* [PATCH v2] platform/x86/amd/pmf: Fix ACPI buffer validation in APTS path
@ 2026-09-30 17:24 Shyam Sundar S K
2026-09-30 22:37 ` Mario Limonciello
0 siblings, 1 reply; 2+ messages in thread
From: Shyam Sundar S K @ 2026-09-30 17:24 UTC (permalink / raw)
To: hansg, ilpo.jarvinen
Cc: platform-driver-x86, Patil.Reddy, mario.limonciello,
Shyam Sundar S K
apts_if_call_store_buffer() reads a u16 size from info->buffer.pointer
without checking that it is non-NULL and at least 2 bytes long, so a
short or NULL firmware buffer can cause an out-of-bounds read.
apmf_if_call_store_buffer() checked the length but not the pointer.
Add a helper, amd_pmf_if_verify_buffer(), that rejects a NULL pointer or
a buffer smaller than the 2-byte header, and call it from both paths, so
the check is done consistently.
Fixes: 3eecb434d7f2 ("platform/x86/amd/pmf: Add support to get sps default APTS index values")
Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
---
v2:
- Add a helper function for NULL pointer and buffer validation. Reuse it in
apmf_if_call_store_buffer() and apts_if_call_store_buffer().
drivers/platform/x86/amd/pmf/acpi.c | 20 ++++++++++++++++----
1 file changed, 16 insertions(+), 4 deletions(-)
diff --git a/drivers/platform/x86/amd/pmf/acpi.c b/drivers/platform/x86/amd/pmf/acpi.c
index 3d94b03cf794..efad01a4ed27 100644
--- a/drivers/platform/x86/amd/pmf/acpi.c
+++ b/drivers/platform/x86/amd/pmf/acpi.c
@@ -50,6 +50,16 @@ static union acpi_object *apmf_if_call(struct amd_pmf_dev *pdev, int fn, struct
return buffer.pointer;
}
+static int amd_pmf_if_verify_buffer(struct amd_pmf_dev *pdev, union acpi_object *info)
+{
+ if (!info->buffer.pointer || info->buffer.length < 2) {
+ dev_err(pdev->dev, "buffer pointer is NULL or too small\n");
+ return -EINVAL;
+ }
+
+ return 0;
+}
+
static int apmf_if_call_store_buffer(struct amd_pmf_dev *pdev, int fn, void *dest, size_t out_sz)
{
union acpi_object *info;
@@ -66,11 +76,9 @@ static int apmf_if_call_store_buffer(struct amd_pmf_dev *pdev, int fn, void *des
goto out;
}
- if (info->buffer.length < 2) {
- dev_err(pdev->dev, "buffer too small\n");
- err = -EINVAL;
+ err = amd_pmf_if_verify_buffer(pdev, info);
+ if (err)
goto out;
- }
size = *(u16 *)info->buffer.pointer;
if (info->buffer.length < size) {
@@ -138,6 +146,10 @@ static int apts_if_call_store_buffer(struct amd_pmf_dev *pdev,
goto out;
}
+ err = amd_pmf_if_verify_buffer(pdev, info);
+ if (err)
+ goto out;
+
size = *(u16 *)info->buffer.pointer;
if (info->buffer.length < size) {
dev_err(pdev->dev, "buffer smaller than header size %u < %zu\n",
--
2.34.1
^ permalink raw reply related [flat|nested] 2+ messages in thread
* Re: [PATCH v2] platform/x86/amd/pmf: Fix ACPI buffer validation in APTS path
2026-09-30 17:24 [PATCH v2] platform/x86/amd/pmf: Fix ACPI buffer validation in APTS path Shyam Sundar S K
@ 2026-09-30 22:37 ` Mario Limonciello
0 siblings, 0 replies; 2+ messages in thread
From: Mario Limonciello @ 2026-09-30 22:37 UTC (permalink / raw)
To: Shyam Sundar S K, hansg, ilpo.jarvinen; +Cc: platform-driver-x86, Patil.Reddy
On 9/30/26 12:24, Shyam Sundar S K wrote:
> apts_if_call_store_buffer() reads a u16 size from info->buffer.pointer
> without checking that it is non-NULL and at least 2 bytes long, so a
> short or NULL firmware buffer can cause an out-of-bounds read.
> apmf_if_call_store_buffer() checked the length but not the pointer.
>
> Add a helper, amd_pmf_if_verify_buffer(), that rejects a NULL pointer or
> a buffer smaller than the 2-byte header, and call it from both paths, so
> the check is done consistently.
>
> Fixes: 3eecb434d7f2 ("platform/x86/amd/pmf: Add support to get sps default APTS index values")
> Signed-off-by: Shyam Sundar S K <Shyam-sundar.S-k@amd.com>
> ---
> v2:
> - Add a helper function for NULL pointer and buffer validation. Reuse it in
> apmf_if_call_store_buffer() and apts_if_call_store_buffer().
>
> drivers/platform/x86/amd/pmf/acpi.c | 20 ++++++++++++++++----
> 1 file changed, 16 insertions(+), 4 deletions(-)
>
> diff --git a/drivers/platform/x86/amd/pmf/acpi.c b/drivers/platform/x86/amd/pmf/acpi.c
> index 3d94b03cf794..efad01a4ed27 100644
> --- a/drivers/platform/x86/amd/pmf/acpi.c
> +++ b/drivers/platform/x86/amd/pmf/acpi.c
> @@ -50,6 +50,16 @@ static union acpi_object *apmf_if_call(struct amd_pmf_dev *pdev, int fn, struct
> return buffer.pointer;
> }
>
> +static int amd_pmf_if_verify_buffer(struct amd_pmf_dev *pdev, union acpi_object *info)
> +{
> + if (!info->buffer.pointer || info->buffer.length < 2) {
> + dev_err(pdev->dev, "buffer pointer is NULL or too small\n");
> + return -EINVAL;
> + }
> +
> + return 0;
> +}
> +
> static int apmf_if_call_store_buffer(struct amd_pmf_dev *pdev, int fn, void *dest, size_t out_sz)
> {
> union acpi_object *info;
> @@ -66,11 +76,9 @@ static int apmf_if_call_store_buffer(struct amd_pmf_dev *pdev, int fn, void *des
> goto out;
> }
>
> - if (info->buffer.length < 2) {
> - dev_err(pdev->dev, "buffer too small\n");
> - err = -EINVAL;
> + err = amd_pmf_if_verify_buffer(pdev, info);
> + if (err)
> goto out;
> - }
>
> size = *(u16 *)info->buffer.pointer;
> if (info->buffer.length < size) {
> @@ -138,6 +146,10 @@ static int apts_if_call_store_buffer(struct amd_pmf_dev *pdev,
> goto out;
> }
>
> + err = amd_pmf_if_verify_buffer(pdev, info);
> + if (err)
> + goto out;
> +
> size = *(u16 *)info->buffer.pointer;
> if (info->buffer.length < size) {
> dev_err(pdev->dev, "buffer smaller than header size %u < %zu\n",
If I'm not mistaken you can actually fold in more to the common check.
if (info->type != ACPI_TYPE_BUFFER) {
and buffer size check
if (info->buffer.length < size) {
and then
if (size < out_sz) {
Then you could just return the size on success from the helper both
apts_if_call_store_buffer calls
1) apts_if_call
2) helper
3) memcpy
4) kfree
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2026-09-30 22:37 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-09-30 17:24 [PATCH v2] platform/x86/amd/pmf: Fix ACPI buffer validation in APTS path Shyam Sundar S K
2026-09-30 22:37 ` Mario Limonciello
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.