[PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Bjorn Andersson]

Firmware that doesn't provide section headers leave both e_shentsize and
e_shnum 0, which obvious isn't compatible with the newly introduced
stricter checks.

Make the section-related checks conditional on either of these values
being non-zero.

Fixes: 9f9967fed9d0 ("soc: qcom: mdt_loader: Ensure we don't read past the ELF header")
Reported-by: Val Packett <val@packett.cool>
Closes: https://lore.kernel.org/all/ece307c3-7d65-440f-babd-88cf9705b908@packett.cool/
Reported-by: Neil Armstrong <neil.armstrong@linaro.org>
Closes: https://lore.kernel.org/all/aec9cd03-6fc2-4dc8-b937-8b7cf7bf4128@linaro.org/
Signed-off-by: Bjorn Andersson <bjorn.andersson@oss.qualcomm.com>
---
 drivers/soc/qcom/mdt_loader.c | 12 +++++++-----
 1 file changed, 7 insertions(+), 5 deletions(-)

diff --git a/drivers/soc/qcom/mdt_loader.c b/drivers/soc/qcom/mdt_loader.c
index 0ca268bdf1f8c579b36a04fd94417be172fc1099..5710ac0c07a8c8a56110e3fc5e30b8610e106449 100644
--- a/drivers/soc/qcom/mdt_loader.c
+++ b/drivers/soc/qcom/mdt_loader.c
@@ -39,12 +39,14 @@ static bool mdt_header_valid(const struct firmware *fw)
 	if (phend > fw->size)
 		return false;
 
-	if (ehdr->e_shentsize != sizeof(struct elf32_shdr))
-		return false;
+	if (ehdr->e_shentsize || ehdr->e_shnum) {
+		if (ehdr->e_shentsize != sizeof(struct elf32_shdr))
+			return false;
 
-	shend = size_add(size_mul(sizeof(struct elf32_shdr), ehdr->e_shnum), ehdr->e_shoff);
-	if (shend > fw->size)
-		return false;
+		shend = size_add(size_mul(sizeof(struct elf32_shdr), ehdr->e_shnum), ehdr->e_shoff);
+		if (shend > fw->size)
+			return false;
+	}
 
 	return true;
 }

---
base-commit: 79fb37f39b77bbf9a56304e9af843cd93a7a1916
change-id: 20250730-mdt-loader-shentsize-zero-8c99653b6343

Best regards,
-- 
Bjorn Andersson <bjorn.andersson@oss.qualcomm.com>

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Dmitry Baryshkov]

On Wed, Jul 30, 2025 at 03:51:51PM -0500, Bjorn Andersson wrote:
> Firmware that doesn't provide section headers leave both e_shentsize and
> e_shnum 0, which obvious isn't compatible with the newly introduced
> stricter checks.
> 
> Make the section-related checks conditional on either of these values
> being non-zero.
> 
> Fixes: 9f9967fed9d0 ("soc: qcom: mdt_loader: Ensure we don't read past the ELF header")
> Reported-by: Val Packett <val@packett.cool>
> Closes: https://lore.kernel.org/all/ece307c3-7d65-440f-babd-88cf9705b908@packett.cool/
> Reported-by: Neil Armstrong <neil.armstrong@linaro.org>
> Closes: https://lore.kernel.org/all/aec9cd03-6fc2-4dc8-b937-8b7cf7bf4128@linaro.org/
> Signed-off-by: Bjorn Andersson <bjorn.andersson@oss.qualcomm.com>
> ---
>  drivers/soc/qcom/mdt_loader.c | 12 +++++++-----
>  1 file changed, 7 insertions(+), 5 deletions(-)
> 

Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@oss.qualcomm.com>


-- 
With best wishes
Dmitry

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Neil Armstrong]

On 30/07/2025 22:51, Bjorn Andersson wrote:
> Firmware that doesn't provide section headers leave both e_shentsize and
> e_shnum 0, which obvious isn't compatible with the newly introduced
> stricter checks.
> 
> Make the section-related checks conditional on either of these values
> being non-zero.
> 
Missing:
Fixes: 9f35ab0e53cc ("soc: qcom: mdt_loader: Fix error return values in mdt_header_valid()")

Or it won't apply on stable kernels

> Fixes: 9f9967fed9d0 ("soc: qcom: mdt_loader: Ensure we don't read past the ELF header")
> Reported-by: Val Packett <val@packett.cool>
> Closes: https://lore.kernel.org/all/ece307c3-7d65-440f-babd-88cf9705b908@packett.cool/
> Reported-by: Neil Armstrong <neil.armstrong@linaro.org>
> Closes: https://lore.kernel.org/all/aec9cd03-6fc2-4dc8-b937-8b7cf7bf4128@linaro.org/
> Signed-off-by: Bjorn Andersson <bjorn.andersson@oss.qualcomm.com>
> ---
>   drivers/soc/qcom/mdt_loader.c | 12 +++++++-----
>   1 file changed, 7 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/soc/qcom/mdt_loader.c b/drivers/soc/qcom/mdt_loader.c
> index 0ca268bdf1f8c579b36a04fd94417be172fc1099..5710ac0c07a8c8a56110e3fc5e30b8610e106449 100644
> --- a/drivers/soc/qcom/mdt_loader.c
> +++ b/drivers/soc/qcom/mdt_loader.c
> @@ -39,12 +39,14 @@ static bool mdt_header_valid(const struct firmware *fw)
>   	if (phend > fw->size)
>   		return false;
>   
> -	if (ehdr->e_shentsize != sizeof(struct elf32_shdr))
> -		return false;
> +	if (ehdr->e_shentsize || ehdr->e_shnum) {
> +		if (ehdr->e_shentsize != sizeof(struct elf32_shdr))
> +			return false;
>   
> -	shend = size_add(size_mul(sizeof(struct elf32_shdr), ehdr->e_shnum), ehdr->e_shoff);
> -	if (shend > fw->size)
> -		return false;
> +		shend = size_add(size_mul(sizeof(struct elf32_shdr), ehdr->e_shnum), ehdr->e_shoff);
> +		if (shend > fw->size)
> +			return false;
> +	}
>   
>   	return true;
>   }
> 
> ---
> base-commit: 79fb37f39b77bbf9a56304e9af843cd93a7a1916
> change-id: 20250730-mdt-loader-shentsize-zero-8c99653b6343
> 
> Best regards,

Tested-by: Neil Armstrong <neil.armstrong@linaro.org> # on SM8650-QRD

Thanks,
Neil

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Dmitry Baryshkov]

On Thu, Jul 31, 2025 at 04:58:09PM +0200, Neil Armstrong wrote:
> On 30/07/2025 22:51, Bjorn Andersson wrote:
> > Firmware that doesn't provide section headers leave both e_shentsize and
> > e_shnum 0, which obvious isn't compatible with the newly introduced
> > stricter checks.
> > 
> > Make the section-related checks conditional on either of these values
> > being non-zero.
> > 
> Missing:
> Fixes: 9f35ab0e53cc ("soc: qcom: mdt_loader: Fix error return values in mdt_header_valid()")
> 
> Or it won't apply on stable kernels

Then it should also be:

Cc: stable@vger.kernel.org


-- 
With best wishes
Dmitry

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Bjorn Andersson]

On Thu, Jul 31, 2025 at 11:23:25PM +0300, Dmitry Baryshkov wrote:
> On Thu, Jul 31, 2025 at 04:58:09PM +0200, Neil Armstrong wrote:
> > On 30/07/2025 22:51, Bjorn Andersson wrote:
> > > Firmware that doesn't provide section headers leave both e_shentsize and
> > > e_shnum 0, which obvious isn't compatible with the newly introduced
> > > stricter checks.
> > > 
> > > Make the section-related checks conditional on either of these values
> > > being non-zero.
> > > 
> > Missing:
> > Fixes: 9f35ab0e53cc ("soc: qcom: mdt_loader: Fix error return values in mdt_header_valid()")
> > 
> > Or it won't apply on stable kernels
> 
> Then it should also be:
> 
> Cc: stable@vger.kernel.org
> 

Given that the fixed commit is Cc: stable, it makes sense to include
both.

Thanks,
Bjorn

> 
> -- 
> With best wishes
> Dmitry

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Neil Armstrong]

On 30/07/2025 22:51, Bjorn Andersson wrote:
> Firmware that doesn't provide section headers leave both e_shentsize and
> e_shnum 0, which obvious isn't compatible with the newly introduced
> stricter checks.
> 
> Make the section-related checks conditional on either of these values
> being non-zero.
> 
> Fixes: 9f9967fed9d0 ("soc: qcom: mdt_loader: Ensure we don't read past the ELF header")
> Reported-by: Val Packett <val@packett.cool>
> Closes: https://lore.kernel.org/all/ece307c3-7d65-440f-babd-88cf9705b908@packett.cool/
> Reported-by: Neil Armstrong <neil.armstrong@linaro.org>
> Closes: https://lore.kernel.org/all/aec9cd03-6fc2-4dc8-b937-8b7cf7bf4128@linaro.org/
> Signed-off-by: Bjorn Andersson <bjorn.andersson@oss.qualcomm.com>
> ---
>   drivers/soc/qcom/mdt_loader.c | 12 +++++++-----
>   1 file changed, 7 insertions(+), 5 deletions(-)
> 

The issue solved by this fix is still present on v6.17-rc1, even if we reported
a breakage over -next on the 22th of July.

Neil

Re: [PATCH] soc: qcom: mdt_loader: Deal with zero e_shentsize

[Bjorn Andersson]

On Wed, 30 Jul 2025 15:51:51 -0500, Bjorn Andersson wrote:
> Firmware that doesn't provide section headers leave both e_shentsize and
> e_shnum 0, which obvious isn't compatible with the newly introduced
> stricter checks.
> 
> Make the section-related checks conditional on either of these values
> being non-zero.
> 
> [...]

Applied, thanks!

[1/1] soc: qcom: mdt_loader: Deal with zero e_shentsize
      commit: 25daf9af0ac1bf12490b723b5efaf8dcc85980bc

Best regards,
-- 
Bjorn Andersson <andersson@kernel.org>