2024-04-27 05:28:47

by linke li

[permalink] [raw]
Subject: [PATCH] x86/efi: mark racy access on efi_rts_work.efi_rts_id

In efi_crash_gracefully_on_page_fault(), efi_rts_work.efi_rts_id can by
changed by other thread from the comment. Mark possible data race on
efi_rts_work.efi_rts_id as benign using READ_ONCE.

This patch is aimed at reducing the number of benign races reported by
KCSAN in order to focus future debugging effort on harmful races.

Signed-off-by: linke li <[email protected]>
---
arch/x86/platform/efi/quirks.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/arch/x86/platform/efi/quirks.c b/arch/x86/platform/efi/quirks.c
index f0cc00032751..4acb81700caf 100644
--- a/arch/x86/platform/efi/quirks.c
+++ b/arch/x86/platform/efi/quirks.c
@@ -751,7 +751,7 @@ void efi_crash_gracefully_on_page_fault(unsigned long phys_addr)
* because this case occurs *very* rarely and hence could be improved
* on a need by basis.
*/
- if (efi_rts_work.efi_rts_id == EFI_RESET_SYSTEM) {
+ if (READ_ONCE(efi_rts_work.efi_rts_id) == EFI_RESET_SYSTEM) {
pr_info("efi_reset_system() buggy! Reboot through BIOS\n");
machine_real_restart(MRR_BIOS);
return;
--
2.39.3 (Apple Git-146)



2024-04-29 15:40:58

by Ard Biesheuvel

[permalink] [raw]
Subject: Re: [PATCH] x86/efi: mark racy access on efi_rts_work.efi_rts_id

On Sat, 27 Apr 2024 at 07:28, linke li <[email protected]> wrote:
>
> In efi_crash_gracefully_on_page_fault(), efi_rts_work.efi_rts_id can by
> changed by other thread from the comment. Mark possible data race on
> efi_rts_work.efi_rts_id as benign using READ_ONCE.
>
> This patch is aimed at reducing the number of benign races reported by
> KCSAN in order to focus future debugging effort on harmful races.
>
> Signed-off-by: linke li <[email protected]>
> ---
> arch/x86/platform/efi/quirks.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/arch/x86/platform/efi/quirks.c b/arch/x86/platform/efi/quirks.c
> index f0cc00032751..4acb81700caf 100644
> --- a/arch/x86/platform/efi/quirks.c
> +++ b/arch/x86/platform/efi/quirks.c
> @@ -751,7 +751,7 @@ void efi_crash_gracefully_on_page_fault(unsigned long phys_addr)
> * because this case occurs *very* rarely and hence could be improved
> * on a need by basis.
> */
> - if (efi_rts_work.efi_rts_id == EFI_RESET_SYSTEM) {
> + if (READ_ONCE(efi_rts_work.efi_rts_id) == EFI_RESET_SYSTEM) {
> pr_info("efi_reset_system() buggy! Reboot through BIOS\n");
> machine_real_restart(MRR_BIOS);
> return;

Why is this the only reference that needs an annotation?

2024-05-01 03:00:02

by linke li

[permalink] [raw]
Subject: Re: [PATCH] x86/efi: mark racy access on efi_rts_work.efi_rts_id

> Why is this the only reference that needs an annotation?

In the function efi_runtime_services_init(), the if check on
efi_rts_work.efi_rts_id in line 728 uses READ_ONCE(), and the comment
explains that this is to reduce reports from KTSAN.

I propose that the check in line 754 undergoes a similar treatment for
consistency and to potentially reduce sanitizer reports.

Thank you.