[PATCH v2 1/1] drm/mediatek: Fix potential KP on 0 bytes nvmem cell read

[PATCH v2 1/1] drm/mediatek: Fix potential KP on 0 bytes nvmem cell read

[Liankun Yang]

If the len is 0, kernel crash will occur when performing operations.
Add the len check conditions to prevent kernel crash.

Fixes: f70ac097a2cf ("drm/mediatek: Add MT8195 Embedded DisplayPort driver")
Signed-off-by: Liankun Yang <liankun.yang@mediatek.com>
---
Changes in V2:
- Fix the commit title.
- Remove blank line between the Fixes and Signe-off-by.
- Modify the judgment writing method.
Per suggestion from the previous thread:
https://patchwork.kernel.org/project/linux-mediatek/patch/20240925084116.28848-1-liankun.yang@mediatek.com/
---
 drivers/gpu/drm/mediatek/mtk_dp.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/gpu/drm/mediatek/mtk_dp.c b/drivers/gpu/drm/mediatek/mtk_dp.c
index d8796a904eca..9ecdf62398ee 100644
--- a/drivers/gpu/drm/mediatek/mtk_dp.c
+++ b/drivers/gpu/drm/mediatek/mtk_dp.c
@@ -1082,7 +1082,7 @@ static void mtk_dp_get_calibration_data(struct mtk_dp *mtk_dp)
 	buf = (u32 *)nvmem_cell_read(cell, &len);
 	nvmem_cell_put(cell);
 
-	if (IS_ERR(buf) || ((len / sizeof(u32)) != 4)) {
+	if (IS_ERR(buf) || !len || ((len / sizeof(u32)) != 4)) {
 		dev_warn(dev, "Failed to read nvmem_cell_read\n");
 
 		if (!IS_ERR(buf))
-- 
2.45.2

Re: [PATCH v2 1/1] drm/mediatek: Fix potential KP on 0 bytes nvmem cell read

[Guillaume Ranquet]

On Fri, 27 Sep 2024 11:03, Liankun Yang <liankun.yang@mediatek.com> wrote:
>If the len is 0, kernel crash will occur when performing operations.
>Add the len check conditions to prevent kernel crash.
>
>Fixes: f70ac097a2cf ("drm/mediatek: Add MT8195 Embedded DisplayPort driver")
>Signed-off-by: Liankun Yang <liankun.yang@mediatek.com>
>---
>Changes in V2:
>- Fix the commit title.
>- Remove blank line between the Fixes and Signe-off-by.
>- Modify the judgment writing method.
>Per suggestion from the previous thread:
>https://patchwork.kernel.org/project/linux-mediatek/patch/20240925084116.28848-1-liankun.yang@mediatek.com/
>---
> drivers/gpu/drm/mediatek/mtk_dp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>diff --git a/drivers/gpu/drm/mediatek/mtk_dp.c b/drivers/gpu/drm/mediatek/mtk_dp.c
>index d8796a904eca..9ecdf62398ee 100644
>--- a/drivers/gpu/drm/mediatek/mtk_dp.c
>+++ b/drivers/gpu/drm/mediatek/mtk_dp.c
>@@ -1082,7 +1082,7 @@ static void mtk_dp_get_calibration_data(struct mtk_dp *mtk_dp)
> 	buf = (u32 *)nvmem_cell_read(cell, &len);
> 	nvmem_cell_put(cell);
>
>-	if (IS_ERR(buf) || ((len / sizeof(u32)) != 4)) {
>+	if (IS_ERR(buf) || !len || ((len / sizeof(u32)) != 4)) {
> 		dev_warn(dev, "Failed to read nvmem_cell_read\n");

Hello Liankun,
Would you be able to describe the crash with more details?

I'm afraid I don't understand the fix?

if len is 0, dividing 0 by sizeof(u32) is 0 and thus != 4
So to me, checking for len != 0 is redundant?

Thx,
Guillaume.
>
> 		if (!IS_ERR(buf))
>--
>2.45.2

Re: [PATCH v2 1/1] drm/mediatek: Fix potential KP on 0 bytes nvmem cell read

[CK Hu (胡俊光)]

Hi, Liankun:

On Fri, 2024-09-27 at 17:03 +0800, Liankun Yang wrote:
> If the len is 0, kernel crash will occur when performing operations.
> Add the len check conditions to prevent kernel crash.
> 
> Fixes: f70ac097a2cf ("drm/mediatek: Add MT8195 Embedded DisplayPort driver")
> Signed-off-by: Liankun Yang <liankun.yang@mediatek.com>
> ---
> Changes in V2:
> - Fix the commit title.
> - Remove blank line between the Fixes and Signe-off-by.
> - Modify the judgment writing method.
> Per suggestion from the previous thread:
> https://patchwork.kernel.org/project/linux-mediatek/patch/20240925084116.28848-1-liankun.yang@mediatek.com/
> ---
>  drivers/gpu/drm/mediatek/mtk_dp.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/gpu/drm/mediatek/mtk_dp.c b/drivers/gpu/drm/mediatek/mtk_dp.c
> index d8796a904eca..9ecdf62398ee 100644
> --- a/drivers/gpu/drm/mediatek/mtk_dp.c
> +++ b/drivers/gpu/drm/mediatek/mtk_dp.c
> @@ -1082,7 +1082,7 @@ static void mtk_dp_get_calibration_data(struct mtk_dp *mtk_dp)
>  	buf = (u32 *)nvmem_cell_read(cell, &len);
>  	nvmem_cell_put(cell);
>  
> -	if (IS_ERR(buf) || ((len / sizeof(u32)) != 4)) {

In original code, when len = 0, no kernel panic occur.
Nacked by me.

Regards,
CK


> +	if (IS_ERR(buf) || !len || ((len / sizeof(u32)) != 4)) {
>  		dev_warn(dev, "Failed to read nvmem_cell_read\n");
>  
>  		if (!IS_ERR(buf))

Re: [PATCH v2 1/1] drm/mediatek: Fix potential KP on 0 bytes nvmem cell read

[LIANKUN YANG (杨连坤)]

On Wed, 2024-10-09 at 08:41 -0700, Guillaume Ranquet wrote:
>  	 
> External email : Please do not click links or open attachments until
> you have verified the sender or the content.
>  On Fri, 27 Sep 2024 11:03, Liankun Yang <liankun.yang@mediatek.com>
> wrote:
> >If the len is 0, kernel crash will occur when performing operations.
> >Add the len check conditions to prevent kernel crash.
> >
> >Fixes: f70ac097a2cf ("drm/mediatek: Add MT8195 Embedded DisplayPort
> driver")
> >Signed-off-by: Liankun Yang <liankun.yang@mediatek.com>
> >---
> >Changes in V2:
> >- Fix the commit title.
> >- Remove blank line between the Fixes and Signe-off-by.
> >- Modify the judgment writing method.
> >Per suggestion from the previous thread:
> >
> https://patchwork.kernel.org/project/linux-mediatek/patch/20240925084116.28848-1-liankun.yang@mediatek.com/
> >---
> > drivers/gpu/drm/mediatek/mtk_dp.c | 2 +-
> > 1 file changed, 1 insertion(+), 1 deletion(-)
> >
> >diff --git a/drivers/gpu/drm/mediatek/mtk_dp.c
> b/drivers/gpu/drm/mediatek/mtk_dp.c
> >index d8796a904eca..9ecdf62398ee 100644
> >--- a/drivers/gpu/drm/mediatek/mtk_dp.c
> >+++ b/drivers/gpu/drm/mediatek/mtk_dp.c
> >@@ -1082,7 +1082,7 @@ static void mtk_dp_get_calibration_data(struct
> mtk_dp *mtk_dp)
> > buf = (u32 *)nvmem_cell_read(cell, &len);
> > nvmem_cell_put(cell);
> >
> >-if (IS_ERR(buf) || ((len / sizeof(u32)) != 4)) {
> >+if (IS_ERR(buf) || !len || ((len / sizeof(u32)) != 4)) {
> > dev_warn(dev, "Failed to read nvmem_cell_read\n");
> 
> Hello Liankun,
> Would you be able to describe the crash with more details?
> 
> I'm afraid I don't understand the fix?
> 
> if len is 0, dividing 0 by sizeof(u32) is 0 and thus != 4
> So to me, checking for len != 0 is redundant?
> 
> Thx,
> Guillaume.

Hello Guillaume,

Thank you for your correction.

The check for len != 0 is redundant.

I did not review this patch carefully, and will review it again next
time.

Thx,
Liankun.

> >
> > if (!IS_ERR(buf))
> >--
> >2.45.2