Re: [RFC] ext3/jbd race: releasing in-use journal_heads

[Stephen C. Tweedie]

Hi,

On Wed, 2005-03-09 at 13:28, Jan Kara wrote:

>   Hmm. I see for example a place at jbd/commit.c, line 287 (which you
> did not change in your patch) which does this and doesn't seem to be
> protected against journal_unmap_buffer() (but maybe I miss something).
> Not that I'd find that race probable but in theory...

Indeed; I can't see why that wouldn't trigger (at least without the
existing, low-risk journal_unmap_buffer() patch.)

Andrew, I think we just go with that simple patch for now --- it catches
the cases we actually see in testing.  And rather than mess with
temporary states where b_transaction goes NULL, I think the *correct*
long-term fix is to hide those states using locking rather than by list
tricks.  Merging the bh_journal_head and bh_state locks really seems
like the safe solution here, as the latter seems to be held nearly
everywhere where we need protection against journal_put_journal_head()
(and where it's not held --- as it wasn't in journal_unmap_buffer() ---
it's a bug.)

--Stephen

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/