1 /******************************************************************************
2  * mcelog.c
3  * Driver for receiving and transferring machine check error infomation
4  *
5  * Copyright (c) 2012 Intel Corporation
6  * Author: Liu, Jinsong <jinsong.liu@intel.com>
7  * Author: Jiang, Yunhong <yunhong.jiang@intel.com>
8  * Author: Ke, Liping <liping.ke@intel.com>
9  *
10  * This program is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU General Public License version 2
12  * as published by the Free Software Foundation; or, when distributed
13  * separately from the Linux kernel or incorporated into other
14  * software packages, subject to the following license:
15  *
16  * Permission is hereby granted, free of charge, to any person obtaining a copy
17  * of this source file (the "Software"), to deal in the Software without
18  * restriction, including without limitation the rights to use, copy, modify,
19  * merge, publish, distribute, sublicense, and/or sell copies of the Software,
20  * and to permit persons to whom the Software is furnished to do so, subject to
21  * the following conditions:
22  *
23  * The above copyright notice and this permission notice shall be included in
24  * all copies or substantial portions of the Software.
25  *
26  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
27  * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
28  * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
29  * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
30  * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
31  * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
32  * IN THE SOFTWARE.
33  */
34 
35 #define pr_fmt(fmt) "xen_mcelog: " fmt
36 
37 #include <linux/init.h>
38 #include <linux/types.h>
39 #include <linux/kernel.h>
40 #include <linux/slab.h>
41 #include <linux/fs.h>
42 #include <linux/device.h>
43 #include <linux/miscdevice.h>
44 #include <linux/uaccess.h>
45 #include <linux/capability.h>
46 #include <linux/poll.h>
47 #include <linux/sched.h>
48 
49 #include <xen/interface/xen.h>
50 #include <xen/events.h>
51 #include <xen/interface/vcpu.h>
52 #include <xen/xen.h>
53 #include <asm/xen/hypercall.h>
54 #include <asm/xen/hypervisor.h>
55 
56 static struct mc_info g_mi;
57 static struct mcinfo_logical_cpu *g_physinfo;
58 static uint32_t ncpus;
59 
60 static DEFINE_MUTEX(mcelog_lock);
61 
62 static struct xen_mce_log xen_mcelog = {
63 	.signature	= XEN_MCE_LOG_SIGNATURE,
64 	.len		= XEN_MCE_LOG_LEN,
65 	.recordlen	= sizeof(struct xen_mce),
66 };
67 
68 static DEFINE_SPINLOCK(xen_mce_chrdev_state_lock);
69 static int xen_mce_chrdev_open_count;	/* #times opened */
70 static int xen_mce_chrdev_open_exclu;	/* already open exclusive? */
71 
72 static DECLARE_WAIT_QUEUE_HEAD(xen_mce_chrdev_wait);
73 
xen_mce_chrdev_open(struct inode * inode,struct file * file)74 static int xen_mce_chrdev_open(struct inode *inode, struct file *file)
75 {
76 	spin_lock(&xen_mce_chrdev_state_lock);
77 
78 	if (xen_mce_chrdev_open_exclu ||
79 	    (xen_mce_chrdev_open_count && (file->f_flags & O_EXCL))) {
80 		spin_unlock(&xen_mce_chrdev_state_lock);
81 
82 		return -EBUSY;
83 	}
84 
85 	if (file->f_flags & O_EXCL)
86 		xen_mce_chrdev_open_exclu = 1;
87 	xen_mce_chrdev_open_count++;
88 
89 	spin_unlock(&xen_mce_chrdev_state_lock);
90 
91 	return nonseekable_open(inode, file);
92 }
93 
xen_mce_chrdev_release(struct inode * inode,struct file * file)94 static int xen_mce_chrdev_release(struct inode *inode, struct file *file)
95 {
96 	spin_lock(&xen_mce_chrdev_state_lock);
97 
98 	xen_mce_chrdev_open_count--;
99 	xen_mce_chrdev_open_exclu = 0;
100 
101 	spin_unlock(&xen_mce_chrdev_state_lock);
102 
103 	return 0;
104 }
105 
xen_mce_chrdev_read(struct file * filp,char __user * ubuf,size_t usize,loff_t * off)106 static ssize_t xen_mce_chrdev_read(struct file *filp, char __user *ubuf,
107 				size_t usize, loff_t *off)
108 {
109 	char __user *buf = ubuf;
110 	unsigned num;
111 	int i, err;
112 
113 	mutex_lock(&mcelog_lock);
114 
115 	num = xen_mcelog.next;
116 
117 	/* Only supports full reads right now */
118 	err = -EINVAL;
119 	if (*off != 0 || usize < XEN_MCE_LOG_LEN*sizeof(struct xen_mce))
120 		goto out;
121 
122 	err = 0;
123 	for (i = 0; i < num; i++) {
124 		struct xen_mce *m = &xen_mcelog.entry[i];
125 
126 		err |= copy_to_user(buf, m, sizeof(*m));
127 		buf += sizeof(*m);
128 	}
129 
130 	memset(xen_mcelog.entry, 0, num * sizeof(struct xen_mce));
131 	xen_mcelog.next = 0;
132 
133 	if (err)
134 		err = -EFAULT;
135 
136 out:
137 	mutex_unlock(&mcelog_lock);
138 
139 	return err ? err : buf - ubuf;
140 }
141 
xen_mce_chrdev_poll(struct file * file,poll_table * wait)142 static __poll_t xen_mce_chrdev_poll(struct file *file, poll_table *wait)
143 {
144 	poll_wait(file, &xen_mce_chrdev_wait, wait);
145 
146 	if (xen_mcelog.next)
147 		return EPOLLIN | EPOLLRDNORM;
148 
149 	return 0;
150 }
151 
xen_mce_chrdev_ioctl(struct file * f,unsigned int cmd,unsigned long arg)152 static long xen_mce_chrdev_ioctl(struct file *f, unsigned int cmd,
153 				unsigned long arg)
154 {
155 	int __user *p = (int __user *)arg;
156 
157 	if (!capable(CAP_SYS_ADMIN))
158 		return -EPERM;
159 
160 	switch (cmd) {
161 	case MCE_GET_RECORD_LEN:
162 		return put_user(sizeof(struct xen_mce), p);
163 	case MCE_GET_LOG_LEN:
164 		return put_user(XEN_MCE_LOG_LEN, p);
165 	case MCE_GETCLEAR_FLAGS: {
166 		unsigned flags;
167 
168 		do {
169 			flags = xen_mcelog.flags;
170 		} while (cmpxchg(&xen_mcelog.flags, flags, 0) != flags);
171 
172 		return put_user(flags, p);
173 	}
174 	default:
175 		return -ENOTTY;
176 	}
177 }
178 
179 static const struct file_operations xen_mce_chrdev_ops = {
180 	.open			= xen_mce_chrdev_open,
181 	.release		= xen_mce_chrdev_release,
182 	.read			= xen_mce_chrdev_read,
183 	.poll			= xen_mce_chrdev_poll,
184 	.unlocked_ioctl		= xen_mce_chrdev_ioctl,
185 	.llseek			= no_llseek,
186 };
187 
188 static struct miscdevice xen_mce_chrdev_device = {
189 	MISC_MCELOG_MINOR,
190 	"mcelog",
191 	&xen_mce_chrdev_ops,
192 };
193 
194 /*
195  * Caller should hold the mcelog_lock
196  */
xen_mce_log(struct xen_mce * mce)197 static void xen_mce_log(struct xen_mce *mce)
198 {
199 	unsigned entry;
200 
201 	entry = xen_mcelog.next;
202 
203 	/*
204 	 * When the buffer fills up discard new entries.
205 	 * Assume that the earlier errors are the more
206 	 * interesting ones:
207 	 */
208 	if (entry >= XEN_MCE_LOG_LEN) {
209 		set_bit(XEN_MCE_OVERFLOW,
210 			(unsigned long *)&xen_mcelog.flags);
211 		return;
212 	}
213 
214 	memcpy(xen_mcelog.entry + entry, mce, sizeof(struct xen_mce));
215 
216 	xen_mcelog.next++;
217 }
218 
convert_log(struct mc_info * mi)219 static int convert_log(struct mc_info *mi)
220 {
221 	struct mcinfo_common *mic;
222 	struct mcinfo_global *mc_global;
223 	struct mcinfo_bank *mc_bank;
224 	struct xen_mce m;
225 	unsigned int i, j;
226 
227 	mic = NULL;
228 	x86_mcinfo_lookup(&mic, mi, MC_TYPE_GLOBAL);
229 	if (unlikely(!mic)) {
230 		pr_warn("Failed to find global error info\n");
231 		return -ENODEV;
232 	}
233 
234 	memset(&m, 0, sizeof(struct xen_mce));
235 
236 	mc_global = (struct mcinfo_global *)mic;
237 	m.mcgstatus = mc_global->mc_gstatus;
238 	m.apicid = mc_global->mc_apicid;
239 
240 	for (i = 0; i < ncpus; i++)
241 		if (g_physinfo[i].mc_apicid == m.apicid)
242 			break;
243 	if (unlikely(i == ncpus)) {
244 		pr_warn("Failed to match cpu with apicid %d\n", m.apicid);
245 		return -ENODEV;
246 	}
247 
248 	m.socketid = g_physinfo[i].mc_chipid;
249 	m.cpu = m.extcpu = g_physinfo[i].mc_cpunr;
250 	m.cpuvendor = (__u8)g_physinfo[i].mc_vendor;
251 	for (j = 0; j < g_physinfo[i].mc_nmsrvals; ++j)
252 		switch (g_physinfo[i].mc_msrvalues[j].reg) {
253 		case MSR_IA32_MCG_CAP:
254 			m.mcgcap = g_physinfo[i].mc_msrvalues[j].value;
255 			break;
256 
257 		case MSR_PPIN:
258 		case MSR_AMD_PPIN:
259 			m.ppin = g_physinfo[i].mc_msrvalues[j].value;
260 			break;
261 		}
262 
263 	mic = NULL;
264 	x86_mcinfo_lookup(&mic, mi, MC_TYPE_BANK);
265 	if (unlikely(!mic)) {
266 		pr_warn("Fail to find bank error info\n");
267 		return -ENODEV;
268 	}
269 
270 	do {
271 		if ((!mic) || (mic->size == 0) ||
272 		    (mic->type != MC_TYPE_GLOBAL   &&
273 		     mic->type != MC_TYPE_BANK     &&
274 		     mic->type != MC_TYPE_EXTENDED &&
275 		     mic->type != MC_TYPE_RECOVERY))
276 			break;
277 
278 		if (mic->type == MC_TYPE_BANK) {
279 			mc_bank = (struct mcinfo_bank *)mic;
280 			m.misc = mc_bank->mc_misc;
281 			m.status = mc_bank->mc_status;
282 			m.addr = mc_bank->mc_addr;
283 			m.tsc = mc_bank->mc_tsc;
284 			m.bank = mc_bank->mc_bank;
285 			m.finished = 1;
286 			/*log this record*/
287 			xen_mce_log(&m);
288 		}
289 		mic = x86_mcinfo_next(mic);
290 	} while (1);
291 
292 	return 0;
293 }
294 
mc_queue_handle(uint32_t flags)295 static int mc_queue_handle(uint32_t flags)
296 {
297 	struct xen_mc mc_op;
298 	int ret = 0;
299 
300 	mc_op.cmd = XEN_MC_fetch;
301 	set_xen_guest_handle(mc_op.u.mc_fetch.data, &g_mi);
302 	do {
303 		mc_op.u.mc_fetch.flags = flags;
304 		ret = HYPERVISOR_mca(&mc_op);
305 		if (ret) {
306 			pr_err("Failed to fetch %surgent error log\n",
307 			       flags == XEN_MC_URGENT ? "" : "non");
308 			break;
309 		}
310 
311 		if (mc_op.u.mc_fetch.flags & XEN_MC_NODATA ||
312 		    mc_op.u.mc_fetch.flags & XEN_MC_FETCHFAILED)
313 			break;
314 		else {
315 			ret = convert_log(&g_mi);
316 			if (ret)
317 				pr_warn("Failed to convert this error log, continue acking it anyway\n");
318 
319 			mc_op.u.mc_fetch.flags = flags | XEN_MC_ACK;
320 			ret = HYPERVISOR_mca(&mc_op);
321 			if (ret) {
322 				pr_err("Failed to ack previous error log\n");
323 				break;
324 			}
325 		}
326 	} while (1);
327 
328 	return ret;
329 }
330 
331 /* virq handler for machine check error info*/
xen_mce_work_fn(struct work_struct * work)332 static void xen_mce_work_fn(struct work_struct *work)
333 {
334 	int err;
335 
336 	mutex_lock(&mcelog_lock);
337 
338 	/* urgent mc_info */
339 	err = mc_queue_handle(XEN_MC_URGENT);
340 	if (err)
341 		pr_err("Failed to handle urgent mc_info queue, continue handling nonurgent mc_info queue anyway\n");
342 
343 	/* nonurgent mc_info */
344 	err = mc_queue_handle(XEN_MC_NONURGENT);
345 	if (err)
346 		pr_err("Failed to handle nonurgent mc_info queue\n");
347 
348 	/* wake processes polling /dev/mcelog */
349 	wake_up_interruptible(&xen_mce_chrdev_wait);
350 
351 	mutex_unlock(&mcelog_lock);
352 }
353 static DECLARE_WORK(xen_mce_work, xen_mce_work_fn);
354 
xen_mce_interrupt(int irq,void * dev_id)355 static irqreturn_t xen_mce_interrupt(int irq, void *dev_id)
356 {
357 	schedule_work(&xen_mce_work);
358 	return IRQ_HANDLED;
359 }
360 
bind_virq_for_mce(void)361 static int bind_virq_for_mce(void)
362 {
363 	int ret;
364 	struct xen_mc mc_op;
365 
366 	memset(&mc_op, 0, sizeof(struct xen_mc));
367 
368 	/* Fetch physical CPU Numbers */
369 	mc_op.cmd = XEN_MC_physcpuinfo;
370 	set_xen_guest_handle(mc_op.u.mc_physcpuinfo.info, g_physinfo);
371 	ret = HYPERVISOR_mca(&mc_op);
372 	if (ret) {
373 		pr_err("Failed to get CPU numbers\n");
374 		return ret;
375 	}
376 
377 	/* Fetch each CPU Physical Info for later reference*/
378 	ncpus = mc_op.u.mc_physcpuinfo.ncpus;
379 	g_physinfo = kcalloc(ncpus, sizeof(struct mcinfo_logical_cpu),
380 			     GFP_KERNEL);
381 	if (!g_physinfo)
382 		return -ENOMEM;
383 	set_xen_guest_handle(mc_op.u.mc_physcpuinfo.info, g_physinfo);
384 	ret = HYPERVISOR_mca(&mc_op);
385 	if (ret) {
386 		pr_err("Failed to get CPU info\n");
387 		kfree(g_physinfo);
388 		return ret;
389 	}
390 
391 	ret  = bind_virq_to_irqhandler(VIRQ_MCA, 0,
392 				       xen_mce_interrupt, 0, "mce", NULL);
393 	if (ret < 0) {
394 		pr_err("Failed to bind virq\n");
395 		kfree(g_physinfo);
396 		return ret;
397 	}
398 
399 	return 0;
400 }
401 
xen_late_init_mcelog(void)402 static int __init xen_late_init_mcelog(void)
403 {
404 	int ret;
405 
406 	/* Only DOM0 is responsible for MCE logging */
407 	if (!xen_initial_domain())
408 		return -ENODEV;
409 
410 	/* register character device /dev/mcelog for xen mcelog */
411 	ret = misc_register(&xen_mce_chrdev_device);
412 	if (ret)
413 		return ret;
414 
415 	ret = bind_virq_for_mce();
416 	if (ret)
417 		goto deregister;
418 
419 	pr_info("/dev/mcelog registered by Xen\n");
420 
421 	return 0;
422 
423 deregister:
424 	misc_deregister(&xen_mce_chrdev_device);
425 	return ret;
426 }
427 device_initcall(xen_late_init_mcelog);
428