Re: [PATCH v1] mm: fix div by zero in bdi_ratio_from_pages

[Stefan Roesch <shr@devkernel.io>]

David Hildenbrand <david@redhat.com> writes:

> On 04.01.25 02:20, Stefan Roesch wrote:
>> During testing it has been detected, that it is possible to get div by
>> zero error in bdi_set_min_bytes. The error is caused by the function
>> bdi_ratio_from_pages(). bdi_ratio_from_pages() calls
>> global_dirty_limits. If the dirty threshold is 0, the div by zero is
>> raised. This can happen if the root user is setting:
>> echo 0 > /proc/sys/vm/dirty_ration.
>> The following is a test case:
>> echo 0 > /proc/sys/vm/dirty_ratio
>> cd /sys/class/bdi/<device>
>> echo 1 > strict_limit
>> echo 8192 > min_bytes
>> ==> error is raised.
>> The problem is addressed by returning -EINVAL if dirty_ratio or
>> dirty_bytes is set to 0.
>> Reported-by: cheung wall <zzqq0103.hey@gmail.com>
>> Closes: https://lore.kernel.org/linux-mm/87pll35yd0.fsf@devkernel.io/T/#t
>> Signed-off-by: Stefan Roesch <shr@devkernel.io>
>> ---
>>   mm/page-writeback.c | 2 ++
>>   1 file changed, 2 insertions(+)
>> diff --git a/mm/page-writeback.c b/mm/page-writeback.c
>> index d213ead95675..91aa7a5c0078 100644
>> --- a/mm/page-writeback.c
>> +++ b/mm/page-writeback.c
>> @@ -692,6 +692,8 @@ static unsigned long bdi_ratio_from_pages(unsigned long pages)
>>   	unsigned long ratio;
>>     	global_dirty_limits(&background_thresh, &dirty_thresh);
>> +	if (!dirty_thresh)
>> +		return -EINVAL;
>>   	ratio = div64_u64(pages * 100ULL * BDI_RATIO_SCALE, dirty_thresh);
>>     	return ratio;
>
> bdi_set_min_bytes() calls bdi_ratio_from_pages() and passes the result to
> __bdi_set_min_ratio().
>
> __bdi_set_min_ratio() expects an "unsigned int min_ratio". I assume this will
> work because "max_ratio > 100 * BDI_RATIO_SCALE", but it is rather confusing ...
>
> Maybe we want something like:
>
> /* Use 101% to indicate "invalid" */
> #define BDI_RATIO_INVALID (101 * BDI_RATIO_SCALE)
>
> Or alternatively, just handle it in the callers of bdi_ratio_from_pages(),
> checking for -EINVAL manually.

David, I prefer the second option, its a bit easier to follow.