You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

Fengguang Wu 8bc3be2751 writeback: speed up writeback of big dirty files ... After making dirty a 100M file, the normal behavior is to start the writeback for all data after 30s delays. But sometimes the following happens instead: - after 30s: ~4M - after 5s: ~4M - after 5s: all remaining 92M Some analyze shows that the internal io dispatch queues goes like this: s_io s_more_io ------------------------- 1) 100M,1K 0 2) 1K 96M 3) 0 96M 1) initial state with a 100M file and a 1K file 2) 4M written, nr_to_write <= 0, so write more 3) 1K written, nr_to_write > 0, no more writes(BUG) nr_to_write > 0 in (3) fools the upper layer to think that data have all been written out. The big dirty file is actually still sitting in s_more_io. We cannot simply splice s_more_io back to s_io as soon as s_io becomes empty, and let the loop in generic_sync_sb_inodes() continue: this may starve newly expired inodes in s_dirty. It is also not an option to draw inodes from both s_more_io and s_dirty, an let the loop go on: this might lead to live locks, and might also starve other superblocks in sync time(well kupdate may still starve some superblocks, that's another bug). We have to return when a full scan of s_io completes. So nr_to_write > 0 does not necessarily mean that "all data are written". This patch introduces a flag writeback_control.more_io to indicate that more io should be done. With it the big dirty file no longer has to wait for the next kupdate invokation 5s later. In sync_sb_inodes() we only set more_io on super_blocks we actually visited. This avoids the interaction between two pdflush deamons. Also in __sync_single_inode() we don't blindly keep requeuing the io if the filesystem cannot progress. Failing to do so may lead to 100% iowait. Tested-by: Mike Snitzer <snitzer@gmail.com> Signed-off-by: Fengguang Wu <wfg@mail.ustc.edu.cn> Cc: Michael Rubin <mrubin@google.com> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>		17 years ago
..
9p	9p: use copy of the options value instead of original	17 years ago
adfs	Slab API: remove useless ctor parameter and reorder parameters	17 years ago
affs	fs: mark nibblemap const	17 years ago
afs	vfs: Add 64 bit i_version support	17 years ago
autofs	Use task_pid_nr() instead of pid_nr(task_pid())	17 years ago
autofs4	pid namespaces: round up the API	17 years ago
befs	fs/: Spelling fixes	17 years ago
bfs	regression: bfs endianness bug	17 years ago
cifs	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
coda	coda: convert struct class_device to struct device	17 years ago
configfs	configfs: file.c fix possible recursive locking	17 years ago
cramfs	fs/cramfs/inode.c: replace hardcoded value with preprocessor constant	17 years ago
debugfs	Kobject: convert fs/* from kobject_unregister() to kobject_put()	17 years ago
devpts	…
dlm	dlm: static initialization improvements	17 years ago
ecryptfs	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
efs	exportfs: make struct export_operations const	17 years ago
exportfs	exportfs: update documentation	17 years ago
ext2	ext2: Fix the max file size for ext2 file system.	17 years ago
ext3	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
ext4	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
fat	fat: optimize fat_count_free_clusters()	17 years ago
freevxfs	fs/: Spelling fixes	17 years ago
fuse	Kobject: convert fs/* from kobject_unregister() to kobject_put()	17 years ago
gfs2	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
hfs	hfs: fix coverity-found null deref	17 years ago
hfsplus	Slab API: remove useless ctor parameter and reorder parameters	17 years ago
hostfs	uml: fix hostfs style	17 years ago
hpfs	Slab API: remove useless ctor parameter and reorder parameters	17 years ago
hppfs	…
hugetlbfs	hugetlb: allow sticky directory mount option	17 years ago
isofs	exportfs: make struct export_operations const	17 years ago
jbd	spinlock: lockbreak cleanup	17 years ago
jbd2	spinlock: lockbreak cleanup	17 years ago
jffs2	Typoes: "whith" -> "with"	17 years ago
jfs	Spelling fixes: lenght->length	17 years ago
lockd	NLM: tear down RPC clients in nlm_shutdown_hosts	17 years ago
minix	limit minixfs printks on corrupted dir i_size	17 years ago
msdos	…
ncpfs	vm audit: add VM_DONTEXPAND to mmap for drivers that need it	17 years ago
nfs	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
nfs_common	…
nfsd	nfsd: more careful input validation in nfsctl write methods	17 years ago
nls	sparse pointer use of zero as null	17 years ago
ntfs	is_vmalloc_addr(): Check if an address is within the vmalloc boundaries	17 years ago
ocfs2	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
openpromfs	[SPARC]: Constify function pointer tables.	17 years ago
partitions	Kobject: convert fs/* from kobject_unregister() to kobject_put()	17 years ago
proc	Fix /proc dcache deadlock in do_exit	17 years ago
qnx4	Slab API: remove useless ctor parameter and reorder parameters	17 years ago
ramfs	Remove valueless definition of hard-selected RAMFS option	17 years ago
reiserfs	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
romfs	fs/romfs/inode.c: trivial improvements	17 years ago
smbfs	Merge branch 'task_killable' of git://git.kernel.org/pub/scm/linux/kernel/git/willy/misc	17 years ago
sysfs	Merge git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi-misc-2.6	17 years ago
sysv	Slab API: remove useless ctor parameter and reorder parameters	17 years ago
udf	fs/udf/balloc.c: mark a variable as uninitialized_var()	17 years ago
ufs	ufs: fix nexstep dir block size	17 years ago
vfat	…
xfs	is_vmalloc_addr(): Check if an address is within the vmalloc boundaries	17 years ago
Kconfig	nfsd: select CONFIG_PROC_FS in nfsv4 and gss server cases	17 years ago
Kconfig.binfmt	x86: compat_binfmt_elf Kconfig	17 years ago
Makefile	x86: compat_binfmt_elf Kconfig	17 years ago
aio.c	core: remove last users of empty FASTCALL macro	17 years ago
anon_inodes.c	anon-inodes use open coded atomic_inc for the shared inode	17 years ago
attr.c	VFS: make notify_change pass ATTR_KILL_S*ID to setattr operations	17 years ago
bad_inode.c	…
binfmt_aout.c	mm: fix exit_mmap BUG() on a.out binary exit	17 years ago
binfmt_elf.c	fs/binfmt_elf.c: spello fix	17 years ago
binfmt_elf_fdpic.c	pid namespaces: changes to show virtual ids to user	17 years ago
binfmt_em86.c	Convert files to UTF-8 and some cleanups	17 years ago
binfmt_flat.c	binfmt_flat: warning fixes	17 years ago
binfmt_misc.c	Convert files to UTF-8 and some cleanups	17 years ago
binfmt_script.c	Convert files to UTF-8 and some cleanups	17 years ago
binfmt_som.c	core_pattern: ignore RLIMIT_CORE if core_pattern is a pipe	17 years ago
bio.c	__bio_clone: don't calculate hw/phys segment counts	17 years ago
block_dev.c	Driver core: convert block from raw kobjects to core devices	17 years ago
buffer.c	bufferhead: revert constructor removal	17 years ago
char_dev.c	Kobject: rename kobject_init_ng() to kobject_init()	17 years ago
compat.c	timerfd: new timerfd API	17 years ago
compat_binfmt_elf.c	x86: compat_binfmt_elf	17 years ago
compat_ioctl.c	remove __attribute_used__	17 years ago
dcache.c	dcache: don't expose uninitialized memory in /proc/<pid>/fd/<fd>	17 years ago
dcookies.c	…
direct-io.c	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
dnotify.c	…
dquot.c	Don't send quota messages repeatedly when hardlimit reached	17 years ago
drop_caches.c	…
eventfd.c	…
eventpoll.c	lockdep: annotate epoll	17 years ago
exec.c	exec: rework the group exit and fix the race with kill	17 years ago
fcntl.c	pid namespaces: changes to show virtual ids to user	17 years ago
fifo.c	…
file.c	…
file_table.c	fs/file_table.c: use list_for_each_entry() instead of list_for_each()	17 years ago
filesystems.c	…
fs-writeback.c	writeback: speed up writeback of big dirty files	17 years ago
generic_acl.c	…
inode.c	ext4: Add inode version support in ext4	17 years ago
inotify.c	[PATCH] new helper - inotify_evict_watch()	17 years ago
inotify_user.c	change inotifyfs magic as the same magic is used for futexfs	17 years ago
internal.h	…
ioctl.c	…
ioprio.c	cfq-iosched: relax IOPRIO_CLASS_IDLE restrictions	17 years ago
libfs.c	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
locks.c	pid-namespaces-vs-locks-interaction	17 years ago
mbcache.c	fs: Fix to correct the mbcache entries counter	17 years ago
mpage.c	Pagecache zeroing: zero_user_segment, zero_user_segments and zero_user	17 years ago
namei.c	Use access mode instead of open flags to determine needed permissions	17 years ago
namespace.c	kobject: convert main fs kobject to use kobject_create	17 years ago
nfsctl.c	…
no-block.c	…
open.c	mark sys_open/sys_read exports unused	17 years ago
pipe.c	sched: affine sync wakeups	17 years ago
pnode.c	…
pnode.h	[PATCH] new helpers - collect_mounts() and release_collected_mounts()	17 years ago
posix_acl.c	…
quota.c	…
quota_v1.c	…
quota_v2.c	…
read_write.c	ext4: export iov_shorten from kernel for ext4's use	17 years ago
read_write.h	…
readdir.c	Use mutex_lock_killable in vfs_readdir	17 years ago
select.c	fs/select, remove unused macros	17 years ago
seq_file.c	[FS] seq_file: Introduce the seq_open_private()	17 years ago
signalfd.c	Fix a small number of "memeber" typoes.	17 years ago
splice.c	splice: always updated atime in direct splice	17 years ago
stack.c	…
stat.c	…
super.c	Convert files to UTF-8 and some cleanups	17 years ago
sync.c	…
timerfd.c	timerfd: new timerfd API	17 years ago
utimes.c	VFS: check nanoseconds in utimensat	17 years ago
xattr.c	[PATCH] pass dentry to audit_inode()/audit_inode_child()	17 years ago
xattr_acl.c	…