On 2015/9/17 3:45, Jan Kara wrote:
> On Mon 31-08-15 17:46:24, Zhang Zhen wrote:
>> In ext4_mb_discard_group_preallocations(), if free is always less than needed,
>> and some PAs are always used in every loop, it will be endless loop.
>>
>> Here we pick a random value to limit the max number of loop.
>
> Were you able to trigger this in practice or is it just a theoretical
> concern?
>
We found the problem in a stress test.
It lead to system hungtask.
Thanks!
> My slight concern is that in theory we could prematurely declare ENOSPC
> with this patch since ext4_mb_discard_preallocations() doesn't reliably
> discard all the preallocations anymore. But probably that's acceptable.
> But we should add a comment before ext4_mb_discard_group_preallocations()
> saying that the functions needn't free all the preallocations.
>
> Honza
>>
>> Signed-off-by: Zhang Zhen <zhenzhang.zhang@xxxxxxxxxx>
>> ---
>> fs/ext4/mballoc.c | 7 +++++--
>> 1 file changed, 5 insertions(+), 2 deletions(-)
>>
>> diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
>> index 34b610e..553fbde 100644
>> --- a/fs/ext4/mballoc.c
>> +++ b/fs/ext4/mballoc.c
>> @@ -3836,6 +3836,7 @@ ext4_mb_discard_group_preallocations(struct super_block *sb,
>> int err;
>> int busy = 0;
>> int free = 0;
>> + int tried = 0;
>>
>> mb_debug(1, "discard preallocation for group %u\n", group);
>>
>> @@ -3886,9 +3887,11 @@ repeat:
>> list_add(&pa->u.pa_tmp_list, &list);
>> }
>>
>> - /* if we still need more blocks and some PAs were used, try again */
>> - if (free < needed && busy) {
>> + /* if we still need more blocks and some PAs were used, try again,
>> + here 20 is a ramdon value. */
>> + if (free < needed && busy && tried < 20) {
>> busy = 0;
>> + tried++;
>> ext4_unlock_group(sb, group);
>> cond_resched();
>> goto repeat;
>> --
>> 1.9.1
>>
>>
>> .
>>
>>
>>
>>
--
To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
