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   - add support for sysfs
  32   - possibly remove procfs support
  33 */
  34
  35 #include <linux/module.h>
  36 #include <linux/types.h>
  37 #include <linux/errno.h>
  38 #include <linux/ioport.h>
  39 #include <linux/dma-mapping.h>
  40 #include <linux/kernel.h>
  41 #include <linux/netdevice.h>
  42 #include <linux/etherdevice.h>
  43 #include <linux/skbuff.h>
  44 #include <linux/init.h>
  45 #include <linux/delay.h>
  46 #include <linux/mm.h>
  47 #include <linux/ethtool.h>
  48 #include <linux/proc_fs.h>
  49 #include <linux/in.h>
  50 #include <linux/ip.h>
  51 #include <net/net_namespace.h>
  52 #include <asm/semaphore.h>
  53 #include <asm/hvcall.h>
  54 #include <asm/atomic.h>
  55 #include <asm/vio.h>
  56 #include <asm/uaccess.h>
  57 #include <linux/seq_file.h>
  58
  59 #include "ibmveth.h"
  60
  61 #undef DEBUG
  62
  63 #define ibmveth_printk(fmt, args...) \
  64   printk(KERN_DEBUG "%s: " fmt, __FILE__, ## args)
  65
  66 #define ibmveth_error_printk(fmt, args...) \
  67   printk(KERN_ERR "(%s:%3.3d ua:%x) ERROR: " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  68
  69 #ifdef DEBUG
  70 #define ibmveth_debug_printk_no_adapter(fmt, args...) \
  71   printk(KERN_DEBUG "(%s:%3.3d): " fmt, __FILE__, __LINE__ , ## args)
  72 #define ibmveth_debug_printk(fmt, args...) \
  73   printk(KERN_DEBUG "(%s:%3.3d ua:%x): " fmt, __FILE__, __LINE__ , adapter->vdev->unit_address, ## args)
  74 #define ibmveth_assert(expr) \
  75   if(!(expr)) {                                   \
  76     printk(KERN_DEBUG "assertion failed (%s:%3.3d ua:%x): %s\n", __FILE__, __LINE__, adapter->vdev->unit_address, #expr); \
  77     BUG(); \
  78   }
  79 #else
  80 #define ibmveth_debug_printk_no_adapter(fmt, args...)
  81 #define ibmveth_debug_printk(fmt, args...)
  82 #define ibmveth_assert(expr)
  83 #endif
  84
  85 static int ibmveth_open(struct net_device *dev);
  86 static int ibmveth_close(struct net_device *dev);
  87 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
  88 static int ibmveth_poll(struct napi_struct *napi, int budget);
  89 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev);
  90 static void ibmveth_set_multicast_list(struct net_device *dev);
  91 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu);
  92 static void ibmveth_proc_register_driver(void);
  93 static void ibmveth_proc_unregister_driver(void);
  94 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter);
  95 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter);
  96 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance);
  97 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter);
  98 static struct kobj_type ktype_veth_pool;
  99
 100 #ifdef CONFIG_PROC_FS
 101 #define IBMVETH_PROC_DIR "ibmveth"
 102 static struct proc_dir_entry *ibmveth_proc_dir;
 103 #endif
 104
 105 static const char ibmveth_driver_name[] = "ibmveth";
 106 static const char ibmveth_driver_string[] = "IBM i/pSeries Virtual Ethernet Driver";
 107 #define ibmveth_driver_version "1.03"
 108
 109 MODULE_AUTHOR("Santiago Leon <santil@us.ibm.com>");
 110 MODULE_DESCRIPTION("IBM i/pSeries Virtual Ethernet Driver");
 111 MODULE_LICENSE("GPL");
 112 MODULE_VERSION(ibmveth_driver_version);
 113
 114 struct ibmveth_stat {
 115         char name[ETH_GSTRING_LEN];
 116         int offset;
 117 };
 118
 119 #define IBMVETH_STAT_OFF(stat) offsetof(struct ibmveth_adapter, stat)
 120 #define IBMVETH_GET_STAT(a, off) *((u64 *)(((unsigned long)(a)) + off))
 121
 122 struct ibmveth_stat ibmveth_stats[] = {
 123         { "replenish_task_cycles", IBMVETH_STAT_OFF(replenish_task_cycles) },
 124         { "replenish_no_mem", IBMVETH_STAT_OFF(replenish_no_mem) },
 125         { "replenish_add_buff_failure", IBMVETH_STAT_OFF(replenish_add_buff_failure) },
 126         { "replenish_add_buff_success", IBMVETH_STAT_OFF(replenish_add_buff_success) },
 127         { "rx_invalid_buffer", IBMVETH_STAT_OFF(rx_invalid_buffer) },
 128         { "rx_no_buffer", IBMVETH_STAT_OFF(rx_no_buffer) },
 129         { "tx_map_failed", IBMVETH_STAT_OFF(tx_map_failed) },
 130         { "tx_send_failed", IBMVETH_STAT_OFF(tx_send_failed) },
 131 };
 132
 133 /* simple methods of getting data from the current rxq entry */
 134 static inline u32 ibmveth_rxq_flags(struct ibmveth_adapter *adapter)
 135 {
 136         return adapter->rx_queue.queue_addr[adapter->rx_queue.index].flags_off;
 137 }
 138
 139 static inline int ibmveth_rxq_toggle(struct ibmveth_adapter *adapter)
 140 {
 141         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_TOGGLE) >> IBMVETH_RXQ_TOGGLE_SHIFT;
 142 }
 143
 144 static inline int ibmveth_rxq_pending_buffer(struct ibmveth_adapter *adapter)
 145 {
 146         return (ibmveth_rxq_toggle(adapter) == adapter->rx_queue.toggle);
 147 }
 148
 149 static inline int ibmveth_rxq_buffer_valid(struct ibmveth_adapter *adapter)
 150 {
 151         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_VALID);
 152 }
 153
 154 static inline int ibmveth_rxq_frame_offset(struct ibmveth_adapter *adapter)
 155 {
 156         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_OFF_MASK);
 157 }
 158
 159 static inline int ibmveth_rxq_frame_length(struct ibmveth_adapter *adapter)
 160 {
 161         return (adapter->rx_queue.queue_addr[adapter->rx_queue.index].length);
 162 }
 163
 164 static inline int ibmveth_rxq_csum_good(struct ibmveth_adapter *adapter)
 165 {
 166         return (ibmveth_rxq_flags(adapter) & IBMVETH_RXQ_CSUM_GOOD);
 167 }
 168
 169 /* setup the initial settings for a buffer pool */
 170 static void ibmveth_init_buffer_pool(struct ibmveth_buff_pool *pool, u32 pool_index, u32 pool_size, u32 buff_size, u32 pool_active)
 171 {
 172         pool->size = pool_size;
 173         pool->index = pool_index;
 174         pool->buff_size = buff_size;
 175         pool->threshold = pool_size / 2;
 176         pool->active = pool_active;
 177 }
 178
 179 /* allocate and setup an buffer pool - called during open */
 180 static int ibmveth_alloc_buffer_pool(struct ibmveth_buff_pool *pool)
 181 {
 182         int i;
 183
 184         pool->free_map = kmalloc(sizeof(u16) * pool->size, GFP_KERNEL);
 185
 186         if(!pool->free_map) {
 187                 return -1;
 188         }
 189
 190         pool->dma_addr = kmalloc(sizeof(dma_addr_t) * pool->size, GFP_KERNEL);
 191         if(!pool->dma_addr) {
 192                 kfree(pool->free_map);
 193                 pool->free_map = NULL;
 194                 return -1;
 195         }
 196
 197         pool->skbuff = kmalloc(sizeof(void*) * pool->size, GFP_KERNEL);
 198
 199         if(!pool->skbuff) {
 200                 kfree(pool->dma_addr);
 201                 pool->dma_addr = NULL;
 202
 203                 kfree(pool->free_map);
 204                 pool->free_map = NULL;
 205                 return -1;
 206         }
 207
 208         memset(pool->skbuff, 0, sizeof(void*) * pool->size);
 209         memset(pool->dma_addr, 0, sizeof(dma_addr_t) * pool->size);
 210
 211         for(i = 0; i < pool->size; ++i) {
 212                 pool->free_map[i] = i;
 213         }
 214
 215         atomic_set(&pool->available, 0);
 216         pool->producer_index = 0;
 217         pool->consumer_index = 0;
 218
 219         return 0;
 220 }
 221
 222 /* replenish the buffers for a pool.  note that we don't need to
 223  * skb_reserve these since they are used for incoming...
 224  */
 225 static void ibmveth_replenish_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 226 {
 227         u32 i;
 228         u32 count = pool->size - atomic_read(&pool->available);
 229         u32 buffers_added = 0;
 230
 231         mb();
 232
 233         for(i = 0; i < count; ++i) {
 234                 struct sk_buff *skb;
 235                 unsigned int free_index, index;
 236                 u64 correlator;
 237                 union ibmveth_buf_desc desc;
 238                 unsigned long lpar_rc;
 239                 dma_addr_t dma_addr;
 240
 241                 skb = alloc_skb(pool->buff_size, GFP_ATOMIC);
 242
 243                 if(!skb) {
 244                         ibmveth_debug_printk("replenish: unable to allocate skb\n");
 245                         adapter->replenish_no_mem++;
 246                         break;
 247                 }
 248
 249                 free_index = pool->consumer_index;
 250                 pool->consumer_index = (pool->consumer_index + 1) % pool->size;
 251                 index = pool->free_map[free_index];
 252
 253                 ibmveth_assert(index != IBM_VETH_INVALID_MAP);
 254                 ibmveth_assert(pool->skbuff[index] == NULL);
 255
 256                 dma_addr = dma_map_single(&adapter->vdev->dev, skb->data,
 257                                 pool->buff_size, DMA_FROM_DEVICE);
 258
 259                 pool->free_map[free_index] = IBM_VETH_INVALID_MAP;
 260                 pool->dma_addr[index] = dma_addr;
 261                 pool->skbuff[index] = skb;
 262
 263                 correlator = ((u64)pool->index << 32) | index;
 264                 *(u64*)skb->data = correlator;
 265
 266                 desc.fields.flags_len = IBMVETH_BUF_VALID | pool->buff_size;
 267                 desc.fields.address = dma_addr;
 268
 269                 lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 270
 271                 if(lpar_rc != H_SUCCESS) {
 272                         pool->free_map[free_index] = index;
 273                         pool->skbuff[index] = NULL;
 274                         if (pool->consumer_index == 0)
 275                                 pool->consumer_index = pool->size - 1;
 276                         else
 277                                 pool->consumer_index--;
 278                         dma_unmap_single(&adapter->vdev->dev,
 279                                         pool->dma_addr[index], pool->buff_size,
 280                                         DMA_FROM_DEVICE);
 281                         dev_kfree_skb_any(skb);
 282                         adapter->replenish_add_buff_failure++;
 283                         break;
 284                 } else {
 285                         buffers_added++;
 286                         adapter->replenish_add_buff_success++;
 287                 }
 288         }
 289
 290         mb();
 291         atomic_add(buffers_added, &(pool->available));
 292 }
 293
 294 /* replenish routine */
 295 static void ibmveth_replenish_task(struct ibmveth_adapter *adapter)
 296 {
 297         int i;
 298
 299         adapter->replenish_task_cycles++;
 300
 301         for(i = 0; i < IbmVethNumBufferPools; i++)
 302                 if(adapter->rx_buff_pool[i].active)
 303                         ibmveth_replenish_buffer_pool(adapter,
 304                                                      &adapter->rx_buff_pool[i]);
 305
 306         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 307 }
 308
 309 /* empty and free ana buffer pool - also used to do cleanup in error paths */
 310 static void ibmveth_free_buffer_pool(struct ibmveth_adapter *adapter, struct ibmveth_buff_pool *pool)
 311 {
 312         int i;
 313
 314         kfree(pool->free_map);
 315         pool->free_map = NULL;
 316
 317         if(pool->skbuff && pool->dma_addr) {
 318                 for(i = 0; i < pool->size; ++i) {
 319                         struct sk_buff *skb = pool->skbuff[i];
 320                         if(skb) {
 321                                 dma_unmap_single(&adapter->vdev->dev,
 322                                                  pool->dma_addr[i],
 323                                                  pool->buff_size,
 324                                                  DMA_FROM_DEVICE);
 325                                 dev_kfree_skb_any(skb);
 326                                 pool->skbuff[i] = NULL;
 327                         }
 328                 }
 329         }
 330
 331         if(pool->dma_addr) {
 332                 kfree(pool->dma_addr);
 333                 pool->dma_addr = NULL;
 334         }
 335
 336         if(pool->skbuff) {
 337                 kfree(pool->skbuff);
 338                 pool->skbuff = NULL;
 339         }
 340 }
 341
 342 /* remove a buffer from a pool */
 343 static void ibmveth_remove_buffer_from_pool(struct ibmveth_adapter *adapter, u64 correlator)
 344 {
 345         unsigned int pool  = correlator >> 32;
 346         unsigned int index = correlator & 0xffffffffUL;
 347         unsigned int free_index;
 348         struct sk_buff *skb;
 349
 350         ibmveth_assert(pool < IbmVethNumBufferPools);
 351         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 352
 353         skb = adapter->rx_buff_pool[pool].skbuff[index];
 354
 355         ibmveth_assert(skb != NULL);
 356
 357         adapter->rx_buff_pool[pool].skbuff[index] = NULL;
 358
 359         dma_unmap_single(&adapter->vdev->dev,
 360                          adapter->rx_buff_pool[pool].dma_addr[index],
 361                          adapter->rx_buff_pool[pool].buff_size,
 362                          DMA_FROM_DEVICE);
 363
 364         free_index = adapter->rx_buff_pool[pool].producer_index;
 365         adapter->rx_buff_pool[pool].producer_index
 366                 = (adapter->rx_buff_pool[pool].producer_index + 1)
 367                 % adapter->rx_buff_pool[pool].size;
 368         adapter->rx_buff_pool[pool].free_map[free_index] = index;
 369
 370         mb();
 371
 372         atomic_dec(&(adapter->rx_buff_pool[pool].available));
 373 }
 374
 375 /* get the current buffer on the rx queue */
 376 static inline struct sk_buff *ibmveth_rxq_get_buffer(struct ibmveth_adapter *adapter)
 377 {
 378         u64 correlator = adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator;
 379         unsigned int pool = correlator >> 32;
 380         unsigned int index = correlator & 0xffffffffUL;
 381
 382         ibmveth_assert(pool < IbmVethNumBufferPools);
 383         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 384
 385         return adapter->rx_buff_pool[pool].skbuff[index];
 386 }
 387
 388 /* recycle the current buffer on the rx queue */
 389 static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
 390 {
 391         u32 q_index = adapter->rx_queue.index;
 392         u64 correlator = adapter->rx_queue.queue_addr[q_index].correlator;
 393         unsigned int pool = correlator >> 32;
 394         unsigned int index = correlator & 0xffffffffUL;
 395         union ibmveth_buf_desc desc;
 396         unsigned long lpar_rc;
 397
 398         ibmveth_assert(pool < IbmVethNumBufferPools);
 399         ibmveth_assert(index < adapter->rx_buff_pool[pool].size);
 400
 401         if(!adapter->rx_buff_pool[pool].active) {
 402                 ibmveth_rxq_harvest_buffer(adapter);
 403                 ibmveth_free_buffer_pool(adapter, &adapter->rx_buff_pool[pool]);
 404                 return;
 405         }
 406
 407         desc.fields.flags_len = IBMVETH_BUF_VALID |
 408                 adapter->rx_buff_pool[pool].buff_size;
 409         desc.fields.address = adapter->rx_buff_pool[pool].dma_addr[index];
 410
 411         lpar_rc = h_add_logical_lan_buffer(adapter->vdev->unit_address, desc.desc);
 412
 413         if(lpar_rc != H_SUCCESS) {
 414                 ibmveth_debug_printk("h_add_logical_lan_buffer failed during recycle rc=%ld", lpar_rc);
 415                 ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 416         }
 417
 418         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 419                 adapter->rx_queue.index = 0;
 420                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 421         }
 422 }
 423
 424 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter)
 425 {
 426         ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
 427
 428         if(++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
 429                 adapter->rx_queue.index = 0;
 430                 adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
 431         }
 432 }
 433
 434 static void ibmveth_cleanup(struct ibmveth_adapter *adapter)
 435 {
 436         int i;
 437
 438         if(adapter->buffer_list_addr != NULL) {
 439                 if(!dma_mapping_error(adapter->buffer_list_dma)) {
 440                         dma_unmap_single(&adapter->vdev->dev,
 441                                         adapter->buffer_list_dma, 4096,
 442                                         DMA_BIDIRECTIONAL);
 443                         adapter->buffer_list_dma = DMA_ERROR_CODE;
 444                 }
 445                 free_page((unsigned long)adapter->buffer_list_addr);
 446                 adapter->buffer_list_addr = NULL;
 447         }
 448
 449         if(adapter->filter_list_addr != NULL) {
 450                 if(!dma_mapping_error(adapter->filter_list_dma)) {
 451                         dma_unmap_single(&adapter->vdev->dev,
 452                                         adapter->filter_list_dma, 4096,
 453                                         DMA_BIDIRECTIONAL);
 454                         adapter->filter_list_dma = DMA_ERROR_CODE;
 455                 }
 456                 free_page((unsigned long)adapter->filter_list_addr);
 457                 adapter->filter_list_addr = NULL;
 458         }
 459
 460         if(adapter->rx_queue.queue_addr != NULL) {
 461                 if(!dma_mapping_error(adapter->rx_queue.queue_dma)) {
 462                         dma_unmap_single(&adapter->vdev->dev,
 463                                         adapter->rx_queue.queue_dma,
 464                                         adapter->rx_queue.queue_len,
 465                                         DMA_BIDIRECTIONAL);
 466                         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
 467                 }
 468                 kfree(adapter->rx_queue.queue_addr);
 469                 adapter->rx_queue.queue_addr = NULL;
 470         }
 471
 472         for(i = 0; i<IbmVethNumBufferPools; i++)
 473                 if (adapter->rx_buff_pool[i].active)
 474                         ibmveth_free_buffer_pool(adapter,
 475                                                  &adapter->rx_buff_pool[i]);
 476 }
 477
 478 static int ibmveth_register_logical_lan(struct ibmveth_adapter *adapter,
 479         union ibmveth_buf_desc rxq_desc, u64 mac_address)
 480 {
 481         int rc, try_again = 1;
 482
 483         /* After a kexec the adapter will still be open, so our attempt to
 484         * open it will fail. So if we get a failure we free the adapter and
 485         * try again, but only once. */
 486 retry:
 487         rc = h_register_logical_lan(adapter->vdev->unit_address,
 488                                     adapter->buffer_list_dma, rxq_desc.desc,
 489                                     adapter->filter_list_dma, mac_address);
 490
 491         if (rc != H_SUCCESS && try_again) {
 492                 do {
 493                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 494                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 495
 496                 try_again = 0;
 497                 goto retry;
 498         }
 499
 500         return rc;
 501 }
 502
 503 static int ibmveth_open(struct net_device *netdev)
 504 {
 505         struct ibmveth_adapter *adapter = netdev->priv;
 506         u64 mac_address = 0;
 507         int rxq_entries = 1;
 508         unsigned long lpar_rc;
 509         int rc;
 510         union ibmveth_buf_desc rxq_desc;
 511         int i;
 512
 513         ibmveth_debug_printk("open starting\n");
 514
 515         napi_enable(&adapter->napi);
 516
 517         for(i = 0; i<IbmVethNumBufferPools; i++)
 518                 rxq_entries += adapter->rx_buff_pool[i].size;
 519
 520         adapter->buffer_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 521         adapter->filter_list_addr = (void*) get_zeroed_page(GFP_KERNEL);
 522
 523         if(!adapter->buffer_list_addr || !adapter->filter_list_addr) {
 524                 ibmveth_error_printk("unable to allocate filter or buffer list pages\n");
 525                 ibmveth_cleanup(adapter);
 526                 napi_disable(&adapter->napi);
 527                 return -ENOMEM;
 528         }
 529
 530         adapter->rx_queue.queue_len = sizeof(struct ibmveth_rx_q_entry) * rxq_entries;
 531         adapter->rx_queue.queue_addr = kmalloc(adapter->rx_queue.queue_len, GFP_KERNEL);
 532
 533         if(!adapter->rx_queue.queue_addr) {
 534                 ibmveth_error_printk("unable to allocate rx queue pages\n");
 535                 ibmveth_cleanup(adapter);
 536                 napi_disable(&adapter->napi);
 537                 return -ENOMEM;
 538         }
 539
 540         adapter->buffer_list_dma = dma_map_single(&adapter->vdev->dev,
 541                         adapter->buffer_list_addr, 4096, DMA_BIDIRECTIONAL);
 542         adapter->filter_list_dma = dma_map_single(&adapter->vdev->dev,
 543                         adapter->filter_list_addr, 4096, DMA_BIDIRECTIONAL);
 544         adapter->rx_queue.queue_dma = dma_map_single(&adapter->vdev->dev,
 545                         adapter->rx_queue.queue_addr,
 546                         adapter->rx_queue.queue_len, DMA_BIDIRECTIONAL);
 547
 548         if((dma_mapping_error(adapter->buffer_list_dma) ) ||
 549            (dma_mapping_error(adapter->filter_list_dma)) ||
 550            (dma_mapping_error(adapter->rx_queue.queue_dma))) {
 551                 ibmveth_error_printk("unable to map filter or buffer list pages\n");
 552                 ibmveth_cleanup(adapter);
 553                 napi_disable(&adapter->napi);
 554                 return -ENOMEM;
 555         }
 556
 557         adapter->rx_queue.index = 0;
 558         adapter->rx_queue.num_slots = rxq_entries;
 559         adapter->rx_queue.toggle = 1;
 560
 561         memcpy(&mac_address, netdev->dev_addr, netdev->addr_len);
 562         mac_address = mac_address >> 16;
 563
 564         rxq_desc.fields.flags_len = IBMVETH_BUF_VALID | adapter->rx_queue.queue_len;
 565         rxq_desc.fields.address = adapter->rx_queue.queue_dma;
 566
 567         ibmveth_debug_printk("buffer list @ 0x%p\n", adapter->buffer_list_addr);
 568         ibmveth_debug_printk("filter list @ 0x%p\n", adapter->filter_list_addr);
 569         ibmveth_debug_printk("receive q   @ 0x%p\n", adapter->rx_queue.queue_addr);
 570
 571         h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE);
 572
 573         lpar_rc = ibmveth_register_logical_lan(adapter, rxq_desc, mac_address);
 574
 575         if(lpar_rc != H_SUCCESS) {
 576                 ibmveth_error_printk("h_register_logical_lan failed with %ld\n", lpar_rc);
 577                 ibmveth_error_printk("buffer TCE:0x%lx filter TCE:0x%lx rxq desc:0x%lx MAC:0x%lx\n",
 578                                      adapter->buffer_list_dma,
 579                                      adapter->filter_list_dma,
 580                                      rxq_desc.desc,
 581                                      mac_address);
 582                 ibmveth_cleanup(adapter);
 583                 napi_disable(&adapter->napi);
 584                 return -ENONET;
 585         }
 586
 587         for(i = 0; i<IbmVethNumBufferPools; i++) {
 588                 if(!adapter->rx_buff_pool[i].active)
 589                         continue;
 590                 if (ibmveth_alloc_buffer_pool(&adapter->rx_buff_pool[i])) {
 591                         ibmveth_error_printk("unable to alloc pool\n");
 592                         adapter->rx_buff_pool[i].active = 0;
 593                         ibmveth_cleanup(adapter);
 594                         napi_disable(&adapter->napi);
 595                         return -ENOMEM ;
 596                 }
 597         }
 598
 599         ibmveth_debug_printk("registering irq 0x%x\n", netdev->irq);
 600         if((rc = request_irq(netdev->irq, &ibmveth_interrupt, 0, netdev->name, netdev)) != 0) {
 601                 ibmveth_error_printk("unable to request irq 0x%x, rc %d\n", netdev->irq, rc);
 602                 do {
 603                         rc = h_free_logical_lan(adapter->vdev->unit_address);
 604                 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
 605
 606                 ibmveth_cleanup(adapter);
 607                 napi_disable(&adapter->napi);
 608                 return rc;
 609         }
 610
 611         ibmveth_debug_printk("initial replenish cycle\n");
 612         ibmveth_interrupt(netdev->irq, netdev);
 613
 614         netif_start_queue(netdev);
 615
 616         ibmveth_debug_printk("open complete\n");
 617
 618         return 0;
 619 }
 620
 621 static int ibmveth_close(struct net_device *netdev)
 622 {
 623         struct ibmveth_adapter *adapter = netdev->priv;
 624         long lpar_rc;
 625
 626         ibmveth_debug_printk("close starting\n");
 627
 628         napi_disable(&adapter->napi);
 629
 630         if (!adapter->pool_config)
 631                 netif_stop_queue(netdev);
 632
 633         free_irq(netdev->irq, netdev);
 634
 635         do {
 636                 lpar_rc = h_free_logical_lan(adapter->vdev->unit_address);
 637         } while (H_IS_LONG_BUSY(lpar_rc) || (lpar_rc == H_BUSY));
 638
 639         if(lpar_rc != H_SUCCESS)
 640         {
 641                 ibmveth_error_printk("h_free_logical_lan failed with %lx, continuing with close\n",
 642                                      lpar_rc);
 643         }
 644
 645         adapter->rx_no_buffer = *(u64*)(((char*)adapter->buffer_list_addr) + 4096 - 8);
 646
 647         ibmveth_cleanup(adapter);
 648
 649         ibmveth_debug_printk("close complete\n");
 650
 651         return 0;
 652 }
 653
 654 static int netdev_get_settings(struct net_device *dev, struct ethtool_cmd *cmd) {
 655         cmd->supported = (SUPPORTED_1000baseT_Full | SUPPORTED_Autoneg | SUPPORTED_FIBRE);
 656         cmd->advertising = (ADVERTISED_1000baseT_Full | ADVERTISED_Autoneg | ADVERTISED_FIBRE);
 657         cmd->speed = SPEED_1000;
 658         cmd->duplex = DUPLEX_FULL;
 659         cmd->port = PORT_FIBRE;
 660         cmd->phy_address = 0;
 661         cmd->transceiver = XCVR_INTERNAL;
 662         cmd->autoneg = AUTONEG_ENABLE;
 663         cmd->maxtxpkt = 0;
 664         cmd->maxrxpkt = 1;
 665         return 0;
 666 }
 667
 668 static void netdev_get_drvinfo (struct net_device *dev, struct ethtool_drvinfo *info) {
 669         strncpy(info->driver, ibmveth_driver_name, sizeof(info->driver) - 1);
 670         strncpy(info->version, ibmveth_driver_version, sizeof(info->version) - 1);
 671 }
 672
 673 static u32 netdev_get_link(struct net_device *dev) {
 674         return 1;
 675 }
 676
 677 static void ibmveth_set_rx_csum_flags(struct net_device *dev, u32 data)
 678 {
 679         struct ibmveth_adapter *adapter = dev->priv;
 680
 681         if (data)
 682                 adapter->rx_csum = 1;
 683         else {
 684                 /*
 685                  * Since the ibmveth firmware interface does not have the concept of
 686                  * separate tx/rx checksum offload enable, if rx checksum is disabled
 687                  * we also have to disable tx checksum offload. Once we disable rx
 688                  * checksum offload, we are no longer allowed to send tx buffers that
 689                  * are not properly checksummed.
 690                  */
 691                 adapter->rx_csum = 0;
 692                 dev->features &= ~NETIF_F_IP_CSUM;
 693         }
 694 }
 695
 696 static void ibmveth_set_tx_csum_flags(struct net_device *dev, u32 data)
 697 {
 698         struct ibmveth_adapter *adapter = dev->priv;
 699
 700         if (data) {
 701                 dev->features |= NETIF_F_IP_CSUM;
 702                 adapter->rx_csum = 1;
 703         } else
 704                 dev->features &= ~NETIF_F_IP_CSUM;
 705 }
 706
 707 static int ibmveth_set_csum_offload(struct net_device *dev, u32 data,
 708                                     void (*done) (struct net_device *, u32))
 709 {
 710         struct ibmveth_adapter *adapter = dev->priv;
 711         u64 set_attr, clr_attr, ret_attr;
 712         long ret;
 713         int rc1 = 0, rc2 = 0;
 714         int restart = 0;
 715
 716         if (netif_running(dev)) {
 717                 restart = 1;
 718                 adapter->pool_config = 1;
 719                 ibmveth_close(dev);
 720                 adapter->pool_config = 0;
 721         }
 722
 723         set_attr = 0;
 724         clr_attr = 0;
 725
 726         if (data)
 727                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 728         else
 729                 clr_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
 730
 731         ret = h_illan_attributes(adapter->vdev->unit_address, 0, 0, &ret_attr);
 732
 733         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
 734             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
 735             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
 736                 ret = h_illan_attributes(adapter->vdev->unit_address, clr_attr,
 737                                          set_attr, &ret_attr);
 738
 739                 if (ret != H_SUCCESS) {
 740                         rc1 = -EIO;
 741                         ibmveth_error_printk("unable to change checksum offload settings."
 742                                              " %d rc=%ld\n", data, ret);
 743
 744                         ret = h_illan_attributes(adapter->vdev->unit_address,
 745                                                  set_attr, clr_attr, &ret_attr);
 746                 } else
 747                         done(dev, data);
 748         } else {
 749                 rc1 = -EIO;
 750                 ibmveth_error_printk("unable to change checksum offload settings."
 751                                      " %d rc=%ld ret_attr=%lx\n", data, ret, ret_attr);
 752         }
 753
 754         if (restart)
 755                 rc2 = ibmveth_open(dev);
 756
 757         return rc1 ? rc1 : rc2;
 758 }
 759
 760 static int ibmveth_set_rx_csum(struct net_device *dev, u32 data)
 761 {
 762         struct ibmveth_adapter *adapter = dev->priv;
 763
 764         if ((data && adapter->rx_csum) || (!data && !adapter->rx_csum))
 765                 return 0;
 766
 767         return ibmveth_set_csum_offload(dev, data, ibmveth_set_rx_csum_flags);
 768 }
 769
 770 static int ibmveth_set_tx_csum(struct net_device *dev, u32 data)
 771 {
 772         struct ibmveth_adapter *adapter = dev->priv;
 773         int rc = 0;
 774
 775         if (data && (dev->features & NETIF_F_IP_CSUM))
 776                 return 0;
 777         if (!data && !(dev->features & NETIF_F_IP_CSUM))
 778                 return 0;
 779
 780         if (data && !adapter->rx_csum)
 781                 rc = ibmveth_set_csum_offload(dev, data, ibmveth_set_tx_csum_flags);
 782         else
 783                 ibmveth_set_tx_csum_flags(dev, data);
 784
 785         return rc;
 786 }
 787
 788 static u32 ibmveth_get_rx_csum(struct net_device *dev)
 789 {
 790         struct ibmveth_adapter *adapter = dev->priv;
 791         return adapter->rx_csum;
 792 }
 793
 794 static void ibmveth_get_strings(struct net_device *dev, u32 stringset, u8 *data)
 795 {
 796         int i;
 797
 798         if (stringset != ETH_SS_STATS)
 799                 return;
 800
 801         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++, data += ETH_GSTRING_LEN)
 802                 memcpy(data, ibmveth_stats[i].name, ETH_GSTRING_LEN);
 803 }
 804
 805 static int ibmveth_get_stats_count(struct net_device *dev)
 806 {
 807         return ARRAY_SIZE(ibmveth_stats);
 808 }
 809
 810 static void ibmveth_get_ethtool_stats(struct net_device *dev,
 811                                       struct ethtool_stats *stats, u64 *data)
 812 {
 813         int i;
 814         struct ibmveth_adapter *adapter = dev->priv;
 815
 816         for (i = 0; i < ARRAY_SIZE(ibmveth_stats); i++)
 817                 data[i] = IBMVETH_GET_STAT(adapter, ibmveth_stats[i].offset);
 818 }
 819
 820 static const struct ethtool_ops netdev_ethtool_ops = {
 821         .get_drvinfo            = netdev_get_drvinfo,
 822         .get_settings           = netdev_get_settings,
 823         .get_link               = netdev_get_link,
 824         .set_tx_csum            = ibmveth_set_tx_csum,
 825         .get_rx_csum            = ibmveth_get_rx_csum,
 826         .set_rx_csum            = ibmveth_set_rx_csum,
 827         .get_strings            = ibmveth_get_strings,
 828         .get_stats_count        = ibmveth_get_stats_count,
 829         .get_ethtool_stats      = ibmveth_get_ethtool_stats,
 830 };
 831
 832 static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 833 {
 834         return -EOPNOTSUPP;
 835 }
 836
 837 #define page_offset(v) ((unsigned long)(v) & ((1 << 12) - 1))
 838
 839 static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 840 {
 841         struct ibmveth_adapter *adapter = netdev->priv;
 842         union ibmveth_buf_desc desc;
 843         unsigned long lpar_rc;
 844         unsigned long correlator;
 845         unsigned long flags;
 846         unsigned int retry_count;
 847         unsigned int tx_dropped = 0;
 848         unsigned int tx_bytes = 0;
 849         unsigned int tx_packets = 0;
 850         unsigned int tx_send_failed = 0;
 851         unsigned int tx_map_failed = 0;
 852
 853         desc.fields.flags_len = IBMVETH_BUF_VALID | skb->len;
 854         desc.fields.address = dma_map_single(&adapter->vdev->dev, skb->data,
 855                                              skb->len, DMA_TO_DEVICE);
 856
 857         if (skb->ip_summed == CHECKSUM_PARTIAL &&
 858             ip_hdr(skb)->protocol != IPPROTO_TCP && skb_checksum_help(skb)) {
 859                 ibmveth_error_printk("tx: failed to checksum packet\n");
 860                 tx_dropped++;
 861                 goto out;
 862         }
 863
 864         if (skb->ip_summed == CHECKSUM_PARTIAL) {
 865                 unsigned char *buf = skb_transport_header(skb) + skb->csum_offset;
 866
 867                 desc.fields.flags_len |= (IBMVETH_BUF_NO_CSUM | IBMVETH_BUF_CSUM_GOOD);
 868
 869                 /* Need to zero out the checksum */
 870                 buf[0] = 0;
 871                 buf[1] = 0;
 872         }
 873
 874         if (dma_mapping_error(desc.fields.address)) {
 875                 ibmveth_error_printk("tx: unable to map xmit buffer\n");
 876                 tx_map_failed++;
 877                 tx_dropped++;
 878                 goto out;
 879         }
 880
 881         /* send the frame. Arbitrarily set retrycount to 1024 */
 882         correlator = 0;
 883         retry_count = 1024;
 884         do {
 885                 lpar_rc = h_send_logical_lan(adapter->vdev->unit_address,
 886                                              desc.desc, 0, 0, 0, 0, 0,
 887                                              correlator, &correlator);
 888         } while ((lpar_rc == H_BUSY) && (retry_count--));
 889
 890         if(lpar_rc != H_SUCCESS && lpar_rc != H_DROPPED) {
 891                 ibmveth_error_printk("tx: h_send_logical_lan failed with rc=%ld\n", lpar_rc);
 892                 ibmveth_error_printk("tx: valid=%d, len=%d, address=0x%08x\n",
 893                                      (desc.fields.flags_len & IBMVETH_BUF_VALID) ? 1 : 0,
 894                                      skb->len, desc.fields.address);
 895                 tx_send_failed++;
 896                 tx_dropped++;
 897         } else {
 898                 tx_packets++;
 899                 tx_bytes += skb->len;
 900                 netdev->trans_start = jiffies;
 901         }
 902
 903         dma_unmap_single(&adapter->vdev->dev, desc.fields.address,
 904                          skb->len, DMA_TO_DEVICE);
 905
 906 out:    spin_lock_irqsave(&adapter->stats_lock, flags);
 907         netdev->stats.tx_dropped += tx_dropped;
 908         netdev->stats.tx_bytes += tx_bytes;
 909         netdev->stats.tx_packets += tx_packets;
 910         adapter->tx_send_failed += tx_send_failed;
 911         adapter->tx_map_failed += tx_map_failed;
 912         spin_unlock_irqrestore(&adapter->stats_lock, flags);
 913
 914         dev_kfree_skb(skb);
 915         return 0;
 916 }
 917
 918 static int ibmveth_poll(struct napi_struct *napi, int budget)
 919 {
 920         struct ibmveth_adapter *adapter = container_of(napi, struct ibmveth_adapter, napi);
 921         struct net_device *netdev = adapter->netdev;
 922         int frames_processed = 0;
 923         unsigned long lpar_rc;
 924
 925  restart_poll:
 926         do {
 927                 struct sk_buff *skb;
 928
 929                 if (!ibmveth_rxq_pending_buffer(adapter))
 930                         break;
 931
 932                 rmb();
 933                 if (!ibmveth_rxq_buffer_valid(adapter)) {
 934                         wmb(); /* suggested by larson1 */
 935                         adapter->rx_invalid_buffer++;
 936                         ibmveth_debug_printk("recycling invalid buffer\n");
 937                         ibmveth_rxq_recycle_buffer(adapter);
 938                 } else {
 939                         int length = ibmveth_rxq_frame_length(adapter);
 940                         int offset = ibmveth_rxq_frame_offset(adapter);
 941                         int csum_good = ibmveth_rxq_csum_good(adapter);
 942
 943                         skb = ibmveth_rxq_get_buffer(adapter);
 944                         if (csum_good)
 945                                 skb->ip_summed = CHECKSUM_UNNECESSARY;
 946
 947                         ibmveth_rxq_harvest_buffer(adapter);
 948
 949                         skb_reserve(skb, offset);
 950                         skb_put(skb, length);
 951                         skb->protocol = eth_type_trans(skb, netdev);
 952
 953                         netif_receive_skb(skb); /* send it up */
 954
 955                         netdev->stats.rx_packets++;
 956                         netdev->stats.rx_bytes += length;
 957                         frames_processed++;
 958                         netdev->last_rx = jiffies;
 959                 }
 960         } while (frames_processed < budget);
 961
 962         ibmveth_replenish_task(adapter);
 963
 964         if (frames_processed < budget) {
 965                 /* We think we are done - reenable interrupts,
 966                  * then check once more to make sure we are done.
 967                  */
 968                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 969                                        VIO_IRQ_ENABLE);
 970
 971                 ibmveth_assert(lpar_rc == H_SUCCESS);
 972
 973                 netif_rx_complete(netdev, napi);
 974
 975                 if (ibmveth_rxq_pending_buffer(adapter) &&
 976                     netif_rx_reschedule(netdev, napi)) {
 977                         lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 978                                                VIO_IRQ_DISABLE);
 979                         goto restart_poll;
 980                 }
 981         }
 982
 983         return frames_processed;
 984 }
 985
 986 static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
 987 {
 988         struct net_device *netdev = dev_instance;
 989         struct ibmveth_adapter *adapter = netdev->priv;
 990         unsigned long lpar_rc;
 991
 992         if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
 993                 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
 994                                        VIO_IRQ_DISABLE);
 995                 ibmveth_assert(lpar_rc == H_SUCCESS);
 996                 __netif_rx_schedule(netdev, &adapter->napi);
 997         }
 998         return IRQ_HANDLED;
 999 }
1000
1001 static void ibmveth_set_multicast_list(struct net_device *netdev)
1002 {
1003         struct ibmveth_adapter *adapter = netdev->priv;
1004         unsigned long lpar_rc;
1005
1006         if((netdev->flags & IFF_PROMISC) || (netdev->mc_count > adapter->mcastFilterSize)) {
1007                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1008                                            IbmVethMcastEnableRecv |
1009                                            IbmVethMcastDisableFiltering,
1010                                            0);
1011                 if(lpar_rc != H_SUCCESS) {
1012                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when entering promisc mode\n", lpar_rc);
1013                 }
1014         } else {
1015                 struct dev_mc_list *mclist = netdev->mc_list;
1016                 int i;
1017                 /* clear the filter table & disable filtering */
1018                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1019                                            IbmVethMcastEnableRecv |
1020                                            IbmVethMcastDisableFiltering |
1021                                            IbmVethMcastClearFilterTable,
1022                                            0);
1023                 if(lpar_rc != H_SUCCESS) {
1024                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when attempting to clear filter table\n", lpar_rc);
1025                 }
1026                 /* add the addresses to the filter table */
1027                 for(i = 0; i < netdev->mc_count; ++i, mclist = mclist->next) {
1028                         // add the multicast address to the filter table
1029                         unsigned long mcast_addr = 0;
1030                         memcpy(((char *)&mcast_addr)+2, mclist->dmi_addr, 6);
1031                         lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1032                                                    IbmVethMcastAddFilter,
1033                                                    mcast_addr);
1034                         if(lpar_rc != H_SUCCESS) {
1035                                 ibmveth_error_printk("h_multicast_ctrl rc=%ld when adding an entry to the filter table\n", lpar_rc);
1036                         }
1037                 }
1038
1039                 /* re-enable filtering */
1040                 lpar_rc = h_multicast_ctrl(adapter->vdev->unit_address,
1041                                            IbmVethMcastEnableFiltering,
1042                                            0);
1043                 if(lpar_rc != H_SUCCESS) {
1044                         ibmveth_error_printk("h_multicast_ctrl rc=%ld when enabling filtering\n", lpar_rc);
1045                 }
1046         }
1047 }
1048
1049 static int ibmveth_change_mtu(struct net_device *dev, int new_mtu)
1050 {
1051         struct ibmveth_adapter *adapter = dev->priv;
1052         int new_mtu_oh = new_mtu + IBMVETH_BUFF_OH;
1053         int reinit = 0;
1054         int i, rc;
1055
1056         if (new_mtu < IBMVETH_MAX_MTU)
1057                 return -EINVAL;
1058
1059         for (i = 0; i < IbmVethNumBufferPools; i++)
1060                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size)
1061                         break;
1062
1063         if (i == IbmVethNumBufferPools)
1064                 return -EINVAL;
1065
1066         /* Look for an active buffer pool that can hold the new MTU */
1067         for(i = 0; i<IbmVethNumBufferPools; i++) {
1068                 if (!adapter->rx_buff_pool[i].active) {
1069                         adapter->rx_buff_pool[i].active = 1;
1070                         reinit = 1;
1071                 }
1072
1073                 if (new_mtu_oh < adapter->rx_buff_pool[i].buff_size) {
1074                         if (reinit && netif_running(adapter->netdev)) {
1075                                 adapter->pool_config = 1;
1076                                 ibmveth_close(adapter->netdev);
1077                                 adapter->pool_config = 0;
1078                                 dev->mtu = new_mtu;
1079                                 if ((rc = ibmveth_open(adapter->netdev)))
1080                                         return rc;
1081                         } else
1082                                 dev->mtu = new_mtu;
1083                         return 0;
1084                 }
1085         }
1086         return -EINVAL;
1087 }
1088
1089 #ifdef CONFIG_NET_POLL_CONTROLLER
1090 static void ibmveth_poll_controller(struct net_device *dev)
1091 {
1092         ibmveth_replenish_task(dev->priv);
1093         ibmveth_interrupt(dev->irq, dev);
1094 }
1095 #endif
1096
1097 static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_id *id)
1098 {
1099         int rc, i;
1100         long ret;
1101         struct net_device *netdev;
1102         struct ibmveth_adapter *adapter;
1103         u64 set_attr, ret_attr;
1104
1105         unsigned char *mac_addr_p;
1106         unsigned int *mcastFilterSize_p;
1107
1108
1109         ibmveth_debug_printk_no_adapter("entering ibmveth_probe for UA 0x%x\n",
1110                                         dev->unit_address);
1111
1112         mac_addr_p = (unsigned char *) vio_get_attribute(dev,
1113                                                 VETH_MAC_ADDR, NULL);
1114         if(!mac_addr_p) {
1115                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find VETH_MAC_ADDR "
1116                                 "attribute\n", __FILE__, __LINE__);
1117                 return 0;
1118         }
1119
1120         mcastFilterSize_p = (unsigned int *) vio_get_attribute(dev,
1121                                                 VETH_MCAST_FILTER_SIZE, NULL);
1122         if(!mcastFilterSize_p) {
1123                 printk(KERN_ERR "(%s:%3.3d) ERROR: Can't find "
1124                                 "VETH_MCAST_FILTER_SIZE attribute\n",
1125                                 __FILE__, __LINE__);
1126                 return 0;
1127         }
1128
1129         netdev = alloc_etherdev(sizeof(struct ibmveth_adapter));
1130
1131         if(!netdev)
1132                 return -ENOMEM;
1133
1134         adapter = netdev->priv;
1135         dev->dev.driver_data = netdev;
1136
1137         adapter->vdev = dev;
1138         adapter->netdev = netdev;
1139         adapter->mcastFilterSize= *mcastFilterSize_p;
1140         adapter->pool_config = 0;
1141
1142         netif_napi_add(netdev, &adapter->napi, ibmveth_poll, 16);
1143
1144         /*      Some older boxes running PHYP non-natively have an OF that
1145                 returns a 8-byte local-mac-address field (and the first
1146                 2 bytes have to be ignored) while newer boxes' OF return
1147                 a 6-byte field. Note that IEEE 1275 specifies that
1148                 local-mac-address must be a 6-byte field.
1149                 The RPA doc specifies that the first byte must be 10b, so
1150                 we'll just look for it to solve this 8 vs. 6 byte field issue */
1151
1152         if ((*mac_addr_p & 0x3) != 0x02)
1153                 mac_addr_p += 2;
1154
1155         adapter->mac_addr = 0;
1156         memcpy(&adapter->mac_addr, mac_addr_p, 6);
1157
1158         netdev->irq = dev->irq;
1159         netdev->open               = ibmveth_open;
1160         netdev->stop               = ibmveth_close;
1161         netdev->hard_start_xmit    = ibmveth_start_xmit;
1162         netdev->set_multicast_list = ibmveth_set_multicast_list;
1163         netdev->do_ioctl           = ibmveth_ioctl;
1164         netdev->ethtool_ops           = &netdev_ethtool_ops;
1165         netdev->change_mtu         = ibmveth_change_mtu;
1166         SET_NETDEV_DEV(netdev, &dev->dev);
1167 #ifdef CONFIG_NET_POLL_CONTROLLER
1168         netdev->poll_controller = ibmveth_poll_controller;
1169 #endif
1170         netdev->features |= NETIF_F_LLTX;
1171         spin_lock_init(&adapter->stats_lock);
1172
1173         memcpy(&netdev->dev_addr, &adapter->mac_addr, netdev->addr_len);
1174
1175         for(i = 0; i<IbmVethNumBufferPools; i++) {
1176                 struct kobject *kobj = &adapter->rx_buff_pool[i].kobj;
1177                 ibmveth_init_buffer_pool(&adapter->rx_buff_pool[i], i,
1178                                          pool_count[i], pool_size[i],
1179                                          pool_active[i]);
1180                 kobj->parent = &dev->dev.kobj;
1181                 sprintf(kobj->name, "pool%d", i);
1182                 kobj->ktype = &ktype_veth_pool;
1183                 kobject_register(kobj);
1184         }
1185
1186         ibmveth_debug_printk("adapter @ 0x%p\n", adapter);
1187
1188         adapter->buffer_list_dma = DMA_ERROR_CODE;
1189         adapter->filter_list_dma = DMA_ERROR_CODE;
1190         adapter->rx_queue.queue_dma = DMA_ERROR_CODE;
1191
1192         ibmveth_debug_printk("registering netdev...\n");
1193
1194         ret = h_illan_attributes(dev->unit_address, 0, 0, &ret_attr);
1195
1196         if (ret == H_SUCCESS && !(ret_attr & IBMVETH_ILLAN_ACTIVE_TRUNK) &&
1197             !(ret_attr & IBMVETH_ILLAN_TRUNK_PRI_MASK) &&
1198             (ret_attr & IBMVETH_ILLAN_PADDED_PKT_CSUM)) {
1199                 set_attr = IBMVETH_ILLAN_IPV4_TCP_CSUM;
1200
1201                 ret = h_illan_attributes(dev->unit_address, 0, set_attr, &ret_attr);
1202
1203                 if (ret == H_SUCCESS) {
1204                         adapter->rx_csum = 1;
1205                         netdev->features |= NETIF_F_IP_CSUM;
1206                 } else
1207                         ret = h_illan_attributes(dev->unit_address, set_attr, 0, &ret_attr);
1208         }
1209
1210         rc = register_netdev(netdev);
1211
1212         if(rc) {
1213                 ibmveth_debug_printk("failed to register netdev rc=%d\n", rc);
1214                 free_netdev(netdev);
1215                 return rc;
1216         }
1217
1218         ibmveth_debug_printk("registered\n");
1219
1220         ibmveth_proc_register_adapter(adapter);
1221
1222         return 0;
1223 }
1224
1225 static int __devexit ibmveth_remove(struct vio_dev *dev)
1226 {
1227         struct net_device *netdev = dev->dev.driver_data;
1228         struct ibmveth_adapter *adapter = netdev->priv;
1229         int i;
1230
1231         for(i = 0; i<IbmVethNumBufferPools; i++)
1232                 kobject_unregister(&adapter->rx_buff_pool[i].kobj);
1233
1234         unregister_netdev(netdev);
1235
1236         ibmveth_proc_unregister_adapter(adapter);
1237
1238         free_netdev(netdev);
1239         return 0;
1240 }
1241
1242 #ifdef CONFIG_PROC_FS
1243 static void ibmveth_proc_register_driver(void)
1244 {
1245         ibmveth_proc_dir = proc_mkdir(IBMVETH_PROC_DIR, init_net.proc_net);
1246         if (ibmveth_proc_dir) {
1247         }
1248 }
1249
1250 static void ibmveth_proc_unregister_driver(void)
1251 {
1252         remove_proc_entry(IBMVETH_PROC_DIR, init_net.proc_net);
1253 }
1254
1255 static void *ibmveth_seq_start(struct seq_file *seq, loff_t *pos)
1256 {
1257         if (*pos == 0) {
1258                 return (void *)1;
1259         } else {
1260                 return NULL;
1261         }
1262 }
1263
1264 static void *ibmveth_seq_next(struct seq_file *seq, void *v, loff_t *pos)
1265 {
1266         ++*pos;
1267         return NULL;
1268 }
1269
1270 static void ibmveth_seq_stop(struct seq_file *seq, void *v)
1271 {
1272 }
1273
1274 static int ibmveth_seq_show(struct seq_file *seq, void *v)
1275 {
1276         struct ibmveth_adapter *adapter = seq->private;
1277         char *current_mac = ((char*) &adapter->netdev->dev_addr);
1278         char *firmware_mac = ((char*) &adapter->mac_addr) ;
1279
1280         seq_printf(seq, "%s %s\n\n", ibmveth_driver_string, ibmveth_driver_version);
1281
1282         seq_printf(seq, "Unit Address:    0x%x\n", adapter->vdev->unit_address);
1283         seq_printf(seq, "Current MAC:     %02X:%02X:%02X:%02X:%02X:%02X\n",
1284                    current_mac[0], current_mac[1], current_mac[2],
1285                    current_mac[3], current_mac[4], current_mac[5]);
1286         seq_printf(seq, "Firmware MAC:    %02X:%02X:%02X:%02X:%02X:%02X\n",
1287                    firmware_mac[0], firmware_mac[1], firmware_mac[2],
1288                    firmware_mac[3], firmware_mac[4], firmware_mac[5]);
1289
1290         seq_printf(seq, "\nAdapter Statistics:\n");
1291         seq_printf(seq, "  TX:  vio_map_single failres:      %ld\n", adapter->tx_map_failed);
1292         seq_printf(seq, "       send failures:               %ld\n", adapter->tx_send_failed);
1293         seq_printf(seq, "  RX:  replenish task cycles:       %ld\n", adapter->replenish_task_cycles);
1294         seq_printf(seq, "       alloc_skb_failures:          %ld\n", adapter->replenish_no_mem);
1295         seq_printf(seq, "       add buffer failures:         %ld\n", adapter->replenish_add_buff_failure);
1296         seq_printf(seq, "       invalid buffers:             %ld\n", adapter->rx_invalid_buffer);
1297         seq_printf(seq, "       no buffers:                  %ld\n", adapter->rx_no_buffer);
1298
1299         return 0;
1300 }
1301 static struct seq_operations ibmveth_seq_ops = {
1302         .start = ibmveth_seq_start,
1303         .next  = ibmveth_seq_next,
1304         .stop  = ibmveth_seq_stop,
1305         .show  = ibmveth_seq_show,
1306 };
1307
1308 static int ibmveth_proc_open(struct inode *inode, struct file *file)
1309 {
1310         struct seq_file *seq;
1311         struct proc_dir_entry *proc;
1312         int rc;
1313
1314         rc = seq_open(file, &ibmveth_seq_ops);
1315         if (!rc) {
1316                 /* recover the pointer buried in proc_dir_entry data */
1317                 seq = file->private_data;
1318                 proc = PDE(inode);
1319                 seq->private = proc->data;
1320         }
1321         return rc;
1322 }
1323
1324 static const struct file_operations ibmveth_proc_fops = {
1325         .owner   = THIS_MODULE,
1326         .open    = ibmveth_proc_open,
1327         .read    = seq_read,
1328         .llseek  = seq_lseek,
1329         .release = seq_release,
1330 };
1331
1332 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1333 {
1334         struct proc_dir_entry *entry;
1335         if (ibmveth_proc_dir) {
1336                 char u_addr[10];
1337                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1338                 entry = create_proc_entry(u_addr, S_IFREG, ibmveth_proc_dir);
1339                 if (!entry) {
1340                         ibmveth_error_printk("Cannot create adapter proc entry");
1341                 } else {
1342                         entry->data = (void *) adapter;
1343                         entry->proc_fops = &ibmveth_proc_fops;
1344                 }
1345         }
1346         return;
1347 }
1348
1349 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1350 {
1351         if (ibmveth_proc_dir) {
1352                 char u_addr[10];
1353                 sprintf(u_addr, "%x", adapter->vdev->unit_address);
1354                 remove_proc_entry(u_addr, ibmveth_proc_dir);
1355         }
1356 }
1357
1358 #else /* CONFIG_PROC_FS */
1359 static void ibmveth_proc_register_adapter(struct ibmveth_adapter *adapter)
1360 {
1361 }
1362
1363 static void ibmveth_proc_unregister_adapter(struct ibmveth_adapter *adapter)
1364 {
1365 }
1366 static void ibmveth_proc_register_driver(void)
1367 {
1368 }
1369
1370 static void ibmveth_proc_unregister_driver(void)
1371 {
1372 }
1373 #endif /* CONFIG_PROC_FS */
1374
1375 static struct attribute veth_active_attr;
1376 static struct attribute veth_num_attr;
1377 static struct attribute veth_size_attr;
1378
1379 static ssize_t veth_pool_show(struct kobject * kobj,
1380                               struct attribute * attr, char * buf)
1381 {
1382         struct ibmveth_buff_pool *pool = container_of(kobj,
1383                                                       struct ibmveth_buff_pool,
1384                                                       kobj);
1385
1386         if (attr == &veth_active_attr)
1387                 return sprintf(buf, "%d\n", pool->active);
1388         else if (attr == &veth_num_attr)
1389                 return sprintf(buf, "%d\n", pool->size);
1390         else if (attr == &veth_size_attr)
1391                 return sprintf(buf, "%d\n", pool->buff_size);
1392         return 0;
1393 }
1394
1395 static ssize_t veth_pool_store(struct kobject * kobj, struct attribute * attr,
1396 const char * buf, size_t count)
1397 {
1398         struct ibmveth_buff_pool *pool = container_of(kobj,
1399                                                       struct ibmveth_buff_pool,
1400                                                       kobj);
1401         struct net_device *netdev =
1402             container_of(kobj->parent, struct device, kobj)->driver_data;
1403         struct ibmveth_adapter *adapter = netdev->priv;
1404         long value = simple_strtol(buf, NULL, 10);
1405         long rc;
1406
1407         if (attr == &veth_active_attr) {
1408                 if (value && !pool->active) {
1409                         if (netif_running(netdev)) {
1410                                 if(ibmveth_alloc_buffer_pool(pool)) {
1411                                         ibmveth_error_printk("unable to alloc pool\n");
1412                                         return -ENOMEM;
1413                                 }
1414                                 pool->active = 1;
1415                                 adapter->pool_config = 1;
1416                                 ibmveth_close(netdev);
1417                                 adapter->pool_config = 0;
1418                                 if ((rc = ibmveth_open(netdev)))
1419                                         return rc;
1420                         } else
1421                                 pool->active = 1;
1422                 } else if (!value && pool->active) {
1423                         int mtu = netdev->mtu + IBMVETH_BUFF_OH;
1424                         int i;
1425                         /* Make sure there is a buffer pool with buffers that
1426                            can hold a packet of the size of the MTU */
1427                         for (i = 0; i < IbmVethNumBufferPools; i++) {
1428                                 if (pool == &adapter->rx_buff_pool[i])
1429                                         continue;
1430                                 if (!adapter->rx_buff_pool[i].active)
1431                                         continue;
1432                                 if (mtu <= adapter->rx_buff_pool[i].buff_size)
1433                                         break;
1434                         }
1435
1436                         if (i == IbmVethNumBufferPools) {
1437                                 ibmveth_error_printk("no active pool >= MTU\n");
1438                                 return -EPERM;
1439                         }
1440
1441                         pool->active = 0;
1442                         if (netif_running(netdev)) {
1443                                 adapter->pool_config = 1;
1444                                 ibmveth_close(netdev);
1445                                 adapter->pool_config = 0;
1446                                 if ((rc = ibmveth_open(netdev)))
1447                                         return rc;
1448                         }
1449                 }
1450         } else if (attr == &veth_num_attr) {
1451                 if (value <= 0 || value > IBMVETH_MAX_POOL_COUNT)
1452                         return -EINVAL;
1453                 else {
1454                         if (netif_running(netdev)) {
1455                                 adapter->pool_config = 1;
1456                                 ibmveth_close(netdev);
1457                                 adapter->pool_config = 0;
1458                                 pool->size = value;
1459                                 if ((rc = ibmveth_open(netdev)))
1460                                         return rc;
1461                         } else
1462                                 pool->size = value;
1463                 }
1464         } else if (attr == &veth_size_attr) {
1465                 if (value <= IBMVETH_BUFF_OH || value > IBMVETH_MAX_BUF_SIZE)
1466                         return -EINVAL;
1467                 else {
1468                         if (netif_running(netdev)) {
1469                                 adapter->pool_config = 1;
1470                                 ibmveth_close(netdev);
1471                                 adapter->pool_config = 0;
1472                                 pool->buff_size = value;
1473                                 if ((rc = ibmveth_open(netdev)))
1474                                         return rc;
1475                         } else
1476                                 pool->buff_size = value;
1477                 }
1478         }
1479
1480         /* kick the interrupt handler to allocate/deallocate pools */
1481         ibmveth_interrupt(netdev->irq, netdev);
1482         return count;
1483 }
1484
1485
1486 #define ATTR(_name, _mode)      \
1487         struct attribute veth_##_name##_attr = {               \
1488         .name = __stringify(_name), .mode = _mode, \
1489         };
1490
1491 static ATTR(active, 0644);
1492 static ATTR(num, 0644);
1493 static ATTR(size, 0644);
1494
1495 static struct attribute * veth_pool_attrs[] = {
1496         &veth_active_attr,
1497         &veth_num_attr,
1498         &veth_size_attr,
1499         NULL,
1500 };
1501
1502 static struct sysfs_ops veth_pool_ops = {
1503         .show   = veth_pool_show,
1504         .store  = veth_pool_store,
1505 };
1506
1507 static struct kobj_type ktype_veth_pool = {
1508         .release        = NULL,
1509         .sysfs_ops      = &veth_pool_ops,
1510         .default_attrs  = veth_pool_attrs,
1511 };
1512
1513
1514 static struct vio_device_id ibmveth_device_table[] __devinitdata= {
1515         { "network", "IBM,l-lan"},
1516         { "", "" }
1517 };
1518 MODULE_DEVICE_TABLE(vio, ibmveth_device_table);
1519
1520 static struct vio_driver ibmveth_driver = {
1521         .id_table       = ibmveth_device_table,
1522         .probe          = ibmveth_probe,
1523         .remove         = ibmveth_remove,
1524         .driver         = {
1525                 .name   = ibmveth_driver_name,
1526                 .owner  = THIS_MODULE,
1527         }
1528 };
1529
1530 static int __init ibmveth_module_init(void)
1531 {
1532         ibmveth_printk("%s: %s %s\n", ibmveth_driver_name, ibmveth_driver_string, ibmveth_driver_version);
1533
1534         ibmveth_proc_register_driver();
1535
1536         return vio_register_driver(&ibmveth_driver);
1537 }
1538
1539 static void __exit ibmveth_module_exit(void)
1540 {
1541         vio_unregister_driver(&ibmveth_driver);
1542         ibmveth_proc_unregister_driver();
1543 }
1544
1545 module_init(ibmveth_module_init);
1546 module_exit(ibmveth_module_exit);