1 /*
2  * Copyright (c) 2005 Cisco Systems.  All rights reserved.
3  *
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:
9  *
10  *     Redistribution and use in source and binary forms, with or
11  *     without modification, are permitted provided that the following
12  *     conditions are met:
13  *
14  *      - Redistributions of source code must retain the above
15  *        copyright notice, this list of conditions and the following
16  *        disclaimer.
17  *
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.
22  *
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
30  * SOFTWARE.
31  */
32 
33 #include <linux/jiffies.h>
34 #include <linux/module.h>
35 #include <linux/timer.h>
36 #include <linux/workqueue.h>
37 
38 #include "mthca_dev.h"
39 
40 enum {
41 	MTHCA_CATAS_POLL_INTERVAL	= 5 * HZ,
42 
43 	MTHCA_CATAS_TYPE_INTERNAL	= 0,
44 	MTHCA_CATAS_TYPE_UPLINK		= 3,
45 	MTHCA_CATAS_TYPE_DDR		= 4,
46 	MTHCA_CATAS_TYPE_PARITY		= 5,
47 };
48 
49 static DEFINE_SPINLOCK(catas_lock);
50 
51 static LIST_HEAD(catas_list);
52 static struct workqueue_struct *catas_wq;
53 static struct work_struct catas_work;
54 
55 static int catas_reset_disable;
56 module_param_named(catas_reset_disable, catas_reset_disable, int, 0644);
57 MODULE_PARM_DESC(catas_reset_disable, "disable reset on catastrophic event if nonzero");
58 
catas_reset(struct work_struct * work)59 static void catas_reset(struct work_struct *work)
60 {
61 	struct mthca_dev *dev, *tmpdev;
62 	LIST_HEAD(tlist);
63 	int ret;
64 
65 	mutex_lock(&mthca_device_mutex);
66 
67 	spin_lock_irq(&catas_lock);
68 	list_splice_init(&catas_list, &tlist);
69 	spin_unlock_irq(&catas_lock);
70 
71 	list_for_each_entry_safe(dev, tmpdev, &tlist, catas_err.list) {
72 		struct pci_dev *pdev = dev->pdev;
73 		ret = __mthca_restart_one(dev->pdev);
74 		/* 'dev' now is not valid */
75 		if (ret)
76 			printk(KERN_ERR "mthca %s: Reset failed (%d)\n",
77 			       pci_name(pdev), ret);
78 		else {
79 			struct mthca_dev *d = pci_get_drvdata(pdev);
80 			mthca_dbg(d, "Reset succeeded\n");
81 		}
82 	}
83 
84 	mutex_unlock(&mthca_device_mutex);
85 }
86 
handle_catas(struct mthca_dev * dev)87 static void handle_catas(struct mthca_dev *dev)
88 {
89 	struct ib_event event;
90 	unsigned long flags;
91 	const char *type;
92 	int i;
93 
94 	event.device = &dev->ib_dev;
95 	event.event  = IB_EVENT_DEVICE_FATAL;
96 	event.element.port_num = 0;
97 	dev->active = false;
98 
99 	ib_dispatch_event(&event);
100 
101 	switch (swab32(readl(dev->catas_err.map)) >> 24) {
102 	case MTHCA_CATAS_TYPE_INTERNAL:
103 		type = "internal error";
104 		break;
105 	case MTHCA_CATAS_TYPE_UPLINK:
106 		type = "uplink bus error";
107 		break;
108 	case MTHCA_CATAS_TYPE_DDR:
109 		type = "DDR data error";
110 		break;
111 	case MTHCA_CATAS_TYPE_PARITY:
112 		type = "internal parity error";
113 		break;
114 	default:
115 		type = "unknown error";
116 		break;
117 	}
118 
119 	mthca_err(dev, "Catastrophic error detected: %s\n", type);
120 	for (i = 0; i < dev->catas_err.size; ++i)
121 		mthca_err(dev, "  buf[%02x]: %08x\n",
122 			  i, swab32(readl(dev->catas_err.map + i)));
123 
124 	if (catas_reset_disable)
125 		return;
126 
127 	spin_lock_irqsave(&catas_lock, flags);
128 	list_add(&dev->catas_err.list, &catas_list);
129 	queue_work(catas_wq, &catas_work);
130 	spin_unlock_irqrestore(&catas_lock, flags);
131 }
132 
poll_catas(struct timer_list * t)133 static void poll_catas(struct timer_list *t)
134 {
135 	struct mthca_dev *dev = from_timer(dev, t, catas_err.timer);
136 	int i;
137 
138 	for (i = 0; i < dev->catas_err.size; ++i)
139 		if (readl(dev->catas_err.map + i)) {
140 			handle_catas(dev);
141 			return;
142 		}
143 
144 	mod_timer(&dev->catas_err.timer,
145 		  round_jiffies(jiffies + MTHCA_CATAS_POLL_INTERVAL));
146 }
147 
mthca_start_catas_poll(struct mthca_dev * dev)148 void mthca_start_catas_poll(struct mthca_dev *dev)
149 {
150 	phys_addr_t addr;
151 
152 	timer_setup(&dev->catas_err.timer, poll_catas, 0);
153 	dev->catas_err.map  = NULL;
154 
155 	addr = pci_resource_start(dev->pdev, 0) +
156 		((pci_resource_len(dev->pdev, 0) - 1) &
157 		 dev->catas_err.addr);
158 
159 	dev->catas_err.map = ioremap(addr, dev->catas_err.size * 4);
160 	if (!dev->catas_err.map) {
161 		mthca_warn(dev, "couldn't map catastrophic error region "
162 			   "at 0x%llx/0x%x\n", (unsigned long long) addr,
163 			   dev->catas_err.size * 4);
164 		return;
165 	}
166 
167 	dev->catas_err.timer.expires  = jiffies + MTHCA_CATAS_POLL_INTERVAL;
168 	INIT_LIST_HEAD(&dev->catas_err.list);
169 	add_timer(&dev->catas_err.timer);
170 }
171 
mthca_stop_catas_poll(struct mthca_dev * dev)172 void mthca_stop_catas_poll(struct mthca_dev *dev)
173 {
174 	del_timer_sync(&dev->catas_err.timer);
175 
176 	if (dev->catas_err.map)
177 		iounmap(dev->catas_err.map);
178 
179 	spin_lock_irq(&catas_lock);
180 	list_del(&dev->catas_err.list);
181 	spin_unlock_irq(&catas_lock);
182 }
183 
mthca_catas_init(void)184 int __init mthca_catas_init(void)
185 {
186 	INIT_WORK(&catas_work, catas_reset);
187 
188 	catas_wq = alloc_ordered_workqueue("mthca_catas", WQ_MEM_RECLAIM);
189 	if (!catas_wq)
190 		return -ENOMEM;
191 
192 	return 0;
193 }
194 
mthca_catas_cleanup(void)195 void mthca_catas_cleanup(void)
196 {
197 	destroy_workqueue(catas_wq);
198 }
199