2 * Copyright (c) 2013-2015, 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
33 #include <linux/kernel.h>
34 #include <linux/module.h>
35 #include <linux/random.h>
36 #include <linux/vmalloc.h>
37 #include <linux/hardirq.h>
38 #include <linux/mlx5/driver.h>
39 #include <linux/mlx5/cmd.h>
40 #include "mlx5_core.h"
45 MLX5_HEALTH_POLL_INTERVAL = 2 * HZ,
50 MLX5_HEALTH_SYNDR_FW_ERR = 0x1,
51 MLX5_HEALTH_SYNDR_IRISC_ERR = 0x7,
52 MLX5_HEALTH_SYNDR_HW_UNRECOVERABLE_ERR = 0x8,
53 MLX5_HEALTH_SYNDR_CRC_ERR = 0x9,
54 MLX5_HEALTH_SYNDR_FETCH_PCI_ERR = 0xa,
55 MLX5_HEALTH_SYNDR_HW_FTL_ERR = 0xb,
56 MLX5_HEALTH_SYNDR_ASYNC_EQ_OVERRUN_ERR = 0xc,
57 MLX5_HEALTH_SYNDR_EQ_ERR = 0xd,
58 MLX5_HEALTH_SYNDR_EQ_INV = 0xe,
59 MLX5_HEALTH_SYNDR_FFSER_ERR = 0xf,
60 MLX5_HEALTH_SYNDR_HIGH_TEMP = 0x10
64 MLX5_DROP_NEW_HEALTH_WORK,
65 MLX5_DROP_NEW_RECOVERY_WORK,
68 u8 mlx5_get_nic_state(struct mlx5_core_dev *dev)
70 return (ioread32be(&dev->iseg->cmdq_addr_l_sz) >> 8) & 3;
73 void mlx5_set_nic_state(struct mlx5_core_dev *dev, u8 state)
75 u32 cur_cmdq_addr_l_sz;
77 cur_cmdq_addr_l_sz = ioread32be(&dev->iseg->cmdq_addr_l_sz);
78 iowrite32be((cur_cmdq_addr_l_sz & 0xFFFFF000) |
79 state << MLX5_NIC_IFC_OFFSET,
80 &dev->iseg->cmdq_addr_l_sz);
83 static int in_fatal(struct mlx5_core_dev *dev)
85 struct mlx5_core_health *health = &dev->priv.health;
86 struct health_buffer __iomem *h = health->health;
88 if (mlx5_get_nic_state(dev) == MLX5_NIC_IFC_DISABLED)
91 if (ioread32be(&h->fw_ver) == 0xffffffff)
97 void mlx5_enter_error_state(struct mlx5_core_dev *dev, bool force)
99 mutex_lock(&dev->intf_state_mutex);
100 if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
103 mlx5_core_err(dev, "start\n");
104 if (pci_channel_offline(dev->pdev) || in_fatal(dev) || force) {
105 dev->state = MLX5_DEVICE_STATE_INTERNAL_ERROR;
106 mlx5_cmd_trigger_completions(dev);
109 mlx5_notifier_call_chain(dev->priv.events, MLX5_DEV_EVENT_SYS_ERROR, (void *)1);
110 mlx5_core_err(dev, "end\n");
113 mutex_unlock(&dev->intf_state_mutex);
116 static void mlx5_handle_bad_state(struct mlx5_core_dev *dev)
118 u8 nic_interface = mlx5_get_nic_state(dev);
120 switch (nic_interface) {
121 case MLX5_NIC_IFC_FULL:
122 mlx5_core_warn(dev, "Expected to see disabled NIC but it is full driver\n");
125 case MLX5_NIC_IFC_DISABLED:
126 mlx5_core_warn(dev, "starting teardown\n");
129 case MLX5_NIC_IFC_NO_DRAM_NIC:
130 mlx5_core_warn(dev, "Expected to see disabled NIC but it is no dram nic\n");
133 mlx5_core_warn(dev, "Expected to see disabled NIC but it is has invalid value %d\n",
137 mlx5_disable_device(dev);
140 static void health_recover(struct work_struct *work)
142 struct mlx5_core_health *health;
143 struct delayed_work *dwork;
144 struct mlx5_core_dev *dev;
145 struct mlx5_priv *priv;
148 dwork = container_of(work, struct delayed_work, work);
149 health = container_of(dwork, struct mlx5_core_health, recover_work);
150 priv = container_of(health, struct mlx5_priv, health);
151 dev = container_of(priv, struct mlx5_core_dev, priv);
153 nic_state = mlx5_get_nic_state(dev);
154 if (nic_state == MLX5_NIC_IFC_INVALID) {
155 dev_err(&dev->pdev->dev, "health recovery flow aborted since the nic state is invalid\n");
159 dev_err(&dev->pdev->dev, "starting health recovery flow\n");
160 mlx5_recover_device(dev);
163 /* How much time to wait until health resetting the driver (in msecs) */
164 #define MLX5_RECOVERY_DELAY_MSECS 60000
165 static void health_care(struct work_struct *work)
167 unsigned long recover_delay = msecs_to_jiffies(MLX5_RECOVERY_DELAY_MSECS);
168 struct mlx5_core_health *health;
169 struct mlx5_core_dev *dev;
170 struct mlx5_priv *priv;
173 health = container_of(work, struct mlx5_core_health, work);
174 priv = container_of(health, struct mlx5_priv, health);
175 dev = container_of(priv, struct mlx5_core_dev, priv);
176 mlx5_core_warn(dev, "handling bad device here\n");
177 mlx5_handle_bad_state(dev);
179 spin_lock_irqsave(&health->wq_lock, flags);
180 if (!test_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags))
181 schedule_delayed_work(&health->recover_work, recover_delay);
183 dev_err(&dev->pdev->dev,
184 "new health works are not permitted at this stage\n");
185 spin_unlock_irqrestore(&health->wq_lock, flags);
188 static const char *hsynd_str(u8 synd)
191 case MLX5_HEALTH_SYNDR_FW_ERR:
192 return "firmware internal error";
193 case MLX5_HEALTH_SYNDR_IRISC_ERR:
194 return "irisc not responding";
195 case MLX5_HEALTH_SYNDR_HW_UNRECOVERABLE_ERR:
196 return "unrecoverable hardware error";
197 case MLX5_HEALTH_SYNDR_CRC_ERR:
198 return "firmware CRC error";
199 case MLX5_HEALTH_SYNDR_FETCH_PCI_ERR:
200 return "ICM fetch PCI error";
201 case MLX5_HEALTH_SYNDR_HW_FTL_ERR:
202 return "HW fatal error\n";
203 case MLX5_HEALTH_SYNDR_ASYNC_EQ_OVERRUN_ERR:
204 return "async EQ buffer overrun";
205 case MLX5_HEALTH_SYNDR_EQ_ERR:
207 case MLX5_HEALTH_SYNDR_EQ_INV:
208 return "Invalid EQ referenced";
209 case MLX5_HEALTH_SYNDR_FFSER_ERR:
210 return "FFSER error";
211 case MLX5_HEALTH_SYNDR_HIGH_TEMP:
212 return "High temperature";
214 return "unrecognized error";
218 static void print_health_info(struct mlx5_core_dev *dev)
220 struct mlx5_core_health *health = &dev->priv.health;
221 struct health_buffer __iomem *h = health->health;
226 /* If the syndrome is 0, the device is OK and no need to print buffer */
227 if (!ioread8(&h->synd))
230 for (i = 0; i < ARRAY_SIZE(h->assert_var); i++)
231 dev_err(&dev->pdev->dev, "assert_var[%d] 0x%08x\n", i, ioread32be(h->assert_var + i));
233 dev_err(&dev->pdev->dev, "assert_exit_ptr 0x%08x\n", ioread32be(&h->assert_exit_ptr));
234 dev_err(&dev->pdev->dev, "assert_callra 0x%08x\n", ioread32be(&h->assert_callra));
235 sprintf(fw_str, "%d.%d.%d", fw_rev_maj(dev), fw_rev_min(dev), fw_rev_sub(dev));
236 dev_err(&dev->pdev->dev, "fw_ver %s\n", fw_str);
237 dev_err(&dev->pdev->dev, "hw_id 0x%08x\n", ioread32be(&h->hw_id));
238 dev_err(&dev->pdev->dev, "irisc_index %d\n", ioread8(&h->irisc_index));
239 dev_err(&dev->pdev->dev, "synd 0x%x: %s\n", ioread8(&h->synd), hsynd_str(ioread8(&h->synd)));
240 dev_err(&dev->pdev->dev, "ext_synd 0x%04x\n", ioread16be(&h->ext_synd));
241 fw = ioread32be(&h->fw_ver);
242 dev_err(&dev->pdev->dev, "raw fw_ver 0x%08x\n", fw);
245 static unsigned long get_next_poll_jiffies(void)
249 get_random_bytes(&next, sizeof(next));
251 next += jiffies + MLX5_HEALTH_POLL_INTERVAL;
256 void mlx5_trigger_health_work(struct mlx5_core_dev *dev)
258 struct mlx5_core_health *health = &dev->priv.health;
261 spin_lock_irqsave(&health->wq_lock, flags);
262 if (!test_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags))
263 queue_work(health->wq, &health->work);
265 dev_err(&dev->pdev->dev,
266 "new health works are not permitted at this stage\n");
267 spin_unlock_irqrestore(&health->wq_lock, flags);
270 static void poll_health(struct timer_list *t)
272 struct mlx5_core_dev *dev = from_timer(dev, t, priv.health.timer);
273 struct mlx5_core_health *health = &dev->priv.health;
276 if (dev->state == MLX5_DEVICE_STATE_INTERNAL_ERROR)
279 count = ioread32be(health->health_counter);
280 if (count == health->prev)
281 ++health->miss_counter;
283 health->miss_counter = 0;
285 health->prev = count;
286 if (health->miss_counter == MAX_MISSES) {
287 dev_err(&dev->pdev->dev, "device's health compromised - reached miss count\n");
288 print_health_info(dev);
291 if (in_fatal(dev) && !health->sick) {
293 print_health_info(dev);
294 mlx5_trigger_health_work(dev);
298 mod_timer(&health->timer, get_next_poll_jiffies());
301 void mlx5_start_health_poll(struct mlx5_core_dev *dev)
303 struct mlx5_core_health *health = &dev->priv.health;
305 timer_setup(&health->timer, poll_health, 0);
307 clear_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
308 clear_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
309 health->health = &dev->iseg->health;
310 health->health_counter = &dev->iseg->health_counter;
312 health->timer.expires = round_jiffies(jiffies + MLX5_HEALTH_POLL_INTERVAL);
313 add_timer(&health->timer);
316 void mlx5_stop_health_poll(struct mlx5_core_dev *dev, bool disable_health)
318 struct mlx5_core_health *health = &dev->priv.health;
321 if (disable_health) {
322 spin_lock_irqsave(&health->wq_lock, flags);
323 set_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
324 set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
325 spin_unlock_irqrestore(&health->wq_lock, flags);
328 del_timer_sync(&health->timer);
331 void mlx5_drain_health_wq(struct mlx5_core_dev *dev)
333 struct mlx5_core_health *health = &dev->priv.health;
336 spin_lock_irqsave(&health->wq_lock, flags);
337 set_bit(MLX5_DROP_NEW_HEALTH_WORK, &health->flags);
338 set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
339 spin_unlock_irqrestore(&health->wq_lock, flags);
340 cancel_delayed_work_sync(&health->recover_work);
341 cancel_work_sync(&health->work);
344 void mlx5_drain_health_recovery(struct mlx5_core_dev *dev)
346 struct mlx5_core_health *health = &dev->priv.health;
349 spin_lock_irqsave(&health->wq_lock, flags);
350 set_bit(MLX5_DROP_NEW_RECOVERY_WORK, &health->flags);
351 spin_unlock_irqrestore(&health->wq_lock, flags);
352 cancel_delayed_work_sync(&dev->priv.health.recover_work);
355 void mlx5_health_flush(struct mlx5_core_dev *dev)
357 struct mlx5_core_health *health = &dev->priv.health;
359 flush_workqueue(health->wq);
362 void mlx5_health_cleanup(struct mlx5_core_dev *dev)
364 struct mlx5_core_health *health = &dev->priv.health;
366 destroy_workqueue(health->wq);
369 int mlx5_health_init(struct mlx5_core_dev *dev)
371 struct mlx5_core_health *health;
374 health = &dev->priv.health;
375 name = kmalloc(64, GFP_KERNEL);
379 strcpy(name, "mlx5_health");
380 strcat(name, dev->priv.name);
381 health->wq = create_singlethread_workqueue(name);
385 spin_lock_init(&health->wq_lock);
386 INIT_WORK(&health->work, health_care);
387 INIT_DELAYED_WORK(&health->recover_work, health_recover);