From ef989f46136dace8d44e7b4346dfb8530f6ad56b Mon Sep 17 00:00:00 2001 From: Andreas Dilger Date: Mon, 19 Nov 2012 12:30:53 -0700 Subject: [PATCH] LU-20 kernel: add descriptions to kernel patches Add some descriptive text to kernel patches so that it is more clear what the patches are needed for, and whether they can simply be dropped, or if they should be pushed upstream. Signed-off-by: Andreas Dilger Change-Id: Ie19de1feedc8bedeb7ae890886aab73815500c1e Reviewed-on: http://review.whamcloud.com/4613 Tested-by: Hudson Tested-by: Maloo Reviewed-by: James Simmons Reviewed-by: Keith Mannthey --- .../kernel_patches/patches/bh_lru_size_config.patch | 19 ++++++++++++++++++- .../patches/export-2.6.32-vanilla.patch | 8 ++++---- 2 files changed, 22 insertions(+), 5 deletions(-) diff --git a/lustre/kernel_patches/patches/bh_lru_size_config.patch b/lustre/kernel_patches/patches/bh_lru_size_config.patch index e84edba..4db0102 100644 --- a/lustre/kernel_patches/patches/bh_lru_size_config.patch +++ b/lustre/kernel_patches/patches/bh_lru_size_config.patch @@ -1,3 +1,20 @@ +Allow increasing the buffer-head per-CPU LRU size to allow efficient +filesystem operations that access many blocks for each transaction. +For example, creating a file in a large ext4 directory with quota +enabled will accesses multiple buffer heads and will overflow the LRU +at the default 8-block LRU size: + +* parent directory inode table block (ctime, nlinks for subdirs) +* new inode bitmap +* inode table block +* 2 quota blocks +* directory leaf block (not reused, but pollutes one cache entry) +* 2 levels htree blocks (only one is reused, other pollutes cache) +* 2 levels indirect/index blocks (only one is reused) + +Signed-off-by: Liang Zhen +Signed-off-by: Andreas Dilger + --- linux-2.6.32-131.6.1/fs/buffer.c +++ linux-2.6.32-131.6.1-pdo/fs/buffer.c @@ -1250,8 +1250,7 @@ @@ -19,7 +36,7 @@ +config BH_LRU_SIZE + int "buffer head per-CPU LRU size" + range 8 64 -+ default "8" ++ default "16" + help + This sets the per-CPU LRU size for buffer heads in memory. + More complex filesystems may be modiyfing multiple blocks diff --git a/lustre/kernel_patches/patches/export-2.6.32-vanilla.patch b/lustre/kernel_patches/patches/export-2.6.32-vanilla.patch index 0cb7884..95157b3 100644 --- a/lustre/kernel_patches/patches/export-2.6.32-vanilla.patch +++ b/lustre/kernel_patches/patches/export-2.6.32-vanilla.patch @@ -1,7 +1,7 @@ -security_inode_unlink() is used in filter_vfs_unlink() -to avoid lock ordering problems. I'm not sure if this -is still needed with ext4, and it definitely looks to -be gone with DMU changes. +security_inode_unlink() was used in filter_vfs_unlink() to avoid +lock ordering problems. This is no longer needed for Lustre 2.4 +with the change to osd-ldiskfs, but is kept for compatibility with +older server code so they can share the same kernel. Index: linux+rh+chaos/security/security.c =================================================================== -- 1.8.3.1