patch: NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM as fatal - Githubissues

AmpereComputing / ampere-lts-kernel---DEPRECATED

Linux 5.4 and 5.10 Longterm kernel (LTS) with Ampere patches

21 stars 18 forks source link

patch: NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM as fatal #103

Open JasonTian518 opened 3 years ago

JasonTian518 commented 3 years ago

patch: NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM as fatal gitlab commit id: amp-centos/-/commit/78f9fdd47972a7113dd1c50ef7baa5f08c362318

5.15 kernel: ACPI/AEST: Initial AEST driver trace, ras: add ARM RAS extension trace event NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM a… …s fatal
5.10 kernel: 4787e43e1e7b NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM as fatal
5.4 kernel: abe9b616563e NOUPSTREAM: arm64/ras: Treat single-bit ECC error in the MMU TC RAM as fatal

According to below comments in the code:

Workaround for ARM N1 errata where transient single-bit
error in the MMU TC RAM might cause corruption. Treat
these CEs as uncontainable UEs.
This is needed for lts code as this comes from ARM N1 errata We haven't test method as the EINJ error injection can't generate a transient single-bit ECC error(err_misc0&0x3 != 0x2)

adamliyi commented 3 years ago

NOUPSTREAM arm64ras Treat single-bit ECC error in the MMU TC RAM as fatal.patch.txt

Backport this patch to 5.4 and 5.10 kernel. Make sure kernel/ras.c can be compiled.