393 lines · c
1/* SPDX-License-Identifier: MIT */2/******************************************************************************3 * arch-x86/mca.h4 * Guest OS machine check interface to x86 Xen.5 *6 * Contributed by Advanced Micro Devices, Inc.7 * Author: Christoph Egger <Christoph.Egger@amd.com>8 *9 * Updated by Intel Corporation10 * Author: Liu, Jinsong <jinsong.liu@intel.com>11 *12 * Permission is hereby granted, free of charge, to any person obtaining a copy13 * of this software and associated documentation files (the "Software"), to14 * deal in the Software without restriction, including without limitation the15 * rights to use, copy, modify, merge, publish, distribute, sublicense, and/or16 * sell copies of the Software, and to permit persons to whom the Software is17 * furnished to do so, subject to the following conditions:18 *19 * The above copyright notice and this permission notice shall be included in20 * all copies or substantial portions of the Software.21 *22 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR23 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,24 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE25 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER26 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING27 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER28 * DEALINGS IN THE SOFTWARE.29 */30 31#ifndef __XEN_PUBLIC_ARCH_X86_MCA_H__32#define __XEN_PUBLIC_ARCH_X86_MCA_H__33 34/* Hypercall */35#define __HYPERVISOR_mca __HYPERVISOR_arch_036 37#define XEN_MCA_INTERFACE_VERSION 0x01ecc00338 39/* IN: Dom0 calls hypercall to retrieve nonurgent error log entry */40#define XEN_MC_NONURGENT 0x141/* IN: Dom0 calls hypercall to retrieve urgent error log entry */42#define XEN_MC_URGENT 0x243/* IN: Dom0 acknowledges previosly-fetched error log entry */44#define XEN_MC_ACK 0x445 46/* OUT: All is ok */47#define XEN_MC_OK 0x048/* OUT: Domain could not fetch data. */49#define XEN_MC_FETCHFAILED 0x150/* OUT: There was no machine check data to fetch. */51#define XEN_MC_NODATA 0x252 53#ifndef __ASSEMBLY__54/* vIRQ injected to Dom0 */55#define VIRQ_MCA VIRQ_ARCH_056 57/*58 * mc_info entry types59 * mca machine check info are recorded in mc_info entries.60 * when fetch mca info, it can use MC_TYPE_... to distinguish61 * different mca info.62 */63#define MC_TYPE_GLOBAL 064#define MC_TYPE_BANK 165#define MC_TYPE_EXTENDED 266#define MC_TYPE_RECOVERY 367 68struct mcinfo_common {69 uint16_t type; /* structure type */70 uint16_t size; /* size of this struct in bytes */71};72 73#define MC_FLAG_CORRECTABLE (1 << 0)74#define MC_FLAG_UNCORRECTABLE (1 << 1)75#define MC_FLAG_RECOVERABLE (1 << 2)76#define MC_FLAG_POLLED (1 << 3)77#define MC_FLAG_RESET (1 << 4)78#define MC_FLAG_CMCI (1 << 5)79#define MC_FLAG_MCE (1 << 6)80 81/* contains x86 global mc information */82struct mcinfo_global {83 struct mcinfo_common common;84 85 uint16_t mc_domid; /* running domain at the time in error */86 uint16_t mc_vcpuid; /* virtual cpu scheduled for mc_domid */87 uint32_t mc_socketid; /* physical socket of the physical core */88 uint16_t mc_coreid; /* physical impacted core */89 uint16_t mc_core_threadid; /* core thread of physical core */90 uint32_t mc_apicid;91 uint32_t mc_flags;92 uint64_t mc_gstatus; /* global status */93};94 95/* contains x86 bank mc information */96struct mcinfo_bank {97 struct mcinfo_common common;98 99 uint16_t mc_bank; /* bank nr */100 uint16_t mc_domid; /* domain referenced by mc_addr if valid */101 uint64_t mc_status; /* bank status */102 uint64_t mc_addr; /* bank address */103 uint64_t mc_misc;104 uint64_t mc_ctrl2;105 uint64_t mc_tsc;106};107 108struct mcinfo_msr {109 uint64_t reg; /* MSR */110 uint64_t value; /* MSR value */111};112 113/* contains mc information from other or additional mc MSRs */114struct mcinfo_extended {115 struct mcinfo_common common;116 uint32_t mc_msrs; /* Number of msr with valid values. */117 /*118 * Currently Intel extended MSR (32/64) include all gp registers119 * and E(R)FLAGS, E(R)IP, E(R)MISC, up to 11/19 of them might be120 * useful at present. So expand this array to 16/32 to leave room.121 */122 struct mcinfo_msr mc_msr[sizeof(void *) * 4];123};124 125/* Recovery Action flags. Giving recovery result information to DOM0 */126 127/* Xen takes successful recovery action, the error is recovered */128#define REC_ACTION_RECOVERED (0x1 << 0)129/* No action is performed by XEN */130#define REC_ACTION_NONE (0x1 << 1)131/* It's possible DOM0 might take action ownership in some case */132#define REC_ACTION_NEED_RESET (0x1 << 2)133 134/*135 * Different Recovery Action types, if the action is performed successfully,136 * REC_ACTION_RECOVERED flag will be returned.137 */138 139/* Page Offline Action */140#define MC_ACTION_PAGE_OFFLINE (0x1 << 0)141/* CPU offline Action */142#define MC_ACTION_CPU_OFFLINE (0x1 << 1)143/* L3 cache disable Action */144#define MC_ACTION_CACHE_SHRINK (0x1 << 2)145 146/*147 * Below interface used between XEN/DOM0 for passing XEN's recovery action148 * information to DOM0.149 */150struct page_offline_action {151 /* Params for passing the offlined page number to DOM0 */152 uint64_t mfn;153 uint64_t status;154};155 156struct cpu_offline_action {157 /* Params for passing the identity of the offlined CPU to DOM0 */158 uint32_t mc_socketid;159 uint16_t mc_coreid;160 uint16_t mc_core_threadid;161};162 163#define MAX_UNION_SIZE 16164struct mcinfo_recovery {165 struct mcinfo_common common;166 uint16_t mc_bank; /* bank nr */167 uint8_t action_flags;168 uint8_t action_types;169 union {170 struct page_offline_action page_retire;171 struct cpu_offline_action cpu_offline;172 uint8_t pad[MAX_UNION_SIZE];173 } action_info;174};175 176 177#define MCINFO_MAXSIZE 768178struct mc_info {179 /* Number of mcinfo_* entries in mi_data */180 uint32_t mi_nentries;181 uint32_t flags;182 uint64_t mi_data[(MCINFO_MAXSIZE - 1) / 8];183};184DEFINE_GUEST_HANDLE_STRUCT(mc_info);185 186#define __MC_MSR_ARRAYSIZE 8187#define __MC_NMSRS 1188#define MC_NCAPS 7189struct mcinfo_logical_cpu {190 uint32_t mc_cpunr;191 uint32_t mc_chipid;192 uint16_t mc_coreid;193 uint16_t mc_threadid;194 uint32_t mc_apicid;195 uint32_t mc_clusterid;196 uint32_t mc_ncores;197 uint32_t mc_ncores_active;198 uint32_t mc_nthreads;199 uint32_t mc_cpuid_level;200 uint32_t mc_family;201 uint32_t mc_vendor;202 uint32_t mc_model;203 uint32_t mc_step;204 char mc_vendorid[16];205 char mc_brandid[64];206 uint32_t mc_cpu_caps[MC_NCAPS];207 uint32_t mc_cache_size;208 uint32_t mc_cache_alignment;209 uint32_t mc_nmsrvals;210 struct mcinfo_msr mc_msrvalues[__MC_MSR_ARRAYSIZE];211};212DEFINE_GUEST_HANDLE_STRUCT(mcinfo_logical_cpu);213 214/*215 * Prototype:216 * uint32_t x86_mcinfo_nentries(struct mc_info *mi);217 */218#define x86_mcinfo_nentries(_mi) \219 ((_mi)->mi_nentries)220/*221 * Prototype:222 * struct mcinfo_common *x86_mcinfo_first(struct mc_info *mi);223 */224#define x86_mcinfo_first(_mi) \225 ((struct mcinfo_common *)(_mi)->mi_data)226/*227 * Prototype:228 * struct mcinfo_common *x86_mcinfo_next(struct mcinfo_common *mic);229 */230#define x86_mcinfo_next(_mic) \231 ((struct mcinfo_common *)((uint8_t *)(_mic) + (_mic)->size))232 233/*234 * Prototype:235 * void x86_mcinfo_lookup(void *ret, struct mc_info *mi, uint16_t type);236 */237static inline void x86_mcinfo_lookup(struct mcinfo_common **ret,238 struct mc_info *mi, uint16_t type)239{240 uint32_t i;241 struct mcinfo_common *mic;242 bool found = 0;243 244 if (!ret || !mi)245 return;246 247 mic = x86_mcinfo_first(mi);248 for (i = 0; i < x86_mcinfo_nentries(mi); i++) {249 if (mic->type == type) {250 found = 1;251 break;252 }253 mic = x86_mcinfo_next(mic);254 }255 256 *ret = found ? mic : NULL;257}258 259/*260 * Fetch machine check data from hypervisor.261 */262#define XEN_MC_fetch 1263struct xen_mc_fetch {264 /*265 * IN: XEN_MC_NONURGENT, XEN_MC_URGENT,266 * XEN_MC_ACK if ack'king an earlier fetch267 * OUT: XEN_MC_OK, XEN_MC_FETCHAILED, XEN_MC_NODATA268 */269 uint32_t flags;270 uint32_t _pad0;271 /* OUT: id for ack, IN: id we are ack'ing */272 uint64_t fetch_id;273 274 /* OUT variables. */275 GUEST_HANDLE(mc_info) data;276};277DEFINE_GUEST_HANDLE_STRUCT(xen_mc_fetch);278 279 280/*281 * This tells the hypervisor to notify a DomU about the machine check error282 */283#define XEN_MC_notifydomain 2284struct xen_mc_notifydomain {285 /* IN variables */286 uint16_t mc_domid; /* The unprivileged domain to notify */287 uint16_t mc_vcpuid; /* The vcpu in mc_domid to notify */288 289 /* IN/OUT variables */290 uint32_t flags;291};292DEFINE_GUEST_HANDLE_STRUCT(xen_mc_notifydomain);293 294#define XEN_MC_physcpuinfo 3295struct xen_mc_physcpuinfo {296 /* IN/OUT */297 uint32_t ncpus;298 uint32_t _pad0;299 /* OUT */300 GUEST_HANDLE(mcinfo_logical_cpu) info;301};302 303#define XEN_MC_msrinject 4304#define MC_MSRINJ_MAXMSRS 8305struct xen_mc_msrinject {306 /* IN */307 uint32_t mcinj_cpunr; /* target processor id */308 uint32_t mcinj_flags; /* see MC_MSRINJ_F_* below */309 uint32_t mcinj_count; /* 0 .. count-1 in array are valid */310 uint32_t _pad0;311 struct mcinfo_msr mcinj_msr[MC_MSRINJ_MAXMSRS];312};313 314/* Flags for mcinj_flags above; bits 16-31 are reserved */315#define MC_MSRINJ_F_INTERPOSE 0x1316 317#define XEN_MC_mceinject 5318struct xen_mc_mceinject {319 unsigned int mceinj_cpunr; /* target processor id */320};321 322struct xen_mc {323 uint32_t cmd;324 uint32_t interface_version; /* XEN_MCA_INTERFACE_VERSION */325 union {326 struct xen_mc_fetch mc_fetch;327 struct xen_mc_notifydomain mc_notifydomain;328 struct xen_mc_physcpuinfo mc_physcpuinfo;329 struct xen_mc_msrinject mc_msrinject;330 struct xen_mc_mceinject mc_mceinject;331 } u;332};333DEFINE_GUEST_HANDLE_STRUCT(xen_mc);334 335/*336 * Fields are zero when not available. Also, this struct is shared with337 * userspace mcelog and thus must keep existing fields at current offsets.338 * Only add new fields to the end of the structure339 */340struct xen_mce {341 __u64 status;342 __u64 misc;343 __u64 addr;344 __u64 mcgstatus;345 __u64 ip;346 __u64 tsc; /* cpu time stamp counter */347 __u64 time; /* wall time_t when error was detected */348 __u8 cpuvendor; /* cpu vendor as encoded in system.h */349 __u8 inject_flags; /* software inject flags */350 __u16 pad;351 __u32 cpuid; /* CPUID 1 EAX */352 __u8 cs; /* code segment */353 __u8 bank; /* machine check bank */354 __u8 cpu; /* cpu number; obsolete; use extcpu now */355 __u8 finished; /* entry is valid */356 __u32 extcpu; /* linux cpu number that detected the error */357 __u32 socketid; /* CPU socket ID */358 __u32 apicid; /* CPU initial apic ID */359 __u64 mcgcap; /* MCGCAP MSR: machine check capabilities of CPU */360 __u64 synd; /* MCA_SYND MSR: only valid on SMCA systems */361 __u64 ipid; /* MCA_IPID MSR: only valid on SMCA systems */362 __u64 ppin; /* Protected Processor Inventory Number */363};364 365/*366 * This structure contains all data related to the MCE log. Also367 * carries a signature to make it easier to find from external368 * debugging tools. Each entry is only valid when its finished flag369 * is set.370 */371 372#define XEN_MCE_LOG_LEN 32373 374struct xen_mce_log {375 char signature[12]; /* "MACHINECHECK" */376 unsigned len; /* = XEN_MCE_LOG_LEN */377 unsigned next;378 unsigned flags;379 unsigned recordlen; /* length of struct xen_mce */380 struct xen_mce entry[XEN_MCE_LOG_LEN];381};382 383#define XEN_MCE_OVERFLOW 0 /* bit 0 in flags means overflow */384 385#define XEN_MCE_LOG_SIGNATURE "MACHINECHECK"386 387#define MCE_GET_RECORD_LEN _IOR('M', 1, int)388#define MCE_GET_LOG_LEN _IOR('M', 2, int)389#define MCE_GETCLEAR_FLAGS _IOR('M', 3, int)390 391#endif /* __ASSEMBLY__ */392#endif /* __XEN_PUBLIC_ARCH_X86_MCA_H__ */393