>
> On Fri 20-11-15 13:34:36, Namjae Jeon wrote:
> > Actually, I can quickly create memleak with simple testcase like:
> > while(1000)
> > {
> > Random seek(max 500M) -> write 1M -> fsync ->truncate(random size)
> > }
> >
> > As there is significant memleak in this situation also, I started focusing
> > on this simple testcase only. When use this testcase, I never saw EBUSY path
> > in jbd2_journal_invalidatepage() -> journal_unmap_buffer().
> >
> > On the other hand(), for memleak pages jbd2_journal_invalidatepage() ->
> > journal_unmap_buffer() was almost bailing out every time from
> > if (!buffer_dirty(bh)) {
> > /* bdflush has written it. We can drop it now */
> > goto zap_buffer;
> > }
> >
> > As per debugging, NULL mapping page or buffer is created in below scenario:
> > Write(or Kjournald) -> jbd2_journal_commit_transaction
> > -> BH_JBDDirty buffer is added to forget list.
> > -> Buffer is added to new checkpoint,
> > additional reference count is taken on b_jcount which
> > keeps buffer busy until remove checkpoint.
> > -> Buffer is unfiled, removed from forget list, BH_JBDDirty
> > is cleared and BH_Dirty is set, it is exposed to VM.
>
> Ah, I see! Thanks for the easy reproduction testcase and for the debugging.
> Attached patch fixes the issue (at least the simple truncate case) for me.
> Can you check whether it fixes the issue for you as well? Thanks!
This patch looks good to me! I checked that all leak issues are fixed
with your patch.
Thanks for your help!
>
> Honza
> --
> Jan Kara <jack@xxxxxxxx>
> SUSE Labs, CR
--
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
