Re: [PATCH 3/9] xfs: convert buffer cache to use high order folios

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

 



On Tue, Mar 19, 2024 at 02:32:04PM -0700, Christoph Hellwig wrote:
> On Tue, Mar 19, 2024 at 10:29:09AM -0700, Darrick J. Wong wrote:
> > So.... does that mean a 128K folio for a 68k xattr remote value buffer?
> 
> I though 64k was the maximum xattr size?

64k is the maximum xattr value size, yes.  But remote xattr value blocks
now have block headers complete with owner/uuid/magic/etc.  Each block
can only store $blksz-56 bytes now.  Hence that 64k value needs
ceil(65536 / 4040) == 17 blocks on a 4k fsb filesystem.

(On a 64k fsb filesystem, that bloats up to 128k.)

Luckily we bwrite remote blocks to disk directly, which is why we've
never blown out the buffer log item bitmap.

> > I've been noticing the 4k merkle tree blobs consume 2 fsb in the xattr
> > tree, which isn't awesome.
> 
> Maybe that's a question for the fsverity thread, but how do we end
> up with these weird sizes?

Same reason.  Merkle tree blocks are $blksz by default, but storing a
4096 blob in the xattrs requires ceil(4096 / 4040) == 2 blocks.  Most of
that second block is wasted.

--D




[Index of Archives]     [XFS Filesystem Development (older mail)]     [Linux Filesystem Development]     [Linux Audio Users]     [Yosemite Trails]     [Linux Kernel]     [Linux RAID]     [Linux SCSI]


  Powered by Linux