[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ext3+acl


On Mon, Jun 18, 2001 at 06:49:01AM -0700, Jay Weber wrote:

> I'm cc'ing the ext3 list also as I get the feeling this is still an issue
> in stock ext3-0.0.7a patch as you've stated it works for you after
> applying the LFS patch.

Stock ext3 on 2.2 doesn't have the LFS stuff but I've now pushed it
all into CVS locally to prepare for an LFS branch.

> Stephen may not be aware and better yet, maybe we're just masking the
> issue in our case.  The assertion it would generate prior was
> h->buffer_credits > 0 or such.

Right.  ext3's journaling layer needs to be really careful about
transaction sizes: there's no point starting a transaction and finding
there is not enough room in the journal to finish it, so we need
transactions to make an educated guess in advance about the maximum
number of log buffers they will need.  If you modify an inode for LFS
and have to update the superblock too (to set the flag indicating that
there are large files on the partition), then you MUST have both
blocks reserved in the journal_start.  The behaviour you were
describing is perfectly expected, and the solution is correct.


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]