162306a36Sopenharmony_ci/*
262306a36Sopenharmony_ci * Copyright (c) 2005 Cisco Systems.  All rights reserved.
362306a36Sopenharmony_ci *
462306a36Sopenharmony_ci * This software is available to you under a choice of one of two
562306a36Sopenharmony_ci * licenses.  You may choose to be licensed under the terms of the GNU
662306a36Sopenharmony_ci * General Public License (GPL) Version 2, available from the file
762306a36Sopenharmony_ci * COPYING in the main directory of this source tree, or the
862306a36Sopenharmony_ci * OpenIB.org BSD license below:
962306a36Sopenharmony_ci *
1062306a36Sopenharmony_ci *     Redistribution and use in source and binary forms, with or
1162306a36Sopenharmony_ci *     without modification, are permitted provided that the following
1262306a36Sopenharmony_ci *     conditions are met:
1362306a36Sopenharmony_ci *
1462306a36Sopenharmony_ci *      - Redistributions of source code must retain the above
1562306a36Sopenharmony_ci *        copyright notice, this list of conditions and the following
1662306a36Sopenharmony_ci *        disclaimer.
1762306a36Sopenharmony_ci *
1862306a36Sopenharmony_ci *      - Redistributions in binary form must reproduce the above
1962306a36Sopenharmony_ci *        copyright notice, this list of conditions and the following
2062306a36Sopenharmony_ci *        disclaimer in the documentation and/or other materials
2162306a36Sopenharmony_ci *        provided with the distribution.
2262306a36Sopenharmony_ci *
2362306a36Sopenharmony_ci * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
2462306a36Sopenharmony_ci * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
2562306a36Sopenharmony_ci * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
2662306a36Sopenharmony_ci * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
2762306a36Sopenharmony_ci * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
2862306a36Sopenharmony_ci * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
2962306a36Sopenharmony_ci * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
3062306a36Sopenharmony_ci * SOFTWARE.
3162306a36Sopenharmony_ci */
3262306a36Sopenharmony_ci
3362306a36Sopenharmony_ci#include <linux/jiffies.h>
3462306a36Sopenharmony_ci#include <linux/module.h>
3562306a36Sopenharmony_ci#include <linux/timer.h>
3662306a36Sopenharmony_ci#include <linux/workqueue.h>
3762306a36Sopenharmony_ci
3862306a36Sopenharmony_ci#include "mthca_dev.h"
3962306a36Sopenharmony_ci
4062306a36Sopenharmony_cienum {
4162306a36Sopenharmony_ci	MTHCA_CATAS_POLL_INTERVAL	= 5 * HZ,
4262306a36Sopenharmony_ci
4362306a36Sopenharmony_ci	MTHCA_CATAS_TYPE_INTERNAL	= 0,
4462306a36Sopenharmony_ci	MTHCA_CATAS_TYPE_UPLINK		= 3,
4562306a36Sopenharmony_ci	MTHCA_CATAS_TYPE_DDR		= 4,
4662306a36Sopenharmony_ci	MTHCA_CATAS_TYPE_PARITY		= 5,
4762306a36Sopenharmony_ci};
4862306a36Sopenharmony_ci
4962306a36Sopenharmony_cistatic DEFINE_SPINLOCK(catas_lock);
5062306a36Sopenharmony_ci
5162306a36Sopenharmony_cistatic LIST_HEAD(catas_list);
5262306a36Sopenharmony_cistatic struct workqueue_struct *catas_wq;
5362306a36Sopenharmony_cistatic struct work_struct catas_work;
5462306a36Sopenharmony_ci
5562306a36Sopenharmony_cistatic int catas_reset_disable;
5662306a36Sopenharmony_cimodule_param_named(catas_reset_disable, catas_reset_disable, int, 0644);
5762306a36Sopenharmony_ciMODULE_PARM_DESC(catas_reset_disable, "disable reset on catastrophic event if nonzero");
5862306a36Sopenharmony_ci
5962306a36Sopenharmony_cistatic void catas_reset(struct work_struct *work)
6062306a36Sopenharmony_ci{
6162306a36Sopenharmony_ci	struct mthca_dev *dev, *tmpdev;
6262306a36Sopenharmony_ci	LIST_HEAD(tlist);
6362306a36Sopenharmony_ci	int ret;
6462306a36Sopenharmony_ci
6562306a36Sopenharmony_ci	mutex_lock(&mthca_device_mutex);
6662306a36Sopenharmony_ci
6762306a36Sopenharmony_ci	spin_lock_irq(&catas_lock);
6862306a36Sopenharmony_ci	list_splice_init(&catas_list, &tlist);
6962306a36Sopenharmony_ci	spin_unlock_irq(&catas_lock);
7062306a36Sopenharmony_ci
7162306a36Sopenharmony_ci	list_for_each_entry_safe(dev, tmpdev, &tlist, catas_err.list) {
7262306a36Sopenharmony_ci		struct pci_dev *pdev = dev->pdev;
7362306a36Sopenharmony_ci		ret = __mthca_restart_one(dev->pdev);
7462306a36Sopenharmony_ci		/* 'dev' now is not valid */
7562306a36Sopenharmony_ci		if (ret)
7662306a36Sopenharmony_ci			printk(KERN_ERR "mthca %s: Reset failed (%d)\n",
7762306a36Sopenharmony_ci			       pci_name(pdev), ret);
7862306a36Sopenharmony_ci		else {
7962306a36Sopenharmony_ci			struct mthca_dev *d = pci_get_drvdata(pdev);
8062306a36Sopenharmony_ci			mthca_dbg(d, "Reset succeeded\n");
8162306a36Sopenharmony_ci		}
8262306a36Sopenharmony_ci	}
8362306a36Sopenharmony_ci
8462306a36Sopenharmony_ci	mutex_unlock(&mthca_device_mutex);
8562306a36Sopenharmony_ci}
8662306a36Sopenharmony_ci
8762306a36Sopenharmony_cistatic void handle_catas(struct mthca_dev *dev)
8862306a36Sopenharmony_ci{
8962306a36Sopenharmony_ci	struct ib_event event;
9062306a36Sopenharmony_ci	unsigned long flags;
9162306a36Sopenharmony_ci	const char *type;
9262306a36Sopenharmony_ci	int i;
9362306a36Sopenharmony_ci
9462306a36Sopenharmony_ci	event.device = &dev->ib_dev;
9562306a36Sopenharmony_ci	event.event  = IB_EVENT_DEVICE_FATAL;
9662306a36Sopenharmony_ci	event.element.port_num = 0;
9762306a36Sopenharmony_ci	dev->active = false;
9862306a36Sopenharmony_ci
9962306a36Sopenharmony_ci	ib_dispatch_event(&event);
10062306a36Sopenharmony_ci
10162306a36Sopenharmony_ci	switch (swab32(readl(dev->catas_err.map)) >> 24) {
10262306a36Sopenharmony_ci	case MTHCA_CATAS_TYPE_INTERNAL:
10362306a36Sopenharmony_ci		type = "internal error";
10462306a36Sopenharmony_ci		break;
10562306a36Sopenharmony_ci	case MTHCA_CATAS_TYPE_UPLINK:
10662306a36Sopenharmony_ci		type = "uplink bus error";
10762306a36Sopenharmony_ci		break;
10862306a36Sopenharmony_ci	case MTHCA_CATAS_TYPE_DDR:
10962306a36Sopenharmony_ci		type = "DDR data error";
11062306a36Sopenharmony_ci		break;
11162306a36Sopenharmony_ci	case MTHCA_CATAS_TYPE_PARITY:
11262306a36Sopenharmony_ci		type = "internal parity error";
11362306a36Sopenharmony_ci		break;
11462306a36Sopenharmony_ci	default:
11562306a36Sopenharmony_ci		type = "unknown error";
11662306a36Sopenharmony_ci		break;
11762306a36Sopenharmony_ci	}
11862306a36Sopenharmony_ci
11962306a36Sopenharmony_ci	mthca_err(dev, "Catastrophic error detected: %s\n", type);
12062306a36Sopenharmony_ci	for (i = 0; i < dev->catas_err.size; ++i)
12162306a36Sopenharmony_ci		mthca_err(dev, "  buf[%02x]: %08x\n",
12262306a36Sopenharmony_ci			  i, swab32(readl(dev->catas_err.map + i)));
12362306a36Sopenharmony_ci
12462306a36Sopenharmony_ci	if (catas_reset_disable)
12562306a36Sopenharmony_ci		return;
12662306a36Sopenharmony_ci
12762306a36Sopenharmony_ci	spin_lock_irqsave(&catas_lock, flags);
12862306a36Sopenharmony_ci	list_add(&dev->catas_err.list, &catas_list);
12962306a36Sopenharmony_ci	queue_work(catas_wq, &catas_work);
13062306a36Sopenharmony_ci	spin_unlock_irqrestore(&catas_lock, flags);
13162306a36Sopenharmony_ci}
13262306a36Sopenharmony_ci
13362306a36Sopenharmony_cistatic void poll_catas(struct timer_list *t)
13462306a36Sopenharmony_ci{
13562306a36Sopenharmony_ci	struct mthca_dev *dev = from_timer(dev, t, catas_err.timer);
13662306a36Sopenharmony_ci	int i;
13762306a36Sopenharmony_ci
13862306a36Sopenharmony_ci	for (i = 0; i < dev->catas_err.size; ++i)
13962306a36Sopenharmony_ci		if (readl(dev->catas_err.map + i)) {
14062306a36Sopenharmony_ci			handle_catas(dev);
14162306a36Sopenharmony_ci			return;
14262306a36Sopenharmony_ci		}
14362306a36Sopenharmony_ci
14462306a36Sopenharmony_ci	mod_timer(&dev->catas_err.timer,
14562306a36Sopenharmony_ci		  round_jiffies(jiffies + MTHCA_CATAS_POLL_INTERVAL));
14662306a36Sopenharmony_ci}
14762306a36Sopenharmony_ci
14862306a36Sopenharmony_civoid mthca_start_catas_poll(struct mthca_dev *dev)
14962306a36Sopenharmony_ci{
15062306a36Sopenharmony_ci	phys_addr_t addr;
15162306a36Sopenharmony_ci
15262306a36Sopenharmony_ci	timer_setup(&dev->catas_err.timer, poll_catas, 0);
15362306a36Sopenharmony_ci	dev->catas_err.map  = NULL;
15462306a36Sopenharmony_ci
15562306a36Sopenharmony_ci	addr = pci_resource_start(dev->pdev, 0) +
15662306a36Sopenharmony_ci		((pci_resource_len(dev->pdev, 0) - 1) &
15762306a36Sopenharmony_ci		 dev->catas_err.addr);
15862306a36Sopenharmony_ci
15962306a36Sopenharmony_ci	dev->catas_err.map = ioremap(addr, dev->catas_err.size * 4);
16062306a36Sopenharmony_ci	if (!dev->catas_err.map) {
16162306a36Sopenharmony_ci		mthca_warn(dev, "couldn't map catastrophic error region "
16262306a36Sopenharmony_ci			   "at 0x%llx/0x%x\n", (unsigned long long) addr,
16362306a36Sopenharmony_ci			   dev->catas_err.size * 4);
16462306a36Sopenharmony_ci		return;
16562306a36Sopenharmony_ci	}
16662306a36Sopenharmony_ci
16762306a36Sopenharmony_ci	dev->catas_err.timer.expires  = jiffies + MTHCA_CATAS_POLL_INTERVAL;
16862306a36Sopenharmony_ci	INIT_LIST_HEAD(&dev->catas_err.list);
16962306a36Sopenharmony_ci	add_timer(&dev->catas_err.timer);
17062306a36Sopenharmony_ci}
17162306a36Sopenharmony_ci
17262306a36Sopenharmony_civoid mthca_stop_catas_poll(struct mthca_dev *dev)
17362306a36Sopenharmony_ci{
17462306a36Sopenharmony_ci	del_timer_sync(&dev->catas_err.timer);
17562306a36Sopenharmony_ci
17662306a36Sopenharmony_ci	if (dev->catas_err.map)
17762306a36Sopenharmony_ci		iounmap(dev->catas_err.map);
17862306a36Sopenharmony_ci
17962306a36Sopenharmony_ci	spin_lock_irq(&catas_lock);
18062306a36Sopenharmony_ci	list_del(&dev->catas_err.list);
18162306a36Sopenharmony_ci	spin_unlock_irq(&catas_lock);
18262306a36Sopenharmony_ci}
18362306a36Sopenharmony_ci
18462306a36Sopenharmony_ciint __init mthca_catas_init(void)
18562306a36Sopenharmony_ci{
18662306a36Sopenharmony_ci	INIT_WORK(&catas_work, catas_reset);
18762306a36Sopenharmony_ci
18862306a36Sopenharmony_ci	catas_wq = alloc_ordered_workqueue("mthca_catas", WQ_MEM_RECLAIM);
18962306a36Sopenharmony_ci	if (!catas_wq)
19062306a36Sopenharmony_ci		return -ENOMEM;
19162306a36Sopenharmony_ci
19262306a36Sopenharmony_ci	return 0;
19362306a36Sopenharmony_ci}
19462306a36Sopenharmony_ci
19562306a36Sopenharmony_civoid mthca_catas_cleanup(void)
19662306a36Sopenharmony_ci{
19762306a36Sopenharmony_ci	destroy_workqueue(catas_wq);
19862306a36Sopenharmony_ci}
199