Linux kernel mirror (for testing)
git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel
os
linux
Merge tag 'linus-mce-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras
Pull a machine check recovery fix from Tony Luck.
I really don't like how the MCE code does some of the things it does,
but this does seem to be an improvement.
* tag 'linus-mce-fix' of git://git.kernel.org/pub/scm/linux/kernel/git/ras/ras:
x86/mce: Only restart instruction after machine check recovery if it is safe
+11
-3
+11
-3
arch/x86/kernel/cpu/mcheck/mce.c
+11
-3
arch/x86/kernel/cpu/mcheck/mce.c
···
945
945
atomic_t inuse;
946
946
struct task_struct *t;
947
947
__u64 paddr;
948
+
int restartable;
948
949
} mce_info[MCE_INFO_MAX];
949
950
950
-
static void mce_save_info(__u64 addr)
951
+
static void mce_save_info(__u64 addr, int c)
951
952
{
952
953
struct mce_info *mi;
953
954
···
956
955
if (atomic_cmpxchg(&mi->inuse, 0, 1) == 0) {
957
956
mi->t = current;
958
957
mi->paddr = addr;
958
+
mi->restartable = c;
959
959
return;
960
960
}
961
961
}
···
1132
1130
mce_panic("Fatal machine check on current CPU", &m, msg);
1133
1131
if (worst == MCE_AR_SEVERITY) {
1134
1132
/* schedule action before return to userland */
1135
-
mce_save_info(m.addr);
1133
+
mce_save_info(m.addr, m.mcgstatus & MCG_STATUS_RIPV);
1136
1134
set_thread_flag(TIF_MCE_NOTIFY);
1137
1135
} else if (kill_it) {
1138
1136
force_sig(SIGBUS, current);
···
1181
1179
1182
1180
pr_err("Uncorrected hardware memory error in user-access at %llx",
1183
1181
mi->paddr);
1184
-
if (memory_failure(pfn, MCE_VECTOR, MF_ACTION_REQUIRED) < 0) {
1182
+
/*
1183
+
* We must call memory_failure() here even if the current process is
1184
+
* doomed. We still need to mark the page as poisoned and alert any
1185
+
* other users of the page.
1186
+
*/
1187
+
if (memory_failure(pfn, MCE_VECTOR, MF_ACTION_REQUIRED) < 0 ||
1188
+
mi->restartable == 0) {
1185
1189
pr_err("Memory error not recovered");
1186
1190
force_sig(SIGBUS, current);
1187
1191
}