xref: /openbmc/linux/drivers/net/ethernet/ibm/ibmvnic.c (revision 800e26b8)
1 // SPDX-License-Identifier: GPL-2.0-or-later
2 /**************************************************************************/
3 /*                                                                        */
4 /*  IBM System i and System p Virtual NIC Device Driver                   */
5 /*  Copyright (C) 2014 IBM Corp.                                          */
6 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
7 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
8 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
9 /*                                                                        */
10 /*                                                                        */
11 /* This module contains the implementation of a virtual ethernet device   */
12 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
13 /* option of the RS/6000 Platform Architecture to interface with virtual  */
14 /* ethernet NICs that are presented to the partition by the hypervisor.   */
15 /*									   */
16 /* Messages are passed between the VNIC driver and the VNIC server using  */
17 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
18 /* issue and receive commands that initiate communication with the server */
19 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
20 /* are used by the driver to notify the server that a packet is           */
21 /* ready for transmission or that a buffer has been added to receive a    */
22 /* packet. Subsequently, sCRQs are used by the server to notify the       */
23 /* driver that a packet transmission has been completed or that a packet  */
24 /* has been received and placed in a waiting buffer.                      */
25 /*                                                                        */
26 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
27 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
28 /* or receive has been completed, the VNIC driver is required to use      */
29 /* "long term mapping". This entails that large, continuous DMA mapped    */
30 /* buffers are allocated on driver initialization and these buffers are   */
31 /* then continuously reused to pass skbs to and from the VNIC server.     */
32 /*                                                                        */
33 /**************************************************************************/
34 
35 #include <linux/module.h>
36 #include <linux/moduleparam.h>
37 #include <linux/types.h>
38 #include <linux/errno.h>
39 #include <linux/completion.h>
40 #include <linux/ioport.h>
41 #include <linux/dma-mapping.h>
42 #include <linux/kernel.h>
43 #include <linux/netdevice.h>
44 #include <linux/etherdevice.h>
45 #include <linux/skbuff.h>
46 #include <linux/init.h>
47 #include <linux/delay.h>
48 #include <linux/mm.h>
49 #include <linux/ethtool.h>
50 #include <linux/proc_fs.h>
51 #include <linux/if_arp.h>
52 #include <linux/in.h>
53 #include <linux/ip.h>
54 #include <linux/ipv6.h>
55 #include <linux/irq.h>
56 #include <linux/kthread.h>
57 #include <linux/seq_file.h>
58 #include <linux/interrupt.h>
59 #include <net/net_namespace.h>
60 #include <asm/hvcall.h>
61 #include <linux/atomic.h>
62 #include <asm/vio.h>
63 #include <asm/iommu.h>
64 #include <linux/uaccess.h>
65 #include <asm/firmware.h>
66 #include <linux/workqueue.h>
67 #include <linux/if_vlan.h>
68 #include <linux/utsname.h>
69 
70 #include "ibmvnic.h"
71 
72 static const char ibmvnic_driver_name[] = "ibmvnic";
73 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
74 
75 MODULE_AUTHOR("Santiago Leon");
76 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
77 MODULE_LICENSE("GPL");
78 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
79 
80 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
81 static int ibmvnic_remove(struct vio_dev *);
82 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
83 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
84 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
85 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
86 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
87 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
88 		       union sub_crq *sub_crq);
89 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
90 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
91 static int enable_scrq_irq(struct ibmvnic_adapter *,
92 			   struct ibmvnic_sub_crq_queue *);
93 static int disable_scrq_irq(struct ibmvnic_adapter *,
94 			    struct ibmvnic_sub_crq_queue *);
95 static int pending_scrq(struct ibmvnic_adapter *,
96 			struct ibmvnic_sub_crq_queue *);
97 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
98 					struct ibmvnic_sub_crq_queue *);
99 static int ibmvnic_poll(struct napi_struct *napi, int data);
100 static void send_map_query(struct ibmvnic_adapter *adapter);
101 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
102 static int send_request_unmap(struct ibmvnic_adapter *, u8);
103 static int send_login(struct ibmvnic_adapter *adapter);
104 static void send_cap_queries(struct ibmvnic_adapter *adapter);
105 static int init_sub_crqs(struct ibmvnic_adapter *);
106 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
107 static int ibmvnic_init(struct ibmvnic_adapter *);
108 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
109 static void release_crq_queue(struct ibmvnic_adapter *);
110 static int __ibmvnic_set_mac(struct net_device *, u8 *);
111 static int init_crq_queue(struct ibmvnic_adapter *adapter);
112 static int send_query_phys_parms(struct ibmvnic_adapter *adapter);
113 
114 struct ibmvnic_stat {
115 	char name[ETH_GSTRING_LEN];
116 	int offset;
117 };
118 
119 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
120 			     offsetof(struct ibmvnic_statistics, stat))
121 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
122 
123 static const struct ibmvnic_stat ibmvnic_stats[] = {
124 	{"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
125 	{"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
126 	{"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
127 	{"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
128 	{"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
129 	{"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
130 	{"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
131 	{"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
132 	{"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
133 	{"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
134 	{"align_errors", IBMVNIC_STAT_OFF(align_errors)},
135 	{"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
136 	{"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
137 	{"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
138 	{"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
139 	{"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
140 	{"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
141 	{"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
142 	{"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
143 	{"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
144 	{"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
145 	{"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
146 };
147 
148 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
149 			  unsigned long length, unsigned long *number,
150 			  unsigned long *irq)
151 {
152 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
153 	long rc;
154 
155 	rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
156 	*number = retbuf[0];
157 	*irq = retbuf[1];
158 
159 	return rc;
160 }
161 
162 /**
163  * ibmvnic_wait_for_completion - Check device state and wait for completion
164  * @adapter: private device data
165  * @comp_done: completion structure to wait for
166  * @timeout: time to wait in milliseconds
167  *
168  * Wait for a completion signal or until the timeout limit is reached
169  * while checking that the device is still active.
170  */
171 static int ibmvnic_wait_for_completion(struct ibmvnic_adapter *adapter,
172 				       struct completion *comp_done,
173 				       unsigned long timeout)
174 {
175 	struct net_device *netdev;
176 	unsigned long div_timeout;
177 	u8 retry;
178 
179 	netdev = adapter->netdev;
180 	retry = 5;
181 	div_timeout = msecs_to_jiffies(timeout / retry);
182 	while (true) {
183 		if (!adapter->crq.active) {
184 			netdev_err(netdev, "Device down!\n");
185 			return -ENODEV;
186 		}
187 		if (!retry--)
188 			break;
189 		if (wait_for_completion_timeout(comp_done, div_timeout))
190 			return 0;
191 	}
192 	netdev_err(netdev, "Operation timed out.\n");
193 	return -ETIMEDOUT;
194 }
195 
196 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
197 				struct ibmvnic_long_term_buff *ltb, int size)
198 {
199 	struct device *dev = &adapter->vdev->dev;
200 	int rc;
201 
202 	ltb->size = size;
203 	ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
204 				       GFP_KERNEL);
205 
206 	if (!ltb->buff) {
207 		dev_err(dev, "Couldn't alloc long term buffer\n");
208 		return -ENOMEM;
209 	}
210 	ltb->map_id = adapter->map_id;
211 	adapter->map_id++;
212 
213 	mutex_lock(&adapter->fw_lock);
214 	adapter->fw_done_rc = 0;
215 	reinit_completion(&adapter->fw_done);
216 	rc = send_request_map(adapter, ltb->addr,
217 			      ltb->size, ltb->map_id);
218 	if (rc) {
219 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
220 		mutex_unlock(&adapter->fw_lock);
221 		return rc;
222 	}
223 
224 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
225 	if (rc) {
226 		dev_err(dev,
227 			"Long term map request aborted or timed out,rc = %d\n",
228 			rc);
229 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
230 		mutex_unlock(&adapter->fw_lock);
231 		return rc;
232 	}
233 
234 	if (adapter->fw_done_rc) {
235 		dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
236 			adapter->fw_done_rc);
237 		dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
238 		mutex_unlock(&adapter->fw_lock);
239 		return -1;
240 	}
241 	mutex_unlock(&adapter->fw_lock);
242 	return 0;
243 }
244 
245 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
246 				struct ibmvnic_long_term_buff *ltb)
247 {
248 	struct device *dev = &adapter->vdev->dev;
249 
250 	if (!ltb->buff)
251 		return;
252 
253 	if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
254 	    adapter->reset_reason != VNIC_RESET_MOBILITY)
255 		send_request_unmap(adapter, ltb->map_id);
256 	dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
257 }
258 
259 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
260 				struct ibmvnic_long_term_buff *ltb)
261 {
262 	struct device *dev = &adapter->vdev->dev;
263 	int rc;
264 
265 	memset(ltb->buff, 0, ltb->size);
266 
267 	mutex_lock(&adapter->fw_lock);
268 	adapter->fw_done_rc = 0;
269 
270 	reinit_completion(&adapter->fw_done);
271 	rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
272 	if (rc) {
273 		mutex_unlock(&adapter->fw_lock);
274 		return rc;
275 	}
276 
277 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
278 	if (rc) {
279 		dev_info(dev,
280 			 "Reset failed, long term map request timed out or aborted\n");
281 		mutex_unlock(&adapter->fw_lock);
282 		return rc;
283 	}
284 
285 	if (adapter->fw_done_rc) {
286 		dev_info(dev,
287 			 "Reset failed, attempting to free and reallocate buffer\n");
288 		free_long_term_buff(adapter, ltb);
289 		mutex_unlock(&adapter->fw_lock);
290 		return alloc_long_term_buff(adapter, ltb, ltb->size);
291 	}
292 	mutex_unlock(&adapter->fw_lock);
293 	return 0;
294 }
295 
296 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
297 {
298 	int i;
299 
300 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
301 	     i++)
302 		adapter->rx_pool[i].active = 0;
303 }
304 
305 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
306 			      struct ibmvnic_rx_pool *pool)
307 {
308 	int count = pool->size - atomic_read(&pool->available);
309 	struct device *dev = &adapter->vdev->dev;
310 	int buffers_added = 0;
311 	unsigned long lpar_rc;
312 	union sub_crq sub_crq;
313 	struct sk_buff *skb;
314 	unsigned int offset;
315 	dma_addr_t dma_addr;
316 	unsigned char *dst;
317 	u64 *handle_array;
318 	int shift = 0;
319 	int index;
320 	int i;
321 
322 	if (!pool->active)
323 		return;
324 
325 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
326 				      be32_to_cpu(adapter->login_rsp_buf->
327 				      off_rxadd_subcrqs));
328 
329 	for (i = 0; i < count; ++i) {
330 		skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
331 		if (!skb) {
332 			dev_err(dev, "Couldn't replenish rx buff\n");
333 			adapter->replenish_no_mem++;
334 			break;
335 		}
336 
337 		index = pool->free_map[pool->next_free];
338 
339 		if (pool->rx_buff[index].skb)
340 			dev_err(dev, "Inconsistent free_map!\n");
341 
342 		/* Copy the skb to the long term mapped DMA buffer */
343 		offset = index * pool->buff_size;
344 		dst = pool->long_term_buff.buff + offset;
345 		memset(dst, 0, pool->buff_size);
346 		dma_addr = pool->long_term_buff.addr + offset;
347 		pool->rx_buff[index].data = dst;
348 
349 		pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
350 		pool->rx_buff[index].dma = dma_addr;
351 		pool->rx_buff[index].skb = skb;
352 		pool->rx_buff[index].pool_index = pool->index;
353 		pool->rx_buff[index].size = pool->buff_size;
354 
355 		memset(&sub_crq, 0, sizeof(sub_crq));
356 		sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
357 		sub_crq.rx_add.correlator =
358 		    cpu_to_be64((u64)&pool->rx_buff[index]);
359 		sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
360 		sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
361 
362 		/* The length field of the sCRQ is defined to be 24 bits so the
363 		 * buffer size needs to be left shifted by a byte before it is
364 		 * converted to big endian to prevent the last byte from being
365 		 * truncated.
366 		 */
367 #ifdef __LITTLE_ENDIAN__
368 		shift = 8;
369 #endif
370 		sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
371 
372 		lpar_rc = send_subcrq(adapter, handle_array[pool->index],
373 				      &sub_crq);
374 		if (lpar_rc != H_SUCCESS)
375 			goto failure;
376 
377 		buffers_added++;
378 		adapter->replenish_add_buff_success++;
379 		pool->next_free = (pool->next_free + 1) % pool->size;
380 	}
381 	atomic_add(buffers_added, &pool->available);
382 	return;
383 
384 failure:
385 	if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
386 		dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
387 	pool->free_map[pool->next_free] = index;
388 	pool->rx_buff[index].skb = NULL;
389 
390 	dev_kfree_skb_any(skb);
391 	adapter->replenish_add_buff_failure++;
392 	atomic_add(buffers_added, &pool->available);
393 
394 	if (lpar_rc == H_CLOSED || adapter->failover_pending) {
395 		/* Disable buffer pool replenishment and report carrier off if
396 		 * queue is closed or pending failover.
397 		 * Firmware guarantees that a signal will be sent to the
398 		 * driver, triggering a reset.
399 		 */
400 		deactivate_rx_pools(adapter);
401 		netif_carrier_off(adapter->netdev);
402 	}
403 }
404 
405 static void replenish_pools(struct ibmvnic_adapter *adapter)
406 {
407 	int i;
408 
409 	adapter->replenish_task_cycles++;
410 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
411 	     i++) {
412 		if (adapter->rx_pool[i].active)
413 			replenish_rx_pool(adapter, &adapter->rx_pool[i]);
414 	}
415 }
416 
417 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
418 {
419 	kfree(adapter->tx_stats_buffers);
420 	kfree(adapter->rx_stats_buffers);
421 	adapter->tx_stats_buffers = NULL;
422 	adapter->rx_stats_buffers = NULL;
423 }
424 
425 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
426 {
427 	adapter->tx_stats_buffers =
428 				kcalloc(IBMVNIC_MAX_QUEUES,
429 					sizeof(struct ibmvnic_tx_queue_stats),
430 					GFP_KERNEL);
431 	if (!adapter->tx_stats_buffers)
432 		return -ENOMEM;
433 
434 	adapter->rx_stats_buffers =
435 				kcalloc(IBMVNIC_MAX_QUEUES,
436 					sizeof(struct ibmvnic_rx_queue_stats),
437 					GFP_KERNEL);
438 	if (!adapter->rx_stats_buffers)
439 		return -ENOMEM;
440 
441 	return 0;
442 }
443 
444 static void release_stats_token(struct ibmvnic_adapter *adapter)
445 {
446 	struct device *dev = &adapter->vdev->dev;
447 
448 	if (!adapter->stats_token)
449 		return;
450 
451 	dma_unmap_single(dev, adapter->stats_token,
452 			 sizeof(struct ibmvnic_statistics),
453 			 DMA_FROM_DEVICE);
454 	adapter->stats_token = 0;
455 }
456 
457 static int init_stats_token(struct ibmvnic_adapter *adapter)
458 {
459 	struct device *dev = &adapter->vdev->dev;
460 	dma_addr_t stok;
461 
462 	stok = dma_map_single(dev, &adapter->stats,
463 			      sizeof(struct ibmvnic_statistics),
464 			      DMA_FROM_DEVICE);
465 	if (dma_mapping_error(dev, stok)) {
466 		dev_err(dev, "Couldn't map stats buffer\n");
467 		return -1;
468 	}
469 
470 	adapter->stats_token = stok;
471 	netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
472 	return 0;
473 }
474 
475 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
476 {
477 	struct ibmvnic_rx_pool *rx_pool;
478 	int rx_scrqs;
479 	int i, j, rc;
480 	u64 *size_array;
481 
482 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
483 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
484 
485 	rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
486 	for (i = 0; i < rx_scrqs; i++) {
487 		rx_pool = &adapter->rx_pool[i];
488 
489 		netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
490 
491 		if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
492 			free_long_term_buff(adapter, &rx_pool->long_term_buff);
493 			rx_pool->buff_size = be64_to_cpu(size_array[i]);
494 			rc = alloc_long_term_buff(adapter,
495 						  &rx_pool->long_term_buff,
496 						  rx_pool->size *
497 						  rx_pool->buff_size);
498 		} else {
499 			rc = reset_long_term_buff(adapter,
500 						  &rx_pool->long_term_buff);
501 		}
502 
503 		if (rc)
504 			return rc;
505 
506 		for (j = 0; j < rx_pool->size; j++)
507 			rx_pool->free_map[j] = j;
508 
509 		memset(rx_pool->rx_buff, 0,
510 		       rx_pool->size * sizeof(struct ibmvnic_rx_buff));
511 
512 		atomic_set(&rx_pool->available, 0);
513 		rx_pool->next_alloc = 0;
514 		rx_pool->next_free = 0;
515 		rx_pool->active = 1;
516 	}
517 
518 	return 0;
519 }
520 
521 static void release_rx_pools(struct ibmvnic_adapter *adapter)
522 {
523 	struct ibmvnic_rx_pool *rx_pool;
524 	int i, j;
525 
526 	if (!adapter->rx_pool)
527 		return;
528 
529 	for (i = 0; i < adapter->num_active_rx_pools; i++) {
530 		rx_pool = &adapter->rx_pool[i];
531 
532 		netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
533 
534 		kfree(rx_pool->free_map);
535 		free_long_term_buff(adapter, &rx_pool->long_term_buff);
536 
537 		if (!rx_pool->rx_buff)
538 			continue;
539 
540 		for (j = 0; j < rx_pool->size; j++) {
541 			if (rx_pool->rx_buff[j].skb) {
542 				dev_kfree_skb_any(rx_pool->rx_buff[j].skb);
543 				rx_pool->rx_buff[j].skb = NULL;
544 			}
545 		}
546 
547 		kfree(rx_pool->rx_buff);
548 	}
549 
550 	kfree(adapter->rx_pool);
551 	adapter->rx_pool = NULL;
552 	adapter->num_active_rx_pools = 0;
553 }
554 
555 static int init_rx_pools(struct net_device *netdev)
556 {
557 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
558 	struct device *dev = &adapter->vdev->dev;
559 	struct ibmvnic_rx_pool *rx_pool;
560 	int rxadd_subcrqs;
561 	u64 *size_array;
562 	int i, j;
563 
564 	rxadd_subcrqs =
565 		be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
566 	size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
567 		be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
568 
569 	adapter->rx_pool = kcalloc(rxadd_subcrqs,
570 				   sizeof(struct ibmvnic_rx_pool),
571 				   GFP_KERNEL);
572 	if (!adapter->rx_pool) {
573 		dev_err(dev, "Failed to allocate rx pools\n");
574 		return -1;
575 	}
576 
577 	adapter->num_active_rx_pools = rxadd_subcrqs;
578 
579 	for (i = 0; i < rxadd_subcrqs; i++) {
580 		rx_pool = &adapter->rx_pool[i];
581 
582 		netdev_dbg(adapter->netdev,
583 			   "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
584 			   i, adapter->req_rx_add_entries_per_subcrq,
585 			   be64_to_cpu(size_array[i]));
586 
587 		rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
588 		rx_pool->index = i;
589 		rx_pool->buff_size = be64_to_cpu(size_array[i]);
590 		rx_pool->active = 1;
591 
592 		rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
593 					    GFP_KERNEL);
594 		if (!rx_pool->free_map) {
595 			release_rx_pools(adapter);
596 			return -1;
597 		}
598 
599 		rx_pool->rx_buff = kcalloc(rx_pool->size,
600 					   sizeof(struct ibmvnic_rx_buff),
601 					   GFP_KERNEL);
602 		if (!rx_pool->rx_buff) {
603 			dev_err(dev, "Couldn't alloc rx buffers\n");
604 			release_rx_pools(adapter);
605 			return -1;
606 		}
607 
608 		if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
609 					 rx_pool->size * rx_pool->buff_size)) {
610 			release_rx_pools(adapter);
611 			return -1;
612 		}
613 
614 		for (j = 0; j < rx_pool->size; ++j)
615 			rx_pool->free_map[j] = j;
616 
617 		atomic_set(&rx_pool->available, 0);
618 		rx_pool->next_alloc = 0;
619 		rx_pool->next_free = 0;
620 	}
621 
622 	return 0;
623 }
624 
625 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
626 			     struct ibmvnic_tx_pool *tx_pool)
627 {
628 	int rc, i;
629 
630 	rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
631 	if (rc)
632 		return rc;
633 
634 	memset(tx_pool->tx_buff, 0,
635 	       tx_pool->num_buffers *
636 	       sizeof(struct ibmvnic_tx_buff));
637 
638 	for (i = 0; i < tx_pool->num_buffers; i++)
639 		tx_pool->free_map[i] = i;
640 
641 	tx_pool->consumer_index = 0;
642 	tx_pool->producer_index = 0;
643 
644 	return 0;
645 }
646 
647 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
648 {
649 	int tx_scrqs;
650 	int i, rc;
651 
652 	tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
653 	for (i = 0; i < tx_scrqs; i++) {
654 		rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
655 		if (rc)
656 			return rc;
657 		rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
658 		if (rc)
659 			return rc;
660 	}
661 
662 	return 0;
663 }
664 
665 static void release_vpd_data(struct ibmvnic_adapter *adapter)
666 {
667 	if (!adapter->vpd)
668 		return;
669 
670 	kfree(adapter->vpd->buff);
671 	kfree(adapter->vpd);
672 
673 	adapter->vpd = NULL;
674 }
675 
676 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
677 				struct ibmvnic_tx_pool *tx_pool)
678 {
679 	kfree(tx_pool->tx_buff);
680 	kfree(tx_pool->free_map);
681 	free_long_term_buff(adapter, &tx_pool->long_term_buff);
682 }
683 
684 static void release_tx_pools(struct ibmvnic_adapter *adapter)
685 {
686 	int i;
687 
688 	if (!adapter->tx_pool)
689 		return;
690 
691 	for (i = 0; i < adapter->num_active_tx_pools; i++) {
692 		release_one_tx_pool(adapter, &adapter->tx_pool[i]);
693 		release_one_tx_pool(adapter, &adapter->tso_pool[i]);
694 	}
695 
696 	kfree(adapter->tx_pool);
697 	adapter->tx_pool = NULL;
698 	kfree(adapter->tso_pool);
699 	adapter->tso_pool = NULL;
700 	adapter->num_active_tx_pools = 0;
701 }
702 
703 static int init_one_tx_pool(struct net_device *netdev,
704 			    struct ibmvnic_tx_pool *tx_pool,
705 			    int num_entries, int buf_size)
706 {
707 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
708 	int i;
709 
710 	tx_pool->tx_buff = kcalloc(num_entries,
711 				   sizeof(struct ibmvnic_tx_buff),
712 				   GFP_KERNEL);
713 	if (!tx_pool->tx_buff)
714 		return -1;
715 
716 	if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
717 				 num_entries * buf_size))
718 		return -1;
719 
720 	tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
721 	if (!tx_pool->free_map)
722 		return -1;
723 
724 	for (i = 0; i < num_entries; i++)
725 		tx_pool->free_map[i] = i;
726 
727 	tx_pool->consumer_index = 0;
728 	tx_pool->producer_index = 0;
729 	tx_pool->num_buffers = num_entries;
730 	tx_pool->buf_size = buf_size;
731 
732 	return 0;
733 }
734 
735 static int init_tx_pools(struct net_device *netdev)
736 {
737 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
738 	int tx_subcrqs;
739 	int i, rc;
740 
741 	tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
742 	adapter->tx_pool = kcalloc(tx_subcrqs,
743 				   sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
744 	if (!adapter->tx_pool)
745 		return -1;
746 
747 	adapter->tso_pool = kcalloc(tx_subcrqs,
748 				    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
749 	if (!adapter->tso_pool)
750 		return -1;
751 
752 	adapter->num_active_tx_pools = tx_subcrqs;
753 
754 	for (i = 0; i < tx_subcrqs; i++) {
755 		rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
756 				      adapter->req_tx_entries_per_subcrq,
757 				      adapter->req_mtu + VLAN_HLEN);
758 		if (rc) {
759 			release_tx_pools(adapter);
760 			return rc;
761 		}
762 
763 		rc = init_one_tx_pool(netdev, &adapter->tso_pool[i],
764 				      IBMVNIC_TSO_BUFS,
765 				      IBMVNIC_TSO_BUF_SZ);
766 		if (rc) {
767 			release_tx_pools(adapter);
768 			return rc;
769 		}
770 	}
771 
772 	return 0;
773 }
774 
775 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
776 {
777 	int i;
778 
779 	if (adapter->napi_enabled)
780 		return;
781 
782 	for (i = 0; i < adapter->req_rx_queues; i++)
783 		napi_enable(&adapter->napi[i]);
784 
785 	adapter->napi_enabled = true;
786 }
787 
788 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
789 {
790 	int i;
791 
792 	if (!adapter->napi_enabled)
793 		return;
794 
795 	for (i = 0; i < adapter->req_rx_queues; i++) {
796 		netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
797 		napi_disable(&adapter->napi[i]);
798 	}
799 
800 	adapter->napi_enabled = false;
801 }
802 
803 static int init_napi(struct ibmvnic_adapter *adapter)
804 {
805 	int i;
806 
807 	adapter->napi = kcalloc(adapter->req_rx_queues,
808 				sizeof(struct napi_struct), GFP_KERNEL);
809 	if (!adapter->napi)
810 		return -ENOMEM;
811 
812 	for (i = 0; i < adapter->req_rx_queues; i++) {
813 		netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
814 		netif_napi_add(adapter->netdev, &adapter->napi[i],
815 			       ibmvnic_poll, NAPI_POLL_WEIGHT);
816 	}
817 
818 	adapter->num_active_rx_napi = adapter->req_rx_queues;
819 	return 0;
820 }
821 
822 static void release_napi(struct ibmvnic_adapter *adapter)
823 {
824 	int i;
825 
826 	if (!adapter->napi)
827 		return;
828 
829 	for (i = 0; i < adapter->num_active_rx_napi; i++) {
830 		netdev_dbg(adapter->netdev, "Releasing napi[%d]\n", i);
831 		netif_napi_del(&adapter->napi[i]);
832 	}
833 
834 	kfree(adapter->napi);
835 	adapter->napi = NULL;
836 	adapter->num_active_rx_napi = 0;
837 	adapter->napi_enabled = false;
838 }
839 
840 static int ibmvnic_login(struct net_device *netdev)
841 {
842 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
843 	unsigned long timeout = msecs_to_jiffies(30000);
844 	int retry_count = 0;
845 	int retries = 10;
846 	bool retry;
847 	int rc;
848 
849 	do {
850 		retry = false;
851 		if (retry_count > retries) {
852 			netdev_warn(netdev, "Login attempts exceeded\n");
853 			return -1;
854 		}
855 
856 		adapter->init_done_rc = 0;
857 		reinit_completion(&adapter->init_done);
858 		rc = send_login(adapter);
859 		if (rc) {
860 			netdev_warn(netdev, "Unable to login\n");
861 			return rc;
862 		}
863 
864 		if (!wait_for_completion_timeout(&adapter->init_done,
865 						 timeout)) {
866 			netdev_warn(netdev, "Login timed out, retrying...\n");
867 			retry = true;
868 			adapter->init_done_rc = 0;
869 			retry_count++;
870 			continue;
871 		}
872 
873 		if (adapter->init_done_rc == ABORTED) {
874 			netdev_warn(netdev, "Login aborted, retrying...\n");
875 			retry = true;
876 			adapter->init_done_rc = 0;
877 			retry_count++;
878 			/* FW or device may be busy, so
879 			 * wait a bit before retrying login
880 			 */
881 			msleep(500);
882 		} else if (adapter->init_done_rc == PARTIALSUCCESS) {
883 			retry_count++;
884 			release_sub_crqs(adapter, 1);
885 
886 			retry = true;
887 			netdev_dbg(netdev,
888 				   "Received partial success, retrying...\n");
889 			adapter->init_done_rc = 0;
890 			reinit_completion(&adapter->init_done);
891 			send_cap_queries(adapter);
892 			if (!wait_for_completion_timeout(&adapter->init_done,
893 							 timeout)) {
894 				netdev_warn(netdev,
895 					    "Capabilities query timed out\n");
896 				return -1;
897 			}
898 
899 			rc = init_sub_crqs(adapter);
900 			if (rc) {
901 				netdev_warn(netdev,
902 					    "SCRQ initialization failed\n");
903 				return -1;
904 			}
905 
906 			rc = init_sub_crq_irqs(adapter);
907 			if (rc) {
908 				netdev_warn(netdev,
909 					    "SCRQ irq initialization failed\n");
910 				return -1;
911 			}
912 		} else if (adapter->init_done_rc) {
913 			netdev_warn(netdev, "Adapter login failed\n");
914 			return -1;
915 		}
916 	} while (retry);
917 
918 	__ibmvnic_set_mac(netdev, adapter->mac_addr);
919 
920 	return 0;
921 }
922 
923 static void release_login_buffer(struct ibmvnic_adapter *adapter)
924 {
925 	kfree(adapter->login_buf);
926 	adapter->login_buf = NULL;
927 }
928 
929 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
930 {
931 	kfree(adapter->login_rsp_buf);
932 	adapter->login_rsp_buf = NULL;
933 }
934 
935 static void release_resources(struct ibmvnic_adapter *adapter)
936 {
937 	release_vpd_data(adapter);
938 
939 	release_tx_pools(adapter);
940 	release_rx_pools(adapter);
941 
942 	release_napi(adapter);
943 	release_login_rsp_buffer(adapter);
944 }
945 
946 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
947 {
948 	struct net_device *netdev = adapter->netdev;
949 	unsigned long timeout = msecs_to_jiffies(30000);
950 	union ibmvnic_crq crq;
951 	bool resend;
952 	int rc;
953 
954 	netdev_dbg(netdev, "setting link state %d\n", link_state);
955 
956 	memset(&crq, 0, sizeof(crq));
957 	crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
958 	crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
959 	crq.logical_link_state.link_state = link_state;
960 
961 	do {
962 		resend = false;
963 
964 		reinit_completion(&adapter->init_done);
965 		rc = ibmvnic_send_crq(adapter, &crq);
966 		if (rc) {
967 			netdev_err(netdev, "Failed to set link state\n");
968 			return rc;
969 		}
970 
971 		if (!wait_for_completion_timeout(&adapter->init_done,
972 						 timeout)) {
973 			netdev_err(netdev, "timeout setting link state\n");
974 			return -1;
975 		}
976 
977 		if (adapter->init_done_rc == 1) {
978 			/* Partuial success, delay and re-send */
979 			mdelay(1000);
980 			resend = true;
981 		} else if (adapter->init_done_rc) {
982 			netdev_warn(netdev, "Unable to set link state, rc=%d\n",
983 				    adapter->init_done_rc);
984 			return adapter->init_done_rc;
985 		}
986 	} while (resend);
987 
988 	return 0;
989 }
990 
991 static int set_real_num_queues(struct net_device *netdev)
992 {
993 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
994 	int rc;
995 
996 	netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
997 		   adapter->req_tx_queues, adapter->req_rx_queues);
998 
999 	rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
1000 	if (rc) {
1001 		netdev_err(netdev, "failed to set the number of tx queues\n");
1002 		return rc;
1003 	}
1004 
1005 	rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
1006 	if (rc)
1007 		netdev_err(netdev, "failed to set the number of rx queues\n");
1008 
1009 	return rc;
1010 }
1011 
1012 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
1013 {
1014 	struct device *dev = &adapter->vdev->dev;
1015 	union ibmvnic_crq crq;
1016 	int len = 0;
1017 	int rc;
1018 
1019 	if (adapter->vpd->buff)
1020 		len = adapter->vpd->len;
1021 
1022 	mutex_lock(&adapter->fw_lock);
1023 	adapter->fw_done_rc = 0;
1024 	reinit_completion(&adapter->fw_done);
1025 
1026 	crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
1027 	crq.get_vpd_size.cmd = GET_VPD_SIZE;
1028 	rc = ibmvnic_send_crq(adapter, &crq);
1029 	if (rc) {
1030 		mutex_unlock(&adapter->fw_lock);
1031 		return rc;
1032 	}
1033 
1034 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1035 	if (rc) {
1036 		dev_err(dev, "Could not retrieve VPD size, rc = %d\n", rc);
1037 		mutex_unlock(&adapter->fw_lock);
1038 		return rc;
1039 	}
1040 	mutex_unlock(&adapter->fw_lock);
1041 
1042 	if (!adapter->vpd->len)
1043 		return -ENODATA;
1044 
1045 	if (!adapter->vpd->buff)
1046 		adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
1047 	else if (adapter->vpd->len != len)
1048 		adapter->vpd->buff =
1049 			krealloc(adapter->vpd->buff,
1050 				 adapter->vpd->len, GFP_KERNEL);
1051 
1052 	if (!adapter->vpd->buff) {
1053 		dev_err(dev, "Could allocate VPD buffer\n");
1054 		return -ENOMEM;
1055 	}
1056 
1057 	adapter->vpd->dma_addr =
1058 		dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
1059 			       DMA_FROM_DEVICE);
1060 	if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
1061 		dev_err(dev, "Could not map VPD buffer\n");
1062 		kfree(adapter->vpd->buff);
1063 		adapter->vpd->buff = NULL;
1064 		return -ENOMEM;
1065 	}
1066 
1067 	mutex_lock(&adapter->fw_lock);
1068 	adapter->fw_done_rc = 0;
1069 	reinit_completion(&adapter->fw_done);
1070 
1071 	crq.get_vpd.first = IBMVNIC_CRQ_CMD;
1072 	crq.get_vpd.cmd = GET_VPD;
1073 	crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
1074 	crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1075 	rc = ibmvnic_send_crq(adapter, &crq);
1076 	if (rc) {
1077 		kfree(adapter->vpd->buff);
1078 		adapter->vpd->buff = NULL;
1079 		mutex_unlock(&adapter->fw_lock);
1080 		return rc;
1081 	}
1082 
1083 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1084 	if (rc) {
1085 		dev_err(dev, "Unable to retrieve VPD, rc = %d\n", rc);
1086 		kfree(adapter->vpd->buff);
1087 		adapter->vpd->buff = NULL;
1088 		mutex_unlock(&adapter->fw_lock);
1089 		return rc;
1090 	}
1091 
1092 	mutex_unlock(&adapter->fw_lock);
1093 	return 0;
1094 }
1095 
1096 static int init_resources(struct ibmvnic_adapter *adapter)
1097 {
1098 	struct net_device *netdev = adapter->netdev;
1099 	int rc;
1100 
1101 	rc = set_real_num_queues(netdev);
1102 	if (rc)
1103 		return rc;
1104 
1105 	adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1106 	if (!adapter->vpd)
1107 		return -ENOMEM;
1108 
1109 	/* Vital Product Data (VPD) */
1110 	rc = ibmvnic_get_vpd(adapter);
1111 	if (rc) {
1112 		netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1113 		return rc;
1114 	}
1115 
1116 	adapter->map_id = 1;
1117 
1118 	rc = init_napi(adapter);
1119 	if (rc)
1120 		return rc;
1121 
1122 	send_map_query(adapter);
1123 
1124 	rc = init_rx_pools(netdev);
1125 	if (rc)
1126 		return rc;
1127 
1128 	rc = init_tx_pools(netdev);
1129 	return rc;
1130 }
1131 
1132 static int __ibmvnic_open(struct net_device *netdev)
1133 {
1134 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1135 	enum vnic_state prev_state = adapter->state;
1136 	int i, rc;
1137 
1138 	adapter->state = VNIC_OPENING;
1139 	replenish_pools(adapter);
1140 	ibmvnic_napi_enable(adapter);
1141 
1142 	/* We're ready to receive frames, enable the sub-crq interrupts and
1143 	 * set the logical link state to up
1144 	 */
1145 	for (i = 0; i < adapter->req_rx_queues; i++) {
1146 		netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1147 		if (prev_state == VNIC_CLOSED)
1148 			enable_irq(adapter->rx_scrq[i]->irq);
1149 		enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1150 	}
1151 
1152 	for (i = 0; i < adapter->req_tx_queues; i++) {
1153 		netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1154 		if (prev_state == VNIC_CLOSED)
1155 			enable_irq(adapter->tx_scrq[i]->irq);
1156 		enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1157 	}
1158 
1159 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1160 	if (rc) {
1161 		for (i = 0; i < adapter->req_rx_queues; i++)
1162 			napi_disable(&adapter->napi[i]);
1163 		release_resources(adapter);
1164 		return rc;
1165 	}
1166 
1167 	netif_tx_start_all_queues(netdev);
1168 
1169 	if (prev_state == VNIC_CLOSED) {
1170 		for (i = 0; i < adapter->req_rx_queues; i++)
1171 			napi_schedule(&adapter->napi[i]);
1172 	}
1173 
1174 	adapter->state = VNIC_OPEN;
1175 	return rc;
1176 }
1177 
1178 static int ibmvnic_open(struct net_device *netdev)
1179 {
1180 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1181 	int rc;
1182 
1183 	/* If device failover is pending, just set device state and return.
1184 	 * Device operation will be handled by reset routine.
1185 	 */
1186 	if (adapter->failover_pending) {
1187 		adapter->state = VNIC_OPEN;
1188 		return 0;
1189 	}
1190 
1191 	if (adapter->state != VNIC_CLOSED) {
1192 		rc = ibmvnic_login(netdev);
1193 		if (rc)
1194 			return rc;
1195 
1196 		rc = init_resources(adapter);
1197 		if (rc) {
1198 			netdev_err(netdev, "failed to initialize resources\n");
1199 			release_resources(adapter);
1200 			return rc;
1201 		}
1202 	}
1203 
1204 	rc = __ibmvnic_open(netdev);
1205 
1206 	return rc;
1207 }
1208 
1209 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1210 {
1211 	struct ibmvnic_rx_pool *rx_pool;
1212 	struct ibmvnic_rx_buff *rx_buff;
1213 	u64 rx_entries;
1214 	int rx_scrqs;
1215 	int i, j;
1216 
1217 	if (!adapter->rx_pool)
1218 		return;
1219 
1220 	rx_scrqs = adapter->num_active_rx_pools;
1221 	rx_entries = adapter->req_rx_add_entries_per_subcrq;
1222 
1223 	/* Free any remaining skbs in the rx buffer pools */
1224 	for (i = 0; i < rx_scrqs; i++) {
1225 		rx_pool = &adapter->rx_pool[i];
1226 		if (!rx_pool || !rx_pool->rx_buff)
1227 			continue;
1228 
1229 		netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1230 		for (j = 0; j < rx_entries; j++) {
1231 			rx_buff = &rx_pool->rx_buff[j];
1232 			if (rx_buff && rx_buff->skb) {
1233 				dev_kfree_skb_any(rx_buff->skb);
1234 				rx_buff->skb = NULL;
1235 			}
1236 		}
1237 	}
1238 }
1239 
1240 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1241 			      struct ibmvnic_tx_pool *tx_pool)
1242 {
1243 	struct ibmvnic_tx_buff *tx_buff;
1244 	u64 tx_entries;
1245 	int i;
1246 
1247 	if (!tx_pool || !tx_pool->tx_buff)
1248 		return;
1249 
1250 	tx_entries = tx_pool->num_buffers;
1251 
1252 	for (i = 0; i < tx_entries; i++) {
1253 		tx_buff = &tx_pool->tx_buff[i];
1254 		if (tx_buff && tx_buff->skb) {
1255 			dev_kfree_skb_any(tx_buff->skb);
1256 			tx_buff->skb = NULL;
1257 		}
1258 	}
1259 }
1260 
1261 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1262 {
1263 	int tx_scrqs;
1264 	int i;
1265 
1266 	if (!adapter->tx_pool || !adapter->tso_pool)
1267 		return;
1268 
1269 	tx_scrqs = adapter->num_active_tx_pools;
1270 
1271 	/* Free any remaining skbs in the tx buffer pools */
1272 	for (i = 0; i < tx_scrqs; i++) {
1273 		netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1274 		clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1275 		clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1276 	}
1277 }
1278 
1279 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1280 {
1281 	struct net_device *netdev = adapter->netdev;
1282 	int i;
1283 
1284 	if (adapter->tx_scrq) {
1285 		for (i = 0; i < adapter->req_tx_queues; i++)
1286 			if (adapter->tx_scrq[i]->irq) {
1287 				netdev_dbg(netdev,
1288 					   "Disabling tx_scrq[%d] irq\n", i);
1289 				disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1290 				disable_irq(adapter->tx_scrq[i]->irq);
1291 			}
1292 	}
1293 
1294 	if (adapter->rx_scrq) {
1295 		for (i = 0; i < adapter->req_rx_queues; i++) {
1296 			if (adapter->rx_scrq[i]->irq) {
1297 				netdev_dbg(netdev,
1298 					   "Disabling rx_scrq[%d] irq\n", i);
1299 				disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1300 				disable_irq(adapter->rx_scrq[i]->irq);
1301 			}
1302 		}
1303 	}
1304 }
1305 
1306 static void ibmvnic_cleanup(struct net_device *netdev)
1307 {
1308 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1309 
1310 	/* ensure that transmissions are stopped if called by do_reset */
1311 	if (test_bit(0, &adapter->resetting))
1312 		netif_tx_disable(netdev);
1313 	else
1314 		netif_tx_stop_all_queues(netdev);
1315 
1316 	ibmvnic_napi_disable(adapter);
1317 	ibmvnic_disable_irqs(adapter);
1318 
1319 	clean_rx_pools(adapter);
1320 	clean_tx_pools(adapter);
1321 }
1322 
1323 static int __ibmvnic_close(struct net_device *netdev)
1324 {
1325 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1326 	int rc = 0;
1327 
1328 	adapter->state = VNIC_CLOSING;
1329 	rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1330 	if (rc)
1331 		return rc;
1332 	adapter->state = VNIC_CLOSED;
1333 	return 0;
1334 }
1335 
1336 static int ibmvnic_close(struct net_device *netdev)
1337 {
1338 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1339 	int rc;
1340 
1341 	/* If device failover is pending, just set device state and return.
1342 	 * Device operation will be handled by reset routine.
1343 	 */
1344 	if (adapter->failover_pending) {
1345 		adapter->state = VNIC_CLOSED;
1346 		return 0;
1347 	}
1348 
1349 	rc = __ibmvnic_close(netdev);
1350 	ibmvnic_cleanup(netdev);
1351 
1352 	return rc;
1353 }
1354 
1355 /**
1356  * build_hdr_data - creates L2/L3/L4 header data buffer
1357  * @hdr_field - bitfield determining needed headers
1358  * @skb - socket buffer
1359  * @hdr_len - array of header lengths
1360  * @tot_len - total length of data
1361  *
1362  * Reads hdr_field to determine which headers are needed by firmware.
1363  * Builds a buffer containing these headers.  Saves individual header
1364  * lengths and total buffer length to be used to build descriptors.
1365  */
1366 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1367 			  int *hdr_len, u8 *hdr_data)
1368 {
1369 	int len = 0;
1370 	u8 *hdr;
1371 
1372 	if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1373 		hdr_len[0] = sizeof(struct vlan_ethhdr);
1374 	else
1375 		hdr_len[0] = sizeof(struct ethhdr);
1376 
1377 	if (skb->protocol == htons(ETH_P_IP)) {
1378 		hdr_len[1] = ip_hdr(skb)->ihl * 4;
1379 		if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1380 			hdr_len[2] = tcp_hdrlen(skb);
1381 		else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1382 			hdr_len[2] = sizeof(struct udphdr);
1383 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1384 		hdr_len[1] = sizeof(struct ipv6hdr);
1385 		if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1386 			hdr_len[2] = tcp_hdrlen(skb);
1387 		else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1388 			hdr_len[2] = sizeof(struct udphdr);
1389 	} else if (skb->protocol == htons(ETH_P_ARP)) {
1390 		hdr_len[1] = arp_hdr_len(skb->dev);
1391 		hdr_len[2] = 0;
1392 	}
1393 
1394 	memset(hdr_data, 0, 120);
1395 	if ((hdr_field >> 6) & 1) {
1396 		hdr = skb_mac_header(skb);
1397 		memcpy(hdr_data, hdr, hdr_len[0]);
1398 		len += hdr_len[0];
1399 	}
1400 
1401 	if ((hdr_field >> 5) & 1) {
1402 		hdr = skb_network_header(skb);
1403 		memcpy(hdr_data + len, hdr, hdr_len[1]);
1404 		len += hdr_len[1];
1405 	}
1406 
1407 	if ((hdr_field >> 4) & 1) {
1408 		hdr = skb_transport_header(skb);
1409 		memcpy(hdr_data + len, hdr, hdr_len[2]);
1410 		len += hdr_len[2];
1411 	}
1412 	return len;
1413 }
1414 
1415 /**
1416  * create_hdr_descs - create header and header extension descriptors
1417  * @hdr_field - bitfield determining needed headers
1418  * @data - buffer containing header data
1419  * @len - length of data buffer
1420  * @hdr_len - array of individual header lengths
1421  * @scrq_arr - descriptor array
1422  *
1423  * Creates header and, if needed, header extension descriptors and
1424  * places them in a descriptor array, scrq_arr
1425  */
1426 
1427 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1428 			    union sub_crq *scrq_arr)
1429 {
1430 	union sub_crq hdr_desc;
1431 	int tmp_len = len;
1432 	int num_descs = 0;
1433 	u8 *data, *cur;
1434 	int tmp;
1435 
1436 	while (tmp_len > 0) {
1437 		cur = hdr_data + len - tmp_len;
1438 
1439 		memset(&hdr_desc, 0, sizeof(hdr_desc));
1440 		if (cur != hdr_data) {
1441 			data = hdr_desc.hdr_ext.data;
1442 			tmp = tmp_len > 29 ? 29 : tmp_len;
1443 			hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1444 			hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1445 			hdr_desc.hdr_ext.len = tmp;
1446 		} else {
1447 			data = hdr_desc.hdr.data;
1448 			tmp = tmp_len > 24 ? 24 : tmp_len;
1449 			hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1450 			hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1451 			hdr_desc.hdr.len = tmp;
1452 			hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1453 			hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1454 			hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1455 			hdr_desc.hdr.flag = hdr_field << 1;
1456 		}
1457 		memcpy(data, cur, tmp);
1458 		tmp_len -= tmp;
1459 		*scrq_arr = hdr_desc;
1460 		scrq_arr++;
1461 		num_descs++;
1462 	}
1463 
1464 	return num_descs;
1465 }
1466 
1467 /**
1468  * build_hdr_descs_arr - build a header descriptor array
1469  * @skb - socket buffer
1470  * @num_entries - number of descriptors to be sent
1471  * @subcrq - first TX descriptor
1472  * @hdr_field - bit field determining which headers will be sent
1473  *
1474  * This function will build a TX descriptor array with applicable
1475  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1476  */
1477 
1478 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1479 				int *num_entries, u8 hdr_field)
1480 {
1481 	int hdr_len[3] = {0, 0, 0};
1482 	int tot_len;
1483 	u8 *hdr_data = txbuff->hdr_data;
1484 
1485 	tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1486 				 txbuff->hdr_data);
1487 	*num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1488 			 txbuff->indir_arr + 1);
1489 }
1490 
1491 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1492 				    struct net_device *netdev)
1493 {
1494 	/* For some backing devices, mishandling of small packets
1495 	 * can result in a loss of connection or TX stall. Device
1496 	 * architects recommend that no packet should be smaller
1497 	 * than the minimum MTU value provided to the driver, so
1498 	 * pad any packets to that length
1499 	 */
1500 	if (skb->len < netdev->min_mtu)
1501 		return skb_put_padto(skb, netdev->min_mtu);
1502 
1503 	return 0;
1504 }
1505 
1506 static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1507 {
1508 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1509 	int queue_num = skb_get_queue_mapping(skb);
1510 	u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1511 	struct device *dev = &adapter->vdev->dev;
1512 	struct ibmvnic_tx_buff *tx_buff = NULL;
1513 	struct ibmvnic_sub_crq_queue *tx_scrq;
1514 	struct ibmvnic_tx_pool *tx_pool;
1515 	unsigned int tx_send_failed = 0;
1516 	unsigned int tx_map_failed = 0;
1517 	unsigned int tx_dropped = 0;
1518 	unsigned int tx_packets = 0;
1519 	unsigned int tx_bytes = 0;
1520 	dma_addr_t data_dma_addr;
1521 	struct netdev_queue *txq;
1522 	unsigned long lpar_rc;
1523 	union sub_crq tx_crq;
1524 	unsigned int offset;
1525 	int num_entries = 1;
1526 	unsigned char *dst;
1527 	u64 *handle_array;
1528 	int index = 0;
1529 	u8 proto = 0;
1530 	netdev_tx_t ret = NETDEV_TX_OK;
1531 
1532 	if (test_bit(0, &adapter->resetting)) {
1533 		if (!netif_subqueue_stopped(netdev, skb))
1534 			netif_stop_subqueue(netdev, queue_num);
1535 		dev_kfree_skb_any(skb);
1536 
1537 		tx_send_failed++;
1538 		tx_dropped++;
1539 		ret = NETDEV_TX_OK;
1540 		goto out;
1541 	}
1542 
1543 	if (ibmvnic_xmit_workarounds(skb, netdev)) {
1544 		tx_dropped++;
1545 		tx_send_failed++;
1546 		ret = NETDEV_TX_OK;
1547 		goto out;
1548 	}
1549 	if (skb_is_gso(skb))
1550 		tx_pool = &adapter->tso_pool[queue_num];
1551 	else
1552 		tx_pool = &adapter->tx_pool[queue_num];
1553 
1554 	tx_scrq = adapter->tx_scrq[queue_num];
1555 	txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1556 	handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1557 		be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1558 
1559 	index = tx_pool->free_map[tx_pool->consumer_index];
1560 
1561 	if (index == IBMVNIC_INVALID_MAP) {
1562 		dev_kfree_skb_any(skb);
1563 		tx_send_failed++;
1564 		tx_dropped++;
1565 		ret = NETDEV_TX_OK;
1566 		goto out;
1567 	}
1568 
1569 	tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1570 
1571 	offset = index * tx_pool->buf_size;
1572 	dst = tx_pool->long_term_buff.buff + offset;
1573 	memset(dst, 0, tx_pool->buf_size);
1574 	data_dma_addr = tx_pool->long_term_buff.addr + offset;
1575 
1576 	if (skb_shinfo(skb)->nr_frags) {
1577 		int cur, i;
1578 
1579 		/* Copy the head */
1580 		skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1581 		cur = skb_headlen(skb);
1582 
1583 		/* Copy the frags */
1584 		for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1585 			const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1586 
1587 			memcpy(dst + cur,
1588 			       page_address(skb_frag_page(frag)) +
1589 			       skb_frag_off(frag), skb_frag_size(frag));
1590 			cur += skb_frag_size(frag);
1591 		}
1592 	} else {
1593 		skb_copy_from_linear_data(skb, dst, skb->len);
1594 	}
1595 
1596 	tx_pool->consumer_index =
1597 	    (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1598 
1599 	tx_buff = &tx_pool->tx_buff[index];
1600 	tx_buff->skb = skb;
1601 	tx_buff->data_dma[0] = data_dma_addr;
1602 	tx_buff->data_len[0] = skb->len;
1603 	tx_buff->index = index;
1604 	tx_buff->pool_index = queue_num;
1605 	tx_buff->last_frag = true;
1606 
1607 	memset(&tx_crq, 0, sizeof(tx_crq));
1608 	tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1609 	tx_crq.v1.type = IBMVNIC_TX_DESC;
1610 	tx_crq.v1.n_crq_elem = 1;
1611 	tx_crq.v1.n_sge = 1;
1612 	tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1613 
1614 	if (skb_is_gso(skb))
1615 		tx_crq.v1.correlator =
1616 			cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1617 	else
1618 		tx_crq.v1.correlator = cpu_to_be32(index);
1619 	tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1620 	tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1621 	tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1622 
1623 	if (adapter->vlan_header_insertion && skb_vlan_tag_present(skb)) {
1624 		tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1625 		tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1626 	}
1627 
1628 	if (skb->protocol == htons(ETH_P_IP)) {
1629 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1630 		proto = ip_hdr(skb)->protocol;
1631 	} else if (skb->protocol == htons(ETH_P_IPV6)) {
1632 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1633 		proto = ipv6_hdr(skb)->nexthdr;
1634 	}
1635 
1636 	if (proto == IPPROTO_TCP)
1637 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1638 	else if (proto == IPPROTO_UDP)
1639 		tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1640 
1641 	if (skb->ip_summed == CHECKSUM_PARTIAL) {
1642 		tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1643 		hdrs += 2;
1644 	}
1645 	if (skb_is_gso(skb)) {
1646 		tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1647 		tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1648 		hdrs += 2;
1649 	}
1650 	/* determine if l2/3/4 headers are sent to firmware */
1651 	if ((*hdrs >> 7) & 1) {
1652 		build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1653 		tx_crq.v1.n_crq_elem = num_entries;
1654 		tx_buff->num_entries = num_entries;
1655 		tx_buff->indir_arr[0] = tx_crq;
1656 		tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1657 						    sizeof(tx_buff->indir_arr),
1658 						    DMA_TO_DEVICE);
1659 		if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1660 			dev_kfree_skb_any(skb);
1661 			tx_buff->skb = NULL;
1662 			if (!firmware_has_feature(FW_FEATURE_CMO))
1663 				dev_err(dev, "tx: unable to map descriptor array\n");
1664 			tx_map_failed++;
1665 			tx_dropped++;
1666 			ret = NETDEV_TX_OK;
1667 			goto tx_err_out;
1668 		}
1669 		lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1670 					       (u64)tx_buff->indir_dma,
1671 					       (u64)num_entries);
1672 		dma_unmap_single(dev, tx_buff->indir_dma,
1673 				 sizeof(tx_buff->indir_arr), DMA_TO_DEVICE);
1674 	} else {
1675 		tx_buff->num_entries = num_entries;
1676 		lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1677 				      &tx_crq);
1678 	}
1679 	if (lpar_rc != H_SUCCESS) {
1680 		if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1681 			dev_err_ratelimited(dev, "tx: send failed\n");
1682 		dev_kfree_skb_any(skb);
1683 		tx_buff->skb = NULL;
1684 
1685 		if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1686 			/* Disable TX and report carrier off if queue is closed
1687 			 * or pending failover.
1688 			 * Firmware guarantees that a signal will be sent to the
1689 			 * driver, triggering a reset or some other action.
1690 			 */
1691 			netif_tx_stop_all_queues(netdev);
1692 			netif_carrier_off(netdev);
1693 		}
1694 
1695 		tx_send_failed++;
1696 		tx_dropped++;
1697 		ret = NETDEV_TX_OK;
1698 		goto tx_err_out;
1699 	}
1700 
1701 	if (atomic_add_return(num_entries, &tx_scrq->used)
1702 					>= adapter->req_tx_entries_per_subcrq) {
1703 		netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1704 		netif_stop_subqueue(netdev, queue_num);
1705 	}
1706 
1707 	tx_packets++;
1708 	tx_bytes += skb->len;
1709 	txq->trans_start = jiffies;
1710 	ret = NETDEV_TX_OK;
1711 	goto out;
1712 
1713 tx_err_out:
1714 	/* roll back consumer index and map array*/
1715 	if (tx_pool->consumer_index == 0)
1716 		tx_pool->consumer_index =
1717 			tx_pool->num_buffers - 1;
1718 	else
1719 		tx_pool->consumer_index--;
1720 	tx_pool->free_map[tx_pool->consumer_index] = index;
1721 out:
1722 	netdev->stats.tx_dropped += tx_dropped;
1723 	netdev->stats.tx_bytes += tx_bytes;
1724 	netdev->stats.tx_packets += tx_packets;
1725 	adapter->tx_send_failed += tx_send_failed;
1726 	adapter->tx_map_failed += tx_map_failed;
1727 	adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1728 	adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1729 	adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1730 
1731 	return ret;
1732 }
1733 
1734 static void ibmvnic_set_multi(struct net_device *netdev)
1735 {
1736 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1737 	struct netdev_hw_addr *ha;
1738 	union ibmvnic_crq crq;
1739 
1740 	memset(&crq, 0, sizeof(crq));
1741 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
1742 	crq.request_capability.cmd = REQUEST_CAPABILITY;
1743 
1744 	if (netdev->flags & IFF_PROMISC) {
1745 		if (!adapter->promisc_supported)
1746 			return;
1747 	} else {
1748 		if (netdev->flags & IFF_ALLMULTI) {
1749 			/* Accept all multicast */
1750 			memset(&crq, 0, sizeof(crq));
1751 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1752 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1753 			crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1754 			ibmvnic_send_crq(adapter, &crq);
1755 		} else if (netdev_mc_empty(netdev)) {
1756 			/* Reject all multicast */
1757 			memset(&crq, 0, sizeof(crq));
1758 			crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1759 			crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1760 			crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1761 			ibmvnic_send_crq(adapter, &crq);
1762 		} else {
1763 			/* Accept one or more multicast(s) */
1764 			netdev_for_each_mc_addr(ha, netdev) {
1765 				memset(&crq, 0, sizeof(crq));
1766 				crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1767 				crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1768 				crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1769 				ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1770 						ha->addr);
1771 				ibmvnic_send_crq(adapter, &crq);
1772 			}
1773 		}
1774 	}
1775 }
1776 
1777 static int __ibmvnic_set_mac(struct net_device *netdev, u8 *dev_addr)
1778 {
1779 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1780 	union ibmvnic_crq crq;
1781 	int rc;
1782 
1783 	if (!is_valid_ether_addr(dev_addr)) {
1784 		rc = -EADDRNOTAVAIL;
1785 		goto err;
1786 	}
1787 
1788 	memset(&crq, 0, sizeof(crq));
1789 	crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1790 	crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1791 	ether_addr_copy(&crq.change_mac_addr.mac_addr[0], dev_addr);
1792 
1793 	mutex_lock(&adapter->fw_lock);
1794 	adapter->fw_done_rc = 0;
1795 	reinit_completion(&adapter->fw_done);
1796 
1797 	rc = ibmvnic_send_crq(adapter, &crq);
1798 	if (rc) {
1799 		rc = -EIO;
1800 		mutex_unlock(&adapter->fw_lock);
1801 		goto err;
1802 	}
1803 
1804 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
1805 	/* netdev->dev_addr is changed in handle_change_mac_rsp function */
1806 	if (rc || adapter->fw_done_rc) {
1807 		rc = -EIO;
1808 		mutex_unlock(&adapter->fw_lock);
1809 		goto err;
1810 	}
1811 	mutex_unlock(&adapter->fw_lock);
1812 	return 0;
1813 err:
1814 	ether_addr_copy(adapter->mac_addr, netdev->dev_addr);
1815 	return rc;
1816 }
1817 
1818 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1819 {
1820 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1821 	struct sockaddr *addr = p;
1822 	int rc;
1823 
1824 	rc = 0;
1825 	ether_addr_copy(adapter->mac_addr, addr->sa_data);
1826 	if (adapter->state != VNIC_PROBED)
1827 		rc = __ibmvnic_set_mac(netdev, addr->sa_data);
1828 
1829 	return rc;
1830 }
1831 
1832 /**
1833  * do_change_param_reset returns zero if we are able to keep processing reset
1834  * events, or non-zero if we hit a fatal error and must halt.
1835  */
1836 static int do_change_param_reset(struct ibmvnic_adapter *adapter,
1837 				 struct ibmvnic_rwi *rwi,
1838 				 u32 reset_state)
1839 {
1840 	struct net_device *netdev = adapter->netdev;
1841 	int i, rc;
1842 
1843 	netdev_dbg(adapter->netdev, "Change param resetting driver (%d)\n",
1844 		   rwi->reset_reason);
1845 
1846 	netif_carrier_off(netdev);
1847 	adapter->reset_reason = rwi->reset_reason;
1848 
1849 	ibmvnic_cleanup(netdev);
1850 
1851 	if (reset_state == VNIC_OPEN) {
1852 		rc = __ibmvnic_close(netdev);
1853 		if (rc)
1854 			return rc;
1855 	}
1856 
1857 	release_resources(adapter);
1858 	release_sub_crqs(adapter, 1);
1859 	release_crq_queue(adapter);
1860 
1861 	adapter->state = VNIC_PROBED;
1862 
1863 	rc = init_crq_queue(adapter);
1864 
1865 	if (rc) {
1866 		netdev_err(adapter->netdev,
1867 			   "Couldn't initialize crq. rc=%d\n", rc);
1868 		return rc;
1869 	}
1870 
1871 	rc = ibmvnic_reset_init(adapter);
1872 	if (rc)
1873 		return IBMVNIC_INIT_FAILED;
1874 
1875 	/* If the adapter was in PROBE state prior to the reset,
1876 	 * exit here.
1877 	 */
1878 	if (reset_state == VNIC_PROBED)
1879 		return 0;
1880 
1881 	rc = ibmvnic_login(netdev);
1882 	if (rc) {
1883 		adapter->state = reset_state;
1884 		return rc;
1885 	}
1886 
1887 	rc = init_resources(adapter);
1888 	if (rc)
1889 		return rc;
1890 
1891 	ibmvnic_disable_irqs(adapter);
1892 
1893 	adapter->state = VNIC_CLOSED;
1894 
1895 	if (reset_state == VNIC_CLOSED)
1896 		return 0;
1897 
1898 	rc = __ibmvnic_open(netdev);
1899 	if (rc)
1900 		return IBMVNIC_OPEN_FAILED;
1901 
1902 	/* refresh device's multicast list */
1903 	ibmvnic_set_multi(netdev);
1904 
1905 	/* kick napi */
1906 	for (i = 0; i < adapter->req_rx_queues; i++)
1907 		napi_schedule(&adapter->napi[i]);
1908 
1909 	return 0;
1910 }
1911 
1912 /**
1913  * do_reset returns zero if we are able to keep processing reset events, or
1914  * non-zero if we hit a fatal error and must halt.
1915  */
1916 static int do_reset(struct ibmvnic_adapter *adapter,
1917 		    struct ibmvnic_rwi *rwi, u32 reset_state)
1918 {
1919 	u64 old_num_rx_queues, old_num_tx_queues;
1920 	u64 old_num_rx_slots, old_num_tx_slots;
1921 	struct net_device *netdev = adapter->netdev;
1922 	int i, rc;
1923 
1924 	netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1925 		   rwi->reset_reason);
1926 
1927 	rtnl_lock();
1928 
1929 	netif_carrier_off(netdev);
1930 	adapter->reset_reason = rwi->reset_reason;
1931 
1932 	old_num_rx_queues = adapter->req_rx_queues;
1933 	old_num_tx_queues = adapter->req_tx_queues;
1934 	old_num_rx_slots = adapter->req_rx_add_entries_per_subcrq;
1935 	old_num_tx_slots = adapter->req_tx_entries_per_subcrq;
1936 
1937 	ibmvnic_cleanup(netdev);
1938 
1939 	if (reset_state == VNIC_OPEN &&
1940 	    adapter->reset_reason != VNIC_RESET_MOBILITY &&
1941 	    adapter->reset_reason != VNIC_RESET_FAILOVER) {
1942 		adapter->state = VNIC_CLOSING;
1943 
1944 		/* Release the RTNL lock before link state change and
1945 		 * re-acquire after the link state change to allow
1946 		 * linkwatch_event to grab the RTNL lock and run during
1947 		 * a reset.
1948 		 */
1949 		rtnl_unlock();
1950 		rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1951 		rtnl_lock();
1952 		if (rc)
1953 			goto out;
1954 
1955 		if (adapter->state != VNIC_CLOSING) {
1956 			rc = -1;
1957 			goto out;
1958 		}
1959 
1960 		adapter->state = VNIC_CLOSED;
1961 	}
1962 
1963 	if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1964 		/* remove the closed state so when we call open it appears
1965 		 * we are coming from the probed state.
1966 		 */
1967 		adapter->state = VNIC_PROBED;
1968 
1969 		if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1970 			rc = ibmvnic_reenable_crq_queue(adapter);
1971 			release_sub_crqs(adapter, 1);
1972 		} else {
1973 			rc = ibmvnic_reset_crq(adapter);
1974 			if (!rc)
1975 				rc = vio_enable_interrupts(adapter->vdev);
1976 		}
1977 
1978 		if (rc) {
1979 			netdev_err(adapter->netdev,
1980 				   "Couldn't initialize crq. rc=%d\n", rc);
1981 			goto out;
1982 		}
1983 
1984 		rc = ibmvnic_reset_init(adapter);
1985 		if (rc) {
1986 			rc = IBMVNIC_INIT_FAILED;
1987 			goto out;
1988 		}
1989 
1990 		/* If the adapter was in PROBE state prior to the reset,
1991 		 * exit here.
1992 		 */
1993 		if (reset_state == VNIC_PROBED) {
1994 			rc = 0;
1995 			goto out;
1996 		}
1997 
1998 		rc = ibmvnic_login(netdev);
1999 		if (rc) {
2000 			adapter->state = reset_state;
2001 			goto out;
2002 		}
2003 
2004 		if (adapter->req_rx_queues != old_num_rx_queues ||
2005 		    adapter->req_tx_queues != old_num_tx_queues ||
2006 		    adapter->req_rx_add_entries_per_subcrq !=
2007 		    old_num_rx_slots ||
2008 		    adapter->req_tx_entries_per_subcrq !=
2009 		    old_num_tx_slots) {
2010 			release_rx_pools(adapter);
2011 			release_tx_pools(adapter);
2012 			release_napi(adapter);
2013 			release_vpd_data(adapter);
2014 
2015 			rc = init_resources(adapter);
2016 			if (rc)
2017 				goto out;
2018 
2019 		} else {
2020 			rc = reset_tx_pools(adapter);
2021 			if (rc)
2022 				goto out;
2023 
2024 			rc = reset_rx_pools(adapter);
2025 			if (rc)
2026 				goto out;
2027 		}
2028 		ibmvnic_disable_irqs(adapter);
2029 	}
2030 	adapter->state = VNIC_CLOSED;
2031 
2032 	if (reset_state == VNIC_CLOSED) {
2033 		rc = 0;
2034 		goto out;
2035 	}
2036 
2037 	rc = __ibmvnic_open(netdev);
2038 	if (rc) {
2039 		rc = IBMVNIC_OPEN_FAILED;
2040 		goto out;
2041 	}
2042 
2043 	/* refresh device's multicast list */
2044 	ibmvnic_set_multi(netdev);
2045 
2046 	/* kick napi */
2047 	for (i = 0; i < adapter->req_rx_queues; i++)
2048 		napi_schedule(&adapter->napi[i]);
2049 
2050 	if (adapter->reset_reason != VNIC_RESET_FAILOVER)
2051 		call_netdevice_notifiers(NETDEV_NOTIFY_PEERS, netdev);
2052 
2053 	rc = 0;
2054 
2055 out:
2056 	rtnl_unlock();
2057 
2058 	return rc;
2059 }
2060 
2061 static int do_hard_reset(struct ibmvnic_adapter *adapter,
2062 			 struct ibmvnic_rwi *rwi, u32 reset_state)
2063 {
2064 	struct net_device *netdev = adapter->netdev;
2065 	int rc;
2066 
2067 	netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
2068 		   rwi->reset_reason);
2069 
2070 	netif_carrier_off(netdev);
2071 	adapter->reset_reason = rwi->reset_reason;
2072 
2073 	ibmvnic_cleanup(netdev);
2074 	release_resources(adapter);
2075 	release_sub_crqs(adapter, 0);
2076 	release_crq_queue(adapter);
2077 
2078 	/* remove the closed state so when we call open it appears
2079 	 * we are coming from the probed state.
2080 	 */
2081 	adapter->state = VNIC_PROBED;
2082 
2083 	reinit_completion(&adapter->init_done);
2084 	rc = init_crq_queue(adapter);
2085 	if (rc) {
2086 		netdev_err(adapter->netdev,
2087 			   "Couldn't initialize crq. rc=%d\n", rc);
2088 		return rc;
2089 	}
2090 
2091 	rc = ibmvnic_init(adapter);
2092 	if (rc)
2093 		return rc;
2094 
2095 	/* If the adapter was in PROBE state prior to the reset,
2096 	 * exit here.
2097 	 */
2098 	if (reset_state == VNIC_PROBED)
2099 		return 0;
2100 
2101 	rc = ibmvnic_login(netdev);
2102 	if (rc) {
2103 		adapter->state = VNIC_PROBED;
2104 		return 0;
2105 	}
2106 
2107 	rc = init_resources(adapter);
2108 	if (rc)
2109 		return rc;
2110 
2111 	ibmvnic_disable_irqs(adapter);
2112 	adapter->state = VNIC_CLOSED;
2113 
2114 	if (reset_state == VNIC_CLOSED)
2115 		return 0;
2116 
2117 	rc = __ibmvnic_open(netdev);
2118 	if (rc)
2119 		return IBMVNIC_OPEN_FAILED;
2120 
2121 	return 0;
2122 }
2123 
2124 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
2125 {
2126 	struct ibmvnic_rwi *rwi;
2127 	unsigned long flags;
2128 
2129 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2130 
2131 	if (!list_empty(&adapter->rwi_list)) {
2132 		rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
2133 				       list);
2134 		list_del(&rwi->list);
2135 	} else {
2136 		rwi = NULL;
2137 	}
2138 
2139 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2140 	return rwi;
2141 }
2142 
2143 static void free_all_rwi(struct ibmvnic_adapter *adapter)
2144 {
2145 	struct ibmvnic_rwi *rwi;
2146 
2147 	rwi = get_next_rwi(adapter);
2148 	while (rwi) {
2149 		kfree(rwi);
2150 		rwi = get_next_rwi(adapter);
2151 	}
2152 }
2153 
2154 static void __ibmvnic_reset(struct work_struct *work)
2155 {
2156 	struct ibmvnic_rwi *rwi;
2157 	struct ibmvnic_adapter *adapter;
2158 	bool saved_state = false;
2159 	unsigned long flags;
2160 	u32 reset_state;
2161 	int rc = 0;
2162 
2163 	adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
2164 
2165 	if (test_and_set_bit_lock(0, &adapter->resetting)) {
2166 		schedule_delayed_work(&adapter->ibmvnic_delayed_reset,
2167 				      IBMVNIC_RESET_DELAY);
2168 		return;
2169 	}
2170 
2171 	rwi = get_next_rwi(adapter);
2172 	while (rwi) {
2173 		spin_lock_irqsave(&adapter->state_lock, flags);
2174 
2175 		if (adapter->state == VNIC_REMOVING ||
2176 		    adapter->state == VNIC_REMOVED) {
2177 			spin_unlock_irqrestore(&adapter->state_lock, flags);
2178 			kfree(rwi);
2179 			rc = EBUSY;
2180 			break;
2181 		}
2182 
2183 		if (!saved_state) {
2184 			reset_state = adapter->state;
2185 			adapter->state = VNIC_RESETTING;
2186 			saved_state = true;
2187 		}
2188 		spin_unlock_irqrestore(&adapter->state_lock, flags);
2189 
2190 		if (rwi->reset_reason == VNIC_RESET_CHANGE_PARAM) {
2191 			/* CHANGE_PARAM requestor holds rtnl_lock */
2192 			rc = do_change_param_reset(adapter, rwi, reset_state);
2193 		} else if (adapter->force_reset_recovery) {
2194 			/* Transport event occurred during previous reset */
2195 			if (adapter->wait_for_reset) {
2196 				/* Previous was CHANGE_PARAM; caller locked */
2197 				adapter->force_reset_recovery = false;
2198 				rc = do_hard_reset(adapter, rwi, reset_state);
2199 			} else {
2200 				rtnl_lock();
2201 				adapter->force_reset_recovery = false;
2202 				rc = do_hard_reset(adapter, rwi, reset_state);
2203 				rtnl_unlock();
2204 			}
2205 		} else if (!(rwi->reset_reason == VNIC_RESET_FATAL &&
2206 				adapter->from_passive_init)) {
2207 			rc = do_reset(adapter, rwi, reset_state);
2208 		}
2209 		kfree(rwi);
2210 		if (rc == IBMVNIC_OPEN_FAILED) {
2211 			if (list_empty(&adapter->rwi_list))
2212 				adapter->state = VNIC_CLOSED;
2213 			else
2214 				adapter->state = reset_state;
2215 			rc = 0;
2216 		} else if (rc && rc != IBMVNIC_INIT_FAILED &&
2217 		    !adapter->force_reset_recovery)
2218 			break;
2219 
2220 		rwi = get_next_rwi(adapter);
2221 
2222 		if (rwi && (rwi->reset_reason == VNIC_RESET_FAILOVER ||
2223 			    rwi->reset_reason == VNIC_RESET_MOBILITY))
2224 			adapter->force_reset_recovery = true;
2225 	}
2226 
2227 	if (adapter->wait_for_reset) {
2228 		adapter->reset_done_rc = rc;
2229 		complete(&adapter->reset_done);
2230 	}
2231 
2232 	if (rc) {
2233 		netdev_dbg(adapter->netdev, "Reset failed\n");
2234 		free_all_rwi(adapter);
2235 	}
2236 
2237 	clear_bit_unlock(0, &adapter->resetting);
2238 }
2239 
2240 static void __ibmvnic_delayed_reset(struct work_struct *work)
2241 {
2242 	struct ibmvnic_adapter *adapter;
2243 
2244 	adapter = container_of(work, struct ibmvnic_adapter,
2245 			       ibmvnic_delayed_reset.work);
2246 	__ibmvnic_reset(&adapter->ibmvnic_reset);
2247 }
2248 
2249 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2250 			 enum ibmvnic_reset_reason reason)
2251 {
2252 	struct list_head *entry, *tmp_entry;
2253 	struct ibmvnic_rwi *rwi, *tmp;
2254 	struct net_device *netdev = adapter->netdev;
2255 	unsigned long flags;
2256 	int ret;
2257 
2258 	if (adapter->state == VNIC_REMOVING ||
2259 	    adapter->state == VNIC_REMOVED ||
2260 	    adapter->failover_pending) {
2261 		ret = EBUSY;
2262 		netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2263 		goto err;
2264 	}
2265 
2266 	if (adapter->state == VNIC_PROBING) {
2267 		netdev_warn(netdev, "Adapter reset during probe\n");
2268 		ret = adapter->init_done_rc = EAGAIN;
2269 		goto err;
2270 	}
2271 
2272 	spin_lock_irqsave(&adapter->rwi_lock, flags);
2273 
2274 	list_for_each(entry, &adapter->rwi_list) {
2275 		tmp = list_entry(entry, struct ibmvnic_rwi, list);
2276 		if (tmp->reset_reason == reason) {
2277 			netdev_dbg(netdev, "Skipping matching reset\n");
2278 			spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2279 			ret = EBUSY;
2280 			goto err;
2281 		}
2282 	}
2283 
2284 	rwi = kzalloc(sizeof(*rwi), GFP_ATOMIC);
2285 	if (!rwi) {
2286 		spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2287 		ibmvnic_close(netdev);
2288 		ret = ENOMEM;
2289 		goto err;
2290 	}
2291 	/* if we just received a transport event,
2292 	 * flush reset queue and process this reset
2293 	 */
2294 	if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2295 		list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2296 			list_del(entry);
2297 	}
2298 	rwi->reset_reason = reason;
2299 	list_add_tail(&rwi->list, &adapter->rwi_list);
2300 	spin_unlock_irqrestore(&adapter->rwi_lock, flags);
2301 	netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2302 	schedule_work(&adapter->ibmvnic_reset);
2303 
2304 	return 0;
2305 err:
2306 	return -ret;
2307 }
2308 
2309 static void ibmvnic_tx_timeout(struct net_device *dev, unsigned int txqueue)
2310 {
2311 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2312 
2313 	ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2314 }
2315 
2316 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2317 				  struct ibmvnic_rx_buff *rx_buff)
2318 {
2319 	struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2320 
2321 	rx_buff->skb = NULL;
2322 
2323 	pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2324 	pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2325 
2326 	atomic_dec(&pool->available);
2327 }
2328 
2329 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2330 {
2331 	struct net_device *netdev = napi->dev;
2332 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2333 	int scrq_num = (int)(napi - adapter->napi);
2334 	int frames_processed = 0;
2335 
2336 restart_poll:
2337 	while (frames_processed < budget) {
2338 		struct sk_buff *skb;
2339 		struct ibmvnic_rx_buff *rx_buff;
2340 		union sub_crq *next;
2341 		u32 length;
2342 		u16 offset;
2343 		u8 flags = 0;
2344 
2345 		if (unlikely(test_bit(0, &adapter->resetting) &&
2346 			     adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2347 			enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2348 			napi_complete_done(napi, frames_processed);
2349 			return frames_processed;
2350 		}
2351 
2352 		if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2353 			break;
2354 		next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2355 		rx_buff =
2356 		    (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2357 							  rx_comp.correlator);
2358 		/* do error checking */
2359 		if (next->rx_comp.rc) {
2360 			netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2361 				   be16_to_cpu(next->rx_comp.rc));
2362 			/* free the entry */
2363 			next->rx_comp.first = 0;
2364 			dev_kfree_skb_any(rx_buff->skb);
2365 			remove_buff_from_pool(adapter, rx_buff);
2366 			continue;
2367 		} else if (!rx_buff->skb) {
2368 			/* free the entry */
2369 			next->rx_comp.first = 0;
2370 			remove_buff_from_pool(adapter, rx_buff);
2371 			continue;
2372 		}
2373 
2374 		length = be32_to_cpu(next->rx_comp.len);
2375 		offset = be16_to_cpu(next->rx_comp.off_frame_data);
2376 		flags = next->rx_comp.flags;
2377 		skb = rx_buff->skb;
2378 		skb_copy_to_linear_data(skb, rx_buff->data + offset,
2379 					length);
2380 
2381 		/* VLAN Header has been stripped by the system firmware and
2382 		 * needs to be inserted by the driver
2383 		 */
2384 		if (adapter->rx_vlan_header_insertion &&
2385 		    (flags & IBMVNIC_VLAN_STRIPPED))
2386 			__vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2387 					       ntohs(next->rx_comp.vlan_tci));
2388 
2389 		/* free the entry */
2390 		next->rx_comp.first = 0;
2391 		remove_buff_from_pool(adapter, rx_buff);
2392 
2393 		skb_put(skb, length);
2394 		skb->protocol = eth_type_trans(skb, netdev);
2395 		skb_record_rx_queue(skb, scrq_num);
2396 
2397 		if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2398 		    flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2399 			skb->ip_summed = CHECKSUM_UNNECESSARY;
2400 		}
2401 
2402 		length = skb->len;
2403 		napi_gro_receive(napi, skb); /* send it up */
2404 		netdev->stats.rx_packets++;
2405 		netdev->stats.rx_bytes += length;
2406 		adapter->rx_stats_buffers[scrq_num].packets++;
2407 		adapter->rx_stats_buffers[scrq_num].bytes += length;
2408 		frames_processed++;
2409 	}
2410 
2411 	if (adapter->state != VNIC_CLOSING)
2412 		replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2413 
2414 	if (frames_processed < budget) {
2415 		enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2416 		napi_complete_done(napi, frames_processed);
2417 		if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2418 		    napi_reschedule(napi)) {
2419 			disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2420 			goto restart_poll;
2421 		}
2422 	}
2423 	return frames_processed;
2424 }
2425 
2426 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2427 {
2428 	int rc, ret;
2429 
2430 	adapter->fallback.mtu = adapter->req_mtu;
2431 	adapter->fallback.rx_queues = adapter->req_rx_queues;
2432 	adapter->fallback.tx_queues = adapter->req_tx_queues;
2433 	adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2434 	adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2435 
2436 	reinit_completion(&adapter->reset_done);
2437 	adapter->wait_for_reset = true;
2438 	rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2439 
2440 	if (rc) {
2441 		ret = rc;
2442 		goto out;
2443 	}
2444 	rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done, 60000);
2445 	if (rc) {
2446 		ret = -ENODEV;
2447 		goto out;
2448 	}
2449 
2450 	ret = 0;
2451 	if (adapter->reset_done_rc) {
2452 		ret = -EIO;
2453 		adapter->desired.mtu = adapter->fallback.mtu;
2454 		adapter->desired.rx_queues = adapter->fallback.rx_queues;
2455 		adapter->desired.tx_queues = adapter->fallback.tx_queues;
2456 		adapter->desired.rx_entries = adapter->fallback.rx_entries;
2457 		adapter->desired.tx_entries = adapter->fallback.tx_entries;
2458 
2459 		reinit_completion(&adapter->reset_done);
2460 		adapter->wait_for_reset = true;
2461 		rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2462 		if (rc) {
2463 			ret = rc;
2464 			goto out;
2465 		}
2466 		rc = ibmvnic_wait_for_completion(adapter, &adapter->reset_done,
2467 						 60000);
2468 		if (rc) {
2469 			ret = -ENODEV;
2470 			goto out;
2471 		}
2472 	}
2473 out:
2474 	adapter->wait_for_reset = false;
2475 
2476 	return ret;
2477 }
2478 
2479 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2480 {
2481 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2482 
2483 	adapter->desired.mtu = new_mtu + ETH_HLEN;
2484 
2485 	return wait_for_reset(adapter);
2486 }
2487 
2488 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2489 						struct net_device *dev,
2490 						netdev_features_t features)
2491 {
2492 	/* Some backing hardware adapters can not
2493 	 * handle packets with a MSS less than 224
2494 	 * or with only one segment.
2495 	 */
2496 	if (skb_is_gso(skb)) {
2497 		if (skb_shinfo(skb)->gso_size < 224 ||
2498 		    skb_shinfo(skb)->gso_segs == 1)
2499 			features &= ~NETIF_F_GSO_MASK;
2500 	}
2501 
2502 	return features;
2503 }
2504 
2505 static const struct net_device_ops ibmvnic_netdev_ops = {
2506 	.ndo_open		= ibmvnic_open,
2507 	.ndo_stop		= ibmvnic_close,
2508 	.ndo_start_xmit		= ibmvnic_xmit,
2509 	.ndo_set_rx_mode	= ibmvnic_set_multi,
2510 	.ndo_set_mac_address	= ibmvnic_set_mac,
2511 	.ndo_validate_addr	= eth_validate_addr,
2512 	.ndo_tx_timeout		= ibmvnic_tx_timeout,
2513 	.ndo_change_mtu		= ibmvnic_change_mtu,
2514 	.ndo_features_check     = ibmvnic_features_check,
2515 };
2516 
2517 /* ethtool functions */
2518 
2519 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2520 				      struct ethtool_link_ksettings *cmd)
2521 {
2522 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2523 	int rc;
2524 
2525 	rc = send_query_phys_parms(adapter);
2526 	if (rc) {
2527 		adapter->speed = SPEED_UNKNOWN;
2528 		adapter->duplex = DUPLEX_UNKNOWN;
2529 	}
2530 	cmd->base.speed = adapter->speed;
2531 	cmd->base.duplex = adapter->duplex;
2532 	cmd->base.port = PORT_FIBRE;
2533 	cmd->base.phy_address = 0;
2534 	cmd->base.autoneg = AUTONEG_ENABLE;
2535 
2536 	return 0;
2537 }
2538 
2539 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2540 				struct ethtool_drvinfo *info)
2541 {
2542 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2543 
2544 	strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2545 	strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2546 	strlcpy(info->fw_version, adapter->fw_version,
2547 		sizeof(info->fw_version));
2548 }
2549 
2550 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2551 {
2552 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2553 
2554 	return adapter->msg_enable;
2555 }
2556 
2557 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2558 {
2559 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2560 
2561 	adapter->msg_enable = data;
2562 }
2563 
2564 static u32 ibmvnic_get_link(struct net_device *netdev)
2565 {
2566 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2567 
2568 	/* Don't need to send a query because we request a logical link up at
2569 	 * init and then we wait for link state indications
2570 	 */
2571 	return adapter->logical_link_state;
2572 }
2573 
2574 static void ibmvnic_get_ringparam(struct net_device *netdev,
2575 				  struct ethtool_ringparam *ring)
2576 {
2577 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2578 
2579 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2580 		ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2581 		ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2582 	} else {
2583 		ring->rx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2584 		ring->tx_max_pending = IBMVNIC_MAX_QUEUE_SZ;
2585 	}
2586 	ring->rx_mini_max_pending = 0;
2587 	ring->rx_jumbo_max_pending = 0;
2588 	ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2589 	ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2590 	ring->rx_mini_pending = 0;
2591 	ring->rx_jumbo_pending = 0;
2592 }
2593 
2594 static int ibmvnic_set_ringparam(struct net_device *netdev,
2595 				 struct ethtool_ringparam *ring)
2596 {
2597 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2598 	int ret;
2599 
2600 	ret = 0;
2601 	adapter->desired.rx_entries = ring->rx_pending;
2602 	adapter->desired.tx_entries = ring->tx_pending;
2603 
2604 	ret = wait_for_reset(adapter);
2605 
2606 	if (!ret &&
2607 	    (adapter->req_rx_add_entries_per_subcrq != ring->rx_pending ||
2608 	     adapter->req_tx_entries_per_subcrq != ring->tx_pending))
2609 		netdev_info(netdev,
2610 			    "Could not match full ringsize request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2611 			    ring->rx_pending, ring->tx_pending,
2612 			    adapter->req_rx_add_entries_per_subcrq,
2613 			    adapter->req_tx_entries_per_subcrq);
2614 	return ret;
2615 }
2616 
2617 static void ibmvnic_get_channels(struct net_device *netdev,
2618 				 struct ethtool_channels *channels)
2619 {
2620 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2621 
2622 	if (adapter->priv_flags & IBMVNIC_USE_SERVER_MAXES) {
2623 		channels->max_rx = adapter->max_rx_queues;
2624 		channels->max_tx = adapter->max_tx_queues;
2625 	} else {
2626 		channels->max_rx = IBMVNIC_MAX_QUEUES;
2627 		channels->max_tx = IBMVNIC_MAX_QUEUES;
2628 	}
2629 
2630 	channels->max_other = 0;
2631 	channels->max_combined = 0;
2632 	channels->rx_count = adapter->req_rx_queues;
2633 	channels->tx_count = adapter->req_tx_queues;
2634 	channels->other_count = 0;
2635 	channels->combined_count = 0;
2636 }
2637 
2638 static int ibmvnic_set_channels(struct net_device *netdev,
2639 				struct ethtool_channels *channels)
2640 {
2641 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2642 	int ret;
2643 
2644 	ret = 0;
2645 	adapter->desired.rx_queues = channels->rx_count;
2646 	adapter->desired.tx_queues = channels->tx_count;
2647 
2648 	ret = wait_for_reset(adapter);
2649 
2650 	if (!ret &&
2651 	    (adapter->req_rx_queues != channels->rx_count ||
2652 	     adapter->req_tx_queues != channels->tx_count))
2653 		netdev_info(netdev,
2654 			    "Could not match full channels request. Requested: RX %d, TX %d; Allowed: RX %llu, TX %llu\n",
2655 			    channels->rx_count, channels->tx_count,
2656 			    adapter->req_rx_queues, adapter->req_tx_queues);
2657 	return ret;
2658 
2659 }
2660 
2661 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2662 {
2663 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2664 	int i;
2665 
2666 	switch (stringset) {
2667 	case ETH_SS_STATS:
2668 		for (i = 0; i < ARRAY_SIZE(ibmvnic_stats);
2669 				i++, data += ETH_GSTRING_LEN)
2670 			memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2671 
2672 		for (i = 0; i < adapter->req_tx_queues; i++) {
2673 			snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2674 			data += ETH_GSTRING_LEN;
2675 
2676 			snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2677 			data += ETH_GSTRING_LEN;
2678 
2679 			snprintf(data, ETH_GSTRING_LEN,
2680 				 "tx%d_dropped_packets", i);
2681 			data += ETH_GSTRING_LEN;
2682 		}
2683 
2684 		for (i = 0; i < adapter->req_rx_queues; i++) {
2685 			snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2686 			data += ETH_GSTRING_LEN;
2687 
2688 			snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2689 			data += ETH_GSTRING_LEN;
2690 
2691 			snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2692 			data += ETH_GSTRING_LEN;
2693 		}
2694 		break;
2695 
2696 	case ETH_SS_PRIV_FLAGS:
2697 		for (i = 0; i < ARRAY_SIZE(ibmvnic_priv_flags); i++)
2698 			strcpy(data + i * ETH_GSTRING_LEN,
2699 			       ibmvnic_priv_flags[i]);
2700 		break;
2701 	default:
2702 		return;
2703 	}
2704 }
2705 
2706 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2707 {
2708 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2709 
2710 	switch (sset) {
2711 	case ETH_SS_STATS:
2712 		return ARRAY_SIZE(ibmvnic_stats) +
2713 		       adapter->req_tx_queues * NUM_TX_STATS +
2714 		       adapter->req_rx_queues * NUM_RX_STATS;
2715 	case ETH_SS_PRIV_FLAGS:
2716 		return ARRAY_SIZE(ibmvnic_priv_flags);
2717 	default:
2718 		return -EOPNOTSUPP;
2719 	}
2720 }
2721 
2722 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2723 				      struct ethtool_stats *stats, u64 *data)
2724 {
2725 	struct ibmvnic_adapter *adapter = netdev_priv(dev);
2726 	union ibmvnic_crq crq;
2727 	int i, j;
2728 	int rc;
2729 
2730 	memset(&crq, 0, sizeof(crq));
2731 	crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2732 	crq.request_statistics.cmd = REQUEST_STATISTICS;
2733 	crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2734 	crq.request_statistics.len =
2735 	    cpu_to_be32(sizeof(struct ibmvnic_statistics));
2736 
2737 	/* Wait for data to be written */
2738 	reinit_completion(&adapter->stats_done);
2739 	rc = ibmvnic_send_crq(adapter, &crq);
2740 	if (rc)
2741 		return;
2742 	rc = ibmvnic_wait_for_completion(adapter, &adapter->stats_done, 10000);
2743 	if (rc)
2744 		return;
2745 
2746 	for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2747 		data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2748 						ibmvnic_stats[i].offset));
2749 
2750 	for (j = 0; j < adapter->req_tx_queues; j++) {
2751 		data[i] = adapter->tx_stats_buffers[j].packets;
2752 		i++;
2753 		data[i] = adapter->tx_stats_buffers[j].bytes;
2754 		i++;
2755 		data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2756 		i++;
2757 	}
2758 
2759 	for (j = 0; j < adapter->req_rx_queues; j++) {
2760 		data[i] = adapter->rx_stats_buffers[j].packets;
2761 		i++;
2762 		data[i] = adapter->rx_stats_buffers[j].bytes;
2763 		i++;
2764 		data[i] = adapter->rx_stats_buffers[j].interrupts;
2765 		i++;
2766 	}
2767 }
2768 
2769 static u32 ibmvnic_get_priv_flags(struct net_device *netdev)
2770 {
2771 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2772 
2773 	return adapter->priv_flags;
2774 }
2775 
2776 static int ibmvnic_set_priv_flags(struct net_device *netdev, u32 flags)
2777 {
2778 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2779 	bool which_maxes = !!(flags & IBMVNIC_USE_SERVER_MAXES);
2780 
2781 	if (which_maxes)
2782 		adapter->priv_flags |= IBMVNIC_USE_SERVER_MAXES;
2783 	else
2784 		adapter->priv_flags &= ~IBMVNIC_USE_SERVER_MAXES;
2785 
2786 	return 0;
2787 }
2788 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2789 	.get_drvinfo		= ibmvnic_get_drvinfo,
2790 	.get_msglevel		= ibmvnic_get_msglevel,
2791 	.set_msglevel		= ibmvnic_set_msglevel,
2792 	.get_link		= ibmvnic_get_link,
2793 	.get_ringparam		= ibmvnic_get_ringparam,
2794 	.set_ringparam		= ibmvnic_set_ringparam,
2795 	.get_channels		= ibmvnic_get_channels,
2796 	.set_channels		= ibmvnic_set_channels,
2797 	.get_strings            = ibmvnic_get_strings,
2798 	.get_sset_count         = ibmvnic_get_sset_count,
2799 	.get_ethtool_stats	= ibmvnic_get_ethtool_stats,
2800 	.get_link_ksettings	= ibmvnic_get_link_ksettings,
2801 	.get_priv_flags		= ibmvnic_get_priv_flags,
2802 	.set_priv_flags		= ibmvnic_set_priv_flags,
2803 };
2804 
2805 /* Routines for managing CRQs/sCRQs  */
2806 
2807 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2808 				   struct ibmvnic_sub_crq_queue *scrq)
2809 {
2810 	int rc;
2811 
2812 	if (scrq->irq) {
2813 		free_irq(scrq->irq, scrq);
2814 		irq_dispose_mapping(scrq->irq);
2815 		scrq->irq = 0;
2816 	}
2817 
2818 	memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2819 	atomic_set(&scrq->used, 0);
2820 	scrq->cur = 0;
2821 
2822 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2823 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2824 	return rc;
2825 }
2826 
2827 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2828 {
2829 	int i, rc;
2830 
2831 	for (i = 0; i < adapter->req_tx_queues; i++) {
2832 		netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2833 		rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2834 		if (rc)
2835 			return rc;
2836 	}
2837 
2838 	for (i = 0; i < adapter->req_rx_queues; i++) {
2839 		netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2840 		rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2841 		if (rc)
2842 			return rc;
2843 	}
2844 
2845 	return rc;
2846 }
2847 
2848 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2849 				  struct ibmvnic_sub_crq_queue *scrq,
2850 				  bool do_h_free)
2851 {
2852 	struct device *dev = &adapter->vdev->dev;
2853 	long rc;
2854 
2855 	netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2856 
2857 	if (do_h_free) {
2858 		/* Close the sub-crqs */
2859 		do {
2860 			rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2861 						adapter->vdev->unit_address,
2862 						scrq->crq_num);
2863 		} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2864 
2865 		if (rc) {
2866 			netdev_err(adapter->netdev,
2867 				   "Failed to release sub-CRQ %16lx, rc = %ld\n",
2868 				   scrq->crq_num, rc);
2869 		}
2870 	}
2871 
2872 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2873 			 DMA_BIDIRECTIONAL);
2874 	free_pages((unsigned long)scrq->msgs, 2);
2875 	kfree(scrq);
2876 }
2877 
2878 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2879 							*adapter)
2880 {
2881 	struct device *dev = &adapter->vdev->dev;
2882 	struct ibmvnic_sub_crq_queue *scrq;
2883 	int rc;
2884 
2885 	scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2886 	if (!scrq)
2887 		return NULL;
2888 
2889 	scrq->msgs =
2890 		(union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2891 	if (!scrq->msgs) {
2892 		dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2893 		goto zero_page_failed;
2894 	}
2895 
2896 	scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2897 					 DMA_BIDIRECTIONAL);
2898 	if (dma_mapping_error(dev, scrq->msg_token)) {
2899 		dev_warn(dev, "Couldn't map crq queue messages page\n");
2900 		goto map_failed;
2901 	}
2902 
2903 	rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2904 			   4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2905 
2906 	if (rc == H_RESOURCE)
2907 		rc = ibmvnic_reset_crq(adapter);
2908 
2909 	if (rc == H_CLOSED) {
2910 		dev_warn(dev, "Partner adapter not ready, waiting.\n");
2911 	} else if (rc) {
2912 		dev_warn(dev, "Error %d registering sub-crq\n", rc);
2913 		goto reg_failed;
2914 	}
2915 
2916 	scrq->adapter = adapter;
2917 	scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2918 	spin_lock_init(&scrq->lock);
2919 
2920 	netdev_dbg(adapter->netdev,
2921 		   "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2922 		   scrq->crq_num, scrq->hw_irq, scrq->irq);
2923 
2924 	return scrq;
2925 
2926 reg_failed:
2927 	dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2928 			 DMA_BIDIRECTIONAL);
2929 map_failed:
2930 	free_pages((unsigned long)scrq->msgs, 2);
2931 zero_page_failed:
2932 	kfree(scrq);
2933 
2934 	return NULL;
2935 }
2936 
2937 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2938 {
2939 	int i;
2940 
2941 	if (adapter->tx_scrq) {
2942 		for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2943 			if (!adapter->tx_scrq[i])
2944 				continue;
2945 
2946 			netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2947 				   i);
2948 			if (adapter->tx_scrq[i]->irq) {
2949 				free_irq(adapter->tx_scrq[i]->irq,
2950 					 adapter->tx_scrq[i]);
2951 				irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2952 				adapter->tx_scrq[i]->irq = 0;
2953 			}
2954 
2955 			release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2956 					      do_h_free);
2957 		}
2958 
2959 		kfree(adapter->tx_scrq);
2960 		adapter->tx_scrq = NULL;
2961 		adapter->num_active_tx_scrqs = 0;
2962 	}
2963 
2964 	if (adapter->rx_scrq) {
2965 		for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2966 			if (!adapter->rx_scrq[i])
2967 				continue;
2968 
2969 			netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2970 				   i);
2971 			if (adapter->rx_scrq[i]->irq) {
2972 				free_irq(adapter->rx_scrq[i]->irq,
2973 					 adapter->rx_scrq[i]);
2974 				irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2975 				adapter->rx_scrq[i]->irq = 0;
2976 			}
2977 
2978 			release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2979 					      do_h_free);
2980 		}
2981 
2982 		kfree(adapter->rx_scrq);
2983 		adapter->rx_scrq = NULL;
2984 		adapter->num_active_rx_scrqs = 0;
2985 	}
2986 }
2987 
2988 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2989 			    struct ibmvnic_sub_crq_queue *scrq)
2990 {
2991 	struct device *dev = &adapter->vdev->dev;
2992 	unsigned long rc;
2993 
2994 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2995 				H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2996 	if (rc)
2997 		dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2998 			scrq->hw_irq, rc);
2999 	return rc;
3000 }
3001 
3002 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
3003 			   struct ibmvnic_sub_crq_queue *scrq)
3004 {
3005 	struct device *dev = &adapter->vdev->dev;
3006 	unsigned long rc;
3007 
3008 	if (scrq->hw_irq > 0x100000000ULL) {
3009 		dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
3010 		return 1;
3011 	}
3012 
3013 	if (test_bit(0, &adapter->resetting) &&
3014 	    adapter->reset_reason == VNIC_RESET_MOBILITY) {
3015 		u64 val = (0xff000000) | scrq->hw_irq;
3016 
3017 		rc = plpar_hcall_norets(H_EOI, val);
3018 		/* H_EOI would fail with rc = H_FUNCTION when running
3019 		 * in XIVE mode which is expected, but not an error.
3020 		 */
3021 		if (rc && (rc != H_FUNCTION))
3022 			dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
3023 				val, rc);
3024 	}
3025 
3026 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
3027 				H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
3028 	if (rc)
3029 		dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
3030 			scrq->hw_irq, rc);
3031 	return rc;
3032 }
3033 
3034 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
3035 			       struct ibmvnic_sub_crq_queue *scrq)
3036 {
3037 	struct device *dev = &adapter->vdev->dev;
3038 	struct ibmvnic_tx_pool *tx_pool;
3039 	struct ibmvnic_tx_buff *txbuff;
3040 	union sub_crq *next;
3041 	int index;
3042 	int i, j;
3043 
3044 restart_loop:
3045 	while (pending_scrq(adapter, scrq)) {
3046 		unsigned int pool = scrq->pool_index;
3047 		int num_entries = 0;
3048 
3049 		next = ibmvnic_next_scrq(adapter, scrq);
3050 		for (i = 0; i < next->tx_comp.num_comps; i++) {
3051 			if (next->tx_comp.rcs[i]) {
3052 				dev_err(dev, "tx error %x\n",
3053 					next->tx_comp.rcs[i]);
3054 				continue;
3055 			}
3056 			index = be32_to_cpu(next->tx_comp.correlators[i]);
3057 			if (index & IBMVNIC_TSO_POOL_MASK) {
3058 				tx_pool = &adapter->tso_pool[pool];
3059 				index &= ~IBMVNIC_TSO_POOL_MASK;
3060 			} else {
3061 				tx_pool = &adapter->tx_pool[pool];
3062 			}
3063 
3064 			txbuff = &tx_pool->tx_buff[index];
3065 
3066 			for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
3067 				if (!txbuff->data_dma[j])
3068 					continue;
3069 
3070 				txbuff->data_dma[j] = 0;
3071 			}
3072 
3073 			if (txbuff->last_frag) {
3074 				dev_kfree_skb_any(txbuff->skb);
3075 				txbuff->skb = NULL;
3076 			}
3077 
3078 			num_entries += txbuff->num_entries;
3079 
3080 			tx_pool->free_map[tx_pool->producer_index] = index;
3081 			tx_pool->producer_index =
3082 				(tx_pool->producer_index + 1) %
3083 					tx_pool->num_buffers;
3084 		}
3085 		/* remove tx_comp scrq*/
3086 		next->tx_comp.first = 0;
3087 
3088 		if (atomic_sub_return(num_entries, &scrq->used) <=
3089 		    (adapter->req_tx_entries_per_subcrq / 2) &&
3090 		    __netif_subqueue_stopped(adapter->netdev,
3091 					     scrq->pool_index)) {
3092 			netif_wake_subqueue(adapter->netdev, scrq->pool_index);
3093 			netdev_dbg(adapter->netdev, "Started queue %d\n",
3094 				   scrq->pool_index);
3095 		}
3096 	}
3097 
3098 	enable_scrq_irq(adapter, scrq);
3099 
3100 	if (pending_scrq(adapter, scrq)) {
3101 		disable_scrq_irq(adapter, scrq);
3102 		goto restart_loop;
3103 	}
3104 
3105 	return 0;
3106 }
3107 
3108 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
3109 {
3110 	struct ibmvnic_sub_crq_queue *scrq = instance;
3111 	struct ibmvnic_adapter *adapter = scrq->adapter;
3112 
3113 	disable_scrq_irq(adapter, scrq);
3114 	ibmvnic_complete_tx(adapter, scrq);
3115 
3116 	return IRQ_HANDLED;
3117 }
3118 
3119 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
3120 {
3121 	struct ibmvnic_sub_crq_queue *scrq = instance;
3122 	struct ibmvnic_adapter *adapter = scrq->adapter;
3123 
3124 	/* When booting a kdump kernel we can hit pending interrupts
3125 	 * prior to completing driver initialization.
3126 	 */
3127 	if (unlikely(adapter->state != VNIC_OPEN))
3128 		return IRQ_NONE;
3129 
3130 	adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
3131 
3132 	if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
3133 		disable_scrq_irq(adapter, scrq);
3134 		__napi_schedule(&adapter->napi[scrq->scrq_num]);
3135 	}
3136 
3137 	return IRQ_HANDLED;
3138 }
3139 
3140 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
3141 {
3142 	struct device *dev = &adapter->vdev->dev;
3143 	struct ibmvnic_sub_crq_queue *scrq;
3144 	int i = 0, j = 0;
3145 	int rc = 0;
3146 
3147 	for (i = 0; i < adapter->req_tx_queues; i++) {
3148 		netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
3149 			   i);
3150 		scrq = adapter->tx_scrq[i];
3151 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3152 
3153 		if (!scrq->irq) {
3154 			rc = -EINVAL;
3155 			dev_err(dev, "Error mapping irq\n");
3156 			goto req_tx_irq_failed;
3157 		}
3158 
3159 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-tx%d",
3160 			 adapter->vdev->unit_address, i);
3161 		rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
3162 				 0, scrq->name, scrq);
3163 
3164 		if (rc) {
3165 			dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
3166 				scrq->irq, rc);
3167 			irq_dispose_mapping(scrq->irq);
3168 			goto req_tx_irq_failed;
3169 		}
3170 	}
3171 
3172 	for (i = 0; i < adapter->req_rx_queues; i++) {
3173 		netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
3174 			   i);
3175 		scrq = adapter->rx_scrq[i];
3176 		scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
3177 		if (!scrq->irq) {
3178 			rc = -EINVAL;
3179 			dev_err(dev, "Error mapping irq\n");
3180 			goto req_rx_irq_failed;
3181 		}
3182 		snprintf(scrq->name, sizeof(scrq->name), "ibmvnic-%x-rx%d",
3183 			 adapter->vdev->unit_address, i);
3184 		rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
3185 				 0, scrq->name, scrq);
3186 		if (rc) {
3187 			dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
3188 				scrq->irq, rc);
3189 			irq_dispose_mapping(scrq->irq);
3190 			goto req_rx_irq_failed;
3191 		}
3192 	}
3193 	return rc;
3194 
3195 req_rx_irq_failed:
3196 	for (j = 0; j < i; j++) {
3197 		free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
3198 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3199 	}
3200 	i = adapter->req_tx_queues;
3201 req_tx_irq_failed:
3202 	for (j = 0; j < i; j++) {
3203 		free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
3204 		irq_dispose_mapping(adapter->rx_scrq[j]->irq);
3205 	}
3206 	release_sub_crqs(adapter, 1);
3207 	return rc;
3208 }
3209 
3210 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
3211 {
3212 	struct device *dev = &adapter->vdev->dev;
3213 	struct ibmvnic_sub_crq_queue **allqueues;
3214 	int registered_queues = 0;
3215 	int total_queues;
3216 	int more = 0;
3217 	int i;
3218 
3219 	total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
3220 
3221 	allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
3222 	if (!allqueues)
3223 		return -1;
3224 
3225 	for (i = 0; i < total_queues; i++) {
3226 		allqueues[i] = init_sub_crq_queue(adapter);
3227 		if (!allqueues[i]) {
3228 			dev_warn(dev, "Couldn't allocate all sub-crqs\n");
3229 			break;
3230 		}
3231 		registered_queues++;
3232 	}
3233 
3234 	/* Make sure we were able to register the minimum number of queues */
3235 	if (registered_queues <
3236 	    adapter->min_tx_queues + adapter->min_rx_queues) {
3237 		dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
3238 		goto tx_failed;
3239 	}
3240 
3241 	/* Distribute the failed allocated queues*/
3242 	for (i = 0; i < total_queues - registered_queues + more ; i++) {
3243 		netdev_dbg(adapter->netdev, "Reducing number of queues\n");
3244 		switch (i % 3) {
3245 		case 0:
3246 			if (adapter->req_rx_queues > adapter->min_rx_queues)
3247 				adapter->req_rx_queues--;
3248 			else
3249 				more++;
3250 			break;
3251 		case 1:
3252 			if (adapter->req_tx_queues > adapter->min_tx_queues)
3253 				adapter->req_tx_queues--;
3254 			else
3255 				more++;
3256 			break;
3257 		}
3258 	}
3259 
3260 	adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
3261 				   sizeof(*adapter->tx_scrq), GFP_KERNEL);
3262 	if (!adapter->tx_scrq)
3263 		goto tx_failed;
3264 
3265 	for (i = 0; i < adapter->req_tx_queues; i++) {
3266 		adapter->tx_scrq[i] = allqueues[i];
3267 		adapter->tx_scrq[i]->pool_index = i;
3268 		adapter->num_active_tx_scrqs++;
3269 	}
3270 
3271 	adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3272 				   sizeof(*adapter->rx_scrq), GFP_KERNEL);
3273 	if (!adapter->rx_scrq)
3274 		goto rx_failed;
3275 
3276 	for (i = 0; i < adapter->req_rx_queues; i++) {
3277 		adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3278 		adapter->rx_scrq[i]->scrq_num = i;
3279 		adapter->num_active_rx_scrqs++;
3280 	}
3281 
3282 	kfree(allqueues);
3283 	return 0;
3284 
3285 rx_failed:
3286 	kfree(adapter->tx_scrq);
3287 	adapter->tx_scrq = NULL;
3288 tx_failed:
3289 	for (i = 0; i < registered_queues; i++)
3290 		release_sub_crq_queue(adapter, allqueues[i], 1);
3291 	kfree(allqueues);
3292 	return -1;
3293 }
3294 
3295 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3296 {
3297 	struct device *dev = &adapter->vdev->dev;
3298 	union ibmvnic_crq crq;
3299 	int max_entries;
3300 
3301 	if (!retry) {
3302 		/* Sub-CRQ entries are 32 byte long */
3303 		int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3304 
3305 		if (adapter->min_tx_entries_per_subcrq > entries_page ||
3306 		    adapter->min_rx_add_entries_per_subcrq > entries_page) {
3307 			dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3308 			return;
3309 		}
3310 
3311 		if (adapter->desired.mtu)
3312 			adapter->req_mtu = adapter->desired.mtu;
3313 		else
3314 			adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3315 
3316 		if (!adapter->desired.tx_entries)
3317 			adapter->desired.tx_entries =
3318 					adapter->max_tx_entries_per_subcrq;
3319 		if (!adapter->desired.rx_entries)
3320 			adapter->desired.rx_entries =
3321 					adapter->max_rx_add_entries_per_subcrq;
3322 
3323 		max_entries = IBMVNIC_MAX_LTB_SIZE /
3324 			      (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3325 
3326 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3327 			adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3328 			adapter->desired.tx_entries = max_entries;
3329 		}
3330 
3331 		if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3332 			adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3333 			adapter->desired.rx_entries = max_entries;
3334 		}
3335 
3336 		if (adapter->desired.tx_entries)
3337 			adapter->req_tx_entries_per_subcrq =
3338 					adapter->desired.tx_entries;
3339 		else
3340 			adapter->req_tx_entries_per_subcrq =
3341 					adapter->max_tx_entries_per_subcrq;
3342 
3343 		if (adapter->desired.rx_entries)
3344 			adapter->req_rx_add_entries_per_subcrq =
3345 					adapter->desired.rx_entries;
3346 		else
3347 			adapter->req_rx_add_entries_per_subcrq =
3348 					adapter->max_rx_add_entries_per_subcrq;
3349 
3350 		if (adapter->desired.tx_queues)
3351 			adapter->req_tx_queues =
3352 					adapter->desired.tx_queues;
3353 		else
3354 			adapter->req_tx_queues =
3355 					adapter->opt_tx_comp_sub_queues;
3356 
3357 		if (adapter->desired.rx_queues)
3358 			adapter->req_rx_queues =
3359 					adapter->desired.rx_queues;
3360 		else
3361 			adapter->req_rx_queues =
3362 					adapter->opt_rx_comp_queues;
3363 
3364 		adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3365 	}
3366 
3367 	memset(&crq, 0, sizeof(crq));
3368 	crq.request_capability.first = IBMVNIC_CRQ_CMD;
3369 	crq.request_capability.cmd = REQUEST_CAPABILITY;
3370 
3371 	crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3372 	crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3373 	atomic_inc(&adapter->running_cap_crqs);
3374 	ibmvnic_send_crq(adapter, &crq);
3375 
3376 	crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3377 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3378 	atomic_inc(&adapter->running_cap_crqs);
3379 	ibmvnic_send_crq(adapter, &crq);
3380 
3381 	crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3382 	crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3383 	atomic_inc(&adapter->running_cap_crqs);
3384 	ibmvnic_send_crq(adapter, &crq);
3385 
3386 	crq.request_capability.capability =
3387 	    cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3388 	crq.request_capability.number =
3389 	    cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3390 	atomic_inc(&adapter->running_cap_crqs);
3391 	ibmvnic_send_crq(adapter, &crq);
3392 
3393 	crq.request_capability.capability =
3394 	    cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3395 	crq.request_capability.number =
3396 	    cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3397 	atomic_inc(&adapter->running_cap_crqs);
3398 	ibmvnic_send_crq(adapter, &crq);
3399 
3400 	crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3401 	crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3402 	atomic_inc(&adapter->running_cap_crqs);
3403 	ibmvnic_send_crq(adapter, &crq);
3404 
3405 	if (adapter->netdev->flags & IFF_PROMISC) {
3406 		if (adapter->promisc_supported) {
3407 			crq.request_capability.capability =
3408 			    cpu_to_be16(PROMISC_REQUESTED);
3409 			crq.request_capability.number = cpu_to_be64(1);
3410 			atomic_inc(&adapter->running_cap_crqs);
3411 			ibmvnic_send_crq(adapter, &crq);
3412 		}
3413 	} else {
3414 		crq.request_capability.capability =
3415 		    cpu_to_be16(PROMISC_REQUESTED);
3416 		crq.request_capability.number = cpu_to_be64(0);
3417 		atomic_inc(&adapter->running_cap_crqs);
3418 		ibmvnic_send_crq(adapter, &crq);
3419 	}
3420 }
3421 
3422 static int pending_scrq(struct ibmvnic_adapter *adapter,
3423 			struct ibmvnic_sub_crq_queue *scrq)
3424 {
3425 	union sub_crq *entry = &scrq->msgs[scrq->cur];
3426 
3427 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3428 		return 1;
3429 	else
3430 		return 0;
3431 }
3432 
3433 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3434 					struct ibmvnic_sub_crq_queue *scrq)
3435 {
3436 	union sub_crq *entry;
3437 	unsigned long flags;
3438 
3439 	spin_lock_irqsave(&scrq->lock, flags);
3440 	entry = &scrq->msgs[scrq->cur];
3441 	if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3442 		if (++scrq->cur == scrq->size)
3443 			scrq->cur = 0;
3444 	} else {
3445 		entry = NULL;
3446 	}
3447 	spin_unlock_irqrestore(&scrq->lock, flags);
3448 
3449 	return entry;
3450 }
3451 
3452 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3453 {
3454 	struct ibmvnic_crq_queue *queue = &adapter->crq;
3455 	union ibmvnic_crq *crq;
3456 
3457 	crq = &queue->msgs[queue->cur];
3458 	if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3459 		if (++queue->cur == queue->size)
3460 			queue->cur = 0;
3461 	} else {
3462 		crq = NULL;
3463 	}
3464 
3465 	return crq;
3466 }
3467 
3468 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3469 {
3470 	switch (rc) {
3471 	case H_PARAMETER:
3472 		dev_warn_ratelimited(dev,
3473 				     "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3474 				     func, rc);
3475 		break;
3476 	case H_CLOSED:
3477 		dev_warn_ratelimited(dev,
3478 				     "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3479 				     func, rc);
3480 		break;
3481 	default:
3482 		dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3483 		break;
3484 	}
3485 }
3486 
3487 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3488 		       union sub_crq *sub_crq)
3489 {
3490 	unsigned int ua = adapter->vdev->unit_address;
3491 	struct device *dev = &adapter->vdev->dev;
3492 	u64 *u64_crq = (u64 *)sub_crq;
3493 	int rc;
3494 
3495 	netdev_dbg(adapter->netdev,
3496 		   "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3497 		   (unsigned long int)cpu_to_be64(remote_handle),
3498 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3499 		   (unsigned long int)cpu_to_be64(u64_crq[1]),
3500 		   (unsigned long int)cpu_to_be64(u64_crq[2]),
3501 		   (unsigned long int)cpu_to_be64(u64_crq[3]));
3502 
3503 	/* Make sure the hypervisor sees the complete request */
3504 	mb();
3505 
3506 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3507 				cpu_to_be64(remote_handle),
3508 				cpu_to_be64(u64_crq[0]),
3509 				cpu_to_be64(u64_crq[1]),
3510 				cpu_to_be64(u64_crq[2]),
3511 				cpu_to_be64(u64_crq[3]));
3512 
3513 	if (rc)
3514 		print_subcrq_error(dev, rc, __func__);
3515 
3516 	return rc;
3517 }
3518 
3519 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3520 				u64 remote_handle, u64 ioba, u64 num_entries)
3521 {
3522 	unsigned int ua = adapter->vdev->unit_address;
3523 	struct device *dev = &adapter->vdev->dev;
3524 	int rc;
3525 
3526 	/* Make sure the hypervisor sees the complete request */
3527 	mb();
3528 	rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3529 				cpu_to_be64(remote_handle),
3530 				ioba, num_entries);
3531 
3532 	if (rc)
3533 		print_subcrq_error(dev, rc, __func__);
3534 
3535 	return rc;
3536 }
3537 
3538 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3539 			    union ibmvnic_crq *crq)
3540 {
3541 	unsigned int ua = adapter->vdev->unit_address;
3542 	struct device *dev = &adapter->vdev->dev;
3543 	u64 *u64_crq = (u64 *)crq;
3544 	int rc;
3545 
3546 	netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3547 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
3548 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
3549 
3550 	if (!adapter->crq.active &&
3551 	    crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3552 		dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3553 		return -EINVAL;
3554 	}
3555 
3556 	/* Make sure the hypervisor sees the complete request */
3557 	mb();
3558 
3559 	rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3560 				cpu_to_be64(u64_crq[0]),
3561 				cpu_to_be64(u64_crq[1]));
3562 
3563 	if (rc) {
3564 		if (rc == H_CLOSED) {
3565 			dev_warn(dev, "CRQ Queue closed\n");
3566 			if (test_bit(0, &adapter->resetting))
3567 				ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3568 		}
3569 
3570 		dev_warn(dev, "Send error (rc=%d)\n", rc);
3571 	}
3572 
3573 	return rc;
3574 }
3575 
3576 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3577 {
3578 	union ibmvnic_crq crq;
3579 
3580 	memset(&crq, 0, sizeof(crq));
3581 	crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3582 	crq.generic.cmd = IBMVNIC_CRQ_INIT;
3583 	netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3584 
3585 	return ibmvnic_send_crq(adapter, &crq);
3586 }
3587 
3588 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3589 {
3590 	union ibmvnic_crq crq;
3591 
3592 	memset(&crq, 0, sizeof(crq));
3593 	crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3594 	crq.version_exchange.cmd = VERSION_EXCHANGE;
3595 	crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3596 
3597 	return ibmvnic_send_crq(adapter, &crq);
3598 }
3599 
3600 struct vnic_login_client_data {
3601 	u8	type;
3602 	__be16	len;
3603 	char	name[];
3604 } __packed;
3605 
3606 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3607 {
3608 	int len;
3609 
3610 	/* Calculate the amount of buffer space needed for the
3611 	 * vnic client data in the login buffer. There are four entries,
3612 	 * OS name, LPAR name, device name, and a null last entry.
3613 	 */
3614 	len = 4 * sizeof(struct vnic_login_client_data);
3615 	len += 6; /* "Linux" plus NULL */
3616 	len += strlen(utsname()->nodename) + 1;
3617 	len += strlen(adapter->netdev->name) + 1;
3618 
3619 	return len;
3620 }
3621 
3622 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3623 				 struct vnic_login_client_data *vlcd)
3624 {
3625 	const char *os_name = "Linux";
3626 	int len;
3627 
3628 	/* Type 1 - LPAR OS */
3629 	vlcd->type = 1;
3630 	len = strlen(os_name) + 1;
3631 	vlcd->len = cpu_to_be16(len);
3632 	strncpy(vlcd->name, os_name, len);
3633 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3634 
3635 	/* Type 2 - LPAR name */
3636 	vlcd->type = 2;
3637 	len = strlen(utsname()->nodename) + 1;
3638 	vlcd->len = cpu_to_be16(len);
3639 	strncpy(vlcd->name, utsname()->nodename, len);
3640 	vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3641 
3642 	/* Type 3 - device name */
3643 	vlcd->type = 3;
3644 	len = strlen(adapter->netdev->name) + 1;
3645 	vlcd->len = cpu_to_be16(len);
3646 	strncpy(vlcd->name, adapter->netdev->name, len);
3647 }
3648 
3649 static int send_login(struct ibmvnic_adapter *adapter)
3650 {
3651 	struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3652 	struct ibmvnic_login_buffer *login_buffer;
3653 	struct device *dev = &adapter->vdev->dev;
3654 	dma_addr_t rsp_buffer_token;
3655 	dma_addr_t buffer_token;
3656 	size_t rsp_buffer_size;
3657 	union ibmvnic_crq crq;
3658 	size_t buffer_size;
3659 	__be64 *tx_list_p;
3660 	__be64 *rx_list_p;
3661 	int client_data_len;
3662 	struct vnic_login_client_data *vlcd;
3663 	int i;
3664 
3665 	if (!adapter->tx_scrq || !adapter->rx_scrq) {
3666 		netdev_err(adapter->netdev,
3667 			   "RX or TX queues are not allocated, device login failed\n");
3668 		return -1;
3669 	}
3670 
3671 	release_login_rsp_buffer(adapter);
3672 	client_data_len = vnic_client_data_len(adapter);
3673 
3674 	buffer_size =
3675 	    sizeof(struct ibmvnic_login_buffer) +
3676 	    sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3677 	    client_data_len;
3678 
3679 	login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3680 	if (!login_buffer)
3681 		goto buf_alloc_failed;
3682 
3683 	buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3684 				      DMA_TO_DEVICE);
3685 	if (dma_mapping_error(dev, buffer_token)) {
3686 		dev_err(dev, "Couldn't map login buffer\n");
3687 		goto buf_map_failed;
3688 	}
3689 
3690 	rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3691 			  sizeof(u64) * adapter->req_tx_queues +
3692 			  sizeof(u64) * adapter->req_rx_queues +
3693 			  sizeof(u64) * adapter->req_rx_queues +
3694 			  sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3695 
3696 	login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3697 	if (!login_rsp_buffer)
3698 		goto buf_rsp_alloc_failed;
3699 
3700 	rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3701 					  rsp_buffer_size, DMA_FROM_DEVICE);
3702 	if (dma_mapping_error(dev, rsp_buffer_token)) {
3703 		dev_err(dev, "Couldn't map login rsp buffer\n");
3704 		goto buf_rsp_map_failed;
3705 	}
3706 
3707 	adapter->login_buf = login_buffer;
3708 	adapter->login_buf_token = buffer_token;
3709 	adapter->login_buf_sz = buffer_size;
3710 	adapter->login_rsp_buf = login_rsp_buffer;
3711 	adapter->login_rsp_buf_token = rsp_buffer_token;
3712 	adapter->login_rsp_buf_sz = rsp_buffer_size;
3713 
3714 	login_buffer->len = cpu_to_be32(buffer_size);
3715 	login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3716 	login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3717 	login_buffer->off_txcomp_subcrqs =
3718 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3719 	login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3720 	login_buffer->off_rxcomp_subcrqs =
3721 	    cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3722 			sizeof(u64) * adapter->req_tx_queues);
3723 	login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3724 	login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3725 
3726 	tx_list_p = (__be64 *)((char *)login_buffer +
3727 				      sizeof(struct ibmvnic_login_buffer));
3728 	rx_list_p = (__be64 *)((char *)login_buffer +
3729 				      sizeof(struct ibmvnic_login_buffer) +
3730 				      sizeof(u64) * adapter->req_tx_queues);
3731 
3732 	for (i = 0; i < adapter->req_tx_queues; i++) {
3733 		if (adapter->tx_scrq[i]) {
3734 			tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3735 						   crq_num);
3736 		}
3737 	}
3738 
3739 	for (i = 0; i < adapter->req_rx_queues; i++) {
3740 		if (adapter->rx_scrq[i]) {
3741 			rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3742 						   crq_num);
3743 		}
3744 	}
3745 
3746 	/* Insert vNIC login client data */
3747 	vlcd = (struct vnic_login_client_data *)
3748 		((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3749 	login_buffer->client_data_offset =
3750 			cpu_to_be32((char *)vlcd - (char *)login_buffer);
3751 	login_buffer->client_data_len = cpu_to_be32(client_data_len);
3752 
3753 	vnic_add_client_data(adapter, vlcd);
3754 
3755 	netdev_dbg(adapter->netdev, "Login Buffer:\n");
3756 	for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3757 		netdev_dbg(adapter->netdev, "%016lx\n",
3758 			   ((unsigned long int *)(adapter->login_buf))[i]);
3759 	}
3760 
3761 	memset(&crq, 0, sizeof(crq));
3762 	crq.login.first = IBMVNIC_CRQ_CMD;
3763 	crq.login.cmd = LOGIN;
3764 	crq.login.ioba = cpu_to_be32(buffer_token);
3765 	crq.login.len = cpu_to_be32(buffer_size);
3766 	ibmvnic_send_crq(adapter, &crq);
3767 
3768 	return 0;
3769 
3770 buf_rsp_map_failed:
3771 	kfree(login_rsp_buffer);
3772 buf_rsp_alloc_failed:
3773 	dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3774 buf_map_failed:
3775 	kfree(login_buffer);
3776 buf_alloc_failed:
3777 	return -1;
3778 }
3779 
3780 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3781 			    u32 len, u8 map_id)
3782 {
3783 	union ibmvnic_crq crq;
3784 
3785 	memset(&crq, 0, sizeof(crq));
3786 	crq.request_map.first = IBMVNIC_CRQ_CMD;
3787 	crq.request_map.cmd = REQUEST_MAP;
3788 	crq.request_map.map_id = map_id;
3789 	crq.request_map.ioba = cpu_to_be32(addr);
3790 	crq.request_map.len = cpu_to_be32(len);
3791 	return ibmvnic_send_crq(adapter, &crq);
3792 }
3793 
3794 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3795 {
3796 	union ibmvnic_crq crq;
3797 
3798 	memset(&crq, 0, sizeof(crq));
3799 	crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3800 	crq.request_unmap.cmd = REQUEST_UNMAP;
3801 	crq.request_unmap.map_id = map_id;
3802 	return ibmvnic_send_crq(adapter, &crq);
3803 }
3804 
3805 static void send_map_query(struct ibmvnic_adapter *adapter)
3806 {
3807 	union ibmvnic_crq crq;
3808 
3809 	memset(&crq, 0, sizeof(crq));
3810 	crq.query_map.first = IBMVNIC_CRQ_CMD;
3811 	crq.query_map.cmd = QUERY_MAP;
3812 	ibmvnic_send_crq(adapter, &crq);
3813 }
3814 
3815 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3816 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3817 {
3818 	union ibmvnic_crq crq;
3819 
3820 	atomic_set(&adapter->running_cap_crqs, 0);
3821 	memset(&crq, 0, sizeof(crq));
3822 	crq.query_capability.first = IBMVNIC_CRQ_CMD;
3823 	crq.query_capability.cmd = QUERY_CAPABILITY;
3824 
3825 	crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3826 	atomic_inc(&adapter->running_cap_crqs);
3827 	ibmvnic_send_crq(adapter, &crq);
3828 
3829 	crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3830 	atomic_inc(&adapter->running_cap_crqs);
3831 	ibmvnic_send_crq(adapter, &crq);
3832 
3833 	crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3834 	atomic_inc(&adapter->running_cap_crqs);
3835 	ibmvnic_send_crq(adapter, &crq);
3836 
3837 	crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3838 	atomic_inc(&adapter->running_cap_crqs);
3839 	ibmvnic_send_crq(adapter, &crq);
3840 
3841 	crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3842 	atomic_inc(&adapter->running_cap_crqs);
3843 	ibmvnic_send_crq(adapter, &crq);
3844 
3845 	crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3846 	atomic_inc(&adapter->running_cap_crqs);
3847 	ibmvnic_send_crq(adapter, &crq);
3848 
3849 	crq.query_capability.capability =
3850 	    cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3851 	atomic_inc(&adapter->running_cap_crqs);
3852 	ibmvnic_send_crq(adapter, &crq);
3853 
3854 	crq.query_capability.capability =
3855 	    cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3856 	atomic_inc(&adapter->running_cap_crqs);
3857 	ibmvnic_send_crq(adapter, &crq);
3858 
3859 	crq.query_capability.capability =
3860 	    cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3861 	atomic_inc(&adapter->running_cap_crqs);
3862 	ibmvnic_send_crq(adapter, &crq);
3863 
3864 	crq.query_capability.capability =
3865 	    cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3866 	atomic_inc(&adapter->running_cap_crqs);
3867 	ibmvnic_send_crq(adapter, &crq);
3868 
3869 	crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3870 	atomic_inc(&adapter->running_cap_crqs);
3871 	ibmvnic_send_crq(adapter, &crq);
3872 
3873 	crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3874 	atomic_inc(&adapter->running_cap_crqs);
3875 	ibmvnic_send_crq(adapter, &crq);
3876 
3877 	crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3878 	atomic_inc(&adapter->running_cap_crqs);
3879 	ibmvnic_send_crq(adapter, &crq);
3880 
3881 	crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3882 	atomic_inc(&adapter->running_cap_crqs);
3883 	ibmvnic_send_crq(adapter, &crq);
3884 
3885 	crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3886 	atomic_inc(&adapter->running_cap_crqs);
3887 	ibmvnic_send_crq(adapter, &crq);
3888 
3889 	crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3890 	atomic_inc(&adapter->running_cap_crqs);
3891 	ibmvnic_send_crq(adapter, &crq);
3892 
3893 	crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3894 	atomic_inc(&adapter->running_cap_crqs);
3895 	ibmvnic_send_crq(adapter, &crq);
3896 
3897 	crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3898 	atomic_inc(&adapter->running_cap_crqs);
3899 	ibmvnic_send_crq(adapter, &crq);
3900 
3901 	crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3902 	atomic_inc(&adapter->running_cap_crqs);
3903 	ibmvnic_send_crq(adapter, &crq);
3904 
3905 	crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3906 	atomic_inc(&adapter->running_cap_crqs);
3907 	ibmvnic_send_crq(adapter, &crq);
3908 
3909 	crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3910 	atomic_inc(&adapter->running_cap_crqs);
3911 	ibmvnic_send_crq(adapter, &crq);
3912 
3913 	crq.query_capability.capability =
3914 			cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3915 	atomic_inc(&adapter->running_cap_crqs);
3916 	ibmvnic_send_crq(adapter, &crq);
3917 
3918 	crq.query_capability.capability =
3919 			cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3920 	atomic_inc(&adapter->running_cap_crqs);
3921 	ibmvnic_send_crq(adapter, &crq);
3922 
3923 	crq.query_capability.capability =
3924 			cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3925 	atomic_inc(&adapter->running_cap_crqs);
3926 	ibmvnic_send_crq(adapter, &crq);
3927 
3928 	crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3929 	atomic_inc(&adapter->running_cap_crqs);
3930 	ibmvnic_send_crq(adapter, &crq);
3931 }
3932 
3933 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3934 				struct ibmvnic_adapter *adapter)
3935 {
3936 	struct device *dev = &adapter->vdev->dev;
3937 
3938 	if (crq->get_vpd_size_rsp.rc.code) {
3939 		dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3940 			crq->get_vpd_size_rsp.rc.code);
3941 		complete(&adapter->fw_done);
3942 		return;
3943 	}
3944 
3945 	adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3946 	complete(&adapter->fw_done);
3947 }
3948 
3949 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3950 			   struct ibmvnic_adapter *adapter)
3951 {
3952 	struct device *dev = &adapter->vdev->dev;
3953 	unsigned char *substr = NULL;
3954 	u8 fw_level_len = 0;
3955 
3956 	memset(adapter->fw_version, 0, 32);
3957 
3958 	dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3959 			 DMA_FROM_DEVICE);
3960 
3961 	if (crq->get_vpd_rsp.rc.code) {
3962 		dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3963 			crq->get_vpd_rsp.rc.code);
3964 		goto complete;
3965 	}
3966 
3967 	/* get the position of the firmware version info
3968 	 * located after the ASCII 'RM' substring in the buffer
3969 	 */
3970 	substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3971 	if (!substr) {
3972 		dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3973 		goto complete;
3974 	}
3975 
3976 	/* get length of firmware level ASCII substring */
3977 	if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3978 		fw_level_len = *(substr + 2);
3979 	} else {
3980 		dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3981 		goto complete;
3982 	}
3983 
3984 	/* copy firmware version string from vpd into adapter */
3985 	if ((substr + 3 + fw_level_len) <
3986 	    (adapter->vpd->buff + adapter->vpd->len)) {
3987 		strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3988 	} else {
3989 		dev_info(dev, "FW substr extrapolated VPD buff\n");
3990 	}
3991 
3992 complete:
3993 	if (adapter->fw_version[0] == '\0')
3994 		strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3995 	complete(&adapter->fw_done);
3996 }
3997 
3998 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3999 {
4000 	struct device *dev = &adapter->vdev->dev;
4001 	struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
4002 	netdev_features_t old_hw_features = 0;
4003 	union ibmvnic_crq crq;
4004 	int i;
4005 
4006 	dma_unmap_single(dev, adapter->ip_offload_tok,
4007 			 sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
4008 
4009 	netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
4010 	for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
4011 		netdev_dbg(adapter->netdev, "%016lx\n",
4012 			   ((unsigned long int *)(buf))[i]);
4013 
4014 	netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
4015 	netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
4016 	netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
4017 		   buf->tcp_ipv4_chksum);
4018 	netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
4019 		   buf->tcp_ipv6_chksum);
4020 	netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
4021 		   buf->udp_ipv4_chksum);
4022 	netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
4023 		   buf->udp_ipv6_chksum);
4024 	netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
4025 		   buf->large_tx_ipv4);
4026 	netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
4027 		   buf->large_tx_ipv6);
4028 	netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
4029 		   buf->large_rx_ipv4);
4030 	netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
4031 		   buf->large_rx_ipv6);
4032 	netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
4033 		   buf->max_ipv4_header_size);
4034 	netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
4035 		   buf->max_ipv6_header_size);
4036 	netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
4037 		   buf->max_tcp_header_size);
4038 	netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
4039 		   buf->max_udp_header_size);
4040 	netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
4041 		   buf->max_large_tx_size);
4042 	netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
4043 		   buf->max_large_rx_size);
4044 	netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
4045 		   buf->ipv6_extension_header);
4046 	netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
4047 		   buf->tcp_pseudosum_req);
4048 	netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
4049 		   buf->num_ipv6_ext_headers);
4050 	netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
4051 		   buf->off_ipv6_ext_headers);
4052 
4053 	adapter->ip_offload_ctrl_tok =
4054 	    dma_map_single(dev, &adapter->ip_offload_ctrl,
4055 			   sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
4056 
4057 	if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
4058 		dev_err(dev, "Couldn't map ip offload control buffer\n");
4059 		return;
4060 	}
4061 
4062 	adapter->ip_offload_ctrl.len =
4063 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4064 	adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
4065 	adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
4066 	adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
4067 	adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
4068 	adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
4069 	adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
4070 	adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
4071 	adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
4072 	adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
4073 
4074 	/* large_rx disabled for now, additional features needed */
4075 	adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
4076 	adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
4077 
4078 	if (adapter->state != VNIC_PROBING) {
4079 		old_hw_features = adapter->netdev->hw_features;
4080 		adapter->netdev->hw_features = 0;
4081 	}
4082 
4083 	adapter->netdev->hw_features = NETIF_F_SG | NETIF_F_GSO | NETIF_F_GRO;
4084 
4085 	if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
4086 		adapter->netdev->hw_features |= NETIF_F_IP_CSUM;
4087 
4088 	if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
4089 		adapter->netdev->hw_features |= NETIF_F_IPV6_CSUM;
4090 
4091 	if ((adapter->netdev->features &
4092 	    (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
4093 		adapter->netdev->hw_features |= NETIF_F_RXCSUM;
4094 
4095 	if (buf->large_tx_ipv4)
4096 		adapter->netdev->hw_features |= NETIF_F_TSO;
4097 	if (buf->large_tx_ipv6)
4098 		adapter->netdev->hw_features |= NETIF_F_TSO6;
4099 
4100 	if (adapter->state == VNIC_PROBING) {
4101 		adapter->netdev->features |= adapter->netdev->hw_features;
4102 	} else if (old_hw_features != adapter->netdev->hw_features) {
4103 		netdev_features_t tmp = 0;
4104 
4105 		/* disable features no longer supported */
4106 		adapter->netdev->features &= adapter->netdev->hw_features;
4107 		/* turn on features now supported if previously enabled */
4108 		tmp = (old_hw_features ^ adapter->netdev->hw_features) &
4109 			adapter->netdev->hw_features;
4110 		adapter->netdev->features |=
4111 				tmp & adapter->netdev->wanted_features;
4112 	}
4113 
4114 	memset(&crq, 0, sizeof(crq));
4115 	crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
4116 	crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
4117 	crq.control_ip_offload.len =
4118 	    cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
4119 	crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
4120 	ibmvnic_send_crq(adapter, &crq);
4121 }
4122 
4123 static const char *ibmvnic_fw_err_cause(u16 cause)
4124 {
4125 	switch (cause) {
4126 	case ADAPTER_PROBLEM:
4127 		return "adapter problem";
4128 	case BUS_PROBLEM:
4129 		return "bus problem";
4130 	case FW_PROBLEM:
4131 		return "firmware problem";
4132 	case DD_PROBLEM:
4133 		return "device driver problem";
4134 	case EEH_RECOVERY:
4135 		return "EEH recovery";
4136 	case FW_UPDATED:
4137 		return "firmware updated";
4138 	case LOW_MEMORY:
4139 		return "low Memory";
4140 	default:
4141 		return "unknown";
4142 	}
4143 }
4144 
4145 static void handle_error_indication(union ibmvnic_crq *crq,
4146 				    struct ibmvnic_adapter *adapter)
4147 {
4148 	struct device *dev = &adapter->vdev->dev;
4149 	u16 cause;
4150 
4151 	cause = be16_to_cpu(crq->error_indication.error_cause);
4152 
4153 	dev_warn_ratelimited(dev,
4154 			     "Firmware reports %serror, cause: %s. Starting recovery...\n",
4155 			     crq->error_indication.flags
4156 				& IBMVNIC_FATAL_ERROR ? "FATAL " : "",
4157 			     ibmvnic_fw_err_cause(cause));
4158 
4159 	if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
4160 		ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4161 	else
4162 		ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
4163 }
4164 
4165 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
4166 				 struct ibmvnic_adapter *adapter)
4167 {
4168 	struct net_device *netdev = adapter->netdev;
4169 	struct device *dev = &adapter->vdev->dev;
4170 	long rc;
4171 
4172 	rc = crq->change_mac_addr_rsp.rc.code;
4173 	if (rc) {
4174 		dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
4175 		goto out;
4176 	}
4177 	ether_addr_copy(netdev->dev_addr,
4178 			&crq->change_mac_addr_rsp.mac_addr[0]);
4179 out:
4180 	complete(&adapter->fw_done);
4181 	return rc;
4182 }
4183 
4184 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
4185 				   struct ibmvnic_adapter *adapter)
4186 {
4187 	struct device *dev = &adapter->vdev->dev;
4188 	u64 *req_value;
4189 	char *name;
4190 
4191 	atomic_dec(&adapter->running_cap_crqs);
4192 	switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
4193 	case REQ_TX_QUEUES:
4194 		req_value = &adapter->req_tx_queues;
4195 		name = "tx";
4196 		break;
4197 	case REQ_RX_QUEUES:
4198 		req_value = &adapter->req_rx_queues;
4199 		name = "rx";
4200 		break;
4201 	case REQ_RX_ADD_QUEUES:
4202 		req_value = &adapter->req_rx_add_queues;
4203 		name = "rx_add";
4204 		break;
4205 	case REQ_TX_ENTRIES_PER_SUBCRQ:
4206 		req_value = &adapter->req_tx_entries_per_subcrq;
4207 		name = "tx_entries_per_subcrq";
4208 		break;
4209 	case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
4210 		req_value = &adapter->req_rx_add_entries_per_subcrq;
4211 		name = "rx_add_entries_per_subcrq";
4212 		break;
4213 	case REQ_MTU:
4214 		req_value = &adapter->req_mtu;
4215 		name = "mtu";
4216 		break;
4217 	case PROMISC_REQUESTED:
4218 		req_value = &adapter->promisc;
4219 		name = "promisc";
4220 		break;
4221 	default:
4222 		dev_err(dev, "Got invalid cap request rsp %d\n",
4223 			crq->request_capability.capability);
4224 		return;
4225 	}
4226 
4227 	switch (crq->request_capability_rsp.rc.code) {
4228 	case SUCCESS:
4229 		break;
4230 	case PARTIALSUCCESS:
4231 		dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
4232 			 *req_value,
4233 			 (long int)be64_to_cpu(crq->request_capability_rsp.
4234 					       number), name);
4235 
4236 		if (be16_to_cpu(crq->request_capability_rsp.capability) ==
4237 		    REQ_MTU) {
4238 			pr_err("mtu of %llu is not supported. Reverting.\n",
4239 			       *req_value);
4240 			*req_value = adapter->fallback.mtu;
4241 		} else {
4242 			*req_value =
4243 				be64_to_cpu(crq->request_capability_rsp.number);
4244 		}
4245 
4246 		ibmvnic_send_req_caps(adapter, 1);
4247 		return;
4248 	default:
4249 		dev_err(dev, "Error %d in request cap rsp\n",
4250 			crq->request_capability_rsp.rc.code);
4251 		return;
4252 	}
4253 
4254 	/* Done receiving requested capabilities, query IP offload support */
4255 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4256 		union ibmvnic_crq newcrq;
4257 		int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
4258 		struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
4259 		    &adapter->ip_offload_buf;
4260 
4261 		adapter->wait_capability = false;
4262 		adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
4263 							 buf_sz,
4264 							 DMA_FROM_DEVICE);
4265 
4266 		if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
4267 			if (!firmware_has_feature(FW_FEATURE_CMO))
4268 				dev_err(dev, "Couldn't map offload buffer\n");
4269 			return;
4270 		}
4271 
4272 		memset(&newcrq, 0, sizeof(newcrq));
4273 		newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
4274 		newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
4275 		newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
4276 		newcrq.query_ip_offload.ioba =
4277 		    cpu_to_be32(adapter->ip_offload_tok);
4278 
4279 		ibmvnic_send_crq(adapter, &newcrq);
4280 	}
4281 }
4282 
4283 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
4284 			    struct ibmvnic_adapter *adapter)
4285 {
4286 	struct device *dev = &adapter->vdev->dev;
4287 	struct net_device *netdev = adapter->netdev;
4288 	struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4289 	struct ibmvnic_login_buffer *login = adapter->login_buf;
4290 	int i;
4291 
4292 	dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4293 			 DMA_TO_DEVICE);
4294 	dma_unmap_single(dev, adapter->login_rsp_buf_token,
4295 			 adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4296 
4297 	/* If the number of queues requested can't be allocated by the
4298 	 * server, the login response will return with code 1. We will need
4299 	 * to resend the login buffer with fewer queues requested.
4300 	 */
4301 	if (login_rsp_crq->generic.rc.code) {
4302 		adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4303 		complete(&adapter->init_done);
4304 		return 0;
4305 	}
4306 
4307 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
4308 
4309 	netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4310 	for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4311 		netdev_dbg(adapter->netdev, "%016lx\n",
4312 			   ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4313 	}
4314 
4315 	/* Sanity checks */
4316 	if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4317 	    (be32_to_cpu(login->num_rxcomp_subcrqs) *
4318 	     adapter->req_rx_add_queues !=
4319 	     be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4320 		dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4321 		ibmvnic_remove(adapter->vdev);
4322 		return -EIO;
4323 	}
4324 	release_login_buffer(adapter);
4325 	complete(&adapter->init_done);
4326 
4327 	return 0;
4328 }
4329 
4330 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4331 				     struct ibmvnic_adapter *adapter)
4332 {
4333 	struct device *dev = &adapter->vdev->dev;
4334 	long rc;
4335 
4336 	rc = crq->request_unmap_rsp.rc.code;
4337 	if (rc)
4338 		dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4339 }
4340 
4341 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4342 				 struct ibmvnic_adapter *adapter)
4343 {
4344 	struct net_device *netdev = adapter->netdev;
4345 	struct device *dev = &adapter->vdev->dev;
4346 	long rc;
4347 
4348 	rc = crq->query_map_rsp.rc.code;
4349 	if (rc) {
4350 		dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4351 		return;
4352 	}
4353 	netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4354 		   crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4355 		   crq->query_map_rsp.free_pages);
4356 }
4357 
4358 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4359 				 struct ibmvnic_adapter *adapter)
4360 {
4361 	struct net_device *netdev = adapter->netdev;
4362 	struct device *dev = &adapter->vdev->dev;
4363 	long rc;
4364 
4365 	atomic_dec(&adapter->running_cap_crqs);
4366 	netdev_dbg(netdev, "Outstanding queries: %d\n",
4367 		   atomic_read(&adapter->running_cap_crqs));
4368 	rc = crq->query_capability.rc.code;
4369 	if (rc) {
4370 		dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4371 		goto out;
4372 	}
4373 
4374 	switch (be16_to_cpu(crq->query_capability.capability)) {
4375 	case MIN_TX_QUEUES:
4376 		adapter->min_tx_queues =
4377 		    be64_to_cpu(crq->query_capability.number);
4378 		netdev_dbg(netdev, "min_tx_queues = %lld\n",
4379 			   adapter->min_tx_queues);
4380 		break;
4381 	case MIN_RX_QUEUES:
4382 		adapter->min_rx_queues =
4383 		    be64_to_cpu(crq->query_capability.number);
4384 		netdev_dbg(netdev, "min_rx_queues = %lld\n",
4385 			   adapter->min_rx_queues);
4386 		break;
4387 	case MIN_RX_ADD_QUEUES:
4388 		adapter->min_rx_add_queues =
4389 		    be64_to_cpu(crq->query_capability.number);
4390 		netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4391 			   adapter->min_rx_add_queues);
4392 		break;
4393 	case MAX_TX_QUEUES:
4394 		adapter->max_tx_queues =
4395 		    be64_to_cpu(crq->query_capability.number);
4396 		netdev_dbg(netdev, "max_tx_queues = %lld\n",
4397 			   adapter->max_tx_queues);
4398 		break;
4399 	case MAX_RX_QUEUES:
4400 		adapter->max_rx_queues =
4401 		    be64_to_cpu(crq->query_capability.number);
4402 		netdev_dbg(netdev, "max_rx_queues = %lld\n",
4403 			   adapter->max_rx_queues);
4404 		break;
4405 	case MAX_RX_ADD_QUEUES:
4406 		adapter->max_rx_add_queues =
4407 		    be64_to_cpu(crq->query_capability.number);
4408 		netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4409 			   adapter->max_rx_add_queues);
4410 		break;
4411 	case MIN_TX_ENTRIES_PER_SUBCRQ:
4412 		adapter->min_tx_entries_per_subcrq =
4413 		    be64_to_cpu(crq->query_capability.number);
4414 		netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4415 			   adapter->min_tx_entries_per_subcrq);
4416 		break;
4417 	case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4418 		adapter->min_rx_add_entries_per_subcrq =
4419 		    be64_to_cpu(crq->query_capability.number);
4420 		netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4421 			   adapter->min_rx_add_entries_per_subcrq);
4422 		break;
4423 	case MAX_TX_ENTRIES_PER_SUBCRQ:
4424 		adapter->max_tx_entries_per_subcrq =
4425 		    be64_to_cpu(crq->query_capability.number);
4426 		netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4427 			   adapter->max_tx_entries_per_subcrq);
4428 		break;
4429 	case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4430 		adapter->max_rx_add_entries_per_subcrq =
4431 		    be64_to_cpu(crq->query_capability.number);
4432 		netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4433 			   adapter->max_rx_add_entries_per_subcrq);
4434 		break;
4435 	case TCP_IP_OFFLOAD:
4436 		adapter->tcp_ip_offload =
4437 		    be64_to_cpu(crq->query_capability.number);
4438 		netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4439 			   adapter->tcp_ip_offload);
4440 		break;
4441 	case PROMISC_SUPPORTED:
4442 		adapter->promisc_supported =
4443 		    be64_to_cpu(crq->query_capability.number);
4444 		netdev_dbg(netdev, "promisc_supported = %lld\n",
4445 			   adapter->promisc_supported);
4446 		break;
4447 	case MIN_MTU:
4448 		adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4449 		netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4450 		netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4451 		break;
4452 	case MAX_MTU:
4453 		adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4454 		netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4455 		netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4456 		break;
4457 	case MAX_MULTICAST_FILTERS:
4458 		adapter->max_multicast_filters =
4459 		    be64_to_cpu(crq->query_capability.number);
4460 		netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4461 			   adapter->max_multicast_filters);
4462 		break;
4463 	case VLAN_HEADER_INSERTION:
4464 		adapter->vlan_header_insertion =
4465 		    be64_to_cpu(crq->query_capability.number);
4466 		if (adapter->vlan_header_insertion)
4467 			netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4468 		netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4469 			   adapter->vlan_header_insertion);
4470 		break;
4471 	case RX_VLAN_HEADER_INSERTION:
4472 		adapter->rx_vlan_header_insertion =
4473 		    be64_to_cpu(crq->query_capability.number);
4474 		netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4475 			   adapter->rx_vlan_header_insertion);
4476 		break;
4477 	case MAX_TX_SG_ENTRIES:
4478 		adapter->max_tx_sg_entries =
4479 		    be64_to_cpu(crq->query_capability.number);
4480 		netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4481 			   adapter->max_tx_sg_entries);
4482 		break;
4483 	case RX_SG_SUPPORTED:
4484 		adapter->rx_sg_supported =
4485 		    be64_to_cpu(crq->query_capability.number);
4486 		netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4487 			   adapter->rx_sg_supported);
4488 		break;
4489 	case OPT_TX_COMP_SUB_QUEUES:
4490 		adapter->opt_tx_comp_sub_queues =
4491 		    be64_to_cpu(crq->query_capability.number);
4492 		netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4493 			   adapter->opt_tx_comp_sub_queues);
4494 		break;
4495 	case OPT_RX_COMP_QUEUES:
4496 		adapter->opt_rx_comp_queues =
4497 		    be64_to_cpu(crq->query_capability.number);
4498 		netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4499 			   adapter->opt_rx_comp_queues);
4500 		break;
4501 	case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4502 		adapter->opt_rx_bufadd_q_per_rx_comp_q =
4503 		    be64_to_cpu(crq->query_capability.number);
4504 		netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4505 			   adapter->opt_rx_bufadd_q_per_rx_comp_q);
4506 		break;
4507 	case OPT_TX_ENTRIES_PER_SUBCRQ:
4508 		adapter->opt_tx_entries_per_subcrq =
4509 		    be64_to_cpu(crq->query_capability.number);
4510 		netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4511 			   adapter->opt_tx_entries_per_subcrq);
4512 		break;
4513 	case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4514 		adapter->opt_rxba_entries_per_subcrq =
4515 		    be64_to_cpu(crq->query_capability.number);
4516 		netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4517 			   adapter->opt_rxba_entries_per_subcrq);
4518 		break;
4519 	case TX_RX_DESC_REQ:
4520 		adapter->tx_rx_desc_req = crq->query_capability.number;
4521 		netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4522 			   adapter->tx_rx_desc_req);
4523 		break;
4524 
4525 	default:
4526 		netdev_err(netdev, "Got invalid cap rsp %d\n",
4527 			   crq->query_capability.capability);
4528 	}
4529 
4530 out:
4531 	if (atomic_read(&adapter->running_cap_crqs) == 0) {
4532 		adapter->wait_capability = false;
4533 		ibmvnic_send_req_caps(adapter, 0);
4534 	}
4535 }
4536 
4537 static int send_query_phys_parms(struct ibmvnic_adapter *adapter)
4538 {
4539 	union ibmvnic_crq crq;
4540 	int rc;
4541 
4542 	memset(&crq, 0, sizeof(crq));
4543 	crq.query_phys_parms.first = IBMVNIC_CRQ_CMD;
4544 	crq.query_phys_parms.cmd = QUERY_PHYS_PARMS;
4545 
4546 	mutex_lock(&adapter->fw_lock);
4547 	adapter->fw_done_rc = 0;
4548 	reinit_completion(&adapter->fw_done);
4549 
4550 	rc = ibmvnic_send_crq(adapter, &crq);
4551 	if (rc) {
4552 		mutex_unlock(&adapter->fw_lock);
4553 		return rc;
4554 	}
4555 
4556 	rc = ibmvnic_wait_for_completion(adapter, &adapter->fw_done, 10000);
4557 	if (rc) {
4558 		mutex_unlock(&adapter->fw_lock);
4559 		return rc;
4560 	}
4561 
4562 	mutex_unlock(&adapter->fw_lock);
4563 	return adapter->fw_done_rc ? -EIO : 0;
4564 }
4565 
4566 static int handle_query_phys_parms_rsp(union ibmvnic_crq *crq,
4567 				       struct ibmvnic_adapter *adapter)
4568 {
4569 	struct net_device *netdev = adapter->netdev;
4570 	int rc;
4571 	__be32 rspeed = cpu_to_be32(crq->query_phys_parms_rsp.speed);
4572 
4573 	rc = crq->query_phys_parms_rsp.rc.code;
4574 	if (rc) {
4575 		netdev_err(netdev, "Error %d in QUERY_PHYS_PARMS\n", rc);
4576 		return rc;
4577 	}
4578 	switch (rspeed) {
4579 	case IBMVNIC_10MBPS:
4580 		adapter->speed = SPEED_10;
4581 		break;
4582 	case IBMVNIC_100MBPS:
4583 		adapter->speed = SPEED_100;
4584 		break;
4585 	case IBMVNIC_1GBPS:
4586 		adapter->speed = SPEED_1000;
4587 		break;
4588 	case IBMVNIC_10GBP:
4589 		adapter->speed = SPEED_10000;
4590 		break;
4591 	case IBMVNIC_25GBPS:
4592 		adapter->speed = SPEED_25000;
4593 		break;
4594 	case IBMVNIC_40GBPS:
4595 		adapter->speed = SPEED_40000;
4596 		break;
4597 	case IBMVNIC_50GBPS:
4598 		adapter->speed = SPEED_50000;
4599 		break;
4600 	case IBMVNIC_100GBPS:
4601 		adapter->speed = SPEED_100000;
4602 		break;
4603 	default:
4604 		if (netif_carrier_ok(netdev))
4605 			netdev_warn(netdev, "Unknown speed 0x%08x\n", rspeed);
4606 		adapter->speed = SPEED_UNKNOWN;
4607 	}
4608 	if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_FULL_DUPLEX)
4609 		adapter->duplex = DUPLEX_FULL;
4610 	else if (crq->query_phys_parms_rsp.flags1 & IBMVNIC_HALF_DUPLEX)
4611 		adapter->duplex = DUPLEX_HALF;
4612 	else
4613 		adapter->duplex = DUPLEX_UNKNOWN;
4614 
4615 	return rc;
4616 }
4617 
4618 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4619 			       struct ibmvnic_adapter *adapter)
4620 {
4621 	struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4622 	struct net_device *netdev = adapter->netdev;
4623 	struct device *dev = &adapter->vdev->dev;
4624 	u64 *u64_crq = (u64 *)crq;
4625 	long rc;
4626 
4627 	netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4628 		   (unsigned long int)cpu_to_be64(u64_crq[0]),
4629 		   (unsigned long int)cpu_to_be64(u64_crq[1]));
4630 	switch (gen_crq->first) {
4631 	case IBMVNIC_CRQ_INIT_RSP:
4632 		switch (gen_crq->cmd) {
4633 		case IBMVNIC_CRQ_INIT:
4634 			dev_info(dev, "Partner initialized\n");
4635 			adapter->from_passive_init = true;
4636 			adapter->failover_pending = false;
4637 			if (!completion_done(&adapter->init_done)) {
4638 				complete(&adapter->init_done);
4639 				adapter->init_done_rc = -EIO;
4640 			}
4641 			ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4642 			break;
4643 		case IBMVNIC_CRQ_INIT_COMPLETE:
4644 			dev_info(dev, "Partner initialization complete\n");
4645 			adapter->crq.active = true;
4646 			send_version_xchg(adapter);
4647 			break;
4648 		default:
4649 			dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4650 		}
4651 		return;
4652 	case IBMVNIC_CRQ_XPORT_EVENT:
4653 		netif_carrier_off(netdev);
4654 		adapter->crq.active = false;
4655 		/* terminate any thread waiting for a response
4656 		 * from the device
4657 		 */
4658 		if (!completion_done(&adapter->fw_done)) {
4659 			adapter->fw_done_rc = -EIO;
4660 			complete(&adapter->fw_done);
4661 		}
4662 		if (!completion_done(&adapter->stats_done))
4663 			complete(&adapter->stats_done);
4664 		if (test_bit(0, &adapter->resetting))
4665 			adapter->force_reset_recovery = true;
4666 		if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4667 			dev_info(dev, "Migrated, re-enabling adapter\n");
4668 			ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4669 		} else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4670 			dev_info(dev, "Backing device failover detected\n");
4671 			adapter->failover_pending = true;
4672 		} else {
4673 			/* The adapter lost the connection */
4674 			dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4675 				gen_crq->cmd);
4676 			ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4677 		}
4678 		return;
4679 	case IBMVNIC_CRQ_CMD_RSP:
4680 		break;
4681 	default:
4682 		dev_err(dev, "Got an invalid msg type 0x%02x\n",
4683 			gen_crq->first);
4684 		return;
4685 	}
4686 
4687 	switch (gen_crq->cmd) {
4688 	case VERSION_EXCHANGE_RSP:
4689 		rc = crq->version_exchange_rsp.rc.code;
4690 		if (rc) {
4691 			dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4692 			break;
4693 		}
4694 		ibmvnic_version =
4695 			    be16_to_cpu(crq->version_exchange_rsp.version);
4696 		dev_info(dev, "Partner protocol version is %d\n",
4697 			 ibmvnic_version);
4698 		send_cap_queries(adapter);
4699 		break;
4700 	case QUERY_CAPABILITY_RSP:
4701 		handle_query_cap_rsp(crq, adapter);
4702 		break;
4703 	case QUERY_MAP_RSP:
4704 		handle_query_map_rsp(crq, adapter);
4705 		break;
4706 	case REQUEST_MAP_RSP:
4707 		adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4708 		complete(&adapter->fw_done);
4709 		break;
4710 	case REQUEST_UNMAP_RSP:
4711 		handle_request_unmap_rsp(crq, adapter);
4712 		break;
4713 	case REQUEST_CAPABILITY_RSP:
4714 		handle_request_cap_rsp(crq, adapter);
4715 		break;
4716 	case LOGIN_RSP:
4717 		netdev_dbg(netdev, "Got Login Response\n");
4718 		handle_login_rsp(crq, adapter);
4719 		break;
4720 	case LOGICAL_LINK_STATE_RSP:
4721 		netdev_dbg(netdev,
4722 			   "Got Logical Link State Response, state: %d rc: %d\n",
4723 			   crq->logical_link_state_rsp.link_state,
4724 			   crq->logical_link_state_rsp.rc.code);
4725 		adapter->logical_link_state =
4726 		    crq->logical_link_state_rsp.link_state;
4727 		adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4728 		complete(&adapter->init_done);
4729 		break;
4730 	case LINK_STATE_INDICATION:
4731 		netdev_dbg(netdev, "Got Logical Link State Indication\n");
4732 		adapter->phys_link_state =
4733 		    crq->link_state_indication.phys_link_state;
4734 		adapter->logical_link_state =
4735 		    crq->link_state_indication.logical_link_state;
4736 		if (adapter->phys_link_state && adapter->logical_link_state)
4737 			netif_carrier_on(netdev);
4738 		else
4739 			netif_carrier_off(netdev);
4740 		break;
4741 	case CHANGE_MAC_ADDR_RSP:
4742 		netdev_dbg(netdev, "Got MAC address change Response\n");
4743 		adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4744 		break;
4745 	case ERROR_INDICATION:
4746 		netdev_dbg(netdev, "Got Error Indication\n");
4747 		handle_error_indication(crq, adapter);
4748 		break;
4749 	case REQUEST_STATISTICS_RSP:
4750 		netdev_dbg(netdev, "Got Statistics Response\n");
4751 		complete(&adapter->stats_done);
4752 		break;
4753 	case QUERY_IP_OFFLOAD_RSP:
4754 		netdev_dbg(netdev, "Got Query IP offload Response\n");
4755 		handle_query_ip_offload_rsp(adapter);
4756 		break;
4757 	case MULTICAST_CTRL_RSP:
4758 		netdev_dbg(netdev, "Got multicast control Response\n");
4759 		break;
4760 	case CONTROL_IP_OFFLOAD_RSP:
4761 		netdev_dbg(netdev, "Got Control IP offload Response\n");
4762 		dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4763 				 sizeof(adapter->ip_offload_ctrl),
4764 				 DMA_TO_DEVICE);
4765 		complete(&adapter->init_done);
4766 		break;
4767 	case COLLECT_FW_TRACE_RSP:
4768 		netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4769 		complete(&adapter->fw_done);
4770 		break;
4771 	case GET_VPD_SIZE_RSP:
4772 		handle_vpd_size_rsp(crq, adapter);
4773 		break;
4774 	case GET_VPD_RSP:
4775 		handle_vpd_rsp(crq, adapter);
4776 		break;
4777 	case QUERY_PHYS_PARMS_RSP:
4778 		adapter->fw_done_rc = handle_query_phys_parms_rsp(crq, adapter);
4779 		complete(&adapter->fw_done);
4780 		break;
4781 	default:
4782 		netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4783 			   gen_crq->cmd);
4784 	}
4785 }
4786 
4787 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4788 {
4789 	struct ibmvnic_adapter *adapter = instance;
4790 
4791 	tasklet_schedule(&adapter->tasklet);
4792 	return IRQ_HANDLED;
4793 }
4794 
4795 static void ibmvnic_tasklet(void *data)
4796 {
4797 	struct ibmvnic_adapter *adapter = data;
4798 	struct ibmvnic_crq_queue *queue = &adapter->crq;
4799 	union ibmvnic_crq *crq;
4800 	unsigned long flags;
4801 	bool done = false;
4802 
4803 	spin_lock_irqsave(&queue->lock, flags);
4804 	while (!done) {
4805 		/* Pull all the valid messages off the CRQ */
4806 		while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4807 			ibmvnic_handle_crq(crq, adapter);
4808 			crq->generic.first = 0;
4809 		}
4810 
4811 		/* remain in tasklet until all
4812 		 * capabilities responses are received
4813 		 */
4814 		if (!adapter->wait_capability)
4815 			done = true;
4816 	}
4817 	/* if capabilities CRQ's were sent in this tasklet, the following
4818 	 * tasklet must wait until all responses are received
4819 	 */
4820 	if (atomic_read(&adapter->running_cap_crqs) != 0)
4821 		adapter->wait_capability = true;
4822 	spin_unlock_irqrestore(&queue->lock, flags);
4823 }
4824 
4825 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4826 {
4827 	struct vio_dev *vdev = adapter->vdev;
4828 	int rc;
4829 
4830 	do {
4831 		rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4832 	} while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4833 
4834 	if (rc)
4835 		dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4836 
4837 	return rc;
4838 }
4839 
4840 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4841 {
4842 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4843 	struct device *dev = &adapter->vdev->dev;
4844 	struct vio_dev *vdev = adapter->vdev;
4845 	int rc;
4846 
4847 	/* Close the CRQ */
4848 	do {
4849 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4850 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4851 
4852 	/* Clean out the queue */
4853 	memset(crq->msgs, 0, PAGE_SIZE);
4854 	crq->cur = 0;
4855 	crq->active = false;
4856 
4857 	/* And re-open it again */
4858 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4859 				crq->msg_token, PAGE_SIZE);
4860 
4861 	if (rc == H_CLOSED)
4862 		/* Adapter is good, but other end is not ready */
4863 		dev_warn(dev, "Partner adapter not ready\n");
4864 	else if (rc != 0)
4865 		dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4866 
4867 	return rc;
4868 }
4869 
4870 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4871 {
4872 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4873 	struct vio_dev *vdev = adapter->vdev;
4874 	long rc;
4875 
4876 	if (!crq->msgs)
4877 		return;
4878 
4879 	netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4880 	free_irq(vdev->irq, adapter);
4881 	tasklet_kill(&adapter->tasklet);
4882 	do {
4883 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4884 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4885 
4886 	dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4887 			 DMA_BIDIRECTIONAL);
4888 	free_page((unsigned long)crq->msgs);
4889 	crq->msgs = NULL;
4890 	crq->active = false;
4891 }
4892 
4893 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4894 {
4895 	struct ibmvnic_crq_queue *crq = &adapter->crq;
4896 	struct device *dev = &adapter->vdev->dev;
4897 	struct vio_dev *vdev = adapter->vdev;
4898 	int rc, retrc = -ENOMEM;
4899 
4900 	if (crq->msgs)
4901 		return 0;
4902 
4903 	crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4904 	/* Should we allocate more than one page? */
4905 
4906 	if (!crq->msgs)
4907 		return -ENOMEM;
4908 
4909 	crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4910 	crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4911 					DMA_BIDIRECTIONAL);
4912 	if (dma_mapping_error(dev, crq->msg_token))
4913 		goto map_failed;
4914 
4915 	rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4916 				crq->msg_token, PAGE_SIZE);
4917 
4918 	if (rc == H_RESOURCE)
4919 		/* maybe kexecing and resource is busy. try a reset */
4920 		rc = ibmvnic_reset_crq(adapter);
4921 	retrc = rc;
4922 
4923 	if (rc == H_CLOSED) {
4924 		dev_warn(dev, "Partner adapter not ready\n");
4925 	} else if (rc) {
4926 		dev_warn(dev, "Error %d opening adapter\n", rc);
4927 		goto reg_crq_failed;
4928 	}
4929 
4930 	retrc = 0;
4931 
4932 	tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4933 		     (unsigned long)adapter);
4934 
4935 	netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4936 	snprintf(crq->name, sizeof(crq->name), "ibmvnic-%x",
4937 		 adapter->vdev->unit_address);
4938 	rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, crq->name, adapter);
4939 	if (rc) {
4940 		dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4941 			vdev->irq, rc);
4942 		goto req_irq_failed;
4943 	}
4944 
4945 	rc = vio_enable_interrupts(vdev);
4946 	if (rc) {
4947 		dev_err(dev, "Error %d enabling interrupts\n", rc);
4948 		goto req_irq_failed;
4949 	}
4950 
4951 	crq->cur = 0;
4952 	spin_lock_init(&crq->lock);
4953 
4954 	return retrc;
4955 
4956 req_irq_failed:
4957 	tasklet_kill(&adapter->tasklet);
4958 	do {
4959 		rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4960 	} while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4961 reg_crq_failed:
4962 	dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4963 map_failed:
4964 	free_page((unsigned long)crq->msgs);
4965 	crq->msgs = NULL;
4966 	return retrc;
4967 }
4968 
4969 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4970 {
4971 	struct device *dev = &adapter->vdev->dev;
4972 	unsigned long timeout = msecs_to_jiffies(30000);
4973 	u64 old_num_rx_queues, old_num_tx_queues;
4974 	int rc;
4975 
4976 	adapter->from_passive_init = false;
4977 
4978 	old_num_rx_queues = adapter->req_rx_queues;
4979 	old_num_tx_queues = adapter->req_tx_queues;
4980 
4981 	reinit_completion(&adapter->init_done);
4982 	adapter->init_done_rc = 0;
4983 	ibmvnic_send_crq_init(adapter);
4984 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4985 		dev_err(dev, "Initialization sequence timed out\n");
4986 		return -1;
4987 	}
4988 
4989 	if (adapter->init_done_rc) {
4990 		release_crq_queue(adapter);
4991 		return adapter->init_done_rc;
4992 	}
4993 
4994 	if (adapter->from_passive_init) {
4995 		adapter->state = VNIC_OPEN;
4996 		adapter->from_passive_init = false;
4997 		return -1;
4998 	}
4999 
5000 	if (test_bit(0, &adapter->resetting) && !adapter->wait_for_reset &&
5001 	    adapter->reset_reason != VNIC_RESET_MOBILITY) {
5002 		if (adapter->req_rx_queues != old_num_rx_queues ||
5003 		    adapter->req_tx_queues != old_num_tx_queues) {
5004 			release_sub_crqs(adapter, 0);
5005 			rc = init_sub_crqs(adapter);
5006 		} else {
5007 			rc = reset_sub_crq_queues(adapter);
5008 		}
5009 	} else {
5010 		rc = init_sub_crqs(adapter);
5011 	}
5012 
5013 	if (rc) {
5014 		dev_err(dev, "Initialization of sub crqs failed\n");
5015 		release_crq_queue(adapter);
5016 		return rc;
5017 	}
5018 
5019 	rc = init_sub_crq_irqs(adapter);
5020 	if (rc) {
5021 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5022 		release_crq_queue(adapter);
5023 	}
5024 
5025 	return rc;
5026 }
5027 
5028 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
5029 {
5030 	struct device *dev = &adapter->vdev->dev;
5031 	unsigned long timeout = msecs_to_jiffies(30000);
5032 	int rc;
5033 
5034 	adapter->from_passive_init = false;
5035 
5036 	adapter->init_done_rc = 0;
5037 	ibmvnic_send_crq_init(adapter);
5038 	if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
5039 		dev_err(dev, "Initialization sequence timed out\n");
5040 		return -1;
5041 	}
5042 
5043 	if (adapter->init_done_rc) {
5044 		release_crq_queue(adapter);
5045 		return adapter->init_done_rc;
5046 	}
5047 
5048 	if (adapter->from_passive_init) {
5049 		adapter->state = VNIC_OPEN;
5050 		adapter->from_passive_init = false;
5051 		return -1;
5052 	}
5053 
5054 	rc = init_sub_crqs(adapter);
5055 	if (rc) {
5056 		dev_err(dev, "Initialization of sub crqs failed\n");
5057 		release_crq_queue(adapter);
5058 		return rc;
5059 	}
5060 
5061 	rc = init_sub_crq_irqs(adapter);
5062 	if (rc) {
5063 		dev_err(dev, "Failed to initialize sub crq irqs\n");
5064 		release_crq_queue(adapter);
5065 	}
5066 
5067 	return rc;
5068 }
5069 
5070 static struct device_attribute dev_attr_failover;
5071 
5072 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
5073 {
5074 	struct ibmvnic_adapter *adapter;
5075 	struct net_device *netdev;
5076 	unsigned char *mac_addr_p;
5077 	int rc;
5078 
5079 	dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
5080 		dev->unit_address);
5081 
5082 	mac_addr_p = (unsigned char *)vio_get_attribute(dev,
5083 							VETH_MAC_ADDR, NULL);
5084 	if (!mac_addr_p) {
5085 		dev_err(&dev->dev,
5086 			"(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
5087 			__FILE__, __LINE__);
5088 		return 0;
5089 	}
5090 
5091 	netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
5092 				   IBMVNIC_MAX_QUEUES);
5093 	if (!netdev)
5094 		return -ENOMEM;
5095 
5096 	adapter = netdev_priv(netdev);
5097 	adapter->state = VNIC_PROBING;
5098 	dev_set_drvdata(&dev->dev, netdev);
5099 	adapter->vdev = dev;
5100 	adapter->netdev = netdev;
5101 
5102 	ether_addr_copy(adapter->mac_addr, mac_addr_p);
5103 	ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
5104 	netdev->irq = dev->irq;
5105 	netdev->netdev_ops = &ibmvnic_netdev_ops;
5106 	netdev->ethtool_ops = &ibmvnic_ethtool_ops;
5107 	SET_NETDEV_DEV(netdev, &dev->dev);
5108 
5109 	spin_lock_init(&adapter->stats_lock);
5110 
5111 	INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
5112 	INIT_DELAYED_WORK(&adapter->ibmvnic_delayed_reset,
5113 			  __ibmvnic_delayed_reset);
5114 	INIT_LIST_HEAD(&adapter->rwi_list);
5115 	spin_lock_init(&adapter->rwi_lock);
5116 	spin_lock_init(&adapter->state_lock);
5117 	mutex_init(&adapter->fw_lock);
5118 	init_completion(&adapter->init_done);
5119 	init_completion(&adapter->fw_done);
5120 	init_completion(&adapter->reset_done);
5121 	init_completion(&adapter->stats_done);
5122 	clear_bit(0, &adapter->resetting);
5123 
5124 	do {
5125 		rc = init_crq_queue(adapter);
5126 		if (rc) {
5127 			dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
5128 				rc);
5129 			goto ibmvnic_init_fail;
5130 		}
5131 
5132 		rc = ibmvnic_init(adapter);
5133 		if (rc && rc != EAGAIN)
5134 			goto ibmvnic_init_fail;
5135 	} while (rc == EAGAIN);
5136 
5137 	rc = init_stats_buffers(adapter);
5138 	if (rc)
5139 		goto ibmvnic_init_fail;
5140 
5141 	rc = init_stats_token(adapter);
5142 	if (rc)
5143 		goto ibmvnic_stats_fail;
5144 
5145 	netdev->mtu = adapter->req_mtu - ETH_HLEN;
5146 	netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
5147 	netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
5148 
5149 	rc = device_create_file(&dev->dev, &dev_attr_failover);
5150 	if (rc)
5151 		goto ibmvnic_dev_file_err;
5152 
5153 	netif_carrier_off(netdev);
5154 	rc = register_netdev(netdev);
5155 	if (rc) {
5156 		dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
5157 		goto ibmvnic_register_fail;
5158 	}
5159 	dev_info(&dev->dev, "ibmvnic registered\n");
5160 
5161 	adapter->state = VNIC_PROBED;
5162 
5163 	adapter->wait_for_reset = false;
5164 
5165 	return 0;
5166 
5167 ibmvnic_register_fail:
5168 	device_remove_file(&dev->dev, &dev_attr_failover);
5169 
5170 ibmvnic_dev_file_err:
5171 	release_stats_token(adapter);
5172 
5173 ibmvnic_stats_fail:
5174 	release_stats_buffers(adapter);
5175 
5176 ibmvnic_init_fail:
5177 	release_sub_crqs(adapter, 1);
5178 	release_crq_queue(adapter);
5179 	mutex_destroy(&adapter->fw_lock);
5180 	free_netdev(netdev);
5181 
5182 	return rc;
5183 }
5184 
5185 static int ibmvnic_remove(struct vio_dev *dev)
5186 {
5187 	struct net_device *netdev = dev_get_drvdata(&dev->dev);
5188 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5189 	unsigned long flags;
5190 
5191 	spin_lock_irqsave(&adapter->state_lock, flags);
5192 	if (adapter->state == VNIC_RESETTING) {
5193 		spin_unlock_irqrestore(&adapter->state_lock, flags);
5194 		return -EBUSY;
5195 	}
5196 
5197 	adapter->state = VNIC_REMOVING;
5198 	spin_unlock_irqrestore(&adapter->state_lock, flags);
5199 
5200 	flush_work(&adapter->ibmvnic_reset);
5201 	flush_delayed_work(&adapter->ibmvnic_delayed_reset);
5202 
5203 	rtnl_lock();
5204 	unregister_netdevice(netdev);
5205 
5206 	release_resources(adapter);
5207 	release_sub_crqs(adapter, 1);
5208 	release_crq_queue(adapter);
5209 
5210 	release_stats_token(adapter);
5211 	release_stats_buffers(adapter);
5212 
5213 	adapter->state = VNIC_REMOVED;
5214 
5215 	rtnl_unlock();
5216 	mutex_destroy(&adapter->fw_lock);
5217 	device_remove_file(&dev->dev, &dev_attr_failover);
5218 	free_netdev(netdev);
5219 	dev_set_drvdata(&dev->dev, NULL);
5220 
5221 	return 0;
5222 }
5223 
5224 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
5225 			      const char *buf, size_t count)
5226 {
5227 	struct net_device *netdev = dev_get_drvdata(dev);
5228 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5229 	unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
5230 	__be64 session_token;
5231 	long rc;
5232 
5233 	if (!sysfs_streq(buf, "1"))
5234 		return -EINVAL;
5235 
5236 	rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
5237 			 H_GET_SESSION_TOKEN, 0, 0, 0);
5238 	if (rc) {
5239 		netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
5240 			   rc);
5241 		return -EINVAL;
5242 	}
5243 
5244 	session_token = (__be64)retbuf[0];
5245 	netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
5246 		   be64_to_cpu(session_token));
5247 	rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
5248 				H_SESSION_ERR_DETECTED, session_token, 0, 0);
5249 	if (rc) {
5250 		netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
5251 			   rc);
5252 		return -EINVAL;
5253 	}
5254 
5255 	return count;
5256 }
5257 
5258 static DEVICE_ATTR_WO(failover);
5259 
5260 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
5261 {
5262 	struct net_device *netdev = dev_get_drvdata(&vdev->dev);
5263 	struct ibmvnic_adapter *adapter;
5264 	struct iommu_table *tbl;
5265 	unsigned long ret = 0;
5266 	int i;
5267 
5268 	tbl = get_iommu_table_base(&vdev->dev);
5269 
5270 	/* netdev inits at probe time along with the structures we need below*/
5271 	if (!netdev)
5272 		return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
5273 
5274 	adapter = netdev_priv(netdev);
5275 
5276 	ret += PAGE_SIZE; /* the crq message queue */
5277 	ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
5278 
5279 	for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
5280 		ret += 4 * PAGE_SIZE; /* the scrq message queue */
5281 
5282 	for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
5283 	     i++)
5284 		ret += adapter->rx_pool[i].size *
5285 		    IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
5286 
5287 	return ret;
5288 }
5289 
5290 static int ibmvnic_resume(struct device *dev)
5291 {
5292 	struct net_device *netdev = dev_get_drvdata(dev);
5293 	struct ibmvnic_adapter *adapter = netdev_priv(netdev);
5294 
5295 	if (adapter->state != VNIC_OPEN)
5296 		return 0;
5297 
5298 	tasklet_schedule(&adapter->tasklet);
5299 
5300 	return 0;
5301 }
5302 
5303 static const struct vio_device_id ibmvnic_device_table[] = {
5304 	{"network", "IBM,vnic"},
5305 	{"", "" }
5306 };
5307 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
5308 
5309 static const struct dev_pm_ops ibmvnic_pm_ops = {
5310 	.resume = ibmvnic_resume
5311 };
5312 
5313 static struct vio_driver ibmvnic_driver = {
5314 	.id_table       = ibmvnic_device_table,
5315 	.probe          = ibmvnic_probe,
5316 	.remove         = ibmvnic_remove,
5317 	.get_desired_dma = ibmvnic_get_desired_dma,
5318 	.name		= ibmvnic_driver_name,
5319 	.pm		= &ibmvnic_pm_ops,
5320 };
5321 
5322 /* module functions */
5323 static int __init ibmvnic_module_init(void)
5324 {
5325 	pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
5326 		IBMVNIC_DRIVER_VERSION);
5327 
5328 	return vio_register_driver(&ibmvnic_driver);
5329 }
5330 
5331 static void __exit ibmvnic_module_exit(void)
5332 {
5333 	vio_unregister_driver(&ibmvnic_driver);
5334 }
5335 
5336 module_init(ibmvnic_module_init);
5337 module_exit(ibmvnic_module_exit);
5338