On 09/01/2020 10:33, Vlastimil Babka wrote:
> On 1/8/20 4:21 PM, Sergey Dyasli wrote:
>> From: Ross Lagerwall <ross.lagerwall@xxxxxxxxxx>
>>
>> When KASAN (or SLUB_DEBUG) is turned on, the normal expectation that
>> allocations are aligned to the next power of 2 of the size does not
>> hold.
>
> Hmm, really? They should after 59bb47985c1d ("mm, sl[aou]b: guarantee
> natural alignment for kmalloc(power-of-two)"), i.e. since 5.4.
>
> But actually the guarantee is only for precise power of two sizes given
> to kmalloc(). Allocations of sizes that end up using the 96 or 192 bytes
> kmalloc cache have no such guarantee. But those might then cross page
> boundary also without SLUB_DEBUG.
That's interesting to know. It's certainly not the case for 4.19 kernel
for which PV KASAN was initially developed. But I guess this means that
only patch description needs updating.
>
>> Therefore, handle grant copies that cross page boundaries.
>>
>> Signed-off-by: Ross Lagerwall <ross.lagerwall@xxxxxxxxxx>
>> Signed-off-by: Sergey Dyasli <sergey.dyasli@xxxxxxxxxx>
--
Thanks,
Sergey
