When BIOS reports memory errors to Linux using the ACPI/APEI
error reporting method Linux creates a "struct mce" to pass
to the normal reporting code path.
The constructed record doesn't include a value for the "misc"
field of the structure, and so mce_usable_address() says this
record doesn't include a valid address.
Net result is that functions like uc_decode_notifier() will
just ignore this record instead of taking action to offline
a page.
Signed-off-by: Tony Luck <tony.luck@intel.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Link: https://lkml.kernel.org/r/20210527222846.931851-1-tony.luck@intel.com
mce_setup(&m);
m.bank = -1;
/* Fake a memory read error with unknown channel */
- m.status = MCI_STATUS_VAL | MCI_STATUS_EN | MCI_STATUS_ADDRV | 0x9f;
+ m.status = MCI_STATUS_VAL | MCI_STATUS_EN | MCI_STATUS_ADDRV | MCI_STATUS_MISCV | 0x9f;
+ m.misc = (MCI_MISC_ADDR_PHYS << 6) | PAGE_SHIFT;
if (severity >= GHES_SEV_RECOVERABLE)
m.status |= MCI_STATUS_UC;