Re: [PATCH v4 06/10] ACPI: APEI: GHES: move CXL CPER helpers

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 3 potential issue(s) to consider:
- [High] Deadlock risk due to missing interrupt disable in registration functions.
- [High] Out-of-bounds read in `cxl_cper_post_event()` due to unconditional maximum-size `memcpy`.
- [High] Out-of-bounds read in `cxl_cper_setup_prot_err_work_data()` due to unvalidated `dvsec_len`.
--

commit b9b10da731ac810e776dc0f6489817a12018762f
Author: Ahmed Tiba <ahmed.tiba@arm.com>

ACPI: APEI: GHES: move CXL CPER helpers

Move the CXL CPER handling paths out of ghes.c and into ghes_cper.c so the
helpers can be reused. The code is moved as-is, with the public
prototypes updated so GHES keeps calling into the new translation unit.

> diff --git a/drivers/acpi/apei/ghes_cper.c b/drivers/acpi/apei/ghes_cper.c
> index 131980d360641..d7a666a163c3a 100644
> --- a/drivers/acpi/apei/ghes_cper.c
> +++ b/drivers/acpi/apei/ghes_cper.c
[ ... ]
> @@ -321,6 +323,138 @@ void ghes_defer_non_standard_event(struct acpi_hest_generic_data *gdata,
> +void cxl_cper_post_prot_err(struct cxl_cper_sec_prot_err *prot_err,
> +			    int severity)
> +{
> +#ifdef CONFIG_ACPI_APEI_PCIEAER
> +	struct cxl_cper_prot_err_work_data wd;
> +
> +	if (cxl_cper_sec_prot_err_valid(prot_err))
> +		return;
> +
> +	guard(spinlock_irqsave)(&cxl_cper_prot_err_work_lock);
> +
> +	if (!cxl_cper_prot_err_work)
> +		return;
> +
> +	if (cxl_cper_setup_prot_err_work_data(&wd, prot_err, severity))
> +		return;

This isn't a new bug introduced by this patch, but could this call lead to an
out-of-bounds read?

Looking at cxl_cper_setup_prot_err_work_data() in ghes_helpers.c:

int cxl_cper_setup_prot_err_work_data(...)
{
    ...
    dvsec_start = (u8 *)(prot_err + 1);
    cap_start = dvsec_start + prot_err->dvsec_len;
    
    memcpy(&wd->ras_cap, cap_start, sizeof(wd->ras_cap));
    ...
}

The prot_err->dvsec_len field from the firmware doesn't appear to be
validated against the actual payload bounds before being used to calculate
cap_start. If a buggy firmware provides an artificially large dvsec_len,
could the subsequent memcpy() read adjacent kernel memory?

[ ... ]
> +int cxl_cper_register_prot_err_work(struct work_struct *work)
> +{
> +	if (cxl_cper_prot_err_work)
> +		return -EINVAL;
> +
> +	guard(spinlock)(&cxl_cper_prot_err_work_lock);

This isn't a new bug introduced by this patch, but could this lock usage
cause a deadlock?

The event posting functions like cxl_cper_post_prot_err() are called in
hardware IRQ contexts and correctly use guard(spinlock_irqsave).

If an interrupt occurs on the same CPU while this registration function holds
the spinlock in process context, and the IRQ handler calls
cxl_cper_post_prot_err(), won't it attempt to acquire the already-held lock
and spin indefinitely?

Should these registration functions use guard(spinlock_irqsave) instead?

[ ... ]
> +void cxl_cper_post_event(enum cxl_event_type event_type,
> +			 struct cxl_cper_event_rec *rec)
> +{
> +	struct cxl_cper_work_data wd;
> +
> +	if (rec->hdr.length <= sizeof(rec->hdr) ||
> +	    rec->hdr.length > sizeof(*rec)) {
> +		pr_err(FW_WARN "CXL CPER Invalid section length (%u)\n",
> +		       rec->hdr.length);
> +		return;
> +	}
[ ... ]
> +	wd.event_type = event_type;
> +	memcpy(&wd.rec, rec, sizeof(wd.rec));

This isn't a new bug introduced by this patch, but could this memcpy()
over-read the rec buffer?

The length validation above allows rec->hdr.length to be smaller than
sizeof(*rec), but the memcpy unconditionally copies sizeof(wd.rec) bytes.

If the firmware provides a dynamically sized payload smaller than the maximum
structure size, will this read out-of-bounds kernel memory into the work
queue data?

[ ... ]
> +int cxl_cper_register_work(struct work_struct *work)
> +{
> +	if (cxl_cper_work)
> +		return -EINVAL;
> +
> +	guard(spinlock)(&cxl_cper_work_lock);

This isn't a new bug introduced by this patch, but similarly to the protocol
error registration, could this lock usage cause a deadlock if interrupted
by cxl_cper_post_event() running in an IRQ handler?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260518-topics-ahmtib01-ras_ffh_arm_internal_review-v4-0-42698675ba61@arm.com?part=6