Skip to content

RFC dcache scaling part 1

Nick Piggin linux-fsdevel linux-nfs lkml 1 version In review

Get series with b4
b4 am [email protected]

Versions newest first

  1. v1 RFC Latest28 patches 17 repliesIn review

Patches in v1 No applies recorded yet

#Subject and filesSizeRepliesApplied
1/28fs: d_validate fixesfs/dcache.c+7 -185 replies
2/28kernel: kmem_ptr_validate considered harmfulinclude/linux/slab.h mm/slab.c mm/slob.c +2 more+1 -990 replies
3/28fs: dcache documentation cleanupinclude/linux/dcache.h+6 -120 replies
4/28fs: change d_delete semanticsDocumentation/filesystems/porting Documentation/filesystems/vfs.txt arch/ia64/kernel/perfmon.c +17 more+45 -391 reply
5/28cifs: dont overwrite dentry name in d_revalidatefs/cifs/dir.c+24 -190 replies
6/28jfs: dont overwrite dentry name in d_revalidatefs/jfs/namei.c+35 -80 replies
7/28fs: change d_compare for rcu-walkDocumentation/filesystems/Locking Documentation/filesystems/porting Documentation/filesystems/vfs.txt +19 more+228 -1521 reply
8/28fs: change d_hash for rcu-walkDocumentation/filesystems/Locking Documentation/filesystems/porting Documentation/filesystems/vfs.txt +20 more+88 -491 reply
9/28hostfs: simplify lockingfs/dcache.c fs/hostfs/hostfs_kern.c include/linux/dcache.h+24 -170 replies
10/28fs: dcache scale hashfs/dcache.c include/linux/dcache.h+30 -110 replies
11/28fs: dcache scale lrufs/dcache.c+84 -280 replies
12/28fs: dcache scale dentry refcountarch/powerpc/platforms/cell/spufs/inode.c drivers/infiniband/hw/ipath/ipath_fs.c drivers/infiniband/hw/qib/qib_fs.c +18 more+126 -690 replies
13/28fs: dcache scale d_unhashedarch/powerpc/platforms/cell/spufs/inode.c drivers/usb/core/inode.c fs/autofs4/autofs_i.h +7 more+109 -561 reply
14/28fs: dcache scale subdirsdrivers/staging/smbfs/cache.c drivers/usb/core/inode.c fs/autofs4/autofs_i.h +13 more+314 -1341 reply
15/28fs: scale inode alias listfs/9p/vfs_inode.c fs/affs/amigaffs.c fs/cifs/inode.c +6 more+78 -91 reply
16/28fs: Use rename lock and RCU for multi-step operationsdrivers/staging/pohmelfs/path_entry.c fs/autofs4/waitq.c fs/dcache.c +2 more+152 -301 reply
17/28fs: increase d_name lock coveragefs/dcache.c+16 -70 replies
18/28fs: dcache remove dcache_lockDocumentation/filesystems/Locking Documentation/filesystems/dentry-locking.txt Documentation/filesystems/porting +37 more+120 -3150 replies
19/28fs: dcache avoid starvation in dcache multi-step operationsfs/dcache.c+42 -140 replies
20/28fs: dcache reduce dput lockingfs/dcache.c+25 -310 replies
21/28fs: dcache reduce locking in d_allocfs/dcache.c+4 -20 replies
22/28fs: dcache reduce dcache_inode_lockfs/dcache.c+12 -120 replies
23/28fs: dcache rationalise dget variantsarch/powerpc/platforms/cell/spufs/inode.c drivers/infiniband/hw/ipath/ipath_fs.c drivers/infiniband/hw/qib/qib_fs.c +9 more+23 -460 replies
24/28fs: dcache reduce d_parent lockingfs/dcache.c+14 -110 replies
25/28fs: dcache reduce prune_one_dentry lockingfs/dcache.c+15 -120 replies
26/28fs: reduce dcache_inode_lock width in lru scanningfs/dcache.c+8 -80 replies
27/28fs: use RCU in shrink_dentry_list to reduce lock nestingfs/dcache.c+21 -170 replies
28/28fs: consolidate dentry kill sequencefs/dcache.c+62 -750 replies

Reviews trailers per patch

None of the 28 patches has a Reviewed-by or Acked-by of their own; the series has one for all of them.

PatchTim Pepper
SeriesR-b
1/28
2/28
3/28
4/28
5/28
6/28
7/28
8/28
9/28
10/28
11/28
12/28
13/28
14/28
15/28
16/28
17/28
18/28
19/28
20/28
21/28
22/28
23/28
24/28
25/28
26/28
27/28
28/28

Trailers given in replies are filled; those carried in the patches from earlier versions are outlined with dashes.

Read the v1 thread

Cover letter [patch 00/28] [rfc] dcache scaling part 1

There are 3 main parts to dcache scaling. This one primarily adds new locks to take over dcache_lock, and some pre/post prep and streamlining patches.

The second implements fine grained locking, and is rather trivial after part 1.

The third implements rcu-walk. rcu-walk depends on the first part, because it relies on using d_lock to protect the state of the dentry (when converting from rcu-walk to refcounted walk). Without the fine grained locing, we'd need to use dcache_lock for that, which would be a step backwards to put into path walking again.

Comments?