This simplifies the RCU to refcounting code in particular.
I was originally intending to leave this for later, but walking through
all the dput() logic (see previous commit), I realized that the dput()
"might_sleep()" check was misleadingly weak. And I removed it as
misleading, both for performance profiling and for debugging.
However, the might_sleep() debugging case is actually true: the final
dput() can indeed sleep, if the inode of the dentry that you are
releasing ends up sleeping at iput time (see dentry_iput()). So the
problem with the might_sleep() in dput() wasn't that it wasn't true, it
was that it wasn't actually testing and triggering on the interesting
case.
In particular, just about *any* dput() can indeed sleep, if you happen
to race with another thread deleting the file in question, and you then
lose the race to the be the last dput() for that file. But because it's
a very rare race, the debugging code would never trigger it in practice.
Why is this problematic? The new d_rcu_to_refcount() (see commit
15570086b5: "vfs: reimplement d_rcu_to_refcount() using
lockref_get_or_lock()") does a dput() for the failure case, and it does
it under the RCU lock. So potentially sleeping really is a bug.
But there's no way I'm going to fix this with the previous complicated
"lockref_get_or_lock()" interface. And rather than revert to the old
and crufty nested dentry locking code (which did get this right by
delaying the reference count updates until they were verified to be
safe), let's make forward progress.
Cc: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
This is me being a bit OCD after all the dentry optimization work this
merge window: profiles end up showing 'dput()' as a rather expensive
operation, and there were two unrelated bad reasons for that.
The first reason was reading d_lockref.count for debugging purposes,
which touches the lockref cacheline (for reads) before really need to.
More importantly, the debugging test in question is _wrong_, and has
hidden bugs. It's true that we can only sleep when the count goes down
to zero, but the test as-is hides the much more subtle bug that happens
if we race with somebody else deleting the file.
Anyway we _will_ touch that cacheline, but let's do it for a write and
in the right routine (ie in "lockref_put_or_lock()") which annotates the
costs better. So remove the misleading debug code.
The other was an unnecessary access to the cacheline that contains the
d_lru list, just to check whether we already were on the LRU list or
not. This is exactly what we have d_flags for, so that we can avoid
touching extra cache lines for the common case. So just add another bit
for "is this dentry on the LRU".
Finally, mark the tests properly likely/unlikely, so that the common
fast-paths are dense in the instruction stream.
This makes the profiles look much saner.
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs pile 2 (of many) from Al Viro:
"Mostly Miklos' series this time"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs:
constify dcache.c inlined helpers where possible
fuse: drop dentry on failed revalidate
fuse: clean up return in fuse_dentry_revalidate()
fuse: use d_materialise_unique()
sysfs: use check_submounts_and_drop()
nfs: use check_submounts_and_drop()
gfs2: use check_submounts_and_drop()
afs: use check_submounts_and_drop()
vfs: check unlinked ancestors before mount
vfs: check submounts and drop atomically
vfs: add d_walk()
vfs: restructure d_genocide()
Pull namespace changes from Eric Biederman:
"This is an assorted mishmash of small cleanups, enhancements and bug
fixes.
The major theme is user namespace mount restrictions. nsown_capable
is killed as it encourages not thinking about details that need to be
considered. A very hard to hit pid namespace exiting bug was finally
tracked and fixed. A couple of cleanups to the basic namespace
infrastructure.
Finally there is an enhancement that makes per user namespace
capabilities usable as capabilities, and an enhancement that allows
the per userns root to nice other processes in the user namespace"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/ebiederm/user-namespace:
userns: Kill nsown_capable it makes the wrong thing easy
capabilities: allow nice if we are privileged
pidns: Don't have unshare(CLONE_NEWPID) imply CLONE_THREAD
userns: Allow PR_CAPBSET_DROP in a user namespace.
namespaces: Simplify copy_namespaces so it is clear what is going on.
pidns: Fix hang in zap_pid_ns_processes by sending a potentially extra wakeup
sysfs: Restrict mounting sysfs
userns: Better restrictions on when proc and sysfs can be mounted
vfs: Don't copy mount bind mounts of /proc/<pid>/ns/mnt between namespaces
kernel/nsproxy.c: Improving a snippet of code.
proc: Restrict mounting the proc filesystem
vfs: Lock in place mounts from more privileged users
Pull security subsystem updates from James Morris:
"Nothing major for this kernel, just maintenance updates"
* 'next' of git://git.kernel.org/pub/scm/linux/kernel/git/jmorris/linux-security: (21 commits)
apparmor: add the ability to report a sha1 hash of loaded policy
apparmor: export set of capabilities supported by the apparmor module
apparmor: add the profile introspection file to interface
apparmor: add an optional profile attachment string for profiles
apparmor: add interface files for profiles and namespaces
apparmor: allow setting any profile into the unconfined state
apparmor: make free_profile available outside of policy.c
apparmor: rework namespace free path
apparmor: update how unconfined is handled
apparmor: change how profile replacement update is done
apparmor: convert profile lists to RCU based locking
apparmor: provide base for multiple profiles to be replaced at once
apparmor: add a features/policy dir to interface
apparmor: enable users to query whether apparmor is enabled
apparmor: remove minimum size check for vmalloc()
Smack: parse multiple rules per write to load2, up to PAGE_SIZE-1 bytes
Smack: network label match fix
security: smack: add a hash table to quicken smk_find_entry()
security: smack: fix memleak in smk_write_rules_list()
xattr: Constify ->name member of "struct xattr".
...
Pull trivial tree from Jiri Kosina:
"The usual trivial updates all over the tree -- mostly typo fixes and
documentation updates"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jikos/trivial: (52 commits)
doc: Documentation/cputopology.txt fix typo
treewide: Convert retrun typos to return
Fix comment typo for init_cma_reserved_pageblock
Documentation/trace: Correcting and extending tracepoint documentation
mm/hotplug: fix a typo in Documentation/memory-hotplug.txt
power: Documentation: Update s2ram link
doc: fix a typo in Documentation/00-INDEX
Documentation/printk-formats.txt: No casts needed for u64/s64
doc: Fix typo "is is" in Documentations
treewide: Fix printks with 0x%#
zram: doc fixes
Documentation/kmemcheck: update kmemcheck documentation
doc: documentation/hwspinlock.txt fix typo
PM / Hibernate: add section for resume options
doc: filesystems : Fix typo in Documentations/filesystems
scsi/megaraid fixed several typos in comments
ppc: init_32: Fix error typo "CONFIG_START_KERNEL"
treewide: Add __GFP_NOWARN to k.alloc calls with v.alloc fallbacks
page_isolation: Fix a comment typo in test_pages_isolated()
doc: fix a typo about irq affinity
...
Pull ext3, reiserfs, udf & isofs fixes from Jan Kara:
"The contains a bunch of ext3 cleanups and minor improvements, major
reiserfs locking changes which should hopefully fix deadlocks
introduced by BKL removal, and udf/isofs changes to refuse mounting fs
rw instead of mounting it ro automatically which makes eject button
work as expected for all media (see the changelog for why userspace
should be ok with this change)"
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/jack/linux-fs:
jbd: use a single printk for jbd_debug()
reiserfs: locking, release lock around quota operations
reiserfs: locking, handle nested locks properly
reiserfs: locking, push write lock out of xattr code
jbd: relocate assert after state lock in journal_commit_transaction()
udf: Refuse RW mount of the filesystem instead of making it RO
udf: Standardize return values in mount sequence
isofs: Refuse RW mount of the filesystem instead of making it RO
ext3: allow specifying external journal by pathname mount option
jbd: remove unneeded semicolon
Pull f2fs updates from Jaegeuk Kim:
"This patch-set includes the following major enhancement patches:
- support inline xattrs
- add sysfs support to control GCs explicitly
- add proc entry to show the current segment usage information
- improve the GC/SSR performance
The other bug fixes are as follows:
- avoid the overflow on status calculation
- fix some error handling routines
- fix inconsistent xattr states after power-off-recovery
- fix incorrect xattr node offset definition
- fix deadlock condition in fsync
- fix the fdatasync routine for power-off-recovery"
* tag 'for-f2fs-3.12' of git://git.kernel.org/pub/scm/linux/kernel/git/jaegeuk/f2fs: (40 commits)
f2fs: optimize gc for better performance
f2fs: merge more bios of node block writes
f2fs: avoid an overflow during utilization calculation
f2fs: trigger GC when there are prefree segments
f2fs: use strncasecmp() simplify the string comparison
f2fs: fix omitting to update inode page
f2fs: support the inline xattrs
f2fs: add the truncate_xattr_node function
f2fs: introduce __find_xattr for readability
f2fs: reserve the xattr space dynamically
f2fs: add flags for inline xattrs
f2fs: fix error return code in init_f2fs_fs()
f2fs: fix wrong BUG_ON condition
f2fs: fix memory leak when init f2fs filesystem fail
f2fs: fix a compound statement label error
f2fs: avoid writing inode redundantly when creating a file
f2fs: alloc_page() doesn't return an ERR_PTR
f2fs: should cover i_xattr_nid with its xattr node page lock
f2fs: check the free space first in new_node_page
f2fs: clean up the needless end 'return' of void function
...
Pull sparc changes from David Miller:
"Several bug fixes (from Kirill Tkhai, Geery Uytterhoeven, and Alexey
Dobriyan) and some support for Fujitsu sparc64x chips (from Allen
Pais)"
* git://git.kernel.org/pub/scm/linux/kernel/git/davem/sparc:
sparc64: Export flush_ptrace_access() (needed by lustre)
sparc: fix PCI device proc file mmap(2)
sparc64: Remove RWSEM export leftovers
sparc64: Fix off by one in trampoline TLB mapping installation loop.
sparc64: Fix ITLB handler of null page
esp_scsi: Fix tag state corruption when autosensing.
sparc64: Fix not SRA'ed %o5 in 32-bit traced syscall
sparc64: cleanup: Rename ret_from_syscall to ret_from_fork
sparc32: Fix exit flag passed from traced sys_sigreturn
sparc64: Fix wrong syscall return value passed to trace_sys_exit()
support sparc64x chip type in cpumap.c
cpu hw caps support for sparc64x
Drop a subtree when we find that it has moved or been delated. This can be
done as long as there are no submounts under this location.
If the directory was moved and we come across the same directory in a
future lookup it will be reconnected by d_materialise_unique().
Signed-off-by: Anand Avati <avati@redhat.com>
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
On errors unrelated to the filesystem's state (ENOMEM, ENOTCONN) return the
error itself from ->d_revalidate() insted of returning zero (invalid).
Also make a common label for invalidating the dentry. This will be used by
the next patch.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Use d_materialise_unique() instead of d_splice_alias(). This allows dentry
subtrees to be moved to a new place if there moved, even if something is
referencing a dentry in the subtree (open fd, cwd, etc..).
This will also allow us to drop a subtree if it is found to be replaced by
something else. In this case the disconnected subtree can later be
reconnected to its new location.
d_materialise_unique() ensures that a directory entry only ever has one
alias. We keep fc->inst_mutex around the calls for d_materialise_unique()
on directories to prevent a race with mkdir "stealing" the inode.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Do have_submounts(), shrink_dcache_parent() and d_drop() atomically.
check_submounts_and_drop() can deal with negative dentries and
non-directories as well.
Non-directories can also be mounted on. And just like directories we don't
want these to disappear with invalidation.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Acked-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Do have_submounts(), shrink_dcache_parent() and d_drop() atomically.
check_submounts_and_drop() can deal with negative dentries and
non-directories as well.
Non-directories can also be mounted on. And just like directories we don't
want these to disappear with invalidation.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Trond Myklebust <Trond.Myklebust@netapp.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Do have_submounts(), shrink_dcache_parent() and d_drop() atomically.
check_submounts_and_drop() can deal with negative dentries and
non-directories as well.
Non-directories can also be mounted on. And just like directories we don't
want these to disappear with invalidation.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: Steven Whitehouse <swhiteho@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Do have_submounts(), shrink_dcache_parent() and d_drop() atomically.
check_submounts_and_drop() can deal with negative dentries as well.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: David Howells <dhowells@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We check submounts before doing d_drop() on a non-empty directory dentry in
NFS (have_submounts()), but we do not exclude a racing mount. Nor do we
prevent mounts to be added to the disconnected subtree using relative paths
after the d_drop().
This patch fixes these issues by checking for unlinked (unhashed, non-root)
ancestors before proceeding with the mount. This is done with rename
seqlock taken for write and with ->d_lock grabbed on each ancestor in turn,
including our dentry itself. This ensures that the only one of
check_submounts_and_drop() or has_unlinked_ancestor() can succeed.
Signed-off-by: Miklos Szeredi <miklos@szeredi.hu>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We check submounts before doing d_drop() on a non-empty directory dentry in
NFS (have_submounts()), but we do not exclude a racing mount.
Process A: have_submounts() -> returns false
Process B: mount() -> success
Process A: d_drop()
This patch prepares the ground for the fix by doing the following
operations all under the same rename lock:
have_submounts()
shrink_dcache_parent()
d_drop()
This is actually an optimization since have_submounts() and
shrink_dcache_parent() both traverse the same dentry tree separately.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
CC: David Howells <dhowells@redhat.com>
CC: Steven Whitehouse <swhiteho@redhat.com>
CC: Trond Myklebust <Trond.Myklebust@netapp.com>
CC: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This one replaces three instances open coded tree walking (have_submounts,
select_parent, d_genocide) with a common helper.
In addition to slightly reducing the kernel size, this simplifies the
callers and makes them less bug prone.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
It shouldn't matter when we decrement the refcount during the walk as long
as we do it exactly once.
Restructure d_genocide() to do the killing on entering the dentry instead
of when leaving it. This helps creating a common helper for tree walking.
Signed-off-by: Miklos Szeredi <mszeredi@suse.cz>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
Commit 786d7e1612 "Fix rmmod/read/write races in /proc entries"
must have broken mmapping of PCI device proc files on Sparc.
Notice how it adds wrapper around ->mmap but doesn't do it around ->get_unmapped_area.
Add wrapper around ->get_unmapped_area.
Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull vfs pile 1 from Al Viro:
"Unfortunately, this merge window it'll have a be a lot of small piles -
my fault, actually, for not keeping #for-next in anything that would
resemble a sane shape ;-/
This pile: assorted fixes (the first 3 are -stable fodder, IMO) and
cleanups + %pd/%pD formats (dentry/file pathname, up to 4 last
components) + several long-standing patches from various folks.
There definitely will be a lot more (starting with Miklos'
check_submount_and_drop() series)"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/viro/vfs: (26 commits)
direct-io: Handle O_(D)SYNC AIO
direct-io: Implement generic deferred AIO completions
add formats for dentry/file pathnames
kvm eventfd: switch to fdget
powerpc kvm: use fdget
switch fchmod() to fdget
switch epoll_ctl() to fdget
switch copy_module_from_fd() to fdget
git simplify nilfs check for busy subtree
ibmasmfs: don't bother passing superblock when not needed
don't pass superblock to hypfs_{mkdir,create*}
don't pass superblock to hypfs_diag_create_files
don't pass superblock to hypfs_vm_create_files()
oprofile: get rid of pointless forward declarations of struct super_block
oprofilefs_create_...() do not need superblock argument
oprofilefs_mkdir() doesn't need superblock argument
don't bother with passing superblock to oprofile_create_stats_files()
oprofile: don't bother with passing superblock to ->create_files()
don't bother passing sb to oprofile_create_files()
coh901318: don't open-code simple_read_from_buffer()
...
This patch improves the gc efficiency by optimizing the victim
selection policy. With this optimization, the random re-write
performance could increase up to 20%.
For f2fs, when disk is in shortage of free spaces, gc will selects
dirty segments and moves valid blocks around for making more space
available. The gc cost of a segment is determined by the valid blocks
in the segment. The less the valid blocks, the higher the efficiency.
The ideal victim segment is the one that has the most garbage blocks.
Currently, it searches up to 20 dirty segments for a victim segment.
The selected victim is not likely the best victim for gc when there
are much more dirty segments. Why not searching more dirty segments
for a better victim? The cost of searching dirty segments is
negligible in comparison to moving blocks.
In this patch, it enlarges the MAX_VICTIM_SEARCH to 4096 to make
the search more aggressively for a possible better victim. Since
it also applies to victim selection for SSR, it will likely improve
the SSR efficiency as well.
The test case is simple. It creates as many files until the disk full.
The size for each file is 32KB. Then it writes as many as 100000
records of 4KB size to random offsets of random files in sync mode.
The testing was done on a 2GB partition of a SDHC card. Let's see the
test result of f2fs without and with the patch.
---------------------------------------
2GB partition, SDHC
create 52023 files of size 32768 bytes
random re-write 100000 records of 4KB
---------------------------------------
| file creation (s) | rewrite time (s) | gc count | gc garbage blocks |
[no patch] 341 4227 1174 174840
[patched] 324 2958 645 106682
It's obvious that, with the patch, f2fs finishes the test in 20+% less
time than without the patch. And internally it does much less gc with
higher efficiency than before.
Since the performance improvement is related to gc, it might not be so
obvious for other tests that do not trigger gc as often as this one (
This is because f2fs selects dirty segments for SSR use most of the
time when free space is in shortage). The well-known iozone test tool
was not used for benchmarking the patch becuase it seems do not have
a test case that performs random re-write on a full disk.
This patch is the revised version based on the suggestion from
Jaegeuk Kim.
Signed-off-by: Jin Xu <jinuxstyle@gmail.com>
[Jaegeuk Kim: suggested simpler solution]
Reviewed-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Previously, we experience bio traces as follows when running simple sequential
write test.
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500104928, size = 4K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 499922208, size = 368K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 499914752, size = 140K
-> total 512K
The first one is to write an indirect node block, and the others are to write
direct node blocks.
The reason why there are two separate bios for direct node blocks is:
0. initial state
------------------ ------------------
| | |xxxxxxxx |
------------------ ------------------
1. write 368K
------------------ ------------------
| | |xxxxxxxxWWWWWWWW|
------------------ ------------------
2. write 140K
------------------ ------------------
|WWWWWWW | |xxxxxxxxWWWWWWWW|
------------------ ------------------
This is because f2fs_write_node_pages tries to write just 512K totally, so that
we can lose the chance to merge more bios nicely.
After this patch is applied, we can get the following bio traces.
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500103168, size = 8K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500111368, size = 4K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500107272, size = 512K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500108296, size = 512K
f2fs_do_submit_bio: type = NODE, io = no sync, sector = 500109320, size = 500K
And finally, we can improve the sequential write performance,
from 458.775 MB/s to 479.945 MB/s on SSD.
Signed-off-by: Jaegeuk Kim <jaegeuk.kim@samsung.com>
Pull PTR_RET() removal patches from Rusty Russell:
"PTR_RET() is a weird name, and led to some confusing usage. We ended
up with PTR_ERR_OR_ZERO(), and replacing or fixing all the usages.
This has been sitting in linux-next for a whole cycle"
[ There are still some PTR_RET users scattered about, with some of them
possibly being new, but most of them existing in Rusty's tree too. We
have that
#define PTR_RET(p) PTR_ERR_OR_ZERO(p)
thing in <linux/err.h>, so they continue to work for now - Linus ]
* tag 'PTR_RET-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux:
GFS2: Replace PTR_RET with PTR_ERR_OR_ZERO
Btrfs: volume: Replace PTR_RET with PTR_ERR_OR_ZERO
drm/cma: Replace PTR_RET with PTR_ERR_OR_ZERO
sh_veu: Replace PTR_RET with PTR_ERR_OR_ZERO
dma-buf: Replace PTR_RET with PTR_ERR_OR_ZERO
drivers/rtc: Replace PTR_RET with PTR_ERR_OR_ZERO
mm/oom_kill: remove weird use of ERR_PTR()/PTR_ERR().
staging/zcache: don't use PTR_RET().
remoteproc: don't use PTR_RET().
pinctrl: don't use PTR_RET().
acpi: Replace weird use of PTR_RET.
s390: Replace weird use of PTR_RET.
PTR_RET is now PTR_ERR_OR_ZERO(): Replace most.
PTR_RET is now PTR_ERR_OR_ZERO
Linus Torvalds
Anand Avati
Miklos Szeredi
Miklos Szeredi
Alexey Dobriyan
Jin Xu
Jaegeuk Kim