Skip to contentSkip to conversation

XFS development tree branch, xfs-filestreams-lookup, created. xfs-for-linus-3.15-rc1-14835-gb94acd4

XFS updates [email protected] 1 participant 1 message Apr 22, 2014

lore
Share & tools
lore

[email protected]

author Reviewed-by Dave Chinner Reviewed-by Dave Chinner Reviewed-by Dave Chinner Reviewed-by Dave Chinner Reviewed-by Dave Chinner Reviewed-by Dave Chinner Reviewed-by Dave Chinner
#0 lore raw

This is an automated email from the git hooks/post-receive script. It was generated because a ref change was pushed to the repository containing the project "XFS development tree".

The branch, xfs-filestreams-lookup has been created

        at  b94acd4786dce4379e986e6d58bdd74f8986af2f (commit)

- Log ----------------------------------------------------------------- commit b94acd4786dce4379e986e6d58bdd74f8986af2f

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:52 2014 +1000

xfs: add filestream allocator tracepoints

commit 3b8d90766a85e079fefaee74ca9dde43ce75edea

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:52 2014 +1000

xfs: remove xfs_filestream_associate

There is no good reason to create a filestream when a directory entry is created. Delay it until the first allocation happens to simply the code and reduce the amount of mru cache lookups we do.

commit 1919adda0732e661c6163a6505dddb0bc423b8d8

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:51 2014 +1000

xfs: don't create a slab cache for filestream items

We only have very few of these around, and allocation isn't that much of a hot path. Remove the slab cache to simplify the code, and to not waste any resources for the usual case of not having any inodes that use the filestream allocator.

commit 2cd2ef6a300b1ac912bb515b75451585c3d33ea9

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:51 2014 +1000

xfs: rewrite the filestream allocator using the dentry cache

In Linux we will always be able to find a parent inode for file that are undergoing I/O. Use this to simply the file stream allocator by only keeping track of parent inodes.

commit f37211c336d722805493aec8b13afdbb92bbfd98

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:51 2014 +1000

xfs: remove XFS_IFILESTREAM

We never test the flag except in xfs_inode_is_filestream, but that function already tests the on-disk flag or filesystem wide flags, and is used to decide if we want to set XFS_IFILESTREAM in the first place.

commit 22328d712dd7fdc984d17da2121be840d1f844cd

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:51 2014 +1000

xfs: embedd mru_elem into parent structure

There is no need to do a separate allocation for each mru element, just embedd the structure into the parent one in the user. Besides saving a memory allocation and the infrastructure required for it this also simplifies the API.

While we do major surgery on xfs_mru_cache.c also de-typedef it and make struct mru_cache private to the implementation file.

commit ce695c6551f9488e75247ac1eefcd173fda0c029

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:50 2014 +1000

xfs: handle duplicate entries in xfs_mru_cache_insert

The radix tree code can detect and reject duplicate keys at insert time. Make xfs_mru_cache_insert handle this case so that future changes to the filestream allocator can take advantage of this.

commit c977eb1065612bf64e18c61437e290c22183add8

Author: Christoph Hellwig <[email protected]>
Date:   Wed Apr 23 07:11:41 2014 +1000

xfs: split xfs_bmap_btalloc_nullfb

Split xfs_bmap_btalloc_nullfb into one function for filestream allocations and one for everything else that share a few helpers. This dramatically simplifies the control flow.

commit 8b90a33f476436ad6a49b7138d8a00ecbc62f9a6

Author: Christoph Hellwig <[email protected]>
Date:   Mon Apr 14 19:37:42 2014 +1000

xfs: don't try to use the filestream allocator for metadata allocations

xfs_bmap_btalloc_nullfb has two entirely different control flows when using the filestream allocator vs the regular one, but it get the conditionals wrong and ends up mixing the two for metadata allocations. Fix this by adding a missing userdata check and slight refactoring.

-----------------------------------------------------------------------

hooks/post-receive

Signature · 6 lines
XFS development tree

_______________________________________________
xfs mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/xfs