Re: [PATCH v3 6/6] drm/xe/xe_ras: Track offlined pages by firmware to avoid duplicates

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] xe: failed page offline requests are erroneously cached as successful
- [Medium] xe: page offline deduplication fails during initialization due to wrong order
--

commit 9f25f9a9144c678c3a1d0b655504ad65d4f2849b
Author: Riana Tauro <riana.tauro@intel.com>

drm/xe/xe_ras: Track offlined pages by firmware to avoid duplicates

A memory scrubber can report multiple errors at the same address. Track
pages already successfully offlined by the firmware so that subsequent
reports for the same address are removed from the firmware queue instead
of issuing redundant page offline requests.

> diff --git a/drivers/gpu/drm/xe/xe_ras.c b/drivers/gpu/drm/xe/xe_ras.c
> index 749cf25e9148a..d2bcbecd1ba66 100644
> --- a/drivers/gpu/drm/xe/xe_ras.c
> +++ b/drivers/gpu/drm/xe/xe_ras.c
> @@ -249,11 +253,16 @@ static int send_page_offline_cmd(struct xe_device *xe, u64 page_address,
>  		xe_log_err(xe, SYSCTRL, ret, "page offline command failed with status %u\n",
>  			   response.status);
>  
> +	if (action == XE_RAS_PAGE_ACTION_OFFLINE)
> +		xa_store(&state->offlined_pages, page_address >> XE_PTE_SHIFT,
> +			 XE_RAS_PAGE_OFFLINED, GFP_KERNEL);
> +
>  	return ret;
>  }

[Severity: High]
If ras_status_to_errno(response.status) returns an error, the code logs the
failure but execution continues. Does this cause the failed page offline
request to be erroneously cached as successful in state->offlined_pages?

If so, future reports of the same error will trigger a REMOVE action instead
of attempting to OFFLINE again, permanently dropping hardware memory errors
that failed to be offlined.

[ ... ]

> @@ -1145,10 +1177,14 @@ void xe_ras_init(struct xe_device *xe)
>  	if (IS_ENABLED(CONFIG_PCIEAER))
>  		ras_usp_aer_init(xe);
>  
> +	xa_init(&state->offlined_pages);
> +
>  	get_queued_pages(xe);
>  	get_offlined_list(xe);

[Severity: Medium]
Will page offline deduplication fail for queued pages because of the
initialization order?

During get_queued_pages(), pending errors are processed via
handle_page_offline(), which checks the xarray to detect duplicates:

handle_page_offline()
    if (xa_load(&state->offlined_pages, ...))
        action = XE_RAS_PAGE_ACTION_REMOVE;

Since get_offlined_list() has not yet populated the xarray with the existing
offlined pages, won't the xarray be empty at this point?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260928061830.417395-8-riana.tauro@intel.com?part=6