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