xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 15e3ae36)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /*                                                                        */
4 /*  IBM System i and System p Virtual NIC Device Driver                   */
5 /*  Copyright (C) 2014 IBM Corp.                                          */
6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
9 /*                                                                        */
10 /*                                                                        */
11 /* This module contains the implementation of a virtual ethernet device   */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
15 /*									   */
16 /* Messages are passed between the VNIC driver and the VNIC server using  */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
20 /* are used by the driver to notify the server that a packet is           */
21 /* ready for transmission or that a buffer has been added to receive a    */
22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
23 /* driver that a packet transmission has been completed or that a packet  */
24 /* has been received and placed in a waiting buffer.                      */
25 /*                                                                        */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
28 /* or receive has been completed, the VNIC driver is required to use      */
29 /* "long term mapping". This entails that large, continuous DMA mapped    */
30 /* buffers are allocated on driver initialization and these buffers are   */
31 /* then continuously reused to pass skbs to and from the VNIC server.     */
32 /*                                                                        */
33 /**************************************************************************/
34 
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69 
70 #include "ibmvnic.h"
71 
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74 
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79 
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static int ibmvnic_remove(struct vio_dev *);
82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
88 		       union sub_crq *sub_crq);
89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
91 static int enable_scrq_irq(struct ibmvnic_adapter *,
92 			   struct ibmvnic_sub_crq_queue *);
93 static int disable_scrq_irq(struct ibmvnic_adapter *,
94 			    struct ibmvnic_sub_crq_queue *);
95 static int pending_scrq(struct ibmvnic_adapter *,
96 			struct ibmvnic_sub_crq_queue *);
97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
98 					struct ibmvnic_sub_crq_queue *);
99 static int ibmvnic_poll(struct napi_struct *napi, int data);
100 static void send_map_query(struct ibmvnic_adapter *adapter);
101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
103 static int send_login(struct ibmvnic_adapter *adapter);
104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
105 static int init_sub_crqs(struct ibmvnic_adapter *);
106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
107 static int ibmvnic_init(struct ibmvnic_adapter *);
108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
109 static void release_crq_queue(struct ibmvnic_adapter *);
110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
113 
114 struct ibmvnic_stat {
115 	char name[ETH_GSTRING_LEN];
116 	int offset;
117 };
118 
119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
120 			     offsetof(struct ibmvnic_statistics, stat))
121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
122 
123 static const struct ibmvnic_stat ibmvnic_stats[] = {
124 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
125 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
126 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
127 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
128 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
129 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
130 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
131 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
132 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
133 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
134 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
135 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
136 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
137 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
138 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
139 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
140 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
141 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
142 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
143 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
144 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
145 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
146 };
147 
148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
149 			  unsigned long length, unsigned long *number,
150 			  unsigned long *irq)
151 {
152 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
153 	long rc;
154 
155 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
156 	*number = retbuf[0];
157 	*irq = retbuf[1];
158 
159 	return rc;
160 }
161 
162 /**
163  * ibmvnic_wait_for_completion - Check device state and wait for completion
164  * @adapter: private device data
165  * @comp_done: completion structure to wait for
166  * @timeout: time to wait in milliseconds
167  *
168  * Wait for a completion signal or until the timeout limit is reached
169  * while checking that the device is still active.
170  */
171 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
172 				       struct completion *comp_done,
173 				       unsigned long timeout)
174 {
175 	struct net_device *netdev;
176 	unsigned long div_timeout;
177 	u8 retry;
178 
179 	netdev = adapter->netdev;
180 	retry = 5;
181 	div_timeout = msecs_to_jiffies(timeout / retry);
182 	while (true) {
183 		if (!adapter->crq.active) {
184 			netdev_err(netdev, "Device down!\n");
185 			return -ENODEV;
186 		}
187 		if (!retry--)
188 			break;
189 		if (wait_for_completion_timeout(comp_done, div_timeout))
190 			return 0;
191 	}
192 	netdev_err(netdev, "Operation timed out.\n");
193 	return -ETIMEDOUT;
194 }
195 
196 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
197 				struct ibmvnic_long_term_buff *ltb, int size)
198 {
199 	struct device *dev = &adapter->vdev->dev;
200 	int rc;
201 
202 	ltb->size = size;
203 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
204 				       GFP_KERNEL);
205 
206 	if (!ltb->buff) {
207 		dev_err(dev, "Couldn't alloc long term buffer\n");
208 		return -ENOMEM;
209 	}
210 	ltb->map_id = adapter->map_id;
211 	adapter->map_id++;
212 
213 	mutex_lock(&adapter->fw_lock);
214 	adapter->fw_done_rc = 0;
215 	reinit_completion(&adapter->fw_done);
216 	rc = send_request_map(adapter, ltb->addr,
217 			      ltb->size, ltb->map_id);
218 	if (rc) {
219 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
220 		mutex_unlock(&adapter->fw_lock);
221 		return rc;
222 	}
223 
224 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
225 	if (rc) {
226 		dev_err(dev,
227 			"Long term map request aborted or timed out,rc = %d\n",
228 			rc);
229 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
230 		mutex_unlock(&adapter->fw_lock);
231 		return rc;
232 	}
233 
234 	if (adapter->fw_done_rc) {
235 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
236 			adapter->fw_done_rc);
237 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
238 		mutex_unlock(&adapter->fw_lock);
239 		return -1;
240 	}
241 	mutex_unlock(&adapter->fw_lock);
242 	return 0;
243 }
244 
245 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
246 				struct ibmvnic_long_term_buff *ltb)
247 {
248 	struct device *dev = &adapter->vdev->dev;
249 
250 	if (!ltb->buff)
251 		return;
252 
253 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
254 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
255 		send_request_unmap(adapter, ltb->map_id);
256 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
257 }
258 
259 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
260 				struct ibmvnic_long_term_buff *ltb)
261 {
262 	struct device *dev = &adapter->vdev->dev;
263 	int rc;
264 
265 	memset(ltb->buff, 0, ltb->size);
266 
267 	mutex_lock(&adapter->fw_lock);
268 	adapter->fw_done_rc = 0;
269 
270 	reinit_completion(&adapter->fw_done);
271 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
272 	if (rc) {
273 		mutex_unlock(&adapter->fw_lock);
274 		return rc;
275 	}
276 
277 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
278 	if (rc) {
279 		dev_info(dev,
280 			 "Reset failed, long term map request timed out or aborted\n");
281 		mutex_unlock(&adapter->fw_lock);
282 		return rc;
283 	}
284 
285 	if (adapter->fw_done_rc) {
286 		dev_info(dev,
287 			 "Reset failed, attempting to free and reallocate buffer\n");
288 		free_long_term_buff(adapter, ltb);
289 		mutex_unlock(&adapter->fw_lock);
290 		return alloc_long_term_buff(adapter, ltb, ltb->size);
291 	}
292 	mutex_unlock(&adapter->fw_lock);
293 	return 0;
294 }
295 
296 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
297 {
298 	int i;
299 
300 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
301 	     i++)
302 		adapter->rx_pool[i].active = 0;
303 }
304 
305 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
306 			      struct ibmvnic_rx_pool *pool)
307 {
308 	int count = pool->size - atomic_read(&pool->available);
309 	struct device *dev = &adapter->vdev->dev;
310 	int buffers_added = 0;
311 	unsigned long lpar_rc;
312 	union sub_crq sub_crq;
313 	struct sk_buff *skb;
314 	unsigned int offset;
315 	dma_addr_t dma_addr;
316 	unsigned char *dst;
317 	u64 *handle_array;
318 	int shift = 0;
319 	int index;
320 	int i;
321 
322 	if (!pool->active)
323 		return;
324 
325 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
326 				      be32_to_cpu(adapter->login_rsp_buf->
327 				      off_rxadd_subcrqs));
328 
329 	for (i = 0; i < count; ++i) {
330 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
331 		if (!skb) {
332 			dev_err(dev, "Couldn't replenish rx buff\n");
333 			adapter->replenish_no_mem++;
334 			break;
335 		}
336 
337 		index = pool->free_map[pool->next_free];
338 
339 		if (pool->rx_buff[index].skb)
340 			dev_err(dev, "Inconsistent free_map!\n");
341 
342 		/* Copy the skb to the long term mapped DMA buffer */
343 		offset = index * pool->buff_size;
344 		dst = pool->long_term_buff.buff + offset;
345 		memset(dst, 0, pool->buff_size);
346 		dma_addr = pool->long_term_buff.addr + offset;
347 		pool->rx_buff[index].data = dst;
348 
349 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
350 		pool->rx_buff[index].dma = dma_addr;
351 		pool->rx_buff[index].skb = skb;
352 		pool->rx_buff[index].pool_index = pool->index;
353 		pool->rx_buff[index].size = pool->buff_size;
354 
355 		memset(&sub_crq, 0, sizeof(sub_crq));
356 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
357 		sub_crq.rx_add.correlator =
358 		    cpu_to_be64((u64)&pool->rx_buff[index]);
359 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
360 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
361 
362 		/* The length field of the sCRQ is defined to be 24 bits so the
363 		 * buffer size needs to be left shifted by a byte before it is
364 		 * converted to big endian to prevent the last byte from being
365 		 * truncated.
366 		 */
367 #ifdef __LITTLE_ENDIAN__
368 		shift = 8;
369 #endif
370 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
371 
372 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
373 				      &sub_crq);
374 		if (lpar_rc != H_SUCCESS)
375 			goto failure;
376 
377 		buffers_added++;
378 		adapter->replenish_add_buff_success++;
379 		pool->next_free = (pool->next_free + 1) % pool->size;
380 	}
381 	atomic_add(buffers_added, &pool->available);
382 	return;
383 
384 failure:
385 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
386 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
387 	pool->free_map[pool->next_free] = index;
388 	pool->rx_buff[index].skb = NULL;
389 
390 	dev_kfree_skb_any(skb);
391 	adapter->replenish_add_buff_failure++;
392 	atomic_add(buffers_added, &pool->available);
393 
394 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
395 		/* Disable buffer pool replenishment and report carrier off if
396 		 * queue is closed or pending failover.
397 		 * Firmware guarantees that a signal will be sent to the
398 		 * driver, triggering a reset.
399 		 */
400 		deactivate_rx_pools(adapter);
401 		netif_carrier_off(adapter->netdev);
402 	}
403 }
404 
405 static void replenish_pools(struct ibmvnic_adapter *adapter)
406 {
407 	int i;
408 
409 	adapter->replenish_task_cycles++;
410 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
411 	     i++) {
412 		if (adapter->rx_pool[i].active)
413 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
414 	}
415 }
416 
417 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
418 {
419 	kfree(adapter->tx_stats_buffers);
420 	kfree(adapter->rx_stats_buffers);
421 	adapter->tx_stats_buffers = NULL;
422 	adapter->rx_stats_buffers = NULL;
423 }
424 
425 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
426 {
427 	adapter->tx_stats_buffers =
428 				kcalloc(IBMVNIC_MAX_QUEUES,
429 					sizeof(struct ibmvnic_tx_queue_stats),
430 					GFP_KERNEL);
431 	if (!adapter->tx_stats_buffers)
432 		return -ENOMEM;
433 
434 	adapter->rx_stats_buffers =
435 				kcalloc(IBMVNIC_MAX_QUEUES,
436 					sizeof(struct ibmvnic_rx_queue_stats),
437 					GFP_KERNEL);
438 	if (!adapter->rx_stats_buffers)
439 		return -ENOMEM;
440 
441 	return 0;
442 }
443 
444 static void release_stats_token(struct ibmvnic_adapter *adapter)
445 {
446 	struct device *dev = &adapter->vdev->dev;
447 
448 	if (!adapter->stats_token)
449 		return;
450 
451 	dma_unmap_single(dev, adapter->stats_token,
452 			 sizeof(struct ibmvnic_statistics),
453 			 DMA_FROM_DEVICE);
454 	adapter->stats_token = 0;
455 }
456 
457 static int init_stats_token(struct ibmvnic_adapter *adapter)
458 {
459 	struct device *dev = &adapter->vdev->dev;
460 	dma_addr_t stok;
461 
462 	stok = dma_map_single(dev, &adapter->stats,
463 			      sizeof(struct ibmvnic_statistics),
464 			      DMA_FROM_DEVICE);
465 	if (dma_mapping_error(dev, stok)) {
466 		dev_err(dev, "Couldn't map stats buffer\n");
467 		return -1;
468 	}
469 
470 	adapter->stats_token = stok;
471 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
472 	return 0;
473 }
474 
475 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
476 {
477 	struct ibmvnic_rx_pool *rx_pool;
478 	int rx_scrqs;
479 	int i, j, rc;
480 	u64 *size_array;
481 
482 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
483 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
484 
485 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
486 	for (i = 0; i < rx_scrqs; i++) {
487 		rx_pool = &adapter->rx_pool[i];
488 
489 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
490 
491 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
492 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
493 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
494 			rc = alloc_long_term_buff(adapter,
495 						  &rx_pool->long_term_buff,
496 						  rx_pool->size *
497 						  rx_pool->buff_size);
498 		} else {
499 			rc = reset_long_term_buff(adapter,
500 						  &rx_pool->long_term_buff);
501 		}
502 
503 		if (rc)
504 			return rc;
505 
506 		for (j = 0; j < rx_pool->size; j++)
507 			rx_pool->free_map[j] = j;
508 
509 		memset(rx_pool->rx_buff, 0,
510 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
511 
512 		atomic_set(&rx_pool->available, 0);
513 		rx_pool->next_alloc = 0;
514 		rx_pool->next_free = 0;
515 		rx_pool->active = 1;
516 	}
517 
518 	return 0;
519 }
520 
521 static void release_rx_pools(struct ibmvnic_adapter *adapter)
522 {
523 	struct ibmvnic_rx_pool *rx_pool;
524 	int i, j;
525 
526 	if (!adapter->rx_pool)
527 		return;
528 
529 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
530 		rx_pool = &adapter->rx_pool[i];
531 
532 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
533 
534 		kfree(rx_pool->free_map);
535 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
536 
537 		if (!rx_pool->rx_buff)
538 			continue;
539 
540 		for (j = 0; j < rx_pool->size; j++) {
541 			if (rx_pool->rx_buff[j].skb) {
542 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
543 				rx_pool->rx_buff[j].skb = NULL;
544 			}
545 		}
546 
547 		kfree(rx_pool->rx_buff);
548 	}
549 
550 	kfree(adapter->rx_pool);
551 	adapter->rx_pool = NULL;
552 	adapter->num_active_rx_pools = 0;
553 }
554 
555 static int init_rx_pools(struct net_device *netdev)
556 {
557 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
558 	struct device *dev = &adapter->vdev->dev;
559 	struct ibmvnic_rx_pool *rx_pool;
560 	int rxadd_subcrqs;
561 	u64 *size_array;
562 	int i, j;
563 
564 	rxadd_subcrqs =
565 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
566 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
567 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
568 
569 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
570 				   sizeof(struct ibmvnic_rx_pool),
571 				   GFP_KERNEL);
572 	if (!adapter->rx_pool) {
573 		dev_err(dev, "Failed to allocate rx pools\n");
574 		return -1;
575 	}
576 
577 	adapter->num_active_rx_pools = rxadd_subcrqs;
578 
579 	for (i = 0; i < rxadd_subcrqs; i++) {
580 		rx_pool = &adapter->rx_pool[i];
581 
582 		netdev_dbg(adapter->netdev,
583 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
584 			   i, adapter->req_rx_add_entries_per_subcrq,
585 			   be64_to_cpu(size_array[i]));
586 
587 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
588 		rx_pool->index = i;
589 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
590 		rx_pool->active = 1;
591 
592 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
593 					    GFP_KERNEL);
594 		if (!rx_pool->free_map) {
595 			release_rx_pools(adapter);
596 			return -1;
597 		}
598 
599 		rx_pool->rx_buff = kcalloc(rx_pool->size,
600 					   sizeof(struct ibmvnic_rx_buff),
601 					   GFP_KERNEL);
602 		if (!rx_pool->rx_buff) {
603 			dev_err(dev, "Couldn't alloc rx buffers\n");
604 			release_rx_pools(adapter);
605 			return -1;
606 		}
607 
608 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
609 					 rx_pool->size * rx_pool->buff_size)) {
610 			release_rx_pools(adapter);
611 			return -1;
612 		}
613 
614 		for (j = 0; j < rx_pool->size; ++j)
615 			rx_pool->free_map[j] = j;
616 
617 		atomic_set(&rx_pool->available, 0);
618 		rx_pool->next_alloc = 0;
619 		rx_pool->next_free = 0;
620 	}
621 
622 	return 0;
623 }
624 
625 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
626 			     struct ibmvnic_tx_pool *tx_pool)
627 {
628 	int rc, i;
629 
630 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
631 	if (rc)
632 		return rc;
633 
634 	memset(tx_pool->tx_buff, 0,
635 	       tx_pool->num_buffers *
636 	       sizeof(struct ibmvnic_tx_buff));
637 
638 	for (i = 0; i < tx_pool->num_buffers; i++)
639 		tx_pool->free_map[i] = i;
640 
641 	tx_pool->consumer_index = 0;
642 	tx_pool->producer_index = 0;
643 
644 	return 0;
645 }
646 
647 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
648 {
649 	int tx_scrqs;
650 	int i, rc;
651 
652 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
653 	for (i = 0; i < tx_scrqs; i++) {
654 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
655 		if (rc)
656 			return rc;
657 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
658 		if (rc)
659 			return rc;
660 	}
661 
662 	return 0;
663 }
664 
665 static void release_vpd_data(struct ibmvnic_adapter *adapter)
666 {
667 	if (!adapter->vpd)
668 		return;
669 
670 	kfree(adapter->vpd->buff);
671 	kfree(adapter->vpd);
672 
673 	adapter->vpd = NULL;
674 }
675 
676 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
677 				struct ibmvnic_tx_pool *tx_pool)
678 {
679 	kfree(tx_pool->tx_buff);
680 	kfree(tx_pool->free_map);
681 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
682 }
683 
684 static void release_tx_pools(struct ibmvnic_adapter *adapter)
685 {
686 	int i;
687 
688 	if (!adapter->tx_pool)
689 		return;
690 
691 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
692 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
693 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
694 	}
695 
696 	kfree(adapter->tx_pool);
697 	adapter->tx_pool = NULL;
698 	kfree(adapter->tso_pool);
699 	adapter->tso_pool = NULL;
700 	adapter->num_active_tx_pools = 0;
701 }
702 
703 static int init_one_tx_pool(struct net_device *netdev,
704 			    struct ibmvnic_tx_pool *tx_pool,
705 			    int num_entries, int buf_size)
706 {
707 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
708 	int i;
709 
710 	tx_pool->tx_buff = kcalloc(num_entries,
711 				   sizeof(struct ibmvnic_tx_buff),
712 				   GFP_KERNEL);
713 	if (!tx_pool->tx_buff)
714 		return -1;
715 
716 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
717 				 num_entries * buf_size))
718 		return -1;
719 
720 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
721 	if (!tx_pool->free_map)
722 		return -1;
723 
724 	for (i = 0; i < num_entries; i++)
725 		tx_pool->free_map[i] = i;
726 
727 	tx_pool->consumer_index = 0;
728 	tx_pool->producer_index = 0;
729 	tx_pool->num_buffers = num_entries;
730 	tx_pool->buf_size = buf_size;
731 
732 	return 0;
733 }
734 
735 static int init_tx_pools(struct net_device *netdev)
736 {
737 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
738 	int tx_subcrqs;
739 	int i, rc;
740 
741 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
742 	adapter->tx_pool = kcalloc(tx_subcrqs,
743 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
744 	if (!adapter->tx_pool)
745 		return -1;
746 
747 	adapter->tso_pool = kcalloc(tx_subcrqs,
748 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
749 	if (!adapter->tso_pool)
750 		return -1;
751 
752 	adapter->num_active_tx_pools = tx_subcrqs;
753 
754 	for (i = 0; i < tx_subcrqs; i++) {
755 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
756 				      adapter->req_tx_entries_per_subcrq,
757 				      adapter->req_mtu + VLAN_HLEN);
758 		if (rc) {
759 			release_tx_pools(adapter);
760 			return rc;
761 		}
762 
763 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
764 				      IBMVNIC_TSO_BUFS,
765 				      IBMVNIC_TSO_BUF_SZ);
766 		if (rc) {
767 			release_tx_pools(adapter);
768 			return rc;
769 		}
770 	}
771 
772 	return 0;
773 }
774 
775 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
776 {
777 	int i;
778 
779 	if (adapter->napi_enabled)
780 		return;
781 
782 	for (i = 0; i < adapter->req_rx_queues; i++)
783 		napi_enable(&adapter->napi[i]);
784 
785 	adapter->napi_enabled = true;
786 }
787 
788 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
789 {
790 	int i;
791 
792 	if (!adapter->napi_enabled)
793 		return;
794 
795 	for (i = 0; i < adapter->req_rx_queues; i++) {
796 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
797 		napi_disable(&adapter->napi[i]);
798 	}
799 
800 	adapter->napi_enabled = false;
801 }
802 
803 static int init_napi(struct ibmvnic_adapter *adapter)
804 {
805 	int i;
806 
807 	adapter->napi = kcalloc(adapter->req_rx_queues,
808 				sizeof(struct napi_struct), GFP_KERNEL);
809 	if (!adapter->napi)
810 		return -ENOMEM;
811 
812 	for (i = 0; i < adapter->req_rx_queues; i++) {
813 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
814 		netif_napi_add(adapter->netdev, &adapter->napi[i],
815 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
816 	}
817 
818 	adapter->num_active_rx_napi = adapter->req_rx_queues;
819 	return 0;
820 }
821 
822 static void release_napi(struct ibmvnic_adapter *adapter)
823 {
824 	int i;
825 
826 	if (!adapter->napi)
827 		return;
828 
829 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
830 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
831 		netif_napi_del(&adapter->napi[i]);
832 	}
833 
834 	kfree(adapter->napi);
835 	adapter->napi = NULL;
836 	adapter->num_active_rx_napi = 0;
837 	adapter->napi_enabled = false;
838 }
839 
840 static int ibmvnic_login(struct net_device *netdev)
841 {
842 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
843 	unsigned long timeout = msecs_to_jiffies(30000);
844 	int retry_count = 0;
845 	bool retry;
846 	int rc;
847 
848 	do {
849 		retry = false;
850 		if (retry_count > IBMVNIC_MAX_QUEUES) {
851 			netdev_warn(netdev, "Login attempts exceeded\n");
852 			return -1;
853 		}
854 
855 		adapter->init_done_rc = 0;
856 		reinit_completion(&adapter->init_done);
857 		rc = send_login(adapter);
858 		if (rc) {
859 			netdev_warn(netdev, "Unable to login\n");
860 			return rc;
861 		}
862 
863 		if (!wait_for_completion_timeout(&adapter->init_done,
864 						 timeout)) {
865 			netdev_warn(netdev, "Login timed out\n");
866 			return -1;
867 		}
868 
869 		if (adapter->init_done_rc == PARTIALSUCCESS) {
870 			retry_count++;
871 			release_sub_crqs(adapter, 1);
872 
873 			retry = true;
874 			netdev_dbg(netdev,
875 				   "Received partial success, retrying...\n");
876 			adapter->init_done_rc = 0;
877 			reinit_completion(&adapter->init_done);
878 			send_cap_queries(adapter);
879 			if (!wait_for_completion_timeout(&adapter->init_done,
880 							 timeout)) {
881 				netdev_warn(netdev,
882 					    "Capabilities query timed out\n");
883 				return -1;
884 			}
885 
886 			rc = init_sub_crqs(adapter);
887 			if (rc) {
888 				netdev_warn(netdev,
889 					    "SCRQ initialization failed\n");
890 				return -1;
891 			}
892 
893 			rc = init_sub_crq_irqs(adapter);
894 			if (rc) {
895 				netdev_warn(netdev,
896 					    "SCRQ irq initialization failed\n");
897 				return -1;
898 			}
899 		} else if (adapter->init_done_rc) {
900 			netdev_warn(netdev, "Adapter login failed\n");
901 			return -1;
902 		}
903 	} while (retry);
904 
905 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
906 
907 	return 0;
908 }
909 
910 static void release_login_buffer(struct ibmvnic_adapter *adapter)
911 {
912 	kfree(adapter->login_buf);
913 	adapter->login_buf = NULL;
914 }
915 
916 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
917 {
918 	kfree(adapter->login_rsp_buf);
919 	adapter->login_rsp_buf = NULL;
920 }
921 
922 static void release_resources(struct ibmvnic_adapter *adapter)
923 {
924 	release_vpd_data(adapter);
925 
926 	release_tx_pools(adapter);
927 	release_rx_pools(adapter);
928 
929 	release_napi(adapter);
930 	release_login_rsp_buffer(adapter);
931 }
932 
933 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
934 {
935 	struct net_device *netdev = adapter->netdev;
936 	unsigned long timeout = msecs_to_jiffies(30000);
937 	union ibmvnic_crq crq;
938 	bool resend;
939 	int rc;
940 
941 	netdev_dbg(netdev, "setting link state %d\n", link_state);
942 
943 	memset(&crq, 0, sizeof(crq));
944 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
945 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
946 	crq.logical_link_state.link_state = link_state;
947 
948 	do {
949 		resend = false;
950 
951 		reinit_completion(&adapter->init_done);
952 		rc = ibmvnic_send_crq(adapter, &crq);
953 		if (rc) {
954 			netdev_err(netdev, "Failed to set link state\n");
955 			return rc;
956 		}
957 
958 		if (!wait_for_completion_timeout(&adapter->init_done,
959 						 timeout)) {
960 			netdev_err(netdev, "timeout setting link state\n");
961 			return -1;
962 		}
963 
964 		if (adapter->init_done_rc == 1) {
965 			/* Partuial success, delay and re-send */
966 			mdelay(1000);
967 			resend = true;
968 		} else if (adapter->init_done_rc) {
969 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
970 				    adapter->init_done_rc);
971 			return adapter->init_done_rc;
972 		}
973 	} while (resend);
974 
975 	return 0;
976 }
977 
978 static int set_real_num_queues(struct net_device *netdev)
979 {
980 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
981 	int rc;
982 
983 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
984 		   adapter->req_tx_queues, adapter->req_rx_queues);
985 
986 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
987 	if (rc) {
988 		netdev_err(netdev, "failed to set the number of tx queues\n");
989 		return rc;
990 	}
991 
992 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
993 	if (rc)
994 		netdev_err(netdev, "failed to set the number of rx queues\n");
995 
996 	return rc;
997 }
998 
999 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1000 {
1001 	struct device *dev = &adapter->vdev->dev;
1002 	union ibmvnic_crq crq;
1003 	int len = 0;
1004 	int rc;
1005 
1006 	if (adapter->vpd->buff)
1007 		len = adapter->vpd->len;
1008 
1009 	mutex_lock(&adapter->fw_lock);
1010 	adapter->fw_done_rc = 0;
1011 	reinit_completion(&adapter->fw_done);
1012 
1013 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1014 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1015 	rc = ibmvnic_send_crq(adapter, &crq);
1016 	if (rc) {
1017 		mutex_unlock(&adapter->fw_lock);
1018 		return rc;
1019 	}
1020 
1021 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1022 	if (rc) {
1023 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1024 		mutex_unlock(&adapter->fw_lock);
1025 		return rc;
1026 	}
1027 	mutex_unlock(&adapter->fw_lock);
1028 
1029 	if (!adapter->vpd->len)
1030 		return -ENODATA;
1031 
1032 	if (!adapter->vpd->buff)
1033 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1034 	else if (adapter->vpd->len != len)
1035 		adapter->vpd->buff =
1036 			krealloc(adapter->vpd->buff,
1037 				 adapter->vpd->len, GFP_KERNEL);
1038 
1039 	if (!adapter->vpd->buff) {
1040 		dev_err(dev, "Could allocate VPD buffer\n");
1041 		return -ENOMEM;
1042 	}
1043 
1044 	adapter->vpd->dma_addr =
1045 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1046 			       DMA_FROM_DEVICE);
1047 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1048 		dev_err(dev, "Could not map VPD buffer\n");
1049 		kfree(adapter->vpd->buff);
1050 		adapter->vpd->buff = NULL;
1051 		return -ENOMEM;
1052 	}
1053 
1054 	mutex_lock(&adapter->fw_lock);
1055 	adapter->fw_done_rc = 0;
1056 	reinit_completion(&adapter->fw_done);
1057 
1058 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1059 	crq.get_vpd.cmd = GET_VPD;
1060 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1061 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1062 	rc = ibmvnic_send_crq(adapter, &crq);
1063 	if (rc) {
1064 		kfree(adapter->vpd->buff);
1065 		adapter->vpd->buff = NULL;
1066 		mutex_unlock(&adapter->fw_lock);
1067 		return rc;
1068 	}
1069 
1070 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1071 	if (rc) {
1072 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1073 		kfree(adapter->vpd->buff);
1074 		adapter->vpd->buff = NULL;
1075 		mutex_unlock(&adapter->fw_lock);
1076 		return rc;
1077 	}
1078 
1079 	mutex_unlock(&adapter->fw_lock);
1080 	return 0;
1081 }
1082 
1083 static int init_resources(struct ibmvnic_adapter *adapter)
1084 {
1085 	struct net_device *netdev = adapter->netdev;
1086 	int rc;
1087 
1088 	rc = set_real_num_queues(netdev);
1089 	if (rc)
1090 		return rc;
1091 
1092 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1093 	if (!adapter->vpd)
1094 		return -ENOMEM;
1095 
1096 	/* Vital Product Data (VPD) */
1097 	rc = ibmvnic_get_vpd(adapter);
1098 	if (rc) {
1099 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1100 		return rc;
1101 	}
1102 
1103 	adapter->map_id = 1;
1104 
1105 	rc = init_napi(adapter);
1106 	if (rc)
1107 		return rc;
1108 
1109 	send_map_query(adapter);
1110 
1111 	rc = init_rx_pools(netdev);
1112 	if (rc)
1113 		return rc;
1114 
1115 	rc = init_tx_pools(netdev);
1116 	return rc;
1117 }
1118 
1119 static int __ibmvnic_open(struct net_device *netdev)
1120 {
1121 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1122 	enum vnic_state prev_state = adapter->state;
1123 	int i, rc;
1124 
1125 	adapter->state = VNIC_OPENING;
1126 	replenish_pools(adapter);
1127 	ibmvnic_napi_enable(adapter);
1128 
1129 	/* We're ready to receive frames, enable the sub-crq interrupts and
1130 	 * set the logical link state to up
1131 	 */
1132 	for (i = 0; i < adapter->req_rx_queues; i++) {
1133 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1134 		if (prev_state == VNIC_CLOSED)
1135 			enable_irq(adapter->rx_scrq[i]->irq);
1136 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1137 	}
1138 
1139 	for (i = 0; i < adapter->req_tx_queues; i++) {
1140 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1141 		if (prev_state == VNIC_CLOSED)
1142 			enable_irq(adapter->tx_scrq[i]->irq);
1143 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1144 	}
1145 
1146 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1147 	if (rc) {
1148 		for (i = 0; i < adapter->req_rx_queues; i++)
1149 			napi_disable(&adapter->napi[i]);
1150 		release_resources(adapter);
1151 		return rc;
1152 	}
1153 
1154 	netif_tx_start_all_queues(netdev);
1155 
1156 	if (prev_state == VNIC_CLOSED) {
1157 		for (i = 0; i < adapter->req_rx_queues; i++)
1158 			napi_schedule(&adapter->napi[i]);
1159 	}
1160 
1161 	adapter->state = VNIC_OPEN;
1162 	return rc;
1163 }
1164 
1165 static int ibmvnic_open(struct net_device *netdev)
1166 {
1167 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1168 	int rc;
1169 
1170 	/* If device failover is pending, just set device state and return.
1171 	 * Device operation will be handled by reset routine.
1172 	 */
1173 	if (adapter->failover_pending) {
1174 		adapter->state = VNIC_OPEN;
1175 		return 0;
1176 	}
1177 
1178 	if (adapter->state != VNIC_CLOSED) {
1179 		rc = ibmvnic_login(netdev);
1180 		if (rc)
1181 			return rc;
1182 
1183 		rc = init_resources(adapter);
1184 		if (rc) {
1185 			netdev_err(netdev, "failed to initialize resources\n");
1186 			release_resources(adapter);
1187 			return rc;
1188 		}
1189 	}
1190 
1191 	rc = __ibmvnic_open(netdev);
1192 
1193 	return rc;
1194 }
1195 
1196 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1197 {
1198 	struct ibmvnic_rx_pool *rx_pool;
1199 	struct ibmvnic_rx_buff *rx_buff;
1200 	u64 rx_entries;
1201 	int rx_scrqs;
1202 	int i, j;
1203 
1204 	if (!adapter->rx_pool)
1205 		return;
1206 
1207 	rx_scrqs = adapter->num_active_rx_pools;
1208 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1209 
1210 	/* Free any remaining skbs in the rx buffer pools */
1211 	for (i = 0; i < rx_scrqs; i++) {
1212 		rx_pool = &adapter->rx_pool[i];
1213 		if (!rx_pool || !rx_pool->rx_buff)
1214 			continue;
1215 
1216 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1217 		for (j = 0; j < rx_entries; j++) {
1218 			rx_buff = &rx_pool->rx_buff[j];
1219 			if (rx_buff && rx_buff->skb) {
1220 				dev_kfree_skb_any(rx_buff->skb);
1221 				rx_buff->skb = NULL;
1222 			}
1223 		}
1224 	}
1225 }
1226 
1227 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1228 			      struct ibmvnic_tx_pool *tx_pool)
1229 {
1230 	struct ibmvnic_tx_buff *tx_buff;
1231 	u64 tx_entries;
1232 	int i;
1233 
1234 	if (!tx_pool || !tx_pool->tx_buff)
1235 		return;
1236 
1237 	tx_entries = tx_pool->num_buffers;
1238 
1239 	for (i = 0; i < tx_entries; i++) {
1240 		tx_buff = &tx_pool->tx_buff[i];
1241 		if (tx_buff && tx_buff->skb) {
1242 			dev_kfree_skb_any(tx_buff->skb);
1243 			tx_buff->skb = NULL;
1244 		}
1245 	}
1246 }
1247 
1248 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1249 {
1250 	int tx_scrqs;
1251 	int i;
1252 
1253 	if (!adapter->tx_pool || !adapter->tso_pool)
1254 		return;
1255 
1256 	tx_scrqs = adapter->num_active_tx_pools;
1257 
1258 	/* Free any remaining skbs in the tx buffer pools */
1259 	for (i = 0; i < tx_scrqs; i++) {
1260 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1261 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1262 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1263 	}
1264 }
1265 
1266 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1267 {
1268 	struct net_device *netdev = adapter->netdev;
1269 	int i;
1270 
1271 	if (adapter->tx_scrq) {
1272 		for (i = 0; i < adapter->req_tx_queues; i++)
1273 			if (adapter->tx_scrq[i]->irq) {
1274 				netdev_dbg(netdev,
1275 					   "Disabling tx_scrq[%d] irq\n", i);
1276 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1277 				disable_irq(adapter->tx_scrq[i]->irq);
1278 			}
1279 	}
1280 
1281 	if (adapter->rx_scrq) {
1282 		for (i = 0; i < adapter->req_rx_queues; i++) {
1283 			if (adapter->rx_scrq[i]->irq) {
1284 				netdev_dbg(netdev,
1285 					   "Disabling rx_scrq[%d] irq\n", i);
1286 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1287 				disable_irq(adapter->rx_scrq[i]->irq);
1288 			}
1289 		}
1290 	}
1291 }
1292 
1293 static void ibmvnic_cleanup(struct net_device *netdev)
1294 {
1295 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1296 
1297 	/* ensure that transmissions are stopped if called by do_reset */
1298 	if (test_bit(0, &adapter->resetting))
1299 		netif_tx_disable(netdev);
1300 	else
1301 		netif_tx_stop_all_queues(netdev);
1302 
1303 	ibmvnic_napi_disable(adapter);
1304 	ibmvnic_disable_irqs(adapter);
1305 
1306 	clean_rx_pools(adapter);
1307 	clean_tx_pools(adapter);
1308 }
1309 
1310 static int __ibmvnic_close(struct net_device *netdev)
1311 {
1312 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1313 	int rc = 0;
1314 
1315 	adapter->state = VNIC_CLOSING;
1316 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1317 	if (rc)
1318 		return rc;
1319 	adapter->state = VNIC_CLOSED;
1320 	return 0;
1321 }
1322 
1323 static int ibmvnic_close(struct net_device *netdev)
1324 {
1325 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1326 	int rc;
1327 
1328 	/* If device failover is pending, just set device state and return.
1329 	 * Device operation will be handled by reset routine.
1330 	 */
1331 	if (adapter->failover_pending) {
1332 		adapter->state = VNIC_CLOSED;
1333 		return 0;
1334 	}
1335 
1336 	rc = __ibmvnic_close(netdev);
1337 	ibmvnic_cleanup(netdev);
1338 
1339 	return rc;
1340 }
1341 
1342 /**
1343  * build_hdr_data - creates L2/L3/L4 header data buffer
1344  * @hdr_field - bitfield determining needed headers
1345  * @skb - socket buffer
1346  * @hdr_len - array of header lengths
1347  * @tot_len - total length of data
1348  *
1349  * Reads hdr_field to determine which headers are needed by firmware.
1350  * Builds a buffer containing these headers.  Saves individual header
1351  * lengths and total buffer length to be used to build descriptors.
1352  */
1353 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1354 			  int *hdr_len, u8 *hdr_data)
1355 {
1356 	int len = 0;
1357 	u8 *hdr;
1358 
1359 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1360 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1361 	else
1362 		hdr_len[0] = sizeof(struct ethhdr);
1363 
1364 	if (skb->protocol == htons(ETH_P_IP)) {
1365 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1366 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1367 			hdr_len[2] = tcp_hdrlen(skb);
1368 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1369 			hdr_len[2] = sizeof(struct udphdr);
1370 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1371 		hdr_len[1] = sizeof(struct ipv6hdr);
1372 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1373 			hdr_len[2] = tcp_hdrlen(skb);
1374 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1375 			hdr_len[2] = sizeof(struct udphdr);
1376 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1377 		hdr_len[1] = arp_hdr_len(skb->dev);
1378 		hdr_len[2] = 0;
1379 	}
1380 
1381 	memset(hdr_data, 0, 120);
1382 	if ((hdr_field >> 6) & 1) {
1383 		hdr = skb_mac_header(skb);
1384 		memcpy(hdr_data, hdr, hdr_len[0]);
1385 		len += hdr_len[0];
1386 	}
1387 
1388 	if ((hdr_field >> 5) & 1) {
1389 		hdr = skb_network_header(skb);
1390 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1391 		len += hdr_len[1];
1392 	}
1393 
1394 	if ((hdr_field >> 4) & 1) {
1395 		hdr = skb_transport_header(skb);
1396 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1397 		len += hdr_len[2];
1398 	}
1399 	return len;
1400 }
1401 
1402 /**
1403  * create_hdr_descs - create header and header extension descriptors
1404  * @hdr_field - bitfield determining needed headers
1405  * @data - buffer containing header data
1406  * @len - length of data buffer
1407  * @hdr_len - array of individual header lengths
1408  * @scrq_arr - descriptor array
1409  *
1410  * Creates header and, if needed, header extension descriptors and
1411  * places them in a descriptor array, scrq_arr
1412  */
1413 
1414 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1415 			    union sub_crq *scrq_arr)
1416 {
1417 	union sub_crq hdr_desc;
1418 	int tmp_len = len;
1419 	int num_descs = 0;
1420 	u8 *data, *cur;
1421 	int tmp;
1422 
1423 	while (tmp_len > 0) {
1424 		cur = hdr_data + len - tmp_len;
1425 
1426 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1427 		if (cur != hdr_data) {
1428 			data = hdr_desc.hdr_ext.data;
1429 			tmp = tmp_len > 29 ? 29 : tmp_len;
1430 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1431 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1432 			hdr_desc.hdr_ext.len = tmp;
1433 		} else {
1434 			data = hdr_desc.hdr.data;
1435 			tmp = tmp_len > 24 ? 24 : tmp_len;
1436 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1437 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1438 			hdr_desc.hdr.len = tmp;
1439 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1440 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1441 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1442 			hdr_desc.hdr.flag = hdr_field << 1;
1443 		}
1444 		memcpy(data, cur, tmp);
1445 		tmp_len -= tmp;
1446 		*scrq_arr = hdr_desc;
1447 		scrq_arr++;
1448 		num_descs++;
1449 	}
1450 
1451 	return num_descs;
1452 }
1453 
1454 /**
1455  * build_hdr_descs_arr - build a header descriptor array
1456  * @skb - socket buffer
1457  * @num_entries - number of descriptors to be sent
1458  * @subcrq - first TX descriptor
1459  * @hdr_field - bit field determining which headers will be sent
1460  *
1461  * This function will build a TX descriptor array with applicable
1462  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1463  */
1464 
1465 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1466 				int *num_entries, u8 hdr_field)
1467 {
1468 	int hdr_len[3] = {0, 0, 0};
1469 	int tot_len;
1470 	u8 *hdr_data = txbuff->hdr_data;
1471 
1472 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1473 				 txbuff->hdr_data);
1474 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1475 			 txbuff->indir_arr + 1);
1476 }
1477 
1478 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1479 				    struct net_device *netdev)
1480 {
1481 	/* For some backing devices, mishandling of small packets
1482 	 * can result in a loss of connection or TX stall. Device
1483 	 * architects recommend that no packet should be smaller
1484 	 * than the minimum MTU value provided to the driver, so
1485 	 * pad any packets to that length
1486 	 */
1487 	if (skb->len < netdev->min_mtu)
1488 		return skb_put_padto(skb, netdev->min_mtu);
1489 
1490 	return 0;
1491 }
1492 
1493 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1494 {
1495 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1496 	int queue_num = skb_get_queue_mapping(skb);
1497 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1498 	struct device *dev = &adapter->vdev->dev;
1499 	struct ibmvnic_tx_buff *tx_buff = NULL;
1500 	struct ibmvnic_sub_crq_queue *tx_scrq;
1501 	struct ibmvnic_tx_pool *tx_pool;
1502 	unsigned int tx_send_failed = 0;
1503 	unsigned int tx_map_failed = 0;
1504 	unsigned int tx_dropped = 0;
1505 	unsigned int tx_packets = 0;
1506 	unsigned int tx_bytes = 0;
1507 	dma_addr_t data_dma_addr;
1508 	struct netdev_queue *txq;
1509 	unsigned long lpar_rc;
1510 	union sub_crq tx_crq;
1511 	unsigned int offset;
1512 	int num_entries = 1;
1513 	unsigned char *dst;
1514 	u64 *handle_array;
1515 	int index = 0;
1516 	u8 proto = 0;
1517 	netdev_tx_t ret = NETDEV_TX_OK;
1518 
1519 	if (test_bit(0, &adapter->resetting)) {
1520 		if (!netif_subqueue_stopped(netdev, skb))
1521 			netif_stop_subqueue(netdev, queue_num);
1522 		dev_kfree_skb_any(skb);
1523 
1524 		tx_send_failed++;
1525 		tx_dropped++;
1526 		ret = NETDEV_TX_OK;
1527 		goto out;
1528 	}
1529 
1530 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1531 		tx_dropped++;
1532 		tx_send_failed++;
1533 		ret = NETDEV_TX_OK;
1534 		goto out;
1535 	}
1536 	if (skb_is_gso(skb))
1537 		tx_pool = &adapter->tso_pool[queue_num];
1538 	else
1539 		tx_pool = &adapter->tx_pool[queue_num];
1540 
1541 	tx_scrq = adapter->tx_scrq[queue_num];
1542 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1543 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1544 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1545 
1546 	index = tx_pool->free_map[tx_pool->consumer_index];
1547 
1548 	if (index == IBMVNIC_INVALID_MAP) {
1549 		dev_kfree_skb_any(skb);
1550 		tx_send_failed++;
1551 		tx_dropped++;
1552 		ret = NETDEV_TX_OK;
1553 		goto out;
1554 	}
1555 
1556 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1557 
1558 	offset = index * tx_pool->buf_size;
1559 	dst = tx_pool->long_term_buff.buff + offset;
1560 	memset(dst, 0, tx_pool->buf_size);
1561 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1562 
1563 	if (skb_shinfo(skb)->nr_frags) {
1564 		int cur, i;
1565 
1566 		/* Copy the head */
1567 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1568 		cur = skb_headlen(skb);
1569 
1570 		/* Copy the frags */
1571 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1572 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1573 
1574 			memcpy(dst + cur,
1575 			       page_address(skb_frag_page(frag)) +
1576 			       skb_frag_off(frag), skb_frag_size(frag));
1577 			cur += skb_frag_size(frag);
1578 		}
1579 	} else {
1580 		skb_copy_from_linear_data(skb, dst, skb->len);
1581 	}
1582 
1583 	tx_pool->consumer_index =
1584 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1585 
1586 	tx_buff = &tx_pool->tx_buff[index];
1587 	tx_buff->skb = skb;
1588 	tx_buff->data_dma[0] = data_dma_addr;
1589 	tx_buff->data_len[0] = skb->len;
1590 	tx_buff->index = index;
1591 	tx_buff->pool_index = queue_num;
1592 	tx_buff->last_frag = true;
1593 
1594 	memset(&tx_crq, 0, sizeof(tx_crq));
1595 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1596 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1597 	tx_crq.v1.n_crq_elem = 1;
1598 	tx_crq.v1.n_sge = 1;
1599 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1600 
1601 	if (skb_is_gso(skb))
1602 		tx_crq.v1.correlator =
1603 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1604 	else
1605 		tx_crq.v1.correlator = cpu_to_be32(index);
1606 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1607 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1608 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1609 
1610 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1611 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1612 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1613 	}
1614 
1615 	if (skb->protocol == htons(ETH_P_IP)) {
1616 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1617 		proto = ip_hdr(skb)->protocol;
1618 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1619 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1620 		proto = ipv6_hdr(skb)->nexthdr;
1621 	}
1622 
1623 	if (proto == IPPROTO_TCP)
1624 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1625 	else if (proto == IPPROTO_UDP)
1626 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1627 
1628 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1629 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1630 		hdrs += 2;
1631 	}
1632 	if (skb_is_gso(skb)) {
1633 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1634 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1635 		hdrs += 2;
1636 	}
1637 	/* determine if l2/3/4 headers are sent to firmware */
1638 	if ((*hdrs >> 7) & 1) {
1639 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1640 		tx_crq.v1.n_crq_elem = num_entries;
1641 		tx_buff->num_entries = num_entries;
1642 		tx_buff->indir_arr[0] = tx_crq;
1643 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1644 						    sizeof(tx_buff->indir_arr),
1645 						    DMA_TO_DEVICE);
1646 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1647 			dev_kfree_skb_any(skb);
1648 			tx_buff->skb = NULL;
1649 			if (!firmware_has_feature(FW_FEATURE_CMO))
1650 				dev_err(dev, "tx: unable to map descriptor array\n");
1651 			tx_map_failed++;
1652 			tx_dropped++;
1653 			ret = NETDEV_TX_OK;
1654 			goto tx_err_out;
1655 		}
1656 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1657 					       (u64)tx_buff->indir_dma,
1658 					       (u64)num_entries);
1659 		dma_unmap_single(dev, tx_buff->indir_dma,
1660 				 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1661 	} else {
1662 		tx_buff->num_entries = num_entries;
1663 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1664 				      &tx_crq);
1665 	}
1666 	if (lpar_rc != H_SUCCESS) {
1667 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1668 			dev_err_ratelimited(dev, "tx: send failed\n");
1669 		dev_kfree_skb_any(skb);
1670 		tx_buff->skb = NULL;
1671 
1672 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1673 			/* Disable TX and report carrier off if queue is closed
1674 			 * or pending failover.
1675 			 * Firmware guarantees that a signal will be sent to the
1676 			 * driver, triggering a reset or some other action.
1677 			 */
1678 			netif_tx_stop_all_queues(netdev);
1679 			netif_carrier_off(netdev);
1680 		}
1681 
1682 		tx_send_failed++;
1683 		tx_dropped++;
1684 		ret = NETDEV_TX_OK;
1685 		goto tx_err_out;
1686 	}
1687 
1688 	if (atomic_add_return(num_entries, &tx_scrq->used)
1689 					>= adapter->req_tx_entries_per_subcrq) {
1690 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1691 		netif_stop_subqueue(netdev, queue_num);
1692 	}
1693 
1694 	tx_packets++;
1695 	tx_bytes += skb->len;
1696 	txq->trans_start = jiffies;
1697 	ret = NETDEV_TX_OK;
1698 	goto out;
1699 
1700 tx_err_out:
1701 	/* roll back consumer index and map array*/
1702 	if (tx_pool->consumer_index == 0)
1703 		tx_pool->consumer_index =
1704 			tx_pool->num_buffers - 1;
1705 	else
1706 		tx_pool->consumer_index--;
1707 	tx_pool->free_map[tx_pool->consumer_index] = index;
1708 out:
1709 	netdev->stats.tx_dropped += tx_dropped;
1710 	netdev->stats.tx_bytes += tx_bytes;
1711 	netdev->stats.tx_packets += tx_packets;
1712 	adapter->tx_send_failed += tx_send_failed;
1713 	adapter->tx_map_failed += tx_map_failed;
1714 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1715 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1716 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1717 
1718 	return ret;
1719 }
1720 
1721 static void ibmvnic_set_multi(struct net_device *netdev)
1722 {
1723 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1724 	struct netdev_hw_addr *ha;
1725 	union ibmvnic_crq crq;
1726 
1727 	memset(&crq, 0, sizeof(crq));
1728 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1729 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1730 
1731 	if (netdev->flags & IFF_PROMISC) {
1732 		if (!adapter->promisc_supported)
1733 			return;
1734 	} else {
1735 		if (netdev->flags & IFF_ALLMULTI) {
1736 			/* Accept all multicast */
1737 			memset(&crq, 0, sizeof(crq));
1738 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1739 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1740 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1741 			ibmvnic_send_crq(adapter, &crq);
1742 		} else if (netdev_mc_empty(netdev)) {
1743 			/* Reject all multicast */
1744 			memset(&crq, 0, sizeof(crq));
1745 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1746 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1747 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1748 			ibmvnic_send_crq(adapter, &crq);
1749 		} else {
1750 			/* Accept one or more multicast(s) */
1751 			netdev_for_each_mc_addr(ha, netdev) {
1752 				memset(&crq, 0, sizeof(crq));
1753 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1754 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1755 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1756 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1757 						ha->addr);
1758 				ibmvnic_send_crq(adapter, &crq);
1759 			}
1760 		}
1761 	}
1762 }
1763 
1764 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1765 {
1766 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1767 	union ibmvnic_crq crq;
1768 	int rc;
1769 
1770 	if (!is_valid_ether_addr(dev_addr)) {
1771 		rc = -EADDRNOTAVAIL;
1772 		goto err;
1773 	}
1774 
1775 	memset(&crq, 0, sizeof(crq));
1776 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1777 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1778 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1779 
1780 	mutex_lock(&adapter->fw_lock);
1781 	adapter->fw_done_rc = 0;
1782 	reinit_completion(&adapter->fw_done);
1783 
1784 	rc = ibmvnic_send_crq(adapter, &crq);
1785 	if (rc) {
1786 		rc = -EIO;
1787 		mutex_unlock(&adapter->fw_lock);
1788 		goto err;
1789 	}
1790 
1791 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1792 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1793 	if (rc || adapter->fw_done_rc) {
1794 		rc = -EIO;
1795 		mutex_unlock(&adapter->fw_lock);
1796 		goto err;
1797 	}
1798 	mutex_unlock(&adapter->fw_lock);
1799 	return 0;
1800 err:
1801 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1802 	return rc;
1803 }
1804 
1805 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1806 {
1807 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1808 	struct sockaddr *addr = p;
1809 	int rc;
1810 
1811 	rc = 0;
1812 	ether_addr_copy(adapter->mac_addr, addr->sa_data);
1813 	if (adapter->state != VNIC_PROBED)
1814 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1815 
1816 	return rc;
1817 }
1818 
1819 /**
1820  * do_change_param_reset returns zero if we are able to keep processing reset
1821  * events, or non-zero if we hit a fatal error and must halt.
1822  */
1823 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1824 				 struct ibmvnic_rwi *rwi,
1825 				 u32 reset_state)
1826 {
1827 	struct net_device *netdev = adapter->netdev;
1828 	int i, rc;
1829 
1830 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1831 		   rwi->reset_reason);
1832 
1833 	netif_carrier_off(netdev);
1834 	adapter->reset_reason = rwi->reset_reason;
1835 
1836 	ibmvnic_cleanup(netdev);
1837 
1838 	if (reset_state == VNIC_OPEN) {
1839 		rc = __ibmvnic_close(netdev);
1840 		if (rc)
1841 			return rc;
1842 	}
1843 
1844 	release_resources(adapter);
1845 	release_sub_crqs(adapter, 1);
1846 	release_crq_queue(adapter);
1847 
1848 	adapter->state = VNIC_PROBED;
1849 
1850 	rc = init_crq_queue(adapter);
1851 
1852 	if (rc) {
1853 		netdev_err(adapter->netdev,
1854 			   "Couldn't initialize crq. rc=%d\n", rc);
1855 		return rc;
1856 	}
1857 
1858 	rc = ibmvnic_reset_init(adapter);
1859 	if (rc)
1860 		return IBMVNIC_INIT_FAILED;
1861 
1862 	/* If the adapter was in PROBE state prior to the reset,
1863 	 * exit here.
1864 	 */
1865 	if (reset_state == VNIC_PROBED)
1866 		return 0;
1867 
1868 	rc = ibmvnic_login(netdev);
1869 	if (rc) {
1870 		adapter->state = reset_state;
1871 		return rc;
1872 	}
1873 
1874 	rc = init_resources(adapter);
1875 	if (rc)
1876 		return rc;
1877 
1878 	ibmvnic_disable_irqs(adapter);
1879 
1880 	adapter->state = VNIC_CLOSED;
1881 
1882 	if (reset_state == VNIC_CLOSED)
1883 		return 0;
1884 
1885 	rc = __ibmvnic_open(netdev);
1886 	if (rc)
1887 		return IBMVNIC_OPEN_FAILED;
1888 
1889 	/* refresh device's multicast list */
1890 	ibmvnic_set_multi(netdev);
1891 
1892 	/* kick napi */
1893 	for (i = 0; i < adapter->req_rx_queues; i++)
1894 		napi_schedule(&adapter->napi[i]);
1895 
1896 	return 0;
1897 }
1898 
1899 /**
1900  * do_reset returns zero if we are able to keep processing reset events, or
1901  * non-zero if we hit a fatal error and must halt.
1902  */
1903 static int do_reset(struct ibmvnic_adapter *adapter,
1904 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1905 {
1906 	u64 old_num_rx_queues, old_num_tx_queues;
1907 	u64 old_num_rx_slots, old_num_tx_slots;
1908 	struct net_device *netdev = adapter->netdev;
1909 	int i, rc;
1910 
1911 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1912 		   rwi->reset_reason);
1913 
1914 	rtnl_lock();
1915 
1916 	netif_carrier_off(netdev);
1917 	adapter->reset_reason = rwi->reset_reason;
1918 
1919 	old_num_rx_queues = adapter->req_rx_queues;
1920 	old_num_tx_queues = adapter->req_tx_queues;
1921 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1922 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1923 
1924 	ibmvnic_cleanup(netdev);
1925 
1926 	if (reset_state == VNIC_OPEN &&
1927 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
1928 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1929 		adapter->state = VNIC_CLOSING;
1930 
1931 		/* Release the RTNL lock before link state change and
1932 		 * re-acquire after the link state change to allow
1933 		 * linkwatch_event to grab the RTNL lock and run during
1934 		 * a reset.
1935 		 */
1936 		rtnl_unlock();
1937 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1938 		rtnl_lock();
1939 		if (rc)
1940 			goto out;
1941 
1942 		if (adapter->state != VNIC_CLOSING) {
1943 			rc = -1;
1944 			goto out;
1945 		}
1946 
1947 		adapter->state = VNIC_CLOSED;
1948 	}
1949 
1950 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1951 		/* remove the closed state so when we call open it appears
1952 		 * we are coming from the probed state.
1953 		 */
1954 		adapter->state = VNIC_PROBED;
1955 
1956 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1957 			rc = ibmvnic_reenable_crq_queue(adapter);
1958 			release_sub_crqs(adapter, 1);
1959 		} else {
1960 			rc = ibmvnic_reset_crq(adapter);
1961 			if (!rc)
1962 				rc = vio_enable_interrupts(adapter->vdev);
1963 		}
1964 
1965 		if (rc) {
1966 			netdev_err(adapter->netdev,
1967 				   "Couldn't initialize crq. rc=%d\n", rc);
1968 			goto out;
1969 		}
1970 
1971 		rc = ibmvnic_reset_init(adapter);
1972 		if (rc) {
1973 			rc = IBMVNIC_INIT_FAILED;
1974 			goto out;
1975 		}
1976 
1977 		/* If the adapter was in PROBE state prior to the reset,
1978 		 * exit here.
1979 		 */
1980 		if (reset_state == VNIC_PROBED) {
1981 			rc = 0;
1982 			goto out;
1983 		}
1984 
1985 		rc = ibmvnic_login(netdev);
1986 		if (rc) {
1987 			adapter->state = reset_state;
1988 			goto out;
1989 		}
1990 
1991 		if (adapter->req_rx_queues != old_num_rx_queues ||
1992 		    adapter->req_tx_queues != old_num_tx_queues ||
1993 		    adapter->req_rx_add_entries_per_subcrq !=
1994 		    old_num_rx_slots ||
1995 		    adapter->req_tx_entries_per_subcrq !=
1996 		    old_num_tx_slots) {
1997 			release_rx_pools(adapter);
1998 			release_tx_pools(adapter);
1999 			release_napi(adapter);
2000 			release_vpd_data(adapter);
2001 
2002 			rc = init_resources(adapter);
2003 			if (rc)
2004 				goto out;
2005 
2006 		} else {
2007 			rc = reset_tx_pools(adapter);
2008 			if (rc)
2009 				goto out;
2010 
2011 			rc = reset_rx_pools(adapter);
2012 			if (rc)
2013 				goto out;
2014 		}
2015 		ibmvnic_disable_irqs(adapter);
2016 	}
2017 	adapter->state = VNIC_CLOSED;
2018 
2019 	if (reset_state == VNIC_CLOSED) {
2020 		rc = 0;
2021 		goto out;
2022 	}
2023 
2024 	rc = __ibmvnic_open(netdev);
2025 	if (rc) {
2026 		rc = IBMVNIC_OPEN_FAILED;
2027 		goto out;
2028 	}
2029 
2030 	/* refresh device's multicast list */
2031 	ibmvnic_set_multi(netdev);
2032 
2033 	/* kick napi */
2034 	for (i = 0; i < adapter->req_rx_queues; i++)
2035 		napi_schedule(&adapter->napi[i]);
2036 
2037 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2038 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2039 
2040 	rc = 0;
2041 
2042 out:
2043 	rtnl_unlock();
2044 
2045 	return rc;
2046 }
2047 
2048 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2049 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2050 {
2051 	struct net_device *netdev = adapter->netdev;
2052 	int rc;
2053 
2054 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2055 		   rwi->reset_reason);
2056 
2057 	netif_carrier_off(netdev);
2058 	adapter->reset_reason = rwi->reset_reason;
2059 
2060 	ibmvnic_cleanup(netdev);
2061 	release_resources(adapter);
2062 	release_sub_crqs(adapter, 0);
2063 	release_crq_queue(adapter);
2064 
2065 	/* remove the closed state so when we call open it appears
2066 	 * we are coming from the probed state.
2067 	 */
2068 	adapter->state = VNIC_PROBED;
2069 
2070 	reinit_completion(&adapter->init_done);
2071 	rc = init_crq_queue(adapter);
2072 	if (rc) {
2073 		netdev_err(adapter->netdev,
2074 			   "Couldn't initialize crq. rc=%d\n", rc);
2075 		return rc;
2076 	}
2077 
2078 	rc = ibmvnic_init(adapter);
2079 	if (rc)
2080 		return rc;
2081 
2082 	/* If the adapter was in PROBE state prior to the reset,
2083 	 * exit here.
2084 	 */
2085 	if (reset_state == VNIC_PROBED)
2086 		return 0;
2087 
2088 	rc = ibmvnic_login(netdev);
2089 	if (rc) {
2090 		adapter->state = VNIC_PROBED;
2091 		return 0;
2092 	}
2093 
2094 	rc = init_resources(adapter);
2095 	if (rc)
2096 		return rc;
2097 
2098 	ibmvnic_disable_irqs(adapter);
2099 	adapter->state = VNIC_CLOSED;
2100 
2101 	if (reset_state == VNIC_CLOSED)
2102 		return 0;
2103 
2104 	rc = __ibmvnic_open(netdev);
2105 	if (rc)
2106 		return IBMVNIC_OPEN_FAILED;
2107 
2108 	return 0;
2109 }
2110 
2111 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2112 {
2113 	struct ibmvnic_rwi *rwi;
2114 	unsigned long flags;
2115 
2116 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2117 
2118 	if (!list_empty(&adapter->rwi_list)) {
2119 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2120 				       list);
2121 		list_del(&rwi->list);
2122 	} else {
2123 		rwi = NULL;
2124 	}
2125 
2126 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2127 	return rwi;
2128 }
2129 
2130 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2131 {
2132 	struct ibmvnic_rwi *rwi;
2133 
2134 	rwi = get_next_rwi(adapter);
2135 	while (rwi) {
2136 		kfree(rwi);
2137 		rwi = get_next_rwi(adapter);
2138 	}
2139 }
2140 
2141 static void __ibmvnic_reset(struct work_struct *work)
2142 {
2143 	struct ibmvnic_rwi *rwi;
2144 	struct ibmvnic_adapter *adapter;
2145 	bool saved_state = false;
2146 	unsigned long flags;
2147 	u32 reset_state;
2148 	int rc = 0;
2149 
2150 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2151 
2152 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2153 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2154 				      IBMVNIC_RESET_DELAY);
2155 		return;
2156 	}
2157 
2158 	rwi = get_next_rwi(adapter);
2159 	while (rwi) {
2160 		spin_lock_irqsave(&adapter->state_lock, flags);
2161 
2162 		if (adapter->state == VNIC_REMOVING ||
2163 		    adapter->state == VNIC_REMOVED) {
2164 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2165 			kfree(rwi);
2166 			rc = EBUSY;
2167 			break;
2168 		}
2169 
2170 		if (!saved_state) {
2171 			reset_state = adapter->state;
2172 			adapter->state = VNIC_RESETTING;
2173 			saved_state = true;
2174 		}
2175 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2176 
2177 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2178 			/* CHANGE_PARAM requestor holds rtnl_lock */
2179 			rc = do_change_param_reset(adapter, rwi, reset_state);
2180 		} else if (adapter->force_reset_recovery) {
2181 			/* Transport event occurred during previous reset */
2182 			if (adapter->wait_for_reset) {
2183 				/* Previous was CHANGE_PARAM; caller locked */
2184 				adapter->force_reset_recovery = false;
2185 				rc = do_hard_reset(adapter, rwi, reset_state);
2186 			} else {
2187 				rtnl_lock();
2188 				adapter->force_reset_recovery = false;
2189 				rc = do_hard_reset(adapter, rwi, reset_state);
2190 				rtnl_unlock();
2191 			}
2192 		} else {
2193 			rc = do_reset(adapter, rwi, reset_state);
2194 		}
2195 		kfree(rwi);
2196 		if (rc == IBMVNIC_OPEN_FAILED) {
2197 			if (list_empty(&adapter->rwi_list))
2198 				adapter->state = VNIC_CLOSED;
2199 			else
2200 				adapter->state = reset_state;
2201 			rc = 0;
2202 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2203 		    !adapter->force_reset_recovery)
2204 			break;
2205 
2206 		rwi = get_next_rwi(adapter);
2207 
2208 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2209 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2210 			adapter->force_reset_recovery = true;
2211 	}
2212 
2213 	if (adapter->wait_for_reset) {
2214 		adapter->reset_done_rc = rc;
2215 		complete(&adapter->reset_done);
2216 	}
2217 
2218 	if (rc) {
2219 		netdev_dbg(adapter->netdev, "Reset failed\n");
2220 		free_all_rwi(adapter);
2221 	}
2222 
2223 	clear_bit_unlock(0, &adapter->resetting);
2224 }
2225 
2226 static void __ibmvnic_delayed_reset(struct work_struct *work)
2227 {
2228 	struct ibmvnic_adapter *adapter;
2229 
2230 	adapter = container_of(work, struct ibmvnic_adapter,
2231 			       ibmvnic_delayed_reset.work);
2232 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2233 }
2234 
2235 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2236 			 enum ibmvnic_reset_reason reason)
2237 {
2238 	struct list_head *entry, *tmp_entry;
2239 	struct ibmvnic_rwi *rwi, *tmp;
2240 	struct net_device *netdev = adapter->netdev;
2241 	unsigned long flags;
2242 	int ret;
2243 
2244 	if (adapter->state == VNIC_REMOVING ||
2245 	    adapter->state == VNIC_REMOVED ||
2246 	    adapter->failover_pending) {
2247 		ret = EBUSY;
2248 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2249 		goto err;
2250 	}
2251 
2252 	if (adapter->state == VNIC_PROBING) {
2253 		netdev_warn(netdev, "Adapter reset during probe\n");
2254 		ret = adapter->init_done_rc = EAGAIN;
2255 		goto err;
2256 	}
2257 
2258 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2259 
2260 	list_for_each(entry, &adapter->rwi_list) {
2261 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2262 		if (tmp->reset_reason == reason) {
2263 			netdev_dbg(netdev, "Skipping matching reset\n");
2264 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2265 			ret = EBUSY;
2266 			goto err;
2267 		}
2268 	}
2269 
2270 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2271 	if (!rwi) {
2272 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2273 		ibmvnic_close(netdev);
2274 		ret = ENOMEM;
2275 		goto err;
2276 	}
2277 	/* if we just received a transport event,
2278 	 * flush reset queue and process this reset
2279 	 */
2280 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2281 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2282 			list_del(entry);
2283 	}
2284 	rwi->reset_reason = reason;
2285 	list_add_tail(&rwi->list, &adapter->rwi_list);
2286 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2287 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2288 	schedule_work(&adapter->ibmvnic_reset);
2289 
2290 	return 0;
2291 err:
2292 	return -ret;
2293 }
2294 
2295 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2296 {
2297 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2298 
2299 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2300 }
2301 
2302 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2303 				  struct ibmvnic_rx_buff *rx_buff)
2304 {
2305 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2306 
2307 	rx_buff->skb = NULL;
2308 
2309 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2310 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2311 
2312 	atomic_dec(&pool->available);
2313 }
2314 
2315 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2316 {
2317 	struct net_device *netdev = napi->dev;
2318 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2319 	int scrq_num = (int)(napi - adapter->napi);
2320 	int frames_processed = 0;
2321 
2322 restart_poll:
2323 	while (frames_processed < budget) {
2324 		struct sk_buff *skb;
2325 		struct ibmvnic_rx_buff *rx_buff;
2326 		union sub_crq *next;
2327 		u32 length;
2328 		u16 offset;
2329 		u8 flags = 0;
2330 
2331 		if (unlikely(test_bit(0, &adapter->resetting) &&
2332 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2333 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2334 			napi_complete_done(napi, frames_processed);
2335 			return frames_processed;
2336 		}
2337 
2338 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2339 			break;
2340 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2341 		rx_buff =
2342 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2343 							  rx_comp.correlator);
2344 		/* do error checking */
2345 		if (next->rx_comp.rc) {
2346 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2347 				   be16_to_cpu(next->rx_comp.rc));
2348 			/* free the entry */
2349 			next->rx_comp.first = 0;
2350 			dev_kfree_skb_any(rx_buff->skb);
2351 			remove_buff_from_pool(adapter, rx_buff);
2352 			continue;
2353 		} else if (!rx_buff->skb) {
2354 			/* free the entry */
2355 			next->rx_comp.first = 0;
2356 			remove_buff_from_pool(adapter, rx_buff);
2357 			continue;
2358 		}
2359 
2360 		length = be32_to_cpu(next->rx_comp.len);
2361 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2362 		flags = next->rx_comp.flags;
2363 		skb = rx_buff->skb;
2364 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2365 					length);
2366 
2367 		/* VLAN Header has been stripped by the system firmware and
2368 		 * needs to be inserted by the driver
2369 		 */
2370 		if (adapter->rx_vlan_header_insertion &&
2371 		    (flags & IBMVNIC_VLAN_STRIPPED))
2372 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2373 					       ntohs(next->rx_comp.vlan_tci));
2374 
2375 		/* free the entry */
2376 		next->rx_comp.first = 0;
2377 		remove_buff_from_pool(adapter, rx_buff);
2378 
2379 		skb_put(skb, length);
2380 		skb->protocol = eth_type_trans(skb, netdev);
2381 		skb_record_rx_queue(skb, scrq_num);
2382 
2383 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2384 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2385 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2386 		}
2387 
2388 		length = skb->len;
2389 		napi_gro_receive(napi, skb); /* send it up */
2390 		netdev->stats.rx_packets++;
2391 		netdev->stats.rx_bytes += length;
2392 		adapter->rx_stats_buffers[scrq_num].packets++;
2393 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2394 		frames_processed++;
2395 	}
2396 
2397 	if (adapter->state != VNIC_CLOSING)
2398 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2399 
2400 	if (frames_processed < budget) {
2401 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2402 		napi_complete_done(napi, frames_processed);
2403 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2404 		    napi_reschedule(napi)) {
2405 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2406 			goto restart_poll;
2407 		}
2408 	}
2409 	return frames_processed;
2410 }
2411 
2412 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2413 {
2414 	int rc, ret;
2415 
2416 	adapter->fallback.mtu = adapter->req_mtu;
2417 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2418 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2419 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2420 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2421 
2422 	reinit_completion(&adapter->reset_done);
2423 	adapter->wait_for_reset = true;
2424 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2425 
2426 	if (rc) {
2427 		ret = rc;
2428 		goto out;
2429 	}
2430 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2431 	if (rc) {
2432 		ret = -ENODEV;
2433 		goto out;
2434 	}
2435 
2436 	ret = 0;
2437 	if (adapter->reset_done_rc) {
2438 		ret = -EIO;
2439 		adapter->desired.mtu = adapter->fallback.mtu;
2440 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2441 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2442 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2443 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2444 
2445 		reinit_completion(&adapter->reset_done);
2446 		adapter->wait_for_reset = true;
2447 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2448 		if (rc) {
2449 			ret = rc;
2450 			goto out;
2451 		}
2452 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2453 						 60000);
2454 		if (rc) {
2455 			ret = -ENODEV;
2456 			goto out;
2457 		}
2458 	}
2459 out:
2460 	adapter->wait_for_reset = false;
2461 
2462 	return ret;
2463 }
2464 
2465 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2466 {
2467 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2468 
2469 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2470 
2471 	return wait_for_reset(adapter);
2472 }
2473 
2474 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2475 						struct net_device *dev,
2476 						netdev_features_t features)
2477 {
2478 	/* Some backing hardware adapters can not
2479 	 * handle packets with a MSS less than 224
2480 	 * or with only one segment.
2481 	 */
2482 	if (skb_is_gso(skb)) {
2483 		if (skb_shinfo(skb)->gso_size < 224 ||
2484 		    skb_shinfo(skb)->gso_segs == 1)
2485 			features &= ~NETIF_F_GSO_MASK;
2486 	}
2487 
2488 	return features;
2489 }
2490 
2491 static const struct net_device_ops ibmvnic_netdev_ops = {
2492 	.ndo_open		= ibmvnic_open,
2493 	.ndo_stop		= ibmvnic_close,
2494 	.ndo_start_xmit		= ibmvnic_xmit,
2495 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2496 	.ndo_set_mac_address	= ibmvnic_set_mac,
2497 	.ndo_validate_addr	= eth_validate_addr,
2498 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2499 	.ndo_change_mtu		= ibmvnic_change_mtu,
2500 	.ndo_features_check     = ibmvnic_features_check,
2501 };
2502 
2503 /* ethtool functions */
2504 
2505 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2506 				      struct ethtool_link_ksettings *cmd)
2507 {
2508 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2509 	int rc;
2510 
2511 	rc = send_query_phys_parms(adapter);
2512 	if (rc) {
2513 		adapter->speed = SPEED_UNKNOWN;
2514 		adapter->duplex = DUPLEX_UNKNOWN;
2515 	}
2516 	cmd->base.speed = adapter->speed;
2517 	cmd->base.duplex = adapter->duplex;
2518 	cmd->base.port = PORT_FIBRE;
2519 	cmd->base.phy_address = 0;
2520 	cmd->base.autoneg = AUTONEG_ENABLE;
2521 
2522 	return 0;
2523 }
2524 
2525 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2526 				struct ethtool_drvinfo *info)
2527 {
2528 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2529 
2530 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2531 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2532 	strlcpy(info->fw_version, adapter->fw_version,
2533 		sizeof(info->fw_version));
2534 }
2535 
2536 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2537 {
2538 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2539 
2540 	return adapter->msg_enable;
2541 }
2542 
2543 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2544 {
2545 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2546 
2547 	adapter->msg_enable = data;
2548 }
2549 
2550 static u32 ibmvnic_get_link(struct net_device *netdev)
2551 {
2552 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2553 
2554 	/* Don't need to send a query because we request a logical link up at
2555 	 * init and then we wait for link state indications
2556 	 */
2557 	return adapter->logical_link_state;
2558 }
2559 
2560 static void ibmvnic_get_ringparam(struct net_device *netdev,
2561 				  struct ethtool_ringparam *ring)
2562 {
2563 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2564 
2565 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2566 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2567 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2568 	} else {
2569 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2570 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2571 	}
2572 	ring->rx_mini_max_pending = 0;
2573 	ring->rx_jumbo_max_pending = 0;
2574 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2575 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2576 	ring->rx_mini_pending = 0;
2577 	ring->rx_jumbo_pending = 0;
2578 }
2579 
2580 static int ibmvnic_set_ringparam(struct net_device *netdev,
2581 				 struct ethtool_ringparam *ring)
2582 {
2583 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2584 	int ret;
2585 
2586 	ret = 0;
2587 	adapter->desired.rx_entries = ring->rx_pending;
2588 	adapter->desired.tx_entries = ring->tx_pending;
2589 
2590 	ret = wait_for_reset(adapter);
2591 
2592 	if (!ret &&
2593 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2594 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2595 		netdev_info(netdev,
2596 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2597 			    ring->rx_pending, ring->tx_pending,
2598 			    adapter->req_rx_add_entries_per_subcrq,
2599 			    adapter->req_tx_entries_per_subcrq);
2600 	return ret;
2601 }
2602 
2603 static void ibmvnic_get_channels(struct net_device *netdev,
2604 				 struct ethtool_channels *channels)
2605 {
2606 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2607 
2608 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2609 		channels->max_rx = adapter->max_rx_queues;
2610 		channels->max_tx = adapter->max_tx_queues;
2611 	} else {
2612 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2613 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2614 	}
2615 
2616 	channels->max_other = 0;
2617 	channels->max_combined = 0;
2618 	channels->rx_count = adapter->req_rx_queues;
2619 	channels->tx_count = adapter->req_tx_queues;
2620 	channels->other_count = 0;
2621 	channels->combined_count = 0;
2622 }
2623 
2624 static int ibmvnic_set_channels(struct net_device *netdev,
2625 				struct ethtool_channels *channels)
2626 {
2627 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2628 	int ret;
2629 
2630 	ret = 0;
2631 	adapter->desired.rx_queues = channels->rx_count;
2632 	adapter->desired.tx_queues = channels->tx_count;
2633 
2634 	ret = wait_for_reset(adapter);
2635 
2636 	if (!ret &&
2637 	    (adapter->req_rx_queues != channels->rx_count ||
2638 	     adapter->req_tx_queues != channels->tx_count))
2639 		netdev_info(netdev,
2640 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2641 			    channels->rx_count, channels->tx_count,
2642 			    adapter->req_rx_queues, adapter->req_tx_queues);
2643 	return ret;
2644 
2645 }
2646 
2647 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2648 {
2649 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2650 	int i;
2651 
2652 	switch (stringset) {
2653 	case ETH_SS_STATS:
2654 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2655 				i++, data += ETH_GSTRING_LEN)
2656 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2657 
2658 		for (i = 0; i < adapter->req_tx_queues; i++) {
2659 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2660 			data += ETH_GSTRING_LEN;
2661 
2662 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2663 			data += ETH_GSTRING_LEN;
2664 
2665 			snprintf(data, ETH_GSTRING_LEN,
2666 				 "tx%d_dropped_packets", i);
2667 			data += ETH_GSTRING_LEN;
2668 		}
2669 
2670 		for (i = 0; i < adapter->req_rx_queues; i++) {
2671 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2672 			data += ETH_GSTRING_LEN;
2673 
2674 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2675 			data += ETH_GSTRING_LEN;
2676 
2677 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2678 			data += ETH_GSTRING_LEN;
2679 		}
2680 		break;
2681 
2682 	case ETH_SS_PRIV_FLAGS:
2683 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2684 			strcpy(data + i * ETH_GSTRING_LEN,
2685 			       ibmvnic_priv_flags[i]);
2686 		break;
2687 	default:
2688 		return;
2689 	}
2690 }
2691 
2692 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2693 {
2694 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2695 
2696 	switch (sset) {
2697 	case ETH_SS_STATS:
2698 		return ARRAY_SIZE(ibmvnic_stats) +
2699 		       adapter->req_tx_queues * NUM_TX_STATS +
2700 		       adapter->req_rx_queues * NUM_RX_STATS;
2701 	case ETH_SS_PRIV_FLAGS:
2702 		return ARRAY_SIZE(ibmvnic_priv_flags);
2703 	default:
2704 		return -EOPNOTSUPP;
2705 	}
2706 }
2707 
2708 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2709 				      struct ethtool_stats *stats, u64 *data)
2710 {
2711 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2712 	union ibmvnic_crq crq;
2713 	int i, j;
2714 	int rc;
2715 
2716 	memset(&crq, 0, sizeof(crq));
2717 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2718 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2719 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2720 	crq.request_statistics.len =
2721 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2722 
2723 	/* Wait for data to be written */
2724 	reinit_completion(&adapter->stats_done);
2725 	rc = ibmvnic_send_crq(adapter, &crq);
2726 	if (rc)
2727 		return;
2728 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2729 	if (rc)
2730 		return;
2731 
2732 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2733 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2734 						ibmvnic_stats[i].offset));
2735 
2736 	for (j = 0; j < adapter->req_tx_queues; j++) {
2737 		data[i] = adapter->tx_stats_buffers[j].packets;
2738 		i++;
2739 		data[i] = adapter->tx_stats_buffers[j].bytes;
2740 		i++;
2741 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2742 		i++;
2743 	}
2744 
2745 	for (j = 0; j < adapter->req_rx_queues; j++) {
2746 		data[i] = adapter->rx_stats_buffers[j].packets;
2747 		i++;
2748 		data[i] = adapter->rx_stats_buffers[j].bytes;
2749 		i++;
2750 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2751 		i++;
2752 	}
2753 }
2754 
2755 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2756 {
2757 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2758 
2759 	return adapter->priv_flags;
2760 }
2761 
2762 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2763 {
2764 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2765 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2766 
2767 	if (which_maxes)
2768 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2769 	else
2770 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2771 
2772 	return 0;
2773 }
2774 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2775 	.get_drvinfo		= ibmvnic_get_drvinfo,
2776 	.get_msglevel		= ibmvnic_get_msglevel,
2777 	.set_msglevel		= ibmvnic_set_msglevel,
2778 	.get_link		= ibmvnic_get_link,
2779 	.get_ringparam		= ibmvnic_get_ringparam,
2780 	.set_ringparam		= ibmvnic_set_ringparam,
2781 	.get_channels		= ibmvnic_get_channels,
2782 	.set_channels		= ibmvnic_set_channels,
2783 	.get_strings            = ibmvnic_get_strings,
2784 	.get_sset_count         = ibmvnic_get_sset_count,
2785 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2786 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2787 	.get_priv_flags		= ibmvnic_get_priv_flags,
2788 	.set_priv_flags		= ibmvnic_set_priv_flags,
2789 };
2790 
2791 /* Routines for managing CRQs/sCRQs  */
2792 
2793 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2794 				   struct ibmvnic_sub_crq_queue *scrq)
2795 {
2796 	int rc;
2797 
2798 	if (scrq->irq) {
2799 		free_irq(scrq->irq, scrq);
2800 		irq_dispose_mapping(scrq->irq);
2801 		scrq->irq = 0;
2802 	}
2803 
2804 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2805 	atomic_set(&scrq->used, 0);
2806 	scrq->cur = 0;
2807 
2808 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2809 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2810 	return rc;
2811 }
2812 
2813 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2814 {
2815 	int i, rc;
2816 
2817 	for (i = 0; i < adapter->req_tx_queues; i++) {
2818 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2819 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2820 		if (rc)
2821 			return rc;
2822 	}
2823 
2824 	for (i = 0; i < adapter->req_rx_queues; i++) {
2825 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2826 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2827 		if (rc)
2828 			return rc;
2829 	}
2830 
2831 	return rc;
2832 }
2833 
2834 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2835 				  struct ibmvnic_sub_crq_queue *scrq,
2836 				  bool do_h_free)
2837 {
2838 	struct device *dev = &adapter->vdev->dev;
2839 	long rc;
2840 
2841 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2842 
2843 	if (do_h_free) {
2844 		/* Close the sub-crqs */
2845 		do {
2846 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2847 						adapter->vdev->unit_address,
2848 						scrq->crq_num);
2849 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2850 
2851 		if (rc) {
2852 			netdev_err(adapter->netdev,
2853 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2854 				   scrq->crq_num, rc);
2855 		}
2856 	}
2857 
2858 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2859 			 DMA_BIDIRECTIONAL);
2860 	free_pages((unsigned long)scrq->msgs, 2);
2861 	kfree(scrq);
2862 }
2863 
2864 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2865 							*adapter)
2866 {
2867 	struct device *dev = &adapter->vdev->dev;
2868 	struct ibmvnic_sub_crq_queue *scrq;
2869 	int rc;
2870 
2871 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2872 	if (!scrq)
2873 		return NULL;
2874 
2875 	scrq->msgs =
2876 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2877 	if (!scrq->msgs) {
2878 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2879 		goto zero_page_failed;
2880 	}
2881 
2882 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2883 					 DMA_BIDIRECTIONAL);
2884 	if (dma_mapping_error(dev, scrq->msg_token)) {
2885 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2886 		goto map_failed;
2887 	}
2888 
2889 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2890 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2891 
2892 	if (rc == H_RESOURCE)
2893 		rc = ibmvnic_reset_crq(adapter);
2894 
2895 	if (rc == H_CLOSED) {
2896 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2897 	} else if (rc) {
2898 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2899 		goto reg_failed;
2900 	}
2901 
2902 	scrq->adapter = adapter;
2903 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2904 	spin_lock_init(&scrq->lock);
2905 
2906 	netdev_dbg(adapter->netdev,
2907 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2908 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2909 
2910 	return scrq;
2911 
2912 reg_failed:
2913 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2914 			 DMA_BIDIRECTIONAL);
2915 map_failed:
2916 	free_pages((unsigned long)scrq->msgs, 2);
2917 zero_page_failed:
2918 	kfree(scrq);
2919 
2920 	return NULL;
2921 }
2922 
2923 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2924 {
2925 	int i;
2926 
2927 	if (adapter->tx_scrq) {
2928 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2929 			if (!adapter->tx_scrq[i])
2930 				continue;
2931 
2932 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2933 				   i);
2934 			if (adapter->tx_scrq[i]->irq) {
2935 				free_irq(adapter->tx_scrq[i]->irq,
2936 					 adapter->tx_scrq[i]);
2937 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2938 				adapter->tx_scrq[i]->irq = 0;
2939 			}
2940 
2941 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2942 					      do_h_free);
2943 		}
2944 
2945 		kfree(adapter->tx_scrq);
2946 		adapter->tx_scrq = NULL;
2947 		adapter->num_active_tx_scrqs = 0;
2948 	}
2949 
2950 	if (adapter->rx_scrq) {
2951 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2952 			if (!adapter->rx_scrq[i])
2953 				continue;
2954 
2955 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2956 				   i);
2957 			if (adapter->rx_scrq[i]->irq) {
2958 				free_irq(adapter->rx_scrq[i]->irq,
2959 					 adapter->rx_scrq[i]);
2960 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2961 				adapter->rx_scrq[i]->irq = 0;
2962 			}
2963 
2964 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2965 					      do_h_free);
2966 		}
2967 
2968 		kfree(adapter->rx_scrq);
2969 		adapter->rx_scrq = NULL;
2970 		adapter->num_active_rx_scrqs = 0;
2971 	}
2972 }
2973 
2974 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2975 			    struct ibmvnic_sub_crq_queue *scrq)
2976 {
2977 	struct device *dev = &adapter->vdev->dev;
2978 	unsigned long rc;
2979 
2980 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2981 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2982 	if (rc)
2983 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2984 			scrq->hw_irq, rc);
2985 	return rc;
2986 }
2987 
2988 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2989 			   struct ibmvnic_sub_crq_queue *scrq)
2990 {
2991 	struct device *dev = &adapter->vdev->dev;
2992 	unsigned long rc;
2993 
2994 	if (scrq->hw_irq > 0x100000000ULL) {
2995 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2996 		return 1;
2997 	}
2998 
2999 	if (test_bit(0, &adapter->resetting) &&
3000 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3001 		u64 val = (0xff000000) | scrq->hw_irq;
3002 
3003 		rc = plpar_hcall_norets(H_EOI, val);
3004 		/* H_EOI would fail with rc = H_FUNCTION when running
3005 		 * in XIVE mode which is expected, but not an error.
3006 		 */
3007 		if (rc && (rc != H_FUNCTION))
3008 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3009 				val, rc);
3010 	}
3011 
3012 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3013 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3014 	if (rc)
3015 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3016 			scrq->hw_irq, rc);
3017 	return rc;
3018 }
3019 
3020 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3021 			       struct ibmvnic_sub_crq_queue *scrq)
3022 {
3023 	struct device *dev = &adapter->vdev->dev;
3024 	struct ibmvnic_tx_pool *tx_pool;
3025 	struct ibmvnic_tx_buff *txbuff;
3026 	union sub_crq *next;
3027 	int index;
3028 	int i, j;
3029 
3030 restart_loop:
3031 	while (pending_scrq(adapter, scrq)) {
3032 		unsigned int pool = scrq->pool_index;
3033 		int num_entries = 0;
3034 
3035 		next = ibmvnic_next_scrq(adapter, scrq);
3036 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3037 			if (next->tx_comp.rcs[i]) {
3038 				dev_err(dev, "tx error %x\n",
3039 					next->tx_comp.rcs[i]);
3040 				continue;
3041 			}
3042 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3043 			if (index & IBMVNIC_TSO_POOL_MASK) {
3044 				tx_pool = &adapter->tso_pool[pool];
3045 				index &= ~IBMVNIC_TSO_POOL_MASK;
3046 			} else {
3047 				tx_pool = &adapter->tx_pool[pool];
3048 			}
3049 
3050 			txbuff = &tx_pool->tx_buff[index];
3051 
3052 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3053 				if (!txbuff->data_dma[j])
3054 					continue;
3055 
3056 				txbuff->data_dma[j] = 0;
3057 			}
3058 
3059 			if (txbuff->last_frag) {
3060 				dev_kfree_skb_any(txbuff->skb);
3061 				txbuff->skb = NULL;
3062 			}
3063 
3064 			num_entries += txbuff->num_entries;
3065 
3066 			tx_pool->free_map[tx_pool->producer_index] = index;
3067 			tx_pool->producer_index =
3068 				(tx_pool->producer_index + 1) %
3069 					tx_pool->num_buffers;
3070 		}
3071 		/* remove tx_comp scrq*/
3072 		next->tx_comp.first = 0;
3073 
3074 		if (atomic_sub_return(num_entries, &scrq->used) <=
3075 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3076 		    __netif_subqueue_stopped(adapter->netdev,
3077 					     scrq->pool_index)) {
3078 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3079 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3080 				   scrq->pool_index);
3081 		}
3082 	}
3083 
3084 	enable_scrq_irq(adapter, scrq);
3085 
3086 	if (pending_scrq(adapter, scrq)) {
3087 		disable_scrq_irq(adapter, scrq);
3088 		goto restart_loop;
3089 	}
3090 
3091 	return 0;
3092 }
3093 
3094 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3095 {
3096 	struct ibmvnic_sub_crq_queue *scrq = instance;
3097 	struct ibmvnic_adapter *adapter = scrq->adapter;
3098 
3099 	disable_scrq_irq(adapter, scrq);
3100 	ibmvnic_complete_tx(adapter, scrq);
3101 
3102 	return IRQ_HANDLED;
3103 }
3104 
3105 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3106 {
3107 	struct ibmvnic_sub_crq_queue *scrq = instance;
3108 	struct ibmvnic_adapter *adapter = scrq->adapter;
3109 
3110 	/* When booting a kdump kernel we can hit pending interrupts
3111 	 * prior to completing driver initialization.
3112 	 */
3113 	if (unlikely(adapter->state != VNIC_OPEN))
3114 		return IRQ_NONE;
3115 
3116 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3117 
3118 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3119 		disable_scrq_irq(adapter, scrq);
3120 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3121 	}
3122 
3123 	return IRQ_HANDLED;
3124 }
3125 
3126 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3127 {
3128 	struct device *dev = &adapter->vdev->dev;
3129 	struct ibmvnic_sub_crq_queue *scrq;
3130 	int i = 0, j = 0;
3131 	int rc = 0;
3132 
3133 	for (i = 0; i < adapter->req_tx_queues; i++) {
3134 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3135 			   i);
3136 		scrq = adapter->tx_scrq[i];
3137 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3138 
3139 		if (!scrq->irq) {
3140 			rc = -EINVAL;
3141 			dev_err(dev, "Error mapping irq\n");
3142 			goto req_tx_irq_failed;
3143 		}
3144 
3145 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3146 			 adapter->vdev->unit_address, i);
3147 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3148 				 0, scrq->name, scrq);
3149 
3150 		if (rc) {
3151 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3152 				scrq->irq, rc);
3153 			irq_dispose_mapping(scrq->irq);
3154 			goto req_tx_irq_failed;
3155 		}
3156 	}
3157 
3158 	for (i = 0; i < adapter->req_rx_queues; i++) {
3159 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3160 			   i);
3161 		scrq = adapter->rx_scrq[i];
3162 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3163 		if (!scrq->irq) {
3164 			rc = -EINVAL;
3165 			dev_err(dev, "Error mapping irq\n");
3166 			goto req_rx_irq_failed;
3167 		}
3168 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3169 			 adapter->vdev->unit_address, i);
3170 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3171 				 0, scrq->name, scrq);
3172 		if (rc) {
3173 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3174 				scrq->irq, rc);
3175 			irq_dispose_mapping(scrq->irq);
3176 			goto req_rx_irq_failed;
3177 		}
3178 	}
3179 	return rc;
3180 
3181 req_rx_irq_failed:
3182 	for (j = 0; j < i; j++) {
3183 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3184 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3185 	}
3186 	i = adapter->req_tx_queues;
3187 req_tx_irq_failed:
3188 	for (j = 0; j < i; j++) {
3189 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3190 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3191 	}
3192 	release_sub_crqs(adapter, 1);
3193 	return rc;
3194 }
3195 
3196 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3197 {
3198 	struct device *dev = &adapter->vdev->dev;
3199 	struct ibmvnic_sub_crq_queue **allqueues;
3200 	int registered_queues = 0;
3201 	int total_queues;
3202 	int more = 0;
3203 	int i;
3204 
3205 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3206 
3207 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3208 	if (!allqueues)
3209 		return -1;
3210 
3211 	for (i = 0; i < total_queues; i++) {
3212 		allqueues[i] = init_sub_crq_queue(adapter);
3213 		if (!allqueues[i]) {
3214 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3215 			break;
3216 		}
3217 		registered_queues++;
3218 	}
3219 
3220 	/* Make sure we were able to register the minimum number of queues */
3221 	if (registered_queues <
3222 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3223 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3224 		goto tx_failed;
3225 	}
3226 
3227 	/* Distribute the failed allocated queues*/
3228 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3229 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3230 		switch (i % 3) {
3231 		case 0:
3232 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3233 				adapter->req_rx_queues--;
3234 			else
3235 				more++;
3236 			break;
3237 		case 1:
3238 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3239 				adapter->req_tx_queues--;
3240 			else
3241 				more++;
3242 			break;
3243 		}
3244 	}
3245 
3246 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3247 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3248 	if (!adapter->tx_scrq)
3249 		goto tx_failed;
3250 
3251 	for (i = 0; i < adapter->req_tx_queues; i++) {
3252 		adapter->tx_scrq[i] = allqueues[i];
3253 		adapter->tx_scrq[i]->pool_index = i;
3254 		adapter->num_active_tx_scrqs++;
3255 	}
3256 
3257 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3258 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3259 	if (!adapter->rx_scrq)
3260 		goto rx_failed;
3261 
3262 	for (i = 0; i < adapter->req_rx_queues; i++) {
3263 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3264 		adapter->rx_scrq[i]->scrq_num = i;
3265 		adapter->num_active_rx_scrqs++;
3266 	}
3267 
3268 	kfree(allqueues);
3269 	return 0;
3270 
3271 rx_failed:
3272 	kfree(adapter->tx_scrq);
3273 	adapter->tx_scrq = NULL;
3274 tx_failed:
3275 	for (i = 0; i < registered_queues; i++)
3276 		release_sub_crq_queue(adapter, allqueues[i], 1);
3277 	kfree(allqueues);
3278 	return -1;
3279 }
3280 
3281 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3282 {
3283 	struct device *dev = &adapter->vdev->dev;
3284 	union ibmvnic_crq crq;
3285 	int max_entries;
3286 
3287 	if (!retry) {
3288 		/* Sub-CRQ entries are 32 byte long */
3289 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3290 
3291 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3292 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3293 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3294 			return;
3295 		}
3296 
3297 		if (adapter->desired.mtu)
3298 			adapter->req_mtu = adapter->desired.mtu;
3299 		else
3300 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3301 
3302 		if (!adapter->desired.tx_entries)
3303 			adapter->desired.tx_entries =
3304 					adapter->max_tx_entries_per_subcrq;
3305 		if (!adapter->desired.rx_entries)
3306 			adapter->desired.rx_entries =
3307 					adapter->max_rx_add_entries_per_subcrq;
3308 
3309 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3310 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3311 
3312 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3313 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3314 			adapter->desired.tx_entries = max_entries;
3315 		}
3316 
3317 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3318 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3319 			adapter->desired.rx_entries = max_entries;
3320 		}
3321 
3322 		if (adapter->desired.tx_entries)
3323 			adapter->req_tx_entries_per_subcrq =
3324 					adapter->desired.tx_entries;
3325 		else
3326 			adapter->req_tx_entries_per_subcrq =
3327 					adapter->max_tx_entries_per_subcrq;
3328 
3329 		if (adapter->desired.rx_entries)
3330 			adapter->req_rx_add_entries_per_subcrq =
3331 					adapter->desired.rx_entries;
3332 		else
3333 			adapter->req_rx_add_entries_per_subcrq =
3334 					adapter->max_rx_add_entries_per_subcrq;
3335 
3336 		if (adapter->desired.tx_queues)
3337 			adapter->req_tx_queues =
3338 					adapter->desired.tx_queues;
3339 		else
3340 			adapter->req_tx_queues =
3341 					adapter->opt_tx_comp_sub_queues;
3342 
3343 		if (adapter->desired.rx_queues)
3344 			adapter->req_rx_queues =
3345 					adapter->desired.rx_queues;
3346 		else
3347 			adapter->req_rx_queues =
3348 					adapter->opt_rx_comp_queues;
3349 
3350 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3351 	}
3352 
3353 	memset(&crq, 0, sizeof(crq));
3354 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3355 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3356 
3357 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3358 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3359 	atomic_inc(&adapter->running_cap_crqs);
3360 	ibmvnic_send_crq(adapter, &crq);
3361 
3362 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3363 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3364 	atomic_inc(&adapter->running_cap_crqs);
3365 	ibmvnic_send_crq(adapter, &crq);
3366 
3367 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3368 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3369 	atomic_inc(&adapter->running_cap_crqs);
3370 	ibmvnic_send_crq(adapter, &crq);
3371 
3372 	crq.request_capability.capability =
3373 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3374 	crq.request_capability.number =
3375 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3376 	atomic_inc(&adapter->running_cap_crqs);
3377 	ibmvnic_send_crq(adapter, &crq);
3378 
3379 	crq.request_capability.capability =
3380 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3381 	crq.request_capability.number =
3382 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3383 	atomic_inc(&adapter->running_cap_crqs);
3384 	ibmvnic_send_crq(adapter, &crq);
3385 
3386 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3387 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3388 	atomic_inc(&adapter->running_cap_crqs);
3389 	ibmvnic_send_crq(adapter, &crq);
3390 
3391 	if (adapter->netdev->flags & IFF_PROMISC) {
3392 		if (adapter->promisc_supported) {
3393 			crq.request_capability.capability =
3394 			    cpu_to_be16(PROMISC_REQUESTED);
3395 			crq.request_capability.number = cpu_to_be64(1);
3396 			atomic_inc(&adapter->running_cap_crqs);
3397 			ibmvnic_send_crq(adapter, &crq);
3398 		}
3399 	} else {
3400 		crq.request_capability.capability =
3401 		    cpu_to_be16(PROMISC_REQUESTED);
3402 		crq.request_capability.number = cpu_to_be64(0);
3403 		atomic_inc(&adapter->running_cap_crqs);
3404 		ibmvnic_send_crq(adapter, &crq);
3405 	}
3406 }
3407 
3408 static int pending_scrq(struct ibmvnic_adapter *adapter,
3409 			struct ibmvnic_sub_crq_queue *scrq)
3410 {
3411 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3412 
3413 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3414 		return 1;
3415 	else
3416 		return 0;
3417 }
3418 
3419 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3420 					struct ibmvnic_sub_crq_queue *scrq)
3421 {
3422 	union sub_crq *entry;
3423 	unsigned long flags;
3424 
3425 	spin_lock_irqsave(&scrq->lock, flags);
3426 	entry = &scrq->msgs[scrq->cur];
3427 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3428 		if (++scrq->cur == scrq->size)
3429 			scrq->cur = 0;
3430 	} else {
3431 		entry = NULL;
3432 	}
3433 	spin_unlock_irqrestore(&scrq->lock, flags);
3434 
3435 	return entry;
3436 }
3437 
3438 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3439 {
3440 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3441 	union ibmvnic_crq *crq;
3442 
3443 	crq = &queue->msgs[queue->cur];
3444 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3445 		if (++queue->cur == queue->size)
3446 			queue->cur = 0;
3447 	} else {
3448 		crq = NULL;
3449 	}
3450 
3451 	return crq;
3452 }
3453 
3454 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3455 {
3456 	switch (rc) {
3457 	case H_PARAMETER:
3458 		dev_warn_ratelimited(dev,
3459 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3460 				     func, rc);
3461 		break;
3462 	case H_CLOSED:
3463 		dev_warn_ratelimited(dev,
3464 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3465 				     func, rc);
3466 		break;
3467 	default:
3468 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3469 		break;
3470 	}
3471 }
3472 
3473 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3474 		       union sub_crq *sub_crq)
3475 {
3476 	unsigned int ua = adapter->vdev->unit_address;
3477 	struct device *dev = &adapter->vdev->dev;
3478 	u64 *u64_crq = (u64 *)sub_crq;
3479 	int rc;
3480 
3481 	netdev_dbg(adapter->netdev,
3482 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3483 		   (unsigned long int)cpu_to_be64(remote_handle),
3484 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3485 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3486 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3487 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3488 
3489 	/* Make sure the hypervisor sees the complete request */
3490 	mb();
3491 
3492 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3493 				cpu_to_be64(remote_handle),
3494 				cpu_to_be64(u64_crq[0]),
3495 				cpu_to_be64(u64_crq[1]),
3496 				cpu_to_be64(u64_crq[2]),
3497 				cpu_to_be64(u64_crq[3]));
3498 
3499 	if (rc)
3500 		print_subcrq_error(dev, rc, __func__);
3501 
3502 	return rc;
3503 }
3504 
3505 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3506 				u64 remote_handle, u64 ioba, u64 num_entries)
3507 {
3508 	unsigned int ua = adapter->vdev->unit_address;
3509 	struct device *dev = &adapter->vdev->dev;
3510 	int rc;
3511 
3512 	/* Make sure the hypervisor sees the complete request */
3513 	mb();
3514 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3515 				cpu_to_be64(remote_handle),
3516 				ioba, num_entries);
3517 
3518 	if (rc)
3519 		print_subcrq_error(dev, rc, __func__);
3520 
3521 	return rc;
3522 }
3523 
3524 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3525 			    union ibmvnic_crq *crq)
3526 {
3527 	unsigned int ua = adapter->vdev->unit_address;
3528 	struct device *dev = &adapter->vdev->dev;
3529 	u64 *u64_crq = (u64 *)crq;
3530 	int rc;
3531 
3532 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3533 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3534 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3535 
3536 	if (!adapter->crq.active &&
3537 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3538 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3539 		return -EINVAL;
3540 	}
3541 
3542 	/* Make sure the hypervisor sees the complete request */
3543 	mb();
3544 
3545 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3546 				cpu_to_be64(u64_crq[0]),
3547 				cpu_to_be64(u64_crq[1]));
3548 
3549 	if (rc) {
3550 		if (rc == H_CLOSED) {
3551 			dev_warn(dev, "CRQ Queue closed\n");
3552 			if (test_bit(0, &adapter->resetting))
3553 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3554 		}
3555 
3556 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3557 	}
3558 
3559 	return rc;
3560 }
3561 
3562 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3563 {
3564 	union ibmvnic_crq crq;
3565 
3566 	memset(&crq, 0, sizeof(crq));
3567 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3568 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3569 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3570 
3571 	return ibmvnic_send_crq(adapter, &crq);
3572 }
3573 
3574 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3575 {
3576 	union ibmvnic_crq crq;
3577 
3578 	memset(&crq, 0, sizeof(crq));
3579 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3580 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3581 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3582 
3583 	return ibmvnic_send_crq(adapter, &crq);
3584 }
3585 
3586 struct vnic_login_client_data {
3587 	u8	type;
3588 	__be16	len;
3589 	char	name[];
3590 } __packed;
3591 
3592 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3593 {
3594 	int len;
3595 
3596 	/* Calculate the amount of buffer space needed for the
3597 	 * vnic client data in the login buffer. There are four entries,
3598 	 * OS name, LPAR name, device name, and a null last entry.
3599 	 */
3600 	len = 4 * sizeof(struct vnic_login_client_data);
3601 	len += 6; /* "Linux" plus NULL */
3602 	len += strlen(utsname()->nodename) + 1;
3603 	len += strlen(adapter->netdev->name) + 1;
3604 
3605 	return len;
3606 }
3607 
3608 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3609 				 struct vnic_login_client_data *vlcd)
3610 {
3611 	const char *os_name = "Linux";
3612 	int len;
3613 
3614 	/* Type 1 - LPAR OS */
3615 	vlcd->type = 1;
3616 	len = strlen(os_name) + 1;
3617 	vlcd->len = cpu_to_be16(len);
3618 	strncpy(vlcd->name, os_name, len);
3619 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3620 
3621 	/* Type 2 - LPAR name */
3622 	vlcd->type = 2;
3623 	len = strlen(utsname()->nodename) + 1;
3624 	vlcd->len = cpu_to_be16(len);
3625 	strncpy(vlcd->name, utsname()->nodename, len);
3626 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3627 
3628 	/* Type 3 - device name */
3629 	vlcd->type = 3;
3630 	len = strlen(adapter->netdev->name) + 1;
3631 	vlcd->len = cpu_to_be16(len);
3632 	strncpy(vlcd->name, adapter->netdev->name, len);
3633 }
3634 
3635 static int send_login(struct ibmvnic_adapter *adapter)
3636 {
3637 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3638 	struct ibmvnic_login_buffer *login_buffer;
3639 	struct device *dev = &adapter->vdev->dev;
3640 	dma_addr_t rsp_buffer_token;
3641 	dma_addr_t buffer_token;
3642 	size_t rsp_buffer_size;
3643 	union ibmvnic_crq crq;
3644 	size_t buffer_size;
3645 	__be64 *tx_list_p;
3646 	__be64 *rx_list_p;
3647 	int client_data_len;
3648 	struct vnic_login_client_data *vlcd;
3649 	int i;
3650 
3651 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3652 		netdev_err(adapter->netdev,
3653 			   "RX or TX queues are not allocated, device login failed\n");
3654 		return -1;
3655 	}
3656 
3657 	release_login_rsp_buffer(adapter);
3658 	client_data_len = vnic_client_data_len(adapter);
3659 
3660 	buffer_size =
3661 	    sizeof(struct ibmvnic_login_buffer) +
3662 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3663 	    client_data_len;
3664 
3665 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3666 	if (!login_buffer)
3667 		goto buf_alloc_failed;
3668 
3669 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3670 				      DMA_TO_DEVICE);
3671 	if (dma_mapping_error(dev, buffer_token)) {
3672 		dev_err(dev, "Couldn't map login buffer\n");
3673 		goto buf_map_failed;
3674 	}
3675 
3676 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3677 			  sizeof(u64) * adapter->req_tx_queues +
3678 			  sizeof(u64) * adapter->req_rx_queues +
3679 			  sizeof(u64) * adapter->req_rx_queues +
3680 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3681 
3682 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3683 	if (!login_rsp_buffer)
3684 		goto buf_rsp_alloc_failed;
3685 
3686 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3687 					  rsp_buffer_size, DMA_FROM_DEVICE);
3688 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3689 		dev_err(dev, "Couldn't map login rsp buffer\n");
3690 		goto buf_rsp_map_failed;
3691 	}
3692 
3693 	adapter->login_buf = login_buffer;
3694 	adapter->login_buf_token = buffer_token;
3695 	adapter->login_buf_sz = buffer_size;
3696 	adapter->login_rsp_buf = login_rsp_buffer;
3697 	adapter->login_rsp_buf_token = rsp_buffer_token;
3698 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3699 
3700 	login_buffer->len = cpu_to_be32(buffer_size);
3701 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3702 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3703 	login_buffer->off_txcomp_subcrqs =
3704 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3705 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3706 	login_buffer->off_rxcomp_subcrqs =
3707 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3708 			sizeof(u64) * adapter->req_tx_queues);
3709 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3710 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3711 
3712 	tx_list_p = (__be64 *)((char *)login_buffer +
3713 				      sizeof(struct ibmvnic_login_buffer));
3714 	rx_list_p = (__be64 *)((char *)login_buffer +
3715 				      sizeof(struct ibmvnic_login_buffer) +
3716 				      sizeof(u64) * adapter->req_tx_queues);
3717 
3718 	for (i = 0; i < adapter->req_tx_queues; i++) {
3719 		if (adapter->tx_scrq[i]) {
3720 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3721 						   crq_num);
3722 		}
3723 	}
3724 
3725 	for (i = 0; i < adapter->req_rx_queues; i++) {
3726 		if (adapter->rx_scrq[i]) {
3727 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3728 						   crq_num);
3729 		}
3730 	}
3731 
3732 	/* Insert vNIC login client data */
3733 	vlcd = (struct vnic_login_client_data *)
3734 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3735 	login_buffer->client_data_offset =
3736 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3737 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3738 
3739 	vnic_add_client_data(adapter, vlcd);
3740 
3741 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3742 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3743 		netdev_dbg(adapter->netdev, "%016lx\n",
3744 			   ((unsigned long int *)(adapter->login_buf))[i]);
3745 	}
3746 
3747 	memset(&crq, 0, sizeof(crq));
3748 	crq.login.first = IBMVNIC_CRQ_CMD;
3749 	crq.login.cmd = LOGIN;
3750 	crq.login.ioba = cpu_to_be32(buffer_token);
3751 	crq.login.len = cpu_to_be32(buffer_size);
3752 	ibmvnic_send_crq(adapter, &crq);
3753 
3754 	return 0;
3755 
3756 buf_rsp_map_failed:
3757 	kfree(login_rsp_buffer);
3758 buf_rsp_alloc_failed:
3759 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3760 buf_map_failed:
3761 	kfree(login_buffer);
3762 buf_alloc_failed:
3763 	return -1;
3764 }
3765 
3766 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3767 			    u32 len, u8 map_id)
3768 {
3769 	union ibmvnic_crq crq;
3770 
3771 	memset(&crq, 0, sizeof(crq));
3772 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3773 	crq.request_map.cmd = REQUEST_MAP;
3774 	crq.request_map.map_id = map_id;
3775 	crq.request_map.ioba = cpu_to_be32(addr);
3776 	crq.request_map.len = cpu_to_be32(len);
3777 	return ibmvnic_send_crq(adapter, &crq);
3778 }
3779 
3780 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3781 {
3782 	union ibmvnic_crq crq;
3783 
3784 	memset(&crq, 0, sizeof(crq));
3785 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3786 	crq.request_unmap.cmd = REQUEST_UNMAP;
3787 	crq.request_unmap.map_id = map_id;
3788 	return ibmvnic_send_crq(adapter, &crq);
3789 }
3790 
3791 static void send_map_query(struct ibmvnic_adapter *adapter)
3792 {
3793 	union ibmvnic_crq crq;
3794 
3795 	memset(&crq, 0, sizeof(crq));
3796 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3797 	crq.query_map.cmd = QUERY_MAP;
3798 	ibmvnic_send_crq(adapter, &crq);
3799 }
3800 
3801 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3802 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3803 {
3804 	union ibmvnic_crq crq;
3805 
3806 	atomic_set(&adapter->running_cap_crqs, 0);
3807 	memset(&crq, 0, sizeof(crq));
3808 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3809 	crq.query_capability.cmd = QUERY_CAPABILITY;
3810 
3811 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3812 	atomic_inc(&adapter->running_cap_crqs);
3813 	ibmvnic_send_crq(adapter, &crq);
3814 
3815 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3816 	atomic_inc(&adapter->running_cap_crqs);
3817 	ibmvnic_send_crq(adapter, &crq);
3818 
3819 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3820 	atomic_inc(&adapter->running_cap_crqs);
3821 	ibmvnic_send_crq(adapter, &crq);
3822 
3823 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3824 	atomic_inc(&adapter->running_cap_crqs);
3825 	ibmvnic_send_crq(adapter, &crq);
3826 
3827 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3828 	atomic_inc(&adapter->running_cap_crqs);
3829 	ibmvnic_send_crq(adapter, &crq);
3830 
3831 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3832 	atomic_inc(&adapter->running_cap_crqs);
3833 	ibmvnic_send_crq(adapter, &crq);
3834 
3835 	crq.query_capability.capability =
3836 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3837 	atomic_inc(&adapter->running_cap_crqs);
3838 	ibmvnic_send_crq(adapter, &crq);
3839 
3840 	crq.query_capability.capability =
3841 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3842 	atomic_inc(&adapter->running_cap_crqs);
3843 	ibmvnic_send_crq(adapter, &crq);
3844 
3845 	crq.query_capability.capability =
3846 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3847 	atomic_inc(&adapter->running_cap_crqs);
3848 	ibmvnic_send_crq(adapter, &crq);
3849 
3850 	crq.query_capability.capability =
3851 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3852 	atomic_inc(&adapter->running_cap_crqs);
3853 	ibmvnic_send_crq(adapter, &crq);
3854 
3855 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3856 	atomic_inc(&adapter->running_cap_crqs);
3857 	ibmvnic_send_crq(adapter, &crq);
3858 
3859 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3860 	atomic_inc(&adapter->running_cap_crqs);
3861 	ibmvnic_send_crq(adapter, &crq);
3862 
3863 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3864 	atomic_inc(&adapter->running_cap_crqs);
3865 	ibmvnic_send_crq(adapter, &crq);
3866 
3867 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3868 	atomic_inc(&adapter->running_cap_crqs);
3869 	ibmvnic_send_crq(adapter, &crq);
3870 
3871 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3872 	atomic_inc(&adapter->running_cap_crqs);
3873 	ibmvnic_send_crq(adapter, &crq);
3874 
3875 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3876 	atomic_inc(&adapter->running_cap_crqs);
3877 	ibmvnic_send_crq(adapter, &crq);
3878 
3879 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3880 	atomic_inc(&adapter->running_cap_crqs);
3881 	ibmvnic_send_crq(adapter, &crq);
3882 
3883 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3884 	atomic_inc(&adapter->running_cap_crqs);
3885 	ibmvnic_send_crq(adapter, &crq);
3886 
3887 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3888 	atomic_inc(&adapter->running_cap_crqs);
3889 	ibmvnic_send_crq(adapter, &crq);
3890 
3891 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3892 	atomic_inc(&adapter->running_cap_crqs);
3893 	ibmvnic_send_crq(adapter, &crq);
3894 
3895 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3896 	atomic_inc(&adapter->running_cap_crqs);
3897 	ibmvnic_send_crq(adapter, &crq);
3898 
3899 	crq.query_capability.capability =
3900 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3901 	atomic_inc(&adapter->running_cap_crqs);
3902 	ibmvnic_send_crq(adapter, &crq);
3903 
3904 	crq.query_capability.capability =
3905 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3906 	atomic_inc(&adapter->running_cap_crqs);
3907 	ibmvnic_send_crq(adapter, &crq);
3908 
3909 	crq.query_capability.capability =
3910 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3911 	atomic_inc(&adapter->running_cap_crqs);
3912 	ibmvnic_send_crq(adapter, &crq);
3913 
3914 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3915 	atomic_inc(&adapter->running_cap_crqs);
3916 	ibmvnic_send_crq(adapter, &crq);
3917 }
3918 
3919 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3920 				struct ibmvnic_adapter *adapter)
3921 {
3922 	struct device *dev = &adapter->vdev->dev;
3923 
3924 	if (crq->get_vpd_size_rsp.rc.code) {
3925 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3926 			crq->get_vpd_size_rsp.rc.code);
3927 		complete(&adapter->fw_done);
3928 		return;
3929 	}
3930 
3931 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3932 	complete(&adapter->fw_done);
3933 }
3934 
3935 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3936 			   struct ibmvnic_adapter *adapter)
3937 {
3938 	struct device *dev = &adapter->vdev->dev;
3939 	unsigned char *substr = NULL;
3940 	u8 fw_level_len = 0;
3941 
3942 	memset(adapter->fw_version, 0, 32);
3943 
3944 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3945 			 DMA_FROM_DEVICE);
3946 
3947 	if (crq->get_vpd_rsp.rc.code) {
3948 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3949 			crq->get_vpd_rsp.rc.code);
3950 		goto complete;
3951 	}
3952 
3953 	/* get the position of the firmware version info
3954 	 * located after the ASCII 'RM' substring in the buffer
3955 	 */
3956 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3957 	if (!substr) {
3958 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3959 		goto complete;
3960 	}
3961 
3962 	/* get length of firmware level ASCII substring */
3963 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3964 		fw_level_len = *(substr + 2);
3965 	} else {
3966 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3967 		goto complete;
3968 	}
3969 
3970 	/* copy firmware version string from vpd into adapter */
3971 	if ((substr + 3 + fw_level_len) <
3972 	    (adapter->vpd->buff + adapter->vpd->len)) {
3973 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3974 	} else {
3975 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3976 	}
3977 
3978 complete:
3979 	if (adapter->fw_version[0] == '\0')
3980 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3981 	complete(&adapter->fw_done);
3982 }
3983 
3984 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3985 {
3986 	struct device *dev = &adapter->vdev->dev;
3987 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3988 	netdev_features_t old_hw_features = 0;
3989 	union ibmvnic_crq crq;
3990 	int i;
3991 
3992 	dma_unmap_single(dev, adapter->ip_offload_tok,
3993 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3994 
3995 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3996 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3997 		netdev_dbg(adapter->netdev, "%016lx\n",
3998 			   ((unsigned long int *)(buf))[i]);
3999 
4000 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4001 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4002 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4003 		   buf->tcp_ipv4_chksum);
4004 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4005 		   buf->tcp_ipv6_chksum);
4006 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4007 		   buf->udp_ipv4_chksum);
4008 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4009 		   buf->udp_ipv6_chksum);
4010 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4011 		   buf->large_tx_ipv4);
4012 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4013 		   buf->large_tx_ipv6);
4014 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4015 		   buf->large_rx_ipv4);
4016 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4017 		   buf->large_rx_ipv6);
4018 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4019 		   buf->max_ipv4_header_size);
4020 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4021 		   buf->max_ipv6_header_size);
4022 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4023 		   buf->max_tcp_header_size);
4024 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4025 		   buf->max_udp_header_size);
4026 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4027 		   buf->max_large_tx_size);
4028 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4029 		   buf->max_large_rx_size);
4030 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4031 		   buf->ipv6_extension_header);
4032 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4033 		   buf->tcp_pseudosum_req);
4034 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4035 		   buf->num_ipv6_ext_headers);
4036 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4037 		   buf->off_ipv6_ext_headers);
4038 
4039 	adapter->ip_offload_ctrl_tok =
4040 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
4041 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
4042 
4043 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4044 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4045 		return;
4046 	}
4047 
4048 	adapter->ip_offload_ctrl.len =
4049 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4050 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
4051 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
4052 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
4053 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4054 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
4055 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4056 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
4057 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
4058 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
4059 
4060 	/* large_rx disabled for now, additional features needed */
4061 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
4062 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
4063 
4064 	if (adapter->state != VNIC_PROBING) {
4065 		old_hw_features = adapter->netdev->hw_features;
4066 		adapter->netdev->hw_features = 0;
4067 	}
4068 
4069 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4070 
4071 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4072 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4073 
4074 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4075 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4076 
4077 	if ((adapter->netdev->features &
4078 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4079 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4080 
4081 	if (buf->large_tx_ipv4)
4082 		adapter->netdev->hw_features |= NETIF_F_TSO;
4083 	if (buf->large_tx_ipv6)
4084 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4085 
4086 	if (adapter->state == VNIC_PROBING) {
4087 		adapter->netdev->features |= adapter->netdev->hw_features;
4088 	} else if (old_hw_features != adapter->netdev->hw_features) {
4089 		netdev_features_t tmp = 0;
4090 
4091 		/* disable features no longer supported */
4092 		adapter->netdev->features &= adapter->netdev->hw_features;
4093 		/* turn on features now supported if previously enabled */
4094 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4095 			adapter->netdev->hw_features;
4096 		adapter->netdev->features |=
4097 				tmp & adapter->netdev->wanted_features;
4098 	}
4099 
4100 	memset(&crq, 0, sizeof(crq));
4101 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4102 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4103 	crq.control_ip_offload.len =
4104 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4105 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4106 	ibmvnic_send_crq(adapter, &crq);
4107 }
4108 
4109 static const char *ibmvnic_fw_err_cause(u16 cause)
4110 {
4111 	switch (cause) {
4112 	case ADAPTER_PROBLEM:
4113 		return "adapter problem";
4114 	case BUS_PROBLEM:
4115 		return "bus problem";
4116 	case FW_PROBLEM:
4117 		return "firmware problem";
4118 	case DD_PROBLEM:
4119 		return "device driver problem";
4120 	case EEH_RECOVERY:
4121 		return "EEH recovery";
4122 	case FW_UPDATED:
4123 		return "firmware updated";
4124 	case LOW_MEMORY:
4125 		return "low Memory";
4126 	default:
4127 		return "unknown";
4128 	}
4129 }
4130 
4131 static void handle_error_indication(union ibmvnic_crq *crq,
4132 				    struct ibmvnic_adapter *adapter)
4133 {
4134 	struct device *dev = &adapter->vdev->dev;
4135 	u16 cause;
4136 
4137 	cause = be16_to_cpu(crq->error_indication.error_cause);
4138 
4139 	dev_warn_ratelimited(dev,
4140 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4141 			     crq->error_indication.flags
4142 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4143 			     ibmvnic_fw_err_cause(cause));
4144 
4145 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4146 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4147 	else
4148 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4149 }
4150 
4151 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4152 				 struct ibmvnic_adapter *adapter)
4153 {
4154 	struct net_device *netdev = adapter->netdev;
4155 	struct device *dev = &adapter->vdev->dev;
4156 	long rc;
4157 
4158 	rc = crq->change_mac_addr_rsp.rc.code;
4159 	if (rc) {
4160 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4161 		goto out;
4162 	}
4163 	ether_addr_copy(netdev->dev_addr,
4164 			&crq->change_mac_addr_rsp.mac_addr[0]);
4165 out:
4166 	complete(&adapter->fw_done);
4167 	return rc;
4168 }
4169 
4170 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4171 				   struct ibmvnic_adapter *adapter)
4172 {
4173 	struct device *dev = &adapter->vdev->dev;
4174 	u64 *req_value;
4175 	char *name;
4176 
4177 	atomic_dec(&adapter->running_cap_crqs);
4178 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4179 	case REQ_TX_QUEUES:
4180 		req_value = &adapter->req_tx_queues;
4181 		name = "tx";
4182 		break;
4183 	case REQ_RX_QUEUES:
4184 		req_value = &adapter->req_rx_queues;
4185 		name = "rx";
4186 		break;
4187 	case REQ_RX_ADD_QUEUES:
4188 		req_value = &adapter->req_rx_add_queues;
4189 		name = "rx_add";
4190 		break;
4191 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4192 		req_value = &adapter->req_tx_entries_per_subcrq;
4193 		name = "tx_entries_per_subcrq";
4194 		break;
4195 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4196 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4197 		name = "rx_add_entries_per_subcrq";
4198 		break;
4199 	case REQ_MTU:
4200 		req_value = &adapter->req_mtu;
4201 		name = "mtu";
4202 		break;
4203 	case PROMISC_REQUESTED:
4204 		req_value = &adapter->promisc;
4205 		name = "promisc";
4206 		break;
4207 	default:
4208 		dev_err(dev, "Got invalid cap request rsp %d\n",
4209 			crq->request_capability.capability);
4210 		return;
4211 	}
4212 
4213 	switch (crq->request_capability_rsp.rc.code) {
4214 	case SUCCESS:
4215 		break;
4216 	case PARTIALSUCCESS:
4217 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4218 			 *req_value,
4219 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4220 					       number), name);
4221 
4222 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4223 		    REQ_MTU) {
4224 			pr_err("mtu of %llu is not supported. Reverting.\n",
4225 			       *req_value);
4226 			*req_value = adapter->fallback.mtu;
4227 		} else {
4228 			*req_value =
4229 				be64_to_cpu(crq->request_capability_rsp.number);
4230 		}
4231 
4232 		ibmvnic_send_req_caps(adapter, 1);
4233 		return;
4234 	default:
4235 		dev_err(dev, "Error %d in request cap rsp\n",
4236 			crq->request_capability_rsp.rc.code);
4237 		return;
4238 	}
4239 
4240 	/* Done receiving requested capabilities, query IP offload support */
4241 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4242 		union ibmvnic_crq newcrq;
4243 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4244 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4245 		    &adapter->ip_offload_buf;
4246 
4247 		adapter->wait_capability = false;
4248 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4249 							 buf_sz,
4250 							 DMA_FROM_DEVICE);
4251 
4252 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4253 			if (!firmware_has_feature(FW_FEATURE_CMO))
4254 				dev_err(dev, "Couldn't map offload buffer\n");
4255 			return;
4256 		}
4257 
4258 		memset(&newcrq, 0, sizeof(newcrq));
4259 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4260 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4261 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4262 		newcrq.query_ip_offload.ioba =
4263 		    cpu_to_be32(adapter->ip_offload_tok);
4264 
4265 		ibmvnic_send_crq(adapter, &newcrq);
4266 	}
4267 }
4268 
4269 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4270 			    struct ibmvnic_adapter *adapter)
4271 {
4272 	struct device *dev = &adapter->vdev->dev;
4273 	struct net_device *netdev = adapter->netdev;
4274 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4275 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4276 	int i;
4277 
4278 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4279 			 DMA_TO_DEVICE);
4280 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4281 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4282 
4283 	/* If the number of queues requested can't be allocated by the
4284 	 * server, the login response will return with code 1. We will need
4285 	 * to resend the login buffer with fewer queues requested.
4286 	 */
4287 	if (login_rsp_crq->generic.rc.code) {
4288 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4289 		complete(&adapter->init_done);
4290 		return 0;
4291 	}
4292 
4293 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4294 
4295 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4296 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4297 		netdev_dbg(adapter->netdev, "%016lx\n",
4298 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4299 	}
4300 
4301 	/* Sanity checks */
4302 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4303 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4304 	     adapter->req_rx_add_queues !=
4305 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4306 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4307 		ibmvnic_remove(adapter->vdev);
4308 		return -EIO;
4309 	}
4310 	release_login_buffer(adapter);
4311 	complete(&adapter->init_done);
4312 
4313 	return 0;
4314 }
4315 
4316 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4317 				     struct ibmvnic_adapter *adapter)
4318 {
4319 	struct device *dev = &adapter->vdev->dev;
4320 	long rc;
4321 
4322 	rc = crq->request_unmap_rsp.rc.code;
4323 	if (rc)
4324 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4325 }
4326 
4327 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4328 				 struct ibmvnic_adapter *adapter)
4329 {
4330 	struct net_device *netdev = adapter->netdev;
4331 	struct device *dev = &adapter->vdev->dev;
4332 	long rc;
4333 
4334 	rc = crq->query_map_rsp.rc.code;
4335 	if (rc) {
4336 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4337 		return;
4338 	}
4339 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4340 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4341 		   crq->query_map_rsp.free_pages);
4342 }
4343 
4344 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4345 				 struct ibmvnic_adapter *adapter)
4346 {
4347 	struct net_device *netdev = adapter->netdev;
4348 	struct device *dev = &adapter->vdev->dev;
4349 	long rc;
4350 
4351 	atomic_dec(&adapter->running_cap_crqs);
4352 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4353 		   atomic_read(&adapter->running_cap_crqs));
4354 	rc = crq->query_capability.rc.code;
4355 	if (rc) {
4356 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4357 		goto out;
4358 	}
4359 
4360 	switch (be16_to_cpu(crq->query_capability.capability)) {
4361 	case MIN_TX_QUEUES:
4362 		adapter->min_tx_queues =
4363 		    be64_to_cpu(crq->query_capability.number);
4364 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4365 			   adapter->min_tx_queues);
4366 		break;
4367 	case MIN_RX_QUEUES:
4368 		adapter->min_rx_queues =
4369 		    be64_to_cpu(crq->query_capability.number);
4370 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4371 			   adapter->min_rx_queues);
4372 		break;
4373 	case MIN_RX_ADD_QUEUES:
4374 		adapter->min_rx_add_queues =
4375 		    be64_to_cpu(crq->query_capability.number);
4376 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4377 			   adapter->min_rx_add_queues);
4378 		break;
4379 	case MAX_TX_QUEUES:
4380 		adapter->max_tx_queues =
4381 		    be64_to_cpu(crq->query_capability.number);
4382 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4383 			   adapter->max_tx_queues);
4384 		break;
4385 	case MAX_RX_QUEUES:
4386 		adapter->max_rx_queues =
4387 		    be64_to_cpu(crq->query_capability.number);
4388 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4389 			   adapter->max_rx_queues);
4390 		break;
4391 	case MAX_RX_ADD_QUEUES:
4392 		adapter->max_rx_add_queues =
4393 		    be64_to_cpu(crq->query_capability.number);
4394 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4395 			   adapter->max_rx_add_queues);
4396 		break;
4397 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4398 		adapter->min_tx_entries_per_subcrq =
4399 		    be64_to_cpu(crq->query_capability.number);
4400 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4401 			   adapter->min_tx_entries_per_subcrq);
4402 		break;
4403 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4404 		adapter->min_rx_add_entries_per_subcrq =
4405 		    be64_to_cpu(crq->query_capability.number);
4406 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4407 			   adapter->min_rx_add_entries_per_subcrq);
4408 		break;
4409 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4410 		adapter->max_tx_entries_per_subcrq =
4411 		    be64_to_cpu(crq->query_capability.number);
4412 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4413 			   adapter->max_tx_entries_per_subcrq);
4414 		break;
4415 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4416 		adapter->max_rx_add_entries_per_subcrq =
4417 		    be64_to_cpu(crq->query_capability.number);
4418 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4419 			   adapter->max_rx_add_entries_per_subcrq);
4420 		break;
4421 	case TCP_IP_OFFLOAD:
4422 		adapter->tcp_ip_offload =
4423 		    be64_to_cpu(crq->query_capability.number);
4424 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4425 			   adapter->tcp_ip_offload);
4426 		break;
4427 	case PROMISC_SUPPORTED:
4428 		adapter->promisc_supported =
4429 		    be64_to_cpu(crq->query_capability.number);
4430 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4431 			   adapter->promisc_supported);
4432 		break;
4433 	case MIN_MTU:
4434 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4435 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4436 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4437 		break;
4438 	case MAX_MTU:
4439 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4440 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4441 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4442 		break;
4443 	case MAX_MULTICAST_FILTERS:
4444 		adapter->max_multicast_filters =
4445 		    be64_to_cpu(crq->query_capability.number);
4446 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4447 			   adapter->max_multicast_filters);
4448 		break;
4449 	case VLAN_HEADER_INSERTION:
4450 		adapter->vlan_header_insertion =
4451 		    be64_to_cpu(crq->query_capability.number);
4452 		if (adapter->vlan_header_insertion)
4453 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4454 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4455 			   adapter->vlan_header_insertion);
4456 		break;
4457 	case RX_VLAN_HEADER_INSERTION:
4458 		adapter->rx_vlan_header_insertion =
4459 		    be64_to_cpu(crq->query_capability.number);
4460 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4461 			   adapter->rx_vlan_header_insertion);
4462 		break;
4463 	case MAX_TX_SG_ENTRIES:
4464 		adapter->max_tx_sg_entries =
4465 		    be64_to_cpu(crq->query_capability.number);
4466 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4467 			   adapter->max_tx_sg_entries);
4468 		break;
4469 	case RX_SG_SUPPORTED:
4470 		adapter->rx_sg_supported =
4471 		    be64_to_cpu(crq->query_capability.number);
4472 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4473 			   adapter->rx_sg_supported);
4474 		break;
4475 	case OPT_TX_COMP_SUB_QUEUES:
4476 		adapter->opt_tx_comp_sub_queues =
4477 		    be64_to_cpu(crq->query_capability.number);
4478 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4479 			   adapter->opt_tx_comp_sub_queues);
4480 		break;
4481 	case OPT_RX_COMP_QUEUES:
4482 		adapter->opt_rx_comp_queues =
4483 		    be64_to_cpu(crq->query_capability.number);
4484 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4485 			   adapter->opt_rx_comp_queues);
4486 		break;
4487 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4488 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4489 		    be64_to_cpu(crq->query_capability.number);
4490 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4491 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4492 		break;
4493 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4494 		adapter->opt_tx_entries_per_subcrq =
4495 		    be64_to_cpu(crq->query_capability.number);
4496 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4497 			   adapter->opt_tx_entries_per_subcrq);
4498 		break;
4499 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4500 		adapter->opt_rxba_entries_per_subcrq =
4501 		    be64_to_cpu(crq->query_capability.number);
4502 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4503 			   adapter->opt_rxba_entries_per_subcrq);
4504 		break;
4505 	case TX_RX_DESC_REQ:
4506 		adapter->tx_rx_desc_req = crq->query_capability.number;
4507 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4508 			   adapter->tx_rx_desc_req);
4509 		break;
4510 
4511 	default:
4512 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4513 			   crq->query_capability.capability);
4514 	}
4515 
4516 out:
4517 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4518 		adapter->wait_capability = false;
4519 		ibmvnic_send_req_caps(adapter, 0);
4520 	}
4521 }
4522 
4523 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4524 {
4525 	union ibmvnic_crq crq;
4526 	int rc;
4527 
4528 	memset(&crq, 0, sizeof(crq));
4529 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4530 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4531 
4532 	mutex_lock(&adapter->fw_lock);
4533 	adapter->fw_done_rc = 0;
4534 	reinit_completion(&adapter->fw_done);
4535 
4536 	rc = ibmvnic_send_crq(adapter, &crq);
4537 	if (rc) {
4538 		mutex_unlock(&adapter->fw_lock);
4539 		return rc;
4540 	}
4541 
4542 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4543 	if (rc) {
4544 		mutex_unlock(&adapter->fw_lock);
4545 		return rc;
4546 	}
4547 
4548 	mutex_unlock(&adapter->fw_lock);
4549 	return adapter->fw_done_rc ? -EIO : 0;
4550 }
4551 
4552 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4553 				       struct ibmvnic_adapter *adapter)
4554 {
4555 	struct net_device *netdev = adapter->netdev;
4556 	int rc;
4557 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4558 
4559 	rc = crq->query_phys_parms_rsp.rc.code;
4560 	if (rc) {
4561 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4562 		return rc;
4563 	}
4564 	switch (rspeed) {
4565 	case IBMVNIC_10MBPS:
4566 		adapter->speed = SPEED_10;
4567 		break;
4568 	case IBMVNIC_100MBPS:
4569 		adapter->speed = SPEED_100;
4570 		break;
4571 	case IBMVNIC_1GBPS:
4572 		adapter->speed = SPEED_1000;
4573 		break;
4574 	case IBMVNIC_10GBP:
4575 		adapter->speed = SPEED_10000;
4576 		break;
4577 	case IBMVNIC_25GBPS:
4578 		adapter->speed = SPEED_25000;
4579 		break;
4580 	case IBMVNIC_40GBPS:
4581 		adapter->speed = SPEED_40000;
4582 		break;
4583 	case IBMVNIC_50GBPS:
4584 		adapter->speed = SPEED_50000;
4585 		break;
4586 	case IBMVNIC_100GBPS:
4587 		adapter->speed = SPEED_100000;
4588 		break;
4589 	default:
4590 		if (netif_carrier_ok(netdev))
4591 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4592 		adapter->speed = SPEED_UNKNOWN;
4593 	}
4594 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4595 		adapter->duplex = DUPLEX_FULL;
4596 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4597 		adapter->duplex = DUPLEX_HALF;
4598 	else
4599 		adapter->duplex = DUPLEX_UNKNOWN;
4600 
4601 	return rc;
4602 }
4603 
4604 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4605 			       struct ibmvnic_adapter *adapter)
4606 {
4607 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4608 	struct net_device *netdev = adapter->netdev;
4609 	struct device *dev = &adapter->vdev->dev;
4610 	u64 *u64_crq = (u64 *)crq;
4611 	long rc;
4612 
4613 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4614 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4615 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4616 	switch (gen_crq->first) {
4617 	case IBMVNIC_CRQ_INIT_RSP:
4618 		switch (gen_crq->cmd) {
4619 		case IBMVNIC_CRQ_INIT:
4620 			dev_info(dev, "Partner initialized\n");
4621 			adapter->from_passive_init = true;
4622 			adapter->failover_pending = false;
4623 			if (!completion_done(&adapter->init_done)) {
4624 				complete(&adapter->init_done);
4625 				adapter->init_done_rc = -EIO;
4626 			}
4627 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4628 			break;
4629 		case IBMVNIC_CRQ_INIT_COMPLETE:
4630 			dev_info(dev, "Partner initialization complete\n");
4631 			adapter->crq.active = true;
4632 			send_version_xchg(adapter);
4633 			break;
4634 		default:
4635 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4636 		}
4637 		return;
4638 	case IBMVNIC_CRQ_XPORT_EVENT:
4639 		netif_carrier_off(netdev);
4640 		adapter->crq.active = false;
4641 		/* terminate any thread waiting for a response
4642 		 * from the device
4643 		 */
4644 		if (!completion_done(&adapter->fw_done)) {
4645 			adapter->fw_done_rc = -EIO;
4646 			complete(&adapter->fw_done);
4647 		}
4648 		if (!completion_done(&adapter->stats_done))
4649 			complete(&adapter->stats_done);
4650 		if (test_bit(0, &adapter->resetting))
4651 			adapter->force_reset_recovery = true;
4652 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4653 			dev_info(dev, "Migrated, re-enabling adapter\n");
4654 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4655 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4656 			dev_info(dev, "Backing device failover detected\n");
4657 			adapter->failover_pending = true;
4658 		} else {
4659 			/* The adapter lost the connection */
4660 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4661 				gen_crq->cmd);
4662 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4663 		}
4664 		return;
4665 	case IBMVNIC_CRQ_CMD_RSP:
4666 		break;
4667 	default:
4668 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4669 			gen_crq->first);
4670 		return;
4671 	}
4672 
4673 	switch (gen_crq->cmd) {
4674 	case VERSION_EXCHANGE_RSP:
4675 		rc = crq->version_exchange_rsp.rc.code;
4676 		if (rc) {
4677 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4678 			break;
4679 		}
4680 		dev_info(dev, "Partner protocol version is %d\n",
4681 			 crq->version_exchange_rsp.version);
4682 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
4683 		    ibmvnic_version)
4684 			ibmvnic_version =
4685 			    be16_to_cpu(crq->version_exchange_rsp.version);
4686 		send_cap_queries(adapter);
4687 		break;
4688 	case QUERY_CAPABILITY_RSP:
4689 		handle_query_cap_rsp(crq, adapter);
4690 		break;
4691 	case QUERY_MAP_RSP:
4692 		handle_query_map_rsp(crq, adapter);
4693 		break;
4694 	case REQUEST_MAP_RSP:
4695 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4696 		complete(&adapter->fw_done);
4697 		break;
4698 	case REQUEST_UNMAP_RSP:
4699 		handle_request_unmap_rsp(crq, adapter);
4700 		break;
4701 	case REQUEST_CAPABILITY_RSP:
4702 		handle_request_cap_rsp(crq, adapter);
4703 		break;
4704 	case LOGIN_RSP:
4705 		netdev_dbg(netdev, "Got Login Response\n");
4706 		handle_login_rsp(crq, adapter);
4707 		break;
4708 	case LOGICAL_LINK_STATE_RSP:
4709 		netdev_dbg(netdev,
4710 			   "Got Logical Link State Response, state: %d rc: %d\n",
4711 			   crq->logical_link_state_rsp.link_state,
4712 			   crq->logical_link_state_rsp.rc.code);
4713 		adapter->logical_link_state =
4714 		    crq->logical_link_state_rsp.link_state;
4715 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4716 		complete(&adapter->init_done);
4717 		break;
4718 	case LINK_STATE_INDICATION:
4719 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4720 		adapter->phys_link_state =
4721 		    crq->link_state_indication.phys_link_state;
4722 		adapter->logical_link_state =
4723 		    crq->link_state_indication.logical_link_state;
4724 		if (adapter->phys_link_state && adapter->logical_link_state)
4725 			netif_carrier_on(netdev);
4726 		else
4727 			netif_carrier_off(netdev);
4728 		break;
4729 	case CHANGE_MAC_ADDR_RSP:
4730 		netdev_dbg(netdev, "Got MAC address change Response\n");
4731 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4732 		break;
4733 	case ERROR_INDICATION:
4734 		netdev_dbg(netdev, "Got Error Indication\n");
4735 		handle_error_indication(crq, adapter);
4736 		break;
4737 	case REQUEST_STATISTICS_RSP:
4738 		netdev_dbg(netdev, "Got Statistics Response\n");
4739 		complete(&adapter->stats_done);
4740 		break;
4741 	case QUERY_IP_OFFLOAD_RSP:
4742 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4743 		handle_query_ip_offload_rsp(adapter);
4744 		break;
4745 	case MULTICAST_CTRL_RSP:
4746 		netdev_dbg(netdev, "Got multicast control Response\n");
4747 		break;
4748 	case CONTROL_IP_OFFLOAD_RSP:
4749 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4750 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4751 				 sizeof(adapter->ip_offload_ctrl),
4752 				 DMA_TO_DEVICE);
4753 		complete(&adapter->init_done);
4754 		break;
4755 	case COLLECT_FW_TRACE_RSP:
4756 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4757 		complete(&adapter->fw_done);
4758 		break;
4759 	case GET_VPD_SIZE_RSP:
4760 		handle_vpd_size_rsp(crq, adapter);
4761 		break;
4762 	case GET_VPD_RSP:
4763 		handle_vpd_rsp(crq, adapter);
4764 		break;
4765 	case QUERY_PHYS_PARMS_RSP:
4766 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4767 		complete(&adapter->fw_done);
4768 		break;
4769 	default:
4770 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4771 			   gen_crq->cmd);
4772 	}
4773 }
4774 
4775 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4776 {
4777 	struct ibmvnic_adapter *adapter = instance;
4778 
4779 	tasklet_schedule(&adapter->tasklet);
4780 	return IRQ_HANDLED;
4781 }
4782 
4783 static void ibmvnic_tasklet(void *data)
4784 {
4785 	struct ibmvnic_adapter *adapter = data;
4786 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4787 	union ibmvnic_crq *crq;
4788 	unsigned long flags;
4789 	bool done = false;
4790 
4791 	spin_lock_irqsave(&queue->lock, flags);
4792 	while (!done) {
4793 		/* Pull all the valid messages off the CRQ */
4794 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4795 			ibmvnic_handle_crq(crq, adapter);
4796 			crq->generic.first = 0;
4797 		}
4798 
4799 		/* remain in tasklet until all
4800 		 * capabilities responses are received
4801 		 */
4802 		if (!adapter->wait_capability)
4803 			done = true;
4804 	}
4805 	/* if capabilities CRQ's were sent in this tasklet, the following
4806 	 * tasklet must wait until all responses are received
4807 	 */
4808 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4809 		adapter->wait_capability = true;
4810 	spin_unlock_irqrestore(&queue->lock, flags);
4811 }
4812 
4813 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4814 {
4815 	struct vio_dev *vdev = adapter->vdev;
4816 	int rc;
4817 
4818 	do {
4819 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4820 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4821 
4822 	if (rc)
4823 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4824 
4825 	return rc;
4826 }
4827 
4828 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4829 {
4830 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4831 	struct device *dev = &adapter->vdev->dev;
4832 	struct vio_dev *vdev = adapter->vdev;
4833 	int rc;
4834 
4835 	/* Close the CRQ */
4836 	do {
4837 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4838 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4839 
4840 	/* Clean out the queue */
4841 	memset(crq->msgs, 0, PAGE_SIZE);
4842 	crq->cur = 0;
4843 	crq->active = false;
4844 
4845 	/* And re-open it again */
4846 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4847 				crq->msg_token, PAGE_SIZE);
4848 
4849 	if (rc == H_CLOSED)
4850 		/* Adapter is good, but other end is not ready */
4851 		dev_warn(dev, "Partner adapter not ready\n");
4852 	else if (rc != 0)
4853 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4854 
4855 	return rc;
4856 }
4857 
4858 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4859 {
4860 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4861 	struct vio_dev *vdev = adapter->vdev;
4862 	long rc;
4863 
4864 	if (!crq->msgs)
4865 		return;
4866 
4867 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4868 	free_irq(vdev->irq, adapter);
4869 	tasklet_kill(&adapter->tasklet);
4870 	do {
4871 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4872 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4873 
4874 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4875 			 DMA_BIDIRECTIONAL);
4876 	free_page((unsigned long)crq->msgs);
4877 	crq->msgs = NULL;
4878 	crq->active = false;
4879 }
4880 
4881 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4882 {
4883 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4884 	struct device *dev = &adapter->vdev->dev;
4885 	struct vio_dev *vdev = adapter->vdev;
4886 	int rc, retrc = -ENOMEM;
4887 
4888 	if (crq->msgs)
4889 		return 0;
4890 
4891 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4892 	/* Should we allocate more than one page? */
4893 
4894 	if (!crq->msgs)
4895 		return -ENOMEM;
4896 
4897 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4898 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4899 					DMA_BIDIRECTIONAL);
4900 	if (dma_mapping_error(dev, crq->msg_token))
4901 		goto map_failed;
4902 
4903 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4904 				crq->msg_token, PAGE_SIZE);
4905 
4906 	if (rc == H_RESOURCE)
4907 		/* maybe kexecing and resource is busy. try a reset */
4908 		rc = ibmvnic_reset_crq(adapter);
4909 	retrc = rc;
4910 
4911 	if (rc == H_CLOSED) {
4912 		dev_warn(dev, "Partner adapter not ready\n");
4913 	} else if (rc) {
4914 		dev_warn(dev, "Error %d opening adapter\n", rc);
4915 		goto reg_crq_failed;
4916 	}
4917 
4918 	retrc = 0;
4919 
4920 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4921 		     (unsigned long)adapter);
4922 
4923 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4924 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4925 		 adapter->vdev->unit_address);
4926 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4927 	if (rc) {
4928 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4929 			vdev->irq, rc);
4930 		goto req_irq_failed;
4931 	}
4932 
4933 	rc = vio_enable_interrupts(vdev);
4934 	if (rc) {
4935 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4936 		goto req_irq_failed;
4937 	}
4938 
4939 	crq->cur = 0;
4940 	spin_lock_init(&crq->lock);
4941 
4942 	return retrc;
4943 
4944 req_irq_failed:
4945 	tasklet_kill(&adapter->tasklet);
4946 	do {
4947 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4948 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4949 reg_crq_failed:
4950 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4951 map_failed:
4952 	free_page((unsigned long)crq->msgs);
4953 	crq->msgs = NULL;
4954 	return retrc;
4955 }
4956 
4957 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4958 {
4959 	struct device *dev = &adapter->vdev->dev;
4960 	unsigned long timeout = msecs_to_jiffies(30000);
4961 	u64 old_num_rx_queues, old_num_tx_queues;
4962 	int rc;
4963 
4964 	adapter->from_passive_init = false;
4965 
4966 	old_num_rx_queues = adapter->req_rx_queues;
4967 	old_num_tx_queues = adapter->req_tx_queues;
4968 
4969 	reinit_completion(&adapter->init_done);
4970 	adapter->init_done_rc = 0;
4971 	ibmvnic_send_crq_init(adapter);
4972 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4973 		dev_err(dev, "Initialization sequence timed out\n");
4974 		return -1;
4975 	}
4976 
4977 	if (adapter->init_done_rc) {
4978 		release_crq_queue(adapter);
4979 		return adapter->init_done_rc;
4980 	}
4981 
4982 	if (adapter->from_passive_init) {
4983 		adapter->state = VNIC_OPEN;
4984 		adapter->from_passive_init = false;
4985 		return -1;
4986 	}
4987 
4988 	if (test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
4989 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
4990 		if (adapter->req_rx_queues != old_num_rx_queues ||
4991 		    adapter->req_tx_queues != old_num_tx_queues) {
4992 			release_sub_crqs(adapter, 0);
4993 			rc = init_sub_crqs(adapter);
4994 		} else {
4995 			rc = reset_sub_crq_queues(adapter);
4996 		}
4997 	} else {
4998 		rc = init_sub_crqs(adapter);
4999 	}
5000 
5001 	if (rc) {
5002 		dev_err(dev, "Initialization of sub crqs failed\n");
5003 		release_crq_queue(adapter);
5004 		return rc;
5005 	}
5006 
5007 	rc = init_sub_crq_irqs(adapter);
5008 	if (rc) {
5009 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5010 		release_crq_queue(adapter);
5011 	}
5012 
5013 	return rc;
5014 }
5015 
5016 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
5017 {
5018 	struct device *dev = &adapter->vdev->dev;
5019 	unsigned long timeout = msecs_to_jiffies(30000);
5020 	int rc;
5021 
5022 	adapter->from_passive_init = false;
5023 
5024 	adapter->init_done_rc = 0;
5025 	ibmvnic_send_crq_init(adapter);
5026 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5027 		dev_err(dev, "Initialization sequence timed out\n");
5028 		return -1;
5029 	}
5030 
5031 	if (adapter->init_done_rc) {
5032 		release_crq_queue(adapter);
5033 		return adapter->init_done_rc;
5034 	}
5035 
5036 	if (adapter->from_passive_init) {
5037 		adapter->state = VNIC_OPEN;
5038 		adapter->from_passive_init = false;
5039 		return -1;
5040 	}
5041 
5042 	rc = init_sub_crqs(adapter);
5043 	if (rc) {
5044 		dev_err(dev, "Initialization of sub crqs failed\n");
5045 		release_crq_queue(adapter);
5046 		return rc;
5047 	}
5048 
5049 	rc = init_sub_crq_irqs(adapter);
5050 	if (rc) {
5051 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5052 		release_crq_queue(adapter);
5053 	}
5054 
5055 	return rc;
5056 }
5057 
5058 static struct device_attribute dev_attr_failover;
5059 
5060 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5061 {
5062 	struct ibmvnic_adapter *adapter;
5063 	struct net_device *netdev;
5064 	unsigned char *mac_addr_p;
5065 	int rc;
5066 
5067 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5068 		dev->unit_address);
5069 
5070 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5071 							VETH_MAC_ADDR, NULL);
5072 	if (!mac_addr_p) {
5073 		dev_err(&dev->dev,
5074 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5075 			__FILE__, __LINE__);
5076 		return 0;
5077 	}
5078 
5079 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5080 				   IBMVNIC_MAX_QUEUES);
5081 	if (!netdev)
5082 		return -ENOMEM;
5083 
5084 	adapter = netdev_priv(netdev);
5085 	adapter->state = VNIC_PROBING;
5086 	dev_set_drvdata(&dev->dev, netdev);
5087 	adapter->vdev = dev;
5088 	adapter->netdev = netdev;
5089 
5090 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5091 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5092 	netdev->irq = dev->irq;
5093 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5094 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5095 	SET_NETDEV_DEV(netdev, &dev->dev);
5096 
5097 	spin_lock_init(&adapter->stats_lock);
5098 
5099 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5100 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5101 			  __ibmvnic_delayed_reset);
5102 	INIT_LIST_HEAD(&adapter->rwi_list);
5103 	spin_lock_init(&adapter->rwi_lock);
5104 	spin_lock_init(&adapter->state_lock);
5105 	mutex_init(&adapter->fw_lock);
5106 	init_completion(&adapter->init_done);
5107 	init_completion(&adapter->fw_done);
5108 	init_completion(&adapter->reset_done);
5109 	init_completion(&adapter->stats_done);
5110 	clear_bit(0, &adapter->resetting);
5111 
5112 	do {
5113 		rc = init_crq_queue(adapter);
5114 		if (rc) {
5115 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5116 				rc);
5117 			goto ibmvnic_init_fail;
5118 		}
5119 
5120 		rc = ibmvnic_init(adapter);
5121 		if (rc && rc != EAGAIN)
5122 			goto ibmvnic_init_fail;
5123 	} while (rc == EAGAIN);
5124 
5125 	rc = init_stats_buffers(adapter);
5126 	if (rc)
5127 		goto ibmvnic_init_fail;
5128 
5129 	rc = init_stats_token(adapter);
5130 	if (rc)
5131 		goto ibmvnic_stats_fail;
5132 
5133 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5134 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5135 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5136 
5137 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5138 	if (rc)
5139 		goto ibmvnic_dev_file_err;
5140 
5141 	netif_carrier_off(netdev);
5142 	rc = register_netdev(netdev);
5143 	if (rc) {
5144 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5145 		goto ibmvnic_register_fail;
5146 	}
5147 	dev_info(&dev->dev, "ibmvnic registered\n");
5148 
5149 	adapter->state = VNIC_PROBED;
5150 
5151 	adapter->wait_for_reset = false;
5152 
5153 	return 0;
5154 
5155 ibmvnic_register_fail:
5156 	device_remove_file(&dev->dev, &dev_attr_failover);
5157 
5158 ibmvnic_dev_file_err:
5159 	release_stats_token(adapter);
5160 
5161 ibmvnic_stats_fail:
5162 	release_stats_buffers(adapter);
5163 
5164 ibmvnic_init_fail:
5165 	release_sub_crqs(adapter, 1);
5166 	release_crq_queue(adapter);
5167 	mutex_destroy(&adapter->fw_lock);
5168 	free_netdev(netdev);
5169 
5170 	return rc;
5171 }
5172 
5173 static int ibmvnic_remove(struct vio_dev *dev)
5174 {
5175 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5176 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5177 	unsigned long flags;
5178 
5179 	spin_lock_irqsave(&adapter->state_lock, flags);
5180 	if (adapter->state == VNIC_RESETTING) {
5181 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5182 		return -EBUSY;
5183 	}
5184 
5185 	adapter->state = VNIC_REMOVING;
5186 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5187 
5188 	rtnl_lock();
5189 	unregister_netdevice(netdev);
5190 
5191 	release_resources(adapter);
5192 	release_sub_crqs(adapter, 1);
5193 	release_crq_queue(adapter);
5194 
5195 	release_stats_token(adapter);
5196 	release_stats_buffers(adapter);
5197 
5198 	adapter->state = VNIC_REMOVED;
5199 
5200 	rtnl_unlock();
5201 	mutex_destroy(&adapter->fw_lock);
5202 	device_remove_file(&dev->dev, &dev_attr_failover);
5203 	free_netdev(netdev);
5204 	dev_set_drvdata(&dev->dev, NULL);
5205 
5206 	return 0;
5207 }
5208 
5209 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5210 			      const char *buf, size_t count)
5211 {
5212 	struct net_device *netdev = dev_get_drvdata(dev);
5213 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5214 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5215 	__be64 session_token;
5216 	long rc;
5217 
5218 	if (!sysfs_streq(buf, "1"))
5219 		return -EINVAL;
5220 
5221 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5222 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5223 	if (rc) {
5224 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5225 			   rc);
5226 		return -EINVAL;
5227 	}
5228 
5229 	session_token = (__be64)retbuf[0];
5230 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5231 		   be64_to_cpu(session_token));
5232 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5233 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5234 	if (rc) {
5235 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5236 			   rc);
5237 		return -EINVAL;
5238 	}
5239 
5240 	return count;
5241 }
5242 
5243 static DEVICE_ATTR_WO(failover);
5244 
5245 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5246 {
5247 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5248 	struct ibmvnic_adapter *adapter;
5249 	struct iommu_table *tbl;
5250 	unsigned long ret = 0;
5251 	int i;
5252 
5253 	tbl = get_iommu_table_base(&vdev->dev);
5254 
5255 	/* netdev inits at probe time along with the structures we need below*/
5256 	if (!netdev)
5257 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5258 
5259 	adapter = netdev_priv(netdev);
5260 
5261 	ret += PAGE_SIZE; /* the crq message queue */
5262 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5263 
5264 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5265 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5266 
5267 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
5268 	     i++)
5269 		ret += adapter->rx_pool[i].size *
5270 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5271 
5272 	return ret;
5273 }
5274 
5275 static int ibmvnic_resume(struct device *dev)
5276 {
5277 	struct net_device *netdev = dev_get_drvdata(dev);
5278 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5279 
5280 	if (adapter->state != VNIC_OPEN)
5281 		return 0;
5282 
5283 	tasklet_schedule(&adapter->tasklet);
5284 
5285 	return 0;
5286 }
5287 
5288 static const struct vio_device_id ibmvnic_device_table[] = {
5289 	{"network", "IBM,vnic"},
5290 	{"", "" }
5291 };
5292 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5293 
5294 static const struct dev_pm_ops ibmvnic_pm_ops = {
5295 	.resume = ibmvnic_resume
5296 };
5297 
5298 static struct vio_driver ibmvnic_driver = {
5299 	.id_table       = ibmvnic_device_table,
5300 	.probe          = ibmvnic_probe,
5301 	.remove         = ibmvnic_remove,
5302 	.get_desired_dma = ibmvnic_get_desired_dma,
5303 	.name		= ibmvnic_driver_name,
5304 	.pm		= &ibmvnic_pm_ops,
5305 };
5306 
5307 /* module functions */
5308 static int __init ibmvnic_module_init(void)
5309 {
5310 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5311 		IBMVNIC_DRIVER_VERSION);
5312 
5313 	return vio_register_driver(&ibmvnic_driver);
5314 }
5315 
5316 static void __exit ibmvnic_module_exit(void)
5317 {
5318 	vio_unregister_driver(&ibmvnic_driver);
5319 }
5320 
5321 module_init(ibmvnic_module_init);
5322 module_exit(ibmvnic_module_exit);
5323