2 * Copyright (c) 2007 Mellanox Technologies. All rights reserved.
4 * This software is available to you under a choice of one of two
5 * licenses. You may choose to be licensed under the terms of the GNU
6 * General Public License (GPL) Version 2, available from the file
7 * COPYING in the main directory of this source tree, or the
8 * OpenIB.org BSD license below:
10 * Redistribution and use in source and binary forms, with or
11 * without modification, are permitted provided that the following
14 * - Redistributions of source code must retain the above
15 * copyright notice, this list of conditions and the following
18 * - Redistributions in binary form must reproduce the above
19 * copyright notice, this list of conditions and the following
20 * disclaimer in the documentation and/or other materials
21 * provided with the distribution.
23 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
24 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
25 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
26 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
27 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
28 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
29 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
37 #include <linux/bitops.h>
38 #include <linux/compiler.h>
39 #include <linux/list.h>
40 #include <linux/mutex.h>
41 #include <linux/netdevice.h>
42 #include <linux/if_vlan.h>
43 #ifdef CONFIG_MLX4_EN_DCB
44 #include <linux/dcbnl.h>
46 #include <linux/cpu_rmap.h>
48 #include <linux/mlx4/device.h>
49 #include <linux/mlx4/qp.h>
50 #include <linux/mlx4/cq.h>
51 #include <linux/mlx4/srq.h>
52 #include <linux/mlx4/doorbell.h>
53 #include <linux/mlx4/cmd.h>
57 #define DRV_NAME "mlx4_en"
58 #define DRV_VERSION "2.0"
59 #define DRV_RELDATE "Dec 2011"
61 #define MLX4_EN_MSG_LEVEL (NETIF_MSG_LINK | NETIF_MSG_IFDOWN)
68 #define MLX4_EN_PAGE_SHIFT 12
69 #define MLX4_EN_PAGE_SIZE (1 << MLX4_EN_PAGE_SHIFT)
70 #define DEF_RX_RINGS 16
71 #define MAX_RX_RINGS 128
72 #define MIN_RX_RINGS 4
74 #define HEADROOM (2048 / TXBB_SIZE + 1)
75 #define STAMP_STRIDE 64
76 #define STAMP_DWORDS (STAMP_STRIDE / 4)
77 #define STAMP_SHIFT 31
78 #define STAMP_VAL 0x7fffffff
79 #define STATS_DELAY (HZ / 4)
80 #define MAX_NUM_OF_FS_RULES 256
82 #define MLX4_EN_FILTER_HASH_SHIFT 4
83 #define MLX4_EN_FILTER_EXPIRY_QUOTA 60
85 /* Typical TSO descriptor with 16 gather entries is 352 bytes... */
86 #define MAX_DESC_SIZE 512
87 #define MAX_DESC_TXBBS (MAX_DESC_SIZE / TXBB_SIZE)
90 * OS related constants and tunables
93 #define MLX4_EN_WATCHDOG_TIMEOUT (15 * HZ)
95 /* Use the maximum between 16384 and a single page */
96 #define MLX4_EN_ALLOC_SIZE PAGE_ALIGN(16384)
97 #define MLX4_EN_ALLOC_ORDER get_order(MLX4_EN_ALLOC_SIZE)
99 /* Receive fragment sizes; we use at most 4 fragments (for 9600 byte MTU
100 * and 4K allocations) */
102 FRAG_SZ0 = 512 - NET_IP_ALIGN,
105 FRAG_SZ3 = MLX4_EN_ALLOC_SIZE
107 #define MLX4_EN_MAX_RX_FRAGS 4
109 /* Maximum ring sizes */
110 #define MLX4_EN_MAX_TX_SIZE 8192
111 #define MLX4_EN_MAX_RX_SIZE 8192
113 /* Minimum ring size for our page-allocation scheme to work */
114 #define MLX4_EN_MIN_RX_SIZE (MLX4_EN_ALLOC_SIZE / SMP_CACHE_BYTES)
115 #define MLX4_EN_MIN_TX_SIZE (4096 / TXBB_SIZE)
117 #define MLX4_EN_SMALL_PKT_SIZE 64
118 #define MLX4_EN_MAX_TX_RING_P_UP 32
119 #define MLX4_EN_NUM_UP 8
120 #define MLX4_EN_DEF_TX_RING_SIZE 512
121 #define MLX4_EN_DEF_RX_RING_SIZE 1024
122 #define MAX_TX_RINGS (MLX4_EN_MAX_TX_RING_P_UP * \
125 /* Target number of packets to coalesce with interrupt moderation */
126 #define MLX4_EN_RX_COAL_TARGET 44
127 #define MLX4_EN_RX_COAL_TIME 0x10
129 #define MLX4_EN_TX_COAL_PKTS 16
130 #define MLX4_EN_TX_COAL_TIME 0x10
132 #define MLX4_EN_RX_RATE_LOW 400000
133 #define MLX4_EN_RX_COAL_TIME_LOW 0
134 #define MLX4_EN_RX_RATE_HIGH 450000
135 #define MLX4_EN_RX_COAL_TIME_HIGH 128
136 #define MLX4_EN_RX_SIZE_THRESH 1024
137 #define MLX4_EN_RX_RATE_THRESH (1000000 / MLX4_EN_RX_COAL_TIME_HIGH)
138 #define MLX4_EN_SAMPLE_INTERVAL 0
139 #define MLX4_EN_AVG_PKT_SMALL 256
141 #define MLX4_EN_AUTO_CONF 0xffff
143 #define MLX4_EN_DEF_RX_PAUSE 1
144 #define MLX4_EN_DEF_TX_PAUSE 1
146 /* Interval between successive polls in the Tx routine when polling is used
147 instead of interrupts (in per-core Tx rings) - should be power of 2 */
148 #define MLX4_EN_TX_POLL_MODER 16
149 #define MLX4_EN_TX_POLL_TIMEOUT (HZ / 4)
151 #define ETH_LLC_SNAP_SIZE 8
153 #define SMALL_PACKET_SIZE (256 - NET_IP_ALIGN)
154 #define HEADER_COPY_SIZE (128 - NET_IP_ALIGN)
155 #define MLX4_LOOPBACK_TEST_PAYLOAD (HEADER_COPY_SIZE - ETH_HLEN)
157 #define MLX4_EN_MIN_MTU 46
158 #define ETH_BCAST 0xffffffffffffULL
160 #define MLX4_EN_LOOPBACK_RETRIES 5
161 #define MLX4_EN_LOOPBACK_TIMEOUT 100
163 #ifdef MLX4_EN_PERF_STAT
164 /* Number of samples to 'average' */
166 #define AVG_FACTOR 1024
167 #define NUM_PERF_STATS NUM_PERF_COUNTERS
169 #define INC_PERF_COUNTER(cnt) (++(cnt))
170 #define ADD_PERF_COUNTER(cnt, add) ((cnt) += (add))
171 #define AVG_PERF_COUNTER(cnt, sample) \
172 ((cnt) = ((cnt) * (AVG_SIZE - 1) + (sample) * AVG_FACTOR) / AVG_SIZE)
173 #define GET_PERF_COUNTER(cnt) (cnt)
174 #define GET_AVG_PERF_COUNTER(cnt) ((cnt) / AVG_FACTOR)
178 #define NUM_PERF_STATS 0
179 #define INC_PERF_COUNTER(cnt) do {} while (0)
180 #define ADD_PERF_COUNTER(cnt, add) do {} while (0)
181 #define AVG_PERF_COUNTER(cnt, sample) do {} while (0)
182 #define GET_PERF_COUNTER(cnt) (0)
183 #define GET_AVG_PERF_COUNTER(cnt) (0)
184 #endif /* MLX4_EN_PERF_STAT */
199 #define ROUNDUP_LOG2(x) ilog2(roundup_pow_of_two(x))
200 #define XNOR(x, y) (!(x) == !(y))
203 struct mlx4_en_tx_info {
213 #define MLX4_EN_BIT_DESC_OWN 0x80000000
214 #define CTRL_SIZE sizeof(struct mlx4_wqe_ctrl_seg)
215 #define MLX4_EN_MEMTYPE_PAD 0x100
216 #define DS_SIZE sizeof(struct mlx4_wqe_data_seg)
219 struct mlx4_en_tx_desc {
220 struct mlx4_wqe_ctrl_seg ctrl;
222 struct mlx4_wqe_data_seg data; /* at least one data segment */
223 struct mlx4_wqe_lso_seg lso;
224 struct mlx4_wqe_inline_seg inl;
228 #define MLX4_EN_USE_SRQ 0x01000000
230 #define MLX4_EN_CX3_LOW_ID 0x1000
231 #define MLX4_EN_CX3_HIGH_ID 0x1005
233 struct mlx4_en_rx_alloc {
239 struct mlx4_en_tx_ring {
240 struct mlx4_hwq_resources wqres;
241 u32 size ; /* number of TXBBs */
244 u16 cqn; /* index of port CQ associated with this ring */
251 struct mlx4_en_tx_info *tx_info;
255 struct mlx4_qp_context context;
257 enum mlx4_qp_state qp_state;
258 struct mlx4_srq dummy;
260 unsigned long packets;
261 unsigned long tx_csum;
264 struct netdev_queue *tx_queue;
267 struct mlx4_en_rx_desc {
268 /* actual number of entries depends on rx ring stride */
269 struct mlx4_wqe_data_seg data[0];
272 struct mlx4_en_rx_ring {
273 struct mlx4_hwq_resources wqres;
274 struct mlx4_en_rx_alloc page_alloc[MLX4_EN_MAX_RX_FRAGS];
275 u32 size ; /* number of Rx descs*/
280 u16 cqn; /* index of port CQ associated with this ring */
288 unsigned long packets;
289 unsigned long csum_ok;
290 unsigned long csum_none;
295 struct mlx4_hwq_resources wqres;
298 struct net_device *dev;
299 struct napi_struct napi;
306 struct mlx4_cqe *buf;
307 #define MLX4_EN_OPCODE_ERROR 0x1e
310 struct mlx4_en_port_profile {
323 struct mlx4_en_profile {
330 u8 num_tx_rings_p_up;
331 struct mlx4_en_port_profile prof[MLX4_MAX_PORTS + 1];
335 struct mlx4_dev *dev;
336 struct pci_dev *pdev;
337 struct mutex state_lock;
338 struct net_device *pndev[MLX4_MAX_PORTS + 1];
341 struct mlx4_en_profile profile;
343 struct workqueue_struct *workqueue;
344 struct device *dma_device;
345 void __iomem *uar_map;
346 struct mlx4_uar priv_uar;
350 u8 mac_removed[MLX4_MAX_PORTS + 1];
354 struct mlx4_en_rss_map {
356 struct mlx4_qp qps[MAX_RX_RINGS];
357 enum mlx4_qp_state state[MAX_RX_RINGS];
358 struct mlx4_qp indir_qp;
359 enum mlx4_qp_state indir_state;
362 struct mlx4_en_port_state {
368 struct mlx4_en_pkt_stats {
369 unsigned long broadcast;
370 unsigned long rx_prio[8];
371 unsigned long tx_prio[8];
372 #define NUM_PKT_STATS 17
375 struct mlx4_en_port_stats {
376 unsigned long tso_packets;
377 unsigned long queue_stopped;
378 unsigned long wake_queue;
379 unsigned long tx_timeout;
380 unsigned long rx_alloc_failed;
381 unsigned long rx_chksum_good;
382 unsigned long rx_chksum_none;
383 unsigned long tx_chksum_offload;
384 #define NUM_PORT_STATS 8
387 struct mlx4_en_perf_stats {
394 #define NUM_PERF_COUNTERS 6
397 enum mlx4_en_mclist_act {
403 struct mlx4_en_mc_list {
404 struct list_head list;
405 enum mlx4_en_mclist_act action;
410 struct mlx4_en_frag_info {
412 u16 frag_prefix_size;
419 #ifdef CONFIG_MLX4_EN_DCB
420 /* Minimal TC BW - setting to 0 will block traffic */
421 #define MLX4_EN_BW_MIN 1
422 #define MLX4_EN_BW_MAX 100 /* Utilize 100% of the line */
424 #define MLX4_EN_TC_ETS 7
428 struct ethtool_flow_id {
429 struct list_head list;
430 struct ethtool_rx_flow_spec flow_spec;
435 MLX4_EN_FLAG_PROMISC = (1 << 0),
436 MLX4_EN_FLAG_MC_PROMISC = (1 << 1),
437 /* whether we need to enable hardware loopback by putting dmac
440 MLX4_EN_FLAG_ENABLE_HW_LOOPBACK = (1 << 2),
441 /* whether we need to drop packets that hardware loopback-ed */
442 MLX4_EN_FLAG_RX_FILTER_NEEDED = (1 << 3),
443 MLX4_EN_FLAG_FORCE_PROMISC = (1 << 4)
446 #define MLX4_EN_MAC_HASH_SIZE (1 << BITS_PER_BYTE)
447 #define MLX4_EN_MAC_HASH_IDX 5
449 struct mlx4_en_priv {
450 struct mlx4_en_dev *mdev;
451 struct mlx4_en_port_profile *prof;
452 struct net_device *dev;
453 unsigned long active_vlans[BITS_TO_LONGS(VLAN_N_VID)];
454 struct net_device_stats stats;
455 struct net_device_stats ret_stats;
456 struct mlx4_en_port_state port_state;
457 spinlock_t stats_lock;
458 struct ethtool_flow_id ethtool_rules[MAX_NUM_OF_FS_RULES];
459 /* To allow rules removal while port is going down */
460 struct list_head ethtool_list;
462 unsigned long last_moder_packets[MAX_RX_RINGS];
463 unsigned long last_moder_tx_packets;
464 unsigned long last_moder_bytes[MAX_RX_RINGS];
465 unsigned long last_moder_jiffies;
466 int last_moder_time[MAX_RX_RINGS];
476 u16 adaptive_rx_coal;
479 u32 validate_loopback;
481 struct mlx4_hwq_resources res;
489 unsigned char prev_mac[ETH_ALEN + 2];
495 struct mlx4_en_rss_map rss_map;
498 u8 num_tx_rings_p_up;
502 struct mlx4_en_frag_info frag_info[MLX4_EN_MAX_RX_FRAGS];
506 struct mlx4_en_tx_ring *tx_ring;
507 struct mlx4_en_rx_ring rx_ring[MAX_RX_RINGS];
508 struct mlx4_en_cq *tx_cq;
509 struct mlx4_en_cq rx_cq[MAX_RX_RINGS];
510 struct mlx4_qp drop_qp;
511 struct work_struct rx_mode_task;
512 struct work_struct mac_task;
513 struct work_struct watchdog_task;
514 struct work_struct linkstate_task;
515 struct delayed_work stats_task;
516 struct mlx4_en_perf_stats pstats;
517 struct mlx4_en_pkt_stats pkstats;
518 struct mlx4_en_port_stats port_stats;
520 struct list_head mc_list;
521 struct list_head curr_list;
523 struct mlx4_en_stat_out_mbox hw_stats;
528 struct hlist_head mac_hash[MLX4_EN_MAC_HASH_SIZE];
530 #ifdef CONFIG_MLX4_EN_DCB
532 u16 maxrate[IEEE_8021QAZ_MAX_TCS];
534 #ifdef CONFIG_RFS_ACCEL
535 spinlock_t filters_lock;
537 struct list_head filters;
538 struct hlist_head filter_hash[1 << MLX4_EN_FILTER_HASH_SHIFT];
544 MLX4_EN_WOL_MAGIC = (1ULL << 61),
545 MLX4_EN_WOL_ENABLED = (1ULL << 62),
548 struct mlx4_mac_entry {
549 struct hlist_node hlist;
550 unsigned char mac[ETH_ALEN + 2];
555 #define MLX4_EN_WOL_DO_MODIFY (1ULL << 63)
557 void mlx4_en_update_loopback_state(struct net_device *dev,
558 netdev_features_t features);
560 void mlx4_en_destroy_netdev(struct net_device *dev);
561 int mlx4_en_init_netdev(struct mlx4_en_dev *mdev, int port,
562 struct mlx4_en_port_profile *prof);
564 int mlx4_en_start_port(struct net_device *dev);
565 void mlx4_en_stop_port(struct net_device *dev, int detach);
567 void mlx4_en_free_resources(struct mlx4_en_priv *priv);
568 int mlx4_en_alloc_resources(struct mlx4_en_priv *priv);
570 int mlx4_en_create_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
571 int entries, int ring, enum cq_type mode);
572 void mlx4_en_destroy_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
573 int mlx4_en_activate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq,
575 void mlx4_en_deactivate_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
576 int mlx4_en_set_cq_moder(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
577 int mlx4_en_arm_cq(struct mlx4_en_priv *priv, struct mlx4_en_cq *cq);
579 void mlx4_en_tx_irq(struct mlx4_cq *mcq);
580 u16 mlx4_en_select_queue(struct net_device *dev, struct sk_buff *skb);
581 netdev_tx_t mlx4_en_xmit(struct sk_buff *skb, struct net_device *dev);
583 int mlx4_en_create_tx_ring(struct mlx4_en_priv *priv, struct mlx4_en_tx_ring *ring,
584 int qpn, u32 size, u16 stride);
585 void mlx4_en_destroy_tx_ring(struct mlx4_en_priv *priv, struct mlx4_en_tx_ring *ring);
586 int mlx4_en_activate_tx_ring(struct mlx4_en_priv *priv,
587 struct mlx4_en_tx_ring *ring,
588 int cq, int user_prio);
589 void mlx4_en_deactivate_tx_ring(struct mlx4_en_priv *priv,
590 struct mlx4_en_tx_ring *ring);
592 int mlx4_en_create_rx_ring(struct mlx4_en_priv *priv,
593 struct mlx4_en_rx_ring *ring,
594 u32 size, u16 stride);
595 void mlx4_en_destroy_rx_ring(struct mlx4_en_priv *priv,
596 struct mlx4_en_rx_ring *ring,
597 u32 size, u16 stride);
598 int mlx4_en_activate_rx_rings(struct mlx4_en_priv *priv);
599 void mlx4_en_deactivate_rx_ring(struct mlx4_en_priv *priv,
600 struct mlx4_en_rx_ring *ring);
601 int mlx4_en_process_rx_cq(struct net_device *dev,
602 struct mlx4_en_cq *cq,
604 int mlx4_en_poll_rx_cq(struct napi_struct *napi, int budget);
605 void mlx4_en_fill_qp_context(struct mlx4_en_priv *priv, int size, int stride,
606 int is_tx, int rss, int qpn, int cqn, int user_prio,
607 struct mlx4_qp_context *context);
608 void mlx4_en_sqp_event(struct mlx4_qp *qp, enum mlx4_event event);
609 int mlx4_en_map_buffer(struct mlx4_buf *buf);
610 void mlx4_en_unmap_buffer(struct mlx4_buf *buf);
612 void mlx4_en_calc_rx_buf(struct net_device *dev);
613 int mlx4_en_config_rss_steer(struct mlx4_en_priv *priv);
614 void mlx4_en_release_rss_steer(struct mlx4_en_priv *priv);
615 int mlx4_en_create_drop_qp(struct mlx4_en_priv *priv);
616 void mlx4_en_destroy_drop_qp(struct mlx4_en_priv *priv);
617 int mlx4_en_free_tx_buf(struct net_device *dev, struct mlx4_en_tx_ring *ring);
618 void mlx4_en_rx_irq(struct mlx4_cq *mcq);
620 int mlx4_SET_MCAST_FLTR(struct mlx4_dev *dev, u8 port, u64 mac, u64 clear, u8 mode);
621 int mlx4_SET_VLAN_FLTR(struct mlx4_dev *dev, struct mlx4_en_priv *priv);
623 int mlx4_en_DUMP_ETH_STATS(struct mlx4_en_dev *mdev, u8 port, u8 reset);
624 int mlx4_en_QUERY_PORT(struct mlx4_en_dev *mdev, u8 port);
626 #ifdef CONFIG_MLX4_EN_DCB
627 extern const struct dcbnl_rtnl_ops mlx4_en_dcbnl_ops;
630 int mlx4_en_setup_tc(struct net_device *dev, u8 up);
632 #ifdef CONFIG_RFS_ACCEL
633 void mlx4_en_cleanup_filters(struct mlx4_en_priv *priv,
634 struct mlx4_en_rx_ring *rx_ring);
637 #define MLX4_EN_NUM_SELF_TEST 5
638 void mlx4_en_ex_selftest(struct net_device *dev, u32 *flags, u64 *buf);
639 u64 mlx4_en_mac_to_u64(u8 *addr);
644 extern const struct ethtool_ops mlx4_en_ethtool_ops;
649 * printk / logging functions
653 int en_print(const char *level, const struct mlx4_en_priv *priv,
654 const char *format, ...);
656 #define en_dbg(mlevel, priv, format, arg...) \
658 if (NETIF_MSG_##mlevel & priv->msg_enable) \
659 en_print(KERN_DEBUG, priv, format, ##arg); \
661 #define en_warn(priv, format, arg...) \
662 en_print(KERN_WARNING, priv, format, ##arg)
663 #define en_err(priv, format, arg...) \
664 en_print(KERN_ERR, priv, format, ##arg)
665 #define en_info(priv, format, arg...) \
666 en_print(KERN_INFO, priv, format, ## arg)
668 #define mlx4_err(mdev, format, arg...) \
669 pr_err("%s %s: " format, DRV_NAME, \
670 dev_name(&mdev->pdev->dev), ##arg)
671 #define mlx4_info(mdev, format, arg...) \
672 pr_info("%s %s: " format, DRV_NAME, \
673 dev_name(&mdev->pdev->dev), ##arg)
674 #define mlx4_warn(mdev, format, arg...) \
675 pr_warning("%s %s: " format, DRV_NAME, \
676 dev_name(&mdev->pdev->dev), ##arg)