The use in gnttab_map() is latently buggy, as "frames" there is
xen_pfn_t *, not unsigned long *. Adjust the correct use in
gnttab_map_frames_v2() as well, just to avoid the problematic pattern of
sizeof(<type>).

Signed-off-by: Jan Beulich <jbeulich@suse.com>

--- a/drivers/xen/grant-table.c
+++ b/drivers/xen/grant-table.c
@@ -1406,7 +1406,7 @@ static int gnttab_map_frames_v2(xen_pfn_
 	/* No need for kzalloc as it is initialized in following hypercall
 	 * GNTTABOP_get_status_frames.
 	 */
-	sframes = kmalloc_array(nr_sframes, sizeof(uint64_t), GFP_ATOMIC);
+	sframes = kmalloc_array(nr_sframes, sizeof(*sframes), GFP_ATOMIC);
 	if (!sframes)
 		return -ENOMEM;
 
@@ -1478,7 +1478,7 @@ static int gnttab_map(unsigned int start
 	/* No need for kzalloc as it is initialized in following hypercall
 	 * GNTTABOP_setup_table.
 	 */
-	frames = kmalloc_array(nr_gframes, sizeof(unsigned long), GFP_ATOMIC);
+	frames = kmalloc_array(nr_gframes, sizeof(*frames), GFP_ATOMIC);
 	if (!frames)
 		return -ENOMEM;

On 30/07/2026 3:57 pm, Jan Beulich wrote:
> The use in gnttab_map() is latently buggy, as "frames" there is
> xen_pfn_t *, not unsigned long *.

ARM32 has 64bit xen_pfn_t's.  At least it's wrong in the safe direction.

>  Adjust the correct use in
> gnttab_map_frames_v2() as well, just to avoid the problematic pattern of
> sizeof(<type>).
>
> Signed-off-by: Jan Beulich <jbeulich@suse.com>

Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>

On 30.07.2026 17:25, Andrew Cooper wrote:
> On 30/07/2026 3:57 pm, Jan Beulich wrote:
>> The use in gnttab_map() is latently buggy, as "frames" there is
>> xen_pfn_t *, not unsigned long *.
> 
> ARM32 has 64bit xen_pfn_t's.  At least it's wrong in the safe direction.

No, that alone would make it unsafe - we'd have allocated only half the
size. What keeps the problem being latent is that the problem is on the
PV-only path.

>>  Adjust the correct use in
>> gnttab_map_frames_v2() as well, just to avoid the problematic pattern of
>> sizeof(<type>).
>>
>> Signed-off-by: Jan Beulich <jbeulich@suse.com>
> 
> Reviewed-by: Andrew Cooper <andrew.cooper3@citrix.com>

Thanks.

Jan