Have caller pass in a safely-obtained reference to the parent directory
for calculating a dentry's hash valud.
While we're here, simpify the flow through ceph_encode_fh() so that there
is a single exit point and cleanup.
Also fix a bug with the dentry hash calculation: calculate the hash for the
dentry we were given, not its parent.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
Protect d_parent with d_lock. Carry a reference. Simplify the flow so
that there is a single exit point and cleanup.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
d_parent is protected by d_lock: use it when looking up a dentry's parent
directory inode. Also take a reference and drop it in the caller to avoid
a use-after-free.
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
The ->lookup() and prepopulate_readdir() callers are working with unhashed
dentries, so we don't have to worry. The export.c callers, though, need
to initialize something they got back from d_obtain_alias() and are
potentially racing with other callers. Make sure we don't return unless
the dentry is properly initialized (by us or someone else).
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
Curretly ceph_add_cap clears the complete bit if we are newly issued the
FILE_SHARED cap, which is normally the case for a newly issue cap on a new
directory. That means we clear the just-set bit. Move the check that sets
the flag to after the cap is added/updated.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
This should improve the default read performance, as without it
readahead is practically disabled.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
The lease mask is no longer used (and it changed a while back). Instead,
use a non-zero duration to indicate that there is a lease being issued.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We weren't properly calling lookup_instantiate_filp when setting up the
lookup intent, which could lead to file leakage on errors. So:
- use separate helper for the hidden snapdir translation, immediately
following the mds request
- use ceph_finish_lookup for the final dentry/return value dance in the
exit path
- lookup_instantiate_filp on success
Reported-by: Al Viro <viro@ZenIV.linux.org.uk>
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We only need to put these on the directory unsafe list if they have
side effects that fsync(2) should flush out.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We were always getting NULL here because the intent file f_dentry is always
NULL at this point, which means we were always passing NULL to
ceph_mdsc_do_request. In reality, this was fine, since this isn't
currently ever a write operation that needs to get strung on the dir's
unsafe list.
Use the dir explicitly, and only pass it if this open has side-effects that
a dir fsync should flush.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
The generic_file_aio_write call may block on balance_dirty_pages while we
flush data to the OSDs. If we hold a reference to the FILE_WR cap during
that interval revocation by the MDS (e.g., to do a stat(2)) may be very
slow.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We used to go into this branch if i_wrbuffer_ref_head was non-zero. This
was an ancient check from before we were careful about dealing with all
kinds of caps (and not just dirty pages). It is cleaner to only queue a
capsnap if there is an actual dirty cap. If we are racing with...
something...we will end up here with ci->i_wrbuffer_refs but no dirty
caps.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
There are two problems that come up when we try to queue a capsnap while a
write is in progress:
- The FILE_WR cap is held, but not yet dirty, so we may queue a capsnap
with dirty == 0. That will crash later in __ceph_flush_snaps(). Or
on the FILE_WR cap if a write is in progress.
- We may not have i_head_snapc set, which causes problems pretty quickly.
Look to the snaprealm in this case.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
This allows us to force IO through the sync path which you normally only
get when multiple clients are reading/writing to the same file or by
mounting with -o sync. Among other things, this lets test programs verify
correctness with a single mount.
Reviewed-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
Btrfs needs to be able to control how filemap_write_and_wait_range() is called
in fsync to make it less of a painful operation, so push down taking i_mutex and
the calling of filemap_write_and_wait() down into the ->fsync() handlers. Some
file systems can drop taking the i_mutex altogether it seems, like ext3 and
ocfs2. For correctness sake I just pushed everything down in all cases to make
sure that we keep the current behavior the same for everybody, and then each
individual fs maintainer can make up their mind about what to do from there.
Thanks,
Acked-by: Jan Kara <jack@suse.cz>
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
This converts everybody to handle SEEK_HOLE/SEEK_DATA properly. In some cases
we just return -EINVAL, in others we do the normal generic thing, and in others
we're simply making sure that the properly due-dilligence is done. For example
in NFS/CIFS we need to make sure the file size is update properly for the
SEEK_HOLE and SEEK_DATA case, but since it calls the generic llseek stuff itself
that is all we have to do. Thanks,
Signed-off-by: Josef Bacik <josef@redhat.com>
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
its value depends only on inode and does not change; we might as
well store it in ->i_op->check_acl and be done with that.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We were iterating across stripe boundaries properly, but not moving the
write buffer pointer forward. This caused us to rewrite the same data
after the break. Fix by adjusting the data pointer forward, and
recalculating the io and buffer alignment after the break.
Signed-off-by: Sage Weil <sage@newdream.net>
If we request a lock and then abort (e.g., ^C), we need to send a matching
unlock request to the MDS to unwind our lock attempt to avoid indefinitely
blocking other clients.
Reported-by: Brian Chrisman <brchrisman@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
Getting ENOENT is equivalent to reading 0 bytes. Make that correction
before setting up the hit_stripe and was_short flags.
Fixes the following case:
dd if=/dev/zero of=/mnt/fs_depot/dd3 bs=1 seek=1048576 count=0
dd if=/mnt/fs_depot/dd3 of=/root/ddout1 skip=8 bs=500 count=2 iflag=direct
Reported-by: Henry C Chang <henry.cy.chang@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
If we get a short read from the OSD because the object is small, we need to
zero the remainder of the buffer. For O_DIRECT reads, the attempted range
is not trimmed to i_size by the VFS, so we were actually looping
indefinitely.
Fix by trimming by i_size, and the unconditionally zeroing the trailing
range.
Reported-by: Jeff Wu <cpwu@tnsoft.com.cn>
Signed-off-by: Sage Weil <sage@newdream.net>
We should use ihold whenever we already have a stable inode ref, even
when we aren't holding i_lock. This avoids adding new and unnecessary
locking dependencies.
Signed-off-by: Sage Weil <sage@newdream.net>
In e9964c10 we change cap flushing to do a delicate dance because some
inodes on the cap_dirty list could be in a migrating state (got EXPORT but
not IMPORT) in which we couldn't actually flush and move from
dirty->flushing, breaking the while (!empty) { process first } loop
structure. It worked for a single sync thread, but was not reentrant and
triggered infinite loops when multiple syncers came along.
Instead, move inodes with dirty to a separate cap_dirty_migrating list
when in the limbo export-but-no-import state, allowing us to go back to
the simple loop structure (which was reentrant). This is cleaner and more
robust.
Audited the cap_dirty users and this looks fine:
list_empty(&ci->i_dirty_item) is still a reliable indicator of whether we
have dirty caps (which list we're on is irrelevant) and list_del_init()
calls still do the right thing.
Signed-off-by: Sage Weil <sage@newdream.net>
Since we pass the nofail arg, we should never get an error; BUG if we do.
(And fix the function to not return an error if __map_request fails.)
Signed-off-by: Sage Weil <sage@newdream.net>
Both off and fi->offset are unsigned, so the difference is always >= 0.
Compare them directly instead of the sign of the difference.
Signed-off-by: Sage Weil <sage@newdream.net>
If we grab new_cap, retake the lock, and find we already have a cap now
for the given mds, release new_cap.
Signed-off-by: Sage Weil <sage@newdream.net>
We put ourselves on an inode list for the parent directory of metadata
operations so that an fsync on the directory will wait for metadata updates
to commit to disk. We weren't holding a reference to that directory,
however, and under certain workloads (fsstress in this case) the directory
can go away.
Signed-off-by: Sage Weil <sage@newdream.net>
We increments i_wrbuffer_ref when taking the Fb cap. This breaks
the dirty page accounting and causes looping in
__ceph_do_pending_vmtruncate, and ceph client hangs.
This bug can be reproduced occasionally by running blogbench.
Add a new field i_wb_ref to inode and dedicate it to Fb reference
counting.
Signed-off-by: Henry C Chang <henry.cy.chang@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
The mds session, s, could be freed during ceph_put_mds_session.
Move dout before ceph_put_mds_session.
Signed-off-by: Henry C Chang <henry.cy.chang@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
The __mark_dirty_inode helper now takes i_lock as of 250df6ed. Fix the
one ceph callers that held i_lock (__ceph_mark_dirty_caps) to return the
flags value so that the callers can do it outside of i_lock.
Signed-off-by: Sage Weil <sage@newdream.net>
We should unlock the page and return -ENOMEM if ceph_osdc_new_request
failed.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
* git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
libceph: Create a new key type "ceph".
libceph: Get secret from the kernel keys api when mounting with key=NAME.
ceph: Move secret key parsing earlier.
libceph: fix null dereference when unregistering linger requests
ceph: unlock on error in ceph_osdc_start_request()
ceph: fix possible NULL pointer dereference
ceph: flush msgr_wq during mds_client shutdown
This makes the base64 logic be contained in mount option parsing,
and prepares us for replacing the homebew key management with the
kernel key retention service.
Signed-off-by: Tommi Virtanen <tommi.virtanen@dreamhost.com>
Signed-off-by: Sage Weil <sage@newdream.net>
Fix the incorrect use of igrab() inside the i_lock in NFS and Ceph‥
If we are already holding the i_lock, we have a reference to the
inode so we can safely use ihold() to gain an extra reference. This
avoids hangs due to lock recursion on the i_lock now that the
inode_lock is gone and igrab() uses the i_lock itself.
Signed-off-by: Dave Chinner <dchinner@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: linux-fsdevel@vger.kernel.org
Cc: Ryan Mallon <ryan@bluewatersys.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
The release method for mds connections uses a backpointer to the
mds_client, so we need to flush the workqueue of any pending work (and
ceph_connection references) prior to freeing the mds_client. This fixes
an oops easily triggered under UML by
while true ; do mount ... ; umount ... ; done
Also fix an outdated comment: the flush in ceph_destroy_client only flushes
OSD connections out. This bug is basically an artifact of the ceph ->
ceph+libceph conversion.
Signed-off-by: Sage Weil <sage@newdream.net>
In sync_write_wait(), we assume that the newest request is at the
tail of unsafe write list. We should maintain the semantics here.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
The ino32 mount option forces the ceph fs to report 32 bit
ino values. This is useful for 64 bit kernels with 32 bit userspace.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
d_move puts the renamed dentry at the end of d_subdirs, screwing with our
cached dentry directory offsets. We were just clearing I_COMPLETE to avoid
any possibility of trouble. However, assigning the renamed dentry an
offset at the end of the directory (to match it's new d_subdirs position)
is sufficient to maintain correct behavior and hold onto I_COMPLETE.
This is especially important for workloads like rsync, which renames files
into place. Before, we would lose I_COMPLETE and do MDS lookups for each
file. With this patch we only talk to the MDS on create and rename.
Signed-off-by: Sage Weil <sage@newdream.net>
First, this was racy anyway: d_release isn't called until well after the
dentry is unhashed. Second, this runs afoul of the recent dcache change
that clears d_parent prior to calling d_release (949854d0), causing a NULL
pointer dereference.
Signed-off-by: Sage Weil <sage@newdream.net>
This reverts commit 97d79b403e.
This fails to account for d_parent changes due to rename or disconnected
dentries due to submounts or NFS reexports.
Signed-off-by: Sage Weil <sage@newdream.net>
When creating a new dentry we now hold a reference to the parent
inode in the ceph_dentry. This is required due to the new RCU
changes from 949854d0, which set dentry->d_parent to NULL in d_kill before
calling the ->release() callback. If/when that behavior is changed, we can
revert this hack.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
We were forming a dirty list, and then queueing cap_snaps for each realm
_and_ its children, regardless of whether the children were already in the
dirty list. This meant we did it twice for some realms. Which in turn
meant we corrupted mdsc->snap_flush_list when the cap_snap was re-added to
the list it was already on, and could trigger an infinite loop.
We were also using recursion to do reach all the children, a no-no when
stack is limited.
Instead, (re)queue any children on the dirty list, avoiding processing
anything twice and avoiding any recursion.
Signed-off-by: Sage Weil <sage@newdream.net>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
ceph: avoid picking MDS that is not active
ceph: avoid immediate cap check after import
ceph: fix flushing of caps vs cap import
ceph: fix erroneous cap flush to non-auth mds
ceph: fix cap_wanted_delay_{min,max} mount option initialization
ceph: fix xattr rbtree search
ceph: fix getattr on directory when using norbytes
Ignore replication or auth frag data if it indicates an MDS that is not
active. This can happen if the MDS shuts down and the client has stale
data about the namespace distribution across the MDS cluster. If that's
the case, fall back to directing the request based on the auth cap (which
should always be accurate).
Signed-off-by: Sage Weil <sage@newdream.net>
The NODELAY flag avoids the heuristics that delay cap (issued/wanted)
release. There's no reason for that after we import a cap, and it kills
whatever benefit we get from those delays.
Signed-off-by: Sage Weil <sage@newdream.net>
If we are mid-flush and a cap is migrated to another node, we need to
resend the cap flush message to the new MDS, and do so with the original
flush_seq to avoid leaking across a sync boundary. Previously we didn't
redo the flush (we only flushed newly dirty data), which would cause a
later sync to hang forever.
Signed-off-by: Sage Weil <sage@newdream.net>
The int flushing is global and not clear on each iteration of the loop,
which can cause a second flush of caps to any MDSs with ids greater than
the auth.
Signed-off-by: Sage Weil <sage@newdream.net>
Fix xattr name comparison in rbtree search for strings that share a prefix.
The *name argument is null terminated, but the xattr name is not, so we
need to use strncmp, but that means adjusting for the case where name is
a prefix of xattr->name.
The corresponding case in __set_xattr() already handles this properly
(although in that case *name is also not null terminated).
Reported-by: Sergiy Kibrik <sakib@meta.ua>
Signed-off-by: Sage Weil <sage@newdream.net>
The norbytes mount option was broken, and when doing getattr
on a directory it return the rbytes instead of the number of
entities. This commit fixes it.
Signed-off-by: Yehuda Sadeh <yehuda@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client:
rbd: fix cleanup when trying to mount inexistent image
net/ceph: make ceph_msgr_wq non-reentrant
ceph: fsc->*_wq's aren't used in memory reclaim path
ceph: Always free allocated memory in osdmap_decode()
ceph: Makefile: Remove unnessary code
ceph: associate requests with opening sessions
ceph: drop redundant r_mds field
ceph: implement DIRLAYOUTHASH feature to get dir layout from MDS
ceph: add dir_layout to inode
Remove the if and else conditional because the code is in mainline and there
is no need in it being there.
Also, Changed Makefile to use <modules>-y instead of <modules>-objs
because -objs is deprecated and not mentioned in
Documentation/kbuild/makefiles.txt.
Signed-off-by: Tracey Dent <tdent48227@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
Associate request with sessions that aren't yep open. This makes the
debugfs mdsc request list more informative.
Signed-off-by: Sage Weil <sage@newdream.net>
The r_mds field is redundant, since we can find the same information at
r_session->s_mds, and when r_session is NULL then r_mds is meaningless.
Signed-off-by: Sage Weil <sage@newdream.net>
This implements the DIRLAYOUTHASH protocol feature, which passes the dir
layout over the wire from the MDS. This gives the client knowledge
of the correct hash function to use for mapping dentries among dir
fragments.
Note that if this feature is _not_ present on the client but is on the
MDS, the client may misdirect requests. This will result in a forward
and degrade performance. It may also result in inaccurate NFS filehandle
generation, which will prevent fh resolution when the inode is not present
in the client cache and the parent directories have been fragmented.
Signed-off-by: Sage Weil <sage@newdream.net>
Add a ceph_dir_layout to the inode, and calculate dentry hash values based
on the parent directory's specified dir_hash function. This is needed
because the old default Linux dcache hash function is extremely week and
leads to a poor distribution of files among dir fragments.
Signed-off-by: Sage Weil <sage@newdream.net>
Require filesystems be aware of .d_revalidate being called in rcu-walk
mode (nd->flags & LOOKUP_RCU). For now do a simple push down, returning
-ECHILD from all implementations.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Reduce some branches and memory accesses in dcache lookup by adding dentry
flags to indicate common d_ops are set, rather than having to check them.
This saves a pointer memory access (dentry->d_op) in common path lookup
situations, and saves another pointer load and branch in cases where we
have d_op but not the particular operation.
Patched with:
git grep -E '[.>]([[:space:]])*d_op([[:space:]])*=' | xargs sed -e 's/\([^\t ]*\)->d_op = \(.*\);/d_set_d_op(\1, \2);/' -e 's/\([^\t ]*\)\.d_op = \(.*\);/d_set_d_op(\&\1, \2);/' -i
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
RCU free the struct inode. This will allow:
- Subsequent store-free path walking patch. The inode must be consulted for
permissions when walking, so an RCU inode reference is a must.
- sb_inode_list_lock to be moved inside i_lock because sb list walkers who want
to take i_lock no longer need to take sb_inode_list_lock to walk the list in
the first place. This will simplify and optimize locking.
- Could remove some nested trylock loops in dcache code
- Could potentially simplify things a bit in VM land. Do not need to take the
page lock to follow page->mapping.
The downsides of this is the performance cost of using RCU. In a simple
creat/unlink microbenchmark, performance drops by about 10% due to inability to
reuse cache-hot slab objects. As iterations increase and RCU freeing starts
kicking over, this increases to about 20%.
In cases where inode lifetimes are longer (ie. many inodes may be allocated
during the average life span of a single inode), a lot of this cache reuse is
not applicable, so the regression caused by this patch is smaller.
The cache-hot regression could largely be avoided by using SLAB_DESTROY_BY_RCU,
however this adds some complexity to list walking and store-free path walking,
so I prefer to implement this at a later date, if it is shown to be a win in
real situations. I haven't found a regression in any non-micro benchmark so I
doubt it will be a problem.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Protect d_subdirs and d_child with d_lock, except in filesystems that aren't
using dcache_lock for these anyway (eg. using i_mutex).
Note: if we change the locking rule in future so that ->d_child protection is
provided only with ->d_parent->d_lock, it may allow us to reduce some locking.
But it would be an exception to an otherwise regular locking scheme, so we'd
have to see some good results. Probably not worthwhile.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Protect d_unhashed(dentry) condition with d_lock. This means keeping
DCACHE_UNHASHED bit in synch with hash manipulations.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
Make d_count non-atomic and protect it with d_lock. This allows us to ensure a
0 refcount dentry remains 0 without dcache_lock. It is also fairly natural when
we start protecting many other dentry members with d_lock.
Signed-off-by: Nick Piggin <npiggin@kernel.dk>
For read operation, we have to set the argument _write_ of get_user_pages
to 1 since we will write data to pages. Also, we need to SetPageDirty before
releasing these pages.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
The fh_to_dentry etc. methods use ceph_init_dentry(), which assumes that
d_parent is defined. It isn't for those callers, so check!
Signed-off-by: Sage Weil <sage@newdream.net>
The user buffer may be 512-byte aligned, not page-aligned. We were
assuming the buffer was page-aligned and only accounting for
non-page-aligned io offsets.
Signed-off-by: Henry C Chang <henry_c_chang@tcloudcomputing.com>
Signed-off-by: Sage Weil <sage@newdream.net>
Fill in the local lock with response data if appropriate,
and don't call posix_lock_file when reading locks.
Signed-off-by: Herb Shiu <herb_shiu@tcloudcomputing.com>
Acked-by: Greg Farnum <gregf@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
Previously the kernel client incorrectly assumed everything was a directory.
Signed-off-by: Herb Shiu <herb_shiu@tcloudcomputing.com>
Acked-by: Greg Farnum <gregf@hq.newdream.net>
Signed-off-by: Sage Weil <sage@newdream.net>
last may be NULL, but we dereference it in the else branch without
checking. Normally it doesn't trigger because last == NULL when fpos == 2,
but it could happen on a newly opened dir if the user seeks forward.
Reported-by: Dan Carpenter <error27@gmail.com>
Signed-off-by: Sage Weil <sage@newdream.net>
