x86/mce: Use severity table to handle uncorrected errors in kernel
authorTony Luck <tony.luck@intel.com>
Thu, 22 Sep 2022 19:51:35 +0000 (12:51 -0700)
committerBorislav Petkov <bp@suse.de>
Mon, 31 Oct 2022 16:01:19 +0000 (17:01 +0100)
mce_severity_intel() has a special case to promote UC and AR errors
in kernel context to PANIC severity.

The "AR" case is already handled with separate entries in the severity
table for all instruction fetch errors, and those data fetch errors that
are not in a recoverable area of the kernel (i.e. have an extable fixup
entry).

Add an entry to the severity table for UC errors in kernel context that
reports severity = PANIC. Delete the special case code from
mce_severity_intel().

Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lore.kernel.org/r/20220922195136.54575-2-tony.luck@intel.com
arch/x86/kernel/cpu/mce/severity.c

index 00483d1..c447716 100644 (file)
@@ -203,6 +203,11 @@ static struct severity {
                BITSET(MCI_STATUS_OVER|MCI_STATUS_UC)
                ),
        MCESEV(
+               PANIC, "Uncorrected in kernel",
+               BITSET(MCI_STATUS_UC),
+               KERNEL
+               ),
+       MCESEV(
                UC, "Uncorrected",
                BITSET(MCI_STATUS_UC)
                ),
@@ -391,9 +396,6 @@ static noinstr int mce_severity_intel(struct mce *m, struct pt_regs *regs, char
                        *msg = s->msg;
                s->covered = 1;
 
-               if (s->sev >= MCE_UC_SEVERITY && ctx == IN_KERNEL)
-                       return MCE_PANIC_SEVERITY;
-
                return s->sev;
        }
 }