684 
685 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
686 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
687 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
688 	}
689 
690 	kfree(adapter->tx_pool);
691 	adapter->tx_pool = NULL;
692 	kfree(adapter->tso_pool);
693 	adapter->tso_pool = NULL;
694 	adapter->num_active_tx_pools = 0;
695 }
696 
697 static int init_one_tx_pool(struct net_device *netdev,
698 			    struct ibmvnic_tx_pool *tx_pool,
699 			    int num_entries, int buf_size)
700 {
701 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
702 	int i;
703 
704 	tx_pool->tx_buff = kcalloc(num_entries,
705 				   sizeof(struct ibmvnic_tx_buff),
706 				   GFP_KERNEL);
707 	if (!tx_pool->tx_buff)
708 		return -1;
709 
710 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
711 				 num_entries * buf_size))
712 		return -1;
713 
714 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
715 	if (!tx_pool->free_map)
716 		return -1;
717 
718 	for (i = 0; i < num_entries; i++)
719 		tx_pool->free_map[i] = i;
720 
721 	tx_pool->consumer_index = 0;
722 	tx_pool->producer_index = 0;
723 	tx_pool->num_buffers = num_entries;
724 	tx_pool->buf_size = buf_size;
725 
726 	return 0;
727 }
728 
729 static int init_tx_pools(struct net_device *netdev)
730 {
731 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
732 	int tx_subcrqs;
733 	int i, rc;
734 
735 	tx_subcrqs = adapter->num_active_tx_scrqs;
736 	adapter->tx_pool = kcalloc(tx_subcrqs,
737 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
738 	if (!adapter->tx_pool)
739 		return -1;
740 
741 	adapter->tso_pool = kcalloc(tx_subcrqs,
742 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
743 	if (!adapter->tso_pool)
744 		return -1;
745 
746 	adapter->num_active_tx_pools = tx_subcrqs;
747 
748 	for (i = 0; i < tx_subcrqs; i++) {
749 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
750 				      adapter->req_tx_entries_per_subcrq,
751 				      adapter->req_mtu + VLAN_HLEN);
752 		if (rc) {
753 			release_tx_pools(adapter);
754 			return rc;
755 		}
756 
757 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
758 				      IBMVNIC_TSO_BUFS,
759 				      IBMVNIC_TSO_BUF_SZ);
760 		if (rc) {
761 			release_tx_pools(adapter);
762 			return rc;
763 		}
764 	}
765 
766 	return 0;
767 }
768 
769 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
770 {
771 	int i;
772 
773 	if (adapter->napi_enabled)
774 		return;
775 
776 	for (i = 0; i < adapter->req_rx_queues; i++)
777 		napi_enable(&adapter->napi[i]);
778 
779 	adapter->napi_enabled = true;
780 }
781 
782 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
783 {
784 	int i;
785 
786 	if (!adapter->napi_enabled)
787 		return;
788 
789 	for (i = 0; i < adapter->req_rx_queues; i++) {
790 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
791 		napi_disable(&adapter->napi[i]);
792 	}
793 
794 	adapter->napi_enabled = false;
795 }
796 
797 static int init_napi(struct ibmvnic_adapter *adapter)
798 {
799 	int i;
800 
801 	adapter->napi = kcalloc(adapter->req_rx_queues,
802 				sizeof(struct napi_struct), GFP_KERNEL);
803 	if (!adapter->napi)
804 		return -ENOMEM;
805 
806 	for (i = 0; i < adapter->req_rx_queues; i++) {
807 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
808 		netif_napi_add(adapter->netdev, &adapter->napi[i],
809 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
810 	}
811 
812 	adapter->num_active_rx_napi = adapter->req_rx_queues;
813 	return 0;
814 }
815 
816 static void release_napi(struct ibmvnic_adapter *adapter)
817 {
818 	int i;
819 
820 	if (!adapter->napi)
821 		return;
822 
823 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
824 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
825 		netif_napi_del(&adapter->napi[i]);
826 	}
827 
828 	kfree(adapter->napi);
829 	adapter->napi = NULL;
830 	adapter->num_active_rx_napi = 0;
831 	adapter->napi_enabled = false;
832 }
833 
834 static int ibmvnic_login(struct net_device *netdev)
835 {
836 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
837 	unsigned long timeout = msecs_to_jiffies(30000);
838 	int retry_count = 0;
839 	int retries = 10;
840 	bool retry;
841 	int rc;
842 
843 	do {
844 		retry = false;
845 		if (retry_count > retries) {
846 			netdev_warn(netdev, "Login attempts exceeded\n");
847 			return -1;
848 		}
849 
850 		adapter->init_done_rc = 0;
851 		reinit_completion(&adapter->init_done);
852 		rc = send_login(adapter);
853 		if (rc) {
854 			netdev_warn(netdev, "Unable to login\n");
855 			return rc;
856 		}
857 
858 		if (!wait_for_completion_timeout(&adapter->init_done,
859 						 timeout)) {
860 			netdev_warn(netdev, "Login timed out, retrying...\n");
861 			retry = true;
862 			adapter->init_done_rc = 0;
863 			retry_count++;
864 			continue;
865 		}
866 
867 		if (adapter->init_done_rc == ABORTED) {
868 			netdev_warn(netdev, "Login aborted, retrying...\n");
869 			retry = true;
870 			adapter->init_done_rc = 0;
871 			retry_count++;
872 			/* FW or device may be busy, so
873 			 * wait a bit before retrying login
874 			 */
875 			msleep(500);
876 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
877 			retry_count++;
878 			release_sub_crqs(adapter, 1);
879 
880 			retry = true;
881 			netdev_dbg(netdev,
882 				   "Received partial success, retrying...\n");
883 			adapter->init_done_rc = 0;
884 			reinit_completion(&adapter->init_done);
885 			send_query_cap(adapter);
886 			if (!wait_for_completion_timeout(&adapter->init_done,
887 							 timeout)) {
888 				netdev_warn(netdev,
889 					    "Capabilities query timed out\n");
890 				return -1;
891 			}
892 
893 			rc = init_sub_crqs(adapter);
894 			if (rc) {
895 				netdev_warn(netdev,
896 					    "SCRQ initialization failed\n");
897 				return -1;
898 			}
899 
900 			rc = init_sub_crq_irqs(adapter);
901 			if (rc) {
902 				netdev_warn(netdev,
903 					    "SCRQ irq initialization failed\n");
904 				return -1;
905 			}
906 		} else if (adapter->init_done_rc) {
907 			netdev_warn(netdev, "Adapter login failed\n");
908 			return -1;
909 		}
910 	} while (retry);
911 
912 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
913 
914 	return 0;
915 }
916 
917 static void release_login_buffer(struct ibmvnic_adapter *adapter)
918 {
919 	kfree(adapter->login_buf);
920 	adapter->login_buf = NULL;
921 }
922 
923 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
924 {
925 	kfree(adapter->login_rsp_buf);
926 	adapter->login_rsp_buf = NULL;
927 }
928 
929 static void release_resources(struct ibmvnic_adapter *adapter)
930 {
931 	release_vpd_data(adapter);
932 
933 	release_tx_pools(adapter);
934 	release_rx_pools(adapter);
935 
936 	release_napi(adapter);
937 	release_login_rsp_buffer(adapter);
938 }
939 
940 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
941 {
942 	struct net_device *netdev = adapter->netdev;
943 	unsigned long timeout = msecs_to_jiffies(30000);
944 	union ibmvnic_crq crq;
945 	bool resend;
946 	int rc;
947 
948 	netdev_dbg(netdev, "setting link state %d\n", link_state);
949 
950 	memset(&crq, 0, sizeof(crq));
951 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
952 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
953 	crq.logical_link_state.link_state = link_state;
954 
955 	do {
956 		resend = false;
957 
958 		reinit_completion(&adapter->init_done);
959 		rc = ibmvnic_send_crq(adapter, &crq);
960 		if (rc) {
961 			netdev_err(netdev, "Failed to set link state\n");
962 			return rc;
963 		}
964 
965 		if (!wait_for_completion_timeout(&adapter->init_done,
966 						 timeout)) {
967 			netdev_err(netdev, "timeout setting link state\n");
968 			return -1;
969 		}
970 
971 		if (adapter->init_done_rc == PARTIALSUCCESS) {
972 			/* Partuial success, delay and re-send */
973 			mdelay(1000);
974 			resend = true;
975 		} else if (adapter->init_done_rc) {
976 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
977 				    adapter->init_done_rc);
978 			return adapter->init_done_rc;
979 		}
980 	} while (resend);
981 
982 	return 0;
983 }
984 
985 static int set_real_num_queues(struct net_device *netdev)
986 {
987 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
988 	int rc;
989 
990 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
991 		   adapter->req_tx_queues, adapter->req_rx_queues);
992 
993 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
994 	if (rc) {
995 		netdev_err(netdev, "failed to set the number of tx queues\n");
996 		return rc;
997 	}
998 
999 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1000 	if (rc)
1001 		netdev_err(netdev, "failed to set the number of rx queues\n");
1002 
1003 	return rc;
1004 }
1005 
1006 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1007 {
1008 	struct device *dev = &adapter->vdev->dev;
1009 	union ibmvnic_crq crq;
1010 	int len = 0;
1011 	int rc;
1012 
1013 	if (adapter->vpd->buff)
1014 		len = adapter->vpd->len;
1015 
1016 	mutex_lock(&adapter->fw_lock);
1017 	adapter->fw_done_rc = 0;
1018 	reinit_completion(&adapter->fw_done);
1019 
1020 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1021 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1022 	rc = ibmvnic_send_crq(adapter, &crq);
1023 	if (rc) {
1024 		mutex_unlock(&adapter->fw_lock);
1025 		return rc;
1026 	}
1027 
1028 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1029 	if (rc) {
1030 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1031 		mutex_unlock(&adapter->fw_lock);
1032 		return rc;
1033 	}
1034 	mutex_unlock(&adapter->fw_lock);
1035 
1036 	if (!adapter->vpd->len)
1037 		return -ENODATA;
1038 
1039 	if (!adapter->vpd->buff)
1040 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1041 	else if (adapter->vpd->len != len)
1042 		adapter->vpd->buff =
1043 			krealloc(adapter->vpd->buff,
1044 				 adapter->vpd->len, GFP_KERNEL);
1045 
1046 	if (!adapter->vpd->buff) {
1047 		dev_err(dev, "Could allocate VPD buffer\n");
1048 		return -ENOMEM;
1049 	}
1050 
1051 	adapter->vpd->dma_addr =
1052 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1053 			       DMA_FROM_DEVICE);
1054 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1055 		dev_err(dev, "Could not map VPD buffer\n");
1056 		kfree(adapter->vpd->buff);
1057 		adapter->vpd->buff = NULL;
1058 		return -ENOMEM;
1059 	}
1060 
1061 	mutex_lock(&adapter->fw_lock);
1062 	adapter->fw_done_rc = 0;
1063 	reinit_completion(&adapter->fw_done);
1064 
1065 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1066 	crq.get_vpd.cmd = GET_VPD;
1067 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1068 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1069 	rc = ibmvnic_send_crq(adapter, &crq);
1070 	if (rc) {
1071 		kfree(adapter->vpd->buff);
1072 		adapter->vpd->buff = NULL;
1073 		mutex_unlock(&adapter->fw_lock);
1074 		return rc;
1075 	}
1076 
1077 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1078 	if (rc) {
1079 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1080 		kfree(adapter->vpd->buff);
1081 		adapter->vpd->buff = NULL;
1082 		mutex_unlock(&adapter->fw_lock);
1083 		return rc;
1084 	}
1085 
1086 	mutex_unlock(&adapter->fw_lock);
1087 	return 0;
1088 }
1089 
1090 static int init_resources(struct ibmvnic_adapter *adapter)
1091 {
1092 	struct net_device *netdev = adapter->netdev;
1093 	int rc;
1094 
1095 	rc = set_real_num_queues(netdev);
1096 	if (rc)
1097 		return rc;
1098 
1099 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1100 	if (!adapter->vpd)
1101 		return -ENOMEM;
1102 
1103 	/* Vital Product Data (VPD) */
1104 	rc = ibmvnic_get_vpd(adapter);
1105 	if (rc) {
1106 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1107 		return rc;
1108 	}
1109 
1110 	adapter->map_id = 1;
1111 
1112 	rc = init_napi(adapter);
1113 	if (rc)
1114 		return rc;
1115 
1116 	send_query_map(adapter);
1117 
1118 	rc = init_rx_pools(netdev);
1119 	if (rc)
1120 		return rc;
1121 
1122 	rc = init_tx_pools(netdev);
1123 	return rc;
1124 }
1125 
1126 static int __ibmvnic_open(struct net_device *netdev)
1127 {
1128 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1129 	enum vnic_state prev_state = adapter->state;
1130 	int i, rc;
1131 
1132 	adapter->state = VNIC_OPENING;
1133 	replenish_pools(adapter);
1134 	ibmvnic_napi_enable(adapter);
1135 
1136 	/* We're ready to receive frames, enable the sub-crq interrupts and
1137 	 * set the logical link state to up
1138 	 */
1139 	for (i = 0; i < adapter->req_rx_queues; i++) {
1140 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1141 		if (prev_state == VNIC_CLOSED)
1142 			enable_irq(adapter->rx_scrq[i]->irq);
1143 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1144 	}
1145 
1146 	for (i = 0; i < adapter->req_tx_queues; i++) {
1147 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1148 		if (prev_state == VNIC_CLOSED)
1149 			enable_irq(adapter->tx_scrq[i]->irq);
1150 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1151 	}
1152 
1153 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1154 	if (rc) {
1155 		for (i = 0; i < adapter->req_rx_queues; i++)
1156 			napi_disable(&adapter->napi[i]);
1157 		release_resources(adapter);
1158 		return rc;
1159 	}
1160 
1161 	netif_tx_start_all_queues(netdev);
1162 
1163 	if (prev_state == VNIC_CLOSED) {
1164 		for (i = 0; i < adapter->req_rx_queues; i++)
1165 			napi_schedule(&adapter->napi[i]);
1166 	}
1167 
1168 	adapter->state = VNIC_OPEN;
1169 	return rc;
1170 }
1171 
1172 static int ibmvnic_open(struct net_device *netdev)
1173 {
1174 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1175 	int rc;
1176 
1177 	/* If device failover is pending, just set device state and return.
1178 	 * Device operation will be handled by reset routine.
1179 	 */
1180 	if (adapter->failover_pending) {
1181 		adapter->state = VNIC_OPEN;
1182 		return 0;
1183 	}
1184 
1185 	if (adapter->state != VNIC_CLOSED) {
1186 		rc = ibmvnic_login(netdev);
1187 		if (rc)
1188 			return rc;
1189 
1190 		rc = init_resources(adapter);
1191 		if (rc) {
1192 			netdev_err(netdev, "failed to initialize resources\n");
1193 			release_resources(adapter);
1194 			return rc;
1195 		}
1196 	}
1197 
1198 	rc = __ibmvnic_open(netdev);
1199 
1200 	return rc;
1201 }
1202 
1203 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1204 {
1205 	struct ibmvnic_rx_pool *rx_pool;
1206 	struct ibmvnic_rx_buff *rx_buff;
1207 	u64 rx_entries;
1208 	int rx_scrqs;
1209 	int i, j;
1210 
1211 	if (!adapter->rx_pool)
1212 		return;
1213 
1214 	rx_scrqs = adapter->num_active_rx_pools;
1215 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1216 
1217 	/* Free any remaining skbs in the rx buffer pools */
1218 	for (i = 0; i < rx_scrqs; i++) {
1219 		rx_pool = &adapter->rx_pool[i];
1220 		if (!rx_pool || !rx_pool->rx_buff)
1221 			continue;
1222 
1223 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1224 		for (j = 0; j < rx_entries; j++) {
1225 			rx_buff = &rx_pool->rx_buff[j];
1226 			if (rx_buff && rx_buff->skb) {
1227 				dev_kfree_skb_any(rx_buff->skb);
1228 				rx_buff->skb = NULL;
1229 			}
1230 		}
1231 	}
1232 }
1233 
1234 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1235 			      struct ibmvnic_tx_pool *tx_pool)
1236 {
1237 	struct ibmvnic_tx_buff *tx_buff;
1238 	u64 tx_entries;
1239 	int i;
1240 
1241 	if (!tx_pool || !tx_pool->tx_buff)
1242 		return;
1243 
1244 	tx_entries = tx_pool->num_buffers;
1245 
1246 	for (i = 0; i < tx_entries; i++) {
1247 		tx_buff = &tx_pool->tx_buff[i];
1248 		if (tx_buff && tx_buff->skb) {
1249 			dev_kfree_skb_any(tx_buff->skb);
1250 			tx_buff->skb = NULL;
1251 		}
1252 	}
1253 }
1254 
1255 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1256 {
1257 	int tx_scrqs;
1258 	int i;
1259 
1260 	if (!adapter->tx_pool || !adapter->tso_pool)
1261 		return;
1262 
1263 	tx_scrqs = adapter->num_active_tx_pools;
1264 
1265 	/* Free any remaining skbs in the tx buffer pools */
1266 	for (i = 0; i < tx_scrqs; i++) {
1267 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1268 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1269 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1270 	}
1271 }
1272 
1273 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1274 {
1275 	struct net_device *netdev = adapter->netdev;
1276 	int i;
1277 
1278 	if (adapter->tx_scrq) {
1279 		for (i = 0; i < adapter->req_tx_queues; i++)
1280 			if (adapter->tx_scrq[i]->irq) {
1281 				netdev_dbg(netdev,
1282 					   "Disabling tx_scrq[%d] irq\n", i);
1283 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1284 				disable_irq(adapter->tx_scrq[i]->irq);
1285 			}
1286 	}
1287 
1288 	if (adapter->rx_scrq) {
1289 		for (i = 0; i < adapter->req_rx_queues; i++) {
1290 			if (adapter->rx_scrq[i]->irq) {
1291 				netdev_dbg(netdev,
1292 					   "Disabling rx_scrq[%d] irq\n", i);
1293 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1294 				disable_irq(adapter->rx_scrq[i]->irq);
1295 			}
1296 		}
1297 	}
1298 }
1299 
1300 static void ibmvnic_cleanup(struct net_device *netdev)
1301 {
1302 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1303 
1304 	/* ensure that transmissions are stopped if called by do_reset */
1305 	if (test_bit(0, &adapter->resetting))
1306 		netif_tx_disable(netdev);
1307 	else
1308 		netif_tx_stop_all_queues(netdev);
1309 
1310 	ibmvnic_napi_disable(adapter);
1311 	ibmvnic_disable_irqs(adapter);
1312 
1313 	clean_rx_pools(adapter);
1314 	clean_tx_pools(adapter);
1315 }
1316 
1317 static int __ibmvnic_close(struct net_device *netdev)
1318 {
1319 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1320 	int rc = 0;
1321 
1322 	adapter->state = VNIC_CLOSING;
1323 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1324 	if (rc)
1325 		return rc;
1326 	adapter->state = VNIC_CLOSED;
1327 	return 0;
1328 }
1329 
1330 static int ibmvnic_close(struct net_device *netdev)
1331 {
1332 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1333 	int rc;
1334 
1335 	/* If device failover is pending, just set device state and return.
1336 	 * Device operation will be handled by reset routine.
1337 	 */
1338 	if (adapter->failover_pending) {
1339 		adapter->state = VNIC_CLOSED;
1340 		return 0;
1341 	}
1342 
1343 	rc = __ibmvnic_close(netdev);
1344 	ibmvnic_cleanup(netdev);
1345 
1346 	return rc;
1347 }
1348 
1349 /**
1350  * build_hdr_data - creates L2/L3/L4 header data buffer
1351  * @hdr_field - bitfield determining needed headers
1352  * @skb - socket buffer
1353  * @hdr_len - array of header lengths
1354  * @tot_len - total length of data
1355  *
1356  * Reads hdr_field to determine which headers are needed by firmware.
1357  * Builds a buffer containing these headers.  Saves individual header
1358  * lengths and total buffer length to be used to build descriptors.
1359  */
1360 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1361 			  int *hdr_len, u8 *hdr_data)
1362 {
1363 	int len = 0;
1364 	u8 *hdr;
1365 
1366 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1367 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1368 	else
1369 		hdr_len[0] = sizeof(struct ethhdr);
1370 
1371 	if (skb->protocol == htons(ETH_P_IP)) {
1372 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1373 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1374 			hdr_len[2] = tcp_hdrlen(skb);
1375 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1376 			hdr_len[2] = sizeof(struct udphdr);
1377 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1378 		hdr_len[1] = sizeof(struct ipv6hdr);
1379 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1380 			hdr_len[2] = tcp_hdrlen(skb);
1381 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1382 			hdr_len[2] = sizeof(struct udphdr);
1383 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1384 		hdr_len[1] = arp_hdr_len(skb->dev);
1385 		hdr_len[2] = 0;
1386 	}
1387 
1388 	memset(hdr_data, 0, 120);
1389 	if ((hdr_field >> 6) & 1) {
1390 		hdr = skb_mac_header(skb);
1391 		memcpy(hdr_data, hdr, hdr_len[0]);
1392 		len += hdr_len[0];
1393 	}
1394 
1395 	if ((hdr_field >> 5) & 1) {
1396 		hdr = skb_network_header(skb);
1397 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1398 		len += hdr_len[1];
1399 	}
1400 
1401 	if ((hdr_field >> 4) & 1) {
1402 		hdr = skb_transport_header(skb);
1403 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1404 		len += hdr_len[2];
1405 	}
1406 	return len;
1407 }
1408 
1409 /**
1410  * create_hdr_descs - create header and header extension descriptors
1411  * @hdr_field - bitfield determining needed headers
1412  * @data - buffer containing header data
1413  * @len - length of data buffer
1414  * @hdr_len - array of individual header lengths
1415  * @scrq_arr - descriptor array
1416  *
1417  * Creates header and, if needed, header extension descriptors and
1418  * places them in a descriptor array, scrq_arr
1419  */
1420 
1421 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1422 			    union sub_crq *scrq_arr)
1423 {
1424 	union sub_crq hdr_desc;
1425 	int tmp_len = len;
1426 	int num_descs = 0;
1427 	u8 *data, *cur;
1428 	int tmp;
1429 
1430 	while (tmp_len > 0) {
1431 		cur = hdr_data + len - tmp_len;
1432 
1433 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1434 		if (cur != hdr_data) {
1435 			data = hdr_desc.hdr_ext.data;
1436 			tmp = tmp_len > 29 ? 29 : tmp_len;
1437 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1438 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1439 			hdr_desc.hdr_ext.len = tmp;
1440 		} else {
1441 			data = hdr_desc.hdr.data;
1442 			tmp = tmp_len > 24 ? 24 : tmp_len;
1443 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1444 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1445 			hdr_desc.hdr.len = tmp;
1446 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1447 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1448 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1449 			hdr_desc.hdr.flag = hdr_field << 1;
1450 		}
1451 		memcpy(data, cur, tmp);
1452 		tmp_len -= tmp;
1453 		*scrq_arr = hdr_desc;
1454 		scrq_arr++;
1455 		num_descs++;
1456 	}
1457 
1458 	return num_descs;
1459 }
1460 
1461 /**
1462  * build_hdr_descs_arr - build a header descriptor array
1463  * @skb - socket buffer
1464  * @num_entries - number of descriptors to be sent
1465  * @subcrq - first TX descriptor
1466  * @hdr_field - bit field determining which headers will be sent
1467  *
1468  * This function will build a TX descriptor array with applicable
1469  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1470  */
1471 
1472 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1473 				int *num_entries, u8 hdr_field)
1474 {
1475 	int hdr_len[3] = {0, 0, 0};
1476 	int tot_len;
1477 	u8 *hdr_data = txbuff->hdr_data;
1478 
1479 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1480 				 txbuff->hdr_data);
1481 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1482 			 txbuff->indir_arr + 1);
1483 }
1484 
1485 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1486 				    struct net_device *netdev)
1487 {
1488 	/* For some backing devices, mishandling of small packets
1489 	 * can result in a loss of connection or TX stall. Device
1490 	 * architects recommend that no packet should be smaller
1491 	 * than the minimum MTU value provided to the driver, so
1492 	 * pad any packets to that length
1493 	 */
1494 	if (skb->len < netdev->min_mtu)
1495 		return skb_put_padto(skb, netdev->min_mtu);
1496 
1497 	return 0;
1498 }
1499 
1500 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1501 {
1502 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1503 	int queue_num = skb_get_queue_mapping(skb);
1504 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1505 	struct device *dev = &adapter->vdev->dev;
1506 	struct ibmvnic_tx_buff *tx_buff = NULL;
1507 	struct ibmvnic_sub_crq_queue *tx_scrq;
1508 	struct ibmvnic_tx_pool *tx_pool;
1509 	unsigned int tx_send_failed = 0;
1510 	unsigned int tx_map_failed = 0;
1511 	unsigned int tx_dropped = 0;
1512 	unsigned int tx_packets = 0;
1513 	unsigned int tx_bytes = 0;
1514 	dma_addr_t data_dma_addr;
1515 	struct netdev_queue *txq;
1516 	unsigned long lpar_rc;
1517 	union sub_crq tx_crq;
1518 	unsigned int offset;
1519 	int num_entries = 1;
1520 	unsigned char *dst;
1521 	int index = 0;
1522 	u8 proto = 0;
1523 	u64 handle;
1524 	netdev_tx_t ret = NETDEV_TX_OK;
1525 
1526 	if (test_bit(0, &adapter->resetting)) {
1527 		if (!netif_subqueue_stopped(netdev, skb))
1528 			netif_stop_subqueue(netdev, queue_num);
1529 		dev_kfree_skb_any(skb);
1530 
1531 		tx_send_failed++;
1532 		tx_dropped++;
1533 		ret = NETDEV_TX_OK;
1534 		goto out;
1535 	}
1536 
1537 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1538 		tx_dropped++;
1539 		tx_send_failed++;
1540 		ret = NETDEV_TX_OK;
1541 		goto out;
1542 	}
1543 	if (skb_is_gso(skb))
1544 		tx_pool = &adapter->tso_pool[queue_num];
1545 	else
1546 		tx_pool = &adapter->tx_pool[queue_num];
1547 
1548 	tx_scrq = adapter->tx_scrq[queue_num];
1549 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1550 	handle = tx_scrq->handle;
1551 
1552 	index = tx_pool->free_map[tx_pool->consumer_index];
1553 
1554 	if (index == IBMVNIC_INVALID_MAP) {
1555 		dev_kfree_skb_any(skb);
1556 		tx_send_failed++;
1557 		tx_dropped++;
1558 		ret = NETDEV_TX_OK;
1559 		goto out;
1560 	}
1561 
1562 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1563 
1564 	offset = index * tx_pool->buf_size;
1565 	dst = tx_pool->long_term_buff.buff + offset;
1566 	memset(dst, 0, tx_pool->buf_size);
1567 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1568 
1569 	if (skb_shinfo(skb)->nr_frags) {
1570 		int cur, i;
1571 
1572 		/* Copy the head */
1573 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1574 		cur = skb_headlen(skb);
1575 
1576 		/* Copy the frags */
1577 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1578 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1579 
1580 			memcpy(dst + cur,
1581 			       page_address(skb_frag_page(frag)) +
1582 			       skb_frag_off(frag), skb_frag_size(frag));
1583 			cur += skb_frag_size(frag);
1584 		}
1585 	} else {
1586 		skb_copy_from_linear_data(skb, dst, skb->len);
1587 	}
1588 
1589 	tx_pool->consumer_index =
1590 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1591 
1592 	tx_buff = &tx_pool->tx_buff[index];
1593 	tx_buff->skb = skb;
1594 	tx_buff->data_dma[0] = data_dma_addr;
1595 	tx_buff->data_len[0] = skb->len;
1596 	tx_buff->index = index;
1597 	tx_buff->pool_index = queue_num;
1598 	tx_buff->last_frag = true;
1599 
1600 	memset(&tx_crq, 0, sizeof(tx_crq));
1601 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1602 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1603 	tx_crq.v1.n_crq_elem = 1;
1604 	tx_crq.v1.n_sge = 1;
1605 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1606 
1607 	if (skb_is_gso(skb))
1608 		tx_crq.v1.correlator =
1609 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1610 	else
1611 		tx_crq.v1.correlator = cpu_to_be32(index);
1612 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1613 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1614 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1615 
1616 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1617 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1618 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1619 	}
1620 
1621 	if (skb->protocol == htons(ETH_P_IP)) {
1622 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1623 		proto = ip_hdr(skb)->protocol;
1624 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1625 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1626 		proto = ipv6_hdr(skb)->nexthdr;
1627 	}
1628 
1629 	if (proto == IPPROTO_TCP)
1630 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1631 	else if (proto == IPPROTO_UDP)
1632 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1633 
1634 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1635 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1636 		hdrs += 2;
1637 	}
1638 	if (skb_is_gso(skb)) {
1639 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1640 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1641 		hdrs += 2;
1642 	}
1643 	/* determine if l2/3/4 headers are sent to firmware */
1644 	if ((*hdrs >> 7) & 1) {
1645 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1646 		tx_crq.v1.n_crq_elem = num_entries;
1647 		tx_buff->num_entries = num_entries;
1648 		tx_buff->indir_arr[0] = tx_crq;
1649 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1650 						    sizeof(tx_buff->indir_arr),
1651 						    DMA_TO_DEVICE);
1652 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1653 			dev_kfree_skb_any(skb);
1654 			tx_buff->skb = NULL;
1655 			if (!firmware_has_feature(FW_FEATURE_CMO))
1656 				dev_err(dev, "tx: unable to map descriptor array\n");
1657 			tx_map_failed++;
1658 			tx_dropped++;
1659 			ret = NETDEV_TX_OK;
1660 			goto tx_err_out;
1661 		}
1662 		lpar_rc = send_subcrq_indirect(adapter, handle,
1663 					       (u64)tx_buff->indir_dma,
1664 					       (u64)num_entries);
1665 		dma_unmap_single(dev, tx_buff->indir_dma,
1666 				 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1667 	} else {
1668 		tx_buff->num_entries = num_entries;
1669 		lpar_rc = send_subcrq(adapter, handle,
1670 				      &tx_crq);
1671 	}
1672 	if (lpar_rc != H_SUCCESS) {
1673 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1674 			dev_err_ratelimited(dev, "tx: send failed\n");
1675 		dev_kfree_skb_any(skb);
1676 		tx_buff->skb = NULL;
1677 
1678 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1679 			/* Disable TX and report carrier off if queue is closed
1680 			 * or pending failover.
1681 			 * Firmware guarantees that a signal will be sent to the
1682 			 * driver, triggering a reset or some other action.
1683 			 */
1684 			netif_tx_stop_all_queues(netdev);
1685 			netif_carrier_off(netdev);
1686 		}
1687 
1688 		tx_send_failed++;
1689 		tx_dropped++;
1690 		ret = NETDEV_TX_OK;
1691 		goto tx_err_out;
1692 	}
1693 
1694 	if (atomic_add_return(num_entries, &tx_scrq->used)
1695 					>= adapter->req_tx_entries_per_subcrq) {
1696 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1697 		netif_stop_subqueue(netdev, queue_num);
1698 	}
1699 
1700 	tx_packets++;
1701 	tx_bytes += skb->len;
1702 	txq->trans_start = jiffies;
1703 	ret = NETDEV_TX_OK;
1704 	goto out;
1705 
1706 tx_err_out:
1707 	/* roll back consumer index and map array*/
1708 	if (tx_pool->consumer_index == 0)
1709 		tx_pool->consumer_index =
1710 			tx_pool->num_buffers - 1;
1711 	else
1712 		tx_pool->consumer_index--;
1713 	tx_pool->free_map[tx_pool->consumer_index] = index;
1714 out:
1715 	netdev->stats.tx_dropped += tx_dropped;
1716 	netdev->stats.tx_bytes += tx_bytes;
1717 	netdev->stats.tx_packets += tx_packets;
1718 	adapter->tx_send_failed += tx_send_failed;
1719 	adapter->tx_map_failed += tx_map_failed;
1720 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1721 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1722 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1723 
1724 	return ret;
1725 }
1726 
1727 static void ibmvnic_set_multi(struct net_device *netdev)
1728 {
1729 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1730 	struct netdev_hw_addr *ha;
1731 	union ibmvnic_crq crq;
1732 
1733 	memset(&crq, 0, sizeof(crq));
1734 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1735 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1736 
1737 	if (netdev->flags & IFF_PROMISC) {
1738 		if (!adapter->promisc_supported)
1739 			return;
1740 	} else {
1741 		if (netdev->flags & IFF_ALLMULTI) {
1742 			/* Accept all multicast */
1743 			memset(&crq, 0, sizeof(crq));
1744 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1745 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1746 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1747 			ibmvnic_send_crq(adapter, &crq);
1748 		} else if (netdev_mc_empty(netdev)) {
1749 			/* Reject all multicast */
1750 			memset(&crq, 0, sizeof(crq));
1751 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1752 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1753 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1754 			ibmvnic_send_crq(adapter, &crq);
1755 		} else {
1756 			/* Accept one or more multicast(s) */
1757 			netdev_for_each_mc_addr(ha, netdev) {
1758 				memset(&crq, 0, sizeof(crq));
1759 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1760 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1761 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1762 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1763 						ha->addr);
1764 				ibmvnic_send_crq(adapter, &crq);
1765 			}
1766 		}
1767 	}
1768 }
1769 
1770 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1771 {
1772 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1773 	union ibmvnic_crq crq;
1774 	int rc;
1775 
1776 	if (!is_valid_ether_addr(dev_addr)) {
1777 		rc = -EADDRNOTAVAIL;
1778 		goto err;
1779 	}
1780 
1781 	memset(&crq, 0, sizeof(crq));
1782 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1783 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1784 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1785 
1786 	mutex_lock(&adapter->fw_lock);
1787 	adapter->fw_done_rc = 0;
1788 	reinit_completion(&adapter->fw_done);
1789 
1790 	rc = ibmvnic_send_crq(adapter, &crq);
1791 	if (rc) {
1792 		rc = -EIO;
1793 		mutex_unlock(&adapter->fw_lock);
1794 		goto err;
1795 	}
1796 
1797 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1798 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1799 	if (rc || adapter->fw_done_rc) {
1800 		rc = -EIO;
1801 		mutex_unlock(&adapter->fw_lock);
1802 		goto err;
1803 	}
1804 	mutex_unlock(&adapter->fw_lock);
1805 	return 0;
1806 err:
1807 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1808 	return rc;
1809 }
1810 
1811 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1812 {
1813 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1814 	struct sockaddr *addr = p;
1815 	int rc;
1816 
1817 	rc = 0;
1818 	ether_addr_copy(adapter->mac_addr, addr->sa_data);
1819 	if (adapter->state != VNIC_PROBED)
1820 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1821 
1822 	return rc;
1823 }
1824 
1825 /**
1826  * do_change_param_reset returns zero if we are able to keep processing reset
1827  * events, or non-zero if we hit a fatal error and must halt.
1828  */
1829 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1830 				 struct ibmvnic_rwi *rwi,
1831 				 u32 reset_state)
1832 {
1833 	struct net_device *netdev = adapter->netdev;
1834 	int i, rc;
1835 
1836 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1837 		   rwi->reset_reason);
1838 
1839 	netif_carrier_off(netdev);
1840 	adapter->reset_reason = rwi->reset_reason;
1841 
1842 	ibmvnic_cleanup(netdev);
1843 
1844 	if (reset_state == VNIC_OPEN) {
1845 		rc = __ibmvnic_close(netdev);
1846 		if (rc)
1847 			return rc;
1848 	}
1849 
1850 	release_resources(adapter);
1851 	release_sub_crqs(adapter, 1);
1852 	release_crq_queue(adapter);
1853 
1854 	adapter->state = VNIC_PROBED;
1855 
1856 	rc = init_crq_queue(adapter);
1857 
1858 	if (rc) {
1859 		netdev_err(adapter->netdev,
1860 			   "Couldn't initialize crq. rc=%d\n", rc);
1861 		return rc;
1862 	}
1863 
1864 	rc = ibmvnic_reset_init(adapter, true);
1865 	if (rc)
1866 		return IBMVNIC_INIT_FAILED;
1867 
1868 	/* If the adapter was in PROBE state prior to the reset,
1869 	 * exit here.
1870 	 */
1871 	if (reset_state == VNIC_PROBED)
1872 		return 0;
1873 
1874 	rc = ibmvnic_login(netdev);
1875 	if (rc) {
1876 		adapter->state = reset_state;
1877 		return rc;
1878 	}
1879 
1880 	rc = init_resources(adapter);
1881 	if (rc)
1882 		return rc;
1883 
1884 	ibmvnic_disable_irqs(adapter);
1885 
1886 	adapter->state = VNIC_CLOSED;
1887 
1888 	if (reset_state == VNIC_CLOSED)
1889 		return 0;
1890 
1891 	rc = __ibmvnic_open(netdev);
1892 	if (rc)
1893 		return IBMVNIC_OPEN_FAILED;
1894 
1895 	/* refresh device's multicast list */
1896 	ibmvnic_set_multi(netdev);
1897 
1898 	/* kick napi */
1899 	for (i = 0; i < adapter->req_rx_queues; i++)
1900 		napi_schedule(&adapter->napi[i]);
1901 
1902 	return 0;
1903 }
1904 
1905 /**
1906  * do_reset returns zero if we are able to keep processing reset events, or
1907  * non-zero if we hit a fatal error and must halt.
1908  */
1909 static int do_reset(struct ibmvnic_adapter *adapter,
1910 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1911 {
1912 	u64 old_num_rx_queues, old_num_tx_queues;
1913 	u64 old_num_rx_slots, old_num_tx_slots;
1914 	struct net_device *netdev = adapter->netdev;
1915 	int i, rc;
1916 
1917 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1918 		   rwi->reset_reason);
1919 
1920 	rtnl_lock();
1921 
1922 	netif_carrier_off(netdev);
1923 	adapter->reset_reason = rwi->reset_reason;
1924 
1925 	old_num_rx_queues = adapter->req_rx_queues;
1926 	old_num_tx_queues = adapter->req_tx_queues;
1927 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1928 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1929 
1930 	ibmvnic_cleanup(netdev);
1931 
1932 	if (reset_state == VNIC_OPEN &&
1933 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
1934 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1935 		adapter->state = VNIC_CLOSING;
1936 
1937 		/* Release the RTNL lock before link state change and
1938 		 * re-acquire after the link state change to allow
1939 		 * linkwatch_event to grab the RTNL lock and run during
1940 		 * a reset.
1941 		 */
1942 		rtnl_unlock();
1943 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1944 		rtnl_lock();
1945 		if (rc)
1946 			goto out;
1947 
1948 		if (adapter->state != VNIC_CLOSING) {
1949 			rc = -1;
1950 			goto out;
1951 		}
1952 
1953 		adapter->state = VNIC_CLOSED;
1954 	}
1955 
1956 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1957 		/* remove the closed state so when we call open it appears
1958 		 * we are coming from the probed state.
1959 		 */
1960 		adapter->state = VNIC_PROBED;
1961 
1962 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1963 			rc = ibmvnic_reenable_crq_queue(adapter);
1964 			release_sub_crqs(adapter, 1);
1965 		} else {
1966 			rc = ibmvnic_reset_crq(adapter);
1967 			if (rc == H_CLOSED || rc == H_SUCCESS) {
1968 				rc = vio_enable_interrupts(adapter->vdev);
1969 				if (rc)
1970 					netdev_err(adapter->netdev,
1971 						   "Reset failed to enable interrupts. rc=%d\n",
1972 						   rc);
1973 			}
1974 		}
1975 
1976 		if (rc) {
1977 			netdev_err(adapter->netdev,
1978 				   "Reset couldn't initialize crq. rc=%d\n", rc);
1979 			goto out;
1980 		}
1981 
1982 		rc = ibmvnic_reset_init(adapter, true);
1983 		if (rc) {
1984 			rc = IBMVNIC_INIT_FAILED;
1985 			goto out;
1986 		}
1987 
1988 		/* If the adapter was in PROBE state prior to the reset,
1989 		 * exit here.
1990 		 */
1991 		if (reset_state == VNIC_PROBED) {
1992 			rc = 0;
1993 			goto out;
1994 		}
1995 
1996 		rc = ibmvnic_login(netdev);
1997 		if (rc) {
1998 			adapter->state = reset_state;
1999 			goto out;
2000 		}
2001 
2002 		if (adapter->req_rx_queues != old_num_rx_queues ||
2003 		    adapter->req_tx_queues != old_num_tx_queues ||
2004 		    adapter->req_rx_add_entries_per_subcrq !=
2005 		    old_num_rx_slots ||
2006 		    adapter->req_tx_entries_per_subcrq !=
2007 		    old_num_tx_slots ||
2008 		    !adapter->rx_pool ||
2009 		    !adapter->tso_pool ||
2010 		    !adapter->tx_pool) {
2011 			release_rx_pools(adapter);
2012 			release_tx_pools(adapter);
2013 			release_napi(adapter);
2014 			release_vpd_data(adapter);
2015 
2016 			rc = init_resources(adapter);
2017 			if (rc)
2018 				goto out;
2019 
2020 		} else {
2021 			rc = reset_tx_pools(adapter);
2022 			if (rc) {
2023 				netdev_dbg(adapter->netdev, "reset tx pools failed (%d)\n",
2024 						rc);
2025 				goto out;
2026 			}
2027 
2028 			rc = reset_rx_pools(adapter);
2029 			if (rc) {
2030 				netdev_dbg(adapter->netdev, "reset rx pools failed (%d)\n",
2031 						rc);
2032 				goto out;
2033 			}
2034 		}
2035 		ibmvnic_disable_irqs(adapter);
2036 	}
2037 	adapter->state = VNIC_CLOSED;
2038 
2039 	if (reset_state == VNIC_CLOSED) {
2040 		rc = 0;
2041 		goto out;
2042 	}
2043 
2044 	rc = __ibmvnic_open(netdev);
2045 	if (rc) {
2046 		rc = IBMVNIC_OPEN_FAILED;
2047 		goto out;
2048 	}
2049 
2050 	/* refresh device's multicast list */
2051 	ibmvnic_set_multi(netdev);
2052 
2053 	/* kick napi */
2054 	for (i = 0; i < adapter->req_rx_queues; i++)
2055 		napi_schedule(&adapter->napi[i]);
2056 
2057 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2058 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2059 
2060 	rc = 0;
2061 
2062 out:
2063 	rtnl_unlock();
2064 
2065 	return rc;
2066 }
2067 
2068 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2069 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2070 {
2071 	struct net_device *netdev = adapter->netdev;
2072 	int rc;
2073 
2074 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2075 		   rwi->reset_reason);
2076 
2077 	netif_carrier_off(netdev);
2078 	adapter->reset_reason = rwi->reset_reason;
2079 
2080 	ibmvnic_cleanup(netdev);
2081 	release_resources(adapter);
2082 	release_sub_crqs(adapter, 0);
2083 	release_crq_queue(adapter);
2084 
2085 	/* remove the closed state so when we call open it appears
2086 	 * we are coming from the probed state.
2087 	 */
2088 	adapter->state = VNIC_PROBED;
2089 
2090 	reinit_completion(&adapter->init_done);
2091 	rc = init_crq_queue(adapter);
2092 	if (rc) {
2093 		netdev_err(adapter->netdev,
2094 			   "Couldn't initialize crq. rc=%d\n", rc);
2095 		return rc;
2096 	}
2097 
2098 	rc = ibmvnic_reset_init(adapter, false);
2099 	if (rc)
2100 		return rc;
2101 
2102 	/* If the adapter was in PROBE state prior to the reset,
2103 	 * exit here.
2104 	 */
2105 	if (reset_state == VNIC_PROBED)
2106 		return 0;
2107 
2108 	rc = ibmvnic_login(netdev);
2109 	if (rc) {
2110 		adapter->state = VNIC_PROBED;
2111 		return 0;
2112 	}
2113 
2114 	rc = init_resources(adapter);
2115 	if (rc)
2116 		return rc;
2117 
2118 	ibmvnic_disable_irqs(adapter);
2119 	adapter->state = VNIC_CLOSED;
2120 
2121 	if (reset_state == VNIC_CLOSED)
2122 		return 0;
2123 
2124 	rc = __ibmvnic_open(netdev);
2125 	if (rc)
2126 		return IBMVNIC_OPEN_FAILED;
2127 
2128 	return 0;
2129 }
2130 
2131 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2132 {
2133 	struct ibmvnic_rwi *rwi;
2134 	unsigned long flags;
2135 
2136 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2137 
2138 	if (!list_empty(&adapter->rwi_list)) {
2139 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2140 				       list);
2141 		list_del(&rwi->list);
2142 	} else {
2143 		rwi = NULL;
2144 	}
2145 
2146 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2147 	return rwi;
2148 }
2149 
2150 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2151 {
2152 	struct ibmvnic_rwi *rwi;
2153 
2154 	rwi = get_next_rwi(adapter);
2155 	while (rwi) {
2156 		kfree(rwi);
2157 		rwi = get_next_rwi(adapter);
2158 	}
2159 }
2160 
2161 static void __ibmvnic_reset(struct work_struct *work)
2162 {
2163 	struct ibmvnic_rwi *rwi;
2164 	struct ibmvnic_adapter *adapter;
2165 	bool saved_state = false;
2166 	unsigned long flags;
2167 	u32 reset_state;
2168 	int rc = 0;
2169 
2170 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2171 
2172 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2173 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2174 				      IBMVNIC_RESET_DELAY);
2175 		return;
2176 	}
2177 
2178 	rwi = get_next_rwi(adapter);
2179 	while (rwi) {
2180 		spin_lock_irqsave(&adapter->state_lock, flags);
2181 
2182 		if (adapter->state == VNIC_REMOVING ||
2183 		    adapter->state == VNIC_REMOVED) {
2184 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2185 			kfree(rwi);
2186 			rc = EBUSY;
2187 			break;
2188 		}
2189 
2190 		if (!saved_state) {
2191 			reset_state = adapter->state;
2192 			adapter->state = VNIC_RESETTING;
2193 			saved_state = true;
2194 		}
2195 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2196 
2197 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2198 			/* CHANGE_PARAM requestor holds rtnl_lock */
2199 			rc = do_change_param_reset(adapter, rwi, reset_state);
2200 		} else if (adapter->force_reset_recovery) {
2201 			/* Transport event occurred during previous reset */
2202 			if (adapter->wait_for_reset) {
2203 				/* Previous was CHANGE_PARAM; caller locked */
2204 				adapter->force_reset_recovery = false;
2205 				rc = do_hard_reset(adapter, rwi, reset_state);
2206 			} else {
2207 				rtnl_lock();
2208 				adapter->force_reset_recovery = false;
2209 				rc = do_hard_reset(adapter, rwi, reset_state);
2210 				rtnl_unlock();
2211 			}
2212 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2213 				adapter->from_passive_init)) {
2214 			rc = do_reset(adapter, rwi, reset_state);
2215 		}
2216 		kfree(rwi);
2217 		if (rc == IBMVNIC_OPEN_FAILED) {
2218 			if (list_empty(&adapter->rwi_list))
2219 				adapter->state = VNIC_CLOSED;
2220 			else
2221 				adapter->state = reset_state;
2222 			rc = 0;
2223 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2224 		    !adapter->force_reset_recovery)
2225 			break;
2226 
2227 		rwi = get_next_rwi(adapter);
2228 
2229 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2230 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2231 			adapter->force_reset_recovery = true;
2232 	}
2233 
2234 	if (adapter->wait_for_reset) {
2235 		adapter->reset_done_rc = rc;
2236 		complete(&adapter->reset_done);
2237 	}
2238 
2239 	if (rc) {
2240 		netdev_dbg(adapter->netdev, "Reset failed\n");
2241 		free_all_rwi(adapter);
2242 	}
2243 
2244 	clear_bit_unlock(0, &adapter->resetting);
2245 }
2246 
2247 static void __ibmvnic_delayed_reset(struct work_struct *work)
2248 {
2249 	struct ibmvnic_adapter *adapter;
2250 
2251 	adapter = container_of(work, struct ibmvnic_adapter,
2252 			       ibmvnic_delayed_reset.work);
2253 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2254 }
2255 
2256 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2257 			 enum ibmvnic_reset_reason reason)
2258 {
2259 	struct list_head *entry, *tmp_entry;
2260 	struct ibmvnic_rwi *rwi, *tmp;
2261 	struct net_device *netdev = adapter->netdev;
2262 	unsigned long flags;
2263 	int ret;
2264 
2265 	if (adapter->state == VNIC_REMOVING ||
2266 	    adapter->state == VNIC_REMOVED ||
2267 	    adapter->failover_pending) {
2268 		ret = EBUSY;
2269 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2270 		goto err;
2271 	}
2272 
2273 	if (adapter->state == VNIC_PROBING) {
2274 		netdev_warn(netdev, "Adapter reset during probe\n");
2275 		ret = adapter->init_done_rc = EAGAIN;
2276 		goto err;
2277 	}
2278 
2279 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2280 
2281 	list_for_each(entry, &adapter->rwi_list) {
2282 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2283 		if (tmp->reset_reason == reason) {
2284 			netdev_dbg(netdev, "Skipping matching reset\n");
2285 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2286 			ret = EBUSY;
2287 			goto err;
2288 		}
2289 	}
2290 
2291 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2292 	if (!rwi) {
2293 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2294 		ibmvnic_close(netdev);
2295 		ret = ENOMEM;
2296 		goto err;
2297 	}
2298 	/* if we just received a transport event,
2299 	 * flush reset queue and process this reset
2300 	 */
2301 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2302 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2303 			list_del(entry);
2304 	}
2305 	rwi->reset_reason = reason;
2306 	list_add_tail(&rwi->list, &adapter->rwi_list);
2307 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2308 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2309 	schedule_work(&adapter->ibmvnic_reset);
2310 
2311 	return 0;
2312 err:
2313 	return -ret;
2314 }
2315 
2316 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2317 {
2318 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2319 
2320 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2321 }
2322 
2323 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2324 				  struct ibmvnic_rx_buff *rx_buff)
2325 {
2326 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2327 
2328 	rx_buff->skb = NULL;
2329 
2330 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2331 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2332 
2333 	atomic_dec(&pool->available);
2334 }
2335 
2336 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2337 {
2338 	struct net_device *netdev = napi->dev;
2339 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2340 	int scrq_num = (int)(napi - adapter->napi);
2341 	int frames_processed = 0;
2342 
2343 restart_poll:
2344 	while (frames_processed < budget) {
2345 		struct sk_buff *skb;
2346 		struct ibmvnic_rx_buff *rx_buff;
2347 		union sub_crq *next;
2348 		u32 length;
2349 		u16 offset;
2350 		u8 flags = 0;
2351 
2352 		if (unlikely(test_bit(0, &adapter->resetting) &&
2353 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2354 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2355 			napi_complete_done(napi, frames_processed);
2356 			return frames_processed;
2357 		}
2358 
2359 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2360 			break;
2361 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2362 		rx_buff =
2363 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2364 							  rx_comp.correlator);
2365 		/* do error checking */
2366 		if (next->rx_comp.rc) {
2367 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2368 				   be16_to_cpu(next->rx_comp.rc));
2369 			/* free the entry */
2370 			next->rx_comp.first = 0;
2371 			dev_kfree_skb_any(rx_buff->skb);
2372 			remove_buff_from_pool(adapter, rx_buff);
2373 			continue;
2374 		} else if (!rx_buff->skb) {
2375 			/* free the entry */
2376 			next->rx_comp.first = 0;
2377 			remove_buff_from_pool(adapter, rx_buff);
2378 			continue;
2379 		}
2380 
2381 		length = be32_to_cpu(next->rx_comp.len);
2382 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2383 		flags = next->rx_comp.flags;
2384 		skb = rx_buff->skb;
2385 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2386 					length);
2387 
2388 		/* VLAN Header has been stripped by the system firmware and
2389 		 * needs to be inserted by the driver
2390 		 */
2391 		if (adapter->rx_vlan_header_insertion &&
2392 		    (flags & IBMVNIC_VLAN_STRIPPED))
2393 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2394 					       ntohs(next->rx_comp.vlan_tci));
2395 
2396 		/* free the entry */
2397 		next->rx_comp.first = 0;
2398 		remove_buff_from_pool(adapter, rx_buff);
2399 
2400 		skb_put(skb, length);
2401 		skb->protocol = eth_type_trans(skb, netdev);
2402 		skb_record_rx_queue(skb, scrq_num);
2403 
2404 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2405 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2406 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2407 		}
2408 
2409 		length = skb->len;
2410 		napi_gro_receive(napi, skb); /* send it up */
2411 		netdev->stats.rx_packets++;
2412 		netdev->stats.rx_bytes += length;
2413 		adapter->rx_stats_buffers[scrq_num].packets++;
2414 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2415 		frames_processed++;
2416 	}
2417 
2418 	if (adapter->state != VNIC_CLOSING)
2419 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2420 
2421 	if (frames_processed < budget) {
2422 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2423 		napi_complete_done(napi, frames_processed);
2424 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2425 		    napi_reschedule(napi)) {
2426 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2427 			goto restart_poll;
2428 		}
2429 	}
2430 	return frames_processed;
2431 }
2432 
2433 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2434 {
2435 	int rc, ret;
2436 
2437 	adapter->fallback.mtu = adapter->req_mtu;
2438 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2439 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2440 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2441 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2442 
2443 	reinit_completion(&adapter->reset_done);
2444 	adapter->wait_for_reset = true;
2445 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2446 
2447 	if (rc) {
2448 		ret = rc;
2449 		goto out;
2450 	}
2451 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2452 	if (rc) {
2453 		ret = -ENODEV;
2454 		goto out;
2455 	}
2456 
2457 	ret = 0;
2458 	if (adapter->reset_done_rc) {
2459 		ret = -EIO;
2460 		adapter->desired.mtu = adapter->fallback.mtu;
2461 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2462 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2463 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2464 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2465 
2466 		reinit_completion(&adapter->reset_done);
2467 		adapter->wait_for_reset = true;
2468 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2469 		if (rc) {
2470 			ret = rc;
2471 			goto out;
2472 		}
2473 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2474 						 60000);
2475 		if (rc) {
2476 			ret = -ENODEV;
2477 			goto out;
2478 		}
2479 	}
2480 out:
2481 	adapter->wait_for_reset = false;
2482 
2483 	return ret;
2484 }
2485 
2486 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2487 {
2488 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2489 
2490 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2491 
2492 	return wait_for_reset(adapter);
2493 }
2494 
2495 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2496 						struct net_device *dev,
2497 						netdev_features_t features)
2498 {
2499 	/* Some backing hardware adapters can not
2500 	 * handle packets with a MSS less than 224
2501 	 * or with only one segment.
2502 	 */
2503 	if (skb_is_gso(skb)) {
2504 		if (skb_shinfo(skb)->gso_size < 224 ||
2505 		    skb_shinfo(skb)->gso_segs == 1)
2506 			features &= ~NETIF_F_GSO_MASK;
2507 	}
2508 
2509 	return features;
2510 }
2511 
2512 static const struct net_device_ops ibmvnic_netdev_ops = {
2513 	.ndo_open		= ibmvnic_open,
2514 	.ndo_stop		= ibmvnic_close,
2515 	.ndo_start_xmit		= ibmvnic_xmit,
2516 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2517 	.ndo_set_mac_address	= ibmvnic_set_mac,
2518 	.ndo_validate_addr	= eth_validate_addr,
2519 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2520 	.ndo_change_mtu		= ibmvnic_change_mtu,
2521 	.ndo_features_check     = ibmvnic_features_check,
2522 };
2523 
2524 /* ethtool functions */
2525 
2526 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2527 				      struct ethtool_link_ksettings *cmd)
2528 {
2529 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2530 	int rc;
2531 
2532 	rc = send_query_phys_parms(adapter);
2533 	if (rc) {
2534 		adapter->speed = SPEED_UNKNOWN;
2535 		adapter->duplex = DUPLEX_UNKNOWN;
2536 	}
2537 	cmd->base.speed = adapter->speed;
2538 	cmd->base.duplex = adapter->duplex;
2539 	cmd->base.port = PORT_FIBRE;
2540 	cmd->base.phy_address = 0;
2541 	cmd->base.autoneg = AUTONEG_ENABLE;
2542 
2543 	return 0;
2544 }
2545 
2546 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2547 				struct ethtool_drvinfo *info)
2548 {
2549 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2550 
2551 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2552 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2553 	strlcpy(info->fw_version, adapter->fw_version,
2554 		sizeof(info->fw_version));
2555 }
2556 
2557 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2558 {
2559 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2560 
2561 	return adapter->msg_enable;
2562 }
2563 
2564 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2565 {
2566 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2567 
2568 	adapter->msg_enable = data;
2569 }
2570 
2571 static u32 ibmvnic_get_link(struct net_device *netdev)
2572 {
2573 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2574 
2575 	/* Don't need to send a query because we request a logical link up at
2576 	 * init and then we wait for link state indications
2577 	 */
2578 	return adapter->logical_link_state;
2579 }
2580 
2581 static void ibmvnic_get_ringparam(struct net_device *netdev,
2582 				  struct ethtool_ringparam *ring)
2583 {
2584 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2585 
2586 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2587 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2588 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2589 	} else {
2590 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2591 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2592 	}
2593 	ring->rx_mini_max_pending = 0;
2594 	ring->rx_jumbo_max_pending = 0;
2595 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2596 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2597 	ring->rx_mini_pending = 0;
2598 	ring->rx_jumbo_pending = 0;
2599 }
2600 
2601 static int ibmvnic_set_ringparam(struct net_device *netdev,
2602 				 struct ethtool_ringparam *ring)
2603 {
2604 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2605 	int ret;
2606 
2607 	ret = 0;
2608 	adapter->desired.rx_entries = ring->rx_pending;
2609 	adapter->desired.tx_entries = ring->tx_pending;
2610 
2611 	ret = wait_for_reset(adapter);
2612 
2613 	if (!ret &&
2614 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2615 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2616 		netdev_info(netdev,
2617 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2618 			    ring->rx_pending, ring->tx_pending,
2619 			    adapter->req_rx_add_entries_per_subcrq,
2620 			    adapter->req_tx_entries_per_subcrq);
2621 	return ret;
2622 }
2623 
2624 static void ibmvnic_get_channels(struct net_device *netdev,
2625 				 struct ethtool_channels *channels)
2626 {
2627 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2628 
2629 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2630 		channels->max_rx = adapter->max_rx_queues;
2631 		channels->max_tx = adapter->max_tx_queues;
2632 	} else {
2633 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2634 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2635 	}
2636 
2637 	channels->max_other = 0;
2638 	channels->max_combined = 0;
2639 	channels->rx_count = adapter->req_rx_queues;
2640 	channels->tx_count = adapter->req_tx_queues;
2641 	channels->other_count = 0;
2642 	channels->combined_count = 0;
2643 }
2644 
2645 static int ibmvnic_set_channels(struct net_device *netdev,
2646 				struct ethtool_channels *channels)
2647 {
2648 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2649 	int ret;
2650 
2651 	ret = 0;
2652 	adapter->desired.rx_queues = channels->rx_count;
2653 	adapter->desired.tx_queues = channels->tx_count;
2654 
2655 	ret = wait_for_reset(adapter);
2656 
2657 	if (!ret &&
2658 	    (adapter->req_rx_queues != channels->rx_count ||
2659 	     adapter->req_tx_queues != channels->tx_count))
2660 		netdev_info(netdev,
2661 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2662 			    channels->rx_count, channels->tx_count,
2663 			    adapter->req_rx_queues, adapter->req_tx_queues);
2664 	return ret;
2665 
2666 }
2667 
2668 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2669 {
2670 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2671 	int i;
2672 
2673 	switch (stringset) {
2674 	case ETH_SS_STATS:
2675 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2676 				i++, data += ETH_GSTRING_LEN)
2677 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2678 
2679 		for (i = 0; i < adapter->req_tx_queues; i++) {
2680 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2681 			data += ETH_GSTRING_LEN;
2682 
2683 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2684 			data += ETH_GSTRING_LEN;
2685 
2686 			snprintf(data, ETH_GSTRING_LEN,
2687 				 "tx%d_dropped_packets", i);
2688 			data += ETH_GSTRING_LEN;
2689 		}
2690 
2691 		for (i = 0; i < adapter->req_rx_queues; i++) {
2692 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2693 			data += ETH_GSTRING_LEN;
2694 
2695 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2696 			data += ETH_GSTRING_LEN;
2697 
2698 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2699 			data += ETH_GSTRING_LEN;
2700 		}
2701 		break;
2702 
2703 	case ETH_SS_PRIV_FLAGS:
2704 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2705 			strcpy(data + i * ETH_GSTRING_LEN,
2706 			       ibmvnic_priv_flags[i]);
2707 		break;
2708 	default:
2709 		return;
2710 	}
2711 }
2712 
2713 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2714 {
2715 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2716 
2717 	switch (sset) {
2718 	case ETH_SS_STATS:
2719 		return ARRAY_SIZE(ibmvnic_stats) +
2720 		       adapter->req_tx_queues * NUM_TX_STATS +
2721 		       adapter->req_rx_queues * NUM_RX_STATS;
2722 	case ETH_SS_PRIV_FLAGS:
2723 		return ARRAY_SIZE(ibmvnic_priv_flags);
2724 	default:
2725 		return -EOPNOTSUPP;
2726 	}
2727 }
2728 
2729 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2730 				      struct ethtool_stats *stats, u64 *data)
2731 {
2732 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2733 	union ibmvnic_crq crq;
2734 	int i, j;
2735 	int rc;
2736 
2737 	memset(&crq, 0, sizeof(crq));
2738 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2739 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2740 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2741 	crq.request_statistics.len =
2742 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2743 
2744 	/* Wait for data to be written */
2745 	reinit_completion(&adapter->stats_done);
2746 	rc = ibmvnic_send_crq(adapter, &crq);
2747 	if (rc)
2748 		return;
2749 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2750 	if (rc)
2751 		return;
2752 
2753 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2754 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2755 						ibmvnic_stats[i].offset));
2756 
2757 	for (j = 0; j < adapter->req_tx_queues; j++) {
2758 		data[i] = adapter->tx_stats_buffers[j].packets;
2759 		i++;
2760 		data[i] = adapter->tx_stats_buffers[j].bytes;
2761 		i++;
2762 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2763 		i++;
2764 	}
2765 
2766 	for (j = 0; j < adapter->req_rx_queues; j++) {
2767 		data[i] = adapter->rx_stats_buffers[j].packets;
2768 		i++;
2769 		data[i] = adapter->rx_stats_buffers[j].bytes;
2770 		i++;
2771 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2772 		i++;
2773 	}
2774 }
2775 
2776 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2777 {
2778 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2779 
2780 	return adapter->priv_flags;
2781 }
2782 
2783 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2784 {
2785 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2786 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2787 
2788 	if (which_maxes)
2789 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2790 	else
2791 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2792 
2793 	return 0;
2794 }
2795 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2796 	.get_drvinfo		= ibmvnic_get_drvinfo,
2797 	.get_msglevel		= ibmvnic_get_msglevel,
2798 	.set_msglevel		= ibmvnic_set_msglevel,
2799 	.get_link		= ibmvnic_get_link,
2800 	.get_ringparam		= ibmvnic_get_ringparam,
2801 	.set_ringparam		= ibmvnic_set_ringparam,
2802 	.get_channels		= ibmvnic_get_channels,
2803 	.set_channels		= ibmvnic_set_channels,
2804 	.get_strings            = ibmvnic_get_strings,
2805 	.get_sset_count         = ibmvnic_get_sset_count,
2806 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2807 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2808 	.get_priv_flags		= ibmvnic_get_priv_flags,
2809 	.set_priv_flags		= ibmvnic_set_priv_flags,
2810 };
2811 
2812 /* Routines for managing CRQs/sCRQs  */
2813 
2814 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2815 				   struct ibmvnic_sub_crq_queue *scrq)
2816 {
2817 	int rc;
2818 
2819 	if (scrq->irq) {
2820 		free_irq(scrq->irq, scrq);
2821 		irq_dispose_mapping(scrq->irq);
2822 		scrq->irq = 0;
2823 	}
2824 
2825 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2826 	atomic_set(&scrq->used, 0);
2827 	scrq->cur = 0;
2828 
2829 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2830 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2831 	return rc;
2832 }
2833 
2834 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2835 {
2836 	int i, rc;
2837 
2838 	for (i = 0; i < adapter->req_tx_queues; i++) {
2839 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2840 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2841 		if (rc)
2842 			return rc;
2843 	}
2844 
2845 	for (i = 0; i < adapter->req_rx_queues; i++) {
2846 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2847 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2848 		if (rc)
2849 			return rc;
2850 	}
2851 
2852 	return rc;
2853 }
2854 
2855 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2856 				  struct ibmvnic_sub_crq_queue *scrq,
2857 				  bool do_h_free)
2858 {
2859 	struct device *dev = &adapter->vdev->dev;
2860 	long rc;
2861 
2862 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2863 
2864 	if (do_h_free) {
2865 		/* Close the sub-crqs */
2866 		do {
2867 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2868 						adapter->vdev->unit_address,
2869 						scrq->crq_num);
2870 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2871 
2872 		if (rc) {
2873 			netdev_err(adapter->netdev,
2874 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2875 				   scrq->crq_num, rc);
2876 		}
2877 	}
2878 
2879 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2880 			 DMA_BIDIRECTIONAL);
2881 	free_pages((unsigned long)scrq->msgs, 2);
2882 	kfree(scrq);
2883 }
2884 
2885 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2886 							*adapter)
2887 {
2888 	struct device *dev = &adapter->vdev->dev;
2889 	struct ibmvnic_sub_crq_queue *scrq;
2890 	int rc;
2891 
2892 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2893 	if (!scrq)
2894 		return NULL;
2895 
2896 	scrq->msgs =
2897 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2898 	if (!scrq->msgs) {
2899 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2900 		goto zero_page_failed;
2901 	}
2902 
2903 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2904 					 DMA_BIDIRECTIONAL);
2905 	if (dma_mapping_error(dev, scrq->msg_token)) {
2906 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2907 		goto map_failed;
2908 	}
2909 
2910 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2911 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2912 
2913 	if (rc == H_RESOURCE)
2914 		rc = ibmvnic_reset_crq(adapter);
2915 
2916 	if (rc == H_CLOSED) {
2917 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2918 	} else if (rc) {
2919 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2920 		goto reg_failed;
2921 	}
2922 
2923 	scrq->adapter = adapter;
2924 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2925 	spin_lock_init(&scrq->lock);
2926 
2927 	netdev_dbg(adapter->netdev,
2928 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2929 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2930 
2931 	return scrq;
2932 
2933 reg_failed:
2934 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2935 			 DMA_BIDIRECTIONAL);
2936 map_failed:
2937 	free_pages((unsigned long)scrq->msgs, 2);
2938 zero_page_failed:
2939 	kfree(scrq);
2940 
2941 	return NULL;
2942 }
2943 
2944 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2945 {
2946 	int i;
2947 
2948 	if (adapter->tx_scrq) {
2949 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2950 			if (!adapter->tx_scrq[i])
2951 				continue;
2952 
2953 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2954 				   i);
2955 			if (adapter->tx_scrq[i]->irq) {
2956 				free_irq(adapter->tx_scrq[i]->irq,
2957 					 adapter->tx_scrq[i]);
2958 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2959 				adapter->tx_scrq[i]->irq = 0;
2960 			}
2961 
2962 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2963 					      do_h_free);
2964 		}
2965 
2966 		kfree(adapter->tx_scrq);
2967 		adapter->tx_scrq = NULL;
2968 		adapter->num_active_tx_scrqs = 0;
2969 	}
2970 
2971 	if (adapter->rx_scrq) {
2972 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2973 			if (!adapter->rx_scrq[i])
2974 				continue;
2975 
2976 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2977 				   i);
2978 			if (adapter->rx_scrq[i]->irq) {
2979 				free_irq(adapter->rx_scrq[i]->irq,
2980 					 adapter->rx_scrq[i]);
2981 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2982 				adapter->rx_scrq[i]->irq = 0;
2983 			}
2984 
2985 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2986 					      do_h_free);
2987 		}
2988 
2989 		kfree(adapter->rx_scrq);
2990 		adapter->rx_scrq = NULL;
2991 		adapter->num_active_rx_scrqs = 0;
2992 	}
2993 }
2994 
2995 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2996 			    struct ibmvnic_sub_crq_queue *scrq)
2997 {
2998 	struct device *dev = &adapter->vdev->dev;
2999 	unsigned long rc;
3000 
3001 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3002 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3003 	if (rc)
3004 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
3005 			scrq->hw_irq, rc);
3006 	return rc;
3007 }
3008 
3009 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3010 			   struct ibmvnic_sub_crq_queue *scrq)
3011 {
3012 	struct device *dev = &adapter->vdev->dev;
3013 	unsigned long rc;
3014 
3015 	if (scrq->hw_irq > 0x100000000ULL) {
3016 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3017 		return 1;
3018 	}
3019 
3020 	if (test_bit(0, &adapter->resetting) &&
3021 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3022 		u64 val = (0xff000000) | scrq->hw_irq;
3023 
3024 		rc = plpar_hcall_norets(H_EOI, val);
3025 		/* H_EOI would fail with rc = H_FUNCTION when running
3026 		 * in XIVE mode which is expected, but not an error.
3027 		 */
3028 		if (rc && (rc != H_FUNCTION))
3029 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3030 				val, rc);
3031 	}
3032 
3033 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3034 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3035 	if (rc)
3036 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3037 			scrq->hw_irq, rc);
3038 	return rc;
3039 }
3040 
3041 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3042 			       struct ibmvnic_sub_crq_queue *scrq)
3043 {
3044 	struct device *dev = &adapter->vdev->dev;
3045 	struct ibmvnic_tx_pool *tx_pool;
3046 	struct ibmvnic_tx_buff *txbuff;
3047 	union sub_crq *next;
3048 	int index;
3049 	int i, j;
3050 
3051 restart_loop:
3052 	while (pending_scrq(adapter, scrq)) {
3053 		unsigned int pool = scrq->pool_index;
3054 		int num_entries = 0;
3055 
3056 		next = ibmvnic_next_scrq(adapter, scrq);
3057 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3058 			if (next->tx_comp.rcs[i]) {
3059 				dev_err(dev, "tx error %x\n",
3060 					next->tx_comp.rcs[i]);
3061 				continue;
3062 			}
3063 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3064 			if (index & IBMVNIC_TSO_POOL_MASK) {
3065 				tx_pool = &adapter->tso_pool[pool];
3066 				index &= ~IBMVNIC_TSO_POOL_MASK;
3067 			} else {
3068 				tx_pool = &adapter->tx_pool[pool];
3069 			}
3070 
3071 			txbuff = &tx_pool->tx_buff[index];
3072 
3073 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3074 				if (!txbuff->data_dma[j])
3075 					continue;
3076 
3077 				txbuff->data_dma[j] = 0;
3078 			}
3079 
3080 			if (txbuff->last_frag) {
3081 				dev_kfree_skb_any(txbuff->skb);
3082 				txbuff->skb = NULL;
3083 			}
3084 
3085 			num_entries += txbuff->num_entries;
3086 
3087 			tx_pool->free_map[tx_pool->producer_index] = index;
3088 			tx_pool->producer_index =
3089 				(tx_pool->producer_index + 1) %
3090 					tx_pool->num_buffers;
3091 		}
3092 		/* remove tx_comp scrq*/
3093 		next->tx_comp.first = 0;
3094 
3095 		if (atomic_sub_return(num_entries, &scrq->used) <=
3096 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3097 		    __netif_subqueue_stopped(adapter->netdev,
3098 					     scrq->pool_index)) {
3099 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3100 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3101 				   scrq->pool_index);
3102 		}
3103 	}
3104 
3105 	enable_scrq_irq(adapter, scrq);
3106 
3107 	if (pending_scrq(adapter, scrq)) {
3108 		disable_scrq_irq(adapter, scrq);
3109 		goto restart_loop;
3110 	}
3111 
3112 	return 0;
3113 }
3114 
3115 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3116 {
3117 	struct ibmvnic_sub_crq_queue *scrq = instance;
3118 	struct ibmvnic_adapter *adapter = scrq->adapter;
3119 
3120 	disable_scrq_irq(adapter, scrq);
3121 	ibmvnic_complete_tx(adapter, scrq);
3122 
3123 	return IRQ_HANDLED;
3124 }
3125 
3126 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3127 {
3128 	struct ibmvnic_sub_crq_queue *scrq = instance;
3129 	struct ibmvnic_adapter *adapter = scrq->adapter;
3130 
3131 	/* When booting a kdump kernel we can hit pending interrupts
3132 	 * prior to completing driver initialization.
3133 	 */
3134 	if (unlikely(adapter->state != VNIC_OPEN))
3135 		return IRQ_NONE;
3136 
3137 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3138 
3139 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3140 		disable_scrq_irq(adapter, scrq);
3141 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3142 	}
3143 
3144 	return IRQ_HANDLED;
3145 }
3146 
3147 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3148 {
3149 	struct device *dev = &adapter->vdev->dev;
3150 	struct ibmvnic_sub_crq_queue *scrq;
3151 	int i = 0, j = 0;
3152 	int rc = 0;
3153 
3154 	for (i = 0; i < adapter->req_tx_queues; i++) {
3155 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3156 			   i);
3157 		scrq = adapter->tx_scrq[i];
3158 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3159 
3160 		if (!scrq->irq) {
3161 			rc = -EINVAL;
3162 			dev_err(dev, "Error mapping irq\n");
3163 			goto req_tx_irq_failed;
3164 		}
3165 
3166 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3167 			 adapter->vdev->unit_address, i);
3168 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3169 				 0, scrq->name, scrq);
3170 
3171 		if (rc) {
3172 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3173 				scrq->irq, rc);
3174 			irq_dispose_mapping(scrq->irq);
3175 			goto req_tx_irq_failed;
3176 		}
3177 	}
3178 
3179 	for (i = 0; i < adapter->req_rx_queues; i++) {
3180 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3181 			   i);
3182 		scrq = adapter->rx_scrq[i];
3183 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3184 		if (!scrq->irq) {
3185 			rc = -EINVAL;
3186 			dev_err(dev, "Error mapping irq\n");
3187 			goto req_rx_irq_failed;
3188 		}
3189 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3190 			 adapter->vdev->unit_address, i);
3191 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3192 				 0, scrq->name, scrq);
3193 		if (rc) {
3194 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3195 				scrq->irq, rc);
3196 			irq_dispose_mapping(scrq->irq);
3197 			goto req_rx_irq_failed;
3198 		}
3199 	}
3200 	return rc;
3201 
3202 req_rx_irq_failed:
3203 	for (j = 0; j < i; j++) {
3204 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3205 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3206 	}
3207 	i = adapter->req_tx_queues;
3208 req_tx_irq_failed:
3209 	for (j = 0; j < i; j++) {
3210 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3211 		irq_dispose_mapping(adapter->tx_scrq[j]->irq);
3212 	}
3213 	release_sub_crqs(adapter, 1);
3214 	return rc;
3215 }
3216 
3217 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3218 {
3219 	struct device *dev = &adapter->vdev->dev;
3220 	struct ibmvnic_sub_crq_queue **allqueues;
3221 	int registered_queues = 0;
3222 	int total_queues;
3223 	int more = 0;
3224 	int i;
3225 
3226 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3227 
3228 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3229 	if (!allqueues)
3230 		return -1;
3231 
3232 	for (i = 0; i < total_queues; i++) {
3233 		allqueues[i] = init_sub_crq_queue(adapter);
3234 		if (!allqueues[i]) {
3235 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3236 			break;
3237 		}
3238 		registered_queues++;
3239 	}
3240 
3241 	/* Make sure we were able to register the minimum number of queues */
3242 	if (registered_queues <
3243 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3244 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3245 		goto tx_failed;
3246 	}
3247 
3248 	/* Distribute the failed allocated queues*/
3249 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3250 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3251 		switch (i % 3) {
3252 		case 0:
3253 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3254 				adapter->req_rx_queues--;
3255 			else
3256 				more++;
3257 			break;
3258 		case 1:
3259 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3260 				adapter->req_tx_queues--;
3261 			else
3262 				more++;
3263 			break;
3264 		}
3265 	}
3266 
3267 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3268 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3269 	if (!adapter->tx_scrq)
3270 		goto tx_failed;
3271 
3272 	for (i = 0; i < adapter->req_tx_queues; i++) {
3273 		adapter->tx_scrq[i] = allqueues[i];
3274 		adapter->tx_scrq[i]->pool_index = i;
3275 		adapter->num_active_tx_scrqs++;
3276 	}
3277 
3278 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3279 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3280 	if (!adapter->rx_scrq)
3281 		goto rx_failed;
3282 
3283 	for (i = 0; i < adapter->req_rx_queues; i++) {
3284 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3285 		adapter->rx_scrq[i]->scrq_num = i;
3286 		adapter->num_active_rx_scrqs++;
3287 	}
3288 
3289 	kfree(allqueues);
3290 	return 0;
3291 
3292 rx_failed:
3293 	kfree(adapter->tx_scrq);
3294 	adapter->tx_scrq = NULL;
3295 tx_failed:
3296 	for (i = 0; i < registered_queues; i++)
3297 		release_sub_crq_queue(adapter, allqueues[i], 1);
3298 	kfree(allqueues);
3299 	return -1;
3300 }
3301 
3302 static void send_request_cap(struct ibmvnic_adapter *adapter, int retry)
3303 {
3304 	struct device *dev = &adapter->vdev->dev;
3305 	union ibmvnic_crq crq;
3306 	int max_entries;
3307 
3308 	if (!retry) {
3309 		/* Sub-CRQ entries are 32 byte long */
3310 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3311 
3312 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3313 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3314 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3315 			return;
3316 		}
3317 
3318 		if (adapter->desired.mtu)
3319 			adapter->req_mtu = adapter->desired.mtu;
3320 		else
3321 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3322 
3323 		if (!adapter->desired.tx_entries)
3324 			adapter->desired.tx_entries =
3325 					adapter->max_tx_entries_per_subcrq;
3326 		if (!adapter->desired.rx_entries)
3327 			adapter->desired.rx_entries =
3328 					adapter->max_rx_add_entries_per_subcrq;
3329 
3330 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3331 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3332 
3333 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3334 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3335 			adapter->desired.tx_entries = max_entries;
3336 		}
3337 
3338 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3339 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3340 			adapter->desired.rx_entries = max_entries;
3341 		}
3342 
3343 		if (adapter->desired.tx_entries)
3344 			adapter->req_tx_entries_per_subcrq =
3345 					adapter->desired.tx_entries;
3346 		else
3347 			adapter->req_tx_entries_per_subcrq =
3348 					adapter->max_tx_entries_per_subcrq;
3349 
3350 		if (adapter->desired.rx_entries)
3351 			adapter->req_rx_add_entries_per_subcrq =
3352 					adapter->desired.rx_entries;
3353 		else
3354 			adapter->req_rx_add_entries_per_subcrq =
3355 					adapter->max_rx_add_entries_per_subcrq;
3356 
3357 		if (adapter->desired.tx_queues)
3358 			adapter->req_tx_queues =
3359 					adapter->desired.tx_queues;
3360 		else
3361 			adapter->req_tx_queues =
3362 					adapter->opt_tx_comp_sub_queues;
3363 
3364 		if (adapter->desired.rx_queues)
3365 			adapter->req_rx_queues =
3366 					adapter->desired.rx_queues;
3367 		else
3368 			adapter->req_rx_queues =
3369 					adapter->opt_rx_comp_queues;
3370 
3371 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3372 	}
3373 
3374 	memset(&crq, 0, sizeof(crq));
3375 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3376 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3377 
3378 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3379 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3380 	atomic_inc(&adapter->running_cap_crqs);
3381 	ibmvnic_send_crq(adapter, &crq);
3382 
3383 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3384 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3385 	atomic_inc(&adapter->running_cap_crqs);
3386 	ibmvnic_send_crq(adapter, &crq);
3387 
3388 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3389 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3390 	atomic_inc(&adapter->running_cap_crqs);
3391 	ibmvnic_send_crq(adapter, &crq);
3392 
3393 	crq.request_capability.capability =
3394 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3395 	crq.request_capability.number =
3396 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3397 	atomic_inc(&adapter->running_cap_crqs);
3398 	ibmvnic_send_crq(adapter, &crq);
3399 
3400 	crq.request_capability.capability =
3401 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3402 	crq.request_capability.number =
3403 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3404 	atomic_inc(&adapter->running_cap_crqs);
3405 	ibmvnic_send_crq(adapter, &crq);
3406 
3407 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3408 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3409 	atomic_inc(&adapter->running_cap_crqs);
3410 	ibmvnic_send_crq(adapter, &crq);
3411 
3412 	if (adapter->netdev->flags & IFF_PROMISC) {
3413 		if (adapter->promisc_supported) {
3414 			crq.request_capability.capability =
3415 			    cpu_to_be16(PROMISC_REQUESTED);
3416 			crq.request_capability.number = cpu_to_be64(1);
3417 			atomic_inc(&adapter->running_cap_crqs);
3418 			ibmvnic_send_crq(adapter, &crq);
3419 		}
3420 	} else {
3421 		crq.request_capability.capability =
3422 		    cpu_to_be16(PROMISC_REQUESTED);
3423 		crq.request_capability.number = cpu_to_be64(0);
3424 		atomic_inc(&adapter->running_cap_crqs);
3425 		ibmvnic_send_crq(adapter, &crq);
3426 	}
3427 }
3428 
3429 static int pending_scrq(struct ibmvnic_adapter *adapter,
3430 			struct ibmvnic_sub_crq_queue *scrq)
3431 {
3432 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3433 
3434 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3435 		return 1;
3436 	else
3437 		return 0;
3438 }
3439 
3440 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3441 					struct ibmvnic_sub_crq_queue *scrq)
3442 {
3443 	union sub_crq *entry;
3444 	unsigned long flags;
3445 
3446 	spin_lock_irqsave(&scrq->lock, flags);
3447 	entry = &scrq->msgs[scrq->cur];
3448 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3449 		if (++scrq->cur == scrq->size)
3450 			scrq->cur = 0;
3451 	} else {
3452 		entry = NULL;
3453 	}
3454 	spin_unlock_irqrestore(&scrq->lock, flags);
3455 
3456 	return entry;
3457 }
3458 
3459 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3460 {
3461 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3462 	union ibmvnic_crq *crq;
3463 
3464 	crq = &queue->msgs[queue->cur];
3465 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3466 		if (++queue->cur == queue->size)
3467 			queue->cur = 0;
3468 	} else {
3469 		crq = NULL;
3470 	}
3471 
3472 	return crq;
3473 }
3474 
3475 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3476 {
3477 	switch (rc) {
3478 	case H_PARAMETER:
3479 		dev_warn_ratelimited(dev,
3480 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3481 				     func, rc);
3482 		break;
3483 	case H_CLOSED:
3484 		dev_warn_ratelimited(dev,
3485 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3486 				     func, rc);
3487 		break;
3488 	default:
3489 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3490 		break;
3491 	}
3492 }
3493 
3494 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3495 		       union sub_crq *sub_crq)
3496 {
3497 	unsigned int ua = adapter->vdev->unit_address;
3498 	struct device *dev = &adapter->vdev->dev;
3499 	u64 *u64_crq = (u64 *)sub_crq;
3500 	int rc;
3501 
3502 	netdev_dbg(adapter->netdev,
3503 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3504 		   (unsigned long int)cpu_to_be64(remote_handle),
3505 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3506 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3507 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3508 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3509 
3510 	/* Make sure the hypervisor sees the complete request */
3511 	mb();
3512 
3513 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3514 				cpu_to_be64(remote_handle),
3515 				cpu_to_be64(u64_crq[0]),
3516 				cpu_to_be64(u64_crq[1]),
3517 				cpu_to_be64(u64_crq[2]),
3518 				cpu_to_be64(u64_crq[3]));
3519 
3520 	if (rc)
3521 		print_subcrq_error(dev, rc, __func__);
3522 
3523 	return rc;
3524 }
3525 
3526 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3527 				u64 remote_handle, u64 ioba, u64 num_entries)
3528 {
3529 	unsigned int ua = adapter->vdev->unit_address;
3530 	struct device *dev = &adapter->vdev->dev;
3531 	int rc;
3532 
3533 	/* Make sure the hypervisor sees the complete request */
3534 	mb();
3535 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3536 				cpu_to_be64(remote_handle),
3537 				ioba, num_entries);
3538 
3539 	if (rc)
3540 		print_subcrq_error(dev, rc, __func__);
3541 
3542 	return rc;
3543 }
3544 
3545 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3546 			    union ibmvnic_crq *crq)
3547 {
3548 	unsigned int ua = adapter->vdev->unit_address;
3549 	struct device *dev = &adapter->vdev->dev;
3550 	u64 *u64_crq = (u64 *)crq;
3551 	int rc;
3552 
3553 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3554 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3555 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3556 
3557 	if (!adapter->crq.active &&
3558 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3559 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3560 		return -EINVAL;
3561 	}
3562 
3563 	/* Make sure the hypervisor sees the complete request */
3564 	mb();
3565 
3566 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3567 				cpu_to_be64(u64_crq[0]),
3568 				cpu_to_be64(u64_crq[1]));
3569 
3570 	if (rc) {
3571 		if (rc == H_CLOSED) {
3572 			dev_warn(dev, "CRQ Queue closed\n");
3573 			/* do not reset, report the fail, wait for passive init from server */
3574 		}
3575 
3576 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3577 	}
3578 
3579 	return rc;
3580 }
3581 
3582 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3583 {
3584 	struct device *dev = &adapter->vdev->dev;
3585 	union ibmvnic_crq crq;
3586 	int retries = 100;
3587 	int rc;
3588 
3589 	memset(&crq, 0, sizeof(crq));
3590 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3591 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3592 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3593 
3594 	do {
3595 		rc = ibmvnic_send_crq(adapter, &crq);
3596 		if (rc != H_CLOSED)
3597 			break;
3598 		retries--;
3599 		msleep(50);
3600 
3601 	} while (retries > 0);
3602 
3603 	if (rc) {
3604 		dev_err(dev, "Failed to send init request, rc = %d\n", rc);
3605 		return rc;
3606 	}
3607 
3608 	return 0;
3609 }
3610 
3611 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3612 {
3613 	union ibmvnic_crq crq;
3614 
3615 	memset(&crq, 0, sizeof(crq));
3616 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3617 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3618 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3619 
3620 	return ibmvnic_send_crq(adapter, &crq);
3621 }
3622 
3623 struct vnic_login_client_data {
3624 	u8	type;
3625 	__be16	len;
3626 	char	name[];
3627 } __packed;
3628 
3629 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3630 {
3631 	int len;
3632 
3633 	/* Calculate the amount of buffer space needed for the
3634 	 * vnic client data in the login buffer. There are four entries,
3635 	 * OS name, LPAR name, device name, and a null last entry.
3636 	 */
3637 	len = 4 * sizeof(struct vnic_login_client_data);
3638 	len += 6; /* "Linux" plus NULL */
3639 	len += strlen(utsname()->nodename) + 1;
3640 	len += strlen(adapter->netdev->name) + 1;
3641 
3642 	return len;
3643 }
3644 
3645 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3646 				 struct vnic_login_client_data *vlcd)
3647 {
3648 	const char *os_name = "Linux";
3649 	int len;
3650 
3651 	/* Type 1 - LPAR OS */
3652 	vlcd->type = 1;
3653 	len = strlen(os_name) + 1;
3654 	vlcd->len = cpu_to_be16(len);
3655 	strncpy(vlcd->name, os_name, len);
3656 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3657 
3658 	/* Type 2 - LPAR name */
3659 	vlcd->type = 2;
3660 	len = strlen(utsname()->nodename) + 1;
3661 	vlcd->len = cpu_to_be16(len);
3662 	strncpy(vlcd->name, utsname()->nodename, len);
3663 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3664 
3665 	/* Type 3 - device name */
3666 	vlcd->type = 3;
3667 	len = strlen(adapter->netdev->name) + 1;
3668 	vlcd->len = cpu_to_be16(len);
3669 	strncpy(vlcd->name, adapter->netdev->name, len);
3670 }
3671 
3672 static int send_login(struct ibmvnic_adapter *adapter)
3673 {
3674 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3675 	struct ibmvnic_login_buffer *login_buffer;
3676 	struct device *dev = &adapter->vdev->dev;
3677 	dma_addr_t rsp_buffer_token;
3678 	dma_addr_t buffer_token;
3679 	size_t rsp_buffer_size;
3680 	union ibmvnic_crq crq;
3681 	size_t buffer_size;
3682 	__be64 *tx_list_p;
3683 	__be64 *rx_list_p;
3684 	int client_data_len;
3685 	struct vnic_login_client_data *vlcd;
3686 	int i;
3687 
3688 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3689 		netdev_err(adapter->netdev,
3690 			   "RX or TX queues are not allocated, device login failed\n");
3691 		return -1;
3692 	}
3693 
3694 	release_login_rsp_buffer(adapter);
3695 	client_data_len = vnic_client_data_len(adapter);
3696 
3697 	buffer_size =
3698 	    sizeof(struct ibmvnic_login_buffer) +
3699 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3700 	    client_data_len;
3701 
3702 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3703 	if (!login_buffer)
3704 		goto buf_alloc_failed;
3705 
3706 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3707 				      DMA_TO_DEVICE);
3708 	if (dma_mapping_error(dev, buffer_token)) {
3709 		dev_err(dev, "Couldn't map login buffer\n");
3710 		goto buf_map_failed;
3711 	}
3712 
3713 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3714 			  sizeof(u64) * adapter->req_tx_queues +
3715 			  sizeof(u64) * adapter->req_rx_queues +
3716 			  sizeof(u64) * adapter->req_rx_queues +
3717 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3718 
3719 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3720 	if (!login_rsp_buffer)
3721 		goto buf_rsp_alloc_failed;
3722 
3723 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3724 					  rsp_buffer_size, DMA_FROM_DEVICE);
3725 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3726 		dev_err(dev, "Couldn't map login rsp buffer\n");
3727 		goto buf_rsp_map_failed;
3728 	}
3729 
3730 	adapter->login_buf = login_buffer;
3731 	adapter->login_buf_token = buffer_token;
3732 	adapter->login_buf_sz = buffer_size;
3733 	adapter->login_rsp_buf = login_rsp_buffer;
3734 	adapter->login_rsp_buf_token = rsp_buffer_token;
3735 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3736 
3737 	login_buffer->len = cpu_to_be32(buffer_size);
3738 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3739 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3740 	login_buffer->off_txcomp_subcrqs =
3741 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3742 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3743 	login_buffer->off_rxcomp_subcrqs =
3744 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3745 			sizeof(u64) * adapter->req_tx_queues);
3746 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3747 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3748 
3749 	tx_list_p = (__be64 *)((char *)login_buffer +
3750 				      sizeof(struct ibmvnic_login_buffer));
3751 	rx_list_p = (__be64 *)((char *)login_buffer +
3752 				      sizeof(struct ibmvnic_login_buffer) +
3753 				      sizeof(u64) * adapter->req_tx_queues);
3754 
3755 	for (i = 0; i < adapter->req_tx_queues; i++) {
3756 		if (adapter->tx_scrq[i]) {
3757 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3758 						   crq_num);
3759 		}
3760 	}
3761 
3762 	for (i = 0; i < adapter->req_rx_queues; i++) {
3763 		if (adapter->rx_scrq[i]) {
3764 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3765 						   crq_num);
3766 		}
3767 	}
3768 
3769 	/* Insert vNIC login client data */
3770 	vlcd = (struct vnic_login_client_data *)
3771 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3772 	login_buffer->client_data_offset =
3773 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3774 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3775 
3776 	vnic_add_client_data(adapter, vlcd);
3777 
3778 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3779 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3780 		netdev_dbg(adapter->netdev, "%016lx\n",
3781 			   ((unsigned long int *)(adapter->login_buf))[i]);
3782 	}
3783 
3784 	memset(&crq, 0, sizeof(crq));
3785 	crq.login.first = IBMVNIC_CRQ_CMD;
3786 	crq.login.cmd = LOGIN;
3787 	crq.login.ioba = cpu_to_be32(buffer_token);
3788 	crq.login.len = cpu_to_be32(buffer_size);
3789 	ibmvnic_send_crq(adapter, &crq);
3790 
3791 	return 0;
3792 
3793 buf_rsp_map_failed:
3794 	kfree(login_rsp_buffer);
3795 buf_rsp_alloc_failed:
3796 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3797 buf_map_failed:
3798 	kfree(login_buffer);
3799 buf_alloc_failed:
3800 	return -1;
3801 }
3802 
3803 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3804 			    u32 len, u8 map_id)
3805 {
3806 	union ibmvnic_crq crq;
3807 
3808 	memset(&crq, 0, sizeof(crq));
3809 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3810 	crq.request_map.cmd = REQUEST_MAP;
3811 	crq.request_map.map_id = map_id;
3812 	crq.request_map.ioba = cpu_to_be32(addr);
3813 	crq.request_map.len = cpu_to_be32(len);
3814 	return ibmvnic_send_crq(adapter, &crq);
3815 }
3816 
3817 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3818 {
3819 	union ibmvnic_crq crq;
3820 
3821 	memset(&crq, 0, sizeof(crq));
3822 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3823 	crq.request_unmap.cmd = REQUEST_UNMAP;
3824 	crq.request_unmap.map_id = map_id;
3825 	return ibmvnic_send_crq(adapter, &crq);
3826 }
3827 
3828 static void send_query_map(struct ibmvnic_adapter *adapter)
3829 {
3830 	union ibmvnic_crq crq;
3831 
3832 	memset(&crq, 0, sizeof(crq));
3833 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3834 	crq.query_map.cmd = QUERY_MAP;
3835 	ibmvnic_send_crq(adapter, &crq);
3836 }
3837 
3838 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3839 static void send_query_cap(struct ibmvnic_adapter *adapter)
3840 {
3841 	union ibmvnic_crq crq;
3842 
3843 	atomic_set(&adapter->running_cap_crqs, 0);
3844 	memset(&crq, 0, sizeof(crq));
3845 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3846 	crq.query_capability.cmd = QUERY_CAPABILITY;
3847 
3848 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3849 	atomic_inc(&adapter->running_cap_crqs);
3850 	ibmvnic_send_crq(adapter, &crq);
3851 
3852 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3853 	atomic_inc(&adapter->running_cap_crqs);
3854 	ibmvnic_send_crq(adapter, &crq);
3855 
3856 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3857 	atomic_inc(&adapter->running_cap_crqs);
3858 	ibmvnic_send_crq(adapter, &crq);
3859 
3860 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3861 	atomic_inc(&adapter->running_cap_crqs);
3862 	ibmvnic_send_crq(adapter, &crq);
3863 
3864 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3865 	atomic_inc(&adapter->running_cap_crqs);
3866 	ibmvnic_send_crq(adapter, &crq);
3867 
3868 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3869 	atomic_inc(&adapter->running_cap_crqs);
3870 	ibmvnic_send_crq(adapter, &crq);
3871 
3872 	crq.query_capability.capability =
3873 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3874 	atomic_inc(&adapter->running_cap_crqs);
3875 	ibmvnic_send_crq(adapter, &crq);
3876 
3877 	crq.query_capability.capability =
3878 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3879 	atomic_inc(&adapter->running_cap_crqs);
3880 	ibmvnic_send_crq(adapter, &crq);
3881 
3882 	crq.query_capability.capability =
3883 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3884 	atomic_inc(&adapter->running_cap_crqs);
3885 	ibmvnic_send_crq(adapter, &crq);
3886 
3887 	crq.query_capability.capability =
3888 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3889 	atomic_inc(&adapter->running_cap_crqs);
3890 	ibmvnic_send_crq(adapter, &crq);
3891 
3892 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3893 	atomic_inc(&adapter->running_cap_crqs);
3894 	ibmvnic_send_crq(adapter, &crq);
3895 
3896 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3897 	atomic_inc(&adapter->running_cap_crqs);
3898 	ibmvnic_send_crq(adapter, &crq);
3899 
3900 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3901 	atomic_inc(&adapter->running_cap_crqs);
3902 	ibmvnic_send_crq(adapter, &crq);
3903 
3904 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3905 	atomic_inc(&adapter->running_cap_crqs);
3906 	ibmvnic_send_crq(adapter, &crq);
3907 
3908 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3909 	atomic_inc(&adapter->running_cap_crqs);
3910 	ibmvnic_send_crq(adapter, &crq);
3911 
3912 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3913 	atomic_inc(&adapter->running_cap_crqs);
3914 	ibmvnic_send_crq(adapter, &crq);
3915 
3916 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3917 	atomic_inc(&adapter->running_cap_crqs);
3918 	ibmvnic_send_crq(adapter, &crq);
3919 
3920 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3921 	atomic_inc(&adapter->running_cap_crqs);
3922 	ibmvnic_send_crq(adapter, &crq);
3923 
3924 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3925 	atomic_inc(&adapter->running_cap_crqs);
3926 	ibmvnic_send_crq(adapter, &crq);
3927 
3928 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3929 	atomic_inc(&adapter->running_cap_crqs);
3930 	ibmvnic_send_crq(adapter, &crq);
3931 
3932 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3933 	atomic_inc(&adapter->running_cap_crqs);
3934 	ibmvnic_send_crq(adapter, &crq);
3935 
3936 	crq.query_capability.capability =
3937 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3938 	atomic_inc(&adapter->running_cap_crqs);
3939 	ibmvnic_send_crq(adapter, &crq);
3940 
3941 	crq.query_capability.capability =
3942 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3943 	atomic_inc(&adapter->running_cap_crqs);
3944 	ibmvnic_send_crq(adapter, &crq);
3945 
3946 	crq.query_capability.capability =
3947 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3948 	atomic_inc(&adapter->running_cap_crqs);
3949 	ibmvnic_send_crq(adapter, &crq);
3950 
3951 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3952 	atomic_inc(&adapter->running_cap_crqs);
3953 	ibmvnic_send_crq(adapter, &crq);
3954 }
3955 
3956 static void send_query_ip_offload(struct ibmvnic_adapter *adapter)
3957 {
3958 	int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3959 	struct device *dev = &adapter->vdev->dev;
3960 	union ibmvnic_crq crq;
3961 
3962 	adapter->ip_offload_tok =
3963 		dma_map_single(dev,
3964 			       &adapter->ip_offload_buf,
3965 			       buf_sz,
3966 			       DMA_FROM_DEVICE);
3967 
3968 	if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3969 		if (!firmware_has_feature(FW_FEATURE_CMO))
3970 			dev_err(dev, "Couldn't map offload buffer\n");
3971 		return;
3972 	}
3973 
3974 	memset(&crq, 0, sizeof(crq));
3975 	crq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3976 	crq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3977 	crq.query_ip_offload.len = cpu_to_be32(buf_sz);
3978 	crq.query_ip_offload.ioba =
3979 	    cpu_to_be32(adapter->ip_offload_tok);
3980 
3981 	ibmvnic_send_crq(adapter, &crq);
3982 }
3983 
3984 static void send_control_ip_offload(struct ibmvnic_adapter *adapter)
3985 {
3986 	struct ibmvnic_control_ip_offload_buffer *ctrl_buf = &adapter->ip_offload_ctrl;
3987 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3988 	struct device *dev = &adapter->vdev->dev;
3989 	netdev_features_t old_hw_features = 0;
3990 	union ibmvnic_crq crq;
3991 
3992 	adapter->ip_offload_ctrl_tok =
3993 		dma_map_single(dev,
3994 			       ctrl_buf,
3995 			       sizeof(adapter->ip_offload_ctrl),
3996 			       DMA_TO_DEVICE);
3997 
3998 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3999 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4000 		return;
4001 	}
4002 
4003 	ctrl_buf->len = cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4004 	ctrl_buf->version = cpu_to_be32(INITIAL_VERSION_IOB);
4005 	ctrl_buf->ipv4_chksum = buf->ipv4_chksum;
4006 	ctrl_buf->ipv6_chksum = buf->ipv6_chksum;
4007 	ctrl_buf->tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4008 	ctrl_buf->udp_ipv4_chksum = buf->udp_ipv4_chksum;
4009 	ctrl_buf->tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4010 	ctrl_buf->udp_ipv6_chksum = buf->udp_ipv6_chksum;
4011 	ctrl_buf->large_tx_ipv4 = buf->large_tx_ipv4;
4012 	ctrl_buf->large_tx_ipv6 = buf->large_tx_ipv6;
4013 
4014 	/* large_rx disabled for now, additional features needed */
4015 	ctrl_buf->large_rx_ipv4 = 0;
4016 	ctrl_buf->large_rx_ipv6 = 0;
4017 
4018 	if (adapter->state != VNIC_PROBING) {
4019 		old_hw_features = adapter->netdev->hw_features;
4020 		adapter->netdev->hw_features = 0;
4021 	}
4022 
4023 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4024 
4025 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4026 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4027 
4028 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4029 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4030 
4031 	if ((adapter->netdev->features &
4032 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4033 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4034 
4035 	if (buf->large_tx_ipv4)
4036 		adapter->netdev->hw_features |= NETIF_F_TSO;
4037 	if (buf->large_tx_ipv6)
4038 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4039 
4040 	if (adapter->state == VNIC_PROBING) {
4041 		adapter->netdev->features |= adapter->netdev->hw_features;
4042 	} else if (old_hw_features != adapter->netdev->hw_features) {
4043 		netdev_features_t tmp = 0;
4044 
4045 		/* disable features no longer supported */
4046 		adapter->netdev->features &= adapter->netdev->hw_features;
4047 		/* turn on features now supported if previously enabled */
4048 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4049 			adapter->netdev->hw_features;
4050 		adapter->netdev->features |=
4051 				tmp & adapter->netdev->wanted_features;
4052 	}
4053 
4054 	memset(&crq, 0, sizeof(crq));
4055 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4056 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4057 	crq.control_ip_offload.len =
4058 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4059 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4060 	ibmvnic_send_crq(adapter, &crq);
4061 }
4062 
4063 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
4064 				struct ibmvnic_adapter *adapter)
4065 {
4066 	struct device *dev = &adapter->vdev->dev;
4067 
4068 	if (crq->get_vpd_size_rsp.rc.code) {
4069 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
4070 			crq->get_vpd_size_rsp.rc.code);
4071 		complete(&adapter->fw_done);
4072 		return;
4073 	}
4074 
4075 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
4076 	complete(&adapter->fw_done);
4077 }
4078 
4079 static void handle_vpd_rsp(union ibmvnic_crq *crq,
4080 			   struct ibmvnic_adapter *adapter)
4081 {
4082 	struct device *dev = &adapter->vdev->dev;
4083 	unsigned char *substr = NULL;
4084 	u8 fw_level_len = 0;
4085 
4086 	memset(adapter->fw_version, 0, 32);
4087 
4088 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
4089 			 DMA_FROM_DEVICE);
4090 
4091 	if (crq->get_vpd_rsp.rc.code) {
4092 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
4093 			crq->get_vpd_rsp.rc.code);
4094 		goto complete;
4095 	}
4096 
4097 	/* get the position of the firmware version info
4098 	 * located after the ASCII 'RM' substring in the buffer
4099 	 */
4100 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
4101 	if (!substr) {
4102 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
4103 		goto complete;
4104 	}
4105 
4106 	/* get length of firmware level ASCII substring */
4107 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
4108 		fw_level_len = *(substr + 2);
4109 	} else {
4110 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
4111 		goto complete;
4112 	}
4113 
4114 	/* copy firmware version string from vpd into adapter */
4115 	if ((substr + 3 + fw_level_len) <
4116 	    (adapter->vpd->buff + adapter->vpd->len)) {
4117 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
4118 	} else {
4119 		dev_info(dev, "FW substr extrapolated VPD buff\n");
4120 	}
4121 
4122 complete:
4123 	if (adapter->fw_version[0] == '\0')
4124 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
4125 	complete(&adapter->fw_done);
4126 }
4127 
4128 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
4129 {
4130 	struct device *dev = &adapter->vdev->dev;
4131 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4132 	int i;
4133 
4134 	dma_unmap_single(dev, adapter->ip_offload_tok,
4135 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4136 
4137 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4138 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4139 		netdev_dbg(adapter->netdev, "%016lx\n",
4140 			   ((unsigned long int *)(buf))[i]);
4141 
4142 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4143 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4144 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4145 		   buf->tcp_ipv4_chksum);
4146 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4147 		   buf->tcp_ipv6_chksum);
4148 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4149 		   buf->udp_ipv4_chksum);
4150 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4151 		   buf->udp_ipv6_chksum);
4152 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4153 		   buf->large_tx_ipv4);
4154 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4155 		   buf->large_tx_ipv6);
4156 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4157 		   buf->large_rx_ipv4);
4158 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4159 		   buf->large_rx_ipv6);
4160 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4161 		   buf->max_ipv4_header_size);
4162 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4163 		   buf->max_ipv6_header_size);
4164 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4165 		   buf->max_tcp_header_size);
4166 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4167 		   buf->max_udp_header_size);
4168 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4169 		   buf->max_large_tx_size);
4170 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4171 		   buf->max_large_rx_size);
4172 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4173 		   buf->ipv6_extension_header);
4174 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4175 		   buf->tcp_pseudosum_req);
4176 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4177 		   buf->num_ipv6_ext_headers);
4178 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4179 		   buf->off_ipv6_ext_headers);
4180 
4181 	send_control_ip_offload(adapter);
4182 }
4183 
4184 static const char *ibmvnic_fw_err_cause(u16 cause)
4185 {
4186 	switch (cause) {
4187 	case ADAPTER_PROBLEM:
4188 		return "adapter problem";
4189 	case BUS_PROBLEM:
4190 		return "bus problem";
4191 	case FW_PROBLEM:
4192 		return "firmware problem";
4193 	case DD_PROBLEM:
4194 		return "device driver problem";
4195 	case EEH_RECOVERY:
4196 		return "EEH recovery";
4197 	case FW_UPDATED:
4198 		return "firmware updated";
4199 	case LOW_MEMORY:
4200 		return "low Memory";
4201 	default:
4202 		return "unknown";
4203 	}
4204 }
4205 
4206 static void handle_error_indication(union ibmvnic_crq *crq,
4207 				    struct ibmvnic_adapter *adapter)
4208 {
4209 	struct device *dev = &adapter->vdev->dev;
4210 	u16 cause;
4211 
4212 	cause = be16_to_cpu(crq->error_indication.error_cause);
4213 
4214 	dev_warn_ratelimited(dev,
4215 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4216 			     crq->error_indication.flags
4217 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4218 			     ibmvnic_fw_err_cause(cause));
4219 
4220 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4221 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4222 	else
4223 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4224 }
4225 
4226 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4227 				 struct ibmvnic_adapter *adapter)
4228 {
4229 	struct net_device *netdev = adapter->netdev;
4230 	struct device *dev = &adapter->vdev->dev;
4231 	long rc;
4232 
4233 	rc = crq->change_mac_addr_rsp.rc.code;
4234 	if (rc) {
4235 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4236 		goto out;
4237 	}
4238 	ether_addr_copy(netdev->dev_addr,
4239 			&crq->change_mac_addr_rsp.mac_addr[0]);
4240 out:
4241 	complete(&adapter->fw_done);
4242 	return rc;
4243 }
4244 
4245 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4246 				   struct ibmvnic_adapter *adapter)
4247 {
4248 	struct device *dev = &adapter->vdev->dev;
4249 	u64 *req_value;
4250 	char *name;
4251 
4252 	atomic_dec(&adapter->running_cap_crqs);
4253 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4254 	case REQ_TX_QUEUES:
4255 		req_value = &adapter->req_tx_queues;
4256 		name = "tx";
4257 		break;
4258 	case REQ_RX_QUEUES:
4259 		req_value = &adapter->req_rx_queues;
4260 		name = "rx";
4261 		break;
4262 	case REQ_RX_ADD_QUEUES:
4263 		req_value = &adapter->req_rx_add_queues;
4264 		name = "rx_add";
4265 		break;
4266 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4267 		req_value = &adapter->req_tx_entries_per_subcrq;
4268 		name = "tx_entries_per_subcrq";
4269 		break;
4270 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4271 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4272 		name = "rx_add_entries_per_subcrq";
4273 		break;
4274 	case REQ_MTU:
4275 		req_value = &adapter->req_mtu;
4276 		name = "mtu";
4277 		break;
4278 	case PROMISC_REQUESTED:
4279 		req_value = &adapter->promisc;
4280 		name = "promisc";
4281 		break;
4282 	default:
4283 		dev_err(dev, "Got invalid cap request rsp %d\n",
4284 			crq->request_capability.capability);
4285 		return;
4286 	}
4287 
4288 	switch (crq->request_capability_rsp.rc.code) {
4289 	case SUCCESS:
4290 		break;
4291 	case PARTIALSUCCESS:
4292 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4293 			 *req_value,
4294 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4295 					       number), name);
4296 
4297 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4298 		    REQ_MTU) {
4299 			pr_err("mtu of %llu is not supported. Reverting.\n",
4300 			       *req_value);
4301 			*req_value = adapter->fallback.mtu;
4302 		} else {
4303 			*req_value =
4304 				be64_to_cpu(crq->request_capability_rsp.number);
4305 		}
4306 
4307 		send_request_cap(adapter, 1);
4308 		return;
4309 	default:
4310 		dev_err(dev, "Error %d in request cap rsp\n",
4311 			crq->request_capability_rsp.rc.code);
4312 		return;
4313 	}
4314 
4315 	/* Done receiving requested capabilities, query IP offload support */
4316 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4317 		adapter->wait_capability = false;
4318 		send_query_ip_offload(adapter);
4319 	}
4320 }
4321 
4322 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4323 			    struct ibmvnic_adapter *adapter)
4324 {
4325 	struct device *dev = &adapter->vdev->dev;
4326 	struct net_device *netdev = adapter->netdev;
4327 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4328 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4329 	u64 *tx_handle_array;
4330 	u64 *rx_handle_array;
4331 	int num_tx_pools;
4332 	int num_rx_pools;
4333 	u64 *size_array;
4334 	int i;
4335 
4336 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4337 			 DMA_TO_DEVICE);
4338 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4339 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4340 
4341 	/* If the number of queues requested can't be allocated by the
4342 	 * server, the login response will return with code 1. We will need
4343 	 * to resend the login buffer with fewer queues requested.
4344 	 */
4345 	if (login_rsp_crq->generic.rc.code) {
4346 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4347 		complete(&adapter->init_done);
4348 		return 0;
4349 	}
4350 
4351 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4352 
4353 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4354 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4355 		netdev_dbg(adapter->netdev, "%016lx\n",
4356 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4357 	}
4358 
4359 	/* Sanity checks */
4360 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4361 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4362 	     adapter->req_rx_add_queues !=
4363 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4364 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4365 		ibmvnic_remove(adapter->vdev);
4366 		return -EIO;
4367 	}
4368 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4369 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
4370 	/* variable buffer sizes are not supported, so just read the
4371 	 * first entry.
4372 	 */
4373 	adapter->cur_rx_buf_sz = be64_to_cpu(size_array[0]);
4374 
4375 	num_tx_pools = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
4376 	num_rx_pools = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4377 
4378 	tx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4379 				  be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
4380 	rx_handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
4381 				  be32_to_cpu(adapter->login_rsp_buf->off_rxadd_subcrqs));
4382 
4383 	for (i = 0; i < num_tx_pools; i++)
4384 		adapter->tx_scrq[i]->handle = tx_handle_array[i];
4385 
4386 	for (i = 0; i < num_rx_pools; i++)
4387 		adapter->rx_scrq[i]->handle = rx_handle_array[i];
4388 
4389 	adapter->num_active_tx_scrqs = num_tx_pools;
4390 	adapter->num_active_rx_scrqs = num_rx_pools;
4391 	release_login_rsp_buffer(adapter);
4392 	release_login_buffer(adapter);
4393 	complete(&adapter->init_done);
4394 
4395 	return 0;
4396 }
4397 
4398 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4399 				     struct ibmvnic_adapter *adapter)
4400 {
4401 	struct device *dev = &adapter->vdev->dev;
4402 	long rc;
4403 
4404 	rc = crq->request_unmap_rsp.rc.code;
4405 	if (rc)
4406 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4407 }
4408 
4409 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4410 				 struct ibmvnic_adapter *adapter)
4411 {
4412 	struct net_device *netdev = adapter->netdev;
4413 	struct device *dev = &adapter->vdev->dev;
4414 	long rc;
4415 
4416 	rc = crq->query_map_rsp.rc.code;
4417 	if (rc) {
4418 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4419 		return;
4420 	}
4421 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4422 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4423 		   crq->query_map_rsp.free_pages);
4424 }
4425 
4426 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4427 				 struct ibmvnic_adapter *adapter)
4428 {
4429 	struct net_device *netdev = adapter->netdev;
4430 	struct device *dev = &adapter->vdev->dev;
4431 	long rc;
4432 
4433 	atomic_dec(&adapter->running_cap_crqs);
4434 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4435 		   atomic_read(&adapter->running_cap_crqs));
4436 	rc = crq->query_capability.rc.code;
4437 	if (rc) {
4438 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4439 		goto out;
4440 	}
4441 
4442 	switch (be16_to_cpu(crq->query_capability.capability)) {
4443 	case MIN_TX_QUEUES:
4444 		adapter->min_tx_queues =
4445 		    be64_to_cpu(crq->query_capability.number);
4446 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4447 			   adapter->min_tx_queues);
4448 		break;
4449 	case MIN_RX_QUEUES:
4450 		adapter->min_rx_queues =
4451 		    be64_to_cpu(crq->query_capability.number);
4452 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4453 			   adapter->min_rx_queues);
4454 		break;
4455 	case MIN_RX_ADD_QUEUES:
4456 		adapter->min_rx_add_queues =
4457 		    be64_to_cpu(crq->query_capability.number);
4458 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4459 			   adapter->min_rx_add_queues);
4460 		break;
4461 	case MAX_TX_QUEUES:
4462 		adapter->max_tx_queues =
4463 		    be64_to_cpu(crq->query_capability.number);
4464 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4465 			   adapter->max_tx_queues);
4466 		break;
4467 	case MAX_RX_QUEUES:
4468 		adapter->max_rx_queues =
4469 		    be64_to_cpu(crq->query_capability.number);
4470 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4471 			   adapter->max_rx_queues);
4472 		break;
4473 	case MAX_RX_ADD_QUEUES:
4474 		adapter->max_rx_add_queues =
4475 		    be64_to_cpu(crq->query_capability.number);
4476 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4477 			   adapter->max_rx_add_queues);
4478 		break;
4479 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4480 		adapter->min_tx_entries_per_subcrq =
4481 		    be64_to_cpu(crq->query_capability.number);
4482 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4483 			   adapter->min_tx_entries_per_subcrq);
4484 		break;
4485 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4486 		adapter->min_rx_add_entries_per_subcrq =
4487 		    be64_to_cpu(crq->query_capability.number);
4488 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4489 			   adapter->min_rx_add_entries_per_subcrq);
4490 		break;
4491 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4492 		adapter->max_tx_entries_per_subcrq =
4493 		    be64_to_cpu(crq->query_capability.number);
4494 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4495 			   adapter->max_tx_entries_per_subcrq);
4496 		break;
4497 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4498 		adapter->max_rx_add_entries_per_subcrq =
4499 		    be64_to_cpu(crq->query_capability.number);
4500 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4501 			   adapter->max_rx_add_entries_per_subcrq);
4502 		break;
4503 	case TCP_IP_OFFLOAD:
4504 		adapter->tcp_ip_offload =
4505 		    be64_to_cpu(crq->query_capability.number);
4506 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4507 			   adapter->tcp_ip_offload);
4508 		break;
4509 	case PROMISC_SUPPORTED:
4510 		adapter->promisc_supported =
4511 		    be64_to_cpu(crq->query_capability.number);
4512 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4513 			   adapter->promisc_supported);
4514 		break;
4515 	case MIN_MTU:
4516 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4517 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4518 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4519 		break;
4520 	case MAX_MTU:
4521 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4522 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4523 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4524 		break;
4525 	case MAX_MULTICAST_FILTERS:
4526 		adapter->max_multicast_filters =
4527 		    be64_to_cpu(crq->query_capability.number);
4528 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4529 			   adapter->max_multicast_filters);
4530 		break;
4531 	case VLAN_HEADER_INSERTION:
4532 		adapter->vlan_header_insertion =
4533 		    be64_to_cpu(crq->query_capability.number);
4534 		if (adapter->vlan_header_insertion)
4535 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4536 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4537 			   adapter->vlan_header_insertion);
4538 		break;
4539 	case RX_VLAN_HEADER_INSERTION:
4540 		adapter->rx_vlan_header_insertion =
4541 		    be64_to_cpu(crq->query_capability.number);
4542 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4543 			   adapter->rx_vlan_header_insertion);
4544 		break;
4545 	case MAX_TX_SG_ENTRIES:
4546 		adapter->max_tx_sg_entries =
4547 		    be64_to_cpu(crq->query_capability.number);
4548 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4549 			   adapter->max_tx_sg_entries);
4550 		break;
4551 	case RX_SG_SUPPORTED:
4552 		adapter->rx_sg_supported =
4553 		    be64_to_cpu(crq->query_capability.number);
4554 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4555 			   adapter->rx_sg_supported);
4556 		break;
4557 	case OPT_TX_COMP_SUB_QUEUES:
4558 		adapter->opt_tx_comp_sub_queues =
4559 		    be64_to_cpu(crq->query_capability.number);
4560 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4561 			   adapter->opt_tx_comp_sub_queues);
4562 		break;
4563 	case OPT_RX_COMP_QUEUES:
4564 		adapter->opt_rx_comp_queues =
4565 		    be64_to_cpu(crq->query_capability.number);
4566 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4567 			   adapter->opt_rx_comp_queues);
4568 		break;
4569 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4570 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4571 		    be64_to_cpu(crq->query_capability.number);
4572 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4573 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4574 		break;
4575 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4576 		adapter->opt_tx_entries_per_subcrq =
4577 		    be64_to_cpu(crq->query_capability.number);
4578 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4579 			   adapter->opt_tx_entries_per_subcrq);
4580 		break;
4581 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4582 		adapter->opt_rxba_entries_per_subcrq =
4583 		    be64_to_cpu(crq->query_capability.number);
4584 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4585 			   adapter->opt_rxba_entries_per_subcrq);
4586 		break;
4587 	case TX_RX_DESC_REQ:
4588 		adapter->tx_rx_desc_req = crq->query_capability.number;
4589 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4590 			   adapter->tx_rx_desc_req);
4591 		break;
4592 
4593 	default:
4594 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4595 			   crq->query_capability.capability);
4596 	}
4597 
4598 out:
4599 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4600 		adapter->wait_capability = false;
4601 		send_request_cap(adapter, 0);
4602 	}
4603 }
4604 
4605 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4606 {
4607 	union ibmvnic_crq crq;
4608 	int rc;
4609 
4610 	memset(&crq, 0, sizeof(crq));
4611 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4612 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4613 
4614 	mutex_lock(&adapter->fw_lock);
4615 	adapter->fw_done_rc = 0;
4616 	reinit_completion(&adapter->fw_done);
4617 
4618 	rc = ibmvnic_send_crq(adapter, &crq);
4619 	if (rc) {
4620 		mutex_unlock(&adapter->fw_lock);
4621 		return rc;
4622 	}
4623 
4624 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4625 	if (rc) {
4626 		mutex_unlock(&adapter->fw_lock);
4627 		return rc;
4628 	}
4629 
4630 	mutex_unlock(&adapter->fw_lock);
4631 	return adapter->fw_done_rc ? -EIO : 0;
4632 }
4633 
4634 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4635 				       struct ibmvnic_adapter *adapter)
4636 {
4637 	struct net_device *netdev = adapter->netdev;
4638 	int rc;
4639 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4640 
4641 	rc = crq->query_phys_parms_rsp.rc.code;
4642 	if (rc) {
4643 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4644 		return rc;
4645 	}
4646 	switch (rspeed) {
4647 	case IBMVNIC_10MBPS:
4648 		adapter->speed = SPEED_10;
4649 		break;
4650 	case IBMVNIC_100MBPS:
4651 		adapter->speed = SPEED_100;
4652 		break;
4653 	case IBMVNIC_1GBPS:
4654 		adapter->speed = SPEED_1000;
4655 		break;
4656 	case IBMVNIC_10GBPS:
4657 		adapter->speed = SPEED_10000;
4658 		break;
4659 	case IBMVNIC_25GBPS:
4660 		adapter->speed = SPEED_25000;
4661 		break;
4662 	case IBMVNIC_40GBPS:
4663 		adapter->speed = SPEED_40000;
4664 		break;
4665 	case IBMVNIC_50GBPS:
4666 		adapter->speed = SPEED_50000;
4667 		break;
4668 	case IBMVNIC_100GBPS:
4669 		adapter->speed = SPEED_100000;
4670 		break;
4671 	case IBMVNIC_200GBPS:
4672 		adapter->speed = SPEED_200000;
4673 		break;
4674 	default:
4675 		if (netif_carrier_ok(netdev))
4676 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4677 		adapter->speed = SPEED_UNKNOWN;
4678 	}
4679 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4680 		adapter->duplex = DUPLEX_FULL;
4681 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4682 		adapter->duplex = DUPLEX_HALF;
4683 	else
4684 		adapter->duplex = DUPLEX_UNKNOWN;
4685 
4686 	return rc;
4687 }
4688 
4689 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4690 			       struct ibmvnic_adapter *adapter)
4691 {
4692 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4693 	struct net_device *netdev = adapter->netdev;
4694 	struct device *dev = &adapter->vdev->dev;
4695 	u64 *u64_crq = (u64 *)crq;
4696 	long rc;
4697 
4698 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4699 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4700 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4701 	switch (gen_crq->first) {
4702 	case IBMVNIC_CRQ_INIT_RSP:
4703 		switch (gen_crq->cmd) {
4704 		case IBMVNIC_CRQ_INIT:
4705 			dev_info(dev, "Partner initialized\n");
4706 			adapter->from_passive_init = true;
4707 			adapter->failover_pending = false;
4708 			if (!completion_done(&adapter->init_done)) {
4709 				complete(&adapter->init_done);
4710 				adapter->init_done_rc = -EIO;
4711 			}
4712 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4713 			break;
4714 		case IBMVNIC_CRQ_INIT_COMPLETE:
4715 			dev_info(dev, "Partner initialization complete\n");
4716 			adapter->crq.active = true;
4717 			send_version_xchg(adapter);
4718 			break;
4719 		default:
4720 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4721 		}
4722 		return;
4723 	case IBMVNIC_CRQ_XPORT_EVENT:
4724 		netif_carrier_off(netdev);
4725 		adapter->crq.active = false;
4726 		/* terminate any thread waiting for a response
4727 		 * from the device
4728 		 */
4729 		if (!completion_done(&adapter->fw_done)) {
4730 			adapter->fw_done_rc = -EIO;
4731 			complete(&adapter->fw_done);
4732 		}
4733 		if (!completion_done(&adapter->stats_done))
4734 			complete(&adapter->stats_done);
4735 		if (test_bit(0, &adapter->resetting))
4736 			adapter->force_reset_recovery = true;
4737 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4738 			dev_info(dev, "Migrated, re-enabling adapter\n");
4739 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4740 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4741 			dev_info(dev, "Backing device failover detected\n");
4742 			adapter->failover_pending = true;
4743 		} else {
4744 			/* The adapter lost the connection */
4745 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4746 				gen_crq->cmd);
4747 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4748 		}
4749 		return;
4750 	case IBMVNIC_CRQ_CMD_RSP:
4751 		break;
4752 	default:
4753 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4754 			gen_crq->first);
4755 		return;
4756 	}
4757 
4758 	switch (gen_crq->cmd) {
4759 	case VERSION_EXCHANGE_RSP:
4760 		rc = crq->version_exchange_rsp.rc.code;
4761 		if (rc) {
4762 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4763 			break;
4764 		}
4765 		ibmvnic_version =
4766 			    be16_to_cpu(crq->version_exchange_rsp.version);
4767 		dev_info(dev, "Partner protocol version is %d\n",
4768 			 ibmvnic_version);
4769 		send_query_cap(adapter);
4770 		break;
4771 	case QUERY_CAPABILITY_RSP:
4772 		handle_query_cap_rsp(crq, adapter);
4773 		break;
4774 	case QUERY_MAP_RSP:
4775 		handle_query_map_rsp(crq, adapter);
4776 		break;
4777 	case REQUEST_MAP_RSP:
4778 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4779 		complete(&adapter->fw_done);
4780 		break;
4781 	case REQUEST_UNMAP_RSP:
4782 		handle_request_unmap_rsp(crq, adapter);
4783 		break;
4784 	case REQUEST_CAPABILITY_RSP:
4785 		handle_request_cap_rsp(crq, adapter);
4786 		break;
4787 	case LOGIN_RSP:
4788 		netdev_dbg(netdev, "Got Login Response\n");
4789 		handle_login_rsp(crq, adapter);
4790 		break;
4791 	case LOGICAL_LINK_STATE_RSP:
4792 		netdev_dbg(netdev,
4793 			   "Got Logical Link State Response, state: %d rc: %d\n",
4794 			   crq->logical_link_state_rsp.link_state,
4795 			   crq->logical_link_state_rsp.rc.code);
4796 		adapter->logical_link_state =
4797 		    crq->logical_link_state_rsp.link_state;
4798 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4799 		complete(&adapter->init_done);
4800 		break;
4801 	case LINK_STATE_INDICATION:
4802 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4803 		adapter->phys_link_state =
4804 		    crq->link_state_indication.phys_link_state;
4805 		adapter->logical_link_state =
4806 		    crq->link_state_indication.logical_link_state;
4807 		if (adapter->phys_link_state && adapter->logical_link_state)
4808 			netif_carrier_on(netdev);
4809 		else
4810 			netif_carrier_off(netdev);
4811 		break;
4812 	case CHANGE_MAC_ADDR_RSP:
4813 		netdev_dbg(netdev, "Got MAC address change Response\n");
4814 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4815 		break;
4816 	case ERROR_INDICATION:
4817 		netdev_dbg(netdev, "Got Error Indication\n");
4818 		handle_error_indication(crq, adapter);
4819 		break;
4820 	case REQUEST_STATISTICS_RSP:
4821 		netdev_dbg(netdev, "Got Statistics Response\n");
4822 		complete(&adapter->stats_done);
4823 		break;
4824 	case QUERY_IP_OFFLOAD_RSP:
4825 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4826 		handle_query_ip_offload_rsp(adapter);
4827 		break;
4828 	case MULTICAST_CTRL_RSP:
4829 		netdev_dbg(netdev, "Got multicast control Response\n");
4830 		break;
4831 	case CONTROL_IP_OFFLOAD_RSP:
4832 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4833 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4834 				 sizeof(adapter->ip_offload_ctrl),
4835 				 DMA_TO_DEVICE);
4836 		complete(&adapter->init_done);
4837 		break;
4838 	case COLLECT_FW_TRACE_RSP:
4839 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4840 		complete(&adapter->fw_done);
4841 		break;
4842 	case GET_VPD_SIZE_RSP:
4843 		handle_vpd_size_rsp(crq, adapter);
4844 		break;
4845 	case GET_VPD_RSP:
4846 		handle_vpd_rsp(crq, adapter);
4847 		break;
4848 	case QUERY_PHYS_PARMS_RSP:
4849 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4850 		complete(&adapter->fw_done);
4851 		break;
4852 	default:
4853 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4854 			   gen_crq->cmd);
4855 	}
4856 }
4857 
4858 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4859 {
4860 	struct ibmvnic_adapter *adapter = instance;
4861 
4862 	tasklet_schedule(&adapter->tasklet);
4863 	return IRQ_HANDLED;
4864 }
4865 
4866 static void ibmvnic_tasklet(struct tasklet_struct *t)
4867 {
4868 	struct ibmvnic_adapter *adapter = from_tasklet(adapter, t, tasklet);
4869 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4870 	union ibmvnic_crq *crq;
4871 	unsigned long flags;
4872 	bool done = false;
4873 
4874 	spin_lock_irqsave(&queue->lock, flags);
4875 	while (!done) {
4876 		/* Pull all the valid messages off the CRQ */
4877 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4878 			ibmvnic_handle_crq(crq, adapter);
4879 			crq->generic.first = 0;
4880 		}
4881 
4882 		/* remain in tasklet until all
4883 		 * capabilities responses are received
4884 		 */
4885 		if (!adapter->wait_capability)
4886 			done = true;
4887 	}
4888 	/* if capabilities CRQ's were sent in this tasklet, the following
4889 	 * tasklet must wait until all responses are received
4890 	 */
4891 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4892 		adapter->wait_capability = true;
4893 	spin_unlock_irqrestore(&queue->lock, flags);
4894 }
4895 
4896 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4897 {
4898 	struct vio_dev *vdev = adapter->vdev;
4899 	int rc;
4900 
4901 	do {
4902 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4903 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4904 
4905 	if (rc)
4906 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4907 
4908 	return rc;
4909 }
4910 
4911 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4912 {
4913 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4914 	struct device *dev = &adapter->vdev->dev;
4915 	struct vio_dev *vdev = adapter->vdev;
4916 	int rc;
4917 
4918 	/* Close the CRQ */
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 	/* Clean out the queue */
4924 	memset(crq->msgs, 0, PAGE_SIZE);
4925 	crq->cur = 0;
4926 	crq->active = false;
4927 
4928 	/* And re-open it again */
4929 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4930 				crq->msg_token, PAGE_SIZE);
4931 
4932 	if (rc == H_CLOSED)
4933 		/* Adapter is good, but other end is not ready */
4934 		dev_warn(dev, "Partner adapter not ready\n");
4935 	else if (rc != 0)
4936 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4937 
4938 	return rc;
4939 }
4940 
4941 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4942 {
4943 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4944 	struct vio_dev *vdev = adapter->vdev;
4945 	long rc;
4946 
4947 	if (!crq->msgs)
4948 		return;
4949 
4950 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4951 	free_irq(vdev->irq, adapter);
4952 	tasklet_kill(&adapter->tasklet);
4953 	do {
4954 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4955 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4956 
4957 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4958 			 DMA_BIDIRECTIONAL);
4959 	free_page((unsigned long)crq->msgs);
4960 	crq->msgs = NULL;
4961 	crq->active = false;
4962 }
4963 
4964 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4965 {
4966 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4967 	struct device *dev = &adapter->vdev->dev;
4968 	struct vio_dev *vdev = adapter->vdev;
4969 	int rc, retrc = -ENOMEM;
4970 
4971 	if (crq->msgs)
4972 		return 0;
4973 
4974 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4975 	/* Should we allocate more than one page? */
4976 
4977 	if (!crq->msgs)
4978 		return -ENOMEM;
4979 
4980 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4981 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4982 					DMA_BIDIRECTIONAL);
4983 	if (dma_mapping_error(dev, crq->msg_token))
4984 		goto map_failed;
4985 
4986 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4987 				crq->msg_token, PAGE_SIZE);
4988 
4989 	if (rc == H_RESOURCE)
4990 		/* maybe kexecing and resource is busy. try a reset */
4991 		rc = ibmvnic_reset_crq(adapter);
4992 	retrc = rc;
4993 
4994 	if (rc == H_CLOSED) {
4995 		dev_warn(dev, "Partner adapter not ready\n");
4996 	} else if (rc) {
4997 		dev_warn(dev, "Error %d opening adapter\n", rc);
4998 		goto reg_crq_failed;
4999 	}
5000 
5001 	retrc = 0;
5002 
5003 	tasklet_setup(&adapter->tasklet, (void *)ibmvnic_tasklet);
5004 
5005 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
5006 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
5007 		 adapter->vdev->unit_address);
5008 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
5009 	if (rc) {
5010 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
5011 			vdev->irq, rc);
5012 		goto req_irq_failed;
5013 	}
5014 
5015 	rc = vio_enable_interrupts(vdev);
5016 	if (rc) {
5017 		dev_err(dev, "Error %d enabling interrupts\n", rc);
5018 		goto req_irq_failed;
5019 	}
5020 
5021 	crq->cur = 0;
5022 	spin_lock_init(&crq->lock);
5023 
5024 	return retrc;
5025 
5026 req_irq_failed:
5027 	tasklet_kill(&adapter->tasklet);
5028 	do {
5029 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
5030 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
5031 reg_crq_failed:
5032 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
5033 map_failed:
5034 	free_page((unsigned long)crq->msgs);
5035 	crq->msgs = NULL;
5036 	return retrc;
5037 }
5038 
5039 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter, bool reset)
5040 {
5041 	struct device *dev = &adapter->vdev->dev;
5042 	unsigned long timeout = msecs_to_jiffies(30000);
5043 	u64 old_num_rx_queues, old_num_tx_queues;
5044 	int rc;
5045 
5046 	adapter->from_passive_init = false;
5047 
5048 	if (reset) {
5049 		old_num_rx_queues = adapter->req_rx_queues;
5050 		old_num_tx_queues = adapter->req_tx_queues;
5051 		reinit_completion(&adapter->init_done);
5052 	}
5053 
5054 	adapter->init_done_rc = 0;
5055 	rc = ibmvnic_send_crq_init(adapter);
5056 	if (rc) {
5057 		dev_err(dev, "Send crq init failed with error %d\n", rc);
5058 		return rc;
5059 	}
5060 
5061 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5062 		dev_err(dev, "Initialization sequence timed out\n");
5063 		return -1;
5064 	}
5065 
5066 	if (adapter->init_done_rc) {
5067 		release_crq_queue(adapter);
5068 		return adapter->init_done_rc;
5069 	}
5070 
5071 	if (adapter->from_passive_init) {
5072 		adapter->state = VNIC_OPEN;
5073 		adapter->from_passive_init = false;
5074 		return -1;
5075 	}
5076 
5077 	if (reset &&
5078 	    test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5079 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5080 		if (adapter->req_rx_queues != old_num_rx_queues ||
5081 		    adapter->req_tx_queues != old_num_tx_queues) {
5082 			release_sub_crqs(adapter, 0);
5083 			rc = init_sub_crqs(adapter);
5084 		} else {
5085 			rc = reset_sub_crq_queues(adapter);
5086 		}
5087 	} else {
5088 		rc = init_sub_crqs(adapter);
5089 	}
5090 
5091 	if (rc) {
5092 		dev_err(dev, "Initialization of sub crqs failed\n");
5093 		release_crq_queue(adapter);
5094 		return rc;
5095 	}
5096 
5097 	rc = init_sub_crq_irqs(adapter);
5098 	if (rc) {
5099 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5100 		release_crq_queue(adapter);
5101 	}
5102 
5103 	return rc;
5104 }
5105 
5106 static struct device_attribute dev_attr_failover;
5107 
5108 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5109 {
5110 	struct ibmvnic_adapter *adapter;
5111 	struct net_device *netdev;
5112 	unsigned char *mac_addr_p;
5113 	int rc;
5114 
5115 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5116 		dev->unit_address);
5117 
5118 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5119 							VETH_MAC_ADDR, NULL);
5120 	if (!mac_addr_p) {
5121 		dev_err(&dev->dev,
5122 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5123 			__FILE__, __LINE__);
5124 		return 0;
5125 	}
5126 
5127 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5128 				   IBMVNIC_MAX_QUEUES);
5129 	if (!netdev)
5130 		return -ENOMEM;
5131 
5132 	adapter = netdev_priv(netdev);
5133 	adapter->state = VNIC_PROBING;
5134 	dev_set_drvdata(&dev->dev, netdev);
5135 	adapter->vdev = dev;
5136 	adapter->netdev = netdev;
5137 
5138 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5139 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5140 	netdev->irq = dev->irq;
5141 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5142 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5143 	SET_NETDEV_DEV(netdev, &dev->dev);
5144 
5145 	spin_lock_init(&adapter->stats_lock);
5146 
5147 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5148 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5149 			  __ibmvnic_delayed_reset);
5150 	INIT_LIST_HEAD(&adapter->rwi_list);
5151 	spin_lock_init(&adapter->rwi_lock);
5152 	spin_lock_init(&adapter->state_lock);
5153 	mutex_init(&adapter->fw_lock);
5154 	init_completion(&adapter->init_done);
5155 	init_completion(&adapter->fw_done);
5156 	init_completion(&adapter->reset_done);
5157 	init_completion(&adapter->stats_done);
5158 	clear_bit(0, &adapter->resetting);
5159 
5160 	do {
5161 		rc = init_crq_queue(adapter);
5162 		if (rc) {
5163 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5164 				rc);
5165 			goto ibmvnic_init_fail;
5166 		}
5167 
5168 		rc = ibmvnic_reset_init(adapter, false);
5169 		if (rc && rc != EAGAIN)
5170 			goto ibmvnic_init_fail;
5171 	} while (rc == EAGAIN);
5172 
5173 	rc = init_stats_buffers(adapter);
5174 	if (rc)
5175 		goto ibmvnic_init_fail;
5176 
5177 	rc = init_stats_token(adapter);
5178 	if (rc)
5179 		goto ibmvnic_stats_fail;
5180 
5181 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5182 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5183 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5184 
5185 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5186 	if (rc)
5187 		goto ibmvnic_dev_file_err;
5188 
5189 	netif_carrier_off(netdev);
5190 	rc = register_netdev(netdev);
5191 	if (rc) {
5192 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5193 		goto ibmvnic_register_fail;
5194 	}
5195 	dev_info(&dev->dev, "ibmvnic registered\n");
5196 
5197 	adapter->state = VNIC_PROBED;
5198 
5199 	adapter->wait_for_reset = false;
5200 
5201 	return 0;
5202 
5203 ibmvnic_register_fail:
5204 	device_remove_file(&dev->dev, &dev_attr_failover);
5205 
5206 ibmvnic_dev_file_err:
5207 	release_stats_token(adapter);
5208 
5209 ibmvnic_stats_fail:
5210 	release_stats_buffers(adapter);
5211 
5212 ibmvnic_init_fail:
5213 	release_sub_crqs(adapter, 1);
5214 	release_crq_queue(adapter);
5215 	mutex_destroy(&adapter->fw_lock);
5216 	free_netdev(netdev);
5217 
5218 	return rc;
5219 }
5220 
5221 static int ibmvnic_remove(struct vio_dev *dev)
5222 {
5223 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5224 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5225 	unsigned long flags;
5226 
5227 	spin_lock_irqsave(&adapter->state_lock, flags);
5228 	if (adapter->state == VNIC_RESETTING) {
5229 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5230 		return -EBUSY;
5231 	}
5232 
5233 	adapter->state = VNIC_REMOVING;
5234 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5235 
5236 	flush_work(&adapter->ibmvnic_reset);
5237 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5238 
5239 	rtnl_lock();
5240 	unregister_netdevice(netdev);
5241 
5242 	release_resources(adapter);
5243 	release_sub_crqs(adapter, 1);
5244 	release_crq_queue(adapter);
5245 
5246 	release_stats_token(adapter);
5247 	release_stats_buffers(adapter);
5248 
5249 	adapter->state = VNIC_REMOVED;
5250 
5251 	rtnl_unlock();
5252 	mutex_destroy(&adapter->fw_lock);
5253 	device_remove_file(&dev->dev, &dev_attr_failover);
5254 	free_netdev(netdev);
5255 	dev_set_drvdata(&dev->dev, NULL);
5256 
5257 	return 0;
5258 }
5259 
5260 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5261 			      const char *buf, size_t count)
5262 {
5263 	struct net_device *netdev = dev_get_drvdata(dev);
5264 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5265 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5266 	__be64 session_token;
5267 	long rc;
5268 
5269 	if (!sysfs_streq(buf, "1"))
5270 		return -EINVAL;
5271 
5272 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5273 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5274 	if (rc) {
5275 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5276 			   rc);
5277 		return -EINVAL;
5278 	}
5279 
5280 	session_token = (__be64)retbuf[0];
5281 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5282 		   be64_to_cpu(session_token));
5283 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5284 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5285 	if (rc) {
5286 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5287 			   rc);
5288 		return -EINVAL;
5289 	}
5290 
5291 	return count;
5292 }
5293 
5294 static DEVICE_ATTR_WO(failover);
5295 
5296 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5297 {
5298 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5299 	struct ibmvnic_adapter *adapter;
5300 	struct iommu_table *tbl;
5301 	unsigned long ret = 0;
5302 	int i;
5303 
5304 	tbl = get_iommu_table_base(&vdev->dev);
5305 
5306 	/* netdev inits at probe time along with the structures we need below*/
5307 	if (!netdev)
5308 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5309 
5310 	adapter = netdev_priv(netdev);
5311 
5312 	ret += PAGE_SIZE; /* the crq message queue */
5313 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5314 
5315 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5316 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5317 
5318 	for (i = 0; i < adapter->num_active_rx_pools; i++)
5319 		ret += adapter->rx_pool[i].size *
5320 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5321 
5322 	return ret;
5323 }
5324 
5325 static int ibmvnic_resume(struct device *dev)
5326 {
5327 	struct net_device *netdev = dev_get_drvdata(dev);
5328 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5329 
5330 	if (adapter->state != VNIC_OPEN)
5331 		return 0;
5332 
5333 	tasklet_schedule(&adapter->tasklet);
5334 
5335 	return 0;
5336 }
5337 
5338 static const struct vio_device_id ibmvnic_device_table[] = {
5339 	{"network", "IBM,vnic"},
5340 	{"", "" }
5341 };
5342 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5343 
5344 static const struct dev_pm_ops ibmvnic_pm_ops = {
5345 	.resume = ibmvnic_resume
5346 };
5347 
5348 static struct vio_driver ibmvnic_driver = {
5349 	.id_table       = ibmvnic_device_table,
5350 	.probe          = ibmvnic_probe,
5351 	.remove         = ibmvnic_remove,
5352 	.get_desired_dma = ibmvnic_get_desired_dma,
5353 	.name		= ibmvnic_driver_name,
5354 	.pm		= &ibmvnic_pm_ops,
5355 };
5356 
5357 /* module functions */
5358 static int __init ibmvnic_module_init(void)
5359 {
5360 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5361 		IBMVNIC_DRIVER_VERSION);
5362 
5363 	return vio_register_driver(&ibmvnic_driver);
5364 }
5365 
5366 static void __exit ibmvnic_module_exit(void)
5367 {
5368 	vio_unregister_driver(&ibmvnic_driver);
5369 }
5370 
5371 module_init(ibmvnic_module_init);
5372 module_exit(ibmvnic_module_exit);
5373