Skip to content

fuse: An attempt to implement a write-back cache policy

Maxim Patlasov linux-fsdevel lkml 5 versions – In review

Get series with b4
b4 am [email protected]

Versions newest first · linked by author and title

  1. v6 Latest11 patches 9 repliesIn review
  2. v516 patches 11 repliesSuperseded
  3. v414 patches 13 repliesSuperseded
Earlier versions · 2
VersionPostedPatchesRepliesStatus
v314 patches8 repliesSuperseded
v214 patches10 repliesSuperseded

Changes in v6 from the cover letter

- rebased on writepages.v2 branch of the fuse tree (beba4ae4c2fda4e03a813aec640586087fa80a8b)
- added waiting for fuse writeback to ->write_begin fuse address space operation.

Patches in v6 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/11fuse: Linking file to inode helperfs/fuse/file.c+19 -140 replies
2/11fuse: Prepare to handle short readsfs/fuse/file.c+13 -80 replies
3/11fuse: Connection bit for enabling writebackfs/fuse/fuse_i.h+3 -00 replies
4/11fuse: Trust kernel i_size only - v4fs/fuse/dir.c fs/fuse/file.c fs/fuse/inode.c+44 -60 replies
5/11fuse: Trust kernel i_mtime only -v2fs/fuse/dir.c fs/fuse/file.c fs/fuse/fuse_i.h +1 more+138 -204 replies
5/11fuse: Trust kernel i_mtime only -v3fs/fuse/dir.c fs/fuse/file.c fs/fuse/fuse_i.h +1 more+136 -260 replies
6/11fuse: Flush files on wb close -v2fs/fuse/file.c+6 -01 reply
7/11fuse: restructure fuse_readpage()fs/fuse/file.c+37 -184 replies
8/11fuse: Implement write_begin/write_end callbacks -v2fs/fuse/file.c+100 -00 replies
9/11fuse: fuse_flush() should wait on writebackfs/fuse/file.c+9 -00 replies
10/11fuse: Fix O_DIRECT operations vs cached writeback misorder - v2fs/fuse/cuse.c fs/fuse/file.c fs/fuse/fuse_i.h+62 -50 replies
11/11fuse: Turn writeback cache onfs/fuse/file.c fs/fuse/fuse_i.h fs/fuse/inode.c +1 more+24 -00 replies

Reviews trailers per patch

No Reviewed-by, Acked-by or Tested-by trailers for v6 yet.

Read the v6 thread

Cover letter [PATCH v6 00/11] fuse: An attempt to implement a write-back cache policy

Hi,

This is the sixth iteration of Pavel Emelyanov's patch-set implementing write-back policy for FUSE page cache. This version is for the branch "writepages.v2" of the fuse git tree, so a few patches from the previous version are dropped (because they are already in the tree). Initial patch-set description was the following:

One of the problems with the existing FUSE implementation is that it uses the write-through cache policy which results in performance problems on certain workloads. E.g. when copying a big file into a FUSE file the cp pushes every 128k to the userspace synchronously. This becomes a problem when the userspace back-end uses networking for storing the data.

A good solution of this is switching the FUSE page cache into a write-back policy. With this file data are pushed to the userspace with big chunks (depending on the dirty memory limits, but this is much more than 128k) which lets the FUSE daemons handle the size updates in a more efficient manner.

The writeback feature is per-connection and is explicitly configurable at the init stage (is it worth making it CAP_SOMETHING protected?) When the writeback is turned ON:

* still copy writeback pages to temporary buffer when sending a writeback request and finish the page writeback immediately

* make kernel maintain the inode's i_size to avoid frequent i_size synchronization with the user space

* take NR_WRITEBACK_TEMP into account when makeing balance_dirty_pages decision. This protects us from having too many dirty pages on FUSE

The provided patchset survives the fsx test. Performance measurements are not yet all finished, but the mentioned copying of a huge file becomes noticeably faster even on machines with few RAM and doesn't make the system stuck (the dirty pages balancer does its work OK). Applies on top of v3.5-rc4.

We are currently exploring this with our own distributed storage implementation which is heavily oriented on storing big blobs of data with extremely rare meta-data updates (virtual machines' and containers' disk images). With the existing cache policy a typical usage scenario -- copying a big VM disk into a cloud -- takes way too much time to proceed, much longer than if it was simply scp-ed over the same network. The write-back policy (as I mentioned) noticeably improves this scenario. Kirill (in Cc) can share more details about the performance and the storage concepts details if required.

Changed in v2:
- numerous bugfixes:
- fuse_write_begin and fuse_writepages_fill and fuse_writepage_locked must wait on page writeback because page writeback can extend beyond the lifetime of the page-cache page
- fuse_send_writepages can end_page_writeback on original page only after adding request to fi->writepages list; otherwise another writeback may happen inside the gap between end_page_writeback and adding to the list
- fuse_direct_io must wait on page writeback; otherwise data corruption is possible due to reordering requests
- fuse_flush must flush dirty memory and wait for all writeback on given inode before sending FUSE_FLUSH to userspace; otherwise FUSE_FLUSH is not reliable
- fuse_file_fallocate must hold i_mutex around FUSE_FALLOCATE and i_size update; otherwise a race with a writer extending i_size is possible
- fix handling errors in fuse_writepages and fuse_send_writepages
- handle i_mtime intelligently if writeback cache is on (see patch #7 (update i_mtime on buffered writes) for details.
- put enabling writeback cache under fusermount control; (see mount option 'allow_wbcache' introduced by patch #13 (turn writeback cache on))
- rebased on v3.7-rc5

Changed in v3:
- rebased on for-next branch of the fuse tree (fb05f41f5f96f7423c53da4d87913fb44fd0565d)

Changed in v4:
- rebased on for-next branch of the fuse tree (634734b63ac39e137a1c623ba74f3e062b6577db)
- fixed fuse_fillattr() for non-writeback_chace case
- added comments explaining why we cannot trust size from server
- rewrote patch handling i_mtime; it's titled Trust-kernel-i_mtime-only now
- simplified patch titled Flush-files-on-wb-close
- eliminated code duplications from fuse_readpage() ans fuse_prepare_write()
- added comment about "disk full" errors to fuse_write_begin()

Changed in v5:
- rebased on for-next branch of the fuse tree (e5c5f05dca0cf90f0f3bb1aea85dcf658baff185) with "hold i_mutex in fuse_file_fallocate() - v2" patch applied manually
- updated patch descriptions according to Miklos' demand (using From: for patches initially developed by someone else)
- moved ->writepages() to a separate patch and enabled it unconditionally for mmaped writeback
- moved restructuring fuse_readpage to a separate patch
- avoided use of union for fuse_fill_data
- grabbed a ref to the request in fuse_send_writepages

Changed in v6:
- rebased on writepages.v2 branch of the fuse tree (beba4ae4c2fda4e03a813aec640586087fa80a8b)
- added waiting for fuse writeback to ->write_begin fuse address space operation.

Thanks,
Maxim

---

Maxim Patlasov (9):
fuse: Linking file to inode helper
fuse: Connection bit for enabling writeback
fuse: Trust kernel i_size only - v4
fuse: Trust kernel i_mtime only -v2
fuse: Flush files on wb close -v2
fuse: restructure fuse_readpage()
fuse: fuse_flush() should wait on writeback
fuse: Fix O_DIRECT operations vs cached writeback misorder - v2
fuse: Turn writeback cache on

Pavel Emelyanov (2):
fuse: Prepare to handle short reads
fuse: Implement write_begin/write_end callbacks -v2

 fs/fuse/cuse.c            |    5 -
 fs/fuse/dir.c             |  120 +++++++++++++++-
 fs/fuse/file.c            |  328 +++++++++++++++++++++++++++++++++++++++------
 fs/fuse/fuse_i.h          |   26 +++-
 fs/fuse/inode.c           |   37 ++++-
 include/uapi/linux/fuse.h |    2 
 6 files changed, 451 insertions(+), 67 deletions(-)
Signature · 1 line
Signature