[PATCH 1/3] x86/mce: Reset MCA_SYND between bank scans

[PATCH 1/3] x86/mce: Reset MCA_SYND between bank scans

[OptoCloud]

machine_check_poll() and __mc_scan_banks() both reuse a single
struct mce_hw_err while iterating over the MCA banks. The record is
zeroed once before the loop; each iteration then resets only MISC
and ADDR.

On SMCA, mce_read_aux() writes SYND only when MCI_STATUS_SYNDV is
set. Nothing clears it again, so a bank that does not set SYNDV
inherits the SYND value read for the previous bank in the same
scan, and that stale value is printed and handed to the decoders as
if it belonged to the current error.

Reset SYND at the start of each iteration in both loops, alongside
the existing MISC and ADDR resets.

Found by code inspection; not reproduced on hardware.

Fixes: db819d60f672 ("x86/mce: Add support for new MCA_SYND register")
Cc: stable@vger.kernel.org
Signed-off-by: Eirik Bøe <git@optocloud.no>
---
 arch/x86/kernel/cpu/mce/core.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/x86/kernel/cpu/mce/core.c b/arch/x86/kernel/cpu/mce/core.c
index ab469605fc89..16183fa4ddc7 100644
--- a/arch/x86/kernel/cpu/mce/core.c
+++ b/arch/x86/kernel/cpu/mce/core.c
@@ -808,6 +808,7 @@ void machine_check_poll(enum mcp_flags flags, mce_banks_t *b)
 
 		m->misc = 0;
 		m->addr = 0;
+		m->synd = 0;
 		m->bank = i;
 
 		barrier();
@@ -1347,6 +1348,7 @@ __mc_scan_banks(struct mce_hw_err *err, struct pt_regs *regs,
 
 		m->misc = 0;
 		m->addr = 0;
+		m->synd = 0;
 		m->bank = i;
 
 		m->status = mce_rdmsrq(mca_msr_reg(i, MCA_STATUS));
-- 
2.47.3

Re: [PATCH 1/3] x86/mce: Reset MCA_SYND between bank scans

[Yazen Ghannam]

On Sun, Aug 30, 2026 at 04:40:01AM +0000, OptoCloud wrote:
> machine_check_poll() and __mc_scan_banks() both reuse a single
> struct mce_hw_err while iterating over the MCA banks. The record is
> zeroed once before the loop; each iteration then resets only MISC
> and ADDR.
> 
> On SMCA, mce_read_aux() writes SYND only when MCI_STATUS_SYNDV is
> set. Nothing clears it again, so a bank that does not set SYNDV
> inherits the SYND value read for the previous bank in the same
> scan, and that stale value is printed and handed to the decoders as
> if it belonged to the current error.

True, but harmless in practice. If SYNDV is not set, then decoders
should not consider the value of MCA_SYND. It doesn't matter if the
value is '0' or garbage.

> 
> Reset SYND at the start of each iteration in both loops, alongside
> the existing MISC and ADDR resets.
> 
> Found by code inspection; not reproduced on hardware.
> 
> Fixes: db819d60f672 ("x86/mce: Add support for new MCA_SYND register")
> Cc: stable@vger.kernel.org

I don't agree that this needs to go to the stable branches, since it's
not fixing a practical bug.

> Signed-off-by: Eirik Bøe <git@optocloud.no>

Overall, I think this is a fair patch. It does bring the MCA_SYND
behavior inline with the other registers.

Besides the minor nits above this looks good to me.

Reviewed-by: Yazen Ghannam <yazen.ghannam@amd.com>

Thanks,
Yazen