| rjw | 1f88458 | 2022-01-06 17:20:42 +0800 | [diff] [blame] | 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 |  | 
|  | 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 |  | 
|  | 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 |  | 
|  | 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 |  | 
|  | 142 | static unsigned int 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 POLLIN | POLLRDNORM; | 
|  | 148 |  | 
|  | 149 | return 0; | 
|  | 150 | } | 
|  | 151 |  | 
|  | 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 | */ | 
|  | 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 |  | 
|  | 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 | uint32_t i; | 
|  | 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 | m.mcgcap = g_physinfo[i].mc_msrvalues[__MC_MSR_MCGCAP].value; | 
|  | 252 |  | 
|  | 253 | mic = NULL; | 
|  | 254 | x86_mcinfo_lookup(&mic, mi, MC_TYPE_BANK); | 
|  | 255 | if (unlikely(!mic)) { | 
|  | 256 | pr_warn("Fail to find bank error info\n"); | 
|  | 257 | return -ENODEV; | 
|  | 258 | } | 
|  | 259 |  | 
|  | 260 | do { | 
|  | 261 | if ((!mic) || (mic->size == 0) || | 
|  | 262 | (mic->type != MC_TYPE_GLOBAL   && | 
|  | 263 | mic->type != MC_TYPE_BANK     && | 
|  | 264 | mic->type != MC_TYPE_EXTENDED && | 
|  | 265 | mic->type != MC_TYPE_RECOVERY)) | 
|  | 266 | break; | 
|  | 267 |  | 
|  | 268 | if (mic->type == MC_TYPE_BANK) { | 
|  | 269 | mc_bank = (struct mcinfo_bank *)mic; | 
|  | 270 | m.misc = mc_bank->mc_misc; | 
|  | 271 | m.status = mc_bank->mc_status; | 
|  | 272 | m.addr = mc_bank->mc_addr; | 
|  | 273 | m.tsc = mc_bank->mc_tsc; | 
|  | 274 | m.bank = mc_bank->mc_bank; | 
|  | 275 | m.finished = 1; | 
|  | 276 | /*log this record*/ | 
|  | 277 | xen_mce_log(&m); | 
|  | 278 | } | 
|  | 279 | mic = x86_mcinfo_next(mic); | 
|  | 280 | } while (1); | 
|  | 281 |  | 
|  | 282 | return 0; | 
|  | 283 | } | 
|  | 284 |  | 
|  | 285 | static int mc_queue_handle(uint32_t flags) | 
|  | 286 | { | 
|  | 287 | struct xen_mc mc_op; | 
|  | 288 | int ret = 0; | 
|  | 289 |  | 
|  | 290 | mc_op.cmd = XEN_MC_fetch; | 
|  | 291 | mc_op.interface_version = XEN_MCA_INTERFACE_VERSION; | 
|  | 292 | set_xen_guest_handle(mc_op.u.mc_fetch.data, &g_mi); | 
|  | 293 | do { | 
|  | 294 | mc_op.u.mc_fetch.flags = flags; | 
|  | 295 | ret = HYPERVISOR_mca(&mc_op); | 
|  | 296 | if (ret) { | 
|  | 297 | pr_err("Failed to fetch %surgent error log\n", | 
|  | 298 | flags == XEN_MC_URGENT ? "" : "non"); | 
|  | 299 | break; | 
|  | 300 | } | 
|  | 301 |  | 
|  | 302 | if (mc_op.u.mc_fetch.flags & XEN_MC_NODATA || | 
|  | 303 | mc_op.u.mc_fetch.flags & XEN_MC_FETCHFAILED) | 
|  | 304 | break; | 
|  | 305 | else { | 
|  | 306 | ret = convert_log(&g_mi); | 
|  | 307 | if (ret) | 
|  | 308 | pr_warn("Failed to convert this error log, continue acking it anyway\n"); | 
|  | 309 |  | 
|  | 310 | mc_op.u.mc_fetch.flags = flags | XEN_MC_ACK; | 
|  | 311 | ret = HYPERVISOR_mca(&mc_op); | 
|  | 312 | if (ret) { | 
|  | 313 | pr_err("Failed to ack previous error log\n"); | 
|  | 314 | break; | 
|  | 315 | } | 
|  | 316 | } | 
|  | 317 | } while (1); | 
|  | 318 |  | 
|  | 319 | return ret; | 
|  | 320 | } | 
|  | 321 |  | 
|  | 322 | /* virq handler for machine check error info*/ | 
|  | 323 | static void xen_mce_work_fn(struct work_struct *work) | 
|  | 324 | { | 
|  | 325 | int err; | 
|  | 326 |  | 
|  | 327 | mutex_lock(&mcelog_lock); | 
|  | 328 |  | 
|  | 329 | /* urgent mc_info */ | 
|  | 330 | err = mc_queue_handle(XEN_MC_URGENT); | 
|  | 331 | if (err) | 
|  | 332 | pr_err("Failed to handle urgent mc_info queue, continue handling nonurgent mc_info queue anyway\n"); | 
|  | 333 |  | 
|  | 334 | /* nonurgent mc_info */ | 
|  | 335 | err = mc_queue_handle(XEN_MC_NONURGENT); | 
|  | 336 | if (err) | 
|  | 337 | pr_err("Failed to handle nonurgent mc_info queue\n"); | 
|  | 338 |  | 
|  | 339 | /* wake processes polling /dev/mcelog */ | 
|  | 340 | wake_up_interruptible(&xen_mce_chrdev_wait); | 
|  | 341 |  | 
|  | 342 | mutex_unlock(&mcelog_lock); | 
|  | 343 | } | 
|  | 344 | static DECLARE_WORK(xen_mce_work, xen_mce_work_fn); | 
|  | 345 |  | 
|  | 346 | static irqreturn_t xen_mce_interrupt(int irq, void *dev_id) | 
|  | 347 | { | 
|  | 348 | schedule_work(&xen_mce_work); | 
|  | 349 | return IRQ_HANDLED; | 
|  | 350 | } | 
|  | 351 |  | 
|  | 352 | static int bind_virq_for_mce(void) | 
|  | 353 | { | 
|  | 354 | int ret; | 
|  | 355 | struct xen_mc mc_op; | 
|  | 356 |  | 
|  | 357 | memset(&mc_op, 0, sizeof(struct xen_mc)); | 
|  | 358 |  | 
|  | 359 | /* Fetch physical CPU Numbers */ | 
|  | 360 | mc_op.cmd = XEN_MC_physcpuinfo; | 
|  | 361 | mc_op.interface_version = XEN_MCA_INTERFACE_VERSION; | 
|  | 362 | set_xen_guest_handle(mc_op.u.mc_physcpuinfo.info, g_physinfo); | 
|  | 363 | ret = HYPERVISOR_mca(&mc_op); | 
|  | 364 | if (ret) { | 
|  | 365 | pr_err("Failed to get CPU numbers\n"); | 
|  | 366 | return ret; | 
|  | 367 | } | 
|  | 368 |  | 
|  | 369 | /* Fetch each CPU Physical Info for later reference*/ | 
|  | 370 | ncpus = mc_op.u.mc_physcpuinfo.ncpus; | 
|  | 371 | g_physinfo = kcalloc(ncpus, sizeof(struct mcinfo_logical_cpu), | 
|  | 372 | GFP_KERNEL); | 
|  | 373 | if (!g_physinfo) | 
|  | 374 | return -ENOMEM; | 
|  | 375 | set_xen_guest_handle(mc_op.u.mc_physcpuinfo.info, g_physinfo); | 
|  | 376 | ret = HYPERVISOR_mca(&mc_op); | 
|  | 377 | if (ret) { | 
|  | 378 | pr_err("Failed to get CPU info\n"); | 
|  | 379 | kfree(g_physinfo); | 
|  | 380 | return ret; | 
|  | 381 | } | 
|  | 382 |  | 
|  | 383 | ret  = bind_virq_to_irqhandler(VIRQ_MCA, 0, | 
|  | 384 | xen_mce_interrupt, 0, "mce", NULL); | 
|  | 385 | if (ret < 0) { | 
|  | 386 | pr_err("Failed to bind virq\n"); | 
|  | 387 | kfree(g_physinfo); | 
|  | 388 | return ret; | 
|  | 389 | } | 
|  | 390 |  | 
|  | 391 | return 0; | 
|  | 392 | } | 
|  | 393 |  | 
|  | 394 | static int __init xen_late_init_mcelog(void) | 
|  | 395 | { | 
|  | 396 | int ret; | 
|  | 397 |  | 
|  | 398 | /* Only DOM0 is responsible for MCE logging */ | 
|  | 399 | if (!xen_initial_domain()) | 
|  | 400 | return -ENODEV; | 
|  | 401 |  | 
|  | 402 | /* register character device /dev/mcelog for xen mcelog */ | 
|  | 403 | ret = misc_register(&xen_mce_chrdev_device); | 
|  | 404 | if (ret) | 
|  | 405 | return ret; | 
|  | 406 |  | 
|  | 407 | ret = bind_virq_for_mce(); | 
|  | 408 | if (ret) | 
|  | 409 | goto deregister; | 
|  | 410 |  | 
|  | 411 | pr_info("/dev/mcelog registered by Xen\n"); | 
|  | 412 |  | 
|  | 413 | return 0; | 
|  | 414 |  | 
|  | 415 | deregister: | 
|  | 416 | misc_deregister(&xen_mce_chrdev_device); | 
|  | 417 | return ret; | 
|  | 418 | } | 
|  | 419 | device_initcall(xen_late_init_mcelog); |