Re: [PATCH v2] ieee1275 : Add a check for invalid partition number

On 2025-12-20 19:04, Daniel Kiper wrote:
> On Tue, Dec 02, 2025 at 06:29:44PM +0530, Avnish Chouhan wrote:
>> Adding a check for invalid partition number. grub_strtoul() can fail
>> in several scenarios like invalid input, overflow, etc will result in
>> an invalid partition number which could lead to an undefined behavior.
>> 
>> Signed-off-by: Avnish Chouhan <[email protected]>
>> ---
>>  grub-core/kern/ieee1275/openfw.c | 13 ++++++++++++-
>>  1 file changed, 12 insertions(+), 1 deletion(-)
>> 
>> diff --git a/grub-core/kern/ieee1275/openfw.c 
>> b/grub-core/kern/ieee1275/openfw.c
>> index 3b492dd..e82dc34 100644
>> --- a/grub-core/kern/ieee1275/openfw.c
>> +++ b/grub-core/kern/ieee1275/openfw.c
>> @@ -512,7 +512,18 @@ grub_ieee1275_encode_devname (const char *path)
>>      }
>>    if (partition && partition[0])
>>      {
>> -      unsigned int partno = grub_strtoul (partition, 0, 0);
>> +      char *endptr;
>> +      grub_errno = GRUB_ERR_NONE;
> 
> You should do this reset after grub_strtoul() call. The commit 
> 533cd4d68
> (blsuki: Fix grub_errno leakage in blsuki_is_default_entry()) explains 
> why.
> 

Hi Daniel,
Thank you so much for reviewing the patch!

This I have added so that we'll not catch any earlier errors.
I will reset grub_errno after we verify the grub_strtoul as you 
suggested.

>> +      unsigned int partno = grub_strtoul (partition, &endptr, 0);
> 
> Do not cast result immediately to shorter type. You are not able to
> detect overflow then. First assign result to type size equal to type
> returned by the grub_strtoul() function and then check for overflows
> properly.
> 

Sure. I'll use unsigned long!

>> +      if (grub_errno != GRUB_ERR_NONE || *endptr != '\0')
> 
> This check is not reliable. Please take a look at the commit ac8a37dda
> (net/http: Allow use of non-standard TCP/IP ports). It shows how it
> should be done correctly. Even it is reverted now.
> 

This I have added based on the Alec's suggestion in v1, same as you 
suggested. I have used the same check as in the commit ac8a37dda, just 
skipping the range check due to use of "grub_errno != GRUB_ERR_NONE" 
condition.
Would you like me to add range check here?

> By the way, it would be nice if you could verify correctness of
> strtoul()/grub_strtoul()/... calls/checks in the GRUB code after
> the release.

Sure Daniel, I will surly do!
Thank you!

Regards,
Avnish Chouhan

> 
> Daniel

_______________________________________________
Grub-devel mailing list
[email protected]
https://lists.gnu.org/mailman/listinfo/grub-devel