drivers/net/ibmveth.c

   1 /**************************************************************************/
   2 /*                                                                        */
   3 /* IBM eServer i/pSeries Virtual Ethernet Device Driver                   */
   4 /* Copyright (C) 2003 IBM Corp.                                           */
   5 /*  Originally written by Dave Larson (larson1@us.ibm.com)                */
   6 /*  Maintained by Santiago Leon (santil@us.ibm.com)                       */
   7 /*                                                                        */
   8 /*  This program is free software; you can redistribute it and/or modify  */
   9 /*  it under the terms of the GNU General Public License as published by  */
  10 /*  the Free Software Foundation; either version 2 of the License, or     */
  11 /*  (at your option) any later version.                                   */
  12 /*                                                                        */
  13 /*  This program is distributed in the hope that it will be useful,       */
  14 /*  but WITHOUT ANY WARRANTY; without even the implied warranty of        */
  15 /*  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the         */
  16 /*  GNU General Public License for more details.                          */
  17 /*                                                                        */
  18 /*  You should have received a copy of the GNU General Public License     */
  19 /*  along with this program; if not, write to the Free Software           */
  20 /*  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  */
  21 /*                                                                   USA  */
  22 /*                                                                        */
  23 /* This module contains the implementation of a virtual ethernet device   */
  24 /* for use with IBM i/pSeries LPAR Linux.  It utilizes the logical LAN    */
  25 /* option of the RS/6000 Platform Architechture to interface with virtual */
  26 /* ethernet NICs that are presented to the partition by the hypervisor.   */
  27 /*                                                                        */
  28 /**************************************************************************/
  29 /*
  30   TODO:
  31   - remove frag processing code - no longer needed
  32   - add support for sysfs
  33   - possibly remove procfs support
  34 */
  35
  36 #include <linux/module.h>
  37 #include <linux/types.h>
  38 #include <linux/errno.h>
  39 #include <linux/ioport.h>
  40 #include <linux/dma-mapping.h>
  41 #include <linux/kernel.h>
  42 #include <linux/netdevice.h>
  43 #include <linux/etherdevice.h>
  44 #include <linux/skbuff.h>
  45 #include <linux/init.h>
  46 #include <linux/delay.h>
  47 #include <linux/mm.h>
  48 #include <linux/ethtool.h>
  49 #include <linux/proc_fs.h>
  50 #include <linux/in.h>
  51 #include <linux/ip.h>
  52 #include <net/net_namespace.h>
  53 #include <asm/semaphore.h>
  54 #include <asm/hvcall.h>
  55 #include <asm/atomic.h>
  56 #include <asm/vio.h>
  57 #include <asm/uaccess.h>
  58 #include <linux/seq_file.h>
  59
  60 #include "ibmveth.h"
  61
  62 #undef DEBUG
  63
  64 #define ibmveth_printk(fmt, args...) \
  65   printk(KERN_DEBUG "%s: " fmt, __FILE__, ## args)
  66
  67 #define ibmveth_error_printk(fmt, args...) \
  68   printk(KERN_ERR "(%s:%3.3d ua:%x) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  69
  70 #ifdef DEBUG
  71 #define ibmveth_debug_printk_no_adapter(fmt, args...) \
  72   printk(KERN_DEBUG "(%s:%3.3d): " fmt, __FILE__, __LINE__ , ## args)
  73 #define ibmveth_debug_printk(fmt, args...) \
  74   printk(KERN_DEBUG "(%s:%3.3d ua:%x): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  75 #define ibmveth_assert(expr) \
  76   if(!(expr)) {                                   \
  77     printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%x): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
  78     BUG(); \
  79   }
  80 #else
  81 #define ibmveth_debug_printk_no_adapter(fmt, args...)
  82 #define ibmveth_debug_printk(fmt, args...)
  83 #define ibmveth_assert(expr)
  84 #endif
  85
  86 static int ibmveth_open(struct net_device *dev);
  87 static int ibmveth_close(struct net_device *dev);
  88 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
  89 static int ibmveth_poll(struct napi_struct *napi, int budget);
  90 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev);
  91 static struct net_device_stats *ibmveth_get_stats(struct net_device *dev);
  92 static void ibmveth_set_multicast_list(struct net_device *dev);
  93 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu);
  94 static void ibmveth_proc_register_driver(void);
  95 static void ibmveth_proc_unregister_driver(void);
  96 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter);
  97 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter);
  98 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance);
  99 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter);
 100 static struct kobj_type ktype_veth_pool;
 101
 102 #ifdef CONFIG_PROC_FS
 103 #define IBMVETH_PROC_DIR "ibmveth"
 104 static struct proc_dir_entry *ibmveth_proc_dir;
 105 #endif
 106
 107 static const char ibmveth_driver_name[] = "ibmveth";
 108 static const char ibmveth_driver_string[] = "IBM i/pSeries Virtual Ethernet Driver";
 109 #define ibmveth_driver_version "1.03"
 110
 111 MODULE_AUTHOR("Santiago Leon <santil@us.ibm.com>");
 112 MODULE_DESCRIPTION("IBM i/pSeries Virtual Ethernet Driver");
 113 MODULE_LICENSE("GPL");
 114 MODULE_VERSION(ibmveth_driver_version);
 115
 116 /* simple methods of getting data from the current rxq entry */
 117 static inline int ibmveth_rxq_pending_buffer(struct ibmveth_adapter *adapter)
 118 {
 119         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].toggle == adapter->rx_queue.toggle);
 120 }
 121
 122 static inline int ibmveth_rxq_buffer_valid(struct ibmveth_adapter *adapter)
 123 {
 124         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].valid);
 125 }
 126
 127 static inline int ibmveth_rxq_frame_offset(struct ibmveth_adapter *adapter)
 128 {
 129         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].offset);
 130 }
 131
 132 static inline int ibmveth_rxq_frame_length(struct ibmveth_adapter *adapter)
 133 {
 134         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].length);
 135 }
 136
 137 static inline int ibmveth_rxq_csum_good(struct ibmveth_adapter *adapter)
 138 {
 139         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].csum_good);
 140 }
 141
 142 /* setup the initial settings for a buffer pool */
 143 static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool *pool, u32 pool_index, u32 pool_size, u32 buff_size, u32 pool_active)
 144 {
 145         pool->size = pool_size;
 146         pool->index = pool_index;
 147         pool->buff_size = buff_size;
 148         pool->threshold = pool_size / 2;
 149         pool->active = pool_active;
 150 }
 151
 152 /* allocate and setup an buffer pool - called during open */
 153 static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool *pool)
 154 {
 155         int i;
 156
 157         pool->free_map = kmalloc(sizeof(u16) * pool->size, GFP_KERNEL);
 158
 159         if(!pool->free_map) {
 160                 return -1;
 161         }
 162
 163         pool->dma_addr = kmalloc(sizeof(dma_addr_t) * pool->size, GFP_KERNEL);
 164         if(!pool->dma_addr) {
 165                 kfree(pool->free_map);
 166                 pool->free_map = NULL;
 167                 return -1;
 168         }
 169
 170         pool->skbuff = kmalloc(sizeof(void*) * pool->size, GFP_KERNEL);
 171
 172         if(!pool->skbuff) {
 173                 kfree(pool->dma_addr);
 174                 pool->dma_addr = NULL;
 175
 176                 kfree(pool->free_map);
 177                 pool->free_map = NULL;
 178                 return -1;
 179         }
 180
 181         memset(pool->skbuff, 0, sizeof(void*) * pool->size);
 182         memset(pool->dma_addr, 0, sizeof(dma_addr_t) * pool->size);
 183
 184         for(i = 0; i < pool->size; ++i) {
 185                 pool->free_map[i] = i;
 186         }
 187
 188         atomic_set(&pool->available, 0);
 189         pool->producer_index = 0;
 190         pool->consumer_index = 0;
 191
 192         return 0;
 193 }
 194
 195 /* replenish the buffers for a pool.  note that we don't need to
 196  * skb_reserve these since they are used for incoming...
 197  */
 198 static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 199 {
 200         u32 i;
 201         u32 count = pool->size - atomic_read(&pool->available);
 202         u32 buffers_added = 0;
 203
 204         mb();
 205
 206         for(i = 0; i < count; ++i) {
 207                 struct sk_buff *skb;
 208                 unsigned int free_index, index;
 209                 u64 correlator;
 210                 union ibmveth_buf_desc desc;
 211                 unsigned long lpar_rc;
 212                 dma_addr_t dma_addr;
 213
 214                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 215
 216                 if(!skb) {
 217                         ibmveth_debug_printk("replenish: unable to allocate skb\n");
 218                         adapter->replenish_no_mem++;
 219                         break;
 220                 }
 221
 222                 free_index = pool->consumer_index;
 223                 pool->consumer_index = (pool->consumer_index + 1) % pool->size;
 224                 index = pool->free_map[free_index];
 225
 226                 ibmveth_assert(index != IBM_VETH_INVALID_MAP);
 227                 ibmveth_assert(pool->skbuff[index] == NULL);
 228
 229                 dma_addr = dma_map_single(&adapter->vdev->dev, skb->data,
 230                                 pool->buff_size, DMA_FROM_DEVICE);
 231
 232                 pool->free_map[free_index] = IBM_VETH_INVALID_MAP;
 233                 pool->dma_addr[index] = dma_addr;
 234                 pool->skbuff[index] = skb;
 235
 236                 correlator = ((u64)pool->index << 32) | index;
 237                 *(u64*)skb->data = correlator;
 238
 239                 desc.desc = 0;
 240                 desc.fields.valid = 1;
 241                 desc.fields.length = pool->buff_size;
 242                 desc.fields.address = dma_addr;
 243
 244                 lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 245
 246                 if(lpar_rc != H_SUCCESS) {
 247                         pool->free_map[free_index] = index;
 248                         pool->skbuff[index] = NULL;
 249                         if (pool->consumer_index == 0)
 250                                 pool->consumer_index = pool->size - 1;
 251                         else
 252                                 pool->consumer_index--;
 253                         dma_unmap_single(&adapter->vdev->dev,
 254                                         pool->dma_addr[index], pool->buff_size,
 255                                         DMA_FROM_DEVICE);
 256                         dev_kfree_skb_any(skb);
 257                         adapter->replenish_add_buff_failure++;
 258                         break;
 259                 } else {
 260                         buffers_added++;
 261                         adapter->replenish_add_buff_success++;
 262                 }
 263         }
 264
 265         mb();
 266         atomic_add(buffers_added, &(pool->available));
 267 }
 268
 269 /* replenish routine */
 270 static void ibmveth_replenish_task(struct ibmveth_adapter *adapter)
 271 {
 272         int i;
 273
 274         adapter->replenish_task_cycles++;
 275
 276         for(i = 0; i < IbmVethNumBufferPools; i++)
 277                 if(adapter->rx_buff_pool[i].active)
 278                         ibmveth_replenish_buffer_pool(adapter,
 279                                                      &adapter->rx_buff_pool[i]);
 280
 281         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 282 }
 283
 284 /* empty and free ana buffer pool - also used to do cleanup in error paths */
 285 static void ibmveth_free_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 286 {
 287         int i;
 288
 289         kfree(pool->free_map);
 290         pool->free_map = NULL;
 291
 292         if(pool->skbuff && pool->dma_addr) {
 293                 for(i = 0; i < pool->size; ++i) {
 294                         struct sk_buff *skb = pool->skbuff[i];
 295                         if(skb) {
 296                                 dma_unmap_single(&adapter->vdev->dev,
 297                                                  pool->dma_addr[i],
 298                                                  pool->buff_size,
 299                                                  DMA_FROM_DEVICE);
 300                                 dev_kfree_skb_any(skb);
 301                                 pool->skbuff[i] = NULL;
 302                         }
 303                 }
 304         }
 305
 306         if(pool->dma_addr) {
 307                 kfree(pool->dma_addr);
 308                 pool->dma_addr = NULL;
 309         }
 310
 311         if(pool->skbuff) {
 312                 kfree(pool->skbuff);
 313                 pool->skbuff = NULL;
 314         }
 315 }
 316
 317 /* remove a buffer from a pool */
 318 static void ibmveth_remove_buffer_from_pool(struct ibmveth_adapter *adapter, u64 correlator)
 319 {
 320         unsigned int pool  = correlator >> 32;
 321         unsigned int index = correlator & 0xffffffffUL;
 322         unsigned int free_index;
 323         struct sk_buff *skb;
 324
 325         ibmveth_assert(pool < IbmVethNumBufferPools);
 326         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 327
 328         skb = adapter->rx_buff_pool[pool].skbuff[index];
 329
 330         ibmveth_assert(skb != NULL);
 331
 332         adapter->rx_buff_pool[pool].skbuff[index] = NULL;
 333
 334         dma_unmap_single(&adapter->vdev->dev,
 335                          adapter->rx_buff_pool[pool].dma_addr[index],
 336                          adapter->rx_buff_pool[pool].buff_size,
 337                          DMA_FROM_DEVICE);
 338
 339         free_index = adapter->rx_buff_pool[pool].producer_index;
 340         adapter->rx_buff_pool[pool].producer_index
 341                 = (adapter->rx_buff_pool[pool].producer_index + 1)
 342                 % adapter->rx_buff_pool[pool].size;
 343         adapter->rx_buff_pool[pool].free_map[free_index] = index;
 344
 345         mb();
 346
 347         atomic_dec(&(adapter->rx_buff_pool[pool].available));
 348 }
 349
 350 /* get the current buffer on the rx queue */
 351 static inline struct sk_buff *ibmveth_rxq_get_buffer(struct ibmveth_adapter *adapter)
 352 {
 353         u64 correlator = adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator;
 354         unsigned int pool = correlator >> 32;
 355         unsigned int index = correlator & 0xffffffffUL;
 356
 357         ibmveth_assert(pool < IbmVethNumBufferPools);
 358         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 359
 360         return adapter->rx_buff_pool[pool].skbuff[index];
 361 }
 362
 363 /* recycle the current buffer on the rx queue */
 364 static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
 365 {
 366         u32 q_index = adapter->rx_queue.index;
 367         u64 correlator = adapter->rx_queue.queue_addr[q_index].correlator;
 368         unsigned int pool = correlator >> 32;
 369         unsigned int index = correlator & 0xffffffffUL;
 370         union ibmveth_buf_desc desc;
 371         unsigned long lpar_rc;
 372
 373         ibmveth_assert(pool < IbmVethNumBufferPools);
 374         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 375
 376         if(!adapter->rx_buff_pool[pool].active) {
 377                 ibmveth_rxq_harvest_buffer(adapter);
 378                 ibmveth_free_buffer_pool(adapter, &adapter->rx_buff_pool[pool]);
 379                 return;
 380         }
 381
 382         desc.desc = 0;
 383         desc.fields.valid = 1;
 384         desc.fields.length = adapter->rx_buff_pool[pool].buff_size;
 385         desc.fields.address = adapter->rx_buff_pool[pool].dma_addr[index];
 386
 387         lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 388
 389         if(lpar_rc != H_SUCCESS) {
 390                 ibmveth_debug_printk("h_add_logical_lan_buffer failed during recycle rc=%ld", lpar_rc);
 391                 ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 392         }
 393
 394         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 395                 adapter->rx_queue.index = 0;
 396                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 397         }
 398 }
 399
 400 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter)
 401 {
 402         ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 403
 404         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 405                 adapter->rx_queue.index = 0;
 406                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 407         }
 408 }
 409
 410 static void ibmveth_cleanup(struct ibmveth_adapter *adapter)
 411 {
 412         int i;
 413
 414         if(adapter->buffer_list_addr != NULL) {
 415                 if(!dma_mapping_error(adapter->buffer_list_dma)) {
 416                         dma_unmap_single(&adapter->vdev->dev,
 417                                         adapter->buffer_list_dma, 4096,
 418                                         DMA_BIDIRECTIONAL);
 419                         adapter->buffer_list_dma = DMA_ERROR_CODE;
 420                 }
 421                 free_page((unsigned long)adapter->buffer_list_addr);
 422                 adapter->buffer_list_addr = NULL;
 423         }
 424
 425         if(adapter->filter_list_addr != NULL) {
 426                 if(!dma_mapping_error(adapter->filter_list_dma)) {
 427                         dma_unmap_single(&adapter->vdev->dev,
 428                                         adapter->filter_list_dma, 4096,
 429                                         DMA_BIDIRECTIONAL);
 430                         adapter->filter_list_dma = DMA_ERROR_CODE;
 431                 }
 432                 free_page((unsigned long)adapter->filter_list_addr);
 433                 adapter->filter_list_addr = NULL;
 434         }
 435
 436         if(adapter->rx_queue.queue_addr != NULL) {
 437                 if(!dma_mapping_error(adapter->rx_queue.queue_dma)) {
 438                         dma_unmap_single(&adapter->vdev->dev,
 439                                         adapter->rx_queue.queue_dma,
 440                                         adapter->rx_queue.queue_len,
 441                                         DMA_BIDIRECTIONAL);
 442                         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
 443                 }
 444                 kfree(adapter->rx_queue.queue_addr);
 445                 adapter->rx_queue.queue_addr = NULL;
 446         }
 447
 448         for(i = 0; i<IbmVethNumBufferPools; i++)
 449                 if (adapter->rx_buff_pool[i].active)
 450                         ibmveth_free_buffer_pool(adapter,
 451                                                  &adapter->rx_buff_pool[i]);
 452 }
 453
 454 static int ibmveth_register_logical_lan(struct ibmveth_adapter *adapter,
 455         union ibmveth_buf_desc rxq_desc, u64 mac_address)
 456 {
 457         int rc, try_again = 1;
 458
 459         /* After a kexec the adapter will still be open, so our attempt to
 460         * open it will fail. So if we get a failure we free the adapter and
 461         * try again, but only once. */
 462 retry:
 463         rc = h_register_logical_lan(adapter->vdev->unit_address,
 464                                     adapter->buffer_list_dma, rxq_desc.desc,
 465                                     adapter->filter_list_dma, mac_address);
 466
 467         if (rc != H_SUCCESS && try_again) {
 468                 do {
 469                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 470                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 471
 472                 try_again = 0;
 473                 goto retry;
 474         }
 475
 476         return rc;
 477 }
 478
 479 static int ibmveth_open(struct net_device *netdev)
 480 {
 481         struct ibmveth_adapter *adapter = netdev->priv;
 482         u64 mac_address = 0;
 483         int rxq_entries = 1;
 484         unsigned long lpar_rc;
 485         int rc;
 486         union ibmveth_buf_desc rxq_desc;
 487         int i;
 488
 489         ibmveth_debug_printk("open starting\n");
 490
 491         napi_enable(&adapter->napi);
 492
 493         for(i = 0; i<IbmVethNumBufferPools; i++)
 494                 rxq_entries += adapter->rx_buff_pool[i].size;
 495
 496         adapter->buffer_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 497         adapter->filter_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 498
 499         if(!adapter->buffer_list_addr || !adapter->filter_list_addr) {
 500                 ibmveth_error_printk("unable to allocate filter or buffer list pages\n");
 501                 ibmveth_cleanup(adapter);
 502                 napi_disable(&adapter->napi);
 503                 return -ENOMEM;
 504         }
 505
 506         adapter->rx_queue.queue_len = sizeof(struct ibmveth_rx_q_entry) * rxq_entries;
 507         adapter->rx_queue.queue_addr = kmalloc(adapter->rx_queue.queue_len, GFP_KERNEL);
 508
 509         if(!adapter->rx_queue.queue_addr) {
 510                 ibmveth_error_printk("unable to allocate rx queue pages\n");
 511                 ibmveth_cleanup(adapter);
 512                 napi_disable(&adapter->napi);
 513                 return -ENOMEM;
 514         }
 515
 516         adapter->buffer_list_dma = dma_map_single(&adapter->vdev->dev,
 517                         adapter->buffer_list_addr, 4096, DMA_BIDIRECTIONAL);
 518         adapter->filter_list_dma = dma_map_single(&adapter->vdev->dev,
 519                         adapter->filter_list_addr, 4096, DMA_BIDIRECTIONAL);
 520         adapter->rx_queue.queue_dma = dma_map_single(&adapter->vdev->dev,
 521                         adapter->rx_queue.queue_addr,
 522                         adapter->rx_queue.queue_len, DMA_BIDIRECTIONAL);
 523
 524         if((dma_mapping_error(adapter->buffer_list_dma) ) ||
 525            (dma_mapping_error(adapter->filter_list_dma)) ||
 526            (dma_mapping_error(adapter->rx_queue.queue_dma))) {
 527                 ibmveth_error_printk("unable to map filter or buffer list pages\n");
 528                 ibmveth_cleanup(adapter);
 529                 napi_disable(&adapter->napi);
 530                 return -ENOMEM;
 531         }
 532
 533         adapter->rx_queue.index = 0;
 534         adapter->rx_queue.num_slots = rxq_entries;
 535         adapter->rx_queue.toggle = 1;
 536
 537         memcpy(&mac_address, netdev->dev_addr, netdev->addr_len);
 538         mac_address = mac_address >> 16;
 539
 540         rxq_desc.desc = 0;
 541         rxq_desc.fields.valid = 1;
 542         rxq_desc.fields.length = adapter->rx_queue.queue_len;
 543         rxq_desc.fields.address = adapter->rx_queue.queue_dma;
 544
 545         ibmveth_debug_printk("buffer list @ 0x%p\n", adapter->buffer_list_addr);
 546         ibmveth_debug_printk("filter list @ 0x%p\n", adapter->filter_list_addr);
 547         ibmveth_debug_printk("receive q   @ 0x%p\n", adapter->rx_queue.queue_addr);
 548
 549         h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE);
 550
 551         lpar_rc = ibmveth_register_logical_lan(adapter, rxq_desc, mac_address);
 552
 553         if(lpar_rc != H_SUCCESS) {
 554                 ibmveth_error_printk("h_register_logical_lan failed with %ld\n", lpar_rc);
 555                 ibmveth_error_printk("buffer TCE:0x%lx filter TCE:0x%lx rxq desc:0x%lx MAC:0x%lx\n",
 556                                      adapter->buffer_list_dma,
 557                                      adapter->filter_list_dma,
 558                                      rxq_desc.desc,
 559                                      mac_address);
 560                 ibmveth_cleanup(adapter);
 561                 napi_disable(&adapter->napi);
 562                 return -ENONET;
 563         }
 564
 565         for(i = 0; i<IbmVethNumBufferPools; i++) {
 566                 if(!adapter->rx_buff_pool[i].active)
 567                         continue;
 568                 if (ibmveth_alloc_buffer_pool(&adapter->rx_buff_pool[i])) {
 569                         ibmveth_error_printk("unable to alloc pool\n");
 570                         adapter->rx_buff_pool[i].active = 0;
 571                         ibmveth_cleanup(adapter);
 572                         napi_disable(&adapter->napi);
 573                         return -ENOMEM ;
 574                 }
 575         }
 576
 577         ibmveth_debug_printk("registering irq 0x%x\n", netdev->irq);
 578         if((rc = request_irq(netdev->irq, &ibmveth_interrupt, 0, netdev->name, netdev)) != 0) {
 579                 ibmveth_error_printk("unable to request irq 0x%x, rc %d\n", netdev->irq, rc);
 580                 do {
 581                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 582                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 583
 584                 ibmveth_cleanup(adapter);
 585                 napi_disable(&adapter->napi);
 586                 return rc;
 587         }
 588
 589         ibmveth_debug_printk("initial replenish cycle\n");
 590         ibmveth_interrupt(netdev->irq, netdev);
 591
 592         netif_start_queue(netdev);
 593
 594         ibmveth_debug_printk("open complete\n");
 595
 596         return 0;
 597 }
 598
 599 static int ibmveth_close(struct net_device *netdev)
 600 {
 601         struct ibmveth_adapter *adapter = netdev->priv;
 602         long lpar_rc;
 603
 604         ibmveth_debug_printk("close starting\n");
 605
 606         napi_disable(&adapter->napi);
 607
 608         if (!adapter->pool_config)
 609                 netif_stop_queue(netdev);
 610
 611         free_irq(netdev->irq, netdev);
 612
 613         do {
 614                 lpar_rc = h_free_logical_lan(adapter->vdev->unit_address);
 615         } while (H_IS_LONG_BUSY(lpar_rc) || (lpar_rc == H_BUSY));
 616
 617         if(lpar_rc != H_SUCCESS)
 618         {
 619                 ibmveth_error_printk("h_free_logical_lan failed with %lx, continuing with close\n",
 620                                      lpar_rc);
 621         }
 622
 623         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 624
 625         ibmveth_cleanup(adapter);
 626
 627         ibmveth_debug_printk("close complete\n");
 628
 629         return 0;
 630 }
 631
 632 static int netdev_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) {
 633         cmd->supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
 634         cmd->advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg | ADVERTISED_FIBRE);
 635         cmd->speed = SPEED_1000;
 636         cmd->duplex = DUPLEX_FULL;
 637         cmd->port = PORT_FIBRE;
 638         cmd->phy_address = 0;
 639         cmd->transceiver = XCVR_INTERNAL;
 640         cmd->autoneg = AUTONEG_ENABLE;
 641         cmd->maxtxpkt = 0;
 642         cmd->maxrxpkt = 1;
 643         return 0;
 644 }
 645
 646 static void netdev_get_drvinfo (struct net_device *dev, struct ethtool_drvinfo *info) {
 647         strncpy(info->driver, ibmveth_driver_name, sizeof(info->driver) - 1);
 648         strncpy(info->version, ibmveth_driver_version, sizeof(info->version) - 1);
 649 }
 650
 651 static u32 netdev_get_link(struct net_device *dev) {
 652         return 1;
 653 }
 654
 655 static const struct ethtool_ops netdev_ethtool_ops = {
 656         .get_drvinfo            = netdev_get_drvinfo,
 657         .get_settings           = netdev_get_settings,
 658         .get_link               = netdev_get_link,
 659         .get_sg                 = ethtool_op_get_sg,
 660         .get_tx_csum            = ethtool_op_get_tx_csum,
 661 };
 662
 663 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 664 {
 665         return -EOPNOTSUPP;
 666 }
 667
 668 #define page_offset(v) ((unsigned long)(v) & ((1 << 12) - 1))
 669
 670 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 671 {
 672         struct ibmveth_adapter *adapter = netdev->priv;
 673         union ibmveth_buf_desc desc[IbmVethMaxSendFrags];
 674         unsigned long lpar_rc;
 675         int nfrags = 0, curfrag;
 676         unsigned long correlator;
 677         unsigned long flags;
 678         unsigned int retry_count;
 679         unsigned int tx_dropped = 0;
 680         unsigned int tx_bytes = 0;
 681         unsigned int tx_packets = 0;
 682         unsigned int tx_send_failed = 0;
 683         unsigned int tx_map_failed = 0;
 684
 685
 686         if ((skb_shinfo(skb)->nr_frags + 1) > IbmVethMaxSendFrags) {
 687                 tx_dropped++;
 688                 goto out;
 689         }
 690
 691         memset(&desc, 0, sizeof(desc));
 692
 693         /* nfrags = number of frags after the initial fragment */
 694         nfrags = skb_shinfo(skb)->nr_frags;
 695
 696         if(nfrags)
 697                 adapter->tx_multidesc_send++;
 698
 699         /* map the initial fragment */
 700         desc[0].fields.length  = nfrags ? skb->len - skb->data_len : skb->len;
 701         desc[0].fields.address = dma_map_single(&adapter->vdev->dev, skb->data,
 702                                         desc[0].fields.length, DMA_TO_DEVICE);
 703         desc[0].fields.valid   = 1;
 704
 705         if (skb->ip_summed == CHECKSUM_PARTIAL &&
 706             ip_hdr(skb)->protocol != IPPROTO_TCP && skb_checksum_help(skb)) {
 707                 ibmveth_error_printk("tx: failed to checksum packet\n");
 708                 tx_dropped++;
 709                 goto out;
 710         }
 711
 712         if (skb->ip_summed == CHECKSUM_PARTIAL) {
 713                 unsigned char *buf = skb_transport_header(skb) + skb->csum_offset;
 714
 715                 desc[0].fields.no_csum = 1;
 716                 desc[0].fields.csum_good = 1;
 717
 718                 /* Need to zero out the checksum */
 719                 buf[0] = 0;
 720                 buf[1] = 0;
 721         }
 722
 723         if(dma_mapping_error(desc[0].fields.address)) {
 724                 ibmveth_error_printk("tx: unable to map initial fragment\n");
 725                 tx_map_failed++;
 726                 tx_dropped++;
 727                 goto out;
 728         }
 729
 730         curfrag = nfrags;
 731
 732         /* map fragments past the initial portion if there are any */
 733         while(curfrag--) {
 734                 skb_frag_t *frag = &skb_shinfo(skb)->frags[curfrag];
 735                 desc[curfrag+1].fields.address
 736                         = dma_map_single(&adapter->vdev->dev,
 737                                 page_address(frag->page) + frag->page_offset,
 738                                 frag->size, DMA_TO_DEVICE);
 739                 desc[curfrag+1].fields.length = frag->size;
 740                 desc[curfrag+1].fields.valid  = 1;
 741                 if (skb->ip_summed == CHECKSUM_PARTIAL) {
 742                         desc[curfrag+1].fields.no_csum = 1;
 743                         desc[curfrag+1].fields.csum_good = 1;
 744                 }
 745
 746                 if(dma_mapping_error(desc[curfrag+1].fields.address)) {
 747                         ibmveth_error_printk("tx: unable to map fragment %d\n", curfrag);
 748                         tx_map_failed++;
 749                         tx_dropped++;
 750                         /* Free all the mappings we just created */
 751                         while(curfrag < nfrags) {
 752                                 dma_unmap_single(&adapter->vdev->dev,
 753                                                  desc[curfrag+1].fields.address,
 754                                                  desc[curfrag+1].fields.length,
 755                                                  DMA_TO_DEVICE);
 756                                 curfrag++;
 757                         }
 758                         goto out;
 759                 }
 760         }
 761
 762         /* send the frame. Arbitrarily set retrycount to 1024 */
 763         correlator = 0;
 764         retry_count = 1024;
 765         do {
 766                 lpar_rc = h_send_logical_lan(adapter->vdev->unit_address,
 767                                              desc[0].desc,
 768                                              desc[1].desc,
 769                                              desc[2].desc,
 770                                              desc[3].desc,
 771                                              desc[4].desc,
 772                                              desc[5].desc,
 773                                              correlator,
 774                                              &correlator);
 775         } while ((lpar_rc == H_BUSY) && (retry_count--));
 776
 777         if(lpar_rc != H_SUCCESS && lpar_rc != H_DROPPED) {
 778                 int i;
 779                 ibmveth_error_printk("tx: h_send_logical_lan failed with rc=%ld\n", lpar_rc);
 780                 for(i = 0; i < 6; i++) {
 781                         ibmveth_error_printk("tx: desc[%i] valid=%d, len=%d, address=0x%d\n", i,
 782                                              desc[i].fields.valid, desc[i].fields.length, desc[i].fields.address);
 783                 }
 784                 tx_send_failed++;
 785                 tx_dropped++;
 786         } else {
 787                 tx_packets++;
 788                 tx_bytes += skb->len;
 789                 netdev->trans_start = jiffies;
 790         }
 791
 792         do {
 793                 dma_unmap_single(&adapter->vdev->dev,
 794                                 desc[nfrags].fields.address,
 795                                 desc[nfrags].fields.length, DMA_TO_DEVICE);
 796         } while(--nfrags >= 0);
 797
 798 out:    spin_lock_irqsave(&adapter->stats_lock, flags);
 799         adapter->stats.tx_dropped += tx_dropped;
 800         adapter->stats.tx_bytes += tx_bytes;
 801         adapter->stats.tx_packets += tx_packets;
 802         adapter->tx_send_failed += tx_send_failed;
 803         adapter->tx_map_failed += tx_map_failed;
 804         spin_unlock_irqrestore(&adapter->stats_lock, flags);
 805
 806         dev_kfree_skb(skb);
 807         return 0;
 808 }
 809
 810 static int ibmveth_poll(struct napi_struct *napi, int budget)
 811 {
 812         struct ibmveth_adapter *adapter = container_of(napi, struct ibmveth_adapter, napi);
 813         struct net_device *netdev = adapter->netdev;
 814         int frames_processed = 0;
 815         unsigned long lpar_rc;
 816
 817  restart_poll:
 818         do {
 819                 struct sk_buff *skb;
 820
 821                 if (!ibmveth_rxq_pending_buffer(adapter))
 822                         break;
 823
 824                 rmb();
 825                 if (!ibmveth_rxq_buffer_valid(adapter)) {
 826                         wmb(); /* suggested by larson1 */
 827                         adapter->rx_invalid_buffer++;
 828                         ibmveth_debug_printk("recycling invalid buffer\n");
 829                         ibmveth_rxq_recycle_buffer(adapter);
 830                 } else {
 831                         int length = ibmveth_rxq_frame_length(adapter);
 832                         int offset = ibmveth_rxq_frame_offset(adapter);
 833                         int csum_good = ibmveth_rxq_csum_good(adapter);
 834
 835                         skb = ibmveth_rxq_get_buffer(adapter);
 836                         if (csum_good)
 837                                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 838
 839                         ibmveth_rxq_harvest_buffer(adapter);
 840
 841                         skb_reserve(skb, offset);
 842                         skb_put(skb, length);
 843                         skb->protocol = eth_type_trans(skb, netdev);
 844
 845                         netif_receive_skb(skb); /* send it up */
 846
 847                         adapter->stats.rx_packets++;
 848                         adapter->stats.rx_bytes += length;
 849                         frames_processed++;
 850                         netdev->last_rx = jiffies;
 851                 }
 852         } while (frames_processed < budget);
 853
 854         ibmveth_replenish_task(adapter);
 855
 856         if (frames_processed < budget) {
 857                 /* We think we are done - reenable interrupts,
 858                  * then check once more to make sure we are done.
 859                  */
 860                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 861                                        VIO_IRQ_ENABLE);
 862
 863                 ibmveth_assert(lpar_rc == H_SUCCESS);
 864
 865                 netif_rx_complete(netdev, napi);
 866
 867                 if (ibmveth_rxq_pending_buffer(adapter) &&
 868                     netif_rx_reschedule(netdev, napi)) {
 869                         lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 870                                                VIO_IRQ_DISABLE);
 871                         goto restart_poll;
 872                 }
 873         }
 874
 875         return frames_processed;
 876 }
 877
 878 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
 879 {
 880         struct net_device *netdev = dev_instance;
 881         struct ibmveth_adapter *adapter = netdev->priv;
 882         unsigned long lpar_rc;
 883
 884         if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
 885                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 886                                        VIO_IRQ_DISABLE);
 887                 ibmveth_assert(lpar_rc == H_SUCCESS);
 888                 __netif_rx_schedule(netdev, &adapter->napi);
 889         }
 890         return IRQ_HANDLED;
 891 }
 892
 893 static struct net_device_stats *ibmveth_get_stats(struct net_device *dev)
 894 {
 895         struct ibmveth_adapter *adapter = dev->priv;
 896         return &adapter->stats;
 897 }
 898
 899 static void ibmveth_set_multicast_list(struct net_device *netdev)
 900 {
 901         struct ibmveth_adapter *adapter = netdev->priv;
 902         unsigned long lpar_rc;
 903
 904         if((netdev->flags & IFF_PROMISC) || (netdev->mc_count > adapter->mcastFilterSize)) {
 905                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
 906                                            IbmVethMcastEnableRecv |
 907                                            IbmVethMcastDisableFiltering,
 908                                            0);
 909                 if(lpar_rc != H_SUCCESS) {
 910                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when entering promisc mode\n", lpar_rc);
 911                 }
 912         } else {
 913                 struct dev_mc_list *mclist = netdev->mc_list;
 914                 int i;
 915                 /* clear the filter table & disable filtering */
 916                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
 917                                            IbmVethMcastEnableRecv |
 918                                            IbmVethMcastDisableFiltering |
 919                                            IbmVethMcastClearFilterTable,
 920                                            0);
 921                 if(lpar_rc != H_SUCCESS) {
 922                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when attempting to clear filter table\n", lpar_rc);
 923                 }
 924                 /* add the addresses to the filter table */
 925                 for(i = 0; i < netdev->mc_count; ++i, mclist = mclist->next) {
 926                         // add the multicast address to the filter table
 927                         unsigned long mcast_addr = 0;
 928                         memcpy(((char *)&mcast_addr)+2, mclist->dmi_addr, 6);
 929                         lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
 930                                                    IbmVethMcastAddFilter,
 931                                                    mcast_addr);
 932                         if(lpar_rc != H_SUCCESS) {
 933                                 ibmveth_error_printk("h_multicast_ctrl rc=%ld when adding an entry to the filter table\n", lpar_rc);
 934                         }
 935                 }
 936
 937                 /* re-enable filtering */
 938                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
 939                                            IbmVethMcastEnableFiltering,
 940                                            0);
 941                 if(lpar_rc != H_SUCCESS) {
 942                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when enabling filtering\n", lpar_rc);
 943                 }
 944         }
 945 }
 946
 947 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu)
 948 {
 949         struct ibmveth_adapter *adapter = dev->priv;
 950         int new_mtu_oh = new_mtu + IBMVETH_BUFF_OH;
 951         int reinit = 0;
 952         int i, rc;
 953
 954         if (new_mtu < IBMVETH_MAX_MTU)
 955                 return -EINVAL;
 956
 957         for (i = 0; i < IbmVethNumBufferPools; i++)
 958                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size)
 959                         break;
 960
 961         if (i == IbmVethNumBufferPools)
 962                 return -EINVAL;
 963
 964         /* Look for an active buffer pool that can hold the new MTU */
 965         for(i = 0; i<IbmVethNumBufferPools; i++) {
 966                 if (!adapter->rx_buff_pool[i].active) {
 967                         adapter->rx_buff_pool[i].active = 1;
 968                         reinit = 1;
 969                 }
 970
 971                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size) {
 972                         if (reinit && netif_running(adapter->netdev)) {
 973                                 adapter->pool_config = 1;
 974                                 ibmveth_close(adapter->netdev);
 975                                 adapter->pool_config = 0;
 976                                 dev->mtu = new_mtu;
 977                                 if ((rc = ibmveth_open(adapter->netdev)))
 978                                         return rc;
 979                         } else
 980                                 dev->mtu = new_mtu;
 981                         return 0;
 982                 }
 983         }
 984         return -EINVAL;
 985 }
 986
 987 #ifdef CONFIG_NET_POLL_CONTROLLER
 988 static void ibmveth_poll_controller(struct net_device *dev)
 989 {
 990         ibmveth_replenish_task(dev->priv);
 991         ibmveth_interrupt(dev->irq, dev);
 992 }
 993 #endif
 994
 995 static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_id *id)
 996 {
 997         int rc, i;
 998         long ret;
 999         struct net_device *netdev;
1000         struct ibmveth_adapter *adapter;
1001         union ibmveth_illan_attributes set_attr, ret_attr;
1002
1003         unsigned char *mac_addr_p;
1004         unsigned int *mcastFilterSize_p;
1005
1006
1007         ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n",
1008                                         dev->unit_address);
1009
1010         mac_addr_p = (unsigned char *) vio_get_attribute(dev,
1011                                                 VETH_MAC_ADDR, NULL);
1012         if(!mac_addr_p) {
1013                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find VETH_MAC_ADDR "
1014                                 "attribute\n", __FILE__, __LINE__);
1015                 return 0;
1016         }
1017
1018         mcastFilterSize_p = (unsigned int *) vio_get_attribute(dev,
1019                                                 VETH_MCAST_FILTER_SIZE, NULL);
1020         if(!mcastFilterSize_p) {
1021                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find "
1022                                 "VETH_MCAST_FILTER_SIZE attribute\n",
1023                                 __FILE__, __LINE__);
1024                 return 0;
1025         }
1026
1027         netdev = alloc_etherdev(sizeof(struct ibmveth_adapter));
1028
1029         if(!netdev)
1030                 return -ENOMEM;
1031
1032         SET_MODULE_OWNER(netdev);
1033
1034         adapter = netdev->priv;
1035         dev->dev.driver_data = netdev;
1036
1037         adapter->vdev = dev;
1038         adapter->netdev = netdev;
1039         adapter->mcastFilterSize= *mcastFilterSize_p;
1040         adapter->pool_config = 0;
1041
1042         netif_napi_add(netdev, &adapter->napi, ibmveth_poll, 16);
1043
1044         /*      Some older boxes running PHYP non-natively have an OF that
1045                 returns a 8-byte local-mac-address field (and the first
1046                 2 bytes have to be ignored) while newer boxes' OF return
1047                 a 6-byte field. Note that IEEE 1275 specifies that
1048                 local-mac-address must be a 6-byte field.
1049                 The RPA doc specifies that the first byte must be 10b, so
1050                 we'll just look for it to solve this 8 vs. 6 byte field issue */
1051
1052         if ((*mac_addr_p & 0x3) != 0x02)
1053                 mac_addr_p += 2;
1054
1055         adapter->mac_addr = 0;
1056         memcpy(&adapter->mac_addr, mac_addr_p, 6);
1057
1058         netdev->irq = dev->irq;
1059         netdev->open               = ibmveth_open;
1060         netdev->stop               = ibmveth_close;
1061         netdev->hard_start_xmit    = ibmveth_start_xmit;
1062         netdev->get_stats          = ibmveth_get_stats;
1063         netdev->set_multicast_list = ibmveth_set_multicast_list;
1064         netdev->do_ioctl           = ibmveth_ioctl;
1065         netdev->ethtool_ops           = &netdev_ethtool_ops;
1066         netdev->change_mtu         = ibmveth_change_mtu;
1067         SET_NETDEV_DEV(netdev, &dev->dev);
1068 #ifdef CONFIG_NET_POLL_CONTROLLER
1069         netdev->poll_controller = ibmveth_poll_controller;
1070 #endif
1071         netdev->features |= NETIF_F_LLTX;
1072         spin_lock_init(&adapter->stats_lock);
1073
1074         memcpy(&netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
1075
1076         for(i = 0; i<IbmVethNumBufferPools; i++) {
1077                 struct kobject *kobj = &adapter->rx_buff_pool[i].kobj;
1078                 ibmveth_init_buffer_pool(&adapter->rx_buff_pool[i], i,
1079                                          pool_count[i], pool_size[i],
1080                                          pool_active[i]);
1081                 kobj->parent = &dev->dev.kobj;
1082                 sprintf(kobj->name, "pool%d", i);
1083                 kobj->ktype = &ktype_veth_pool;
1084                 kobject_register(kobj);
1085         }
1086
1087         ibmveth_debug_printk("adapter @ 0x%p\n", adapter);
1088
1089         adapter->buffer_list_dma = DMA_ERROR_CODE;
1090         adapter->filter_list_dma = DMA_ERROR_CODE;
1091         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
1092
1093         ibmveth_debug_printk("registering netdev...\n");
1094
1095         ret = h_illan_attributes(dev->unit_address, 0, 0, &ret_attr.desc);
1096
1097         if (ret == H_SUCCESS && !ret_attr.fields.active_trunk &&
1098             !ret_attr.fields.trunk_priority &&
1099             ret_attr.fields.csum_offload_padded_pkt_support) {
1100                 set_attr.desc = 0;
1101                 set_attr.fields.tcp_csum_offload_ipv4 = 1;
1102
1103                 ret = h_illan_attributes(dev->unit_address, 0, set_attr.desc,
1104                                          &ret_attr.desc);
1105
1106                 if (ret == H_SUCCESS)
1107                         netdev->features |= NETIF_F_IP_CSUM;
1108                 else
1109                         ret = h_illan_attributes(dev->unit_address, set_attr.desc,
1110                                                  0, &ret_attr.desc);
1111         }
1112
1113         rc = register_netdev(netdev);
1114
1115         if(rc) {
1116                 ibmveth_debug_printk("failed to register netdev rc=%d\n", rc);
1117                 free_netdev(netdev);
1118                 return rc;
1119         }
1120
1121         ibmveth_debug_printk("registered\n");
1122
1123         ibmveth_proc_register_adapter(adapter);
1124
1125         return 0;
1126 }
1127
1128 static int __devexit ibmveth_remove(struct vio_dev *dev)
1129 {
1130         struct net_device *netdev = dev->dev.driver_data;
1131         struct ibmveth_adapter *adapter = netdev->priv;
1132         int i;
1133
1134         for(i = 0; i<IbmVethNumBufferPools; i++)
1135                 kobject_unregister(&adapter->rx_buff_pool[i].kobj);
1136
1137         unregister_netdev(netdev);
1138
1139         ibmveth_proc_unregister_adapter(adapter);
1140
1141         free_netdev(netdev);
1142         return 0;
1143 }
1144
1145 #ifdef CONFIG_PROC_FS
1146 static void ibmveth_proc_register_driver(void)
1147 {
1148         ibmveth_proc_dir = proc_mkdir(IBMVETH_PROC_DIR, init_net.proc_net);
1149         if (ibmveth_proc_dir) {
1150                 SET_MODULE_OWNER(ibmveth_proc_dir);
1151         }
1152 }
1153
1154 static void ibmveth_proc_unregister_driver(void)
1155 {
1156         remove_proc_entry(IBMVETH_PROC_DIR, init_net.proc_net);
1157 }
1158
1159 static void *ibmveth_seq_start(struct seq_file *seq, loff_t *pos)
1160 {
1161         if (*pos == 0) {
1162                 return (void *)1;
1163         } else {
1164                 return NULL;
1165         }
1166 }
1167
1168 static void *ibmveth_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1169 {
1170         ++*pos;
1171         return NULL;
1172 }
1173
1174 static void ibmveth_seq_stop(struct seq_file *seq, void *v)
1175 {
1176 }
1177
1178 static int ibmveth_seq_show(struct seq_file *seq, void *v)
1179 {
1180         struct ibmveth_adapter *adapter = seq->private;
1181         char *current_mac = ((char*) &adapter->netdev->dev_addr);
1182         char *firmware_mac = ((char*) &adapter->mac_addr) ;
1183
1184         seq_printf(seq, "%s %s\n\n", ibmveth_driver_string, ibmveth_driver_version);
1185
1186         seq_printf(seq, "Unit Address:    0x%x\n", adapter->vdev->unit_address);
1187         seq_printf(seq, "Current MAC:     %02X:%02X:%02X:%02X:%02X:%02X\n",
1188                    current_mac[0], current_mac[1], current_mac[2],
1189                    current_mac[3], current_mac[4], current_mac[5]);
1190         seq_printf(seq, "Firmware MAC:    %02X:%02X:%02X:%02X:%02X:%02X\n",
1191                    firmware_mac[0], firmware_mac[1], firmware_mac[2],
1192                    firmware_mac[3], firmware_mac[4], firmware_mac[5]);
1193
1194         seq_printf(seq, "\nAdapter Statistics:\n");
1195         seq_printf(seq, "  TX:  skbuffs linearized:          %ld\n", adapter->tx_linearized);
1196         seq_printf(seq, "       multi-descriptor sends:      %ld\n", adapter->tx_multidesc_send);
1197         seq_printf(seq, "       skb_linearize failures:      %ld\n", adapter->tx_linearize_failed);
1198         seq_printf(seq, "       vio_map_single failres:      %ld\n", adapter->tx_map_failed);
1199         seq_printf(seq, "       send failures:               %ld\n", adapter->tx_send_failed);
1200         seq_printf(seq, "  RX:  replenish task cycles:       %ld\n", adapter->replenish_task_cycles);
1201         seq_printf(seq, "       alloc_skb_failures:          %ld\n", adapter->replenish_no_mem);
1202         seq_printf(seq, "       add buffer failures:         %ld\n", adapter->replenish_add_buff_failure);
1203         seq_printf(seq, "       invalid buffers:             %ld\n", adapter->rx_invalid_buffer);
1204         seq_printf(seq, "       no buffers:                  %ld\n", adapter->rx_no_buffer);
1205
1206         return 0;
1207 }
1208 static struct seq_operations ibmveth_seq_ops = {
1209         .start = ibmveth_seq_start,
1210         .next  = ibmveth_seq_next,
1211         .stop  = ibmveth_seq_stop,
1212         .show  = ibmveth_seq_show,
1213 };
1214
1215 static int ibmveth_proc_open(struct inode *inode, struct file *file)
1216 {
1217         struct seq_file *seq;
1218         struct proc_dir_entry *proc;
1219         int rc;
1220
1221         rc = seq_open(file, &ibmveth_seq_ops);
1222         if (!rc) {
1223                 /* recover the pointer buried in proc_dir_entry data */
1224                 seq = file->private_data;
1225                 proc = PDE(inode);
1226                 seq->private = proc->data;
1227         }
1228         return rc;
1229 }
1230
1231 static const struct file_operations ibmveth_proc_fops = {
1232         .owner   = THIS_MODULE,
1233         .open    = ibmveth_proc_open,
1234         .read    = seq_read,
1235         .llseek  = seq_lseek,
1236         .release = seq_release,
1237 };
1238
1239 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1240 {
1241         struct proc_dir_entry *entry;
1242         if (ibmveth_proc_dir) {
1243                 char u_addr[10];
1244                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1245                 entry = create_proc_entry(u_addr, S_IFREG, ibmveth_proc_dir);
1246                 if (!entry) {
1247                         ibmveth_error_printk("Cannot create adapter proc entry");
1248                 } else {
1249                         entry->data = (void *) adapter;
1250                         entry->proc_fops = &ibmveth_proc_fops;
1251                         SET_MODULE_OWNER(entry);
1252                 }
1253         }
1254         return;
1255 }
1256
1257 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1258 {
1259         if (ibmveth_proc_dir) {
1260                 char u_addr[10];
1261                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1262                 remove_proc_entry(u_addr, ibmveth_proc_dir);
1263         }
1264 }
1265
1266 #else /* CONFIG_PROC_FS */
1267 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1268 {
1269 }
1270
1271 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1272 {
1273 }
1274 static void ibmveth_proc_register_driver(void)
1275 {
1276 }
1277
1278 static void ibmveth_proc_unregister_driver(void)
1279 {
1280 }
1281 #endif /* CONFIG_PROC_FS */
1282
1283 static struct attribute veth_active_attr;
1284 static struct attribute veth_num_attr;
1285 static struct attribute veth_size_attr;
1286
1287 static ssize_t veth_pool_show(struct kobject * kobj,
1288                               struct attribute * attr, char * buf)
1289 {
1290         struct ibmveth_buff_pool *pool = container_of(kobj,
1291                                                       struct ibmveth_buff_pool,
1292                                                       kobj);
1293
1294         if (attr == &veth_active_attr)
1295                 return sprintf(buf, "%d\n", pool->active);
1296         else if (attr == &veth_num_attr)
1297                 return sprintf(buf, "%d\n", pool->size);
1298         else if (attr == &veth_size_attr)
1299                 return sprintf(buf, "%d\n", pool->buff_size);
1300         return 0;
1301 }
1302
1303 static ssize_t veth_pool_store(struct kobject * kobj, struct attribute * attr,
1304 const char * buf, size_t count)
1305 {
1306         struct ibmveth_buff_pool *pool = container_of(kobj,
1307                                                       struct ibmveth_buff_pool,
1308                                                       kobj);
1309         struct net_device *netdev =
1310             container_of(kobj->parent, struct device, kobj)->driver_data;
1311         struct ibmveth_adapter *adapter = netdev->priv;
1312         long value = simple_strtol(buf, NULL, 10);
1313         long rc;
1314
1315         if (attr == &veth_active_attr) {
1316                 if (value && !pool->active) {
1317                         if (netif_running(netdev)) {
1318                                 if(ibmveth_alloc_buffer_pool(pool)) {
1319                                         ibmveth_error_printk("unable to alloc pool\n");
1320                                         return -ENOMEM;
1321                                 }
1322                                 pool->active = 1;
1323                                 adapter->pool_config = 1;
1324                                 ibmveth_close(netdev);
1325                                 adapter->pool_config = 0;
1326                                 if ((rc = ibmveth_open(netdev)))
1327                                         return rc;
1328                         } else
1329                                 pool->active = 1;
1330                 } else if (!value && pool->active) {
1331                         int mtu = netdev->mtu + IBMVETH_BUFF_OH;
1332                         int i;
1333                         /* Make sure there is a buffer pool with buffers that
1334                            can hold a packet of the size of the MTU */
1335                         for (i = 0; i < IbmVethNumBufferPools; i++) {
1336                                 if (pool == &adapter->rx_buff_pool[i])
1337                                         continue;
1338                                 if (!adapter->rx_buff_pool[i].active)
1339                                         continue;
1340                                 if (mtu <= adapter->rx_buff_pool[i].buff_size)
1341                                         break;
1342                         }
1343
1344                         if (i == IbmVethNumBufferPools) {
1345                                 ibmveth_error_printk("no active pool >= MTU\n");
1346                                 return -EPERM;
1347                         }
1348
1349                         pool->active = 0;
1350                         if (netif_running(netdev)) {
1351                                 adapter->pool_config = 1;
1352                                 ibmveth_close(netdev);
1353                                 adapter->pool_config = 0;
1354                                 if ((rc = ibmveth_open(netdev)))
1355                                         return rc;
1356                         }
1357                 }
1358         } else if (attr == &veth_num_attr) {
1359                 if (value <= 0 || value > IBMVETH_MAX_POOL_COUNT)
1360                         return -EINVAL;
1361                 else {
1362                         if (netif_running(netdev)) {
1363                                 adapter->pool_config = 1;
1364                                 ibmveth_close(netdev);
1365                                 adapter->pool_config = 0;
1366                                 pool->size = value;
1367                                 if ((rc = ibmveth_open(netdev)))
1368                                         return rc;
1369                         } else
1370                                 pool->size = value;
1371                 }
1372         } else if (attr == &veth_size_attr) {
1373                 if (value <= IBMVETH_BUFF_OH || value > IBMVETH_MAX_BUF_SIZE)
1374                         return -EINVAL;
1375                 else {
1376                         if (netif_running(netdev)) {
1377                                 adapter->pool_config = 1;
1378                                 ibmveth_close(netdev);
1379                                 adapter->pool_config = 0;
1380                                 pool->buff_size = value;
1381                                 if ((rc = ibmveth_open(netdev)))
1382                                         return rc;
1383                         } else
1384                                 pool->buff_size = value;
1385                 }
1386         }
1387
1388         /* kick the interrupt handler to allocate/deallocate pools */
1389         ibmveth_interrupt(netdev->irq, netdev);
1390         return count;
1391 }
1392
1393
1394 #define ATTR(_name, _mode)      \
1395         struct attribute veth_##_name##_attr = {               \
1396         .name = __stringify(_name), .mode = _mode, \
1397         };
1398
1399 static ATTR(active, 0644);
1400 static ATTR(num, 0644);
1401 static ATTR(size, 0644);
1402
1403 static struct attribute * veth_pool_attrs[] = {
1404         &veth_active_attr,
1405         &veth_num_attr,
1406         &veth_size_attr,
1407         NULL,
1408 };
1409
1410 static struct sysfs_ops veth_pool_ops = {
1411         .show   = veth_pool_show,
1412         .store  = veth_pool_store,
1413 };
1414
1415 static struct kobj_type ktype_veth_pool = {
1416         .release        = NULL,
1417         .sysfs_ops      = &veth_pool_ops,
1418         .default_attrs  = veth_pool_attrs,
1419 };
1420
1421
1422 static struct vio_device_id ibmveth_device_table[] __devinitdata= {
1423         { "network", "IBM,l-lan"},
1424         { "", "" }
1425 };
1426 MODULE_DEVICE_TABLE(vio, ibmveth_device_table);
1427
1428 static struct vio_driver ibmveth_driver = {
1429         .id_table       = ibmveth_device_table,
1430         .probe          = ibmveth_probe,
1431         .remove         = ibmveth_remove,
1432         .driver         = {
1433                 .name   = ibmveth_driver_name,
1434                 .owner  = THIS_MODULE,
1435         }
1436 };
1437
1438 static int __init ibmveth_module_init(void)
1439 {
1440         ibmveth_printk("%s: %s %s\n", ibmveth_driver_name, ibmveth_driver_string, ibmveth_driver_version);
1441
1442         ibmveth_proc_register_driver();
1443
1444         return vio_register_driver(&ibmveth_driver);
1445 }
1446
1447 static void __exit ibmveth_module_exit(void)
1448 {
1449         vio_unregister_driver(&ibmveth_driver);
1450         ibmveth_proc_unregister_driver();
1451 }
1452
1453 module_init(ibmveth_module_init);
1454 module_exit(ibmveth_module_exit);