xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 28efb0046512e8a13ed9f9bdf0d68d10bbfbe9cf)
1 /**************************************************************************/
2 /*                                                                        */
3 /*  IBM System i and System p Virtual NIC Device Driver                   */
4 /*  Copyright (C) 2014 IBM Corp.                                          */
5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
8 /*                                                                        */
9 /*  This program is free software; you can redistribute it and/or modify  */
10 /*  it under the terms of the GNU General Public License as published by  */
11 /*  the Free Software Foundation; either version 2 of the License, or     */
12 /*  (at your option) any later version.                                   */
13 /*                                                                        */
14 /*  This program is distributed in the hope that it will be useful,       */
15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
17 /*  GNU General Public License for more details.                          */
18 /*                                                                        */
19 /*  You should have received a copy of the GNU General Public License     */
20 /*  along with this program.                                              */
21 /*                                                                        */
22 /* This module contains the implementation of a virtual ethernet device   */
23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
26 /*									   */
27 /* Messages are passed between the VNIC driver and the VNIC server using  */
28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
29 /* issue and receive commands that initiate communication with the server */
30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
31 /* are used by the driver to notify the server that a packet is           */
32 /* ready for transmission or that a buffer has been added to receive a    */
33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
34 /* driver that a packet transmission has been completed or that a packet  */
35 /* has been received and placed in a waiting buffer.                      */
36 /*                                                                        */
37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
39 /* or receive has been completed, the VNIC driver is required to use      */
40 /* "long term mapping". This entails that large, continuous DMA mapped    */
41 /* buffers are allocated on driver initialization and these buffers are   */
42 /* then continuously reused to pass skbs to and from the VNIC server.     */
43 /*                                                                        */
44 /**************************************************************************/
45 
46 #include <linux/module.h>
47 #include <linux/moduleparam.h>
48 #include <linux/types.h>
49 #include <linux/errno.h>
50 #include <linux/completion.h>
51 #include <linux/ioport.h>
52 #include <linux/dma-mapping.h>
53 #include <linux/kernel.h>
54 #include <linux/netdevice.h>
55 #include <linux/etherdevice.h>
56 #include <linux/skbuff.h>
57 #include <linux/init.h>
58 #include <linux/delay.h>
59 #include <linux/mm.h>
60 #include <linux/ethtool.h>
61 #include <linux/proc_fs.h>
62 #include <linux/in.h>
63 #include <linux/ip.h>
64 #include <linux/ipv6.h>
65 #include <linux/irq.h>
66 #include <linux/kthread.h>
67 #include <linux/seq_file.h>
68 #include <linux/interrupt.h>
69 #include <net/net_namespace.h>
70 #include <asm/hvcall.h>
71 #include <linux/atomic.h>
72 #include <asm/vio.h>
73 #include <asm/iommu.h>
74 #include <linux/uaccess.h>
75 #include <asm/firmware.h>
76 #include <linux/workqueue.h>
77 #include <linux/if_vlan.h>
78 
79 #include "ibmvnic.h"
80 
81 static const char ibmvnic_driver_name[] = "ibmvnic";
82 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
83 
84 MODULE_AUTHOR("Santiago Leon");
85 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
86 MODULE_LICENSE("GPL");
87 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
88 
89 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
90 static int ibmvnic_remove(struct vio_dev *);
91 static void release_sub_crqs(struct ibmvnic_adapter *);
92 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
93 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
94 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
95 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
96 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
97 		       union sub_crq *sub_crq);
98 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
99 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
100 static int enable_scrq_irq(struct ibmvnic_adapter *,
101 			   struct ibmvnic_sub_crq_queue *);
102 static int disable_scrq_irq(struct ibmvnic_adapter *,
103 			    struct ibmvnic_sub_crq_queue *);
104 static int pending_scrq(struct ibmvnic_adapter *,
105 			struct ibmvnic_sub_crq_queue *);
106 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
107 					struct ibmvnic_sub_crq_queue *);
108 static int ibmvnic_poll(struct napi_struct *napi, int data);
109 static void send_map_query(struct ibmvnic_adapter *adapter);
110 static void send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
111 static void send_request_unmap(struct ibmvnic_adapter *, u8);
112 static void send_login(struct ibmvnic_adapter *adapter);
113 static void send_cap_queries(struct ibmvnic_adapter *adapter);
114 static int init_sub_crqs(struct ibmvnic_adapter *);
115 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
116 static int ibmvnic_init(struct ibmvnic_adapter *);
117 static void release_crq_queue(struct ibmvnic_adapter *);
118 
119 struct ibmvnic_stat {
120 	char name[ETH_GSTRING_LEN];
121 	int offset;
122 };
123 
124 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
125 			     offsetof(struct ibmvnic_statistics, stat))
126 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
127 
128 static const struct ibmvnic_stat ibmvnic_stats[] = {
129 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
130 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
131 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
132 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
133 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
134 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
135 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
136 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
137 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
138 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
139 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
140 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
141 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
142 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
143 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
144 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
145 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
146 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
147 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
148 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
149 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
150 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
151 };
152 
153 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
154 			  unsigned long length, unsigned long *number,
155 			  unsigned long *irq)
156 {
157 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
158 	long rc;
159 
160 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
161 	*number = retbuf[0];
162 	*irq = retbuf[1];
163 
164 	return rc;
165 }
166 
167 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
168 				struct ibmvnic_long_term_buff *ltb, int size)
169 {
170 	struct device *dev = &adapter->vdev->dev;
171 
172 	ltb->size = size;
173 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
174 				       GFP_KERNEL);
175 
176 	if (!ltb->buff) {
177 		dev_err(dev, "Couldn't alloc long term buffer\n");
178 		return -ENOMEM;
179 	}
180 	ltb->map_id = adapter->map_id;
181 	adapter->map_id++;
182 
183 	init_completion(&adapter->fw_done);
184 	send_request_map(adapter, ltb->addr,
185 			 ltb->size, ltb->map_id);
186 	wait_for_completion(&adapter->fw_done);
187 
188 	if (adapter->fw_done_rc) {
189 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
190 			adapter->fw_done_rc);
191 		return -1;
192 	}
193 	return 0;
194 }
195 
196 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
197 				struct ibmvnic_long_term_buff *ltb)
198 {
199 	struct device *dev = &adapter->vdev->dev;
200 
201 	if (!ltb->buff)
202 		return;
203 
204 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
205 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
206 		send_request_unmap(adapter, ltb->map_id);
207 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
208 }
209 
210 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
211 				struct ibmvnic_long_term_buff *ltb)
212 {
213 	memset(ltb->buff, 0, ltb->size);
214 
215 	init_completion(&adapter->fw_done);
216 	send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
217 	wait_for_completion(&adapter->fw_done);
218 
219 	if (adapter->fw_done_rc) {
220 		dev_info(&adapter->vdev->dev,
221 			 "Reset failed, attempting to free and reallocate buffer\n");
222 		free_long_term_buff(adapter, ltb);
223 		return alloc_long_term_buff(adapter, ltb, ltb->size);
224 	}
225 	return 0;
226 }
227 
228 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
229 {
230 	int i;
231 
232 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
233 	     i++)
234 		adapter->rx_pool[i].active = 0;
235 }
236 
237 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
238 			      struct ibmvnic_rx_pool *pool)
239 {
240 	int count = pool->size - atomic_read(&pool->available);
241 	struct device *dev = &adapter->vdev->dev;
242 	int buffers_added = 0;
243 	unsigned long lpar_rc;
244 	union sub_crq sub_crq;
245 	struct sk_buff *skb;
246 	unsigned int offset;
247 	dma_addr_t dma_addr;
248 	unsigned char *dst;
249 	u64 *handle_array;
250 	int shift = 0;
251 	int index;
252 	int i;
253 
254 	if (!pool->active)
255 		return;
256 
257 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
258 				      be32_to_cpu(adapter->login_rsp_buf->
259 				      off_rxadd_subcrqs));
260 
261 	for (i = 0; i < count; ++i) {
262 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
263 		if (!skb) {
264 			dev_err(dev, "Couldn't replenish rx buff\n");
265 			adapter->replenish_no_mem++;
266 			break;
267 		}
268 
269 		index = pool->free_map[pool->next_free];
270 
271 		if (pool->rx_buff[index].skb)
272 			dev_err(dev, "Inconsistent free_map!\n");
273 
274 		/* Copy the skb to the long term mapped DMA buffer */
275 		offset = index * pool->buff_size;
276 		dst = pool->long_term_buff.buff + offset;
277 		memset(dst, 0, pool->buff_size);
278 		dma_addr = pool->long_term_buff.addr + offset;
279 		pool->rx_buff[index].data = dst;
280 
281 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
282 		pool->rx_buff[index].dma = dma_addr;
283 		pool->rx_buff[index].skb = skb;
284 		pool->rx_buff[index].pool_index = pool->index;
285 		pool->rx_buff[index].size = pool->buff_size;
286 
287 		memset(&sub_crq, 0, sizeof(sub_crq));
288 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
289 		sub_crq.rx_add.correlator =
290 		    cpu_to_be64((u64)&pool->rx_buff[index]);
291 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
292 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
293 
294 		/* The length field of the sCRQ is defined to be 24 bits so the
295 		 * buffer size needs to be left shifted by a byte before it is
296 		 * converted to big endian to prevent the last byte from being
297 		 * truncated.
298 		 */
299 #ifdef __LITTLE_ENDIAN__
300 		shift = 8;
301 #endif
302 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
303 
304 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
305 				      &sub_crq);
306 		if (lpar_rc != H_SUCCESS)
307 			goto failure;
308 
309 		buffers_added++;
310 		adapter->replenish_add_buff_success++;
311 		pool->next_free = (pool->next_free + 1) % pool->size;
312 	}
313 	atomic_add(buffers_added, &pool->available);
314 	return;
315 
316 failure:
317 	dev_info(dev, "replenish pools failure\n");
318 	pool->free_map[pool->next_free] = index;
319 	pool->rx_buff[index].skb = NULL;
320 	if (!dma_mapping_error(dev, dma_addr))
321 		dma_unmap_single(dev, dma_addr, pool->buff_size,
322 				 DMA_FROM_DEVICE);
323 
324 	dev_kfree_skb_any(skb);
325 	adapter->replenish_add_buff_failure++;
326 	atomic_add(buffers_added, &pool->available);
327 
328 	if (lpar_rc == H_CLOSED) {
329 		/* Disable buffer pool replenishment and report carrier off if
330 		 * queue is closed. Firmware guarantees that a signal will
331 		 * be sent to the driver, triggering a reset.
332 		 */
333 		deactivate_rx_pools(adapter);
334 		netif_carrier_off(adapter->netdev);
335 	}
336 }
337 
338 static void replenish_pools(struct ibmvnic_adapter *adapter)
339 {
340 	int i;
341 
342 	adapter->replenish_task_cycles++;
343 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
344 	     i++) {
345 		if (adapter->rx_pool[i].active)
346 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
347 	}
348 }
349 
350 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
351 {
352 	kfree(adapter->tx_stats_buffers);
353 	kfree(adapter->rx_stats_buffers);
354 }
355 
356 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
357 {
358 	adapter->tx_stats_buffers =
359 				kcalloc(adapter->req_tx_queues,
360 					sizeof(struct ibmvnic_tx_queue_stats),
361 					GFP_KERNEL);
362 	if (!adapter->tx_stats_buffers)
363 		return -ENOMEM;
364 
365 	adapter->rx_stats_buffers =
366 				kcalloc(adapter->req_rx_queues,
367 					sizeof(struct ibmvnic_rx_queue_stats),
368 					GFP_KERNEL);
369 	if (!adapter->rx_stats_buffers)
370 		return -ENOMEM;
371 
372 	return 0;
373 }
374 
375 static void release_stats_token(struct ibmvnic_adapter *adapter)
376 {
377 	struct device *dev = &adapter->vdev->dev;
378 
379 	if (!adapter->stats_token)
380 		return;
381 
382 	dma_unmap_single(dev, adapter->stats_token,
383 			 sizeof(struct ibmvnic_statistics),
384 			 DMA_FROM_DEVICE);
385 	adapter->stats_token = 0;
386 }
387 
388 static int init_stats_token(struct ibmvnic_adapter *adapter)
389 {
390 	struct device *dev = &adapter->vdev->dev;
391 	dma_addr_t stok;
392 
393 	stok = dma_map_single(dev, &adapter->stats,
394 			      sizeof(struct ibmvnic_statistics),
395 			      DMA_FROM_DEVICE);
396 	if (dma_mapping_error(dev, stok)) {
397 		dev_err(dev, "Couldn't map stats buffer\n");
398 		return -1;
399 	}
400 
401 	adapter->stats_token = stok;
402 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
403 	return 0;
404 }
405 
406 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
407 {
408 	struct ibmvnic_rx_pool *rx_pool;
409 	int rx_scrqs;
410 	int i, j, rc;
411 
412 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
413 	for (i = 0; i < rx_scrqs; i++) {
414 		rx_pool = &adapter->rx_pool[i];
415 
416 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
417 
418 		rc = reset_long_term_buff(adapter, &rx_pool->long_term_buff);
419 		if (rc)
420 			return rc;
421 
422 		for (j = 0; j < rx_pool->size; j++)
423 			rx_pool->free_map[j] = j;
424 
425 		memset(rx_pool->rx_buff, 0,
426 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
427 
428 		atomic_set(&rx_pool->available, 0);
429 		rx_pool->next_alloc = 0;
430 		rx_pool->next_free = 0;
431 		rx_pool->active = 1;
432 	}
433 
434 	return 0;
435 }
436 
437 static void release_rx_pools(struct ibmvnic_adapter *adapter)
438 {
439 	struct ibmvnic_rx_pool *rx_pool;
440 	int rx_scrqs;
441 	int i, j;
442 
443 	if (!adapter->rx_pool)
444 		return;
445 
446 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
447 	for (i = 0; i < rx_scrqs; i++) {
448 		rx_pool = &adapter->rx_pool[i];
449 
450 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
451 
452 		kfree(rx_pool->free_map);
453 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
454 
455 		if (!rx_pool->rx_buff)
456 			continue;
457 
458 		for (j = 0; j < rx_pool->size; j++) {
459 			if (rx_pool->rx_buff[j].skb) {
460 				dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
461 				rx_pool->rx_buff[i].skb = NULL;
462 			}
463 		}
464 
465 		kfree(rx_pool->rx_buff);
466 	}
467 
468 	kfree(adapter->rx_pool);
469 	adapter->rx_pool = NULL;
470 }
471 
472 static int init_rx_pools(struct net_device *netdev)
473 {
474 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
475 	struct device *dev = &adapter->vdev->dev;
476 	struct ibmvnic_rx_pool *rx_pool;
477 	int rxadd_subcrqs;
478 	u64 *size_array;
479 	int i, j;
480 
481 	rxadd_subcrqs =
482 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
483 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
484 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
485 
486 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
487 				   sizeof(struct ibmvnic_rx_pool),
488 				   GFP_KERNEL);
489 	if (!adapter->rx_pool) {
490 		dev_err(dev, "Failed to allocate rx pools\n");
491 		return -1;
492 	}
493 
494 	for (i = 0; i < rxadd_subcrqs; i++) {
495 		rx_pool = &adapter->rx_pool[i];
496 
497 		netdev_dbg(adapter->netdev,
498 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
499 			   i, adapter->req_rx_add_entries_per_subcrq,
500 			   be64_to_cpu(size_array[i]));
501 
502 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
503 		rx_pool->index = i;
504 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
505 		rx_pool->active = 1;
506 
507 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
508 					    GFP_KERNEL);
509 		if (!rx_pool->free_map) {
510 			release_rx_pools(adapter);
511 			return -1;
512 		}
513 
514 		rx_pool->rx_buff = kcalloc(rx_pool->size,
515 					   sizeof(struct ibmvnic_rx_buff),
516 					   GFP_KERNEL);
517 		if (!rx_pool->rx_buff) {
518 			dev_err(dev, "Couldn't alloc rx buffers\n");
519 			release_rx_pools(adapter);
520 			return -1;
521 		}
522 
523 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
524 					 rx_pool->size * rx_pool->buff_size)) {
525 			release_rx_pools(adapter);
526 			return -1;
527 		}
528 
529 		for (j = 0; j < rx_pool->size; ++j)
530 			rx_pool->free_map[j] = j;
531 
532 		atomic_set(&rx_pool->available, 0);
533 		rx_pool->next_alloc = 0;
534 		rx_pool->next_free = 0;
535 	}
536 
537 	return 0;
538 }
539 
540 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
541 {
542 	struct ibmvnic_tx_pool *tx_pool;
543 	int tx_scrqs;
544 	int i, j, rc;
545 
546 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
547 	for (i = 0; i < tx_scrqs; i++) {
548 		netdev_dbg(adapter->netdev, "Re-setting tx_pool[%d]\n", i);
549 
550 		tx_pool = &adapter->tx_pool[i];
551 
552 		rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
553 		if (rc)
554 			return rc;
555 
556 		memset(tx_pool->tx_buff, 0,
557 		       adapter->req_tx_entries_per_subcrq *
558 		       sizeof(struct ibmvnic_tx_buff));
559 
560 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
561 			tx_pool->free_map[j] = j;
562 
563 		tx_pool->consumer_index = 0;
564 		tx_pool->producer_index = 0;
565 	}
566 
567 	return 0;
568 }
569 
570 static void release_tx_pools(struct ibmvnic_adapter *adapter)
571 {
572 	struct ibmvnic_tx_pool *tx_pool;
573 	int i, tx_scrqs;
574 
575 	if (!adapter->tx_pool)
576 		return;
577 
578 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
579 	for (i = 0; i < tx_scrqs; i++) {
580 		netdev_dbg(adapter->netdev, "Releasing tx_pool[%d]\n", i);
581 		tx_pool = &adapter->tx_pool[i];
582 		kfree(tx_pool->tx_buff);
583 		free_long_term_buff(adapter, &tx_pool->long_term_buff);
584 		kfree(tx_pool->free_map);
585 	}
586 
587 	kfree(adapter->tx_pool);
588 	adapter->tx_pool = NULL;
589 }
590 
591 static int init_tx_pools(struct net_device *netdev)
592 {
593 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
594 	struct device *dev = &adapter->vdev->dev;
595 	struct ibmvnic_tx_pool *tx_pool;
596 	int tx_subcrqs;
597 	int i, j;
598 
599 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
600 	adapter->tx_pool = kcalloc(tx_subcrqs,
601 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
602 	if (!adapter->tx_pool)
603 		return -1;
604 
605 	for (i = 0; i < tx_subcrqs; i++) {
606 		tx_pool = &adapter->tx_pool[i];
607 
608 		netdev_dbg(adapter->netdev,
609 			   "Initializing tx_pool[%d], %lld buffs\n",
610 			   i, adapter->req_tx_entries_per_subcrq);
611 
612 		tx_pool->tx_buff = kcalloc(adapter->req_tx_entries_per_subcrq,
613 					   sizeof(struct ibmvnic_tx_buff),
614 					   GFP_KERNEL);
615 		if (!tx_pool->tx_buff) {
616 			dev_err(dev, "tx pool buffer allocation failed\n");
617 			release_tx_pools(adapter);
618 			return -1;
619 		}
620 
621 		if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
622 					 adapter->req_tx_entries_per_subcrq *
623 					 adapter->req_mtu)) {
624 			release_tx_pools(adapter);
625 			return -1;
626 		}
627 
628 		tx_pool->free_map = kcalloc(adapter->req_tx_entries_per_subcrq,
629 					    sizeof(int), GFP_KERNEL);
630 		if (!tx_pool->free_map) {
631 			release_tx_pools(adapter);
632 			return -1;
633 		}
634 
635 		for (j = 0; j < adapter->req_tx_entries_per_subcrq; j++)
636 			tx_pool->free_map[j] = j;
637 
638 		tx_pool->consumer_index = 0;
639 		tx_pool->producer_index = 0;
640 	}
641 
642 	return 0;
643 }
644 
645 static void release_error_buffers(struct ibmvnic_adapter *adapter)
646 {
647 	struct device *dev = &adapter->vdev->dev;
648 	struct ibmvnic_error_buff *error_buff, *tmp;
649 	unsigned long flags;
650 
651 	spin_lock_irqsave(&adapter->error_list_lock, flags);
652 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list) {
653 		list_del(&error_buff->list);
654 		dma_unmap_single(dev, error_buff->dma, error_buff->len,
655 				 DMA_FROM_DEVICE);
656 		kfree(error_buff->buff);
657 		kfree(error_buff);
658 	}
659 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
660 }
661 
662 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
663 {
664 	int i;
665 
666 	if (adapter->napi_enabled)
667 		return;
668 
669 	for (i = 0; i < adapter->req_rx_queues; i++)
670 		napi_enable(&adapter->napi[i]);
671 
672 	adapter->napi_enabled = true;
673 }
674 
675 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
676 {
677 	int i;
678 
679 	if (!adapter->napi_enabled)
680 		return;
681 
682 	for (i = 0; i < adapter->req_rx_queues; i++) {
683 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
684 		napi_disable(&adapter->napi[i]);
685 	}
686 
687 	adapter->napi_enabled = false;
688 }
689 
690 static int ibmvnic_login(struct net_device *netdev)
691 {
692 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
693 	unsigned long timeout = msecs_to_jiffies(30000);
694 	struct device *dev = &adapter->vdev->dev;
695 	int rc;
696 
697 	do {
698 		if (adapter->renegotiate) {
699 			adapter->renegotiate = false;
700 			release_sub_crqs(adapter);
701 
702 			reinit_completion(&adapter->init_done);
703 			send_cap_queries(adapter);
704 			if (!wait_for_completion_timeout(&adapter->init_done,
705 							 timeout)) {
706 				dev_err(dev, "Capabilities query timeout\n");
707 				return -1;
708 			}
709 			rc = init_sub_crqs(adapter);
710 			if (rc) {
711 				dev_err(dev,
712 					"Initialization of SCRQ's failed\n");
713 				return -1;
714 			}
715 			rc = init_sub_crq_irqs(adapter);
716 			if (rc) {
717 				dev_err(dev,
718 					"Initialization of SCRQ's irqs failed\n");
719 				return -1;
720 			}
721 		}
722 
723 		reinit_completion(&adapter->init_done);
724 		send_login(adapter);
725 		if (!wait_for_completion_timeout(&adapter->init_done,
726 						 timeout)) {
727 			dev_err(dev, "Login timeout\n");
728 			return -1;
729 		}
730 	} while (adapter->renegotiate);
731 
732 	return 0;
733 }
734 
735 static void release_resources(struct ibmvnic_adapter *adapter)
736 {
737 	int i;
738 
739 	release_tx_pools(adapter);
740 	release_rx_pools(adapter);
741 
742 	release_stats_token(adapter);
743 	release_stats_buffers(adapter);
744 	release_error_buffers(adapter);
745 
746 	if (adapter->napi) {
747 		for (i = 0; i < adapter->req_rx_queues; i++) {
748 			if (&adapter->napi[i]) {
749 				netdev_dbg(adapter->netdev,
750 					   "Releasing napi[%d]\n", i);
751 				netif_napi_del(&adapter->napi[i]);
752 			}
753 		}
754 	}
755 }
756 
757 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
758 {
759 	struct net_device *netdev = adapter->netdev;
760 	unsigned long timeout = msecs_to_jiffies(30000);
761 	union ibmvnic_crq crq;
762 	bool resend;
763 	int rc;
764 
765 	netdev_dbg(netdev, "setting link state %d\n", link_state);
766 
767 	memset(&crq, 0, sizeof(crq));
768 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
769 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
770 	crq.logical_link_state.link_state = link_state;
771 
772 	do {
773 		resend = false;
774 
775 		reinit_completion(&adapter->init_done);
776 		rc = ibmvnic_send_crq(adapter, &crq);
777 		if (rc) {
778 			netdev_err(netdev, "Failed to set link state\n");
779 			return rc;
780 		}
781 
782 		if (!wait_for_completion_timeout(&adapter->init_done,
783 						 timeout)) {
784 			netdev_err(netdev, "timeout setting link state\n");
785 			return -1;
786 		}
787 
788 		if (adapter->init_done_rc == 1) {
789 			/* Partuial success, delay and re-send */
790 			mdelay(1000);
791 			resend = true;
792 		}
793 	} while (resend);
794 
795 	return 0;
796 }
797 
798 static int set_real_num_queues(struct net_device *netdev)
799 {
800 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
801 	int rc;
802 
803 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
804 		   adapter->req_tx_queues, adapter->req_rx_queues);
805 
806 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
807 	if (rc) {
808 		netdev_err(netdev, "failed to set the number of tx queues\n");
809 		return rc;
810 	}
811 
812 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
813 	if (rc)
814 		netdev_err(netdev, "failed to set the number of rx queues\n");
815 
816 	return rc;
817 }
818 
819 static int init_resources(struct ibmvnic_adapter *adapter)
820 {
821 	struct net_device *netdev = adapter->netdev;
822 	int i, rc;
823 
824 	rc = set_real_num_queues(netdev);
825 	if (rc)
826 		return rc;
827 
828 	rc = init_stats_buffers(adapter);
829 	if (rc)
830 		return rc;
831 
832 	rc = init_stats_token(adapter);
833 	if (rc)
834 		return rc;
835 
836 	adapter->map_id = 1;
837 	adapter->napi = kcalloc(adapter->req_rx_queues,
838 				sizeof(struct napi_struct), GFP_KERNEL);
839 	if (!adapter->napi)
840 		return -ENOMEM;
841 
842 	for (i = 0; i < adapter->req_rx_queues; i++) {
843 		netdev_dbg(netdev, "Adding napi[%d]\n", i);
844 		netif_napi_add(netdev, &adapter->napi[i], ibmvnic_poll,
845 			       NAPI_POLL_WEIGHT);
846 	}
847 
848 	send_map_query(adapter);
849 
850 	rc = init_rx_pools(netdev);
851 	if (rc)
852 		return rc;
853 
854 	rc = init_tx_pools(netdev);
855 	return rc;
856 }
857 
858 static int __ibmvnic_open(struct net_device *netdev)
859 {
860 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
861 	enum vnic_state prev_state = adapter->state;
862 	int i, rc;
863 
864 	adapter->state = VNIC_OPENING;
865 	replenish_pools(adapter);
866 	ibmvnic_napi_enable(adapter);
867 
868 	/* We're ready to receive frames, enable the sub-crq interrupts and
869 	 * set the logical link state to up
870 	 */
871 	for (i = 0; i < adapter->req_rx_queues; i++) {
872 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
873 		if (prev_state == VNIC_CLOSED)
874 			enable_irq(adapter->rx_scrq[i]->irq);
875 		else
876 			enable_scrq_irq(adapter, adapter->rx_scrq[i]);
877 	}
878 
879 	for (i = 0; i < adapter->req_tx_queues; i++) {
880 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
881 		if (prev_state == VNIC_CLOSED)
882 			enable_irq(adapter->tx_scrq[i]->irq);
883 		else
884 			enable_scrq_irq(adapter, adapter->tx_scrq[i]);
885 	}
886 
887 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
888 	if (rc) {
889 		for (i = 0; i < adapter->req_rx_queues; i++)
890 			napi_disable(&adapter->napi[i]);
891 		release_resources(adapter);
892 		return rc;
893 	}
894 
895 	netif_tx_start_all_queues(netdev);
896 
897 	if (prev_state == VNIC_CLOSED) {
898 		for (i = 0; i < adapter->req_rx_queues; i++)
899 			napi_schedule(&adapter->napi[i]);
900 	}
901 
902 	adapter->state = VNIC_OPEN;
903 	return rc;
904 }
905 
906 static int ibmvnic_open(struct net_device *netdev)
907 {
908 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
909 	int rc;
910 
911 	mutex_lock(&adapter->reset_lock);
912 
913 	if (adapter->state != VNIC_CLOSED) {
914 		rc = ibmvnic_login(netdev);
915 		if (rc) {
916 			mutex_unlock(&adapter->reset_lock);
917 			return rc;
918 		}
919 
920 		rc = init_resources(adapter);
921 		if (rc) {
922 			netdev_err(netdev, "failed to initialize resources\n");
923 			release_resources(adapter);
924 			mutex_unlock(&adapter->reset_lock);
925 			return rc;
926 		}
927 	}
928 
929 	rc = __ibmvnic_open(netdev);
930 	netif_carrier_on(netdev);
931 	mutex_unlock(&adapter->reset_lock);
932 
933 	return rc;
934 }
935 
936 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
937 {
938 	struct ibmvnic_tx_pool *tx_pool;
939 	u64 tx_entries;
940 	int tx_scrqs;
941 	int i, j;
942 
943 	if (!adapter->tx_pool)
944 		return;
945 
946 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
947 	tx_entries = adapter->req_tx_entries_per_subcrq;
948 
949 	/* Free any remaining skbs in the tx buffer pools */
950 	for (i = 0; i < tx_scrqs; i++) {
951 		tx_pool = &adapter->tx_pool[i];
952 		if (!tx_pool)
953 			continue;
954 
955 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
956 		for (j = 0; j < tx_entries; j++) {
957 			if (tx_pool->tx_buff[j].skb) {
958 				dev_kfree_skb_any(tx_pool->tx_buff[j].skb);
959 				tx_pool->tx_buff[j].skb = NULL;
960 			}
961 		}
962 	}
963 }
964 
965 static int __ibmvnic_close(struct net_device *netdev)
966 {
967 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
968 	int rc = 0;
969 	int i;
970 
971 	adapter->state = VNIC_CLOSING;
972 
973 	/* ensure that transmissions are stopped if called by do_reset */
974 	if (adapter->resetting)
975 		netif_tx_disable(netdev);
976 	else
977 		netif_tx_stop_all_queues(netdev);
978 
979 	ibmvnic_napi_disable(adapter);
980 
981 	if (adapter->tx_scrq) {
982 		for (i = 0; i < adapter->req_tx_queues; i++)
983 			if (adapter->tx_scrq[i]->irq) {
984 				netdev_dbg(adapter->netdev,
985 					   "Disabling tx_scrq[%d] irq\n", i);
986 				disable_irq(adapter->tx_scrq[i]->irq);
987 			}
988 	}
989 
990 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
991 	if (rc)
992 		return rc;
993 
994 	if (adapter->rx_scrq) {
995 		for (i = 0; i < adapter->req_rx_queues; i++) {
996 			int retries = 10;
997 
998 			while (pending_scrq(adapter, adapter->rx_scrq[i])) {
999 				retries--;
1000 				mdelay(100);
1001 
1002 				if (retries == 0)
1003 					break;
1004 			}
1005 
1006 			if (adapter->rx_scrq[i]->irq) {
1007 				netdev_dbg(adapter->netdev,
1008 					   "Disabling rx_scrq[%d] irq\n", i);
1009 				disable_irq(adapter->rx_scrq[i]->irq);
1010 			}
1011 		}
1012 	}
1013 
1014 	clean_tx_pools(adapter);
1015 	adapter->state = VNIC_CLOSED;
1016 	return rc;
1017 }
1018 
1019 static int ibmvnic_close(struct net_device *netdev)
1020 {
1021 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1022 	int rc;
1023 
1024 	mutex_lock(&adapter->reset_lock);
1025 	rc = __ibmvnic_close(netdev);
1026 	mutex_unlock(&adapter->reset_lock);
1027 
1028 	return rc;
1029 }
1030 
1031 /**
1032  * build_hdr_data - creates L2/L3/L4 header data buffer
1033  * @hdr_field - bitfield determining needed headers
1034  * @skb - socket buffer
1035  * @hdr_len - array of header lengths
1036  * @tot_len - total length of data
1037  *
1038  * Reads hdr_field to determine which headers are needed by firmware.
1039  * Builds a buffer containing these headers.  Saves individual header
1040  * lengths and total buffer length to be used to build descriptors.
1041  */
1042 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1043 			  int *hdr_len, u8 *hdr_data)
1044 {
1045 	int len = 0;
1046 	u8 *hdr;
1047 
1048 	hdr_len[0] = sizeof(struct ethhdr);
1049 
1050 	if (skb->protocol == htons(ETH_P_IP)) {
1051 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1052 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1053 			hdr_len[2] = tcp_hdrlen(skb);
1054 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1055 			hdr_len[2] = sizeof(struct udphdr);
1056 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1057 		hdr_len[1] = sizeof(struct ipv6hdr);
1058 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1059 			hdr_len[2] = tcp_hdrlen(skb);
1060 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1061 			hdr_len[2] = sizeof(struct udphdr);
1062 	}
1063 
1064 	memset(hdr_data, 0, 120);
1065 	if ((hdr_field >> 6) & 1) {
1066 		hdr = skb_mac_header(skb);
1067 		memcpy(hdr_data, hdr, hdr_len[0]);
1068 		len += hdr_len[0];
1069 	}
1070 
1071 	if ((hdr_field >> 5) & 1) {
1072 		hdr = skb_network_header(skb);
1073 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1074 		len += hdr_len[1];
1075 	}
1076 
1077 	if ((hdr_field >> 4) & 1) {
1078 		hdr = skb_transport_header(skb);
1079 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1080 		len += hdr_len[2];
1081 	}
1082 	return len;
1083 }
1084 
1085 /**
1086  * create_hdr_descs - create header and header extension descriptors
1087  * @hdr_field - bitfield determining needed headers
1088  * @data - buffer containing header data
1089  * @len - length of data buffer
1090  * @hdr_len - array of individual header lengths
1091  * @scrq_arr - descriptor array
1092  *
1093  * Creates header and, if needed, header extension descriptors and
1094  * places them in a descriptor array, scrq_arr
1095  */
1096 
1097 static void create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1098 			     union sub_crq *scrq_arr)
1099 {
1100 	union sub_crq hdr_desc;
1101 	int tmp_len = len;
1102 	u8 *data, *cur;
1103 	int tmp;
1104 
1105 	while (tmp_len > 0) {
1106 		cur = hdr_data + len - tmp_len;
1107 
1108 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1109 		if (cur != hdr_data) {
1110 			data = hdr_desc.hdr_ext.data;
1111 			tmp = tmp_len > 29 ? 29 : tmp_len;
1112 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1113 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1114 			hdr_desc.hdr_ext.len = tmp;
1115 		} else {
1116 			data = hdr_desc.hdr.data;
1117 			tmp = tmp_len > 24 ? 24 : tmp_len;
1118 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1119 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1120 			hdr_desc.hdr.len = tmp;
1121 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1122 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1123 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1124 			hdr_desc.hdr.flag = hdr_field << 1;
1125 		}
1126 		memcpy(data, cur, tmp);
1127 		tmp_len -= tmp;
1128 		*scrq_arr = hdr_desc;
1129 		scrq_arr++;
1130 	}
1131 }
1132 
1133 /**
1134  * build_hdr_descs_arr - build a header descriptor array
1135  * @skb - socket buffer
1136  * @num_entries - number of descriptors to be sent
1137  * @subcrq - first TX descriptor
1138  * @hdr_field - bit field determining which headers will be sent
1139  *
1140  * This function will build a TX descriptor array with applicable
1141  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1142  */
1143 
1144 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1145 				int *num_entries, u8 hdr_field)
1146 {
1147 	int hdr_len[3] = {0, 0, 0};
1148 	int tot_len, len;
1149 	u8 *hdr_data = txbuff->hdr_data;
1150 
1151 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1152 				 txbuff->hdr_data);
1153 	len = tot_len;
1154 	len -= 24;
1155 	if (len > 0)
1156 		num_entries += len % 29 ? len / 29 + 1 : len / 29;
1157 	create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1158 			 txbuff->indir_arr + 1);
1159 }
1160 
1161 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1162 {
1163 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1164 	int queue_num = skb_get_queue_mapping(skb);
1165 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1166 	struct device *dev = &adapter->vdev->dev;
1167 	struct ibmvnic_tx_buff *tx_buff = NULL;
1168 	struct ibmvnic_sub_crq_queue *tx_scrq;
1169 	struct ibmvnic_tx_pool *tx_pool;
1170 	unsigned int tx_send_failed = 0;
1171 	unsigned int tx_map_failed = 0;
1172 	unsigned int tx_dropped = 0;
1173 	unsigned int tx_packets = 0;
1174 	unsigned int tx_bytes = 0;
1175 	dma_addr_t data_dma_addr;
1176 	struct netdev_queue *txq;
1177 	unsigned long lpar_rc;
1178 	union sub_crq tx_crq;
1179 	unsigned int offset;
1180 	int num_entries = 1;
1181 	unsigned char *dst;
1182 	u64 *handle_array;
1183 	int index = 0;
1184 	int ret = 0;
1185 
1186 	if (adapter->resetting) {
1187 		if (!netif_subqueue_stopped(netdev, skb))
1188 			netif_stop_subqueue(netdev, queue_num);
1189 		dev_kfree_skb_any(skb);
1190 
1191 		tx_send_failed++;
1192 		tx_dropped++;
1193 		ret = NETDEV_TX_OK;
1194 		goto out;
1195 	}
1196 
1197 	tx_pool = &adapter->tx_pool[queue_num];
1198 	tx_scrq = adapter->tx_scrq[queue_num];
1199 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1200 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1201 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1202 
1203 	index = tx_pool->free_map[tx_pool->consumer_index];
1204 	offset = index * adapter->req_mtu;
1205 	dst = tx_pool->long_term_buff.buff + offset;
1206 	memset(dst, 0, adapter->req_mtu);
1207 	skb_copy_from_linear_data(skb, dst, skb->len);
1208 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1209 
1210 	tx_pool->consumer_index =
1211 	    (tx_pool->consumer_index + 1) %
1212 		adapter->req_tx_entries_per_subcrq;
1213 
1214 	tx_buff = &tx_pool->tx_buff[index];
1215 	tx_buff->skb = skb;
1216 	tx_buff->data_dma[0] = data_dma_addr;
1217 	tx_buff->data_len[0] = skb->len;
1218 	tx_buff->index = index;
1219 	tx_buff->pool_index = queue_num;
1220 	tx_buff->last_frag = true;
1221 
1222 	memset(&tx_crq, 0, sizeof(tx_crq));
1223 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1224 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1225 	tx_crq.v1.n_crq_elem = 1;
1226 	tx_crq.v1.n_sge = 1;
1227 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1228 	tx_crq.v1.correlator = cpu_to_be32(index);
1229 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1230 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1231 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1232 
1233 	if (adapter->vlan_header_insertion) {
1234 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1235 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1236 	}
1237 
1238 	if (skb->protocol == htons(ETH_P_IP)) {
1239 		if (ip_hdr(skb)->version == 4)
1240 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1241 		else if (ip_hdr(skb)->version == 6)
1242 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1243 
1244 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1245 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1246 		else if (ip_hdr(skb)->protocol != IPPROTO_TCP)
1247 			tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1248 	}
1249 
1250 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1251 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1252 		hdrs += 2;
1253 	}
1254 	/* determine if l2/3/4 headers are sent to firmware */
1255 	if ((*hdrs >> 7) & 1 &&
1256 	    (skb->protocol == htons(ETH_P_IP) ||
1257 	     skb->protocol == htons(ETH_P_IPV6))) {
1258 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1259 		tx_crq.v1.n_crq_elem = num_entries;
1260 		tx_buff->indir_arr[0] = tx_crq;
1261 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1262 						    sizeof(tx_buff->indir_arr),
1263 						    DMA_TO_DEVICE);
1264 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1265 			dev_kfree_skb_any(skb);
1266 			tx_buff->skb = NULL;
1267 			if (!firmware_has_feature(FW_FEATURE_CMO))
1268 				dev_err(dev, "tx: unable to map descriptor array\n");
1269 			tx_map_failed++;
1270 			tx_dropped++;
1271 			ret = NETDEV_TX_OK;
1272 			goto out;
1273 		}
1274 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1275 					       (u64)tx_buff->indir_dma,
1276 					       (u64)num_entries);
1277 	} else {
1278 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1279 				      &tx_crq);
1280 	}
1281 	if (lpar_rc != H_SUCCESS) {
1282 		dev_err(dev, "tx failed with code %ld\n", lpar_rc);
1283 
1284 		if (tx_pool->consumer_index == 0)
1285 			tx_pool->consumer_index =
1286 				adapter->req_tx_entries_per_subcrq - 1;
1287 		else
1288 			tx_pool->consumer_index--;
1289 
1290 		dev_kfree_skb_any(skb);
1291 		tx_buff->skb = NULL;
1292 
1293 		if (lpar_rc == H_CLOSED) {
1294 			/* Disable TX and report carrier off if queue is closed.
1295 			 * Firmware guarantees that a signal will be sent to the
1296 			 * driver, triggering a reset or some other action.
1297 			 */
1298 			netif_tx_stop_all_queues(netdev);
1299 			netif_carrier_off(netdev);
1300 		}
1301 
1302 		tx_send_failed++;
1303 		tx_dropped++;
1304 		ret = NETDEV_TX_OK;
1305 		goto out;
1306 	}
1307 
1308 	if (atomic_inc_return(&tx_scrq->used)
1309 					>= adapter->req_tx_entries_per_subcrq) {
1310 		netdev_info(netdev, "Stopping queue %d\n", queue_num);
1311 		netif_stop_subqueue(netdev, queue_num);
1312 	}
1313 
1314 	tx_packets++;
1315 	tx_bytes += skb->len;
1316 	txq->trans_start = jiffies;
1317 	ret = NETDEV_TX_OK;
1318 
1319 out:
1320 	netdev->stats.tx_dropped += tx_dropped;
1321 	netdev->stats.tx_bytes += tx_bytes;
1322 	netdev->stats.tx_packets += tx_packets;
1323 	adapter->tx_send_failed += tx_send_failed;
1324 	adapter->tx_map_failed += tx_map_failed;
1325 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1326 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1327 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1328 
1329 	return ret;
1330 }
1331 
1332 static void ibmvnic_set_multi(struct net_device *netdev)
1333 {
1334 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1335 	struct netdev_hw_addr *ha;
1336 	union ibmvnic_crq crq;
1337 
1338 	memset(&crq, 0, sizeof(crq));
1339 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1340 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1341 
1342 	if (netdev->flags & IFF_PROMISC) {
1343 		if (!adapter->promisc_supported)
1344 			return;
1345 	} else {
1346 		if (netdev->flags & IFF_ALLMULTI) {
1347 			/* Accept all multicast */
1348 			memset(&crq, 0, sizeof(crq));
1349 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1350 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1351 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1352 			ibmvnic_send_crq(adapter, &crq);
1353 		} else if (netdev_mc_empty(netdev)) {
1354 			/* Reject all multicast */
1355 			memset(&crq, 0, sizeof(crq));
1356 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1357 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1358 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1359 			ibmvnic_send_crq(adapter, &crq);
1360 		} else {
1361 			/* Accept one or more multicast(s) */
1362 			netdev_for_each_mc_addr(ha, netdev) {
1363 				memset(&crq, 0, sizeof(crq));
1364 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1365 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1366 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1367 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1368 						ha->addr);
1369 				ibmvnic_send_crq(adapter, &crq);
1370 			}
1371 		}
1372 	}
1373 }
1374 
1375 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1376 {
1377 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1378 	struct sockaddr *addr = p;
1379 	union ibmvnic_crq crq;
1380 
1381 	if (!is_valid_ether_addr(addr->sa_data))
1382 		return -EADDRNOTAVAIL;
1383 
1384 	memset(&crq, 0, sizeof(crq));
1385 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1386 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1387 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1388 	ibmvnic_send_crq(adapter, &crq);
1389 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1390 	return 0;
1391 }
1392 
1393 /**
1394  * do_reset returns zero if we are able to keep processing reset events, or
1395  * non-zero if we hit a fatal error and must halt.
1396  */
1397 static int do_reset(struct ibmvnic_adapter *adapter,
1398 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1399 {
1400 	struct net_device *netdev = adapter->netdev;
1401 	int i, rc;
1402 
1403 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1404 		   rwi->reset_reason);
1405 
1406 	netif_carrier_off(netdev);
1407 	adapter->reset_reason = rwi->reset_reason;
1408 
1409 	if (rwi->reset_reason == VNIC_RESET_MOBILITY) {
1410 		rc = ibmvnic_reenable_crq_queue(adapter);
1411 		if (rc)
1412 			return 0;
1413 	}
1414 
1415 	rc = __ibmvnic_close(netdev);
1416 	if (rc)
1417 		return rc;
1418 
1419 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1420 		/* remove the closed state so when we call open it appears
1421 		 * we are coming from the probed state.
1422 		 */
1423 		adapter->state = VNIC_PROBED;
1424 
1425 		rc = ibmvnic_init(adapter);
1426 		if (rc)
1427 			return 0;
1428 
1429 		/* If the adapter was in PROBE state prior to the reset,
1430 		 * exit here.
1431 		 */
1432 		if (reset_state == VNIC_PROBED)
1433 			return 0;
1434 
1435 		rc = ibmvnic_login(netdev);
1436 		if (rc) {
1437 			adapter->state = VNIC_PROBED;
1438 			return 0;
1439 		}
1440 
1441 		rc = reset_tx_pools(adapter);
1442 		if (rc)
1443 			return rc;
1444 
1445 		rc = reset_rx_pools(adapter);
1446 		if (rc)
1447 			return rc;
1448 
1449 		if (reset_state == VNIC_CLOSED)
1450 			return 0;
1451 	}
1452 
1453 	rc = __ibmvnic_open(netdev);
1454 	if (rc) {
1455 		if (list_empty(&adapter->rwi_list))
1456 			adapter->state = VNIC_CLOSED;
1457 		else
1458 			adapter->state = reset_state;
1459 
1460 		return 0;
1461 	}
1462 
1463 	netif_carrier_on(netdev);
1464 
1465 	/* kick napi */
1466 	for (i = 0; i < adapter->req_rx_queues; i++)
1467 		napi_schedule(&adapter->napi[i]);
1468 
1469 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
1470 		netdev_notify_peers(netdev);
1471 
1472 	return 0;
1473 }
1474 
1475 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1476 {
1477 	struct ibmvnic_rwi *rwi;
1478 
1479 	mutex_lock(&adapter->rwi_lock);
1480 
1481 	if (!list_empty(&adapter->rwi_list)) {
1482 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1483 				       list);
1484 		list_del(&rwi->list);
1485 	} else {
1486 		rwi = NULL;
1487 	}
1488 
1489 	mutex_unlock(&adapter->rwi_lock);
1490 	return rwi;
1491 }
1492 
1493 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1494 {
1495 	struct ibmvnic_rwi *rwi;
1496 
1497 	rwi = get_next_rwi(adapter);
1498 	while (rwi) {
1499 		kfree(rwi);
1500 		rwi = get_next_rwi(adapter);
1501 	}
1502 }
1503 
1504 static void __ibmvnic_reset(struct work_struct *work)
1505 {
1506 	struct ibmvnic_rwi *rwi;
1507 	struct ibmvnic_adapter *adapter;
1508 	struct net_device *netdev;
1509 	u32 reset_state;
1510 	int rc;
1511 
1512 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1513 	netdev = adapter->netdev;
1514 
1515 	mutex_lock(&adapter->reset_lock);
1516 	adapter->resetting = true;
1517 	reset_state = adapter->state;
1518 
1519 	rwi = get_next_rwi(adapter);
1520 	while (rwi) {
1521 		rc = do_reset(adapter, rwi, reset_state);
1522 		kfree(rwi);
1523 		if (rc)
1524 			break;
1525 
1526 		rwi = get_next_rwi(adapter);
1527 	}
1528 
1529 	if (rc) {
1530 		netdev_dbg(adapter->netdev, "Reset failed\n");
1531 		free_all_rwi(adapter);
1532 		mutex_unlock(&adapter->reset_lock);
1533 		return;
1534 	}
1535 
1536 	adapter->resetting = false;
1537 	mutex_unlock(&adapter->reset_lock);
1538 }
1539 
1540 static void ibmvnic_reset(struct ibmvnic_adapter *adapter,
1541 			  enum ibmvnic_reset_reason reason)
1542 {
1543 	struct ibmvnic_rwi *rwi, *tmp;
1544 	struct net_device *netdev = adapter->netdev;
1545 	struct list_head *entry;
1546 
1547 	if (adapter->state == VNIC_REMOVING ||
1548 	    adapter->state == VNIC_REMOVED) {
1549 		netdev_dbg(netdev, "Adapter removing, skipping reset\n");
1550 		return;
1551 	}
1552 
1553 	if (adapter->state == VNIC_PROBING) {
1554 		netdev_warn(netdev, "Adapter reset during probe\n");
1555 		adapter->init_done_rc = EAGAIN;
1556 		return;
1557 	}
1558 
1559 	mutex_lock(&adapter->rwi_lock);
1560 
1561 	list_for_each(entry, &adapter->rwi_list) {
1562 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
1563 		if (tmp->reset_reason == reason) {
1564 			netdev_dbg(netdev, "Skipping matching reset\n");
1565 			mutex_unlock(&adapter->rwi_lock);
1566 			return;
1567 		}
1568 	}
1569 
1570 	rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
1571 	if (!rwi) {
1572 		mutex_unlock(&adapter->rwi_lock);
1573 		ibmvnic_close(netdev);
1574 		return;
1575 	}
1576 
1577 	rwi->reset_reason = reason;
1578 	list_add_tail(&rwi->list, &adapter->rwi_list);
1579 	mutex_unlock(&adapter->rwi_lock);
1580 
1581 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
1582 	schedule_work(&adapter->ibmvnic_reset);
1583 }
1584 
1585 static void ibmvnic_tx_timeout(struct net_device *dev)
1586 {
1587 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1588 
1589 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
1590 }
1591 
1592 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
1593 				  struct ibmvnic_rx_buff *rx_buff)
1594 {
1595 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
1596 
1597 	rx_buff->skb = NULL;
1598 
1599 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
1600 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
1601 
1602 	atomic_dec(&pool->available);
1603 }
1604 
1605 static int ibmvnic_poll(struct napi_struct *napi, int budget)
1606 {
1607 	struct net_device *netdev = napi->dev;
1608 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1609 	int scrq_num = (int)(napi - adapter->napi);
1610 	int frames_processed = 0;
1611 
1612 restart_poll:
1613 	while (frames_processed < budget) {
1614 		struct sk_buff *skb;
1615 		struct ibmvnic_rx_buff *rx_buff;
1616 		union sub_crq *next;
1617 		u32 length;
1618 		u16 offset;
1619 		u8 flags = 0;
1620 
1621 		if (unlikely(adapter->resetting)) {
1622 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1623 			napi_complete_done(napi, frames_processed);
1624 			return frames_processed;
1625 		}
1626 
1627 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
1628 			break;
1629 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
1630 		rx_buff =
1631 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
1632 							  rx_comp.correlator);
1633 		/* do error checking */
1634 		if (next->rx_comp.rc) {
1635 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
1636 				   be16_to_cpu(next->rx_comp.rc));
1637 			/* free the entry */
1638 			next->rx_comp.first = 0;
1639 			remove_buff_from_pool(adapter, rx_buff);
1640 			continue;
1641 		}
1642 
1643 		length = be32_to_cpu(next->rx_comp.len);
1644 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
1645 		flags = next->rx_comp.flags;
1646 		skb = rx_buff->skb;
1647 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
1648 					length);
1649 
1650 		/* VLAN Header has been stripped by the system firmware and
1651 		 * needs to be inserted by the driver
1652 		 */
1653 		if (adapter->rx_vlan_header_insertion &&
1654 		    (flags & IBMVNIC_VLAN_STRIPPED))
1655 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
1656 					       ntohs(next->rx_comp.vlan_tci));
1657 
1658 		/* free the entry */
1659 		next->rx_comp.first = 0;
1660 		remove_buff_from_pool(adapter, rx_buff);
1661 
1662 		skb_put(skb, length);
1663 		skb->protocol = eth_type_trans(skb, netdev);
1664 		skb_record_rx_queue(skb, scrq_num);
1665 
1666 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
1667 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
1668 			skb->ip_summed = CHECKSUM_UNNECESSARY;
1669 		}
1670 
1671 		length = skb->len;
1672 		napi_gro_receive(napi, skb); /* send it up */
1673 		netdev->stats.rx_packets++;
1674 		netdev->stats.rx_bytes += length;
1675 		adapter->rx_stats_buffers[scrq_num].packets++;
1676 		adapter->rx_stats_buffers[scrq_num].bytes += length;
1677 		frames_processed++;
1678 	}
1679 
1680 	if (adapter->state != VNIC_CLOSING)
1681 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
1682 
1683 	if (frames_processed < budget) {
1684 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1685 		napi_complete_done(napi, frames_processed);
1686 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
1687 		    napi_reschedule(napi)) {
1688 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
1689 			goto restart_poll;
1690 		}
1691 	}
1692 	return frames_processed;
1693 }
1694 
1695 #ifdef CONFIG_NET_POLL_CONTROLLER
1696 static void ibmvnic_netpoll_controller(struct net_device *dev)
1697 {
1698 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1699 	int i;
1700 
1701 	replenish_pools(netdev_priv(dev));
1702 	for (i = 0; i < adapter->req_rx_queues; i++)
1703 		ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
1704 				     adapter->rx_scrq[i]);
1705 }
1706 #endif
1707 
1708 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
1709 {
1710 	return -EOPNOTSUPP;
1711 }
1712 
1713 static const struct net_device_ops ibmvnic_netdev_ops = {
1714 	.ndo_open		= ibmvnic_open,
1715 	.ndo_stop		= ibmvnic_close,
1716 	.ndo_start_xmit		= ibmvnic_xmit,
1717 	.ndo_set_rx_mode	= ibmvnic_set_multi,
1718 	.ndo_set_mac_address	= ibmvnic_set_mac,
1719 	.ndo_validate_addr	= eth_validate_addr,
1720 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
1721 #ifdef CONFIG_NET_POLL_CONTROLLER
1722 	.ndo_poll_controller	= ibmvnic_netpoll_controller,
1723 #endif
1724 	.ndo_change_mtu		= ibmvnic_change_mtu,
1725 };
1726 
1727 /* ethtool functions */
1728 
1729 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
1730 				      struct ethtool_link_ksettings *cmd)
1731 {
1732 	u32 supported, advertising;
1733 
1734 	supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
1735 			  SUPPORTED_FIBRE);
1736 	advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
1737 			    ADVERTISED_FIBRE);
1738 	cmd->base.speed = SPEED_1000;
1739 	cmd->base.duplex = DUPLEX_FULL;
1740 	cmd->base.port = PORT_FIBRE;
1741 	cmd->base.phy_address = 0;
1742 	cmd->base.autoneg = AUTONEG_ENABLE;
1743 
1744 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
1745 						supported);
1746 	ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
1747 						advertising);
1748 
1749 	return 0;
1750 }
1751 
1752 static void ibmvnic_get_drvinfo(struct net_device *dev,
1753 				struct ethtool_drvinfo *info)
1754 {
1755 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
1756 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
1757 }
1758 
1759 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
1760 {
1761 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1762 
1763 	return adapter->msg_enable;
1764 }
1765 
1766 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
1767 {
1768 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1769 
1770 	adapter->msg_enable = data;
1771 }
1772 
1773 static u32 ibmvnic_get_link(struct net_device *netdev)
1774 {
1775 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1776 
1777 	/* Don't need to send a query because we request a logical link up at
1778 	 * init and then we wait for link state indications
1779 	 */
1780 	return adapter->logical_link_state;
1781 }
1782 
1783 static void ibmvnic_get_ringparam(struct net_device *netdev,
1784 				  struct ethtool_ringparam *ring)
1785 {
1786 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1787 
1788 	ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
1789 	ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
1790 	ring->rx_mini_max_pending = 0;
1791 	ring->rx_jumbo_max_pending = 0;
1792 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
1793 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
1794 	ring->rx_mini_pending = 0;
1795 	ring->rx_jumbo_pending = 0;
1796 }
1797 
1798 static void ibmvnic_get_channels(struct net_device *netdev,
1799 				 struct ethtool_channels *channels)
1800 {
1801 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1802 
1803 	channels->max_rx = adapter->max_rx_queues;
1804 	channels->max_tx = adapter->max_tx_queues;
1805 	channels->max_other = 0;
1806 	channels->max_combined = 0;
1807 	channels->rx_count = adapter->req_rx_queues;
1808 	channels->tx_count = adapter->req_tx_queues;
1809 	channels->other_count = 0;
1810 	channels->combined_count = 0;
1811 }
1812 
1813 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
1814 {
1815 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1816 	int i;
1817 
1818 	if (stringset != ETH_SS_STATS)
1819 		return;
1820 
1821 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
1822 		memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
1823 
1824 	for (i = 0; i < adapter->req_tx_queues; i++) {
1825 		snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
1826 		data += ETH_GSTRING_LEN;
1827 
1828 		snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
1829 		data += ETH_GSTRING_LEN;
1830 
1831 		snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
1832 		data += ETH_GSTRING_LEN;
1833 	}
1834 
1835 	for (i = 0; i < adapter->req_rx_queues; i++) {
1836 		snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
1837 		data += ETH_GSTRING_LEN;
1838 
1839 		snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
1840 		data += ETH_GSTRING_LEN;
1841 
1842 		snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
1843 		data += ETH_GSTRING_LEN;
1844 	}
1845 }
1846 
1847 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
1848 {
1849 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1850 
1851 	switch (sset) {
1852 	case ETH_SS_STATS:
1853 		return ARRAY_SIZE(ibmvnic_stats) +
1854 		       adapter->req_tx_queues * NUM_TX_STATS +
1855 		       adapter->req_rx_queues * NUM_RX_STATS;
1856 	default:
1857 		return -EOPNOTSUPP;
1858 	}
1859 }
1860 
1861 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
1862 				      struct ethtool_stats *stats, u64 *data)
1863 {
1864 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
1865 	union ibmvnic_crq crq;
1866 	int i, j;
1867 
1868 	memset(&crq, 0, sizeof(crq));
1869 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
1870 	crq.request_statistics.cmd = REQUEST_STATISTICS;
1871 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
1872 	crq.request_statistics.len =
1873 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
1874 
1875 	/* Wait for data to be written */
1876 	init_completion(&adapter->stats_done);
1877 	ibmvnic_send_crq(adapter, &crq);
1878 	wait_for_completion(&adapter->stats_done);
1879 
1880 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
1881 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
1882 						ibmvnic_stats[i].offset));
1883 
1884 	for (j = 0; j < adapter->req_tx_queues; j++) {
1885 		data[i] = adapter->tx_stats_buffers[j].packets;
1886 		i++;
1887 		data[i] = adapter->tx_stats_buffers[j].bytes;
1888 		i++;
1889 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
1890 		i++;
1891 	}
1892 
1893 	for (j = 0; j < adapter->req_rx_queues; j++) {
1894 		data[i] = adapter->rx_stats_buffers[j].packets;
1895 		i++;
1896 		data[i] = adapter->rx_stats_buffers[j].bytes;
1897 		i++;
1898 		data[i] = adapter->rx_stats_buffers[j].interrupts;
1899 		i++;
1900 	}
1901 }
1902 
1903 static const struct ethtool_ops ibmvnic_ethtool_ops = {
1904 	.get_drvinfo		= ibmvnic_get_drvinfo,
1905 	.get_msglevel		= ibmvnic_get_msglevel,
1906 	.set_msglevel		= ibmvnic_set_msglevel,
1907 	.get_link		= ibmvnic_get_link,
1908 	.get_ringparam		= ibmvnic_get_ringparam,
1909 	.get_channels		= ibmvnic_get_channels,
1910 	.get_strings            = ibmvnic_get_strings,
1911 	.get_sset_count         = ibmvnic_get_sset_count,
1912 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
1913 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
1914 };
1915 
1916 /* Routines for managing CRQs/sCRQs  */
1917 
1918 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
1919 				   struct ibmvnic_sub_crq_queue *scrq)
1920 {
1921 	int rc;
1922 
1923 	if (scrq->irq) {
1924 		free_irq(scrq->irq, scrq);
1925 		irq_dispose_mapping(scrq->irq);
1926 		scrq->irq = 0;
1927 	}
1928 
1929 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
1930 	scrq->cur = 0;
1931 
1932 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
1933 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
1934 	return rc;
1935 }
1936 
1937 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
1938 {
1939 	int i, rc;
1940 
1941 	for (i = 0; i < adapter->req_tx_queues; i++) {
1942 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
1943 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
1944 		if (rc)
1945 			return rc;
1946 	}
1947 
1948 	for (i = 0; i < adapter->req_rx_queues; i++) {
1949 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
1950 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
1951 		if (rc)
1952 			return rc;
1953 	}
1954 
1955 	return rc;
1956 }
1957 
1958 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
1959 				  struct ibmvnic_sub_crq_queue *scrq)
1960 {
1961 	struct device *dev = &adapter->vdev->dev;
1962 	long rc;
1963 
1964 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
1965 
1966 	/* Close the sub-crqs */
1967 	do {
1968 		rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
1969 					adapter->vdev->unit_address,
1970 					scrq->crq_num);
1971 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
1972 
1973 	if (rc) {
1974 		netdev_err(adapter->netdev,
1975 			   "Failed to release sub-CRQ %16lx, rc = %ld\n",
1976 			   scrq->crq_num, rc);
1977 	}
1978 
1979 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
1980 			 DMA_BIDIRECTIONAL);
1981 	free_pages((unsigned long)scrq->msgs, 2);
1982 	kfree(scrq);
1983 }
1984 
1985 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
1986 							*adapter)
1987 {
1988 	struct device *dev = &adapter->vdev->dev;
1989 	struct ibmvnic_sub_crq_queue *scrq;
1990 	int rc;
1991 
1992 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
1993 	if (!scrq)
1994 		return NULL;
1995 
1996 	scrq->msgs =
1997 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
1998 	if (!scrq->msgs) {
1999 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2000 		goto zero_page_failed;
2001 	}
2002 
2003 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2004 					 DMA_BIDIRECTIONAL);
2005 	if (dma_mapping_error(dev, scrq->msg_token)) {
2006 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2007 		goto map_failed;
2008 	}
2009 
2010 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2011 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2012 
2013 	if (rc == H_RESOURCE)
2014 		rc = ibmvnic_reset_crq(adapter);
2015 
2016 	if (rc == H_CLOSED) {
2017 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2018 	} else if (rc) {
2019 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2020 		goto reg_failed;
2021 	}
2022 
2023 	scrq->adapter = adapter;
2024 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2025 	spin_lock_init(&scrq->lock);
2026 
2027 	netdev_dbg(adapter->netdev,
2028 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2029 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2030 
2031 	return scrq;
2032 
2033 reg_failed:
2034 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2035 			 DMA_BIDIRECTIONAL);
2036 map_failed:
2037 	free_pages((unsigned long)scrq->msgs, 2);
2038 zero_page_failed:
2039 	kfree(scrq);
2040 
2041 	return NULL;
2042 }
2043 
2044 static void release_sub_crqs(struct ibmvnic_adapter *adapter)
2045 {
2046 	int i;
2047 
2048 	if (adapter->tx_scrq) {
2049 		for (i = 0; i < adapter->req_tx_queues; i++) {
2050 			if (!adapter->tx_scrq[i])
2051 				continue;
2052 
2053 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2054 				   i);
2055 			if (adapter->tx_scrq[i]->irq) {
2056 				free_irq(adapter->tx_scrq[i]->irq,
2057 					 adapter->tx_scrq[i]);
2058 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2059 				adapter->tx_scrq[i]->irq = 0;
2060 			}
2061 
2062 			release_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2063 		}
2064 
2065 		kfree(adapter->tx_scrq);
2066 		adapter->tx_scrq = NULL;
2067 	}
2068 
2069 	if (adapter->rx_scrq) {
2070 		for (i = 0; i < adapter->req_rx_queues; i++) {
2071 			if (!adapter->rx_scrq[i])
2072 				continue;
2073 
2074 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2075 				   i);
2076 			if (adapter->rx_scrq[i]->irq) {
2077 				free_irq(adapter->rx_scrq[i]->irq,
2078 					 adapter->rx_scrq[i]);
2079 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2080 				adapter->rx_scrq[i]->irq = 0;
2081 			}
2082 
2083 			release_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2084 		}
2085 
2086 		kfree(adapter->rx_scrq);
2087 		adapter->rx_scrq = NULL;
2088 	}
2089 }
2090 
2091 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2092 			    struct ibmvnic_sub_crq_queue *scrq)
2093 {
2094 	struct device *dev = &adapter->vdev->dev;
2095 	unsigned long rc;
2096 
2097 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2098 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2099 	if (rc)
2100 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2101 			scrq->hw_irq, rc);
2102 	return rc;
2103 }
2104 
2105 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2106 			   struct ibmvnic_sub_crq_queue *scrq)
2107 {
2108 	struct device *dev = &adapter->vdev->dev;
2109 	unsigned long rc;
2110 
2111 	if (scrq->hw_irq > 0x100000000ULL) {
2112 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2113 		return 1;
2114 	}
2115 
2116 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2117 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2118 	if (rc)
2119 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2120 			scrq->hw_irq, rc);
2121 	return rc;
2122 }
2123 
2124 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2125 			       struct ibmvnic_sub_crq_queue *scrq)
2126 {
2127 	struct device *dev = &adapter->vdev->dev;
2128 	struct ibmvnic_tx_buff *txbuff;
2129 	union sub_crq *next;
2130 	int index;
2131 	int i, j;
2132 	u8 first;
2133 
2134 restart_loop:
2135 	while (pending_scrq(adapter, scrq)) {
2136 		unsigned int pool = scrq->pool_index;
2137 
2138 		next = ibmvnic_next_scrq(adapter, scrq);
2139 		for (i = 0; i < next->tx_comp.num_comps; i++) {
2140 			if (next->tx_comp.rcs[i]) {
2141 				dev_err(dev, "tx error %x\n",
2142 					next->tx_comp.rcs[i]);
2143 				continue;
2144 			}
2145 			index = be32_to_cpu(next->tx_comp.correlators[i]);
2146 			txbuff = &adapter->tx_pool[pool].tx_buff[index];
2147 
2148 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2149 				if (!txbuff->data_dma[j])
2150 					continue;
2151 
2152 				txbuff->data_dma[j] = 0;
2153 			}
2154 			/* if sub_crq was sent indirectly */
2155 			first = txbuff->indir_arr[0].generic.first;
2156 			if (first == IBMVNIC_CRQ_CMD) {
2157 				dma_unmap_single(dev, txbuff->indir_dma,
2158 						 sizeof(txbuff->indir_arr),
2159 						 DMA_TO_DEVICE);
2160 			}
2161 
2162 			if (txbuff->last_frag) {
2163 				dev_kfree_skb_any(txbuff->skb);
2164 				txbuff->skb = NULL;
2165 			}
2166 
2167 			adapter->tx_pool[pool].free_map[adapter->tx_pool[pool].
2168 						     producer_index] = index;
2169 			adapter->tx_pool[pool].producer_index =
2170 			    (adapter->tx_pool[pool].producer_index + 1) %
2171 			    adapter->req_tx_entries_per_subcrq;
2172 		}
2173 		/* remove tx_comp scrq*/
2174 		next->tx_comp.first = 0;
2175 
2176 		if (atomic_sub_return(next->tx_comp.num_comps, &scrq->used) <=
2177 		    (adapter->req_tx_entries_per_subcrq / 2) &&
2178 		    __netif_subqueue_stopped(adapter->netdev,
2179 					     scrq->pool_index)) {
2180 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2181 			netdev_info(adapter->netdev, "Started queue %d\n",
2182 				    scrq->pool_index);
2183 		}
2184 	}
2185 
2186 	enable_scrq_irq(adapter, scrq);
2187 
2188 	if (pending_scrq(adapter, scrq)) {
2189 		disable_scrq_irq(adapter, scrq);
2190 		goto restart_loop;
2191 	}
2192 
2193 	return 0;
2194 }
2195 
2196 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2197 {
2198 	struct ibmvnic_sub_crq_queue *scrq = instance;
2199 	struct ibmvnic_adapter *adapter = scrq->adapter;
2200 
2201 	disable_scrq_irq(adapter, scrq);
2202 	ibmvnic_complete_tx(adapter, scrq);
2203 
2204 	return IRQ_HANDLED;
2205 }
2206 
2207 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2208 {
2209 	struct ibmvnic_sub_crq_queue *scrq = instance;
2210 	struct ibmvnic_adapter *adapter = scrq->adapter;
2211 
2212 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2213 
2214 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2215 		disable_scrq_irq(adapter, scrq);
2216 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
2217 	}
2218 
2219 	return IRQ_HANDLED;
2220 }
2221 
2222 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2223 {
2224 	struct device *dev = &adapter->vdev->dev;
2225 	struct ibmvnic_sub_crq_queue *scrq;
2226 	int i = 0, j = 0;
2227 	int rc = 0;
2228 
2229 	for (i = 0; i < adapter->req_tx_queues; i++) {
2230 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2231 			   i);
2232 		scrq = adapter->tx_scrq[i];
2233 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2234 
2235 		if (!scrq->irq) {
2236 			rc = -EINVAL;
2237 			dev_err(dev, "Error mapping irq\n");
2238 			goto req_tx_irq_failed;
2239 		}
2240 
2241 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2242 				 0, "ibmvnic_tx", scrq);
2243 
2244 		if (rc) {
2245 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2246 				scrq->irq, rc);
2247 			irq_dispose_mapping(scrq->irq);
2248 			goto req_rx_irq_failed;
2249 		}
2250 	}
2251 
2252 	for (i = 0; i < adapter->req_rx_queues; i++) {
2253 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2254 			   i);
2255 		scrq = adapter->rx_scrq[i];
2256 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2257 		if (!scrq->irq) {
2258 			rc = -EINVAL;
2259 			dev_err(dev, "Error mapping irq\n");
2260 			goto req_rx_irq_failed;
2261 		}
2262 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2263 				 0, "ibmvnic_rx", scrq);
2264 		if (rc) {
2265 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2266 				scrq->irq, rc);
2267 			irq_dispose_mapping(scrq->irq);
2268 			goto req_rx_irq_failed;
2269 		}
2270 	}
2271 	return rc;
2272 
2273 req_rx_irq_failed:
2274 	for (j = 0; j < i; j++) {
2275 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2276 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2277 	}
2278 	i = adapter->req_tx_queues;
2279 req_tx_irq_failed:
2280 	for (j = 0; j < i; j++) {
2281 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2282 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2283 	}
2284 	release_sub_crqs(adapter);
2285 	return rc;
2286 }
2287 
2288 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2289 {
2290 	struct device *dev = &adapter->vdev->dev;
2291 	struct ibmvnic_sub_crq_queue **allqueues;
2292 	int registered_queues = 0;
2293 	int total_queues;
2294 	int more = 0;
2295 	int i;
2296 
2297 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2298 
2299 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2300 	if (!allqueues)
2301 		return -1;
2302 
2303 	for (i = 0; i < total_queues; i++) {
2304 		allqueues[i] = init_sub_crq_queue(adapter);
2305 		if (!allqueues[i]) {
2306 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2307 			break;
2308 		}
2309 		registered_queues++;
2310 	}
2311 
2312 	/* Make sure we were able to register the minimum number of queues */
2313 	if (registered_queues <
2314 	    adapter->min_tx_queues + adapter->min_rx_queues) {
2315 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2316 		goto tx_failed;
2317 	}
2318 
2319 	/* Distribute the failed allocated queues*/
2320 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
2321 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2322 		switch (i % 3) {
2323 		case 0:
2324 			if (adapter->req_rx_queues > adapter->min_rx_queues)
2325 				adapter->req_rx_queues--;
2326 			else
2327 				more++;
2328 			break;
2329 		case 1:
2330 			if (adapter->req_tx_queues > adapter->min_tx_queues)
2331 				adapter->req_tx_queues--;
2332 			else
2333 				more++;
2334 			break;
2335 		}
2336 	}
2337 
2338 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2339 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
2340 	if (!adapter->tx_scrq)
2341 		goto tx_failed;
2342 
2343 	for (i = 0; i < adapter->req_tx_queues; i++) {
2344 		adapter->tx_scrq[i] = allqueues[i];
2345 		adapter->tx_scrq[i]->pool_index = i;
2346 	}
2347 
2348 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
2349 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
2350 	if (!adapter->rx_scrq)
2351 		goto rx_failed;
2352 
2353 	for (i = 0; i < adapter->req_rx_queues; i++) {
2354 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
2355 		adapter->rx_scrq[i]->scrq_num = i;
2356 	}
2357 
2358 	kfree(allqueues);
2359 	return 0;
2360 
2361 rx_failed:
2362 	kfree(adapter->tx_scrq);
2363 	adapter->tx_scrq = NULL;
2364 tx_failed:
2365 	for (i = 0; i < registered_queues; i++)
2366 		release_sub_crq_queue(adapter, allqueues[i]);
2367 	kfree(allqueues);
2368 	return -1;
2369 }
2370 
2371 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
2372 {
2373 	struct device *dev = &adapter->vdev->dev;
2374 	union ibmvnic_crq crq;
2375 
2376 	if (!retry) {
2377 		/* Sub-CRQ entries are 32 byte long */
2378 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
2379 
2380 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
2381 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
2382 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
2383 			return;
2384 		}
2385 
2386 		/* Get the minimum between the queried max and the entries
2387 		 * that fit in our PAGE_SIZE
2388 		 */
2389 		adapter->req_tx_entries_per_subcrq =
2390 		    adapter->max_tx_entries_per_subcrq > entries_page ?
2391 		    entries_page : adapter->max_tx_entries_per_subcrq;
2392 		adapter->req_rx_add_entries_per_subcrq =
2393 		    adapter->max_rx_add_entries_per_subcrq > entries_page ?
2394 		    entries_page : adapter->max_rx_add_entries_per_subcrq;
2395 
2396 		adapter->req_tx_queues = adapter->opt_tx_comp_sub_queues;
2397 		adapter->req_rx_queues = adapter->opt_rx_comp_queues;
2398 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
2399 
2400 		adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
2401 	}
2402 
2403 	memset(&crq, 0, sizeof(crq));
2404 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
2405 	crq.request_capability.cmd = REQUEST_CAPABILITY;
2406 
2407 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
2408 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
2409 	atomic_inc(&adapter->running_cap_crqs);
2410 	ibmvnic_send_crq(adapter, &crq);
2411 
2412 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
2413 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
2414 	atomic_inc(&adapter->running_cap_crqs);
2415 	ibmvnic_send_crq(adapter, &crq);
2416 
2417 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
2418 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
2419 	atomic_inc(&adapter->running_cap_crqs);
2420 	ibmvnic_send_crq(adapter, &crq);
2421 
2422 	crq.request_capability.capability =
2423 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
2424 	crq.request_capability.number =
2425 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
2426 	atomic_inc(&adapter->running_cap_crqs);
2427 	ibmvnic_send_crq(adapter, &crq);
2428 
2429 	crq.request_capability.capability =
2430 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
2431 	crq.request_capability.number =
2432 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
2433 	atomic_inc(&adapter->running_cap_crqs);
2434 	ibmvnic_send_crq(adapter, &crq);
2435 
2436 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
2437 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
2438 	atomic_inc(&adapter->running_cap_crqs);
2439 	ibmvnic_send_crq(adapter, &crq);
2440 
2441 	if (adapter->netdev->flags & IFF_PROMISC) {
2442 		if (adapter->promisc_supported) {
2443 			crq.request_capability.capability =
2444 			    cpu_to_be16(PROMISC_REQUESTED);
2445 			crq.request_capability.number = cpu_to_be64(1);
2446 			atomic_inc(&adapter->running_cap_crqs);
2447 			ibmvnic_send_crq(adapter, &crq);
2448 		}
2449 	} else {
2450 		crq.request_capability.capability =
2451 		    cpu_to_be16(PROMISC_REQUESTED);
2452 		crq.request_capability.number = cpu_to_be64(0);
2453 		atomic_inc(&adapter->running_cap_crqs);
2454 		ibmvnic_send_crq(adapter, &crq);
2455 	}
2456 }
2457 
2458 static int pending_scrq(struct ibmvnic_adapter *adapter,
2459 			struct ibmvnic_sub_crq_queue *scrq)
2460 {
2461 	union sub_crq *entry = &scrq->msgs[scrq->cur];
2462 
2463 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
2464 		return 1;
2465 	else
2466 		return 0;
2467 }
2468 
2469 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
2470 					struct ibmvnic_sub_crq_queue *scrq)
2471 {
2472 	union sub_crq *entry;
2473 	unsigned long flags;
2474 
2475 	spin_lock_irqsave(&scrq->lock, flags);
2476 	entry = &scrq->msgs[scrq->cur];
2477 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2478 		if (++scrq->cur == scrq->size)
2479 			scrq->cur = 0;
2480 	} else {
2481 		entry = NULL;
2482 	}
2483 	spin_unlock_irqrestore(&scrq->lock, flags);
2484 
2485 	return entry;
2486 }
2487 
2488 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
2489 {
2490 	struct ibmvnic_crq_queue *queue = &adapter->crq;
2491 	union ibmvnic_crq *crq;
2492 
2493 	crq = &queue->msgs[queue->cur];
2494 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
2495 		if (++queue->cur == queue->size)
2496 			queue->cur = 0;
2497 	} else {
2498 		crq = NULL;
2499 	}
2500 
2501 	return crq;
2502 }
2503 
2504 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
2505 		       union sub_crq *sub_crq)
2506 {
2507 	unsigned int ua = adapter->vdev->unit_address;
2508 	struct device *dev = &adapter->vdev->dev;
2509 	u64 *u64_crq = (u64 *)sub_crq;
2510 	int rc;
2511 
2512 	netdev_dbg(adapter->netdev,
2513 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
2514 		   (unsigned long int)cpu_to_be64(remote_handle),
2515 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2516 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
2517 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
2518 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
2519 
2520 	/* Make sure the hypervisor sees the complete request */
2521 	mb();
2522 
2523 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
2524 				cpu_to_be64(remote_handle),
2525 				cpu_to_be64(u64_crq[0]),
2526 				cpu_to_be64(u64_crq[1]),
2527 				cpu_to_be64(u64_crq[2]),
2528 				cpu_to_be64(u64_crq[3]));
2529 
2530 	if (rc) {
2531 		if (rc == H_CLOSED)
2532 			dev_warn(dev, "CRQ Queue closed\n");
2533 		dev_err(dev, "Send error (rc=%d)\n", rc);
2534 	}
2535 
2536 	return rc;
2537 }
2538 
2539 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
2540 				u64 remote_handle, u64 ioba, u64 num_entries)
2541 {
2542 	unsigned int ua = adapter->vdev->unit_address;
2543 	struct device *dev = &adapter->vdev->dev;
2544 	int rc;
2545 
2546 	/* Make sure the hypervisor sees the complete request */
2547 	mb();
2548 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
2549 				cpu_to_be64(remote_handle),
2550 				ioba, num_entries);
2551 
2552 	if (rc) {
2553 		if (rc == H_CLOSED)
2554 			dev_warn(dev, "CRQ Queue closed\n");
2555 		dev_err(dev, "Send (indirect) error (rc=%d)\n", rc);
2556 	}
2557 
2558 	return rc;
2559 }
2560 
2561 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
2562 			    union ibmvnic_crq *crq)
2563 {
2564 	unsigned int ua = adapter->vdev->unit_address;
2565 	struct device *dev = &adapter->vdev->dev;
2566 	u64 *u64_crq = (u64 *)crq;
2567 	int rc;
2568 
2569 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
2570 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
2571 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
2572 
2573 	/* Make sure the hypervisor sees the complete request */
2574 	mb();
2575 
2576 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
2577 				cpu_to_be64(u64_crq[0]),
2578 				cpu_to_be64(u64_crq[1]));
2579 
2580 	if (rc) {
2581 		if (rc == H_CLOSED)
2582 			dev_warn(dev, "CRQ Queue closed\n");
2583 		dev_warn(dev, "Send error (rc=%d)\n", rc);
2584 	}
2585 
2586 	return rc;
2587 }
2588 
2589 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
2590 {
2591 	union ibmvnic_crq crq;
2592 
2593 	memset(&crq, 0, sizeof(crq));
2594 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
2595 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
2596 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
2597 
2598 	return ibmvnic_send_crq(adapter, &crq);
2599 }
2600 
2601 static int send_version_xchg(struct ibmvnic_adapter *adapter)
2602 {
2603 	union ibmvnic_crq crq;
2604 
2605 	memset(&crq, 0, sizeof(crq));
2606 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
2607 	crq.version_exchange.cmd = VERSION_EXCHANGE;
2608 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
2609 
2610 	return ibmvnic_send_crq(adapter, &crq);
2611 }
2612 
2613 static void send_login(struct ibmvnic_adapter *adapter)
2614 {
2615 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
2616 	struct ibmvnic_login_buffer *login_buffer;
2617 	struct device *dev = &adapter->vdev->dev;
2618 	dma_addr_t rsp_buffer_token;
2619 	dma_addr_t buffer_token;
2620 	size_t rsp_buffer_size;
2621 	union ibmvnic_crq crq;
2622 	size_t buffer_size;
2623 	__be64 *tx_list_p;
2624 	__be64 *rx_list_p;
2625 	int i;
2626 
2627 	buffer_size =
2628 	    sizeof(struct ibmvnic_login_buffer) +
2629 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues);
2630 
2631 	login_buffer = kmalloc(buffer_size, GFP_ATOMIC);
2632 	if (!login_buffer)
2633 		goto buf_alloc_failed;
2634 
2635 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
2636 				      DMA_TO_DEVICE);
2637 	if (dma_mapping_error(dev, buffer_token)) {
2638 		dev_err(dev, "Couldn't map login buffer\n");
2639 		goto buf_map_failed;
2640 	}
2641 
2642 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
2643 			  sizeof(u64) * adapter->req_tx_queues +
2644 			  sizeof(u64) * adapter->req_rx_queues +
2645 			  sizeof(u64) * adapter->req_rx_queues +
2646 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
2647 
2648 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
2649 	if (!login_rsp_buffer)
2650 		goto buf_rsp_alloc_failed;
2651 
2652 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
2653 					  rsp_buffer_size, DMA_FROM_DEVICE);
2654 	if (dma_mapping_error(dev, rsp_buffer_token)) {
2655 		dev_err(dev, "Couldn't map login rsp buffer\n");
2656 		goto buf_rsp_map_failed;
2657 	}
2658 
2659 	adapter->login_buf = login_buffer;
2660 	adapter->login_buf_token = buffer_token;
2661 	adapter->login_buf_sz = buffer_size;
2662 	adapter->login_rsp_buf = login_rsp_buffer;
2663 	adapter->login_rsp_buf_token = rsp_buffer_token;
2664 	adapter->login_rsp_buf_sz = rsp_buffer_size;
2665 
2666 	login_buffer->len = cpu_to_be32(buffer_size);
2667 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
2668 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
2669 	login_buffer->off_txcomp_subcrqs =
2670 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
2671 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
2672 	login_buffer->off_rxcomp_subcrqs =
2673 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
2674 			sizeof(u64) * adapter->req_tx_queues);
2675 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
2676 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
2677 
2678 	tx_list_p = (__be64 *)((char *)login_buffer +
2679 				      sizeof(struct ibmvnic_login_buffer));
2680 	rx_list_p = (__be64 *)((char *)login_buffer +
2681 				      sizeof(struct ibmvnic_login_buffer) +
2682 				      sizeof(u64) * adapter->req_tx_queues);
2683 
2684 	for (i = 0; i < adapter->req_tx_queues; i++) {
2685 		if (adapter->tx_scrq[i]) {
2686 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
2687 						   crq_num);
2688 		}
2689 	}
2690 
2691 	for (i = 0; i < adapter->req_rx_queues; i++) {
2692 		if (adapter->rx_scrq[i]) {
2693 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
2694 						   crq_num);
2695 		}
2696 	}
2697 
2698 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
2699 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
2700 		netdev_dbg(adapter->netdev, "%016lx\n",
2701 			   ((unsigned long int *)(adapter->login_buf))[i]);
2702 	}
2703 
2704 	memset(&crq, 0, sizeof(crq));
2705 	crq.login.first = IBMVNIC_CRQ_CMD;
2706 	crq.login.cmd = LOGIN;
2707 	crq.login.ioba = cpu_to_be32(buffer_token);
2708 	crq.login.len = cpu_to_be32(buffer_size);
2709 	ibmvnic_send_crq(adapter, &crq);
2710 
2711 	return;
2712 
2713 buf_rsp_map_failed:
2714 	kfree(login_rsp_buffer);
2715 buf_rsp_alloc_failed:
2716 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
2717 buf_map_failed:
2718 	kfree(login_buffer);
2719 buf_alloc_failed:
2720 	return;
2721 }
2722 
2723 static void send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
2724 			     u32 len, u8 map_id)
2725 {
2726 	union ibmvnic_crq crq;
2727 
2728 	memset(&crq, 0, sizeof(crq));
2729 	crq.request_map.first = IBMVNIC_CRQ_CMD;
2730 	crq.request_map.cmd = REQUEST_MAP;
2731 	crq.request_map.map_id = map_id;
2732 	crq.request_map.ioba = cpu_to_be32(addr);
2733 	crq.request_map.len = cpu_to_be32(len);
2734 	ibmvnic_send_crq(adapter, &crq);
2735 }
2736 
2737 static void send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
2738 {
2739 	union ibmvnic_crq crq;
2740 
2741 	memset(&crq, 0, sizeof(crq));
2742 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
2743 	crq.request_unmap.cmd = REQUEST_UNMAP;
2744 	crq.request_unmap.map_id = map_id;
2745 	ibmvnic_send_crq(adapter, &crq);
2746 }
2747 
2748 static void send_map_query(struct ibmvnic_adapter *adapter)
2749 {
2750 	union ibmvnic_crq crq;
2751 
2752 	memset(&crq, 0, sizeof(crq));
2753 	crq.query_map.first = IBMVNIC_CRQ_CMD;
2754 	crq.query_map.cmd = QUERY_MAP;
2755 	ibmvnic_send_crq(adapter, &crq);
2756 }
2757 
2758 /* Send a series of CRQs requesting various capabilities of the VNIC server */
2759 static void send_cap_queries(struct ibmvnic_adapter *adapter)
2760 {
2761 	union ibmvnic_crq crq;
2762 
2763 	atomic_set(&adapter->running_cap_crqs, 0);
2764 	memset(&crq, 0, sizeof(crq));
2765 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
2766 	crq.query_capability.cmd = QUERY_CAPABILITY;
2767 
2768 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
2769 	atomic_inc(&adapter->running_cap_crqs);
2770 	ibmvnic_send_crq(adapter, &crq);
2771 
2772 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
2773 	atomic_inc(&adapter->running_cap_crqs);
2774 	ibmvnic_send_crq(adapter, &crq);
2775 
2776 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
2777 	atomic_inc(&adapter->running_cap_crqs);
2778 	ibmvnic_send_crq(adapter, &crq);
2779 
2780 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
2781 	atomic_inc(&adapter->running_cap_crqs);
2782 	ibmvnic_send_crq(adapter, &crq);
2783 
2784 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
2785 	atomic_inc(&adapter->running_cap_crqs);
2786 	ibmvnic_send_crq(adapter, &crq);
2787 
2788 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
2789 	atomic_inc(&adapter->running_cap_crqs);
2790 	ibmvnic_send_crq(adapter, &crq);
2791 
2792 	crq.query_capability.capability =
2793 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
2794 	atomic_inc(&adapter->running_cap_crqs);
2795 	ibmvnic_send_crq(adapter, &crq);
2796 
2797 	crq.query_capability.capability =
2798 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
2799 	atomic_inc(&adapter->running_cap_crqs);
2800 	ibmvnic_send_crq(adapter, &crq);
2801 
2802 	crq.query_capability.capability =
2803 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
2804 	atomic_inc(&adapter->running_cap_crqs);
2805 	ibmvnic_send_crq(adapter, &crq);
2806 
2807 	crq.query_capability.capability =
2808 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
2809 	atomic_inc(&adapter->running_cap_crqs);
2810 	ibmvnic_send_crq(adapter, &crq);
2811 
2812 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
2813 	atomic_inc(&adapter->running_cap_crqs);
2814 	ibmvnic_send_crq(adapter, &crq);
2815 
2816 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
2817 	atomic_inc(&adapter->running_cap_crqs);
2818 	ibmvnic_send_crq(adapter, &crq);
2819 
2820 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
2821 	atomic_inc(&adapter->running_cap_crqs);
2822 	ibmvnic_send_crq(adapter, &crq);
2823 
2824 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
2825 	atomic_inc(&adapter->running_cap_crqs);
2826 	ibmvnic_send_crq(adapter, &crq);
2827 
2828 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
2829 	atomic_inc(&adapter->running_cap_crqs);
2830 	ibmvnic_send_crq(adapter, &crq);
2831 
2832 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
2833 	atomic_inc(&adapter->running_cap_crqs);
2834 	ibmvnic_send_crq(adapter, &crq);
2835 
2836 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
2837 	atomic_inc(&adapter->running_cap_crqs);
2838 	ibmvnic_send_crq(adapter, &crq);
2839 
2840 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
2841 	atomic_inc(&adapter->running_cap_crqs);
2842 	ibmvnic_send_crq(adapter, &crq);
2843 
2844 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
2845 	atomic_inc(&adapter->running_cap_crqs);
2846 	ibmvnic_send_crq(adapter, &crq);
2847 
2848 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
2849 	atomic_inc(&adapter->running_cap_crqs);
2850 	ibmvnic_send_crq(adapter, &crq);
2851 
2852 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
2853 	atomic_inc(&adapter->running_cap_crqs);
2854 	ibmvnic_send_crq(adapter, &crq);
2855 
2856 	crq.query_capability.capability =
2857 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
2858 	atomic_inc(&adapter->running_cap_crqs);
2859 	ibmvnic_send_crq(adapter, &crq);
2860 
2861 	crq.query_capability.capability =
2862 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
2863 	atomic_inc(&adapter->running_cap_crqs);
2864 	ibmvnic_send_crq(adapter, &crq);
2865 
2866 	crq.query_capability.capability =
2867 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
2868 	atomic_inc(&adapter->running_cap_crqs);
2869 	ibmvnic_send_crq(adapter, &crq);
2870 
2871 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
2872 	atomic_inc(&adapter->running_cap_crqs);
2873 	ibmvnic_send_crq(adapter, &crq);
2874 }
2875 
2876 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
2877 {
2878 	struct device *dev = &adapter->vdev->dev;
2879 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
2880 	union ibmvnic_crq crq;
2881 	int i;
2882 
2883 	dma_unmap_single(dev, adapter->ip_offload_tok,
2884 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
2885 
2886 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
2887 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
2888 		netdev_dbg(adapter->netdev, "%016lx\n",
2889 			   ((unsigned long int *)(buf))[i]);
2890 
2891 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
2892 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
2893 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
2894 		   buf->tcp_ipv4_chksum);
2895 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
2896 		   buf->tcp_ipv6_chksum);
2897 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
2898 		   buf->udp_ipv4_chksum);
2899 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
2900 		   buf->udp_ipv6_chksum);
2901 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
2902 		   buf->large_tx_ipv4);
2903 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
2904 		   buf->large_tx_ipv6);
2905 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
2906 		   buf->large_rx_ipv4);
2907 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
2908 		   buf->large_rx_ipv6);
2909 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
2910 		   buf->max_ipv4_header_size);
2911 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
2912 		   buf->max_ipv6_header_size);
2913 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
2914 		   buf->max_tcp_header_size);
2915 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
2916 		   buf->max_udp_header_size);
2917 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
2918 		   buf->max_large_tx_size);
2919 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
2920 		   buf->max_large_rx_size);
2921 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
2922 		   buf->ipv6_extension_header);
2923 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
2924 		   buf->tcp_pseudosum_req);
2925 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
2926 		   buf->num_ipv6_ext_headers);
2927 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
2928 		   buf->off_ipv6_ext_headers);
2929 
2930 	adapter->ip_offload_ctrl_tok =
2931 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
2932 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
2933 
2934 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
2935 		dev_err(dev, "Couldn't map ip offload control buffer\n");
2936 		return;
2937 	}
2938 
2939 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
2940 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
2941 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
2942 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
2943 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
2944 
2945 	/* large_tx/rx disabled for now, additional features needed */
2946 	adapter->ip_offload_ctrl.large_tx_ipv4 = 0;
2947 	adapter->ip_offload_ctrl.large_tx_ipv6 = 0;
2948 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
2949 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
2950 
2951 	adapter->netdev->features = NETIF_F_GSO;
2952 
2953 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
2954 		adapter->netdev->features |= NETIF_F_IP_CSUM;
2955 
2956 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
2957 		adapter->netdev->features |= NETIF_F_IPV6_CSUM;
2958 
2959 	if ((adapter->netdev->features &
2960 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
2961 		adapter->netdev->features |= NETIF_F_RXCSUM;
2962 
2963 	memset(&crq, 0, sizeof(crq));
2964 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
2965 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
2966 	crq.control_ip_offload.len =
2967 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
2968 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
2969 	ibmvnic_send_crq(adapter, &crq);
2970 }
2971 
2972 static void handle_error_info_rsp(union ibmvnic_crq *crq,
2973 				  struct ibmvnic_adapter *adapter)
2974 {
2975 	struct device *dev = &adapter->vdev->dev;
2976 	struct ibmvnic_error_buff *error_buff, *tmp;
2977 	unsigned long flags;
2978 	bool found = false;
2979 	int i;
2980 
2981 	if (!crq->request_error_rsp.rc.code) {
2982 		dev_info(dev, "Request Error Rsp returned with rc=%x\n",
2983 			 crq->request_error_rsp.rc.code);
2984 		return;
2985 	}
2986 
2987 	spin_lock_irqsave(&adapter->error_list_lock, flags);
2988 	list_for_each_entry_safe(error_buff, tmp, &adapter->errors, list)
2989 		if (error_buff->error_id == crq->request_error_rsp.error_id) {
2990 			found = true;
2991 			list_del(&error_buff->list);
2992 			break;
2993 		}
2994 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
2995 
2996 	if (!found) {
2997 		dev_err(dev, "Couldn't find error id %x\n",
2998 			be32_to_cpu(crq->request_error_rsp.error_id));
2999 		return;
3000 	}
3001 
3002 	dev_err(dev, "Detailed info for error id %x:",
3003 		be32_to_cpu(crq->request_error_rsp.error_id));
3004 
3005 	for (i = 0; i < error_buff->len; i++) {
3006 		pr_cont("%02x", (int)error_buff->buff[i]);
3007 		if (i % 8 == 7)
3008 			pr_cont(" ");
3009 	}
3010 	pr_cont("\n");
3011 
3012 	dma_unmap_single(dev, error_buff->dma, error_buff->len,
3013 			 DMA_FROM_DEVICE);
3014 	kfree(error_buff->buff);
3015 	kfree(error_buff);
3016 }
3017 
3018 static void request_error_information(struct ibmvnic_adapter *adapter,
3019 				      union ibmvnic_crq *err_crq)
3020 {
3021 	struct device *dev = &adapter->vdev->dev;
3022 	struct net_device *netdev = adapter->netdev;
3023 	struct ibmvnic_error_buff *error_buff;
3024 	unsigned long timeout = msecs_to_jiffies(30000);
3025 	union ibmvnic_crq crq;
3026 	unsigned long flags;
3027 	int rc, detail_len;
3028 
3029 	error_buff = kmalloc(sizeof(*error_buff), GFP_ATOMIC);
3030 	if (!error_buff)
3031 		return;
3032 
3033 	detail_len = be32_to_cpu(err_crq->error_indication.detail_error_sz);
3034 	error_buff->buff = kmalloc(detail_len, GFP_ATOMIC);
3035 	if (!error_buff->buff) {
3036 		kfree(error_buff);
3037 		return;
3038 	}
3039 
3040 	error_buff->dma = dma_map_single(dev, error_buff->buff, detail_len,
3041 					 DMA_FROM_DEVICE);
3042 	if (dma_mapping_error(dev, error_buff->dma)) {
3043 		netdev_err(netdev, "Couldn't map error buffer\n");
3044 		kfree(error_buff->buff);
3045 		kfree(error_buff);
3046 		return;
3047 	}
3048 
3049 	error_buff->len = detail_len;
3050 	error_buff->error_id = err_crq->error_indication.error_id;
3051 
3052 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3053 	list_add_tail(&error_buff->list, &adapter->errors);
3054 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3055 
3056 	memset(&crq, 0, sizeof(crq));
3057 	crq.request_error_info.first = IBMVNIC_CRQ_CMD;
3058 	crq.request_error_info.cmd = REQUEST_ERROR_INFO;
3059 	crq.request_error_info.ioba = cpu_to_be32(error_buff->dma);
3060 	crq.request_error_info.len = cpu_to_be32(detail_len);
3061 	crq.request_error_info.error_id = err_crq->error_indication.error_id;
3062 
3063 	rc = ibmvnic_send_crq(adapter, &crq);
3064 	if (rc) {
3065 		netdev_err(netdev, "failed to request error information\n");
3066 		goto err_info_fail;
3067 	}
3068 
3069 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3070 		netdev_err(netdev, "timeout waiting for error information\n");
3071 		goto err_info_fail;
3072 	}
3073 
3074 	return;
3075 
3076 err_info_fail:
3077 	spin_lock_irqsave(&adapter->error_list_lock, flags);
3078 	list_del(&error_buff->list);
3079 	spin_unlock_irqrestore(&adapter->error_list_lock, flags);
3080 
3081 	kfree(error_buff->buff);
3082 	kfree(error_buff);
3083 }
3084 
3085 static void handle_error_indication(union ibmvnic_crq *crq,
3086 				    struct ibmvnic_adapter *adapter)
3087 {
3088 	struct device *dev = &adapter->vdev->dev;
3089 
3090 	dev_err(dev, "Firmware reports %serror id %x, cause %d\n",
3091 		crq->error_indication.flags
3092 			& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3093 		be32_to_cpu(crq->error_indication.error_id),
3094 		be16_to_cpu(crq->error_indication.error_cause));
3095 
3096 	if (be32_to_cpu(crq->error_indication.error_id))
3097 		request_error_information(adapter, crq);
3098 
3099 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3100 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3101 	else
3102 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3103 }
3104 
3105 static void handle_change_mac_rsp(union ibmvnic_crq *crq,
3106 				  struct ibmvnic_adapter *adapter)
3107 {
3108 	struct net_device *netdev = adapter->netdev;
3109 	struct device *dev = &adapter->vdev->dev;
3110 	long rc;
3111 
3112 	rc = crq->change_mac_addr_rsp.rc.code;
3113 	if (rc) {
3114 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3115 		return;
3116 	}
3117 	memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3118 	       ETH_ALEN);
3119 }
3120 
3121 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3122 				   struct ibmvnic_adapter *adapter)
3123 {
3124 	struct device *dev = &adapter->vdev->dev;
3125 	u64 *req_value;
3126 	char *name;
3127 
3128 	atomic_dec(&adapter->running_cap_crqs);
3129 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3130 	case REQ_TX_QUEUES:
3131 		req_value = &adapter->req_tx_queues;
3132 		name = "tx";
3133 		break;
3134 	case REQ_RX_QUEUES:
3135 		req_value = &adapter->req_rx_queues;
3136 		name = "rx";
3137 		break;
3138 	case REQ_RX_ADD_QUEUES:
3139 		req_value = &adapter->req_rx_add_queues;
3140 		name = "rx_add";
3141 		break;
3142 	case REQ_TX_ENTRIES_PER_SUBCRQ:
3143 		req_value = &adapter->req_tx_entries_per_subcrq;
3144 		name = "tx_entries_per_subcrq";
3145 		break;
3146 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3147 		req_value = &adapter->req_rx_add_entries_per_subcrq;
3148 		name = "rx_add_entries_per_subcrq";
3149 		break;
3150 	case REQ_MTU:
3151 		req_value = &adapter->req_mtu;
3152 		name = "mtu";
3153 		break;
3154 	case PROMISC_REQUESTED:
3155 		req_value = &adapter->promisc;
3156 		name = "promisc";
3157 		break;
3158 	default:
3159 		dev_err(dev, "Got invalid cap request rsp %d\n",
3160 			crq->request_capability.capability);
3161 		return;
3162 	}
3163 
3164 	switch (crq->request_capability_rsp.rc.code) {
3165 	case SUCCESS:
3166 		break;
3167 	case PARTIALSUCCESS:
3168 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3169 			 *req_value,
3170 			 (long int)be64_to_cpu(crq->request_capability_rsp.
3171 					       number), name);
3172 		*req_value = be64_to_cpu(crq->request_capability_rsp.number);
3173 		ibmvnic_send_req_caps(adapter, 1);
3174 		return;
3175 	default:
3176 		dev_err(dev, "Error %d in request cap rsp\n",
3177 			crq->request_capability_rsp.rc.code);
3178 		return;
3179 	}
3180 
3181 	/* Done receiving requested capabilities, query IP offload support */
3182 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3183 		union ibmvnic_crq newcrq;
3184 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3185 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3186 		    &adapter->ip_offload_buf;
3187 
3188 		adapter->wait_capability = false;
3189 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3190 							 buf_sz,
3191 							 DMA_FROM_DEVICE);
3192 
3193 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3194 			if (!firmware_has_feature(FW_FEATURE_CMO))
3195 				dev_err(dev, "Couldn't map offload buffer\n");
3196 			return;
3197 		}
3198 
3199 		memset(&newcrq, 0, sizeof(newcrq));
3200 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3201 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3202 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3203 		newcrq.query_ip_offload.ioba =
3204 		    cpu_to_be32(adapter->ip_offload_tok);
3205 
3206 		ibmvnic_send_crq(adapter, &newcrq);
3207 	}
3208 }
3209 
3210 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3211 			    struct ibmvnic_adapter *adapter)
3212 {
3213 	struct device *dev = &adapter->vdev->dev;
3214 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
3215 	struct ibmvnic_login_buffer *login = adapter->login_buf;
3216 	int i;
3217 
3218 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
3219 			 DMA_BIDIRECTIONAL);
3220 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
3221 			 adapter->login_rsp_buf_sz, DMA_BIDIRECTIONAL);
3222 
3223 	/* If the number of queues requested can't be allocated by the
3224 	 * server, the login response will return with code 1. We will need
3225 	 * to resend the login buffer with fewer queues requested.
3226 	 */
3227 	if (login_rsp_crq->generic.rc.code) {
3228 		adapter->renegotiate = true;
3229 		complete(&adapter->init_done);
3230 		return 0;
3231 	}
3232 
3233 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
3234 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
3235 		netdev_dbg(adapter->netdev, "%016lx\n",
3236 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
3237 	}
3238 
3239 	/* Sanity checks */
3240 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
3241 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
3242 	     adapter->req_rx_add_queues !=
3243 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
3244 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
3245 		ibmvnic_remove(adapter->vdev);
3246 		return -EIO;
3247 	}
3248 	complete(&adapter->init_done);
3249 
3250 	return 0;
3251 }
3252 
3253 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
3254 				     struct ibmvnic_adapter *adapter)
3255 {
3256 	struct device *dev = &adapter->vdev->dev;
3257 	long rc;
3258 
3259 	rc = crq->request_unmap_rsp.rc.code;
3260 	if (rc)
3261 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
3262 }
3263 
3264 static void handle_query_map_rsp(union ibmvnic_crq *crq,
3265 				 struct ibmvnic_adapter *adapter)
3266 {
3267 	struct net_device *netdev = adapter->netdev;
3268 	struct device *dev = &adapter->vdev->dev;
3269 	long rc;
3270 
3271 	rc = crq->query_map_rsp.rc.code;
3272 	if (rc) {
3273 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
3274 		return;
3275 	}
3276 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
3277 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
3278 		   crq->query_map_rsp.free_pages);
3279 }
3280 
3281 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
3282 				 struct ibmvnic_adapter *adapter)
3283 {
3284 	struct net_device *netdev = adapter->netdev;
3285 	struct device *dev = &adapter->vdev->dev;
3286 	long rc;
3287 
3288 	atomic_dec(&adapter->running_cap_crqs);
3289 	netdev_dbg(netdev, "Outstanding queries: %d\n",
3290 		   atomic_read(&adapter->running_cap_crqs));
3291 	rc = crq->query_capability.rc.code;
3292 	if (rc) {
3293 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
3294 		goto out;
3295 	}
3296 
3297 	switch (be16_to_cpu(crq->query_capability.capability)) {
3298 	case MIN_TX_QUEUES:
3299 		adapter->min_tx_queues =
3300 		    be64_to_cpu(crq->query_capability.number);
3301 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
3302 			   adapter->min_tx_queues);
3303 		break;
3304 	case MIN_RX_QUEUES:
3305 		adapter->min_rx_queues =
3306 		    be64_to_cpu(crq->query_capability.number);
3307 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
3308 			   adapter->min_rx_queues);
3309 		break;
3310 	case MIN_RX_ADD_QUEUES:
3311 		adapter->min_rx_add_queues =
3312 		    be64_to_cpu(crq->query_capability.number);
3313 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
3314 			   adapter->min_rx_add_queues);
3315 		break;
3316 	case MAX_TX_QUEUES:
3317 		adapter->max_tx_queues =
3318 		    be64_to_cpu(crq->query_capability.number);
3319 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
3320 			   adapter->max_tx_queues);
3321 		break;
3322 	case MAX_RX_QUEUES:
3323 		adapter->max_rx_queues =
3324 		    be64_to_cpu(crq->query_capability.number);
3325 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
3326 			   adapter->max_rx_queues);
3327 		break;
3328 	case MAX_RX_ADD_QUEUES:
3329 		adapter->max_rx_add_queues =
3330 		    be64_to_cpu(crq->query_capability.number);
3331 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
3332 			   adapter->max_rx_add_queues);
3333 		break;
3334 	case MIN_TX_ENTRIES_PER_SUBCRQ:
3335 		adapter->min_tx_entries_per_subcrq =
3336 		    be64_to_cpu(crq->query_capability.number);
3337 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
3338 			   adapter->min_tx_entries_per_subcrq);
3339 		break;
3340 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
3341 		adapter->min_rx_add_entries_per_subcrq =
3342 		    be64_to_cpu(crq->query_capability.number);
3343 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
3344 			   adapter->min_rx_add_entries_per_subcrq);
3345 		break;
3346 	case MAX_TX_ENTRIES_PER_SUBCRQ:
3347 		adapter->max_tx_entries_per_subcrq =
3348 		    be64_to_cpu(crq->query_capability.number);
3349 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
3350 			   adapter->max_tx_entries_per_subcrq);
3351 		break;
3352 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
3353 		adapter->max_rx_add_entries_per_subcrq =
3354 		    be64_to_cpu(crq->query_capability.number);
3355 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
3356 			   adapter->max_rx_add_entries_per_subcrq);
3357 		break;
3358 	case TCP_IP_OFFLOAD:
3359 		adapter->tcp_ip_offload =
3360 		    be64_to_cpu(crq->query_capability.number);
3361 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
3362 			   adapter->tcp_ip_offload);
3363 		break;
3364 	case PROMISC_SUPPORTED:
3365 		adapter->promisc_supported =
3366 		    be64_to_cpu(crq->query_capability.number);
3367 		netdev_dbg(netdev, "promisc_supported = %lld\n",
3368 			   adapter->promisc_supported);
3369 		break;
3370 	case MIN_MTU:
3371 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
3372 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
3373 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
3374 		break;
3375 	case MAX_MTU:
3376 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
3377 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
3378 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
3379 		break;
3380 	case MAX_MULTICAST_FILTERS:
3381 		adapter->max_multicast_filters =
3382 		    be64_to_cpu(crq->query_capability.number);
3383 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
3384 			   adapter->max_multicast_filters);
3385 		break;
3386 	case VLAN_HEADER_INSERTION:
3387 		adapter->vlan_header_insertion =
3388 		    be64_to_cpu(crq->query_capability.number);
3389 		if (adapter->vlan_header_insertion)
3390 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
3391 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
3392 			   adapter->vlan_header_insertion);
3393 		break;
3394 	case RX_VLAN_HEADER_INSERTION:
3395 		adapter->rx_vlan_header_insertion =
3396 		    be64_to_cpu(crq->query_capability.number);
3397 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
3398 			   adapter->rx_vlan_header_insertion);
3399 		break;
3400 	case MAX_TX_SG_ENTRIES:
3401 		adapter->max_tx_sg_entries =
3402 		    be64_to_cpu(crq->query_capability.number);
3403 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
3404 			   adapter->max_tx_sg_entries);
3405 		break;
3406 	case RX_SG_SUPPORTED:
3407 		adapter->rx_sg_supported =
3408 		    be64_to_cpu(crq->query_capability.number);
3409 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
3410 			   adapter->rx_sg_supported);
3411 		break;
3412 	case OPT_TX_COMP_SUB_QUEUES:
3413 		adapter->opt_tx_comp_sub_queues =
3414 		    be64_to_cpu(crq->query_capability.number);
3415 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
3416 			   adapter->opt_tx_comp_sub_queues);
3417 		break;
3418 	case OPT_RX_COMP_QUEUES:
3419 		adapter->opt_rx_comp_queues =
3420 		    be64_to_cpu(crq->query_capability.number);
3421 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
3422 			   adapter->opt_rx_comp_queues);
3423 		break;
3424 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
3425 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
3426 		    be64_to_cpu(crq->query_capability.number);
3427 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
3428 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
3429 		break;
3430 	case OPT_TX_ENTRIES_PER_SUBCRQ:
3431 		adapter->opt_tx_entries_per_subcrq =
3432 		    be64_to_cpu(crq->query_capability.number);
3433 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
3434 			   adapter->opt_tx_entries_per_subcrq);
3435 		break;
3436 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
3437 		adapter->opt_rxba_entries_per_subcrq =
3438 		    be64_to_cpu(crq->query_capability.number);
3439 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
3440 			   adapter->opt_rxba_entries_per_subcrq);
3441 		break;
3442 	case TX_RX_DESC_REQ:
3443 		adapter->tx_rx_desc_req = crq->query_capability.number;
3444 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
3445 			   adapter->tx_rx_desc_req);
3446 		break;
3447 
3448 	default:
3449 		netdev_err(netdev, "Got invalid cap rsp %d\n",
3450 			   crq->query_capability.capability);
3451 	}
3452 
3453 out:
3454 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
3455 		adapter->wait_capability = false;
3456 		ibmvnic_send_req_caps(adapter, 0);
3457 	}
3458 }
3459 
3460 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
3461 			       struct ibmvnic_adapter *adapter)
3462 {
3463 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
3464 	struct net_device *netdev = adapter->netdev;
3465 	struct device *dev = &adapter->vdev->dev;
3466 	u64 *u64_crq = (u64 *)crq;
3467 	long rc;
3468 
3469 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
3470 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3471 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3472 	switch (gen_crq->first) {
3473 	case IBMVNIC_CRQ_INIT_RSP:
3474 		switch (gen_crq->cmd) {
3475 		case IBMVNIC_CRQ_INIT:
3476 			dev_info(dev, "Partner initialized\n");
3477 			adapter->from_passive_init = true;
3478 			complete(&adapter->init_done);
3479 			break;
3480 		case IBMVNIC_CRQ_INIT_COMPLETE:
3481 			dev_info(dev, "Partner initialization complete\n");
3482 			send_version_xchg(adapter);
3483 			break;
3484 		default:
3485 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
3486 		}
3487 		return;
3488 	case IBMVNIC_CRQ_XPORT_EVENT:
3489 		netif_carrier_off(netdev);
3490 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
3491 			dev_info(dev, "Migrated, re-enabling adapter\n");
3492 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
3493 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
3494 			dev_info(dev, "Backing device failover detected\n");
3495 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
3496 		} else {
3497 			/* The adapter lost the connection */
3498 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
3499 				gen_crq->cmd);
3500 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3501 		}
3502 		return;
3503 	case IBMVNIC_CRQ_CMD_RSP:
3504 		break;
3505 	default:
3506 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
3507 			gen_crq->first);
3508 		return;
3509 	}
3510 
3511 	switch (gen_crq->cmd) {
3512 	case VERSION_EXCHANGE_RSP:
3513 		rc = crq->version_exchange_rsp.rc.code;
3514 		if (rc) {
3515 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
3516 			break;
3517 		}
3518 		dev_info(dev, "Partner protocol version is %d\n",
3519 			 crq->version_exchange_rsp.version);
3520 		if (be16_to_cpu(crq->version_exchange_rsp.version) <
3521 		    ibmvnic_version)
3522 			ibmvnic_version =
3523 			    be16_to_cpu(crq->version_exchange_rsp.version);
3524 		send_cap_queries(adapter);
3525 		break;
3526 	case QUERY_CAPABILITY_RSP:
3527 		handle_query_cap_rsp(crq, adapter);
3528 		break;
3529 	case QUERY_MAP_RSP:
3530 		handle_query_map_rsp(crq, adapter);
3531 		break;
3532 	case REQUEST_MAP_RSP:
3533 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
3534 		complete(&adapter->fw_done);
3535 		break;
3536 	case REQUEST_UNMAP_RSP:
3537 		handle_request_unmap_rsp(crq, adapter);
3538 		break;
3539 	case REQUEST_CAPABILITY_RSP:
3540 		handle_request_cap_rsp(crq, adapter);
3541 		break;
3542 	case LOGIN_RSP:
3543 		netdev_dbg(netdev, "Got Login Response\n");
3544 		handle_login_rsp(crq, adapter);
3545 		break;
3546 	case LOGICAL_LINK_STATE_RSP:
3547 		netdev_dbg(netdev,
3548 			   "Got Logical Link State Response, state: %d rc: %d\n",
3549 			   crq->logical_link_state_rsp.link_state,
3550 			   crq->logical_link_state_rsp.rc.code);
3551 		adapter->logical_link_state =
3552 		    crq->logical_link_state_rsp.link_state;
3553 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
3554 		complete(&adapter->init_done);
3555 		break;
3556 	case LINK_STATE_INDICATION:
3557 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
3558 		adapter->phys_link_state =
3559 		    crq->link_state_indication.phys_link_state;
3560 		adapter->logical_link_state =
3561 		    crq->link_state_indication.logical_link_state;
3562 		break;
3563 	case CHANGE_MAC_ADDR_RSP:
3564 		netdev_dbg(netdev, "Got MAC address change Response\n");
3565 		handle_change_mac_rsp(crq, adapter);
3566 		break;
3567 	case ERROR_INDICATION:
3568 		netdev_dbg(netdev, "Got Error Indication\n");
3569 		handle_error_indication(crq, adapter);
3570 		break;
3571 	case REQUEST_ERROR_RSP:
3572 		netdev_dbg(netdev, "Got Error Detail Response\n");
3573 		handle_error_info_rsp(crq, adapter);
3574 		break;
3575 	case REQUEST_STATISTICS_RSP:
3576 		netdev_dbg(netdev, "Got Statistics Response\n");
3577 		complete(&adapter->stats_done);
3578 		break;
3579 	case QUERY_IP_OFFLOAD_RSP:
3580 		netdev_dbg(netdev, "Got Query IP offload Response\n");
3581 		handle_query_ip_offload_rsp(adapter);
3582 		break;
3583 	case MULTICAST_CTRL_RSP:
3584 		netdev_dbg(netdev, "Got multicast control Response\n");
3585 		break;
3586 	case CONTROL_IP_OFFLOAD_RSP:
3587 		netdev_dbg(netdev, "Got Control IP offload Response\n");
3588 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
3589 				 sizeof(adapter->ip_offload_ctrl),
3590 				 DMA_TO_DEVICE);
3591 		complete(&adapter->init_done);
3592 		break;
3593 	case COLLECT_FW_TRACE_RSP:
3594 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
3595 		complete(&adapter->fw_done);
3596 		break;
3597 	default:
3598 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
3599 			   gen_crq->cmd);
3600 	}
3601 }
3602 
3603 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
3604 {
3605 	struct ibmvnic_adapter *adapter = instance;
3606 
3607 	tasklet_schedule(&adapter->tasklet);
3608 	return IRQ_HANDLED;
3609 }
3610 
3611 static void ibmvnic_tasklet(void *data)
3612 {
3613 	struct ibmvnic_adapter *adapter = data;
3614 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3615 	union ibmvnic_crq *crq;
3616 	unsigned long flags;
3617 	bool done = false;
3618 
3619 	spin_lock_irqsave(&queue->lock, flags);
3620 	while (!done) {
3621 		/* Pull all the valid messages off the CRQ */
3622 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
3623 			ibmvnic_handle_crq(crq, adapter);
3624 			crq->generic.first = 0;
3625 		}
3626 
3627 		/* remain in tasklet until all
3628 		 * capabilities responses are received
3629 		 */
3630 		if (!adapter->wait_capability)
3631 			done = true;
3632 	}
3633 	/* if capabilities CRQ's were sent in this tasklet, the following
3634 	 * tasklet must wait until all responses are received
3635 	 */
3636 	if (atomic_read(&adapter->running_cap_crqs) != 0)
3637 		adapter->wait_capability = true;
3638 	spin_unlock_irqrestore(&queue->lock, flags);
3639 }
3640 
3641 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
3642 {
3643 	struct vio_dev *vdev = adapter->vdev;
3644 	int rc;
3645 
3646 	do {
3647 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
3648 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
3649 
3650 	if (rc)
3651 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
3652 
3653 	return rc;
3654 }
3655 
3656 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
3657 {
3658 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3659 	struct device *dev = &adapter->vdev->dev;
3660 	struct vio_dev *vdev = adapter->vdev;
3661 	int rc;
3662 
3663 	/* Close the CRQ */
3664 	do {
3665 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3666 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3667 
3668 	/* Clean out the queue */
3669 	memset(crq->msgs, 0, PAGE_SIZE);
3670 	crq->cur = 0;
3671 
3672 	/* And re-open it again */
3673 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3674 				crq->msg_token, PAGE_SIZE);
3675 
3676 	if (rc == H_CLOSED)
3677 		/* Adapter is good, but other end is not ready */
3678 		dev_warn(dev, "Partner adapter not ready\n");
3679 	else if (rc != 0)
3680 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
3681 
3682 	return rc;
3683 }
3684 
3685 static void release_crq_queue(struct ibmvnic_adapter *adapter)
3686 {
3687 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3688 	struct vio_dev *vdev = adapter->vdev;
3689 	long rc;
3690 
3691 	if (!crq->msgs)
3692 		return;
3693 
3694 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
3695 	free_irq(vdev->irq, adapter);
3696 	tasklet_kill(&adapter->tasklet);
3697 	do {
3698 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3699 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3700 
3701 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
3702 			 DMA_BIDIRECTIONAL);
3703 	free_page((unsigned long)crq->msgs);
3704 	crq->msgs = NULL;
3705 }
3706 
3707 static int init_crq_queue(struct ibmvnic_adapter *adapter)
3708 {
3709 	struct ibmvnic_crq_queue *crq = &adapter->crq;
3710 	struct device *dev = &adapter->vdev->dev;
3711 	struct vio_dev *vdev = adapter->vdev;
3712 	int rc, retrc = -ENOMEM;
3713 
3714 	if (crq->msgs)
3715 		return 0;
3716 
3717 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
3718 	/* Should we allocate more than one page? */
3719 
3720 	if (!crq->msgs)
3721 		return -ENOMEM;
3722 
3723 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
3724 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
3725 					DMA_BIDIRECTIONAL);
3726 	if (dma_mapping_error(dev, crq->msg_token))
3727 		goto map_failed;
3728 
3729 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
3730 				crq->msg_token, PAGE_SIZE);
3731 
3732 	if (rc == H_RESOURCE)
3733 		/* maybe kexecing and resource is busy. try a reset */
3734 		rc = ibmvnic_reset_crq(adapter);
3735 	retrc = rc;
3736 
3737 	if (rc == H_CLOSED) {
3738 		dev_warn(dev, "Partner adapter not ready\n");
3739 	} else if (rc) {
3740 		dev_warn(dev, "Error %d opening adapter\n", rc);
3741 		goto reg_crq_failed;
3742 	}
3743 
3744 	retrc = 0;
3745 
3746 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
3747 		     (unsigned long)adapter);
3748 
3749 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
3750 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
3751 			 adapter);
3752 	if (rc) {
3753 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
3754 			vdev->irq, rc);
3755 		goto req_irq_failed;
3756 	}
3757 
3758 	rc = vio_enable_interrupts(vdev);
3759 	if (rc) {
3760 		dev_err(dev, "Error %d enabling interrupts\n", rc);
3761 		goto req_irq_failed;
3762 	}
3763 
3764 	crq->cur = 0;
3765 	spin_lock_init(&crq->lock);
3766 
3767 	return retrc;
3768 
3769 req_irq_failed:
3770 	tasklet_kill(&adapter->tasklet);
3771 	do {
3772 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
3773 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
3774 reg_crq_failed:
3775 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
3776 map_failed:
3777 	free_page((unsigned long)crq->msgs);
3778 	crq->msgs = NULL;
3779 	return retrc;
3780 }
3781 
3782 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
3783 {
3784 	struct device *dev = &adapter->vdev->dev;
3785 	unsigned long timeout = msecs_to_jiffies(30000);
3786 	int rc;
3787 
3788 	if (adapter->resetting) {
3789 		rc = ibmvnic_reset_crq(adapter);
3790 		if (!rc)
3791 			rc = vio_enable_interrupts(adapter->vdev);
3792 	} else {
3793 		rc = init_crq_queue(adapter);
3794 	}
3795 
3796 	if (rc) {
3797 		dev_err(dev, "Couldn't initialize crq. rc=%d\n", rc);
3798 		return rc;
3799 	}
3800 
3801 	adapter->from_passive_init = false;
3802 
3803 	init_completion(&adapter->init_done);
3804 	adapter->init_done_rc = 0;
3805 	ibmvnic_send_crq_init(adapter);
3806 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
3807 		dev_err(dev, "Initialization sequence timed out\n");
3808 		return -1;
3809 	}
3810 
3811 	if (adapter->init_done_rc) {
3812 		release_crq_queue(adapter);
3813 		return adapter->init_done_rc;
3814 	}
3815 
3816 	if (adapter->from_passive_init) {
3817 		adapter->state = VNIC_OPEN;
3818 		adapter->from_passive_init = false;
3819 		return -1;
3820 	}
3821 
3822 	if (adapter->resetting)
3823 		rc = reset_sub_crq_queues(adapter);
3824 	else
3825 		rc = init_sub_crqs(adapter);
3826 	if (rc) {
3827 		dev_err(dev, "Initialization of sub crqs failed\n");
3828 		release_crq_queue(adapter);
3829 		return rc;
3830 	}
3831 
3832 	rc = init_sub_crq_irqs(adapter);
3833 	if (rc) {
3834 		dev_err(dev, "Failed to initialize sub crq irqs\n");
3835 		release_crq_queue(adapter);
3836 	}
3837 
3838 	return rc;
3839 }
3840 
3841 static struct device_attribute dev_attr_failover;
3842 
3843 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
3844 {
3845 	struct ibmvnic_adapter *adapter;
3846 	struct net_device *netdev;
3847 	unsigned char *mac_addr_p;
3848 	int rc;
3849 
3850 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
3851 		dev->unit_address);
3852 
3853 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
3854 							VETH_MAC_ADDR, NULL);
3855 	if (!mac_addr_p) {
3856 		dev_err(&dev->dev,
3857 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
3858 			__FILE__, __LINE__);
3859 		return 0;
3860 	}
3861 
3862 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
3863 				   IBMVNIC_MAX_TX_QUEUES);
3864 	if (!netdev)
3865 		return -ENOMEM;
3866 
3867 	adapter = netdev_priv(netdev);
3868 	adapter->state = VNIC_PROBING;
3869 	dev_set_drvdata(&dev->dev, netdev);
3870 	adapter->vdev = dev;
3871 	adapter->netdev = netdev;
3872 
3873 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
3874 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
3875 	netdev->irq = dev->irq;
3876 	netdev->netdev_ops = &ibmvnic_netdev_ops;
3877 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
3878 	SET_NETDEV_DEV(netdev, &dev->dev);
3879 
3880 	spin_lock_init(&adapter->stats_lock);
3881 
3882 	INIT_LIST_HEAD(&adapter->errors);
3883 	spin_lock_init(&adapter->error_list_lock);
3884 
3885 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
3886 	INIT_LIST_HEAD(&adapter->rwi_list);
3887 	mutex_init(&adapter->reset_lock);
3888 	mutex_init(&adapter->rwi_lock);
3889 	adapter->resetting = false;
3890 
3891 	do {
3892 		rc = ibmvnic_init(adapter);
3893 		if (rc && rc != EAGAIN)
3894 			goto ibmvnic_init_fail;
3895 	} while (rc == EAGAIN);
3896 
3897 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
3898 
3899 	rc = device_create_file(&dev->dev, &dev_attr_failover);
3900 	if (rc)
3901 		goto ibmvnic_init_fail;
3902 
3903 	netif_carrier_off(netdev);
3904 	rc = register_netdev(netdev);
3905 	if (rc) {
3906 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
3907 		goto ibmvnic_register_fail;
3908 	}
3909 	dev_info(&dev->dev, "ibmvnic registered\n");
3910 
3911 	adapter->state = VNIC_PROBED;
3912 	return 0;
3913 
3914 ibmvnic_register_fail:
3915 	device_remove_file(&dev->dev, &dev_attr_failover);
3916 
3917 ibmvnic_init_fail:
3918 	release_sub_crqs(adapter);
3919 	release_crq_queue(adapter);
3920 	free_netdev(netdev);
3921 
3922 	return rc;
3923 }
3924 
3925 static int ibmvnic_remove(struct vio_dev *dev)
3926 {
3927 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
3928 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3929 
3930 	adapter->state = VNIC_REMOVING;
3931 	unregister_netdev(netdev);
3932 	mutex_lock(&adapter->reset_lock);
3933 
3934 	release_resources(adapter);
3935 	release_sub_crqs(adapter);
3936 	release_crq_queue(adapter);
3937 
3938 	adapter->state = VNIC_REMOVED;
3939 
3940 	mutex_unlock(&adapter->reset_lock);
3941 	device_remove_file(&dev->dev, &dev_attr_failover);
3942 	free_netdev(netdev);
3943 	dev_set_drvdata(&dev->dev, NULL);
3944 
3945 	return 0;
3946 }
3947 
3948 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
3949 			      const char *buf, size_t count)
3950 {
3951 	struct net_device *netdev = dev_get_drvdata(dev);
3952 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
3953 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
3954 	__be64 session_token;
3955 	long rc;
3956 
3957 	if (!sysfs_streq(buf, "1"))
3958 		return -EINVAL;
3959 
3960 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
3961 			 H_GET_SESSION_TOKEN, 0, 0, 0);
3962 	if (rc) {
3963 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
3964 			   rc);
3965 		return -EINVAL;
3966 	}
3967 
3968 	session_token = (__be64)retbuf[0];
3969 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
3970 		   be64_to_cpu(session_token));
3971 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3972 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
3973 	if (rc) {
3974 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
3975 			   rc);
3976 		return -EINVAL;
3977 	}
3978 
3979 	return count;
3980 }
3981 
3982 static DEVICE_ATTR(failover, 0200, NULL, failover_store);
3983 
3984 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
3985 {
3986 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
3987 	struct ibmvnic_adapter *adapter;
3988 	struct iommu_table *tbl;
3989 	unsigned long ret = 0;
3990 	int i;
3991 
3992 	tbl = get_iommu_table_base(&vdev->dev);
3993 
3994 	/* netdev inits at probe time along with the structures we need below*/
3995 	if (!netdev)
3996 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
3997 
3998 	adapter = netdev_priv(netdev);
3999 
4000 	ret += PAGE_SIZE; /* the crq message queue */
4001 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4002 
4003 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4004 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
4005 
4006 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4007 	     i++)
4008 		ret += adapter->rx_pool[i].size *
4009 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4010 
4011 	return ret;
4012 }
4013 
4014 static int ibmvnic_resume(struct device *dev)
4015 {
4016 	struct net_device *netdev = dev_get_drvdata(dev);
4017 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4018 
4019 	if (adapter->state != VNIC_OPEN)
4020 		return 0;
4021 
4022 	tasklet_schedule(&adapter->tasklet);
4023 
4024 	return 0;
4025 }
4026 
4027 static const struct vio_device_id ibmvnic_device_table[] = {
4028 	{"network", "IBM,vnic"},
4029 	{"", "" }
4030 };
4031 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4032 
4033 static const struct dev_pm_ops ibmvnic_pm_ops = {
4034 	.resume = ibmvnic_resume
4035 };
4036 
4037 static struct vio_driver ibmvnic_driver = {
4038 	.id_table       = ibmvnic_device_table,
4039 	.probe          = ibmvnic_probe,
4040 	.remove         = ibmvnic_remove,
4041 	.get_desired_dma = ibmvnic_get_desired_dma,
4042 	.name		= ibmvnic_driver_name,
4043 	.pm		= &ibmvnic_pm_ops,
4044 };
4045 
4046 /* module functions */
4047 static int __init ibmvnic_module_init(void)
4048 {
4049 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4050 		IBMVNIC_DRIVER_VERSION);
4051 
4052 	return vio_register_driver(&ibmvnic_driver);
4053 }
4054 
4055 static void __exit ibmvnic_module_exit(void)
4056 {
4057 	vio_unregister_driver(&ibmvnic_driver);
4058 }
4059 
4060 module_init(ibmvnic_module_init);
4061 module_exit(ibmvnic_module_exit);
4062