[1/2] vfio: Fix overrun after readlink() fills buffer completely

readlink() returns the number of bytes written to the buffer, and it
doesn't write a terminating null byte.  vfio_init() writes it itself.
Overruns the buffer when readlink() filled it completely.

Fix by reserving space for the null byte when calling readlink(), like
we do elsewhere.

Spotted by Coverity.

Signed-off-by: Markus Armbruster <armbru@redhat.com>
---
 hw/misc/vfio.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

On 21 February 2014 16:42, Markus Armbruster <armbru@redhat.com> wrote:
> readlink() returns the number of bytes written to the buffer, and it
> doesn't write a terminating null byte.  vfio_init() writes it itself.
> Overruns the buffer when readlink() filled it completely.
>
> Fix by reserving space for the null byte when calling readlink(), like
> we do elsewhere.
>
> Spotted by Coverity.
>
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> ---
>  hw/misc/vfio.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/hw/misc/vfio.c b/hw/misc/vfio.c
> index 8db182f..8e56785 100644
> --- a/hw/misc/vfio.c
> +++ b/hw/misc/vfio.c
> @@ -3681,7 +3681,7 @@ static int vfio_initfn(PCIDevice *pdev)
>
>      strncat(path, "iommu_group", sizeof(path) - strlen(path) - 1);
>
> -    len = readlink(path, iommu_group_path, PATH_MAX);
> +    len = readlink(path, iommu_group_path, PATH_MAX - 1);
>      if (len <= 0) {
>          error_report("vfio: error no iommu_group for device");
>          return -errno;

"sizeof(iommu_group_path) - 1" would be slightly better,
I think, but PATH_MAX - 1 works too.

Reviewed-by: Peter Maydell <peter.maydell@linaro.org>

(cc stable?)

thanks
-- PMM

On Fri, 2014-02-21 at 17:42 +0100, Markus Armbruster wrote:
> readlink() returns the number of bytes written to the buffer, and it
> doesn't write a terminating null byte.  vfio_init() writes it itself.
> Overruns the buffer when readlink() filled it completely.
> 
> Fix by reserving space for the null byte when calling readlink(), like
> we do elsewhere.
> 
> Spotted by Coverity.
> 
> Signed-off-by: Markus Armbruster <armbru@redhat.com>
> ---
>  hw/misc/vfio.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/hw/misc/vfio.c b/hw/misc/vfio.c
> index 8db182f..8e56785 100644
> --- a/hw/misc/vfio.c
> +++ b/hw/misc/vfio.c
> @@ -3681,7 +3681,7 @@ static int vfio_initfn(PCIDevice *pdev)
>  
>      strncat(path, "iommu_group", sizeof(path) - strlen(path) - 1);
>  
> -    len = readlink(path, iommu_group_path, PATH_MAX);
> +    len = readlink(path, iommu_group_path, PATH_MAX - 1);
>      if (len <= 0) {
>          error_report("vfio: error no iommu_group for device");
>          return -errno;

I'm not sure why we wouldn't follow the same logic as pci-assign here.
If we fill to the length provided we have no idea if we have the full
path.  We certainly expect it to fit well within PATH_MAX, so let's
handle it as an error if we reach PATH_MAX.  Thanks,

Alex

Alex Williamson <alex.williamson@redhat.com> writes:

> On Fri, 2014-02-21 at 17:42 +0100, Markus Armbruster wrote:
>> readlink() returns the number of bytes written to the buffer, and it
>> doesn't write a terminating null byte.  vfio_init() writes it itself.
>> Overruns the buffer when readlink() filled it completely.
>> 
>> Fix by reserving space for the null byte when calling readlink(), like
>> we do elsewhere.
>> 
>> Spotted by Coverity.
>> 
>> Signed-off-by: Markus Armbruster <armbru@redhat.com>
>> ---
>>  hw/misc/vfio.c | 2 +-
>>  1 file changed, 1 insertion(+), 1 deletion(-)
>> 
>> diff --git a/hw/misc/vfio.c b/hw/misc/vfio.c
>> index 8db182f..8e56785 100644
>> --- a/hw/misc/vfio.c
>> +++ b/hw/misc/vfio.c
>> @@ -3681,7 +3681,7 @@ static int vfio_initfn(PCIDevice *pdev)
>>  
>>      strncat(path, "iommu_group", sizeof(path) - strlen(path) - 1);
>>  
>> -    len = readlink(path, iommu_group_path, PATH_MAX);
>> +    len = readlink(path, iommu_group_path, PATH_MAX - 1);
>>      if (len <= 0) {
>>          error_report("vfio: error no iommu_group for device");
>>          return -errno;
>
> I'm not sure why we wouldn't follow the same logic as pci-assign here.
> If we fill to the length provided we have no idea if we have the full
> path.  We certainly expect it to fit well within PATH_MAX, so let's
> handle it as an error if we reach PATH_MAX.  Thanks,

Fair point.  v2 sent.