drivers/net/ethernet/ibm/ibmvnic.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /*  IBM System i and System p Virtual NIC Device Driver                   */
   4 /*  Copyright (C) 2014 IBM Corp.                                          */
   5 /*  Santiago Leon (santi_leon@yahoo.com)                                  */
   6 /*  Thomas Falcon (tlfalcon@linux.vnet.ibm.com)                           */
   7 /*  John Allen (jallen@linux.vnet.ibm.com)                                */
   8 /*                                                                        */
   9 /*  This program is free software; you can redistribute it and/or modify  */
  10 /*  it under the terms of the GNU General Public License as published by  */
  11 /*  the Free Software Foundation; either version 2 of the License, or     */
  12 /*  (at your option) any later version.                                   */
  13 /*                                                                        */
  14 /*  This program is distributed in the hope that it will be useful,       */
  15 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  16 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  17 /*  GNU General Public License for more details.                          */
  18 /*                                                                        */
  19 /*  You should have received a copy of the GNU General Public License     */
  20 /*  along with this program.                                              */
  21 /*                                                                        */
  22 /* This module contains the implementation of a virtual ethernet device   */
  23 /* for use with IBM i/p Series LPAR Linux. It utilizes the logical LAN    */
  24 /* option of the RS/6000 Platform Architecture to interface with virtual  */
  25 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  26 /*                                                                         */
  27 /* Messages are passed between the VNIC driver and the VNIC server using  */
  28 /* Command/Response Queues (CRQs) and sub CRQs (sCRQs). CRQs are used to  */
  29 /* issue and receive commands that initiate communication with the server */
  30 /* on driver initialization. Sub CRQs (sCRQs) are similar to CRQs, but    */
  31 /* are used by the driver to notify the server that a packet is           */
  32 /* ready for transmission or that a buffer has been added to receive a    */
  33 /* packet. Subsequently, sCRQs are used by the server to notify the       */
  34 /* driver that a packet transmission has been completed or that a packet  */
  35 /* has been received and placed in a waiting buffer.                      */
  36 /*                                                                        */
  37 /* In lieu of a more conventional "on-the-fly" DMA mapping strategy in    */
  38 /* which skbs are DMA mapped and immediately unmapped when the transmit   */
  39 /* or receive has been completed, the VNIC driver is required to use      */
  40 /* "long term mapping". This entails that large, continuous DMA mapped    */
  41 /* buffers are allocated on driver initialization and these buffers are   */
  42 /* then continuously reused to pass skbs to and from the VNIC server.     */
  43 /*                                                                        */
  44 /**************************************************************************/
  45
  46 #include <linux/module.h>
  47 #include <linux/moduleparam.h>
  48 #include <linux/types.h>
  49 #include <linux/errno.h>
  50 #include <linux/completion.h>
  51 #include <linux/ioport.h>
  52 #include <linux/dma-mapping.h>
  53 #include <linux/kernel.h>
  54 #include <linux/netdevice.h>
  55 #include <linux/etherdevice.h>
  56 #include <linux/skbuff.h>
  57 #include <linux/init.h>
  58 #include <linux/delay.h>
  59 #include <linux/mm.h>
  60 #include <linux/ethtool.h>
  61 #include <linux/proc_fs.h>
  62 #include <linux/if_arp.h>
  63 #include <linux/in.h>
  64 #include <linux/ip.h>
  65 #include <linux/ipv6.h>
  66 #include <linux/irq.h>
  67 #include <linux/kthread.h>
  68 #include <linux/seq_file.h>
  69 #include <linux/interrupt.h>
  70 #include <net/net_namespace.h>
  71 #include <asm/hvcall.h>
  72 #include <linux/atomic.h>
  73 #include <asm/vio.h>
  74 #include <asm/iommu.h>
  75 #include <linux/uaccess.h>
  76 #include <asm/firmware.h>
  77 #include <linux/workqueue.h>
  78 #include <linux/if_vlan.h>
  79 #include <linux/utsname.h>
  80
  81 #include "ibmvnic.h"
  82
  83 static const char ibmvnic_driver_name[] = "ibmvnic";
  84 static const char ibmvnic_driver_string[] = "IBM System i/p Virtual NIC Driver";
  85
  86 MODULE_AUTHOR("Santiago Leon");
  87 MODULE_DESCRIPTION("IBM System i/p Virtual NIC Driver");
  88 MODULE_LICENSE("GPL");
  89 MODULE_VERSION(IBMVNIC_DRIVER_VERSION);
  90
  91 static int ibmvnic_version = IBMVNIC_INITIAL_VERSION;
  92 static int ibmvnic_remove(struct vio_dev *);
  93 static void release_sub_crqs(struct ibmvnic_adapter *, bool);
  94 static int ibmvnic_reset_crq(struct ibmvnic_adapter *);
  95 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *);
  96 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *);
  97 static int ibmvnic_send_crq(struct ibmvnic_adapter *, union ibmvnic_crq *);
  98 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
  99                        union sub_crq *sub_crq);
 100 static int send_subcrq_indirect(struct ibmvnic_adapter *, u64, u64, u64);
 101 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance);
 102 static int enable_scrq_irq(struct ibmvnic_adapter *,
 103                            struct ibmvnic_sub_crq_queue *);
 104 static int disable_scrq_irq(struct ibmvnic_adapter *,
 105                             struct ibmvnic_sub_crq_queue *);
 106 static int pending_scrq(struct ibmvnic_adapter *,
 107                         struct ibmvnic_sub_crq_queue *);
 108 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *,
 109                                         struct ibmvnic_sub_crq_queue *);
 110 static int ibmvnic_poll(struct napi_struct *napi, int data);
 111 static void send_map_query(struct ibmvnic_adapter *adapter);
 112 static int send_request_map(struct ibmvnic_adapter *, dma_addr_t, __be32, u8);
 113 static int send_request_unmap(struct ibmvnic_adapter *, u8);
 114 static int send_login(struct ibmvnic_adapter *adapter);
 115 static void send_cap_queries(struct ibmvnic_adapter *adapter);
 116 static int init_sub_crqs(struct ibmvnic_adapter *);
 117 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter);
 118 static int ibmvnic_init(struct ibmvnic_adapter *);
 119 static int ibmvnic_reset_init(struct ibmvnic_adapter *);
 120 static void release_crq_queue(struct ibmvnic_adapter *);
 121 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p);
 122 static int init_crq_queue(struct ibmvnic_adapter *adapter);
 123
 124 struct ibmvnic_stat {
 125         char name[ETH_GSTRING_LEN];
 126         int offset;
 127 };
 128
 129 #define IBMVNIC_STAT_OFF(stat) (offsetof(struct ibmvnic_adapter, stats) + \
 130                              offsetof(struct ibmvnic_statistics, stat))
 131 #define IBMVNIC_GET_STAT(a, off) (*((u64 *)(((unsigned long)(a)) + off)))
 132
 133 static const struct ibmvnic_stat ibmvnic_stats[] = {
 134         {"rx_packets", IBMVNIC_STAT_OFF(rx_packets)},
 135         {"rx_bytes", IBMVNIC_STAT_OFF(rx_bytes)},
 136         {"tx_packets", IBMVNIC_STAT_OFF(tx_packets)},
 137         {"tx_bytes", IBMVNIC_STAT_OFF(tx_bytes)},
 138         {"ucast_tx_packets", IBMVNIC_STAT_OFF(ucast_tx_packets)},
 139         {"ucast_rx_packets", IBMVNIC_STAT_OFF(ucast_rx_packets)},
 140         {"mcast_tx_packets", IBMVNIC_STAT_OFF(mcast_tx_packets)},
 141         {"mcast_rx_packets", IBMVNIC_STAT_OFF(mcast_rx_packets)},
 142         {"bcast_tx_packets", IBMVNIC_STAT_OFF(bcast_tx_packets)},
 143         {"bcast_rx_packets", IBMVNIC_STAT_OFF(bcast_rx_packets)},
 144         {"align_errors", IBMVNIC_STAT_OFF(align_errors)},
 145         {"fcs_errors", IBMVNIC_STAT_OFF(fcs_errors)},
 146         {"single_collision_frames", IBMVNIC_STAT_OFF(single_collision_frames)},
 147         {"multi_collision_frames", IBMVNIC_STAT_OFF(multi_collision_frames)},
 148         {"sqe_test_errors", IBMVNIC_STAT_OFF(sqe_test_errors)},
 149         {"deferred_tx", IBMVNIC_STAT_OFF(deferred_tx)},
 150         {"late_collisions", IBMVNIC_STAT_OFF(late_collisions)},
 151         {"excess_collisions", IBMVNIC_STAT_OFF(excess_collisions)},
 152         {"internal_mac_tx_errors", IBMVNIC_STAT_OFF(internal_mac_tx_errors)},
 153         {"carrier_sense", IBMVNIC_STAT_OFF(carrier_sense)},
 154         {"too_long_frames", IBMVNIC_STAT_OFF(too_long_frames)},
 155         {"internal_mac_rx_errors", IBMVNIC_STAT_OFF(internal_mac_rx_errors)},
 156 };
 157
 158 static long h_reg_sub_crq(unsigned long unit_address, unsigned long token,
 159                           unsigned long length, unsigned long *number,
 160                           unsigned long *irq)
 161 {
 162         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
 163         long rc;
 164
 165         rc = plpar_hcall(H_REG_SUB_CRQ, retbuf, unit_address, token, length);
 166         *number = retbuf[0];
 167         *irq = retbuf[1];
 168
 169         return rc;
 170 }
 171
 172 static int alloc_long_term_buff(struct ibmvnic_adapter *adapter,
 173                                 struct ibmvnic_long_term_buff *ltb, int size)
 174 {
 175         struct device *dev = &adapter->vdev->dev;
 176         int rc;
 177
 178         ltb->size = size;
 179         ltb->buff = dma_alloc_coherent(dev, ltb->size, &ltb->addr,
 180                                        GFP_KERNEL);
 181
 182         if (!ltb->buff) {
 183                 dev_err(dev, "Couldn't alloc long term buffer\n");
 184                 return -ENOMEM;
 185         }
 186         ltb->map_id = adapter->map_id;
 187         adapter->map_id++;
 188
 189         init_completion(&adapter->fw_done);
 190         rc = send_request_map(adapter, ltb->addr,
 191                               ltb->size, ltb->map_id);
 192         if (rc) {
 193                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 194                 return rc;
 195         }
 196         wait_for_completion(&adapter->fw_done);
 197
 198         if (adapter->fw_done_rc) {
 199                 dev_err(dev, "Couldn't map long term buffer,rc = %d\n",
 200                         adapter->fw_done_rc);
 201                 dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 202                 return -1;
 203         }
 204         return 0;
 205 }
 206
 207 static void free_long_term_buff(struct ibmvnic_adapter *adapter,
 208                                 struct ibmvnic_long_term_buff *ltb)
 209 {
 210         struct device *dev = &adapter->vdev->dev;
 211
 212         if (!ltb->buff)
 213                 return;
 214
 215         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
 216             adapter->reset_reason != VNIC_RESET_MOBILITY)
 217                 send_request_unmap(adapter, ltb->map_id);
 218         dma_free_coherent(dev, ltb->size, ltb->buff, ltb->addr);
 219 }
 220
 221 static int reset_long_term_buff(struct ibmvnic_adapter *adapter,
 222                                 struct ibmvnic_long_term_buff *ltb)
 223 {
 224         int rc;
 225
 226         memset(ltb->buff, 0, ltb->size);
 227
 228         init_completion(&adapter->fw_done);
 229         rc = send_request_map(adapter, ltb->addr, ltb->size, ltb->map_id);
 230         if (rc)
 231                 return rc;
 232         wait_for_completion(&adapter->fw_done);
 233
 234         if (adapter->fw_done_rc) {
 235                 dev_info(&adapter->vdev->dev,
 236                          "Reset failed, attempting to free and reallocate buffer\n");
 237                 free_long_term_buff(adapter, ltb);
 238                 return alloc_long_term_buff(adapter, ltb, ltb->size);
 239         }
 240         return 0;
 241 }
 242
 243 static void deactivate_rx_pools(struct ibmvnic_adapter *adapter)
 244 {
 245         int i;
 246
 247         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 248              i++)
 249                 adapter->rx_pool[i].active = 0;
 250 }
 251
 252 static void replenish_rx_pool(struct ibmvnic_adapter *adapter,
 253                               struct ibmvnic_rx_pool *pool)
 254 {
 255         int count = pool->size - atomic_read(&pool->available);
 256         struct device *dev = &adapter->vdev->dev;
 257         int buffers_added = 0;
 258         unsigned long lpar_rc;
 259         union sub_crq sub_crq;
 260         struct sk_buff *skb;
 261         unsigned int offset;
 262         dma_addr_t dma_addr;
 263         unsigned char *dst;
 264         u64 *handle_array;
 265         int shift = 0;
 266         int index;
 267         int i;
 268
 269         if (!pool->active)
 270                 return;
 271
 272         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 273                                       be32_to_cpu(adapter->login_rsp_buf->
 274                                       off_rxadd_subcrqs));
 275
 276         for (i = 0; i < count; ++i) {
 277                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 278                 if (!skb) {
 279                         dev_err(dev, "Couldn't replenish rx buff\n");
 280                         adapter->replenish_no_mem++;
 281                         break;
 282                 }
 283
 284                 index = pool->free_map[pool->next_free];
 285
 286                 if (pool->rx_buff[index].skb)
 287                         dev_err(dev, "Inconsistent free_map!\n");
 288
 289                 /* Copy the skb to the long term mapped DMA buffer */
 290                 offset = index * pool->buff_size;
 291                 dst = pool->long_term_buff.buff + offset;
 292                 memset(dst, 0, pool->buff_size);
 293                 dma_addr = pool->long_term_buff.addr + offset;
 294                 pool->rx_buff[index].data = dst;
 295
 296                 pool->free_map[pool->next_free] = IBMVNIC_INVALID_MAP;
 297                 pool->rx_buff[index].dma = dma_addr;
 298                 pool->rx_buff[index].skb = skb;
 299                 pool->rx_buff[index].pool_index = pool->index;
 300                 pool->rx_buff[index].size = pool->buff_size;
 301
 302                 memset(&sub_crq, 0, sizeof(sub_crq));
 303                 sub_crq.rx_add.first = IBMVNIC_CRQ_CMD;
 304                 sub_crq.rx_add.correlator =
 305                     cpu_to_be64((u64)&pool->rx_buff[index]);
 306                 sub_crq.rx_add.ioba = cpu_to_be32(dma_addr);
 307                 sub_crq.rx_add.map_id = pool->long_term_buff.map_id;
 308
 309                 /* The length field of the sCRQ is defined to be 24 bits so the
 310                  * buffer size needs to be left shifted by a byte before it is
 311                  * converted to big endian to prevent the last byte from being
 312                  * truncated.
 313                  */
 314 #ifdef __LITTLE_ENDIAN__
 315                 shift = 8;
 316 #endif
 317                 sub_crq.rx_add.len = cpu_to_be32(pool->buff_size << shift);
 318
 319                 lpar_rc = send_subcrq(adapter, handle_array[pool->index],
 320                                       &sub_crq);
 321                 if (lpar_rc != H_SUCCESS)
 322                         goto failure;
 323
 324                 buffers_added++;
 325                 adapter->replenish_add_buff_success++;
 326                 pool->next_free = (pool->next_free + 1) % pool->size;
 327         }
 328         atomic_add(buffers_added, &pool->available);
 329         return;
 330
 331 failure:
 332         if (lpar_rc != H_PARAMETER && lpar_rc != H_CLOSED)
 333                 dev_err_ratelimited(dev, "rx: replenish packet buffer failed\n");
 334         pool->free_map[pool->next_free] = index;
 335         pool->rx_buff[index].skb = NULL;
 336
 337         dev_kfree_skb_any(skb);
 338         adapter->replenish_add_buff_failure++;
 339         atomic_add(buffers_added, &pool->available);
 340
 341         if (lpar_rc == H_CLOSED || adapter->failover_pending) {
 342                 /* Disable buffer pool replenishment and report carrier off if
 343                  * queue is closed or pending failover.
 344                  * Firmware guarantees that a signal will be sent to the
 345                  * driver, triggering a reset.
 346                  */
 347                 deactivate_rx_pools(adapter);
 348                 netif_carrier_off(adapter->netdev);
 349         }
 350 }
 351
 352 static void replenish_pools(struct ibmvnic_adapter *adapter)
 353 {
 354         int i;
 355
 356         adapter->replenish_task_cycles++;
 357         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 358              i++) {
 359                 if (adapter->rx_pool[i].active)
 360                         replenish_rx_pool(adapter, &adapter->rx_pool[i]);
 361         }
 362 }
 363
 364 static void release_stats_buffers(struct ibmvnic_adapter *adapter)
 365 {
 366         kfree(adapter->tx_stats_buffers);
 367         kfree(adapter->rx_stats_buffers);
 368         adapter->tx_stats_buffers = NULL;
 369         adapter->rx_stats_buffers = NULL;
 370 }
 371
 372 static int init_stats_buffers(struct ibmvnic_adapter *adapter)
 373 {
 374         adapter->tx_stats_buffers =
 375                                 kcalloc(IBMVNIC_MAX_QUEUES,
 376                                         sizeof(struct ibmvnic_tx_queue_stats),
 377                                         GFP_KERNEL);
 378         if (!adapter->tx_stats_buffers)
 379                 return -ENOMEM;
 380
 381         adapter->rx_stats_buffers =
 382                                 kcalloc(IBMVNIC_MAX_QUEUES,
 383                                         sizeof(struct ibmvnic_rx_queue_stats),
 384                                         GFP_KERNEL);
 385         if (!adapter->rx_stats_buffers)
 386                 return -ENOMEM;
 387
 388         return 0;
 389 }
 390
 391 static void release_stats_token(struct ibmvnic_adapter *adapter)
 392 {
 393         struct device *dev = &adapter->vdev->dev;
 394
 395         if (!adapter->stats_token)
 396                 return;
 397
 398         dma_unmap_single(dev, adapter->stats_token,
 399                          sizeof(struct ibmvnic_statistics),
 400                          DMA_FROM_DEVICE);
 401         adapter->stats_token = 0;
 402 }
 403
 404 static int init_stats_token(struct ibmvnic_adapter *adapter)
 405 {
 406         struct device *dev = &adapter->vdev->dev;
 407         dma_addr_t stok;
 408
 409         stok = dma_map_single(dev, &adapter->stats,
 410                               sizeof(struct ibmvnic_statistics),
 411                               DMA_FROM_DEVICE);
 412         if (dma_mapping_error(dev, stok)) {
 413                 dev_err(dev, "Couldn't map stats buffer\n");
 414                 return -1;
 415         }
 416
 417         adapter->stats_token = stok;
 418         netdev_dbg(adapter->netdev, "Stats token initialized (%llx)\n", stok);
 419         return 0;
 420 }
 421
 422 static int reset_rx_pools(struct ibmvnic_adapter *adapter)
 423 {
 424         struct ibmvnic_rx_pool *rx_pool;
 425         int rx_scrqs;
 426         int i, j, rc;
 427         u64 *size_array;
 428
 429         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 430                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 431
 432         rx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 433         for (i = 0; i < rx_scrqs; i++) {
 434                 rx_pool = &adapter->rx_pool[i];
 435
 436                 netdev_dbg(adapter->netdev, "Re-setting rx_pool[%d]\n", i);
 437
 438                 if (rx_pool->buff_size != be64_to_cpu(size_array[i])) {
 439                         free_long_term_buff(adapter, &rx_pool->long_term_buff);
 440                         rx_pool->buff_size = be64_to_cpu(size_array[i]);
 441                         alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 442                                              rx_pool->size *
 443                                              rx_pool->buff_size);
 444                 } else {
 445                         rc = reset_long_term_buff(adapter,
 446                                                   &rx_pool->long_term_buff);
 447                 }
 448
 449                 if (rc)
 450                         return rc;
 451
 452                 for (j = 0; j < rx_pool->size; j++)
 453                         rx_pool->free_map[j] = j;
 454
 455                 memset(rx_pool->rx_buff, 0,
 456                        rx_pool->size * sizeof(struct ibmvnic_rx_buff));
 457
 458                 atomic_set(&rx_pool->available, 0);
 459                 rx_pool->next_alloc = 0;
 460                 rx_pool->next_free = 0;
 461                 rx_pool->active = 1;
 462         }
 463
 464         return 0;
 465 }
 466
 467 static void release_rx_pools(struct ibmvnic_adapter *adapter)
 468 {
 469         struct ibmvnic_rx_pool *rx_pool;
 470         int i, j;
 471
 472         if (!adapter->rx_pool)
 473                 return;
 474
 475         for (i = 0; i < adapter->num_active_rx_pools; i++) {
 476                 rx_pool = &adapter->rx_pool[i];
 477
 478                 netdev_dbg(adapter->netdev, "Releasing rx_pool[%d]\n", i);
 479
 480                 kfree(rx_pool->free_map);
 481                 free_long_term_buff(adapter, &rx_pool->long_term_buff);
 482
 483                 if (!rx_pool->rx_buff)
 484                         continue;
 485
 486                 for (j = 0; j < rx_pool->size; j++) {
 487                         if (rx_pool->rx_buff[j].skb) {
 488                                 dev_kfree_skb_any(rx_pool->rx_buff[i].skb);
 489                                 rx_pool->rx_buff[i].skb = NULL;
 490                         }
 491                 }
 492
 493                 kfree(rx_pool->rx_buff);
 494         }
 495
 496         kfree(adapter->rx_pool);
 497         adapter->rx_pool = NULL;
 498         adapter->num_active_rx_pools = 0;
 499 }
 500
 501 static int init_rx_pools(struct net_device *netdev)
 502 {
 503         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 504         struct device *dev = &adapter->vdev->dev;
 505         struct ibmvnic_rx_pool *rx_pool;
 506         int rxadd_subcrqs;
 507         u64 *size_array;
 508         int i, j;
 509
 510         rxadd_subcrqs =
 511                 be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
 512         size_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
 513                 be32_to_cpu(adapter->login_rsp_buf->off_rxadd_buff_size));
 514
 515         adapter->rx_pool = kcalloc(rxadd_subcrqs,
 516                                    sizeof(struct ibmvnic_rx_pool),
 517                                    GFP_KERNEL);
 518         if (!adapter->rx_pool) {
 519                 dev_err(dev, "Failed to allocate rx pools\n");
 520                 return -1;
 521         }
 522
 523         adapter->num_active_rx_pools = rxadd_subcrqs;
 524
 525         for (i = 0; i < rxadd_subcrqs; i++) {
 526                 rx_pool = &adapter->rx_pool[i];
 527
 528                 netdev_dbg(adapter->netdev,
 529                            "Initializing rx_pool[%d], %lld buffs, %lld bytes each\n",
 530                            i, adapter->req_rx_add_entries_per_subcrq,
 531                            be64_to_cpu(size_array[i]));
 532
 533                 rx_pool->size = adapter->req_rx_add_entries_per_subcrq;
 534                 rx_pool->index = i;
 535                 rx_pool->buff_size = be64_to_cpu(size_array[i]);
 536                 rx_pool->active = 1;
 537
 538                 rx_pool->free_map = kcalloc(rx_pool->size, sizeof(int),
 539                                             GFP_KERNEL);
 540                 if (!rx_pool->free_map) {
 541                         release_rx_pools(adapter);
 542                         return -1;
 543                 }
 544
 545                 rx_pool->rx_buff = kcalloc(rx_pool->size,
 546                                            sizeof(struct ibmvnic_rx_buff),
 547                                            GFP_KERNEL);
 548                 if (!rx_pool->rx_buff) {
 549                         dev_err(dev, "Couldn't alloc rx buffers\n");
 550                         release_rx_pools(adapter);
 551                         return -1;
 552                 }
 553
 554                 if (alloc_long_term_buff(adapter, &rx_pool->long_term_buff,
 555                                          rx_pool->size * rx_pool->buff_size)) {
 556                         release_rx_pools(adapter);
 557                         return -1;
 558                 }
 559
 560                 for (j = 0; j < rx_pool->size; ++j)
 561                         rx_pool->free_map[j] = j;
 562
 563                 atomic_set(&rx_pool->available, 0);
 564                 rx_pool->next_alloc = 0;
 565                 rx_pool->next_free = 0;
 566         }
 567
 568         return 0;
 569 }
 570
 571 static int reset_one_tx_pool(struct ibmvnic_adapter *adapter,
 572                              struct ibmvnic_tx_pool *tx_pool)
 573 {
 574         int rc, i;
 575
 576         rc = reset_long_term_buff(adapter, &tx_pool->long_term_buff);
 577         if (rc)
 578                 return rc;
 579
 580         memset(tx_pool->tx_buff, 0,
 581                tx_pool->num_buffers *
 582                sizeof(struct ibmvnic_tx_buff));
 583
 584         for (i = 0; i < tx_pool->num_buffers; i++)
 585                 tx_pool->free_map[i] = i;
 586
 587         tx_pool->consumer_index = 0;
 588         tx_pool->producer_index = 0;
 589
 590         return 0;
 591 }
 592
 593 static int reset_tx_pools(struct ibmvnic_adapter *adapter)
 594 {
 595         int tx_scrqs;
 596         int i, rc;
 597
 598         tx_scrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 599         for (i = 0; i < tx_scrqs; i++) {
 600                 rc = reset_one_tx_pool(adapter, &adapter->tso_pool[i]);
 601                 if (rc)
 602                         return rc;
 603                 rc = reset_one_tx_pool(adapter, &adapter->tx_pool[i]);
 604                 if (rc)
 605                         return rc;
 606         }
 607
 608         return 0;
 609 }
 610
 611 static void release_vpd_data(struct ibmvnic_adapter *adapter)
 612 {
 613         if (!adapter->vpd)
 614                 return;
 615
 616         kfree(adapter->vpd->buff);
 617         kfree(adapter->vpd);
 618
 619         adapter->vpd = NULL;
 620 }
 621
 622 static void release_one_tx_pool(struct ibmvnic_adapter *adapter,
 623                                 struct ibmvnic_tx_pool *tx_pool)
 624 {
 625         kfree(tx_pool->tx_buff);
 626         kfree(tx_pool->free_map);
 627         free_long_term_buff(adapter, &tx_pool->long_term_buff);
 628 }
 629
 630 static void release_tx_pools(struct ibmvnic_adapter *adapter)
 631 {
 632         int i;
 633
 634         if (!adapter->tx_pool)
 635                 return;
 636
 637         for (i = 0; i < adapter->num_active_tx_pools; i++) {
 638                 release_one_tx_pool(adapter, &adapter->tx_pool[i]);
 639                 release_one_tx_pool(adapter, &adapter->tso_pool[i]);
 640         }
 641
 642         kfree(adapter->tx_pool);
 643         adapter->tx_pool = NULL;
 644         kfree(adapter->tso_pool);
 645         adapter->tso_pool = NULL;
 646         adapter->num_active_tx_pools = 0;
 647 }
 648
 649 static int init_one_tx_pool(struct net_device *netdev,
 650                             struct ibmvnic_tx_pool *tx_pool,
 651                             int num_entries, int buf_size)
 652 {
 653         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 654         int i;
 655
 656         tx_pool->tx_buff = kcalloc(num_entries,
 657                                    sizeof(struct ibmvnic_tx_buff),
 658                                    GFP_KERNEL);
 659         if (!tx_pool->tx_buff)
 660                 return -1;
 661
 662         if (alloc_long_term_buff(adapter, &tx_pool->long_term_buff,
 663                                  num_entries * buf_size))
 664                 return -1;
 665
 666         tx_pool->free_map = kcalloc(num_entries, sizeof(int), GFP_KERNEL);
 667         if (!tx_pool->free_map)
 668                 return -1;
 669
 670         for (i = 0; i < num_entries; i++)
 671                 tx_pool->free_map[i] = i;
 672
 673         tx_pool->consumer_index = 0;
 674         tx_pool->producer_index = 0;
 675         tx_pool->num_buffers = num_entries;
 676         tx_pool->buf_size = buf_size;
 677
 678         return 0;
 679 }
 680
 681 static int init_tx_pools(struct net_device *netdev)
 682 {
 683         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 684         int tx_subcrqs;
 685         int i, rc;
 686
 687         tx_subcrqs = be32_to_cpu(adapter->login_rsp_buf->num_txsubm_subcrqs);
 688         adapter->tx_pool = kcalloc(tx_subcrqs,
 689                                    sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 690         if (!adapter->tx_pool)
 691                 return -1;
 692
 693         adapter->tso_pool = kcalloc(tx_subcrqs,
 694                                     sizeof(struct ibmvnic_tx_pool), GFP_KERNEL);
 695         if (!adapter->tso_pool)
 696                 return -1;
 697
 698         adapter->num_active_tx_pools = tx_subcrqs;
 699
 700         for (i = 0; i < tx_subcrqs; i++) {
 701                 rc = init_one_tx_pool(netdev, &adapter->tx_pool[i],
 702                                       adapter->req_tx_entries_per_subcrq,
 703                                       adapter->req_mtu + VLAN_HLEN);
 704                 if (rc) {
 705                         release_tx_pools(adapter);
 706                         return rc;
 707                 }
 708
 709                 init_one_tx_pool(netdev, &adapter->tso_pool[i],
 710                                  IBMVNIC_TSO_BUFS,
 711                                  IBMVNIC_TSO_BUF_SZ);
 712                 if (rc) {
 713                         release_tx_pools(adapter);
 714                         return rc;
 715                 }
 716         }
 717
 718         return 0;
 719 }
 720
 721 static void ibmvnic_napi_enable(struct ibmvnic_adapter *adapter)
 722 {
 723         int i;
 724
 725         if (adapter->napi_enabled)
 726                 return;
 727
 728         for (i = 0; i < adapter->req_rx_queues; i++)
 729                 napi_enable(&adapter->napi[i]);
 730
 731         adapter->napi_enabled = true;
 732 }
 733
 734 static void ibmvnic_napi_disable(struct ibmvnic_adapter *adapter)
 735 {
 736         int i;
 737
 738         if (!adapter->napi_enabled)
 739                 return;
 740
 741         for (i = 0; i < adapter->req_rx_queues; i++) {
 742                 netdev_dbg(adapter->netdev, "Disabling napi[%d]\n", i);
 743                 napi_disable(&adapter->napi[i]);
 744         }
 745
 746         adapter->napi_enabled = false;
 747 }
 748
 749 static int init_napi(struct ibmvnic_adapter *adapter)
 750 {
 751         int i;
 752
 753         adapter->napi = kcalloc(adapter->req_rx_queues,
 754                                 sizeof(struct napi_struct), GFP_KERNEL);
 755         if (!adapter->napi)
 756                 return -ENOMEM;
 757
 758         for (i = 0; i < adapter->req_rx_queues; i++) {
 759                 netdev_dbg(adapter->netdev, "Adding napi[%d]\n", i);
 760                 netif_napi_add(adapter->netdev, &adapter->napi[i],
 761                                ibmvnic_poll, NAPI_POLL_WEIGHT);
 762         }
 763
 764         adapter->num_active_rx_napi = adapter->req_rx_queues;
 765         return 0;
 766 }
 767
 768 static void release_napi(struct ibmvnic_adapter *adapter)
 769 {
 770         int i;
 771
 772         if (!adapter->napi)
 773                 return;
 774
 775         for (i = 0; i < adapter->num_active_rx_napi; i++) {
 776                 if (&adapter->napi[i]) {
 777                         netdev_dbg(adapter->netdev,
 778                                    "Releasing napi[%d]\n", i);
 779                         netif_napi_del(&adapter->napi[i]);
 780                 }
 781         }
 782
 783         kfree(adapter->napi);
 784         adapter->napi = NULL;
 785         adapter->num_active_rx_napi = 0;
 786         adapter->napi_enabled = false;
 787 }
 788
 789 static int ibmvnic_login(struct net_device *netdev)
 790 {
 791         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 792         unsigned long timeout = msecs_to_jiffies(30000);
 793         int retry_count = 0;
 794         bool retry;
 795         int rc;
 796
 797         do {
 798                 retry = false;
 799                 if (retry_count > IBMVNIC_MAX_QUEUES) {
 800                         netdev_warn(netdev, "Login attempts exceeded\n");
 801                         return -1;
 802                 }
 803
 804                 adapter->init_done_rc = 0;
 805                 reinit_completion(&adapter->init_done);
 806                 rc = send_login(adapter);
 807                 if (rc) {
 808                         netdev_warn(netdev, "Unable to login\n");
 809                         return rc;
 810                 }
 811
 812                 if (!wait_for_completion_timeout(&adapter->init_done,
 813                                                  timeout)) {
 814                         netdev_warn(netdev, "Login timed out\n");
 815                         return -1;
 816                 }
 817
 818                 if (adapter->init_done_rc == PARTIALSUCCESS) {
 819                         retry_count++;
 820                         release_sub_crqs(adapter, 1);
 821
 822                         retry = true;
 823                         netdev_dbg(netdev,
 824                                    "Received partial success, retrying...\n");
 825                         adapter->init_done_rc = 0;
 826                         reinit_completion(&adapter->init_done);
 827                         send_cap_queries(adapter);
 828                         if (!wait_for_completion_timeout(&adapter->init_done,
 829                                                          timeout)) {
 830                                 netdev_warn(netdev,
 831                                             "Capabilities query timed out\n");
 832                                 return -1;
 833                         }
 834
 835                         rc = init_sub_crqs(adapter);
 836                         if (rc) {
 837                                 netdev_warn(netdev,
 838                                             "SCRQ initialization failed\n");
 839                                 return -1;
 840                         }
 841
 842                         rc = init_sub_crq_irqs(adapter);
 843                         if (rc) {
 844                                 netdev_warn(netdev,
 845                                             "SCRQ irq initialization failed\n");
 846                                 return -1;
 847                         }
 848                 } else if (adapter->init_done_rc) {
 849                         netdev_warn(netdev, "Adapter login failed\n");
 850                         return -1;
 851                 }
 852         } while (retry);
 853
 854         /* handle pending MAC address changes after successful login */
 855         if (adapter->mac_change_pending) {
 856                 __ibmvnic_set_mac(netdev, &adapter->desired.mac);
 857                 adapter->mac_change_pending = false;
 858         }
 859
 860         return 0;
 861 }
 862
 863 static void release_login_buffer(struct ibmvnic_adapter *adapter)
 864 {
 865         kfree(adapter->login_buf);
 866         adapter->login_buf = NULL;
 867 }
 868
 869 static void release_login_rsp_buffer(struct ibmvnic_adapter *adapter)
 870 {
 871         kfree(adapter->login_rsp_buf);
 872         adapter->login_rsp_buf = NULL;
 873 }
 874
 875 static void release_resources(struct ibmvnic_adapter *adapter)
 876 {
 877         release_vpd_data(adapter);
 878
 879         release_tx_pools(adapter);
 880         release_rx_pools(adapter);
 881
 882         release_napi(adapter);
 883         release_login_rsp_buffer(adapter);
 884 }
 885
 886 static int set_link_state(struct ibmvnic_adapter *adapter, u8 link_state)
 887 {
 888         struct net_device *netdev = adapter->netdev;
 889         unsigned long timeout = msecs_to_jiffies(30000);
 890         union ibmvnic_crq crq;
 891         bool resend;
 892         int rc;
 893
 894         netdev_dbg(netdev, "setting link state %d\n", link_state);
 895
 896         memset(&crq, 0, sizeof(crq));
 897         crq.logical_link_state.first = IBMVNIC_CRQ_CMD;
 898         crq.logical_link_state.cmd = LOGICAL_LINK_STATE;
 899         crq.logical_link_state.link_state = link_state;
 900
 901         do {
 902                 resend = false;
 903
 904                 reinit_completion(&adapter->init_done);
 905                 rc = ibmvnic_send_crq(adapter, &crq);
 906                 if (rc) {
 907                         netdev_err(netdev, "Failed to set link state\n");
 908                         return rc;
 909                 }
 910
 911                 if (!wait_for_completion_timeout(&adapter->init_done,
 912                                                  timeout)) {
 913                         netdev_err(netdev, "timeout setting link state\n");
 914                         return -1;
 915                 }
 916
 917                 if (adapter->init_done_rc == 1) {
 918                         /* Partuial success, delay and re-send */
 919                         mdelay(1000);
 920                         resend = true;
 921                 } else if (adapter->init_done_rc) {
 922                         netdev_warn(netdev, "Unable to set link state, rc=%d\n",
 923                                     adapter->init_done_rc);
 924                         return adapter->init_done_rc;
 925                 }
 926         } while (resend);
 927
 928         return 0;
 929 }
 930
 931 static int set_real_num_queues(struct net_device *netdev)
 932 {
 933         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
 934         int rc;
 935
 936         netdev_dbg(netdev, "Setting real tx/rx queues (%llx/%llx)\n",
 937                    adapter->req_tx_queues, adapter->req_rx_queues);
 938
 939         rc = netif_set_real_num_tx_queues(netdev, adapter->req_tx_queues);
 940         if (rc) {
 941                 netdev_err(netdev, "failed to set the number of tx queues\n");
 942                 return rc;
 943         }
 944
 945         rc = netif_set_real_num_rx_queues(netdev, adapter->req_rx_queues);
 946         if (rc)
 947                 netdev_err(netdev, "failed to set the number of rx queues\n");
 948
 949         return rc;
 950 }
 951
 952 static int ibmvnic_get_vpd(struct ibmvnic_adapter *adapter)
 953 {
 954         struct device *dev = &adapter->vdev->dev;
 955         union ibmvnic_crq crq;
 956         int len = 0;
 957         int rc;
 958
 959         if (adapter->vpd->buff)
 960                 len = adapter->vpd->len;
 961
 962         init_completion(&adapter->fw_done);
 963         crq.get_vpd_size.first = IBMVNIC_CRQ_CMD;
 964         crq.get_vpd_size.cmd = GET_VPD_SIZE;
 965         rc = ibmvnic_send_crq(adapter, &crq);
 966         if (rc)
 967                 return rc;
 968         wait_for_completion(&adapter->fw_done);
 969
 970         if (!adapter->vpd->len)
 971                 return -ENODATA;
 972
 973         if (!adapter->vpd->buff)
 974                 adapter->vpd->buff = kzalloc(adapter->vpd->len, GFP_KERNEL);
 975         else if (adapter->vpd->len != len)
 976                 adapter->vpd->buff =
 977                         krealloc(adapter->vpd->buff,
 978                                  adapter->vpd->len, GFP_KERNEL);
 979
 980         if (!adapter->vpd->buff) {
 981                 dev_err(dev, "Could allocate VPD buffer\n");
 982                 return -ENOMEM;
 983         }
 984
 985         adapter->vpd->dma_addr =
 986                 dma_map_single(dev, adapter->vpd->buff, adapter->vpd->len,
 987                                DMA_FROM_DEVICE);
 988         if (dma_mapping_error(dev, adapter->vpd->dma_addr)) {
 989                 dev_err(dev, "Could not map VPD buffer\n");
 990                 kfree(adapter->vpd->buff);
 991                 adapter->vpd->buff = NULL;
 992                 return -ENOMEM;
 993         }
 994
 995         reinit_completion(&adapter->fw_done);
 996         crq.get_vpd.first = IBMVNIC_CRQ_CMD;
 997         crq.get_vpd.cmd = GET_VPD;
 998         crq.get_vpd.ioba = cpu_to_be32(adapter->vpd->dma_addr);
 999         crq.get_vpd.len = cpu_to_be32((u32)adapter->vpd->len);
1000         rc = ibmvnic_send_crq(adapter, &crq);
1001         if (rc) {
1002                 kfree(adapter->vpd->buff);
1003                 adapter->vpd->buff = NULL;
1004                 return rc;
1005         }
1006         wait_for_completion(&adapter->fw_done);
1007
1008         return 0;
1009 }
1010
1011 static int init_resources(struct ibmvnic_adapter *adapter)
1012 {
1013         struct net_device *netdev = adapter->netdev;
1014         int rc;
1015
1016         rc = set_real_num_queues(netdev);
1017         if (rc)
1018                 return rc;
1019
1020         adapter->vpd = kzalloc(sizeof(*adapter->vpd), GFP_KERNEL);
1021         if (!adapter->vpd)
1022                 return -ENOMEM;
1023
1024         /* Vital Product Data (VPD) */
1025         rc = ibmvnic_get_vpd(adapter);
1026         if (rc) {
1027                 netdev_err(netdev, "failed to initialize Vital Product Data (VPD)\n");
1028                 return rc;
1029         }
1030
1031         adapter->map_id = 1;
1032
1033         rc = init_napi(adapter);
1034         if (rc)
1035                 return rc;
1036
1037         send_map_query(adapter);
1038
1039         rc = init_rx_pools(netdev);
1040         if (rc)
1041                 return rc;
1042
1043         rc = init_tx_pools(netdev);
1044         return rc;
1045 }
1046
1047 static int __ibmvnic_open(struct net_device *netdev)
1048 {
1049         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1050         enum vnic_state prev_state = adapter->state;
1051         int i, rc;
1052
1053         adapter->state = VNIC_OPENING;
1054         replenish_pools(adapter);
1055         ibmvnic_napi_enable(adapter);
1056
1057         /* We're ready to receive frames, enable the sub-crq interrupts and
1058          * set the logical link state to up
1059          */
1060         for (i = 0; i < adapter->req_rx_queues; i++) {
1061                 netdev_dbg(netdev, "Enabling rx_scrq[%d] irq\n", i);
1062                 if (prev_state == VNIC_CLOSED)
1063                         enable_irq(adapter->rx_scrq[i]->irq);
1064                 enable_scrq_irq(adapter, adapter->rx_scrq[i]);
1065         }
1066
1067         for (i = 0; i < adapter->req_tx_queues; i++) {
1068                 netdev_dbg(netdev, "Enabling tx_scrq[%d] irq\n", i);
1069                 if (prev_state == VNIC_CLOSED)
1070                         enable_irq(adapter->tx_scrq[i]->irq);
1071                 enable_scrq_irq(adapter, adapter->tx_scrq[i]);
1072         }
1073
1074         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_UP);
1075         if (rc) {
1076                 for (i = 0; i < adapter->req_rx_queues; i++)
1077                         napi_disable(&adapter->napi[i]);
1078                 release_resources(adapter);
1079                 return rc;
1080         }
1081
1082         netif_tx_start_all_queues(netdev);
1083
1084         if (prev_state == VNIC_CLOSED) {
1085                 for (i = 0; i < adapter->req_rx_queues; i++)
1086                         napi_schedule(&adapter->napi[i]);
1087         }
1088
1089         adapter->state = VNIC_OPEN;
1090         return rc;
1091 }
1092
1093 static int ibmvnic_open(struct net_device *netdev)
1094 {
1095         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1096         int rc;
1097
1098         /* If device failover is pending, just set device state and return.
1099          * Device operation will be handled by reset routine.
1100          */
1101         if (adapter->failover_pending) {
1102                 adapter->state = VNIC_OPEN;
1103                 return 0;
1104         }
1105
1106         mutex_lock(&adapter->reset_lock);
1107
1108         if (adapter->state != VNIC_CLOSED) {
1109                 rc = ibmvnic_login(netdev);
1110                 if (rc) {
1111                         mutex_unlock(&adapter->reset_lock);
1112                         return rc;
1113                 }
1114
1115                 rc = init_resources(adapter);
1116                 if (rc) {
1117                         netdev_err(netdev, "failed to initialize resources\n");
1118                         release_resources(adapter);
1119                         mutex_unlock(&adapter->reset_lock);
1120                         return rc;
1121                 }
1122         }
1123
1124         rc = __ibmvnic_open(netdev);
1125         netif_carrier_on(netdev);
1126
1127         mutex_unlock(&adapter->reset_lock);
1128
1129         return rc;
1130 }
1131
1132 static void clean_rx_pools(struct ibmvnic_adapter *adapter)
1133 {
1134         struct ibmvnic_rx_pool *rx_pool;
1135         struct ibmvnic_rx_buff *rx_buff;
1136         u64 rx_entries;
1137         int rx_scrqs;
1138         int i, j;
1139
1140         if (!adapter->rx_pool)
1141                 return;
1142
1143         rx_scrqs = adapter->num_active_rx_pools;
1144         rx_entries = adapter->req_rx_add_entries_per_subcrq;
1145
1146         /* Free any remaining skbs in the rx buffer pools */
1147         for (i = 0; i < rx_scrqs; i++) {
1148                 rx_pool = &adapter->rx_pool[i];
1149                 if (!rx_pool || !rx_pool->rx_buff)
1150                         continue;
1151
1152                 netdev_dbg(adapter->netdev, "Cleaning rx_pool[%d]\n", i);
1153                 for (j = 0; j < rx_entries; j++) {
1154                         rx_buff = &rx_pool->rx_buff[j];
1155                         if (rx_buff && rx_buff->skb) {
1156                                 dev_kfree_skb_any(rx_buff->skb);
1157                                 rx_buff->skb = NULL;
1158                         }
1159                 }
1160         }
1161 }
1162
1163 static void clean_one_tx_pool(struct ibmvnic_adapter *adapter,
1164                               struct ibmvnic_tx_pool *tx_pool)
1165 {
1166         struct ibmvnic_tx_buff *tx_buff;
1167         u64 tx_entries;
1168         int i;
1169
1170         if (!tx_pool || !tx_pool->tx_buff)
1171                 return;
1172
1173         tx_entries = tx_pool->num_buffers;
1174
1175         for (i = 0; i < tx_entries; i++) {
1176                 tx_buff = &tx_pool->tx_buff[i];
1177                 if (tx_buff && tx_buff->skb) {
1178                         dev_kfree_skb_any(tx_buff->skb);
1179                         tx_buff->skb = NULL;
1180                 }
1181         }
1182 }
1183
1184 static void clean_tx_pools(struct ibmvnic_adapter *adapter)
1185 {
1186         int tx_scrqs;
1187         int i;
1188
1189         if (!adapter->tx_pool || !adapter->tso_pool)
1190                 return;
1191
1192         tx_scrqs = adapter->num_active_tx_pools;
1193
1194         /* Free any remaining skbs in the tx buffer pools */
1195         for (i = 0; i < tx_scrqs; i++) {
1196                 netdev_dbg(adapter->netdev, "Cleaning tx_pool[%d]\n", i);
1197                 clean_one_tx_pool(adapter, &adapter->tx_pool[i]);
1198                 clean_one_tx_pool(adapter, &adapter->tso_pool[i]);
1199         }
1200 }
1201
1202 static void ibmvnic_disable_irqs(struct ibmvnic_adapter *adapter)
1203 {
1204         struct net_device *netdev = adapter->netdev;
1205         int i;
1206
1207         if (adapter->tx_scrq) {
1208                 for (i = 0; i < adapter->req_tx_queues; i++)
1209                         if (adapter->tx_scrq[i]->irq) {
1210                                 netdev_dbg(netdev,
1211                                            "Disabling tx_scrq[%d] irq\n", i);
1212                                 disable_scrq_irq(adapter, adapter->tx_scrq[i]);
1213                                 disable_irq(adapter->tx_scrq[i]->irq);
1214                         }
1215         }
1216
1217         if (adapter->rx_scrq) {
1218                 for (i = 0; i < adapter->req_rx_queues; i++) {
1219                         if (adapter->rx_scrq[i]->irq) {
1220                                 netdev_dbg(netdev,
1221                                            "Disabling rx_scrq[%d] irq\n", i);
1222                                 disable_scrq_irq(adapter, adapter->rx_scrq[i]);
1223                                 disable_irq(adapter->rx_scrq[i]->irq);
1224                         }
1225                 }
1226         }
1227 }
1228
1229 static void ibmvnic_cleanup(struct net_device *netdev)
1230 {
1231         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1232
1233         /* ensure that transmissions are stopped if called by do_reset */
1234         if (adapter->resetting)
1235                 netif_tx_disable(netdev);
1236         else
1237                 netif_tx_stop_all_queues(netdev);
1238
1239         ibmvnic_napi_disable(adapter);
1240         ibmvnic_disable_irqs(adapter);
1241
1242         clean_rx_pools(adapter);
1243         clean_tx_pools(adapter);
1244 }
1245
1246 static int __ibmvnic_close(struct net_device *netdev)
1247 {
1248         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1249         int rc = 0;
1250
1251         adapter->state = VNIC_CLOSING;
1252         rc = set_link_state(adapter, IBMVNIC_LOGICAL_LNK_DN);
1253         if (rc)
1254                 return rc;
1255         adapter->state = VNIC_CLOSED;
1256         return 0;
1257 }
1258
1259 static int ibmvnic_close(struct net_device *netdev)
1260 {
1261         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1262         int rc;
1263
1264         /* If device failover is pending, just set device state and return.
1265          * Device operation will be handled by reset routine.
1266          */
1267         if (adapter->failover_pending) {
1268                 adapter->state = VNIC_CLOSED;
1269                 return 0;
1270         }
1271
1272         mutex_lock(&adapter->reset_lock);
1273         rc = __ibmvnic_close(netdev);
1274         ibmvnic_cleanup(netdev);
1275         mutex_unlock(&adapter->reset_lock);
1276
1277         return rc;
1278 }
1279
1280 /**
1281  * build_hdr_data - creates L2/L3/L4 header data buffer
1282  * @hdr_field - bitfield determining needed headers
1283  * @skb - socket buffer
1284  * @hdr_len - array of header lengths
1285  * @tot_len - total length of data
1286  *
1287  * Reads hdr_field to determine which headers are needed by firmware.
1288  * Builds a buffer containing these headers.  Saves individual header
1289  * lengths and total buffer length to be used to build descriptors.
1290  */
1291 static int build_hdr_data(u8 hdr_field, struct sk_buff *skb,
1292                           int *hdr_len, u8 *hdr_data)
1293 {
1294         int len = 0;
1295         u8 *hdr;
1296
1297         if (skb_vlan_tagged(skb) && !skb_vlan_tag_present(skb))
1298                 hdr_len[0] = sizeof(struct vlan_ethhdr);
1299         else
1300                 hdr_len[0] = sizeof(struct ethhdr);
1301
1302         if (skb->protocol == htons(ETH_P_IP)) {
1303                 hdr_len[1] = ip_hdr(skb)->ihl * 4;
1304                 if (ip_hdr(skb)->protocol == IPPROTO_TCP)
1305                         hdr_len[2] = tcp_hdrlen(skb);
1306                 else if (ip_hdr(skb)->protocol == IPPROTO_UDP)
1307                         hdr_len[2] = sizeof(struct udphdr);
1308         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1309                 hdr_len[1] = sizeof(struct ipv6hdr);
1310                 if (ipv6_hdr(skb)->nexthdr == IPPROTO_TCP)
1311                         hdr_len[2] = tcp_hdrlen(skb);
1312                 else if (ipv6_hdr(skb)->nexthdr == IPPROTO_UDP)
1313                         hdr_len[2] = sizeof(struct udphdr);
1314         } else if (skb->protocol == htons(ETH_P_ARP)) {
1315                 hdr_len[1] = arp_hdr_len(skb->dev);
1316                 hdr_len[2] = 0;
1317         }
1318
1319         memset(hdr_data, 0, 120);
1320         if ((hdr_field >> 6) & 1) {
1321                 hdr = skb_mac_header(skb);
1322                 memcpy(hdr_data, hdr, hdr_len[0]);
1323                 len += hdr_len[0];
1324         }
1325
1326         if ((hdr_field >> 5) & 1) {
1327                 hdr = skb_network_header(skb);
1328                 memcpy(hdr_data + len, hdr, hdr_len[1]);
1329                 len += hdr_len[1];
1330         }
1331
1332         if ((hdr_field >> 4) & 1) {
1333                 hdr = skb_transport_header(skb);
1334                 memcpy(hdr_data + len, hdr, hdr_len[2]);
1335                 len += hdr_len[2];
1336         }
1337         return len;
1338 }
1339
1340 /**
1341  * create_hdr_descs - create header and header extension descriptors
1342  * @hdr_field - bitfield determining needed headers
1343  * @data - buffer containing header data
1344  * @len - length of data buffer
1345  * @hdr_len - array of individual header lengths
1346  * @scrq_arr - descriptor array
1347  *
1348  * Creates header and, if needed, header extension descriptors and
1349  * places them in a descriptor array, scrq_arr
1350  */
1351
1352 static int create_hdr_descs(u8 hdr_field, u8 *hdr_data, int len, int *hdr_len,
1353                             union sub_crq *scrq_arr)
1354 {
1355         union sub_crq hdr_desc;
1356         int tmp_len = len;
1357         int num_descs = 0;
1358         u8 *data, *cur;
1359         int tmp;
1360
1361         while (tmp_len > 0) {
1362                 cur = hdr_data + len - tmp_len;
1363
1364                 memset(&hdr_desc, 0, sizeof(hdr_desc));
1365                 if (cur != hdr_data) {
1366                         data = hdr_desc.hdr_ext.data;
1367                         tmp = tmp_len > 29 ? 29 : tmp_len;
1368                         hdr_desc.hdr_ext.first = IBMVNIC_CRQ_CMD;
1369                         hdr_desc.hdr_ext.type = IBMVNIC_HDR_EXT_DESC;
1370                         hdr_desc.hdr_ext.len = tmp;
1371                 } else {
1372                         data = hdr_desc.hdr.data;
1373                         tmp = tmp_len > 24 ? 24 : tmp_len;
1374                         hdr_desc.hdr.first = IBMVNIC_CRQ_CMD;
1375                         hdr_desc.hdr.type = IBMVNIC_HDR_DESC;
1376                         hdr_desc.hdr.len = tmp;
1377                         hdr_desc.hdr.l2_len = (u8)hdr_len[0];
1378                         hdr_desc.hdr.l3_len = cpu_to_be16((u16)hdr_len[1]);
1379                         hdr_desc.hdr.l4_len = (u8)hdr_len[2];
1380                         hdr_desc.hdr.flag = hdr_field << 1;
1381                 }
1382                 memcpy(data, cur, tmp);
1383                 tmp_len -= tmp;
1384                 *scrq_arr = hdr_desc;
1385                 scrq_arr++;
1386                 num_descs++;
1387         }
1388
1389         return num_descs;
1390 }
1391
1392 /**
1393  * build_hdr_descs_arr - build a header descriptor array
1394  * @skb - socket buffer
1395  * @num_entries - number of descriptors to be sent
1396  * @subcrq - first TX descriptor
1397  * @hdr_field - bit field determining which headers will be sent
1398  *
1399  * This function will build a TX descriptor array with applicable
1400  * L2/L3/L4 packet header descriptors to be sent by send_subcrq_indirect.
1401  */
1402
1403 static void build_hdr_descs_arr(struct ibmvnic_tx_buff *txbuff,
1404                                 int *num_entries, u8 hdr_field)
1405 {
1406         int hdr_len[3] = {0, 0, 0};
1407         int tot_len;
1408         u8 *hdr_data = txbuff->hdr_data;
1409
1410         tot_len = build_hdr_data(hdr_field, txbuff->skb, hdr_len,
1411                                  txbuff->hdr_data);
1412         *num_entries += create_hdr_descs(hdr_field, hdr_data, tot_len, hdr_len,
1413                          txbuff->indir_arr + 1);
1414 }
1415
1416 static int ibmvnic_xmit_workarounds(struct sk_buff *skb,
1417                                     struct net_device *netdev)
1418 {
1419         /* For some backing devices, mishandling of small packets
1420          * can result in a loss of connection or TX stall. Device
1421          * architects recommend that no packet should be smaller
1422          * than the minimum MTU value provided to the driver, so
1423          * pad any packets to that length
1424          */
1425         if (skb->len < netdev->min_mtu)
1426                 return skb_put_padto(skb, netdev->min_mtu);
1427
1428         return 0;
1429 }
1430
1431 static int ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev)
1432 {
1433         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1434         int queue_num = skb_get_queue_mapping(skb);
1435         u8 *hdrs = (u8 *)&adapter->tx_rx_desc_req;
1436         struct device *dev = &adapter->vdev->dev;
1437         struct ibmvnic_tx_buff *tx_buff = NULL;
1438         struct ibmvnic_sub_crq_queue *tx_scrq;
1439         struct ibmvnic_tx_pool *tx_pool;
1440         unsigned int tx_send_failed = 0;
1441         unsigned int tx_map_failed = 0;
1442         unsigned int tx_dropped = 0;
1443         unsigned int tx_packets = 0;
1444         unsigned int tx_bytes = 0;
1445         dma_addr_t data_dma_addr;
1446         struct netdev_queue *txq;
1447         unsigned long lpar_rc;
1448         union sub_crq tx_crq;
1449         unsigned int offset;
1450         int num_entries = 1;
1451         unsigned char *dst;
1452         u64 *handle_array;
1453         int index = 0;
1454         u8 proto = 0;
1455         int ret = 0;
1456
1457         if (adapter->resetting) {
1458                 if (!netif_subqueue_stopped(netdev, skb))
1459                         netif_stop_subqueue(netdev, queue_num);
1460                 dev_kfree_skb_any(skb);
1461
1462                 tx_send_failed++;
1463                 tx_dropped++;
1464                 ret = NETDEV_TX_OK;
1465                 goto out;
1466         }
1467
1468         if (ibmvnic_xmit_workarounds(skb, netdev)) {
1469                 tx_dropped++;
1470                 tx_send_failed++;
1471                 ret = NETDEV_TX_OK;
1472                 goto out;
1473         }
1474         if (skb_is_gso(skb))
1475                 tx_pool = &adapter->tso_pool[queue_num];
1476         else
1477                 tx_pool = &adapter->tx_pool[queue_num];
1478
1479         tx_scrq = adapter->tx_scrq[queue_num];
1480         txq = netdev_get_tx_queue(netdev, skb_get_queue_mapping(skb));
1481         handle_array = (u64 *)((u8 *)(adapter->login_rsp_buf) +
1482                 be32_to_cpu(adapter->login_rsp_buf->off_txsubm_subcrqs));
1483
1484         index = tx_pool->free_map[tx_pool->consumer_index];
1485
1486         if (index == IBMVNIC_INVALID_MAP) {
1487                 dev_kfree_skb_any(skb);
1488                 tx_send_failed++;
1489                 tx_dropped++;
1490                 ret = NETDEV_TX_OK;
1491                 goto out;
1492         }
1493
1494         tx_pool->free_map[tx_pool->consumer_index] = IBMVNIC_INVALID_MAP;
1495
1496         offset = index * tx_pool->buf_size;
1497         dst = tx_pool->long_term_buff.buff + offset;
1498         memset(dst, 0, tx_pool->buf_size);
1499         data_dma_addr = tx_pool->long_term_buff.addr + offset;
1500
1501         if (skb_shinfo(skb)->nr_frags) {
1502                 int cur, i;
1503
1504                 /* Copy the head */
1505                 skb_copy_from_linear_data(skb, dst, skb_headlen(skb));
1506                 cur = skb_headlen(skb);
1507
1508                 /* Copy the frags */
1509                 for (i = 0; i < skb_shinfo(skb)->nr_frags; i++) {
1510                         const skb_frag_t *frag = &skb_shinfo(skb)->frags[i];
1511
1512                         memcpy(dst + cur,
1513                                page_address(skb_frag_page(frag)) +
1514                                frag->page_offset, skb_frag_size(frag));
1515                         cur += skb_frag_size(frag);
1516                 }
1517         } else {
1518                 skb_copy_from_linear_data(skb, dst, skb->len);
1519         }
1520
1521         tx_pool->consumer_index =
1522             (tx_pool->consumer_index + 1) % tx_pool->num_buffers;
1523
1524         tx_buff = &tx_pool->tx_buff[index];
1525         tx_buff->skb = skb;
1526         tx_buff->data_dma[0] = data_dma_addr;
1527         tx_buff->data_len[0] = skb->len;
1528         tx_buff->index = index;
1529         tx_buff->pool_index = queue_num;
1530         tx_buff->last_frag = true;
1531
1532         memset(&tx_crq, 0, sizeof(tx_crq));
1533         tx_crq.v1.first = IBMVNIC_CRQ_CMD;
1534         tx_crq.v1.type = IBMVNIC_TX_DESC;
1535         tx_crq.v1.n_crq_elem = 1;
1536         tx_crq.v1.n_sge = 1;
1537         tx_crq.v1.flags1 = IBMVNIC_TX_COMP_NEEDED;
1538
1539         if (skb_is_gso(skb))
1540                 tx_crq.v1.correlator =
1541                         cpu_to_be32(index | IBMVNIC_TSO_POOL_MASK);
1542         else
1543                 tx_crq.v1.correlator = cpu_to_be32(index);
1544         tx_crq.v1.dma_reg = cpu_to_be16(tx_pool->long_term_buff.map_id);
1545         tx_crq.v1.sge_len = cpu_to_be32(skb->len);
1546         tx_crq.v1.ioba = cpu_to_be64(data_dma_addr);
1547
1548         if (adapter->vlan_header_insertion) {
1549                 tx_crq.v1.flags2 |= IBMVNIC_TX_VLAN_INSERT;
1550                 tx_crq.v1.vlan_id = cpu_to_be16(skb->vlan_tci);
1551         }
1552
1553         if (skb->protocol == htons(ETH_P_IP)) {
1554                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV4;
1555                 proto = ip_hdr(skb)->protocol;
1556         } else if (skb->protocol == htons(ETH_P_IPV6)) {
1557                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_IPV6;
1558                 proto = ipv6_hdr(skb)->nexthdr;
1559         }
1560
1561         if (proto == IPPROTO_TCP)
1562                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_TCP;
1563         else if (proto == IPPROTO_UDP)
1564                 tx_crq.v1.flags1 |= IBMVNIC_TX_PROT_UDP;
1565
1566         if (skb->ip_summed == CHECKSUM_PARTIAL) {
1567                 tx_crq.v1.flags1 |= IBMVNIC_TX_CHKSUM_OFFLOAD;
1568                 hdrs += 2;
1569         }
1570         if (skb_is_gso(skb)) {
1571                 tx_crq.v1.flags1 |= IBMVNIC_TX_LSO;
1572                 tx_crq.v1.mss = cpu_to_be16(skb_shinfo(skb)->gso_size);
1573                 hdrs += 2;
1574         }
1575         /* determine if l2/3/4 headers are sent to firmware */
1576         if ((*hdrs >> 7) & 1) {
1577                 build_hdr_descs_arr(tx_buff, &num_entries, *hdrs);
1578                 tx_crq.v1.n_crq_elem = num_entries;
1579                 tx_buff->num_entries = num_entries;
1580                 tx_buff->indir_arr[0] = tx_crq;
1581                 tx_buff->indir_dma = dma_map_single(dev, tx_buff->indir_arr,
1582                                                     sizeof(tx_buff->indir_arr),
1583                                                     DMA_TO_DEVICE);
1584                 if (dma_mapping_error(dev, tx_buff->indir_dma)) {
1585                         dev_kfree_skb_any(skb);
1586                         tx_buff->skb = NULL;
1587                         if (!firmware_has_feature(FW_FEATURE_CMO))
1588                                 dev_err(dev, "tx: unable to map descriptor array\n");
1589                         tx_map_failed++;
1590                         tx_dropped++;
1591                         ret = NETDEV_TX_OK;
1592                         goto tx_err_out;
1593                 }
1594                 lpar_rc = send_subcrq_indirect(adapter, handle_array[queue_num],
1595                                                (u64)tx_buff->indir_dma,
1596                                                (u64)num_entries);
1597         } else {
1598                 tx_buff->num_entries = num_entries;
1599                 lpar_rc = send_subcrq(adapter, handle_array[queue_num],
1600                                       &tx_crq);
1601         }
1602         if (lpar_rc != H_SUCCESS) {
1603                 if (lpar_rc != H_CLOSED && lpar_rc != H_PARAMETER)
1604                         dev_err_ratelimited(dev, "tx: send failed\n");
1605                 dev_kfree_skb_any(skb);
1606                 tx_buff->skb = NULL;
1607
1608                 if (lpar_rc == H_CLOSED || adapter->failover_pending) {
1609                         /* Disable TX and report carrier off if queue is closed
1610                          * or pending failover.
1611                          * Firmware guarantees that a signal will be sent to the
1612                          * driver, triggering a reset or some other action.
1613                          */
1614                         netif_tx_stop_all_queues(netdev);
1615                         netif_carrier_off(netdev);
1616                 }
1617
1618                 tx_send_failed++;
1619                 tx_dropped++;
1620                 ret = NETDEV_TX_OK;
1621                 goto tx_err_out;
1622         }
1623
1624         if (atomic_add_return(num_entries, &tx_scrq->used)
1625                                         >= adapter->req_tx_entries_per_subcrq) {
1626                 netdev_dbg(netdev, "Stopping queue %d\n", queue_num);
1627                 netif_stop_subqueue(netdev, queue_num);
1628         }
1629
1630         tx_packets++;
1631         tx_bytes += skb->len;
1632         txq->trans_start = jiffies;
1633         ret = NETDEV_TX_OK;
1634         goto out;
1635
1636 tx_err_out:
1637         /* roll back consumer index and map array*/
1638         if (tx_pool->consumer_index == 0)
1639                 tx_pool->consumer_index =
1640                         tx_pool->num_buffers - 1;
1641         else
1642                 tx_pool->consumer_index--;
1643         tx_pool->free_map[tx_pool->consumer_index] = index;
1644 out:
1645         netdev->stats.tx_dropped += tx_dropped;
1646         netdev->stats.tx_bytes += tx_bytes;
1647         netdev->stats.tx_packets += tx_packets;
1648         adapter->tx_send_failed += tx_send_failed;
1649         adapter->tx_map_failed += tx_map_failed;
1650         adapter->tx_stats_buffers[queue_num].packets += tx_packets;
1651         adapter->tx_stats_buffers[queue_num].bytes += tx_bytes;
1652         adapter->tx_stats_buffers[queue_num].dropped_packets += tx_dropped;
1653
1654         return ret;
1655 }
1656
1657 static void ibmvnic_set_multi(struct net_device *netdev)
1658 {
1659         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1660         struct netdev_hw_addr *ha;
1661         union ibmvnic_crq crq;
1662
1663         memset(&crq, 0, sizeof(crq));
1664         crq.request_capability.first = IBMVNIC_CRQ_CMD;
1665         crq.request_capability.cmd = REQUEST_CAPABILITY;
1666
1667         if (netdev->flags & IFF_PROMISC) {
1668                 if (!adapter->promisc_supported)
1669                         return;
1670         } else {
1671                 if (netdev->flags & IFF_ALLMULTI) {
1672                         /* Accept all multicast */
1673                         memset(&crq, 0, sizeof(crq));
1674                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1675                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1676                         crq.multicast_ctrl.flags = IBMVNIC_ENABLE_ALL;
1677                         ibmvnic_send_crq(adapter, &crq);
1678                 } else if (netdev_mc_empty(netdev)) {
1679                         /* Reject all multicast */
1680                         memset(&crq, 0, sizeof(crq));
1681                         crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1682                         crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1683                         crq.multicast_ctrl.flags = IBMVNIC_DISABLE_ALL;
1684                         ibmvnic_send_crq(adapter, &crq);
1685                 } else {
1686                         /* Accept one or more multicast(s) */
1687                         netdev_for_each_mc_addr(ha, netdev) {
1688                                 memset(&crq, 0, sizeof(crq));
1689                                 crq.multicast_ctrl.first = IBMVNIC_CRQ_CMD;
1690                                 crq.multicast_ctrl.cmd = MULTICAST_CTRL;
1691                                 crq.multicast_ctrl.flags = IBMVNIC_ENABLE_MC;
1692                                 ether_addr_copy(&crq.multicast_ctrl.mac_addr[0],
1693                                                 ha->addr);
1694                                 ibmvnic_send_crq(adapter, &crq);
1695                         }
1696                 }
1697         }
1698 }
1699
1700 static int __ibmvnic_set_mac(struct net_device *netdev, struct sockaddr *p)
1701 {
1702         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1703         struct sockaddr *addr = p;
1704         union ibmvnic_crq crq;
1705         int rc;
1706
1707         if (!is_valid_ether_addr(addr->sa_data))
1708                 return -EADDRNOTAVAIL;
1709
1710         memset(&crq, 0, sizeof(crq));
1711         crq.change_mac_addr.first = IBMVNIC_CRQ_CMD;
1712         crq.change_mac_addr.cmd = CHANGE_MAC_ADDR;
1713         ether_addr_copy(&crq.change_mac_addr.mac_addr[0], addr->sa_data);
1714
1715         init_completion(&adapter->fw_done);
1716         rc = ibmvnic_send_crq(adapter, &crq);
1717         if (rc)
1718                 return rc;
1719         wait_for_completion(&adapter->fw_done);
1720         /* netdev->dev_addr is changed in handle_change_mac_rsp function */
1721         return adapter->fw_done_rc ? -EIO : 0;
1722 }
1723
1724 static int ibmvnic_set_mac(struct net_device *netdev, void *p)
1725 {
1726         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
1727         struct sockaddr *addr = p;
1728         int rc;
1729
1730         if (adapter->state == VNIC_PROBED) {
1731                 memcpy(&adapter->desired.mac, addr, sizeof(struct sockaddr));
1732                 adapter->mac_change_pending = true;
1733                 return 0;
1734         }
1735
1736         rc = __ibmvnic_set_mac(netdev, addr);
1737
1738         return rc;
1739 }
1740
1741 /**
1742  * do_reset returns zero if we are able to keep processing reset events, or
1743  * non-zero if we hit a fatal error and must halt.
1744  */
1745 static int do_reset(struct ibmvnic_adapter *adapter,
1746                     struct ibmvnic_rwi *rwi, u32 reset_state)
1747 {
1748         u64 old_num_rx_queues, old_num_tx_queues;
1749         struct net_device *netdev = adapter->netdev;
1750         int i, rc;
1751
1752         netdev_dbg(adapter->netdev, "Re-setting driver (%d)\n",
1753                    rwi->reset_reason);
1754
1755         netif_carrier_off(netdev);
1756         adapter->reset_reason = rwi->reset_reason;
1757
1758         old_num_rx_queues = adapter->req_rx_queues;
1759         old_num_tx_queues = adapter->req_tx_queues;
1760
1761         ibmvnic_cleanup(netdev);
1762
1763         if (adapter->reset_reason != VNIC_RESET_MOBILITY &&
1764             adapter->reset_reason != VNIC_RESET_FAILOVER) {
1765                 rc = __ibmvnic_close(netdev);
1766                 if (rc)
1767                         return rc;
1768         }
1769
1770         if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1771             adapter->wait_for_reset) {
1772                 release_resources(adapter);
1773                 release_sub_crqs(adapter, 1);
1774                 release_crq_queue(adapter);
1775         }
1776
1777         if (adapter->reset_reason != VNIC_RESET_NON_FATAL) {
1778                 /* remove the closed state so when we call open it appears
1779                  * we are coming from the probed state.
1780                  */
1781                 adapter->state = VNIC_PROBED;
1782
1783                 if (adapter->wait_for_reset) {
1784                         rc = init_crq_queue(adapter);
1785                 } else if (adapter->reset_reason == VNIC_RESET_MOBILITY) {
1786                         rc = ibmvnic_reenable_crq_queue(adapter);
1787                         release_sub_crqs(adapter, 1);
1788                 } else {
1789                         rc = ibmvnic_reset_crq(adapter);
1790                         if (!rc)
1791                                 rc = vio_enable_interrupts(adapter->vdev);
1792                 }
1793
1794                 if (rc) {
1795                         netdev_err(adapter->netdev,
1796                                    "Couldn't initialize crq. rc=%d\n", rc);
1797                         return rc;
1798                 }
1799
1800                 rc = ibmvnic_reset_init(adapter);
1801                 if (rc)
1802                         return IBMVNIC_INIT_FAILED;
1803
1804                 /* If the adapter was in PROBE state prior to the reset,
1805                  * exit here.
1806                  */
1807                 if (reset_state == VNIC_PROBED)
1808                         return 0;
1809
1810                 rc = ibmvnic_login(netdev);
1811                 if (rc) {
1812                         adapter->state = reset_state;
1813                         return rc;
1814                 }
1815
1816                 if (adapter->reset_reason == VNIC_RESET_CHANGE_PARAM ||
1817                     adapter->wait_for_reset) {
1818                         rc = init_resources(adapter);
1819                         if (rc)
1820                                 return rc;
1821                 } else if (adapter->req_rx_queues != old_num_rx_queues ||
1822                            adapter->req_tx_queues != old_num_tx_queues) {
1823                         adapter->map_id = 1;
1824                         release_rx_pools(adapter);
1825                         release_tx_pools(adapter);
1826                         rc = init_rx_pools(netdev);
1827                         if (rc)
1828                                 return rc;
1829                         rc = init_tx_pools(netdev);
1830                         if (rc)
1831                                 return rc;
1832
1833                         release_napi(adapter);
1834                         rc = init_napi(adapter);
1835                         if (rc)
1836                                 return rc;
1837                 } else {
1838                         rc = reset_tx_pools(adapter);
1839                         if (rc)
1840                                 return rc;
1841
1842                         rc = reset_rx_pools(adapter);
1843                         if (rc)
1844                                 return rc;
1845                 }
1846                 ibmvnic_disable_irqs(adapter);
1847         }
1848         adapter->state = VNIC_CLOSED;
1849
1850         if (reset_state == VNIC_CLOSED)
1851                 return 0;
1852
1853         rc = __ibmvnic_open(netdev);
1854         if (rc) {
1855                 if (list_empty(&adapter->rwi_list))
1856                         adapter->state = VNIC_CLOSED;
1857                 else
1858                         adapter->state = reset_state;
1859
1860                 return 0;
1861         }
1862
1863         /* kick napi */
1864         for (i = 0; i < adapter->req_rx_queues; i++)
1865                 napi_schedule(&adapter->napi[i]);
1866
1867         if (adapter->reset_reason != VNIC_RESET_FAILOVER &&
1868             adapter->reset_reason != VNIC_RESET_CHANGE_PARAM)
1869                 netdev_notify_peers(netdev);
1870
1871         netif_carrier_on(netdev);
1872
1873         return 0;
1874 }
1875
1876 static int do_hard_reset(struct ibmvnic_adapter *adapter,
1877                          struct ibmvnic_rwi *rwi, u32 reset_state)
1878 {
1879         struct net_device *netdev = adapter->netdev;
1880         int rc;
1881
1882         netdev_dbg(adapter->netdev, "Hard resetting driver (%d)\n",
1883                    rwi->reset_reason);
1884
1885         netif_carrier_off(netdev);
1886         adapter->reset_reason = rwi->reset_reason;
1887
1888         ibmvnic_cleanup(netdev);
1889         release_resources(adapter);
1890         release_sub_crqs(adapter, 0);
1891         release_crq_queue(adapter);
1892
1893         /* remove the closed state so when we call open it appears
1894          * we are coming from the probed state.
1895          */
1896         adapter->state = VNIC_PROBED;
1897
1898         rc = init_crq_queue(adapter);
1899         if (rc) {
1900                 netdev_err(adapter->netdev,
1901                            "Couldn't initialize crq. rc=%d\n", rc);
1902                 return rc;
1903         }
1904
1905         rc = ibmvnic_init(adapter);
1906         if (rc)
1907                 return rc;
1908
1909         /* If the adapter was in PROBE state prior to the reset,
1910          * exit here.
1911          */
1912         if (reset_state == VNIC_PROBED)
1913                 return 0;
1914
1915         rc = ibmvnic_login(netdev);
1916         if (rc) {
1917                 adapter->state = VNIC_PROBED;
1918                 return 0;
1919         }
1920         /* netif_set_real_num_xx_queues needs to take rtnl lock here
1921          * unless wait_for_reset is set, in which case the rtnl lock
1922          * has already been taken before initializing the reset
1923          */
1924         if (!adapter->wait_for_reset) {
1925                 rtnl_lock();
1926                 rc = init_resources(adapter);
1927                 rtnl_unlock();
1928         } else {
1929                 rc = init_resources(adapter);
1930         }
1931         if (rc)
1932                 return rc;
1933
1934         ibmvnic_disable_irqs(adapter);
1935         adapter->state = VNIC_CLOSED;
1936
1937         if (reset_state == VNIC_CLOSED)
1938                 return 0;
1939
1940         rc = __ibmvnic_open(netdev);
1941         if (rc) {
1942                 if (list_empty(&adapter->rwi_list))
1943                         adapter->state = VNIC_CLOSED;
1944                 else
1945                         adapter->state = reset_state;
1946
1947                 return 0;
1948         }
1949
1950         netif_carrier_on(netdev);
1951
1952         return 0;
1953 }
1954
1955 static struct ibmvnic_rwi *get_next_rwi(struct ibmvnic_adapter *adapter)
1956 {
1957         struct ibmvnic_rwi *rwi;
1958
1959         mutex_lock(&adapter->rwi_lock);
1960
1961         if (!list_empty(&adapter->rwi_list)) {
1962                 rwi = list_first_entry(&adapter->rwi_list, struct ibmvnic_rwi,
1963                                        list);
1964                 list_del(&rwi->list);
1965         } else {
1966                 rwi = NULL;
1967         }
1968
1969         mutex_unlock(&adapter->rwi_lock);
1970         return rwi;
1971 }
1972
1973 static void free_all_rwi(struct ibmvnic_adapter *adapter)
1974 {
1975         struct ibmvnic_rwi *rwi;
1976
1977         rwi = get_next_rwi(adapter);
1978         while (rwi) {
1979                 kfree(rwi);
1980                 rwi = get_next_rwi(adapter);
1981         }
1982 }
1983
1984 static void __ibmvnic_reset(struct work_struct *work)
1985 {
1986         struct ibmvnic_rwi *rwi;
1987         struct ibmvnic_adapter *adapter;
1988         struct net_device *netdev;
1989         u32 reset_state;
1990         int rc = 0;
1991
1992         adapter = container_of(work, struct ibmvnic_adapter, ibmvnic_reset);
1993         netdev = adapter->netdev;
1994
1995         mutex_lock(&adapter->reset_lock);
1996         reset_state = adapter->state;
1997
1998         rwi = get_next_rwi(adapter);
1999         while (rwi) {
2000                 if (adapter->force_reset_recovery) {
2001                         adapter->force_reset_recovery = false;
2002                         rc = do_hard_reset(adapter, rwi, reset_state);
2003                 } else {
2004                         rc = do_reset(adapter, rwi, reset_state);
2005                 }
2006                 kfree(rwi);
2007                 if (rc && rc != IBMVNIC_INIT_FAILED &&
2008                     !adapter->force_reset_recovery)
2009                         break;
2010
2011                 rwi = get_next_rwi(adapter);
2012         }
2013
2014         if (adapter->wait_for_reset) {
2015                 adapter->wait_for_reset = false;
2016                 adapter->reset_done_rc = rc;
2017                 complete(&adapter->reset_done);
2018         }
2019
2020         if (rc) {
2021                 netdev_dbg(adapter->netdev, "Reset failed\n");
2022                 free_all_rwi(adapter);
2023                 mutex_unlock(&adapter->reset_lock);
2024                 return;
2025         }
2026
2027         adapter->resetting = false;
2028         mutex_unlock(&adapter->reset_lock);
2029 }
2030
2031 static int ibmvnic_reset(struct ibmvnic_adapter *adapter,
2032                          enum ibmvnic_reset_reason reason)
2033 {
2034         struct list_head *entry, *tmp_entry;
2035         struct ibmvnic_rwi *rwi, *tmp;
2036         struct net_device *netdev = adapter->netdev;
2037         int ret;
2038
2039         if (adapter->state == VNIC_REMOVING ||
2040             adapter->state == VNIC_REMOVED ||
2041             adapter->failover_pending) {
2042                 ret = EBUSY;
2043                 netdev_dbg(netdev, "Adapter removing or pending failover, skipping reset\n");
2044                 goto err;
2045         }
2046
2047         if (adapter->state == VNIC_PROBING) {
2048                 netdev_warn(netdev, "Adapter reset during probe\n");
2049                 ret = adapter->init_done_rc = EAGAIN;
2050                 goto err;
2051         }
2052
2053         mutex_lock(&adapter->rwi_lock);
2054
2055         list_for_each(entry, &adapter->rwi_list) {
2056                 tmp = list_entry(entry, struct ibmvnic_rwi, list);
2057                 if (tmp->reset_reason == reason) {
2058                         netdev_dbg(netdev, "Skipping matching reset\n");
2059                         mutex_unlock(&adapter->rwi_lock);
2060                         ret = EBUSY;
2061                         goto err;
2062                 }
2063         }
2064
2065         rwi = kzalloc(sizeof(*rwi), GFP_KERNEL);
2066         if (!rwi) {
2067                 mutex_unlock(&adapter->rwi_lock);
2068                 ibmvnic_close(netdev);
2069                 ret = ENOMEM;
2070                 goto err;
2071         }
2072         /* if we just received a transport event,
2073          * flush reset queue and process this reset
2074          */
2075         if (adapter->force_reset_recovery && !list_empty(&adapter->rwi_list)) {
2076                 list_for_each_safe(entry, tmp_entry, &adapter->rwi_list)
2077                         list_del(entry);
2078         }
2079         rwi->reset_reason = reason;
2080         list_add_tail(&rwi->list, &adapter->rwi_list);
2081         mutex_unlock(&adapter->rwi_lock);
2082         adapter->resetting = true;
2083         netdev_dbg(adapter->netdev, "Scheduling reset (reason %d)\n", reason);
2084         schedule_work(&adapter->ibmvnic_reset);
2085
2086         return 0;
2087 err:
2088         if (adapter->wait_for_reset)
2089                 adapter->wait_for_reset = false;
2090         return -ret;
2091 }
2092
2093 static void ibmvnic_tx_timeout(struct net_device *dev)
2094 {
2095         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2096
2097         ibmvnic_reset(adapter, VNIC_RESET_TIMEOUT);
2098 }
2099
2100 static void remove_buff_from_pool(struct ibmvnic_adapter *adapter,
2101                                   struct ibmvnic_rx_buff *rx_buff)
2102 {
2103         struct ibmvnic_rx_pool *pool = &adapter->rx_pool[rx_buff->pool_index];
2104
2105         rx_buff->skb = NULL;
2106
2107         pool->free_map[pool->next_alloc] = (int)(rx_buff - pool->rx_buff);
2108         pool->next_alloc = (pool->next_alloc + 1) % pool->size;
2109
2110         atomic_dec(&pool->available);
2111 }
2112
2113 static int ibmvnic_poll(struct napi_struct *napi, int budget)
2114 {
2115         struct net_device *netdev = napi->dev;
2116         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2117         int scrq_num = (int)(napi - adapter->napi);
2118         int frames_processed = 0;
2119
2120 restart_poll:
2121         while (frames_processed < budget) {
2122                 struct sk_buff *skb;
2123                 struct ibmvnic_rx_buff *rx_buff;
2124                 union sub_crq *next;
2125                 u32 length;
2126                 u16 offset;
2127                 u8 flags = 0;
2128
2129                 if (unlikely(adapter->resetting &&
2130                              adapter->reset_reason != VNIC_RESET_NON_FATAL)) {
2131                         enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2132                         napi_complete_done(napi, frames_processed);
2133                         return frames_processed;
2134                 }
2135
2136                 if (!pending_scrq(adapter, adapter->rx_scrq[scrq_num]))
2137                         break;
2138                 next = ibmvnic_next_scrq(adapter, adapter->rx_scrq[scrq_num]);
2139                 rx_buff =
2140                     (struct ibmvnic_rx_buff *)be64_to_cpu(next->
2141                                                           rx_comp.correlator);
2142                 /* do error checking */
2143                 if (next->rx_comp.rc) {
2144                         netdev_dbg(netdev, "rx buffer returned with rc %x\n",
2145                                    be16_to_cpu(next->rx_comp.rc));
2146                         /* free the entry */
2147                         next->rx_comp.first = 0;
2148                         dev_kfree_skb_any(rx_buff->skb);
2149                         remove_buff_from_pool(adapter, rx_buff);
2150                         continue;
2151                 } else if (!rx_buff->skb) {
2152                         /* free the entry */
2153                         next->rx_comp.first = 0;
2154                         remove_buff_from_pool(adapter, rx_buff);
2155                         continue;
2156                 }
2157
2158                 length = be32_to_cpu(next->rx_comp.len);
2159                 offset = be16_to_cpu(next->rx_comp.off_frame_data);
2160                 flags = next->rx_comp.flags;
2161                 skb = rx_buff->skb;
2162                 skb_copy_to_linear_data(skb, rx_buff->data + offset,
2163                                         length);
2164
2165                 /* VLAN Header has been stripped by the system firmware and
2166                  * needs to be inserted by the driver
2167                  */
2168                 if (adapter->rx_vlan_header_insertion &&
2169                     (flags & IBMVNIC_VLAN_STRIPPED))
2170                         __vlan_hwaccel_put_tag(skb, htons(ETH_P_8021Q),
2171                                                ntohs(next->rx_comp.vlan_tci));
2172
2173                 /* free the entry */
2174                 next->rx_comp.first = 0;
2175                 remove_buff_from_pool(adapter, rx_buff);
2176
2177                 skb_put(skb, length);
2178                 skb->protocol = eth_type_trans(skb, netdev);
2179                 skb_record_rx_queue(skb, scrq_num);
2180
2181                 if (flags & IBMVNIC_IP_CHKSUM_GOOD &&
2182                     flags & IBMVNIC_TCP_UDP_CHKSUM_GOOD) {
2183                         skb->ip_summed = CHECKSUM_UNNECESSARY;
2184                 }
2185
2186                 length = skb->len;
2187                 napi_gro_receive(napi, skb); /* send it up */
2188                 netdev->stats.rx_packets++;
2189                 netdev->stats.rx_bytes += length;
2190                 adapter->rx_stats_buffers[scrq_num].packets++;
2191                 adapter->rx_stats_buffers[scrq_num].bytes += length;
2192                 frames_processed++;
2193         }
2194
2195         if (adapter->state != VNIC_CLOSING)
2196                 replenish_rx_pool(adapter, &adapter->rx_pool[scrq_num]);
2197
2198         if (frames_processed < budget) {
2199                 enable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2200                 napi_complete_done(napi, frames_processed);
2201                 if (pending_scrq(adapter, adapter->rx_scrq[scrq_num]) &&
2202                     napi_reschedule(napi)) {
2203                         disable_scrq_irq(adapter, adapter->rx_scrq[scrq_num]);
2204                         goto restart_poll;
2205                 }
2206         }
2207         return frames_processed;
2208 }
2209
2210 #ifdef CONFIG_NET_POLL_CONTROLLER
2211 static void ibmvnic_netpoll_controller(struct net_device *dev)
2212 {
2213         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2214         int i;
2215
2216         replenish_pools(netdev_priv(dev));
2217         for (i = 0; i < adapter->req_rx_queues; i++)
2218                 ibmvnic_interrupt_rx(adapter->rx_scrq[i]->irq,
2219                                      adapter->rx_scrq[i]);
2220 }
2221 #endif
2222
2223 static int wait_for_reset(struct ibmvnic_adapter *adapter)
2224 {
2225         int rc, ret;
2226
2227         adapter->fallback.mtu = adapter->req_mtu;
2228         adapter->fallback.rx_queues = adapter->req_rx_queues;
2229         adapter->fallback.tx_queues = adapter->req_tx_queues;
2230         adapter->fallback.rx_entries = adapter->req_rx_add_entries_per_subcrq;
2231         adapter->fallback.tx_entries = adapter->req_tx_entries_per_subcrq;
2232
2233         init_completion(&adapter->reset_done);
2234         adapter->wait_for_reset = true;
2235         rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2236         if (rc)
2237                 return rc;
2238         wait_for_completion(&adapter->reset_done);
2239
2240         ret = 0;
2241         if (adapter->reset_done_rc) {
2242                 ret = -EIO;
2243                 adapter->desired.mtu = adapter->fallback.mtu;
2244                 adapter->desired.rx_queues = adapter->fallback.rx_queues;
2245                 adapter->desired.tx_queues = adapter->fallback.tx_queues;
2246                 adapter->desired.rx_entries = adapter->fallback.rx_entries;
2247                 adapter->desired.tx_entries = adapter->fallback.tx_entries;
2248
2249                 init_completion(&adapter->reset_done);
2250                 adapter->wait_for_reset = true;
2251                 rc = ibmvnic_reset(adapter, VNIC_RESET_CHANGE_PARAM);
2252                 if (rc)
2253                         return ret;
2254                 wait_for_completion(&adapter->reset_done);
2255         }
2256         adapter->wait_for_reset = false;
2257
2258         return ret;
2259 }
2260
2261 static int ibmvnic_change_mtu(struct net_device *netdev, int new_mtu)
2262 {
2263         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2264
2265         adapter->desired.mtu = new_mtu + ETH_HLEN;
2266
2267         return wait_for_reset(adapter);
2268 }
2269
2270 static netdev_features_t ibmvnic_features_check(struct sk_buff *skb,
2271                                                 struct net_device *dev,
2272                                                 netdev_features_t features)
2273 {
2274         /* Some backing hardware adapters can not
2275          * handle packets with a MSS less than 224
2276          * or with only one segment.
2277          */
2278         if (skb_is_gso(skb)) {
2279                 if (skb_shinfo(skb)->gso_size < 224 ||
2280                     skb_shinfo(skb)->gso_segs == 1)
2281                         features &= ~NETIF_F_GSO_MASK;
2282         }
2283
2284         return features;
2285 }
2286
2287 static const struct net_device_ops ibmvnic_netdev_ops = {
2288         .ndo_open               = ibmvnic_open,
2289         .ndo_stop               = ibmvnic_close,
2290         .ndo_start_xmit         = ibmvnic_xmit,
2291         .ndo_set_rx_mode        = ibmvnic_set_multi,
2292         .ndo_set_mac_address    = ibmvnic_set_mac,
2293         .ndo_validate_addr      = eth_validate_addr,
2294         .ndo_tx_timeout         = ibmvnic_tx_timeout,
2295 #ifdef CONFIG_NET_POLL_CONTROLLER
2296         .ndo_poll_controller    = ibmvnic_netpoll_controller,
2297 #endif
2298         .ndo_change_mtu         = ibmvnic_change_mtu,
2299         .ndo_features_check     = ibmvnic_features_check,
2300 };
2301
2302 /* ethtool functions */
2303
2304 static int ibmvnic_get_link_ksettings(struct net_device *netdev,
2305                                       struct ethtool_link_ksettings *cmd)
2306 {
2307         u32 supported, advertising;
2308
2309         supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg |
2310                           SUPPORTED_FIBRE);
2311         advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg |
2312                             ADVERTISED_FIBRE);
2313         cmd->base.speed = SPEED_1000;
2314         cmd->base.duplex = DUPLEX_FULL;
2315         cmd->base.port = PORT_FIBRE;
2316         cmd->base.phy_address = 0;
2317         cmd->base.autoneg = AUTONEG_ENABLE;
2318
2319         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.supported,
2320                                                 supported);
2321         ethtool_convert_legacy_u32_to_link_mode(cmd->link_modes.advertising,
2322                                                 advertising);
2323
2324         return 0;
2325 }
2326
2327 static void ibmvnic_get_drvinfo(struct net_device *netdev,
2328                                 struct ethtool_drvinfo *info)
2329 {
2330         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2331
2332         strlcpy(info->driver, ibmvnic_driver_name, sizeof(info->driver));
2333         strlcpy(info->version, IBMVNIC_DRIVER_VERSION, sizeof(info->version));
2334         strlcpy(info->fw_version, adapter->fw_version,
2335                 sizeof(info->fw_version));
2336 }
2337
2338 static u32 ibmvnic_get_msglevel(struct net_device *netdev)
2339 {
2340         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2341
2342         return adapter->msg_enable;
2343 }
2344
2345 static void ibmvnic_set_msglevel(struct net_device *netdev, u32 data)
2346 {
2347         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2348
2349         adapter->msg_enable = data;
2350 }
2351
2352 static u32 ibmvnic_get_link(struct net_device *netdev)
2353 {
2354         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2355
2356         /* Don't need to send a query because we request a logical link up at
2357          * init and then we wait for link state indications
2358          */
2359         return adapter->logical_link_state;
2360 }
2361
2362 static void ibmvnic_get_ringparam(struct net_device *netdev,
2363                                   struct ethtool_ringparam *ring)
2364 {
2365         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2366
2367         ring->rx_max_pending = adapter->max_rx_add_entries_per_subcrq;
2368         ring->tx_max_pending = adapter->max_tx_entries_per_subcrq;
2369         ring->rx_mini_max_pending = 0;
2370         ring->rx_jumbo_max_pending = 0;
2371         ring->rx_pending = adapter->req_rx_add_entries_per_subcrq;
2372         ring->tx_pending = adapter->req_tx_entries_per_subcrq;
2373         ring->rx_mini_pending = 0;
2374         ring->rx_jumbo_pending = 0;
2375 }
2376
2377 static int ibmvnic_set_ringparam(struct net_device *netdev,
2378                                  struct ethtool_ringparam *ring)
2379 {
2380         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2381
2382         if (ring->rx_pending > adapter->max_rx_add_entries_per_subcrq  ||
2383             ring->tx_pending > adapter->max_tx_entries_per_subcrq) {
2384                 netdev_err(netdev, "Invalid request.\n");
2385                 netdev_err(netdev, "Max tx buffers = %llu\n",
2386                            adapter->max_rx_add_entries_per_subcrq);
2387                 netdev_err(netdev, "Max rx buffers = %llu\n",
2388                            adapter->max_tx_entries_per_subcrq);
2389                 return -EINVAL;
2390         }
2391
2392         adapter->desired.rx_entries = ring->rx_pending;
2393         adapter->desired.tx_entries = ring->tx_pending;
2394
2395         return wait_for_reset(adapter);
2396 }
2397
2398 static void ibmvnic_get_channels(struct net_device *netdev,
2399                                  struct ethtool_channels *channels)
2400 {
2401         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2402
2403         channels->max_rx = adapter->max_rx_queues;
2404         channels->max_tx = adapter->max_tx_queues;
2405         channels->max_other = 0;
2406         channels->max_combined = 0;
2407         channels->rx_count = adapter->req_rx_queues;
2408         channels->tx_count = adapter->req_tx_queues;
2409         channels->other_count = 0;
2410         channels->combined_count = 0;
2411 }
2412
2413 static int ibmvnic_set_channels(struct net_device *netdev,
2414                                 struct ethtool_channels *channels)
2415 {
2416         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
2417
2418         adapter->desired.rx_queues = channels->rx_count;
2419         adapter->desired.tx_queues = channels->tx_count;
2420
2421         return wait_for_reset(adapter);
2422 }
2423
2424 static void ibmvnic_get_strings(struct net_device *dev, u32 stringset, u8 *data)
2425 {
2426         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2427         int i;
2428
2429         if (stringset != ETH_SS_STATS)
2430                 return;
2431
2432         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++, data += ETH_GSTRING_LEN)
2433                 memcpy(data, ibmvnic_stats[i].name, ETH_GSTRING_LEN);
2434
2435         for (i = 0; i < adapter->req_tx_queues; i++) {
2436                 snprintf(data, ETH_GSTRING_LEN, "tx%d_packets", i);
2437                 data += ETH_GSTRING_LEN;
2438
2439                 snprintf(data, ETH_GSTRING_LEN, "tx%d_bytes", i);
2440                 data += ETH_GSTRING_LEN;
2441
2442                 snprintf(data, ETH_GSTRING_LEN, "tx%d_dropped_packets", i);
2443                 data += ETH_GSTRING_LEN;
2444         }
2445
2446         for (i = 0; i < adapter->req_rx_queues; i++) {
2447                 snprintf(data, ETH_GSTRING_LEN, "rx%d_packets", i);
2448                 data += ETH_GSTRING_LEN;
2449
2450                 snprintf(data, ETH_GSTRING_LEN, "rx%d_bytes", i);
2451                 data += ETH_GSTRING_LEN;
2452
2453                 snprintf(data, ETH_GSTRING_LEN, "rx%d_interrupts", i);
2454                 data += ETH_GSTRING_LEN;
2455         }
2456 }
2457
2458 static int ibmvnic_get_sset_count(struct net_device *dev, int sset)
2459 {
2460         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2461
2462         switch (sset) {
2463         case ETH_SS_STATS:
2464                 return ARRAY_SIZE(ibmvnic_stats) +
2465                        adapter->req_tx_queues * NUM_TX_STATS +
2466                        adapter->req_rx_queues * NUM_RX_STATS;
2467         default:
2468                 return -EOPNOTSUPP;
2469         }
2470 }
2471
2472 static void ibmvnic_get_ethtool_stats(struct net_device *dev,
2473                                       struct ethtool_stats *stats, u64 *data)
2474 {
2475         struct ibmvnic_adapter *adapter = netdev_priv(dev);
2476         union ibmvnic_crq crq;
2477         int i, j;
2478         int rc;
2479
2480         memset(&crq, 0, sizeof(crq));
2481         crq.request_statistics.first = IBMVNIC_CRQ_CMD;
2482         crq.request_statistics.cmd = REQUEST_STATISTICS;
2483         crq.request_statistics.ioba = cpu_to_be32(adapter->stats_token);
2484         crq.request_statistics.len =
2485             cpu_to_be32(sizeof(struct ibmvnic_statistics));
2486
2487         /* Wait for data to be written */
2488         init_completion(&adapter->stats_done);
2489         rc = ibmvnic_send_crq(adapter, &crq);
2490         if (rc)
2491                 return;
2492         wait_for_completion(&adapter->stats_done);
2493
2494         for (i = 0; i < ARRAY_SIZE(ibmvnic_stats); i++)
2495                 data[i] = be64_to_cpu(IBMVNIC_GET_STAT(adapter,
2496                                                 ibmvnic_stats[i].offset));
2497
2498         for (j = 0; j < adapter->req_tx_queues; j++) {
2499                 data[i] = adapter->tx_stats_buffers[j].packets;
2500                 i++;
2501                 data[i] = adapter->tx_stats_buffers[j].bytes;
2502                 i++;
2503                 data[i] = adapter->tx_stats_buffers[j].dropped_packets;
2504                 i++;
2505         }
2506
2507         for (j = 0; j < adapter->req_rx_queues; j++) {
2508                 data[i] = adapter->rx_stats_buffers[j].packets;
2509                 i++;
2510                 data[i] = adapter->rx_stats_buffers[j].bytes;
2511                 i++;
2512                 data[i] = adapter->rx_stats_buffers[j].interrupts;
2513                 i++;
2514         }
2515 }
2516
2517 static const struct ethtool_ops ibmvnic_ethtool_ops = {
2518         .get_drvinfo            = ibmvnic_get_drvinfo,
2519         .get_msglevel           = ibmvnic_get_msglevel,
2520         .set_msglevel           = ibmvnic_set_msglevel,
2521         .get_link               = ibmvnic_get_link,
2522         .get_ringparam          = ibmvnic_get_ringparam,
2523         .set_ringparam          = ibmvnic_set_ringparam,
2524         .get_channels           = ibmvnic_get_channels,
2525         .set_channels           = ibmvnic_set_channels,
2526         .get_strings            = ibmvnic_get_strings,
2527         .get_sset_count         = ibmvnic_get_sset_count,
2528         .get_ethtool_stats      = ibmvnic_get_ethtool_stats,
2529         .get_link_ksettings     = ibmvnic_get_link_ksettings,
2530 };
2531
2532 /* Routines for managing CRQs/sCRQs  */
2533
2534 static int reset_one_sub_crq_queue(struct ibmvnic_adapter *adapter,
2535                                    struct ibmvnic_sub_crq_queue *scrq)
2536 {
2537         int rc;
2538
2539         if (scrq->irq) {
2540                 free_irq(scrq->irq, scrq);
2541                 irq_dispose_mapping(scrq->irq);
2542                 scrq->irq = 0;
2543         }
2544
2545         memset(scrq->msgs, 0, 4 * PAGE_SIZE);
2546         atomic_set(&scrq->used, 0);
2547         scrq->cur = 0;
2548
2549         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2550                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2551         return rc;
2552 }
2553
2554 static int reset_sub_crq_queues(struct ibmvnic_adapter *adapter)
2555 {
2556         int i, rc;
2557
2558         for (i = 0; i < adapter->req_tx_queues; i++) {
2559                 netdev_dbg(adapter->netdev, "Re-setting tx_scrq[%d]\n", i);
2560                 rc = reset_one_sub_crq_queue(adapter, adapter->tx_scrq[i]);
2561                 if (rc)
2562                         return rc;
2563         }
2564
2565         for (i = 0; i < adapter->req_rx_queues; i++) {
2566                 netdev_dbg(adapter->netdev, "Re-setting rx_scrq[%d]\n", i);
2567                 rc = reset_one_sub_crq_queue(adapter, adapter->rx_scrq[i]);
2568                 if (rc)
2569                         return rc;
2570         }
2571
2572         return rc;
2573 }
2574
2575 static void release_sub_crq_queue(struct ibmvnic_adapter *adapter,
2576                                   struct ibmvnic_sub_crq_queue *scrq,
2577                                   bool do_h_free)
2578 {
2579         struct device *dev = &adapter->vdev->dev;
2580         long rc;
2581
2582         netdev_dbg(adapter->netdev, "Releasing sub-CRQ\n");
2583
2584         if (do_h_free) {
2585                 /* Close the sub-crqs */
2586                 do {
2587                         rc = plpar_hcall_norets(H_FREE_SUB_CRQ,
2588                                                 adapter->vdev->unit_address,
2589                                                 scrq->crq_num);
2590                 } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
2591
2592                 if (rc) {
2593                         netdev_err(adapter->netdev,
2594                                    "Failed to release sub-CRQ %16lx, rc = %ld\n",
2595                                    scrq->crq_num, rc);
2596                 }
2597         }
2598
2599         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2600                          DMA_BIDIRECTIONAL);
2601         free_pages((unsigned long)scrq->msgs, 2);
2602         kfree(scrq);
2603 }
2604
2605 static struct ibmvnic_sub_crq_queue *init_sub_crq_queue(struct ibmvnic_adapter
2606                                                         *adapter)
2607 {
2608         struct device *dev = &adapter->vdev->dev;
2609         struct ibmvnic_sub_crq_queue *scrq;
2610         int rc;
2611
2612         scrq = kzalloc(sizeof(*scrq), GFP_KERNEL);
2613         if (!scrq)
2614                 return NULL;
2615
2616         scrq->msgs =
2617                 (union sub_crq *)__get_free_pages(GFP_KERNEL | __GFP_ZERO, 2);
2618         if (!scrq->msgs) {
2619                 dev_warn(dev, "Couldn't allocate crq queue messages page\n");
2620                 goto zero_page_failed;
2621         }
2622
2623         scrq->msg_token = dma_map_single(dev, scrq->msgs, 4 * PAGE_SIZE,
2624                                          DMA_BIDIRECTIONAL);
2625         if (dma_mapping_error(dev, scrq->msg_token)) {
2626                 dev_warn(dev, "Couldn't map crq queue messages page\n");
2627                 goto map_failed;
2628         }
2629
2630         rc = h_reg_sub_crq(adapter->vdev->unit_address, scrq->msg_token,
2631                            4 * PAGE_SIZE, &scrq->crq_num, &scrq->hw_irq);
2632
2633         if (rc == H_RESOURCE)
2634                 rc = ibmvnic_reset_crq(adapter);
2635
2636         if (rc == H_CLOSED) {
2637                 dev_warn(dev, "Partner adapter not ready, waiting.\n");
2638         } else if (rc) {
2639                 dev_warn(dev, "Error %d registering sub-crq\n", rc);
2640                 goto reg_failed;
2641         }
2642
2643         scrq->adapter = adapter;
2644         scrq->size = 4 * PAGE_SIZE / sizeof(*scrq->msgs);
2645         spin_lock_init(&scrq->lock);
2646
2647         netdev_dbg(adapter->netdev,
2648                    "sub-crq initialized, num %lx, hw_irq=%lx, irq=%x\n",
2649                    scrq->crq_num, scrq->hw_irq, scrq->irq);
2650
2651         return scrq;
2652
2653 reg_failed:
2654         dma_unmap_single(dev, scrq->msg_token, 4 * PAGE_SIZE,
2655                          DMA_BIDIRECTIONAL);
2656 map_failed:
2657         free_pages((unsigned long)scrq->msgs, 2);
2658 zero_page_failed:
2659         kfree(scrq);
2660
2661         return NULL;
2662 }
2663
2664 static void release_sub_crqs(struct ibmvnic_adapter *adapter, bool do_h_free)
2665 {
2666         int i;
2667
2668         if (adapter->tx_scrq) {
2669                 for (i = 0; i < adapter->num_active_tx_scrqs; i++) {
2670                         if (!adapter->tx_scrq[i])
2671                                 continue;
2672
2673                         netdev_dbg(adapter->netdev, "Releasing tx_scrq[%d]\n",
2674                                    i);
2675                         if (adapter->tx_scrq[i]->irq) {
2676                                 free_irq(adapter->tx_scrq[i]->irq,
2677                                          adapter->tx_scrq[i]);
2678                                 irq_dispose_mapping(adapter->tx_scrq[i]->irq);
2679                                 adapter->tx_scrq[i]->irq = 0;
2680                         }
2681
2682                         release_sub_crq_queue(adapter, adapter->tx_scrq[i],
2683                                               do_h_free);
2684                 }
2685
2686                 kfree(adapter->tx_scrq);
2687                 adapter->tx_scrq = NULL;
2688                 adapter->num_active_tx_scrqs = 0;
2689         }
2690
2691         if (adapter->rx_scrq) {
2692                 for (i = 0; i < adapter->num_active_rx_scrqs; i++) {
2693                         if (!adapter->rx_scrq[i])
2694                                 continue;
2695
2696                         netdev_dbg(adapter->netdev, "Releasing rx_scrq[%d]\n",
2697                                    i);
2698                         if (adapter->rx_scrq[i]->irq) {
2699                                 free_irq(adapter->rx_scrq[i]->irq,
2700                                          adapter->rx_scrq[i]);
2701                                 irq_dispose_mapping(adapter->rx_scrq[i]->irq);
2702                                 adapter->rx_scrq[i]->irq = 0;
2703                         }
2704
2705                         release_sub_crq_queue(adapter, adapter->rx_scrq[i],
2706                                               do_h_free);
2707                 }
2708
2709                 kfree(adapter->rx_scrq);
2710                 adapter->rx_scrq = NULL;
2711                 adapter->num_active_rx_scrqs = 0;
2712         }
2713 }
2714
2715 static int disable_scrq_irq(struct ibmvnic_adapter *adapter,
2716                             struct ibmvnic_sub_crq_queue *scrq)
2717 {
2718         struct device *dev = &adapter->vdev->dev;
2719         unsigned long rc;
2720
2721         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2722                                 H_DISABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2723         if (rc)
2724                 dev_err(dev, "Couldn't disable scrq irq 0x%lx. rc=%ld\n",
2725                         scrq->hw_irq, rc);
2726         return rc;
2727 }
2728
2729 static int enable_scrq_irq(struct ibmvnic_adapter *adapter,
2730                            struct ibmvnic_sub_crq_queue *scrq)
2731 {
2732         struct device *dev = &adapter->vdev->dev;
2733         unsigned long rc;
2734
2735         if (scrq->hw_irq > 0x100000000ULL) {
2736                 dev_err(dev, "bad hw_irq = %lx\n", scrq->hw_irq);
2737                 return 1;
2738         }
2739
2740         if (adapter->resetting &&
2741             adapter->reset_reason == VNIC_RESET_MOBILITY) {
2742                 u64 val = (0xff000000) | scrq->hw_irq;
2743
2744                 rc = plpar_hcall_norets(H_EOI, val);
2745                 if (rc)
2746                         dev_err(dev, "H_EOI FAILED irq 0x%llx. rc=%ld\n",
2747                                 val, rc);
2748         }
2749
2750         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
2751                                 H_ENABLE_VIO_INTERRUPT, scrq->hw_irq, 0, 0);
2752         if (rc)
2753                 dev_err(dev, "Couldn't enable scrq irq 0x%lx. rc=%ld\n",
2754                         scrq->hw_irq, rc);
2755         return rc;
2756 }
2757
2758 static int ibmvnic_complete_tx(struct ibmvnic_adapter *adapter,
2759                                struct ibmvnic_sub_crq_queue *scrq)
2760 {
2761         struct device *dev = &adapter->vdev->dev;
2762         struct ibmvnic_tx_pool *tx_pool;
2763         struct ibmvnic_tx_buff *txbuff;
2764         union sub_crq *next;
2765         int index;
2766         int i, j;
2767         u8 *first;
2768
2769 restart_loop:
2770         while (pending_scrq(adapter, scrq)) {
2771                 unsigned int pool = scrq->pool_index;
2772                 int num_entries = 0;
2773
2774                 next = ibmvnic_next_scrq(adapter, scrq);
2775                 for (i = 0; i < next->tx_comp.num_comps; i++) {
2776                         if (next->tx_comp.rcs[i]) {
2777                                 dev_err(dev, "tx error %x\n",
2778                                         next->tx_comp.rcs[i]);
2779                                 continue;
2780                         }
2781                         index = be32_to_cpu(next->tx_comp.correlators[i]);
2782                         if (index & IBMVNIC_TSO_POOL_MASK) {
2783                                 tx_pool = &adapter->tso_pool[pool];
2784                                 index &= ~IBMVNIC_TSO_POOL_MASK;
2785                         } else {
2786                                 tx_pool = &adapter->tx_pool[pool];
2787                         }
2788
2789                         txbuff = &tx_pool->tx_buff[index];
2790
2791                         for (j = 0; j < IBMVNIC_MAX_FRAGS_PER_CRQ; j++) {
2792                                 if (!txbuff->data_dma[j])
2793                                         continue;
2794
2795                                 txbuff->data_dma[j] = 0;
2796                         }
2797                         /* if sub_crq was sent indirectly */
2798                         first = &txbuff->indir_arr[0].generic.first;
2799                         if (*first == IBMVNIC_CRQ_CMD) {
2800                                 dma_unmap_single(dev, txbuff->indir_dma,
2801                                                  sizeof(txbuff->indir_arr),
2802                                                  DMA_TO_DEVICE);
2803                                 *first = 0;
2804                         }
2805
2806                         if (txbuff->last_frag) {
2807                                 dev_kfree_skb_any(txbuff->skb);
2808                                 txbuff->skb = NULL;
2809                         }
2810
2811                         num_entries += txbuff->num_entries;
2812
2813                         tx_pool->free_map[tx_pool->producer_index] = index;
2814                         tx_pool->producer_index =
2815                                 (tx_pool->producer_index + 1) %
2816                                         tx_pool->num_buffers;
2817                 }
2818                 /* remove tx_comp scrq*/
2819                 next->tx_comp.first = 0;
2820
2821                 if (atomic_sub_return(num_entries, &scrq->used) <=
2822                     (adapter->req_tx_entries_per_subcrq / 2) &&
2823                     __netif_subqueue_stopped(adapter->netdev,
2824                                              scrq->pool_index)) {
2825                         netif_wake_subqueue(adapter->netdev, scrq->pool_index);
2826                         netdev_dbg(adapter->netdev, "Started queue %d\n",
2827                                    scrq->pool_index);
2828                 }
2829         }
2830
2831         enable_scrq_irq(adapter, scrq);
2832
2833         if (pending_scrq(adapter, scrq)) {
2834                 disable_scrq_irq(adapter, scrq);
2835                 goto restart_loop;
2836         }
2837
2838         return 0;
2839 }
2840
2841 static irqreturn_t ibmvnic_interrupt_tx(int irq, void *instance)
2842 {
2843         struct ibmvnic_sub_crq_queue *scrq = instance;
2844         struct ibmvnic_adapter *adapter = scrq->adapter;
2845
2846         disable_scrq_irq(adapter, scrq);
2847         ibmvnic_complete_tx(adapter, scrq);
2848
2849         return IRQ_HANDLED;
2850 }
2851
2852 static irqreturn_t ibmvnic_interrupt_rx(int irq, void *instance)
2853 {
2854         struct ibmvnic_sub_crq_queue *scrq = instance;
2855         struct ibmvnic_adapter *adapter = scrq->adapter;
2856
2857         /* When booting a kdump kernel we can hit pending interrupts
2858          * prior to completing driver initialization.
2859          */
2860         if (unlikely(adapter->state != VNIC_OPEN))
2861                 return IRQ_NONE;
2862
2863         adapter->rx_stats_buffers[scrq->scrq_num].interrupts++;
2864
2865         if (napi_schedule_prep(&adapter->napi[scrq->scrq_num])) {
2866                 disable_scrq_irq(adapter, scrq);
2867                 __napi_schedule(&adapter->napi[scrq->scrq_num]);
2868         }
2869
2870         return IRQ_HANDLED;
2871 }
2872
2873 static int init_sub_crq_irqs(struct ibmvnic_adapter *adapter)
2874 {
2875         struct device *dev = &adapter->vdev->dev;
2876         struct ibmvnic_sub_crq_queue *scrq;
2877         int i = 0, j = 0;
2878         int rc = 0;
2879
2880         for (i = 0; i < adapter->req_tx_queues; i++) {
2881                 netdev_dbg(adapter->netdev, "Initializing tx_scrq[%d] irq\n",
2882                            i);
2883                 scrq = adapter->tx_scrq[i];
2884                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2885
2886                 if (!scrq->irq) {
2887                         rc = -EINVAL;
2888                         dev_err(dev, "Error mapping irq\n");
2889                         goto req_tx_irq_failed;
2890                 }
2891
2892                 rc = request_irq(scrq->irq, ibmvnic_interrupt_tx,
2893                                  0, "ibmvnic_tx", scrq);
2894
2895                 if (rc) {
2896                         dev_err(dev, "Couldn't register tx irq 0x%x. rc=%d\n",
2897                                 scrq->irq, rc);
2898                         irq_dispose_mapping(scrq->irq);
2899                         goto req_tx_irq_failed;
2900                 }
2901         }
2902
2903         for (i = 0; i < adapter->req_rx_queues; i++) {
2904                 netdev_dbg(adapter->netdev, "Initializing rx_scrq[%d] irq\n",
2905                            i);
2906                 scrq = adapter->rx_scrq[i];
2907                 scrq->irq = irq_create_mapping(NULL, scrq->hw_irq);
2908                 if (!scrq->irq) {
2909                         rc = -EINVAL;
2910                         dev_err(dev, "Error mapping irq\n");
2911                         goto req_rx_irq_failed;
2912                 }
2913                 rc = request_irq(scrq->irq, ibmvnic_interrupt_rx,
2914                                  0, "ibmvnic_rx", scrq);
2915                 if (rc) {
2916                         dev_err(dev, "Couldn't register rx irq 0x%x. rc=%d\n",
2917                                 scrq->irq, rc);
2918                         irq_dispose_mapping(scrq->irq);
2919                         goto req_rx_irq_failed;
2920                 }
2921         }
2922         return rc;
2923
2924 req_rx_irq_failed:
2925         for (j = 0; j < i; j++) {
2926                 free_irq(adapter->rx_scrq[j]->irq, adapter->rx_scrq[j]);
2927                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2928         }
2929         i = adapter->req_tx_queues;
2930 req_tx_irq_failed:
2931         for (j = 0; j < i; j++) {
2932                 free_irq(adapter->tx_scrq[j]->irq, adapter->tx_scrq[j]);
2933                 irq_dispose_mapping(adapter->rx_scrq[j]->irq);
2934         }
2935         release_sub_crqs(adapter, 1);
2936         return rc;
2937 }
2938
2939 static int init_sub_crqs(struct ibmvnic_adapter *adapter)
2940 {
2941         struct device *dev = &adapter->vdev->dev;
2942         struct ibmvnic_sub_crq_queue **allqueues;
2943         int registered_queues = 0;
2944         int total_queues;
2945         int more = 0;
2946         int i;
2947
2948         total_queues = adapter->req_tx_queues + adapter->req_rx_queues;
2949
2950         allqueues = kcalloc(total_queues, sizeof(*allqueues), GFP_KERNEL);
2951         if (!allqueues)
2952                 return -1;
2953
2954         for (i = 0; i < total_queues; i++) {
2955                 allqueues[i] = init_sub_crq_queue(adapter);
2956                 if (!allqueues[i]) {
2957                         dev_warn(dev, "Couldn't allocate all sub-crqs\n");
2958                         break;
2959                 }
2960                 registered_queues++;
2961         }
2962
2963         /* Make sure we were able to register the minimum number of queues */
2964         if (registered_queues <
2965             adapter->min_tx_queues + adapter->min_rx_queues) {
2966                 dev_err(dev, "Fatal: Couldn't init  min number of sub-crqs\n");
2967                 goto tx_failed;
2968         }
2969
2970         /* Distribute the failed allocated queues*/
2971         for (i = 0; i < total_queues - registered_queues + more ; i++) {
2972                 netdev_dbg(adapter->netdev, "Reducing number of queues\n");
2973                 switch (i % 3) {
2974                 case 0:
2975                         if (adapter->req_rx_queues > adapter->min_rx_queues)
2976                                 adapter->req_rx_queues--;
2977                         else
2978                                 more++;
2979                         break;
2980                 case 1:
2981                         if (adapter->req_tx_queues > adapter->min_tx_queues)
2982                                 adapter->req_tx_queues--;
2983                         else
2984                                 more++;
2985                         break;
2986                 }
2987         }
2988
2989         adapter->tx_scrq = kcalloc(adapter->req_tx_queues,
2990                                    sizeof(*adapter->tx_scrq), GFP_KERNEL);
2991         if (!adapter->tx_scrq)
2992                 goto tx_failed;
2993
2994         for (i = 0; i < adapter->req_tx_queues; i++) {
2995                 adapter->tx_scrq[i] = allqueues[i];
2996                 adapter->tx_scrq[i]->pool_index = i;
2997                 adapter->num_active_tx_scrqs++;
2998         }
2999
3000         adapter->rx_scrq = kcalloc(adapter->req_rx_queues,
3001                                    sizeof(*adapter->rx_scrq), GFP_KERNEL);
3002         if (!adapter->rx_scrq)
3003                 goto rx_failed;
3004
3005         for (i = 0; i < adapter->req_rx_queues; i++) {
3006                 adapter->rx_scrq[i] = allqueues[i + adapter->req_tx_queues];
3007                 adapter->rx_scrq[i]->scrq_num = i;
3008                 adapter->num_active_rx_scrqs++;
3009         }
3010
3011         kfree(allqueues);
3012         return 0;
3013
3014 rx_failed:
3015         kfree(adapter->tx_scrq);
3016         adapter->tx_scrq = NULL;
3017 tx_failed:
3018         for (i = 0; i < registered_queues; i++)
3019                 release_sub_crq_queue(adapter, allqueues[i], 1);
3020         kfree(allqueues);
3021         return -1;
3022 }
3023
3024 static void ibmvnic_send_req_caps(struct ibmvnic_adapter *adapter, int retry)
3025 {
3026         struct device *dev = &adapter->vdev->dev;
3027         union ibmvnic_crq crq;
3028         int max_entries;
3029
3030         if (!retry) {
3031                 /* Sub-CRQ entries are 32 byte long */
3032                 int entries_page = 4 * PAGE_SIZE / (sizeof(u64) * 4);
3033
3034                 if (adapter->min_tx_entries_per_subcrq > entries_page ||
3035                     adapter->min_rx_add_entries_per_subcrq > entries_page) {
3036                         dev_err(dev, "Fatal, invalid entries per sub-crq\n");
3037                         return;
3038                 }
3039
3040                 if (adapter->desired.mtu)
3041                         adapter->req_mtu = adapter->desired.mtu;
3042                 else
3043                         adapter->req_mtu = adapter->netdev->mtu + ETH_HLEN;
3044
3045                 if (!adapter->desired.tx_entries)
3046                         adapter->desired.tx_entries =
3047                                         adapter->max_tx_entries_per_subcrq;
3048                 if (!adapter->desired.rx_entries)
3049                         adapter->desired.rx_entries =
3050                                         adapter->max_rx_add_entries_per_subcrq;
3051
3052                 max_entries = IBMVNIC_MAX_LTB_SIZE /
3053                               (adapter->req_mtu + IBMVNIC_BUFFER_HLEN);
3054
3055                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3056                         adapter->desired.tx_entries > IBMVNIC_MAX_LTB_SIZE) {
3057                         adapter->desired.tx_entries = max_entries;
3058                 }
3059
3060                 if ((adapter->req_mtu + IBMVNIC_BUFFER_HLEN) *
3061                         adapter->desired.rx_entries > IBMVNIC_MAX_LTB_SIZE) {
3062                         adapter->desired.rx_entries = max_entries;
3063                 }
3064
3065                 if (adapter->desired.tx_entries)
3066                         adapter->req_tx_entries_per_subcrq =
3067                                         adapter->desired.tx_entries;
3068                 else
3069                         adapter->req_tx_entries_per_subcrq =
3070                                         adapter->max_tx_entries_per_subcrq;
3071
3072                 if (adapter->desired.rx_entries)
3073                         adapter->req_rx_add_entries_per_subcrq =
3074                                         adapter->desired.rx_entries;
3075                 else
3076                         adapter->req_rx_add_entries_per_subcrq =
3077                                         adapter->max_rx_add_entries_per_subcrq;
3078
3079                 if (adapter->desired.tx_queues)
3080                         adapter->req_tx_queues =
3081                                         adapter->desired.tx_queues;
3082                 else
3083                         adapter->req_tx_queues =
3084                                         adapter->opt_tx_comp_sub_queues;
3085
3086                 if (adapter->desired.rx_queues)
3087                         adapter->req_rx_queues =
3088                                         adapter->desired.rx_queues;
3089                 else
3090                         adapter->req_rx_queues =
3091                                         adapter->opt_rx_comp_queues;
3092
3093                 adapter->req_rx_add_queues = adapter->max_rx_add_queues;
3094         }
3095
3096         memset(&crq, 0, sizeof(crq));
3097         crq.request_capability.first = IBMVNIC_CRQ_CMD;
3098         crq.request_capability.cmd = REQUEST_CAPABILITY;
3099
3100         crq.request_capability.capability = cpu_to_be16(REQ_TX_QUEUES);
3101         crq.request_capability.number = cpu_to_be64(adapter->req_tx_queues);
3102         atomic_inc(&adapter->running_cap_crqs);
3103         ibmvnic_send_crq(adapter, &crq);
3104
3105         crq.request_capability.capability = cpu_to_be16(REQ_RX_QUEUES);
3106         crq.request_capability.number = cpu_to_be64(adapter->req_rx_queues);
3107         atomic_inc(&adapter->running_cap_crqs);
3108         ibmvnic_send_crq(adapter, &crq);
3109
3110         crq.request_capability.capability = cpu_to_be16(REQ_RX_ADD_QUEUES);
3111         crq.request_capability.number = cpu_to_be64(adapter->req_rx_add_queues);
3112         atomic_inc(&adapter->running_cap_crqs);
3113         ibmvnic_send_crq(adapter, &crq);
3114
3115         crq.request_capability.capability =
3116             cpu_to_be16(REQ_TX_ENTRIES_PER_SUBCRQ);
3117         crq.request_capability.number =
3118             cpu_to_be64(adapter->req_tx_entries_per_subcrq);
3119         atomic_inc(&adapter->running_cap_crqs);
3120         ibmvnic_send_crq(adapter, &crq);
3121
3122         crq.request_capability.capability =
3123             cpu_to_be16(REQ_RX_ADD_ENTRIES_PER_SUBCRQ);
3124         crq.request_capability.number =
3125             cpu_to_be64(adapter->req_rx_add_entries_per_subcrq);
3126         atomic_inc(&adapter->running_cap_crqs);
3127         ibmvnic_send_crq(adapter, &crq);
3128
3129         crq.request_capability.capability = cpu_to_be16(REQ_MTU);
3130         crq.request_capability.number = cpu_to_be64(adapter->req_mtu);
3131         atomic_inc(&adapter->running_cap_crqs);
3132         ibmvnic_send_crq(adapter, &crq);
3133
3134         if (adapter->netdev->flags & IFF_PROMISC) {
3135                 if (adapter->promisc_supported) {
3136                         crq.request_capability.capability =
3137                             cpu_to_be16(PROMISC_REQUESTED);
3138                         crq.request_capability.number = cpu_to_be64(1);
3139                         atomic_inc(&adapter->running_cap_crqs);
3140                         ibmvnic_send_crq(adapter, &crq);
3141                 }
3142         } else {
3143                 crq.request_capability.capability =
3144                     cpu_to_be16(PROMISC_REQUESTED);
3145                 crq.request_capability.number = cpu_to_be64(0);
3146                 atomic_inc(&adapter->running_cap_crqs);
3147                 ibmvnic_send_crq(adapter, &crq);
3148         }
3149 }
3150
3151 static int pending_scrq(struct ibmvnic_adapter *adapter,
3152                         struct ibmvnic_sub_crq_queue *scrq)
3153 {
3154         union sub_crq *entry = &scrq->msgs[scrq->cur];
3155
3156         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP)
3157                 return 1;
3158         else
3159                 return 0;
3160 }
3161
3162 static union sub_crq *ibmvnic_next_scrq(struct ibmvnic_adapter *adapter,
3163                                         struct ibmvnic_sub_crq_queue *scrq)
3164 {
3165         union sub_crq *entry;
3166         unsigned long flags;
3167
3168         spin_lock_irqsave(&scrq->lock, flags);
3169         entry = &scrq->msgs[scrq->cur];
3170         if (entry->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3171                 if (++scrq->cur == scrq->size)
3172                         scrq->cur = 0;
3173         } else {
3174                 entry = NULL;
3175         }
3176         spin_unlock_irqrestore(&scrq->lock, flags);
3177
3178         return entry;
3179 }
3180
3181 static union ibmvnic_crq *ibmvnic_next_crq(struct ibmvnic_adapter *adapter)
3182 {
3183         struct ibmvnic_crq_queue *queue = &adapter->crq;
3184         union ibmvnic_crq *crq;
3185
3186         crq = &queue->msgs[queue->cur];
3187         if (crq->generic.first & IBMVNIC_CRQ_CMD_RSP) {
3188                 if (++queue->cur == queue->size)
3189                         queue->cur = 0;
3190         } else {
3191                 crq = NULL;
3192         }
3193
3194         return crq;
3195 }
3196
3197 static void print_subcrq_error(struct device *dev, int rc, const char *func)
3198 {
3199         switch (rc) {
3200         case H_PARAMETER:
3201                 dev_warn_ratelimited(dev,
3202                                      "%s failed: Send request is malformed or adapter failover pending. (rc=%d)\n",
3203                                      func, rc);
3204                 break;
3205         case H_CLOSED:
3206                 dev_warn_ratelimited(dev,
3207                                      "%s failed: Backing queue closed. Adapter is down or failover pending. (rc=%d)\n",
3208                                      func, rc);
3209                 break;
3210         default:
3211                 dev_err_ratelimited(dev, "%s failed: (rc=%d)\n", func, rc);
3212                 break;
3213         }
3214 }
3215
3216 static int send_subcrq(struct ibmvnic_adapter *adapter, u64 remote_handle,
3217                        union sub_crq *sub_crq)
3218 {
3219         unsigned int ua = adapter->vdev->unit_address;
3220         struct device *dev = &adapter->vdev->dev;
3221         u64 *u64_crq = (u64 *)sub_crq;
3222         int rc;
3223
3224         netdev_dbg(adapter->netdev,
3225                    "Sending sCRQ %016lx: %016lx %016lx %016lx %016lx\n",
3226                    (unsigned long int)cpu_to_be64(remote_handle),
3227                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3228                    (unsigned long int)cpu_to_be64(u64_crq[1]),
3229                    (unsigned long int)cpu_to_be64(u64_crq[2]),
3230                    (unsigned long int)cpu_to_be64(u64_crq[3]));
3231
3232         /* Make sure the hypervisor sees the complete request */
3233         mb();
3234
3235         rc = plpar_hcall_norets(H_SEND_SUB_CRQ, ua,
3236                                 cpu_to_be64(remote_handle),
3237                                 cpu_to_be64(u64_crq[0]),
3238                                 cpu_to_be64(u64_crq[1]),
3239                                 cpu_to_be64(u64_crq[2]),
3240                                 cpu_to_be64(u64_crq[3]));
3241
3242         if (rc)
3243                 print_subcrq_error(dev, rc, __func__);
3244
3245         return rc;
3246 }
3247
3248 static int send_subcrq_indirect(struct ibmvnic_adapter *adapter,
3249                                 u64 remote_handle, u64 ioba, u64 num_entries)
3250 {
3251         unsigned int ua = adapter->vdev->unit_address;
3252         struct device *dev = &adapter->vdev->dev;
3253         int rc;
3254
3255         /* Make sure the hypervisor sees the complete request */
3256         mb();
3257         rc = plpar_hcall_norets(H_SEND_SUB_CRQ_INDIRECT, ua,
3258                                 cpu_to_be64(remote_handle),
3259                                 ioba, num_entries);
3260
3261         if (rc)
3262                 print_subcrq_error(dev, rc, __func__);
3263
3264         return rc;
3265 }
3266
3267 static int ibmvnic_send_crq(struct ibmvnic_adapter *adapter,
3268                             union ibmvnic_crq *crq)
3269 {
3270         unsigned int ua = adapter->vdev->unit_address;
3271         struct device *dev = &adapter->vdev->dev;
3272         u64 *u64_crq = (u64 *)crq;
3273         int rc;
3274
3275         netdev_dbg(adapter->netdev, "Sending CRQ: %016lx %016lx\n",
3276                    (unsigned long int)cpu_to_be64(u64_crq[0]),
3277                    (unsigned long int)cpu_to_be64(u64_crq[1]));
3278
3279         if (!adapter->crq.active &&
3280             crq->generic.first != IBMVNIC_CRQ_INIT_CMD) {
3281                 dev_warn(dev, "Invalid request detected while CRQ is inactive, possible device state change during reset\n");
3282                 return -EINVAL;
3283         }
3284
3285         /* Make sure the hypervisor sees the complete request */
3286         mb();
3287
3288         rc = plpar_hcall_norets(H_SEND_CRQ, ua,
3289                                 cpu_to_be64(u64_crq[0]),
3290                                 cpu_to_be64(u64_crq[1]));
3291
3292         if (rc) {
3293                 if (rc == H_CLOSED) {
3294                         dev_warn(dev, "CRQ Queue closed\n");
3295                         if (adapter->resetting)
3296                                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3297                 }
3298
3299                 dev_warn(dev, "Send error (rc=%d)\n", rc);
3300         }
3301
3302         return rc;
3303 }
3304
3305 static int ibmvnic_send_crq_init(struct ibmvnic_adapter *adapter)
3306 {
3307         union ibmvnic_crq crq;
3308
3309         memset(&crq, 0, sizeof(crq));
3310         crq.generic.first = IBMVNIC_CRQ_INIT_CMD;
3311         crq.generic.cmd = IBMVNIC_CRQ_INIT;
3312         netdev_dbg(adapter->netdev, "Sending CRQ init\n");
3313
3314         return ibmvnic_send_crq(adapter, &crq);
3315 }
3316
3317 static int send_version_xchg(struct ibmvnic_adapter *adapter)
3318 {
3319         union ibmvnic_crq crq;
3320
3321         memset(&crq, 0, sizeof(crq));
3322         crq.version_exchange.first = IBMVNIC_CRQ_CMD;
3323         crq.version_exchange.cmd = VERSION_EXCHANGE;
3324         crq.version_exchange.version = cpu_to_be16(ibmvnic_version);
3325
3326         return ibmvnic_send_crq(adapter, &crq);
3327 }
3328
3329 struct vnic_login_client_data {
3330         u8      type;
3331         __be16  len;
3332         char    name[];
3333 } __packed;
3334
3335 static int vnic_client_data_len(struct ibmvnic_adapter *adapter)
3336 {
3337         int len;
3338
3339         /* Calculate the amount of buffer space needed for the
3340          * vnic client data in the login buffer. There are four entries,
3341          * OS name, LPAR name, device name, and a null last entry.
3342          */
3343         len = 4 * sizeof(struct vnic_login_client_data);
3344         len += 6; /* "Linux" plus NULL */
3345         len += strlen(utsname()->nodename) + 1;
3346         len += strlen(adapter->netdev->name) + 1;
3347
3348         return len;
3349 }
3350
3351 static void vnic_add_client_data(struct ibmvnic_adapter *adapter,
3352                                  struct vnic_login_client_data *vlcd)
3353 {
3354         const char *os_name = "Linux";
3355         int len;
3356
3357         /* Type 1 - LPAR OS */
3358         vlcd->type = 1;
3359         len = strlen(os_name) + 1;
3360         vlcd->len = cpu_to_be16(len);
3361         strncpy(vlcd->name, os_name, len);
3362         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3363
3364         /* Type 2 - LPAR name */
3365         vlcd->type = 2;
3366         len = strlen(utsname()->nodename) + 1;
3367         vlcd->len = cpu_to_be16(len);
3368         strncpy(vlcd->name, utsname()->nodename, len);
3369         vlcd = (struct vnic_login_client_data *)(vlcd->name + len);
3370
3371         /* Type 3 - device name */
3372         vlcd->type = 3;
3373         len = strlen(adapter->netdev->name) + 1;
3374         vlcd->len = cpu_to_be16(len);
3375         strncpy(vlcd->name, adapter->netdev->name, len);
3376 }
3377
3378 static int send_login(struct ibmvnic_adapter *adapter)
3379 {
3380         struct ibmvnic_login_rsp_buffer *login_rsp_buffer;
3381         struct ibmvnic_login_buffer *login_buffer;
3382         struct device *dev = &adapter->vdev->dev;
3383         dma_addr_t rsp_buffer_token;
3384         dma_addr_t buffer_token;
3385         size_t rsp_buffer_size;
3386         union ibmvnic_crq crq;
3387         size_t buffer_size;
3388         __be64 *tx_list_p;
3389         __be64 *rx_list_p;
3390         int client_data_len;
3391         struct vnic_login_client_data *vlcd;
3392         int i;
3393
3394         if (!adapter->tx_scrq || !adapter->rx_scrq) {
3395                 netdev_err(adapter->netdev,
3396                            "RX or TX queues are not allocated, device login failed\n");
3397                 return -1;
3398         }
3399
3400         release_login_rsp_buffer(adapter);
3401         client_data_len = vnic_client_data_len(adapter);
3402
3403         buffer_size =
3404             sizeof(struct ibmvnic_login_buffer) +
3405             sizeof(u64) * (adapter->req_tx_queues + adapter->req_rx_queues) +
3406             client_data_len;
3407
3408         login_buffer = kzalloc(buffer_size, GFP_ATOMIC);
3409         if (!login_buffer)
3410                 goto buf_alloc_failed;
3411
3412         buffer_token = dma_map_single(dev, login_buffer, buffer_size,
3413                                       DMA_TO_DEVICE);
3414         if (dma_mapping_error(dev, buffer_token)) {
3415                 dev_err(dev, "Couldn't map login buffer\n");
3416                 goto buf_map_failed;
3417         }
3418
3419         rsp_buffer_size = sizeof(struct ibmvnic_login_rsp_buffer) +
3420                           sizeof(u64) * adapter->req_tx_queues +
3421                           sizeof(u64) * adapter->req_rx_queues +
3422                           sizeof(u64) * adapter->req_rx_queues +
3423                           sizeof(u8) * IBMVNIC_TX_DESC_VERSIONS;
3424
3425         login_rsp_buffer = kmalloc(rsp_buffer_size, GFP_ATOMIC);
3426         if (!login_rsp_buffer)
3427                 goto buf_rsp_alloc_failed;
3428
3429         rsp_buffer_token = dma_map_single(dev, login_rsp_buffer,
3430                                           rsp_buffer_size, DMA_FROM_DEVICE);
3431         if (dma_mapping_error(dev, rsp_buffer_token)) {
3432                 dev_err(dev, "Couldn't map login rsp buffer\n");
3433                 goto buf_rsp_map_failed;
3434         }
3435
3436         adapter->login_buf = login_buffer;
3437         adapter->login_buf_token = buffer_token;
3438         adapter->login_buf_sz = buffer_size;
3439         adapter->login_rsp_buf = login_rsp_buffer;
3440         adapter->login_rsp_buf_token = rsp_buffer_token;
3441         adapter->login_rsp_buf_sz = rsp_buffer_size;
3442
3443         login_buffer->len = cpu_to_be32(buffer_size);
3444         login_buffer->version = cpu_to_be32(INITIAL_VERSION_LB);
3445         login_buffer->num_txcomp_subcrqs = cpu_to_be32(adapter->req_tx_queues);
3446         login_buffer->off_txcomp_subcrqs =
3447             cpu_to_be32(sizeof(struct ibmvnic_login_buffer));
3448         login_buffer->num_rxcomp_subcrqs = cpu_to_be32(adapter->req_rx_queues);
3449         login_buffer->off_rxcomp_subcrqs =
3450             cpu_to_be32(sizeof(struct ibmvnic_login_buffer) +
3451                         sizeof(u64) * adapter->req_tx_queues);
3452         login_buffer->login_rsp_ioba = cpu_to_be32(rsp_buffer_token);
3453         login_buffer->login_rsp_len = cpu_to_be32(rsp_buffer_size);
3454
3455         tx_list_p = (__be64 *)((char *)login_buffer +
3456                                       sizeof(struct ibmvnic_login_buffer));
3457         rx_list_p = (__be64 *)((char *)login_buffer +
3458                                       sizeof(struct ibmvnic_login_buffer) +
3459                                       sizeof(u64) * adapter->req_tx_queues);
3460
3461         for (i = 0; i < adapter->req_tx_queues; i++) {
3462                 if (adapter->tx_scrq[i]) {
3463                         tx_list_p[i] = cpu_to_be64(adapter->tx_scrq[i]->
3464                                                    crq_num);
3465                 }
3466         }
3467
3468         for (i = 0; i < adapter->req_rx_queues; i++) {
3469                 if (adapter->rx_scrq[i]) {
3470                         rx_list_p[i] = cpu_to_be64(adapter->rx_scrq[i]->
3471                                                    crq_num);
3472                 }
3473         }
3474
3475         /* Insert vNIC login client data */
3476         vlcd = (struct vnic_login_client_data *)
3477                 ((char *)rx_list_p + (sizeof(u64) * adapter->req_rx_queues));
3478         login_buffer->client_data_offset =
3479                         cpu_to_be32((char *)vlcd - (char *)login_buffer);
3480         login_buffer->client_data_len = cpu_to_be32(client_data_len);
3481
3482         vnic_add_client_data(adapter, vlcd);
3483
3484         netdev_dbg(adapter->netdev, "Login Buffer:\n");
3485         for (i = 0; i < (adapter->login_buf_sz - 1) / 8 + 1; i++) {
3486                 netdev_dbg(adapter->netdev, "%016lx\n",
3487                            ((unsigned long int *)(adapter->login_buf))[i]);
3488         }
3489
3490         memset(&crq, 0, sizeof(crq));
3491         crq.login.first = IBMVNIC_CRQ_CMD;
3492         crq.login.cmd = LOGIN;
3493         crq.login.ioba = cpu_to_be32(buffer_token);
3494         crq.login.len = cpu_to_be32(buffer_size);
3495         ibmvnic_send_crq(adapter, &crq);
3496
3497         return 0;
3498
3499 buf_rsp_map_failed:
3500         kfree(login_rsp_buffer);
3501 buf_rsp_alloc_failed:
3502         dma_unmap_single(dev, buffer_token, buffer_size, DMA_TO_DEVICE);
3503 buf_map_failed:
3504         kfree(login_buffer);
3505 buf_alloc_failed:
3506         return -1;
3507 }
3508
3509 static int send_request_map(struct ibmvnic_adapter *adapter, dma_addr_t addr,
3510                             u32 len, u8 map_id)
3511 {
3512         union ibmvnic_crq crq;
3513
3514         memset(&crq, 0, sizeof(crq));
3515         crq.request_map.first = IBMVNIC_CRQ_CMD;
3516         crq.request_map.cmd = REQUEST_MAP;
3517         crq.request_map.map_id = map_id;
3518         crq.request_map.ioba = cpu_to_be32(addr);
3519         crq.request_map.len = cpu_to_be32(len);
3520         return ibmvnic_send_crq(adapter, &crq);
3521 }
3522
3523 static int send_request_unmap(struct ibmvnic_adapter *adapter, u8 map_id)
3524 {
3525         union ibmvnic_crq crq;
3526
3527         memset(&crq, 0, sizeof(crq));
3528         crq.request_unmap.first = IBMVNIC_CRQ_CMD;
3529         crq.request_unmap.cmd = REQUEST_UNMAP;
3530         crq.request_unmap.map_id = map_id;
3531         return ibmvnic_send_crq(adapter, &crq);
3532 }
3533
3534 static void send_map_query(struct ibmvnic_adapter *adapter)
3535 {
3536         union ibmvnic_crq crq;
3537
3538         memset(&crq, 0, sizeof(crq));
3539         crq.query_map.first = IBMVNIC_CRQ_CMD;
3540         crq.query_map.cmd = QUERY_MAP;
3541         ibmvnic_send_crq(adapter, &crq);
3542 }
3543
3544 /* Send a series of CRQs requesting various capabilities of the VNIC server */
3545 static void send_cap_queries(struct ibmvnic_adapter *adapter)
3546 {
3547         union ibmvnic_crq crq;
3548
3549         atomic_set(&adapter->running_cap_crqs, 0);
3550         memset(&crq, 0, sizeof(crq));
3551         crq.query_capability.first = IBMVNIC_CRQ_CMD;
3552         crq.query_capability.cmd = QUERY_CAPABILITY;
3553
3554         crq.query_capability.capability = cpu_to_be16(MIN_TX_QUEUES);
3555         atomic_inc(&adapter->running_cap_crqs);
3556         ibmvnic_send_crq(adapter, &crq);
3557
3558         crq.query_capability.capability = cpu_to_be16(MIN_RX_QUEUES);
3559         atomic_inc(&adapter->running_cap_crqs);
3560         ibmvnic_send_crq(adapter, &crq);
3561
3562         crq.query_capability.capability = cpu_to_be16(MIN_RX_ADD_QUEUES);
3563         atomic_inc(&adapter->running_cap_crqs);
3564         ibmvnic_send_crq(adapter, &crq);
3565
3566         crq.query_capability.capability = cpu_to_be16(MAX_TX_QUEUES);
3567         atomic_inc(&adapter->running_cap_crqs);
3568         ibmvnic_send_crq(adapter, &crq);
3569
3570         crq.query_capability.capability = cpu_to_be16(MAX_RX_QUEUES);
3571         atomic_inc(&adapter->running_cap_crqs);
3572         ibmvnic_send_crq(adapter, &crq);
3573
3574         crq.query_capability.capability = cpu_to_be16(MAX_RX_ADD_QUEUES);
3575         atomic_inc(&adapter->running_cap_crqs);
3576         ibmvnic_send_crq(adapter, &crq);
3577
3578         crq.query_capability.capability =
3579             cpu_to_be16(MIN_TX_ENTRIES_PER_SUBCRQ);
3580         atomic_inc(&adapter->running_cap_crqs);
3581         ibmvnic_send_crq(adapter, &crq);
3582
3583         crq.query_capability.capability =
3584             cpu_to_be16(MIN_RX_ADD_ENTRIES_PER_SUBCRQ);
3585         atomic_inc(&adapter->running_cap_crqs);
3586         ibmvnic_send_crq(adapter, &crq);
3587
3588         crq.query_capability.capability =
3589             cpu_to_be16(MAX_TX_ENTRIES_PER_SUBCRQ);
3590         atomic_inc(&adapter->running_cap_crqs);
3591         ibmvnic_send_crq(adapter, &crq);
3592
3593         crq.query_capability.capability =
3594             cpu_to_be16(MAX_RX_ADD_ENTRIES_PER_SUBCRQ);
3595         atomic_inc(&adapter->running_cap_crqs);
3596         ibmvnic_send_crq(adapter, &crq);
3597
3598         crq.query_capability.capability = cpu_to_be16(TCP_IP_OFFLOAD);
3599         atomic_inc(&adapter->running_cap_crqs);
3600         ibmvnic_send_crq(adapter, &crq);
3601
3602         crq.query_capability.capability = cpu_to_be16(PROMISC_SUPPORTED);
3603         atomic_inc(&adapter->running_cap_crqs);
3604         ibmvnic_send_crq(adapter, &crq);
3605
3606         crq.query_capability.capability = cpu_to_be16(MIN_MTU);
3607         atomic_inc(&adapter->running_cap_crqs);
3608         ibmvnic_send_crq(adapter, &crq);
3609
3610         crq.query_capability.capability = cpu_to_be16(MAX_MTU);
3611         atomic_inc(&adapter->running_cap_crqs);
3612         ibmvnic_send_crq(adapter, &crq);
3613
3614         crq.query_capability.capability = cpu_to_be16(MAX_MULTICAST_FILTERS);
3615         atomic_inc(&adapter->running_cap_crqs);
3616         ibmvnic_send_crq(adapter, &crq);
3617
3618         crq.query_capability.capability = cpu_to_be16(VLAN_HEADER_INSERTION);
3619         atomic_inc(&adapter->running_cap_crqs);
3620         ibmvnic_send_crq(adapter, &crq);
3621
3622         crq.query_capability.capability = cpu_to_be16(RX_VLAN_HEADER_INSERTION);
3623         atomic_inc(&adapter->running_cap_crqs);
3624         ibmvnic_send_crq(adapter, &crq);
3625
3626         crq.query_capability.capability = cpu_to_be16(MAX_TX_SG_ENTRIES);
3627         atomic_inc(&adapter->running_cap_crqs);
3628         ibmvnic_send_crq(adapter, &crq);
3629
3630         crq.query_capability.capability = cpu_to_be16(RX_SG_SUPPORTED);
3631         atomic_inc(&adapter->running_cap_crqs);
3632         ibmvnic_send_crq(adapter, &crq);
3633
3634         crq.query_capability.capability = cpu_to_be16(OPT_TX_COMP_SUB_QUEUES);
3635         atomic_inc(&adapter->running_cap_crqs);
3636         ibmvnic_send_crq(adapter, &crq);
3637
3638         crq.query_capability.capability = cpu_to_be16(OPT_RX_COMP_QUEUES);
3639         atomic_inc(&adapter->running_cap_crqs);
3640         ibmvnic_send_crq(adapter, &crq);
3641
3642         crq.query_capability.capability =
3643                         cpu_to_be16(OPT_RX_BUFADD_Q_PER_RX_COMP_Q);
3644         atomic_inc(&adapter->running_cap_crqs);
3645         ibmvnic_send_crq(adapter, &crq);
3646
3647         crq.query_capability.capability =
3648                         cpu_to_be16(OPT_TX_ENTRIES_PER_SUBCRQ);
3649         atomic_inc(&adapter->running_cap_crqs);
3650         ibmvnic_send_crq(adapter, &crq);
3651
3652         crq.query_capability.capability =
3653                         cpu_to_be16(OPT_RXBA_ENTRIES_PER_SUBCRQ);
3654         atomic_inc(&adapter->running_cap_crqs);
3655         ibmvnic_send_crq(adapter, &crq);
3656
3657         crq.query_capability.capability = cpu_to_be16(TX_RX_DESC_REQ);
3658         atomic_inc(&adapter->running_cap_crqs);
3659         ibmvnic_send_crq(adapter, &crq);
3660 }
3661
3662 static void handle_vpd_size_rsp(union ibmvnic_crq *crq,
3663                                 struct ibmvnic_adapter *adapter)
3664 {
3665         struct device *dev = &adapter->vdev->dev;
3666
3667         if (crq->get_vpd_size_rsp.rc.code) {
3668                 dev_err(dev, "Error retrieving VPD size, rc=%x\n",
3669                         crq->get_vpd_size_rsp.rc.code);
3670                 complete(&adapter->fw_done);
3671                 return;
3672         }
3673
3674         adapter->vpd->len = be64_to_cpu(crq->get_vpd_size_rsp.len);
3675         complete(&adapter->fw_done);
3676 }
3677
3678 static void handle_vpd_rsp(union ibmvnic_crq *crq,
3679                            struct ibmvnic_adapter *adapter)
3680 {
3681         struct device *dev = &adapter->vdev->dev;
3682         unsigned char *substr = NULL;
3683         u8 fw_level_len = 0;
3684
3685         memset(adapter->fw_version, 0, 32);
3686
3687         dma_unmap_single(dev, adapter->vpd->dma_addr, adapter->vpd->len,
3688                          DMA_FROM_DEVICE);
3689
3690         if (crq->get_vpd_rsp.rc.code) {
3691                 dev_err(dev, "Error retrieving VPD from device, rc=%x\n",
3692                         crq->get_vpd_rsp.rc.code);
3693                 goto complete;
3694         }
3695
3696         /* get the position of the firmware version info
3697          * located after the ASCII 'RM' substring in the buffer
3698          */
3699         substr = strnstr(adapter->vpd->buff, "RM", adapter->vpd->len);
3700         if (!substr) {
3701                 dev_info(dev, "Warning - No FW level has been provided in the VPD buffer by the VIOS Server\n");
3702                 goto complete;
3703         }
3704
3705         /* get length of firmware level ASCII substring */
3706         if ((substr + 2) < (adapter->vpd->buff + adapter->vpd->len)) {
3707                 fw_level_len = *(substr + 2);
3708         } else {
3709                 dev_info(dev, "Length of FW substr extrapolated VDP buff\n");
3710                 goto complete;
3711         }
3712
3713         /* copy firmware version string from vpd into adapter */
3714         if ((substr + 3 + fw_level_len) <
3715             (adapter->vpd->buff + adapter->vpd->len)) {
3716                 strncpy((char *)adapter->fw_version, substr + 3, fw_level_len);
3717         } else {
3718                 dev_info(dev, "FW substr extrapolated VPD buff\n");
3719         }
3720
3721 complete:
3722         if (adapter->fw_version[0] == '\0')
3723                 strncpy((char *)adapter->fw_version, "N/A", 3 * sizeof(char));
3724         complete(&adapter->fw_done);
3725 }
3726
3727 static void handle_query_ip_offload_rsp(struct ibmvnic_adapter *adapter)
3728 {
3729         struct device *dev = &adapter->vdev->dev;
3730         struct ibmvnic_query_ip_offload_buffer *buf = &adapter->ip_offload_buf;
3731         union ibmvnic_crq crq;
3732         int i;
3733
3734         dma_unmap_single(dev, adapter->ip_offload_tok,
3735                          sizeof(adapter->ip_offload_buf), DMA_FROM_DEVICE);
3736
3737         netdev_dbg(adapter->netdev, "Query IP Offload Buffer:\n");
3738         for (i = 0; i < (sizeof(adapter->ip_offload_buf) - 1) / 8 + 1; i++)
3739                 netdev_dbg(adapter->netdev, "%016lx\n",
3740                            ((unsigned long int *)(buf))[i]);
3741
3742         netdev_dbg(adapter->netdev, "ipv4_chksum = %d\n", buf->ipv4_chksum);
3743         netdev_dbg(adapter->netdev, "ipv6_chksum = %d\n", buf->ipv6_chksum);
3744         netdev_dbg(adapter->netdev, "tcp_ipv4_chksum = %d\n",
3745                    buf->tcp_ipv4_chksum);
3746         netdev_dbg(adapter->netdev, "tcp_ipv6_chksum = %d\n",
3747                    buf->tcp_ipv6_chksum);
3748         netdev_dbg(adapter->netdev, "udp_ipv4_chksum = %d\n",
3749                    buf->udp_ipv4_chksum);
3750         netdev_dbg(adapter->netdev, "udp_ipv6_chksum = %d\n",
3751                    buf->udp_ipv6_chksum);
3752         netdev_dbg(adapter->netdev, "large_tx_ipv4 = %d\n",
3753                    buf->large_tx_ipv4);
3754         netdev_dbg(adapter->netdev, "large_tx_ipv6 = %d\n",
3755                    buf->large_tx_ipv6);
3756         netdev_dbg(adapter->netdev, "large_rx_ipv4 = %d\n",
3757                    buf->large_rx_ipv4);
3758         netdev_dbg(adapter->netdev, "large_rx_ipv6 = %d\n",
3759                    buf->large_rx_ipv6);
3760         netdev_dbg(adapter->netdev, "max_ipv4_hdr_sz = %d\n",
3761                    buf->max_ipv4_header_size);
3762         netdev_dbg(adapter->netdev, "max_ipv6_hdr_sz = %d\n",
3763                    buf->max_ipv6_header_size);
3764         netdev_dbg(adapter->netdev, "max_tcp_hdr_size = %d\n",
3765                    buf->max_tcp_header_size);
3766         netdev_dbg(adapter->netdev, "max_udp_hdr_size = %d\n",
3767                    buf->max_udp_header_size);
3768         netdev_dbg(adapter->netdev, "max_large_tx_size = %d\n",
3769                    buf->max_large_tx_size);
3770         netdev_dbg(adapter->netdev, "max_large_rx_size = %d\n",
3771                    buf->max_large_rx_size);
3772         netdev_dbg(adapter->netdev, "ipv6_ext_hdr = %d\n",
3773                    buf->ipv6_extension_header);
3774         netdev_dbg(adapter->netdev, "tcp_pseudosum_req = %d\n",
3775                    buf->tcp_pseudosum_req);
3776         netdev_dbg(adapter->netdev, "num_ipv6_ext_hd = %d\n",
3777                    buf->num_ipv6_ext_headers);
3778         netdev_dbg(adapter->netdev, "off_ipv6_ext_hd = %d\n",
3779                    buf->off_ipv6_ext_headers);
3780
3781         adapter->ip_offload_ctrl_tok =
3782             dma_map_single(dev, &adapter->ip_offload_ctrl,
3783                            sizeof(adapter->ip_offload_ctrl), DMA_TO_DEVICE);
3784
3785         if (dma_mapping_error(dev, adapter->ip_offload_ctrl_tok)) {
3786                 dev_err(dev, "Couldn't map ip offload control buffer\n");
3787                 return;
3788         }
3789
3790         adapter->ip_offload_ctrl.len =
3791             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3792         adapter->ip_offload_ctrl.version = cpu_to_be32(INITIAL_VERSION_IOB);
3793         adapter->ip_offload_ctrl.ipv4_chksum = buf->ipv4_chksum;
3794         adapter->ip_offload_ctrl.ipv6_chksum = buf->ipv6_chksum;
3795         adapter->ip_offload_ctrl.tcp_ipv4_chksum = buf->tcp_ipv4_chksum;
3796         adapter->ip_offload_ctrl.udp_ipv4_chksum = buf->udp_ipv4_chksum;
3797         adapter->ip_offload_ctrl.tcp_ipv6_chksum = buf->tcp_ipv6_chksum;
3798         adapter->ip_offload_ctrl.udp_ipv6_chksum = buf->udp_ipv6_chksum;
3799         adapter->ip_offload_ctrl.large_tx_ipv4 = buf->large_tx_ipv4;
3800         adapter->ip_offload_ctrl.large_tx_ipv6 = buf->large_tx_ipv6;
3801
3802         /* large_rx disabled for now, additional features needed */
3803         adapter->ip_offload_ctrl.large_rx_ipv4 = 0;
3804         adapter->ip_offload_ctrl.large_rx_ipv6 = 0;
3805
3806         adapter->netdev->features = NETIF_F_SG | NETIF_F_GSO;
3807
3808         if (buf->tcp_ipv4_chksum || buf->udp_ipv4_chksum)
3809                 adapter->netdev->features |= NETIF_F_IP_CSUM;
3810
3811         if (buf->tcp_ipv6_chksum || buf->udp_ipv6_chksum)
3812                 adapter->netdev->features |= NETIF_F_IPV6_CSUM;
3813
3814         if ((adapter->netdev->features &
3815             (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM)))
3816                 adapter->netdev->features |= NETIF_F_RXCSUM;
3817
3818         if (buf->large_tx_ipv4)
3819                 adapter->netdev->features |= NETIF_F_TSO;
3820         if (buf->large_tx_ipv6)
3821                 adapter->netdev->features |= NETIF_F_TSO6;
3822
3823         adapter->netdev->hw_features |= adapter->netdev->features;
3824
3825         memset(&crq, 0, sizeof(crq));
3826         crq.control_ip_offload.first = IBMVNIC_CRQ_CMD;
3827         crq.control_ip_offload.cmd = CONTROL_IP_OFFLOAD;
3828         crq.control_ip_offload.len =
3829             cpu_to_be32(sizeof(adapter->ip_offload_ctrl));
3830         crq.control_ip_offload.ioba = cpu_to_be32(adapter->ip_offload_ctrl_tok);
3831         ibmvnic_send_crq(adapter, &crq);
3832 }
3833
3834 static const char *ibmvnic_fw_err_cause(u16 cause)
3835 {
3836         switch (cause) {
3837         case ADAPTER_PROBLEM:
3838                 return "adapter problem";
3839         case BUS_PROBLEM:
3840                 return "bus problem";
3841         case FW_PROBLEM:
3842                 return "firmware problem";
3843         case DD_PROBLEM:
3844                 return "device driver problem";
3845         case EEH_RECOVERY:
3846                 return "EEH recovery";
3847         case FW_UPDATED:
3848                 return "firmware updated";
3849         case LOW_MEMORY:
3850                 return "low Memory";
3851         default:
3852                 return "unknown";
3853         }
3854 }
3855
3856 static void handle_error_indication(union ibmvnic_crq *crq,
3857                                     struct ibmvnic_adapter *adapter)
3858 {
3859         struct device *dev = &adapter->vdev->dev;
3860         u16 cause;
3861
3862         cause = be16_to_cpu(crq->error_indication.error_cause);
3863
3864         dev_warn_ratelimited(dev,
3865                              "Firmware reports %serror, cause: %s. Starting recovery...\n",
3866                              crq->error_indication.flags
3867                                 & IBMVNIC_FATAL_ERROR ? "FATAL " : "",
3868                              ibmvnic_fw_err_cause(cause));
3869
3870         if (crq->error_indication.flags & IBMVNIC_FATAL_ERROR)
3871                 ibmvnic_reset(adapter, VNIC_RESET_FATAL);
3872         else
3873                 ibmvnic_reset(adapter, VNIC_RESET_NON_FATAL);
3874 }
3875
3876 static int handle_change_mac_rsp(union ibmvnic_crq *crq,
3877                                  struct ibmvnic_adapter *adapter)
3878 {
3879         struct net_device *netdev = adapter->netdev;
3880         struct device *dev = &adapter->vdev->dev;
3881         long rc;
3882
3883         rc = crq->change_mac_addr_rsp.rc.code;
3884         if (rc) {
3885                 dev_err(dev, "Error %ld in CHANGE_MAC_ADDR_RSP\n", rc);
3886                 goto out;
3887         }
3888         memcpy(netdev->dev_addr, &crq->change_mac_addr_rsp.mac_addr[0],
3889                ETH_ALEN);
3890 out:
3891         complete(&adapter->fw_done);
3892         return rc;
3893 }
3894
3895 static void handle_request_cap_rsp(union ibmvnic_crq *crq,
3896                                    struct ibmvnic_adapter *adapter)
3897 {
3898         struct device *dev = &adapter->vdev->dev;
3899         u64 *req_value;
3900         char *name;
3901
3902         atomic_dec(&adapter->running_cap_crqs);
3903         switch (be16_to_cpu(crq->request_capability_rsp.capability)) {
3904         case REQ_TX_QUEUES:
3905                 req_value = &adapter->req_tx_queues;
3906                 name = "tx";
3907                 break;
3908         case REQ_RX_QUEUES:
3909                 req_value = &adapter->req_rx_queues;
3910                 name = "rx";
3911                 break;
3912         case REQ_RX_ADD_QUEUES:
3913                 req_value = &adapter->req_rx_add_queues;
3914                 name = "rx_add";
3915                 break;
3916         case REQ_TX_ENTRIES_PER_SUBCRQ:
3917                 req_value = &adapter->req_tx_entries_per_subcrq;
3918                 name = "tx_entries_per_subcrq";
3919                 break;
3920         case REQ_RX_ADD_ENTRIES_PER_SUBCRQ:
3921                 req_value = &adapter->req_rx_add_entries_per_subcrq;
3922                 name = "rx_add_entries_per_subcrq";
3923                 break;
3924         case REQ_MTU:
3925                 req_value = &adapter->req_mtu;
3926                 name = "mtu";
3927                 break;
3928         case PROMISC_REQUESTED:
3929                 req_value = &adapter->promisc;
3930                 name = "promisc";
3931                 break;
3932         default:
3933                 dev_err(dev, "Got invalid cap request rsp %d\n",
3934                         crq->request_capability.capability);
3935                 return;
3936         }
3937
3938         switch (crq->request_capability_rsp.rc.code) {
3939         case SUCCESS:
3940                 break;
3941         case PARTIALSUCCESS:
3942                 dev_info(dev, "req=%lld, rsp=%ld in %s queue, retrying.\n",
3943                          *req_value,
3944                          (long int)be64_to_cpu(crq->request_capability_rsp.
3945                                                number), name);
3946
3947                 if (be16_to_cpu(crq->request_capability_rsp.capability) ==
3948                     REQ_MTU) {
3949                         pr_err("mtu of %llu is not supported. Reverting.\n",
3950                                *req_value);
3951                         *req_value = adapter->fallback.mtu;
3952                 } else {
3953                         *req_value =
3954                                 be64_to_cpu(crq->request_capability_rsp.number);
3955                 }
3956
3957                 ibmvnic_send_req_caps(adapter, 1);
3958                 return;
3959         default:
3960                 dev_err(dev, "Error %d in request cap rsp\n",
3961                         crq->request_capability_rsp.rc.code);
3962                 return;
3963         }
3964
3965         /* Done receiving requested capabilities, query IP offload support */
3966         if (atomic_read(&adapter->running_cap_crqs) == 0) {
3967                 union ibmvnic_crq newcrq;
3968                 int buf_sz = sizeof(struct ibmvnic_query_ip_offload_buffer);
3969                 struct ibmvnic_query_ip_offload_buffer *ip_offload_buf =
3970                     &adapter->ip_offload_buf;
3971
3972                 adapter->wait_capability = false;
3973                 adapter->ip_offload_tok = dma_map_single(dev, ip_offload_buf,
3974                                                          buf_sz,
3975                                                          DMA_FROM_DEVICE);
3976
3977                 if (dma_mapping_error(dev, adapter->ip_offload_tok)) {
3978                         if (!firmware_has_feature(FW_FEATURE_CMO))
3979                                 dev_err(dev, "Couldn't map offload buffer\n");
3980                         return;
3981                 }
3982
3983                 memset(&newcrq, 0, sizeof(newcrq));
3984                 newcrq.query_ip_offload.first = IBMVNIC_CRQ_CMD;
3985                 newcrq.query_ip_offload.cmd = QUERY_IP_OFFLOAD;
3986                 newcrq.query_ip_offload.len = cpu_to_be32(buf_sz);
3987                 newcrq.query_ip_offload.ioba =
3988                     cpu_to_be32(adapter->ip_offload_tok);
3989
3990                 ibmvnic_send_crq(adapter, &newcrq);
3991         }
3992 }
3993
3994 static int handle_login_rsp(union ibmvnic_crq *login_rsp_crq,
3995                             struct ibmvnic_adapter *adapter)
3996 {
3997         struct device *dev = &adapter->vdev->dev;
3998         struct net_device *netdev = adapter->netdev;
3999         struct ibmvnic_login_rsp_buffer *login_rsp = adapter->login_rsp_buf;
4000         struct ibmvnic_login_buffer *login = adapter->login_buf;
4001         int i;
4002
4003         dma_unmap_single(dev, adapter->login_buf_token, adapter->login_buf_sz,
4004                          DMA_TO_DEVICE);
4005         dma_unmap_single(dev, adapter->login_rsp_buf_token,
4006                          adapter->login_rsp_buf_sz, DMA_FROM_DEVICE);
4007
4008         /* If the number of queues requested can't be allocated by the
4009          * server, the login response will return with code 1. We will need
4010          * to resend the login buffer with fewer queues requested.
4011          */
4012         if (login_rsp_crq->generic.rc.code) {
4013                 adapter->init_done_rc = login_rsp_crq->generic.rc.code;
4014                 complete(&adapter->init_done);
4015                 return 0;
4016         }
4017
4018         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4019
4020         netdev_dbg(adapter->netdev, "Login Response Buffer:\n");
4021         for (i = 0; i < (adapter->login_rsp_buf_sz - 1) / 8 + 1; i++) {
4022                 netdev_dbg(adapter->netdev, "%016lx\n",
4023                            ((unsigned long int *)(adapter->login_rsp_buf))[i]);
4024         }
4025
4026         /* Sanity checks */
4027         if (login->num_txcomp_subcrqs != login_rsp->num_txsubm_subcrqs ||
4028             (be32_to_cpu(login->num_rxcomp_subcrqs) *
4029              adapter->req_rx_add_queues !=
4030              be32_to_cpu(login_rsp->num_rxadd_subcrqs))) {
4031                 dev_err(dev, "FATAL: Inconsistent login and login rsp\n");
4032                 ibmvnic_remove(adapter->vdev);
4033                 return -EIO;
4034         }
4035         release_login_buffer(adapter);
4036         complete(&adapter->init_done);
4037
4038         return 0;
4039 }
4040
4041 static void handle_request_unmap_rsp(union ibmvnic_crq *crq,
4042                                      struct ibmvnic_adapter *adapter)
4043 {
4044         struct device *dev = &adapter->vdev->dev;
4045         long rc;
4046
4047         rc = crq->request_unmap_rsp.rc.code;
4048         if (rc)
4049                 dev_err(dev, "Error %ld in REQUEST_UNMAP_RSP\n", rc);
4050 }
4051
4052 static void handle_query_map_rsp(union ibmvnic_crq *crq,
4053                                  struct ibmvnic_adapter *adapter)
4054 {
4055         struct net_device *netdev = adapter->netdev;
4056         struct device *dev = &adapter->vdev->dev;
4057         long rc;
4058
4059         rc = crq->query_map_rsp.rc.code;
4060         if (rc) {
4061                 dev_err(dev, "Error %ld in QUERY_MAP_RSP\n", rc);
4062                 return;
4063         }
4064         netdev_dbg(netdev, "page_size = %d\ntot_pages = %d\nfree_pages = %d\n",
4065                    crq->query_map_rsp.page_size, crq->query_map_rsp.tot_pages,
4066                    crq->query_map_rsp.free_pages);
4067 }
4068
4069 static void handle_query_cap_rsp(union ibmvnic_crq *crq,
4070                                  struct ibmvnic_adapter *adapter)
4071 {
4072         struct net_device *netdev = adapter->netdev;
4073         struct device *dev = &adapter->vdev->dev;
4074         long rc;
4075
4076         atomic_dec(&adapter->running_cap_crqs);
4077         netdev_dbg(netdev, "Outstanding queries: %d\n",
4078                    atomic_read(&adapter->running_cap_crqs));
4079         rc = crq->query_capability.rc.code;
4080         if (rc) {
4081                 dev_err(dev, "Error %ld in QUERY_CAP_RSP\n", rc);
4082                 goto out;
4083         }
4084
4085         switch (be16_to_cpu(crq->query_capability.capability)) {
4086         case MIN_TX_QUEUES:
4087                 adapter->min_tx_queues =
4088                     be64_to_cpu(crq->query_capability.number);
4089                 netdev_dbg(netdev, "min_tx_queues = %lld\n",
4090                            adapter->min_tx_queues);
4091                 break;
4092         case MIN_RX_QUEUES:
4093                 adapter->min_rx_queues =
4094                     be64_to_cpu(crq->query_capability.number);
4095                 netdev_dbg(netdev, "min_rx_queues = %lld\n",
4096                            adapter->min_rx_queues);
4097                 break;
4098         case MIN_RX_ADD_QUEUES:
4099                 adapter->min_rx_add_queues =
4100                     be64_to_cpu(crq->query_capability.number);
4101                 netdev_dbg(netdev, "min_rx_add_queues = %lld\n",
4102                            adapter->min_rx_add_queues);
4103                 break;
4104         case MAX_TX_QUEUES:
4105                 adapter->max_tx_queues =
4106                     be64_to_cpu(crq->query_capability.number);
4107                 netdev_dbg(netdev, "max_tx_queues = %lld\n",
4108                            adapter->max_tx_queues);
4109                 break;
4110         case MAX_RX_QUEUES:
4111                 adapter->max_rx_queues =
4112                     be64_to_cpu(crq->query_capability.number);
4113                 netdev_dbg(netdev, "max_rx_queues = %lld\n",
4114                            adapter->max_rx_queues);
4115                 break;
4116         case MAX_RX_ADD_QUEUES:
4117                 adapter->max_rx_add_queues =
4118                     be64_to_cpu(crq->query_capability.number);
4119                 netdev_dbg(netdev, "max_rx_add_queues = %lld\n",
4120                            adapter->max_rx_add_queues);
4121                 break;
4122         case MIN_TX_ENTRIES_PER_SUBCRQ:
4123                 adapter->min_tx_entries_per_subcrq =
4124                     be64_to_cpu(crq->query_capability.number);
4125                 netdev_dbg(netdev, "min_tx_entries_per_subcrq = %lld\n",
4126                            adapter->min_tx_entries_per_subcrq);
4127                 break;
4128         case MIN_RX_ADD_ENTRIES_PER_SUBCRQ:
4129                 adapter->min_rx_add_entries_per_subcrq =
4130                     be64_to_cpu(crq->query_capability.number);
4131                 netdev_dbg(netdev, "min_rx_add_entrs_per_subcrq = %lld\n",
4132                            adapter->min_rx_add_entries_per_subcrq);
4133                 break;
4134         case MAX_TX_ENTRIES_PER_SUBCRQ:
4135                 adapter->max_tx_entries_per_subcrq =
4136                     be64_to_cpu(crq->query_capability.number);
4137                 netdev_dbg(netdev, "max_tx_entries_per_subcrq = %lld\n",
4138                            adapter->max_tx_entries_per_subcrq);
4139                 break;
4140         case MAX_RX_ADD_ENTRIES_PER_SUBCRQ:
4141                 adapter->max_rx_add_entries_per_subcrq =
4142                     be64_to_cpu(crq->query_capability.number);
4143                 netdev_dbg(netdev, "max_rx_add_entrs_per_subcrq = %lld\n",
4144                            adapter->max_rx_add_entries_per_subcrq);
4145                 break;
4146         case TCP_IP_OFFLOAD:
4147                 adapter->tcp_ip_offload =
4148                     be64_to_cpu(crq->query_capability.number);
4149                 netdev_dbg(netdev, "tcp_ip_offload = %lld\n",
4150                            adapter->tcp_ip_offload);
4151                 break;
4152         case PROMISC_SUPPORTED:
4153                 adapter->promisc_supported =
4154                     be64_to_cpu(crq->query_capability.number);
4155                 netdev_dbg(netdev, "promisc_supported = %lld\n",
4156                            adapter->promisc_supported);
4157                 break;
4158         case MIN_MTU:
4159                 adapter->min_mtu = be64_to_cpu(crq->query_capability.number);
4160                 netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4161                 netdev_dbg(netdev, "min_mtu = %lld\n", adapter->min_mtu);
4162                 break;
4163         case MAX_MTU:
4164                 adapter->max_mtu = be64_to_cpu(crq->query_capability.number);
4165                 netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4166                 netdev_dbg(netdev, "max_mtu = %lld\n", adapter->max_mtu);
4167                 break;
4168         case MAX_MULTICAST_FILTERS:
4169                 adapter->max_multicast_filters =
4170                     be64_to_cpu(crq->query_capability.number);
4171                 netdev_dbg(netdev, "max_multicast_filters = %lld\n",
4172                            adapter->max_multicast_filters);
4173                 break;
4174         case VLAN_HEADER_INSERTION:
4175                 adapter->vlan_header_insertion =
4176                     be64_to_cpu(crq->query_capability.number);
4177                 if (adapter->vlan_header_insertion)
4178                         netdev->features |= NETIF_F_HW_VLAN_STAG_TX;
4179                 netdev_dbg(netdev, "vlan_header_insertion = %lld\n",
4180                            adapter->vlan_header_insertion);
4181                 break;
4182         case RX_VLAN_HEADER_INSERTION:
4183                 adapter->rx_vlan_header_insertion =
4184                     be64_to_cpu(crq->query_capability.number);
4185                 netdev_dbg(netdev, "rx_vlan_header_insertion = %lld\n",
4186                            adapter->rx_vlan_header_insertion);
4187                 break;
4188         case MAX_TX_SG_ENTRIES:
4189                 adapter->max_tx_sg_entries =
4190                     be64_to_cpu(crq->query_capability.number);
4191                 netdev_dbg(netdev, "max_tx_sg_entries = %lld\n",
4192                            adapter->max_tx_sg_entries);
4193                 break;
4194         case RX_SG_SUPPORTED:
4195                 adapter->rx_sg_supported =
4196                     be64_to_cpu(crq->query_capability.number);
4197                 netdev_dbg(netdev, "rx_sg_supported = %lld\n",
4198                            adapter->rx_sg_supported);
4199                 break;
4200         case OPT_TX_COMP_SUB_QUEUES:
4201                 adapter->opt_tx_comp_sub_queues =
4202                     be64_to_cpu(crq->query_capability.number);
4203                 netdev_dbg(netdev, "opt_tx_comp_sub_queues = %lld\n",
4204                            adapter->opt_tx_comp_sub_queues);
4205                 break;
4206         case OPT_RX_COMP_QUEUES:
4207                 adapter->opt_rx_comp_queues =
4208                     be64_to_cpu(crq->query_capability.number);
4209                 netdev_dbg(netdev, "opt_rx_comp_queues = %lld\n",
4210                            adapter->opt_rx_comp_queues);
4211                 break;
4212         case OPT_RX_BUFADD_Q_PER_RX_COMP_Q:
4213                 adapter->opt_rx_bufadd_q_per_rx_comp_q =
4214                     be64_to_cpu(crq->query_capability.number);
4215                 netdev_dbg(netdev, "opt_rx_bufadd_q_per_rx_comp_q = %lld\n",
4216                            adapter->opt_rx_bufadd_q_per_rx_comp_q);
4217                 break;
4218         case OPT_TX_ENTRIES_PER_SUBCRQ:
4219                 adapter->opt_tx_entries_per_subcrq =
4220                     be64_to_cpu(crq->query_capability.number);
4221                 netdev_dbg(netdev, "opt_tx_entries_per_subcrq = %lld\n",
4222                            adapter->opt_tx_entries_per_subcrq);
4223                 break;
4224         case OPT_RXBA_ENTRIES_PER_SUBCRQ:
4225                 adapter->opt_rxba_entries_per_subcrq =
4226                     be64_to_cpu(crq->query_capability.number);
4227                 netdev_dbg(netdev, "opt_rxba_entries_per_subcrq = %lld\n",
4228                            adapter->opt_rxba_entries_per_subcrq);
4229                 break;
4230         case TX_RX_DESC_REQ:
4231                 adapter->tx_rx_desc_req = crq->query_capability.number;
4232                 netdev_dbg(netdev, "tx_rx_desc_req = %llx\n",
4233                            adapter->tx_rx_desc_req);
4234                 break;
4235
4236         default:
4237                 netdev_err(netdev, "Got invalid cap rsp %d\n",
4238                            crq->query_capability.capability);
4239         }
4240
4241 out:
4242         if (atomic_read(&adapter->running_cap_crqs) == 0) {
4243                 adapter->wait_capability = false;
4244                 ibmvnic_send_req_caps(adapter, 0);
4245         }
4246 }
4247
4248 static void ibmvnic_handle_crq(union ibmvnic_crq *crq,
4249                                struct ibmvnic_adapter *adapter)
4250 {
4251         struct ibmvnic_generic_crq *gen_crq = &crq->generic;
4252         struct net_device *netdev = adapter->netdev;
4253         struct device *dev = &adapter->vdev->dev;
4254         u64 *u64_crq = (u64 *)crq;
4255         long rc;
4256
4257         netdev_dbg(netdev, "Handling CRQ: %016lx %016lx\n",
4258                    (unsigned long int)cpu_to_be64(u64_crq[0]),
4259                    (unsigned long int)cpu_to_be64(u64_crq[1]));
4260         switch (gen_crq->first) {
4261         case IBMVNIC_CRQ_INIT_RSP:
4262                 switch (gen_crq->cmd) {
4263                 case IBMVNIC_CRQ_INIT:
4264                         dev_info(dev, "Partner initialized\n");
4265                         adapter->from_passive_init = true;
4266                         adapter->failover_pending = false;
4267                         if (!completion_done(&adapter->init_done)) {
4268                                 complete(&adapter->init_done);
4269                                 adapter->init_done_rc = -EIO;
4270                         }
4271                         ibmvnic_reset(adapter, VNIC_RESET_FAILOVER);
4272                         break;
4273                 case IBMVNIC_CRQ_INIT_COMPLETE:
4274                         dev_info(dev, "Partner initialization complete\n");
4275                         adapter->crq.active = true;
4276                         send_version_xchg(adapter);
4277                         break;
4278                 default:
4279                         dev_err(dev, "Unknown crq cmd: %d\n", gen_crq->cmd);
4280                 }
4281                 return;
4282         case IBMVNIC_CRQ_XPORT_EVENT:
4283                 netif_carrier_off(netdev);
4284                 adapter->crq.active = false;
4285                 if (adapter->resetting)
4286                         adapter->force_reset_recovery = true;
4287                 if (gen_crq->cmd == IBMVNIC_PARTITION_MIGRATED) {
4288                         dev_info(dev, "Migrated, re-enabling adapter\n");
4289                         ibmvnic_reset(adapter, VNIC_RESET_MOBILITY);
4290                 } else if (gen_crq->cmd == IBMVNIC_DEVICE_FAILOVER) {
4291                         dev_info(dev, "Backing device failover detected\n");
4292                         adapter->failover_pending = true;
4293                 } else {
4294                         /* The adapter lost the connection */
4295                         dev_err(dev, "Virtual Adapter failed (rc=%d)\n",
4296                                 gen_crq->cmd);
4297                         ibmvnic_reset(adapter, VNIC_RESET_FATAL);
4298                 }
4299                 return;
4300         case IBMVNIC_CRQ_CMD_RSP:
4301                 break;
4302         default:
4303                 dev_err(dev, "Got an invalid msg type 0x%02x\n",
4304                         gen_crq->first);
4305                 return;
4306         }
4307
4308         switch (gen_crq->cmd) {
4309         case VERSION_EXCHANGE_RSP:
4310                 rc = crq->version_exchange_rsp.rc.code;
4311                 if (rc) {
4312                         dev_err(dev, "Error %ld in VERSION_EXCHG_RSP\n", rc);
4313                         break;
4314                 }
4315                 dev_info(dev, "Partner protocol version is %d\n",
4316                          crq->version_exchange_rsp.version);
4317                 if (be16_to_cpu(crq->version_exchange_rsp.version) <
4318                     ibmvnic_version)
4319                         ibmvnic_version =
4320                             be16_to_cpu(crq->version_exchange_rsp.version);
4321                 send_cap_queries(adapter);
4322                 break;
4323         case QUERY_CAPABILITY_RSP:
4324                 handle_query_cap_rsp(crq, adapter);
4325                 break;
4326         case QUERY_MAP_RSP:
4327                 handle_query_map_rsp(crq, adapter);
4328                 break;
4329         case REQUEST_MAP_RSP:
4330                 adapter->fw_done_rc = crq->request_map_rsp.rc.code;
4331                 complete(&adapter->fw_done);
4332                 break;
4333         case REQUEST_UNMAP_RSP:
4334                 handle_request_unmap_rsp(crq, adapter);
4335                 break;
4336         case REQUEST_CAPABILITY_RSP:
4337                 handle_request_cap_rsp(crq, adapter);
4338                 break;
4339         case LOGIN_RSP:
4340                 netdev_dbg(netdev, "Got Login Response\n");
4341                 handle_login_rsp(crq, adapter);
4342                 break;
4343         case LOGICAL_LINK_STATE_RSP:
4344                 netdev_dbg(netdev,
4345                            "Got Logical Link State Response, state: %d rc: %d\n",
4346                            crq->logical_link_state_rsp.link_state,
4347                            crq->logical_link_state_rsp.rc.code);
4348                 adapter->logical_link_state =
4349                     crq->logical_link_state_rsp.link_state;
4350                 adapter->init_done_rc = crq->logical_link_state_rsp.rc.code;
4351                 complete(&adapter->init_done);
4352                 break;
4353         case LINK_STATE_INDICATION:
4354                 netdev_dbg(netdev, "Got Logical Link State Indication\n");
4355                 adapter->phys_link_state =
4356                     crq->link_state_indication.phys_link_state;
4357                 adapter->logical_link_state =
4358                     crq->link_state_indication.logical_link_state;
4359                 break;
4360         case CHANGE_MAC_ADDR_RSP:
4361                 netdev_dbg(netdev, "Got MAC address change Response\n");
4362                 adapter->fw_done_rc = handle_change_mac_rsp(crq, adapter);
4363                 break;
4364         case ERROR_INDICATION:
4365                 netdev_dbg(netdev, "Got Error Indication\n");
4366                 handle_error_indication(crq, adapter);
4367                 break;
4368         case REQUEST_STATISTICS_RSP:
4369                 netdev_dbg(netdev, "Got Statistics Response\n");
4370                 complete(&adapter->stats_done);
4371                 break;
4372         case QUERY_IP_OFFLOAD_RSP:
4373                 netdev_dbg(netdev, "Got Query IP offload Response\n");
4374                 handle_query_ip_offload_rsp(adapter);
4375                 break;
4376         case MULTICAST_CTRL_RSP:
4377                 netdev_dbg(netdev, "Got multicast control Response\n");
4378                 break;
4379         case CONTROL_IP_OFFLOAD_RSP:
4380                 netdev_dbg(netdev, "Got Control IP offload Response\n");
4381                 dma_unmap_single(dev, adapter->ip_offload_ctrl_tok,
4382                                  sizeof(adapter->ip_offload_ctrl),
4383                                  DMA_TO_DEVICE);
4384                 complete(&adapter->init_done);
4385                 break;
4386         case COLLECT_FW_TRACE_RSP:
4387                 netdev_dbg(netdev, "Got Collect firmware trace Response\n");
4388                 complete(&adapter->fw_done);
4389                 break;
4390         case GET_VPD_SIZE_RSP:
4391                 handle_vpd_size_rsp(crq, adapter);
4392                 break;
4393         case GET_VPD_RSP:
4394                 handle_vpd_rsp(crq, adapter);
4395                 break;
4396         default:
4397                 netdev_err(netdev, "Got an invalid cmd type 0x%02x\n",
4398                            gen_crq->cmd);
4399         }
4400 }
4401
4402 static irqreturn_t ibmvnic_interrupt(int irq, void *instance)
4403 {
4404         struct ibmvnic_adapter *adapter = instance;
4405
4406         tasklet_schedule(&adapter->tasklet);
4407         return IRQ_HANDLED;
4408 }
4409
4410 static void ibmvnic_tasklet(void *data)
4411 {
4412         struct ibmvnic_adapter *adapter = data;
4413         struct ibmvnic_crq_queue *queue = &adapter->crq;
4414         union ibmvnic_crq *crq;
4415         unsigned long flags;
4416         bool done = false;
4417
4418         spin_lock_irqsave(&queue->lock, flags);
4419         while (!done) {
4420                 /* Pull all the valid messages off the CRQ */
4421                 while ((crq = ibmvnic_next_crq(adapter)) != NULL) {
4422                         ibmvnic_handle_crq(crq, adapter);
4423                         crq->generic.first = 0;
4424                 }
4425
4426                 /* remain in tasklet until all
4427                  * capabilities responses are received
4428                  */
4429                 if (!adapter->wait_capability)
4430                         done = true;
4431         }
4432         /* if capabilities CRQ's were sent in this tasklet, the following
4433          * tasklet must wait until all responses are received
4434          */
4435         if (atomic_read(&adapter->running_cap_crqs) != 0)
4436                 adapter->wait_capability = true;
4437         spin_unlock_irqrestore(&queue->lock, flags);
4438 }
4439
4440 static int ibmvnic_reenable_crq_queue(struct ibmvnic_adapter *adapter)
4441 {
4442         struct vio_dev *vdev = adapter->vdev;
4443         int rc;
4444
4445         do {
4446                 rc = plpar_hcall_norets(H_ENABLE_CRQ, vdev->unit_address);
4447         } while (rc == H_IN_PROGRESS || rc == H_BUSY || H_IS_LONG_BUSY(rc));
4448
4449         if (rc)
4450                 dev_err(&vdev->dev, "Error enabling adapter (rc=%d)\n", rc);
4451
4452         return rc;
4453 }
4454
4455 static int ibmvnic_reset_crq(struct ibmvnic_adapter *adapter)
4456 {
4457         struct ibmvnic_crq_queue *crq = &adapter->crq;
4458         struct device *dev = &adapter->vdev->dev;
4459         struct vio_dev *vdev = adapter->vdev;
4460         int rc;
4461
4462         /* Close the CRQ */
4463         do {
4464                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4465         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4466
4467         /* Clean out the queue */
4468         memset(crq->msgs, 0, PAGE_SIZE);
4469         crq->cur = 0;
4470         crq->active = false;
4471
4472         /* And re-open it again */
4473         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4474                                 crq->msg_token, PAGE_SIZE);
4475
4476         if (rc == H_CLOSED)
4477                 /* Adapter is good, but other end is not ready */
4478                 dev_warn(dev, "Partner adapter not ready\n");
4479         else if (rc != 0)
4480                 dev_warn(dev, "Couldn't register crq (rc=%d)\n", rc);
4481
4482         return rc;
4483 }
4484
4485 static void release_crq_queue(struct ibmvnic_adapter *adapter)
4486 {
4487         struct ibmvnic_crq_queue *crq = &adapter->crq;
4488         struct vio_dev *vdev = adapter->vdev;
4489         long rc;
4490
4491         if (!crq->msgs)
4492                 return;
4493
4494         netdev_dbg(adapter->netdev, "Releasing CRQ\n");
4495         free_irq(vdev->irq, adapter);
4496         tasklet_kill(&adapter->tasklet);
4497         do {
4498                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4499         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4500
4501         dma_unmap_single(&vdev->dev, crq->msg_token, PAGE_SIZE,
4502                          DMA_BIDIRECTIONAL);
4503         free_page((unsigned long)crq->msgs);
4504         crq->msgs = NULL;
4505         crq->active = false;
4506 }
4507
4508 static int init_crq_queue(struct ibmvnic_adapter *adapter)
4509 {
4510         struct ibmvnic_crq_queue *crq = &adapter->crq;
4511         struct device *dev = &adapter->vdev->dev;
4512         struct vio_dev *vdev = adapter->vdev;
4513         int rc, retrc = -ENOMEM;
4514
4515         if (crq->msgs)
4516                 return 0;
4517
4518         crq->msgs = (union ibmvnic_crq *)get_zeroed_page(GFP_KERNEL);
4519         /* Should we allocate more than one page? */
4520
4521         if (!crq->msgs)
4522                 return -ENOMEM;
4523
4524         crq->size = PAGE_SIZE / sizeof(*crq->msgs);
4525         crq->msg_token = dma_map_single(dev, crq->msgs, PAGE_SIZE,
4526                                         DMA_BIDIRECTIONAL);
4527         if (dma_mapping_error(dev, crq->msg_token))
4528                 goto map_failed;
4529
4530         rc = plpar_hcall_norets(H_REG_CRQ, vdev->unit_address,
4531                                 crq->msg_token, PAGE_SIZE);
4532
4533         if (rc == H_RESOURCE)
4534                 /* maybe kexecing and resource is busy. try a reset */
4535                 rc = ibmvnic_reset_crq(adapter);
4536         retrc = rc;
4537
4538         if (rc == H_CLOSED) {
4539                 dev_warn(dev, "Partner adapter not ready\n");
4540         } else if (rc) {
4541                 dev_warn(dev, "Error %d opening adapter\n", rc);
4542                 goto reg_crq_failed;
4543         }
4544
4545         retrc = 0;
4546
4547         tasklet_init(&adapter->tasklet, (void *)ibmvnic_tasklet,
4548                      (unsigned long)adapter);
4549
4550         netdev_dbg(adapter->netdev, "registering irq 0x%x\n", vdev->irq);
4551         rc = request_irq(vdev->irq, ibmvnic_interrupt, 0, IBMVNIC_NAME,
4552                          adapter);
4553         if (rc) {
4554                 dev_err(dev, "Couldn't register irq 0x%x. rc=%d\n",
4555                         vdev->irq, rc);
4556                 goto req_irq_failed;
4557         }
4558
4559         rc = vio_enable_interrupts(vdev);
4560         if (rc) {
4561                 dev_err(dev, "Error %d enabling interrupts\n", rc);
4562                 goto req_irq_failed;
4563         }
4564
4565         crq->cur = 0;
4566         spin_lock_init(&crq->lock);
4567
4568         return retrc;
4569
4570 req_irq_failed:
4571         tasklet_kill(&adapter->tasklet);
4572         do {
4573                 rc = plpar_hcall_norets(H_FREE_CRQ, vdev->unit_address);
4574         } while (rc == H_BUSY || H_IS_LONG_BUSY(rc));
4575 reg_crq_failed:
4576         dma_unmap_single(dev, crq->msg_token, PAGE_SIZE, DMA_BIDIRECTIONAL);
4577 map_failed:
4578         free_page((unsigned long)crq->msgs);
4579         crq->msgs = NULL;
4580         return retrc;
4581 }
4582
4583 static int ibmvnic_reset_init(struct ibmvnic_adapter *adapter)
4584 {
4585         struct device *dev = &adapter->vdev->dev;
4586         unsigned long timeout = msecs_to_jiffies(30000);
4587         u64 old_num_rx_queues, old_num_tx_queues;
4588         int rc;
4589
4590         adapter->from_passive_init = false;
4591
4592         old_num_rx_queues = adapter->req_rx_queues;
4593         old_num_tx_queues = adapter->req_tx_queues;
4594
4595         init_completion(&adapter->init_done);
4596         adapter->init_done_rc = 0;
4597         ibmvnic_send_crq_init(adapter);
4598         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4599                 dev_err(dev, "Initialization sequence timed out\n");
4600                 return -1;
4601         }
4602
4603         if (adapter->init_done_rc) {
4604                 release_crq_queue(adapter);
4605                 return adapter->init_done_rc;
4606         }
4607
4608         if (adapter->from_passive_init) {
4609                 adapter->state = VNIC_OPEN;
4610                 adapter->from_passive_init = false;
4611                 return -1;
4612         }
4613
4614         if (adapter->resetting && !adapter->wait_for_reset &&
4615             adapter->reset_reason != VNIC_RESET_MOBILITY) {
4616                 if (adapter->req_rx_queues != old_num_rx_queues ||
4617                     adapter->req_tx_queues != old_num_tx_queues) {
4618                         release_sub_crqs(adapter, 0);
4619                         rc = init_sub_crqs(adapter);
4620                 } else {
4621                         rc = reset_sub_crq_queues(adapter);
4622                 }
4623         } else {
4624                 rc = init_sub_crqs(adapter);
4625         }
4626
4627         if (rc) {
4628                 dev_err(dev, "Initialization of sub crqs failed\n");
4629                 release_crq_queue(adapter);
4630                 return rc;
4631         }
4632
4633         rc = init_sub_crq_irqs(adapter);
4634         if (rc) {
4635                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4636                 release_crq_queue(adapter);
4637         }
4638
4639         return rc;
4640 }
4641
4642 static int ibmvnic_init(struct ibmvnic_adapter *adapter)
4643 {
4644         struct device *dev = &adapter->vdev->dev;
4645         unsigned long timeout = msecs_to_jiffies(30000);
4646         int rc;
4647
4648         adapter->from_passive_init = false;
4649
4650         init_completion(&adapter->init_done);
4651         adapter->init_done_rc = 0;
4652         ibmvnic_send_crq_init(adapter);
4653         if (!wait_for_completion_timeout(&adapter->init_done, timeout)) {
4654                 dev_err(dev, "Initialization sequence timed out\n");
4655                 return -1;
4656         }
4657
4658         if (adapter->init_done_rc) {
4659                 release_crq_queue(adapter);
4660                 return adapter->init_done_rc;
4661         }
4662
4663         if (adapter->from_passive_init) {
4664                 adapter->state = VNIC_OPEN;
4665                 adapter->from_passive_init = false;
4666                 return -1;
4667         }
4668
4669         rc = init_sub_crqs(adapter);
4670         if (rc) {
4671                 dev_err(dev, "Initialization of sub crqs failed\n");
4672                 release_crq_queue(adapter);
4673                 return rc;
4674         }
4675
4676         rc = init_sub_crq_irqs(adapter);
4677         if (rc) {
4678                 dev_err(dev, "Failed to initialize sub crq irqs\n");
4679                 release_crq_queue(adapter);
4680         }
4681
4682         return rc;
4683 }
4684
4685 static struct device_attribute dev_attr_failover;
4686
4687 static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id)
4688 {
4689         struct ibmvnic_adapter *adapter;
4690         struct net_device *netdev;
4691         unsigned char *mac_addr_p;
4692         int rc;
4693
4694         dev_dbg(&dev->dev, "entering ibmvnic_probe for UA 0x%x\n",
4695                 dev->unit_address);
4696
4697         mac_addr_p = (unsigned char *)vio_get_attribute(dev,
4698                                                         VETH_MAC_ADDR, NULL);
4699         if (!mac_addr_p) {
4700                 dev_err(&dev->dev,
4701                         "(%s:%3.3d) ERROR: Can't find MAC_ADDR attribute\n",
4702                         __FILE__, __LINE__);
4703                 return 0;
4704         }
4705
4706         netdev = alloc_etherdev_mq(sizeof(struct ibmvnic_adapter),
4707                                    IBMVNIC_MAX_QUEUES);
4708         if (!netdev)
4709                 return -ENOMEM;
4710
4711         adapter = netdev_priv(netdev);
4712         adapter->state = VNIC_PROBING;
4713         dev_set_drvdata(&dev->dev, netdev);
4714         adapter->vdev = dev;
4715         adapter->netdev = netdev;
4716
4717         ether_addr_copy(adapter->mac_addr, mac_addr_p);
4718         ether_addr_copy(netdev->dev_addr, adapter->mac_addr);
4719         netdev->irq = dev->irq;
4720         netdev->netdev_ops = &ibmvnic_netdev_ops;
4721         netdev->ethtool_ops = &ibmvnic_ethtool_ops;
4722         SET_NETDEV_DEV(netdev, &dev->dev);
4723
4724         spin_lock_init(&adapter->stats_lock);
4725
4726         INIT_WORK(&adapter->ibmvnic_reset, __ibmvnic_reset);
4727         INIT_LIST_HEAD(&adapter->rwi_list);
4728         mutex_init(&adapter->reset_lock);
4729         mutex_init(&adapter->rwi_lock);
4730         adapter->resetting = false;
4731
4732         adapter->mac_change_pending = false;
4733
4734         do {
4735                 rc = init_crq_queue(adapter);
4736                 if (rc) {
4737                         dev_err(&dev->dev, "Couldn't initialize crq. rc=%d\n",
4738                                 rc);
4739                         goto ibmvnic_init_fail;
4740                 }
4741
4742                 rc = ibmvnic_init(adapter);
4743                 if (rc && rc != EAGAIN)
4744                         goto ibmvnic_init_fail;
4745         } while (rc == EAGAIN);
4746
4747         rc = init_stats_buffers(adapter);
4748         if (rc)
4749                 goto ibmvnic_init_fail;
4750
4751         rc = init_stats_token(adapter);
4752         if (rc)
4753                 goto ibmvnic_stats_fail;
4754
4755         netdev->mtu = adapter->req_mtu - ETH_HLEN;
4756         netdev->min_mtu = adapter->min_mtu - ETH_HLEN;
4757         netdev->max_mtu = adapter->max_mtu - ETH_HLEN;
4758
4759         rc = device_create_file(&dev->dev, &dev_attr_failover);
4760         if (rc)
4761                 goto ibmvnic_dev_file_err;
4762
4763         netif_carrier_off(netdev);
4764         rc = register_netdev(netdev);
4765         if (rc) {
4766                 dev_err(&dev->dev, "failed to register netdev rc=%d\n", rc);
4767                 goto ibmvnic_register_fail;
4768         }
4769         dev_info(&dev->dev, "ibmvnic registered\n");
4770
4771         adapter->state = VNIC_PROBED;
4772
4773         adapter->wait_for_reset = false;
4774
4775         return 0;
4776
4777 ibmvnic_register_fail:
4778         device_remove_file(&dev->dev, &dev_attr_failover);
4779
4780 ibmvnic_dev_file_err:
4781         release_stats_token(adapter);
4782
4783 ibmvnic_stats_fail:
4784         release_stats_buffers(adapter);
4785
4786 ibmvnic_init_fail:
4787         release_sub_crqs(adapter, 1);
4788         release_crq_queue(adapter);
4789         free_netdev(netdev);
4790
4791         return rc;
4792 }
4793
4794 static int ibmvnic_remove(struct vio_dev *dev)
4795 {
4796         struct net_device *netdev = dev_get_drvdata(&dev->dev);
4797         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4798
4799         adapter->state = VNIC_REMOVING;
4800         unregister_netdev(netdev);
4801         mutex_lock(&adapter->reset_lock);
4802
4803         release_resources(adapter);
4804         release_sub_crqs(adapter, 1);
4805         release_crq_queue(adapter);
4806
4807         release_stats_token(adapter);
4808         release_stats_buffers(adapter);
4809
4810         adapter->state = VNIC_REMOVED;
4811
4812         mutex_unlock(&adapter->reset_lock);
4813         device_remove_file(&dev->dev, &dev_attr_failover);
4814         free_netdev(netdev);
4815         dev_set_drvdata(&dev->dev, NULL);
4816
4817         return 0;
4818 }
4819
4820 static ssize_t failover_store(struct device *dev, struct device_attribute *attr,
4821                               const char *buf, size_t count)
4822 {
4823         struct net_device *netdev = dev_get_drvdata(dev);
4824         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4825         unsigned long retbuf[PLPAR_HCALL_BUFSIZE];
4826         __be64 session_token;
4827         long rc;
4828
4829         if (!sysfs_streq(buf, "1"))
4830                 return -EINVAL;
4831
4832         rc = plpar_hcall(H_VIOCTL, retbuf, adapter->vdev->unit_address,
4833                          H_GET_SESSION_TOKEN, 0, 0, 0);
4834         if (rc) {
4835                 netdev_err(netdev, "Couldn't retrieve session token, rc %ld\n",
4836                            rc);
4837                 return -EINVAL;
4838         }
4839
4840         session_token = (__be64)retbuf[0];
4841         netdev_dbg(netdev, "Initiating client failover, session id %llx\n",
4842                    be64_to_cpu(session_token));
4843         rc = plpar_hcall_norets(H_VIOCTL, adapter->vdev->unit_address,
4844                                 H_SESSION_ERR_DETECTED, session_token, 0, 0);
4845         if (rc) {
4846                 netdev_err(netdev, "Client initiated failover failed, rc %ld\n",
4847                            rc);
4848                 return -EINVAL;
4849         }
4850
4851         return count;
4852 }
4853
4854 static DEVICE_ATTR_WO(failover);
4855
4856 static unsigned long ibmvnic_get_desired_dma(struct vio_dev *vdev)
4857 {
4858         struct net_device *netdev = dev_get_drvdata(&vdev->dev);
4859         struct ibmvnic_adapter *adapter;
4860         struct iommu_table *tbl;
4861         unsigned long ret = 0;
4862         int i;
4863
4864         tbl = get_iommu_table_base(&vdev->dev);
4865
4866         /* netdev inits at probe time along with the structures we need below*/
4867         if (!netdev)
4868                 return IOMMU_PAGE_ALIGN(IBMVNIC_IO_ENTITLEMENT_DEFAULT, tbl);
4869
4870         adapter = netdev_priv(netdev);
4871
4872         ret += PAGE_SIZE; /* the crq message queue */
4873         ret += IOMMU_PAGE_ALIGN(sizeof(struct ibmvnic_statistics), tbl);
4874
4875         for (i = 0; i < adapter->req_tx_queues + adapter->req_rx_queues; i++)
4876                 ret += 4 * PAGE_SIZE; /* the scrq message queue */
4877
4878         for (i = 0; i < be32_to_cpu(adapter->login_rsp_buf->num_rxadd_subcrqs);
4879              i++)
4880                 ret += adapter->rx_pool[i].size *
4881                     IOMMU_PAGE_ALIGN(adapter->rx_pool[i].buff_size, tbl);
4882
4883         return ret;
4884 }
4885
4886 static int ibmvnic_resume(struct device *dev)
4887 {
4888         struct net_device *netdev = dev_get_drvdata(dev);
4889         struct ibmvnic_adapter *adapter = netdev_priv(netdev);
4890
4891         if (adapter->state != VNIC_OPEN)
4892                 return 0;
4893
4894         tasklet_schedule(&adapter->tasklet);
4895
4896         return 0;
4897 }
4898
4899 static const struct vio_device_id ibmvnic_device_table[] = {
4900         {"network", "IBM,vnic"},
4901         {"", "" }
4902 };
4903 MODULE_DEVICE_TABLE(vio, ibmvnic_device_table);
4904
4905 static const struct dev_pm_ops ibmvnic_pm_ops = {
4906         .resume = ibmvnic_resume
4907 };
4908
4909 static struct vio_driver ibmvnic_driver = {
4910         .id_table       = ibmvnic_device_table,
4911         .probe          = ibmvnic_probe,
4912         .remove         = ibmvnic_remove,
4913         .get_desired_dma = ibmvnic_get_desired_dma,
4914         .name           = ibmvnic_driver_name,
4915         .pm             = &ibmvnic_pm_ops,
4916 };
4917
4918 /* module functions */
4919 static int __init ibmvnic_module_init(void)
4920 {
4921         pr_info("%s: %s %s\n", ibmvnic_driver_name, ibmvnic_driver_string,
4922                 IBMVNIC_DRIVER_VERSION);
4923
4924         return vio_register_driver(&ibmvnic_driver);
4925 }
4926
4927 static void __exit ibmvnic_module_exit(void)
4928 {
4929         vio_unregister_driver(&ibmvnic_driver);
4930 }
4931
4932 module_init(ibmvnic_module_init);
4933 module_exit(ibmvnic_module_exit);