Some messages sent by the MDS entail a session sequence number
increment, and the MDS will drop certain types of requests on the floor
when the sequence numbers don't match.
In particular, a REQUEST_CLOSE message can cross with one of the
sequence morphing messages from the MDS which can cause the client to
stall, waiting for a response that will never come.
Originally, this meant an up to 5s delay before the recurring workqueue
job kicked in and resent the request, but a recent change made it so
that the client would never resend, causing a 60s stall unmounting and
sometimes a blockisting event.
Add a new helper for incrementing the session sequence and then testing
to see whether a REQUEST_CLOSE needs to be resent, and move the handling
of CEPH_MDS_SESSION_CLOSING into that function. Change all of the
bare sequence counter increments to use the new helper.
Reorganize check_session_state with a switch statement. It should no
longer be called when the session is CLOSING, so throw a warning if it
ever is (but still handle that case sanely).
[ idryomov: whitespace, pr_err() call fixup ]
URL: https://tracker.ceph.com/issues/47563
Fixes: fa99677342 ("ceph: fix potential mdsc use-after-free crash")
Reported-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Push the allocation of the msg and the send into the caller. Rename
the function to encode_cap_msg and make it void return.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
In client for each inode, it may have many opened files and may
have been pinned in more than one MDS servers. And some inodes
are idle, which have no any opened files.
This patch will show these metrics in the debugfs, likes:
item total
-----------------------------------------
opened files / total inodes 14 / 5
pinned i_caps / total inodes 7 / 5
opened inodes / total inodes 3 / 5
Will send these metrics to ceph, which will be used by the `fs top`,
later.
[ jlayton: drop unrelated hunk, count hashed inodes instead of
allocated ones ]
URL: https://tracker.ceph.com/issues/47005
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This will help simplify the code.
[ jlayton: fix minor merge conflict in quota.c ]
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Tuan and Ulrich mentioned that they were hitting a problem on s390x,
which has a 32-bit ino_t value, even though it's a 64-bit arch (for
historical reasons).
I think the current handling of inode numbers in the ceph driver is
wrong. It tries to use 32-bit inode numbers on 32-bit arches, but that's
actually not a problem. 32-bit arches can deal with 64-bit inode numbers
just fine when userland code is compiled with LFS support (the common
case these days).
What we really want to do is just use 64-bit numbers everywhere, unless
someone has mounted with the ino32 mount option. In that case, we want
to ensure that we hash the inode number down to something that will fit
in 32 bits before presenting the value to userland.
Add new helper functions that do this, and only do the conversion before
presenting these values to userland in getattr and readdir.
The inode table hashvalue is changed to just cast the inode number to
unsigned long, as low-order bits are the most likely to vary anyway.
While it's not strictly required, we do want to put something in
inode->i_ino. Instead of basing it on BITS_PER_LONG, however, base it on
the size of the ino_t type.
NOTE: This is a user-visible change on 32-bit arches:
1/ inode numbers will be seen to have changed between kernel versions.
32-bit arches will see large inode numbers now instead of the hashed
ones they saw before.
2/ any really old software not built with LFS support may start failing
stat() calls with -EOVERFLOW on inode numbers >2^32. Nothing much we
can do about these, but hopefully the intersection of people running
such code on ceph will be very small.
The workaround for both problems is to mount with "-o ino32".
[ idryomov: changelog tweak ]
URL: https://tracker.ceph.com/issues/46828
Reported-by: Ulrich Weigand <Ulrich.Weigand@de.ibm.com>
Reported-and-Tested-by: Tuan Hoang1 <Tuan.Hoang1@ibm.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Make this loop look a bit more sane. Also optimize away the spinlock
release/reacquire if we can't get an inode reference.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This will help to reduce using the global mdsc->mutex lock in many
places.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
It make no sense to check the caps when reconnecting to mds. And
for the async dirop caps, they will be put by its _cb() function,
so when releasing the requests, it will make no sense too.
URL: https://tracker.ceph.com/issues/45635
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The mdsc->cap_dirty_lock is not held while walking the list in
ceph_kick_flushing_caps, which is not safe.
ceph_early_kick_flushing_caps does something similar, but the
s_mutex is held while it's called and I think that guards against
changes to the list.
Ensure we hold the s_mutex when calling ceph_kick_flushing_caps,
and add some clarifying comments.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When flushing a lot of caps to the MDS's at once (e.g. for syncfs),
we can end up waiting a substantial amount of time for MDS replies, due
to the fact that it may delay some of them so that it can batch them up
together in a single journal transaction. This can lead to stalls when
calling sync or syncfs.
What we'd really like to do is request expedited service on the _last_
cap we're flushing back to the server. If the CHECK_CAPS_FLUSH flag is
set on the request and the current inode was the last one on the
session->s_cap_dirty list, then mark the request with
CEPH_CLIENT_CAPS_SYNC.
Note that this heuristic is not perfect. New inodes can race onto the
list after we've started flushing, but it does seem to fix some common
use cases.
URL: https://tracker.ceph.com/issues/44744
Reported-by: Jan Fajerski <jfajerski@suse.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This is a per-sb list now, but that makes it difficult to tell when
the cap is the last dirty one associated with the session. Switch
this to be a per-session list, but continue using the
mdsc->cap_dirty_lock to protect the lists.
This list is only ever walked in ceph_flush_dirty_caps, so change that
to walk the sessions array and then flush the caps for inodes on each
session's list.
If the auth cap ever changes while the inode has dirty caps, then
move the inode to the appropriate session for the new auth_cap. Also,
ensure that we never remove an auth cap while the inode is still on the
s_cap_dirty list.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
write can stuck at waiting for larger max_size in following sequence of
events:
- client opens a file and writes to position 'A' (larger than unit of
max size increment)
- client closes the file handle and updates wanted caps (not wanting
file write caps)
- client opens and truncates the file, writes to position 'A' again.
At the 1st event, client set inode's requested_max_size to 'A'. At the
2nd event, mds removes client's writable range, but client does not reset
requested_max_size. At the 3rd event, client does not request max size
because requested_max_size is already larger than 'A'.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Nothing ensures that session will still be valid by the time we
dereference the pointer. Take and put a reference.
In principle, we should always be able to get a reference here, but
throw a warning if that's ever not the case.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Just take it before calling it. This means we have to do a couple of
minor in-memory operations under the spinlock now, but those shouldn't
be an issue.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
There's no reason to do this here. Just have the caller handle it.
Also, add a lockdep assertion.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This function takes a mdsc argument or ci argument, but if both are
passed in, it ignores the ci arg. Fortunately, nothing does that, but
there's no good reason to have the same function handle both cases.
Also, get rid of some branches and just use |= to set the wake_* vals.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Get rid of the __releases annotation by breaking it up into two
functions: __prep_cap which is done under the spinlock and __send_cap
that is done outside it. Add new fields to cap_msg_args for the wake
boolean and old_xattr_buf pointer.
Nothing checks the return value from __send_cap, so make it void
return.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Count hits and misses in the caps cache. If the client has all of
the necessary caps when a task needs references, then it's counted
as a hit. Any other situation is a miss.
URL: https://tracker.ceph.com/issues/43215
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
It's possible for the VFS to completely forget about an inode, but for
it to still be sitting on the cap release queue. If the MDS sends the
client a cap message for such an inode, it just ignores it today, which
can lead to a stall of up to 5s until the cap release queue is flushed.
If we get a cap message for an inode that can't be located, then go
ahead and flush the cap release queue.
Cc: stable@vger.kernel.org
URL: https://tracker.ceph.com/issues/45532
Fixes: 1e9c2eb681 ("ceph: delete stale dentry when last reference is dropped")
Reported-and-Tested-by: Andrej Filipčič <andrej.filipcic@ijs.si>
Suggested-by: Yan, Zheng <zyan@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If the ceph_mdsc_open_export_target_session() return fails, it will
do a "goto retry", but the session mutex has already been unlocked.
Re-lock the mutex in that case to ensure that we don't unlock it
twice.
Signed-off-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
There are 3 speical error codes: -EAGAIN/-EFBIG/-ESTALE.
After calling try_get_cap_refs, ceph_try_get_caps test for the
-EAGAIN twice. Ensure that it tests for -ESTALE instead.
Signed-off-by: Wu Bo <wubo40@huawei.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
ceph_check_caps() can't request new max size for async creating inode.
This may make ceph_get_caps() loop busily until getting reply of the
async create. Also, wait for async creating reply before calling
ceph_renew_caps().
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
1. try_get_cap_refs() fails to get caps and finds that mds_wanted
does not include what it wants. It returns -ESTALE.
2. ceph_get_caps() calls ceph_renew_caps(). ceph_renew_caps() finds
that inode has cap, so it calls ceph_check_caps().
3. ceph_check_caps() finds that issued caps (without checking if it's
stale) already includes caps wanted by open file, so it skips
updating wanted caps.
Above events can cause an infinite loop inside ceph_get_caps().
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When there is no auth cap, check_max_size() can't do anything and may
cause an infinite loop inside ceph_get_caps().
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Returns 0 if caps were not able to be acquired (yet), 1 if cap
acquisition succeeded, or a negative error code. There are 3 special
error codes:
-EAGAIN: need to sleep but non-blocking is specified
-EFBIG: ask caller to call check_max_size() and try again.
-ESTALE: ask caller to call ceph_renew_caps() and try again.
[ jlayton: add WARN_ON_ONCE check for -EAGAIN ]
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If an inode has caps from multiple mds's, the following can happen:
- non-auth mds revokes Fsc. Fcb is used, so page writeback is queued.
- when writeback finishes, ceph_check_caps() is called with auth only
flag. ceph_check_caps() invalidates pagecache, but skips checking any
non-auth caps.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Originally, calling ceph_get_fmode() for open files is by thread that
handles request reply. There is a small window between updating caps and
and waking the request initiator. We need to prevent ceph_check_caps()
from releasing wanted caps in the window.
Previous patches made fill_inode() call __ceph_touch_fmode() for open file
requests. This prevented ceph_check_caps() from releasing wanted caps for
'caps_wanted_delay_min' seconds, enough for request initiator to get
woken up and call ceph_get_fmode().
This allows us to now call ceph_get_fmode() in ceph_open() instead.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
__ceph_caps_file_wanted() already checks 'caps_wanted_delay_min' and
'caps_wanted_delay_max'. There is no need to duplicate the logic in
ceph_check_caps() and __send_cap()
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Add i_last_rd and i_last_wr to ceph_inode_info. These fields are
used to track the last time the client acquired read/write caps for
the inode.
If there is no read/write on an inode for 'caps_wanted_delay_max'
seconds, __ceph_caps_file_wanted() does not request caps for read/write
even there are open files.
Call __ceph_touch_fmode() for dir operations. __ceph_caps_file_wanted()
calculates dir's wanted caps according to last dir read/modification. If
there is recent dir read, dir inode wants CEPH_CAP_ANY_SHARED caps. If
there is recent dir modification, also wants CEPH_CAP_FILE_EXCL.
Readdir is a special case. Dir inode wants CEPH_CAP_FILE_EXCL after
readdir, as with that, modifications do not need to release
CEPH_CAP_FILE_SHARED or invalidate all dentry leases issued by readdir.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Original code only renews caps for inodes with CEPH_I_CAP_DROPPED flag,
which indicates that mds has closed the session and caps were dropped.
Remove this flag in preparation for not requesting caps for idle open
files.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If a create is done, then typically we'll end up writing to the file
soon afterward. We don't want to wait for the reply before doing that
when doing an async create, so that means we need the layout for the
new file before we've gotten the response from the MDS.
All files created in a directory will initially inherit the same layout,
so copy off the requisite info from the first synchronous create in the
directory, and save it in a new i_cached_layout field. Zero out the
layout when we lose Dc caps in the dir.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If we don't have all of the cap bits for the want mask in
try_get_cap_refs, then just take refs on the need bits.
Signed-off-by: "Yan, Zheng" <ukernel@gmail.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Track and correctly handle directory caps for asynchronous operations.
Add aliases for Frc caps that we now designate at Dcu caps (when dealing
with directories).
Unlike file caps, we don't reclaim these when the session goes away, and
instead preemptively release them. In-flight async dirops are instead
handled during reconnect phase. The client needs to re-do a synchronous
operation in order to re-get directory caps.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Rename it to ceph_take_cap_refs and make it available to other files.
Also replace a comment with a lockdep assertion.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When we issue an async create, we must ensure that any later on-the-wire
requests involving it wait for the create reply.
Expand i_ceph_flags to be an unsigned long, and add a new bit that
MDS requests can wait on. If the bit is set in the inode when sending
caps, then don't send it and just return that it has been delayed.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The last thing that this function does is release i_ceph_lock, so
have the caller do that instead. Add a lockdep assertion to
ensure that the function is always called with i_ceph_lock held.
Change the prototype to take a ceph_inode_info pointer and drop
the separate mdsc argument as we can get that from the session.
While at it, make it non-static. We'll need this to kick any
flushing caps once the create reply comes in.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
These bits will have new meaning for directory inodes.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
In future patches we'll be taking and relying on Fx caps. Add proper
refcounting for them.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
For the caps having no any subset mask requested we shouldn't touch
them.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
__ceph_is_any_caps is a duplicate helper.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Add some visibility of tasks that are waiting for caps to the "caps"
debugfs file. Display the tgid of the waiting task, inode number, and
the caps the task needs and wants.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
KASAN reports a use-after-free when running xfstest generic/531, with the
following trace:
[ 293.903362] kasan_report+0xe/0x20
[ 293.903365] rb_erase+0x1f/0x790
[ 293.903370] __ceph_remove_cap+0x201/0x370
[ 293.903375] __ceph_remove_caps+0x4b/0x70
[ 293.903380] ceph_evict_inode+0x4e/0x360
[ 293.903386] evict+0x169/0x290
[ 293.903390] __dentry_kill+0x16f/0x250
[ 293.903394] dput+0x1c6/0x440
[ 293.903398] __fput+0x184/0x330
[ 293.903404] task_work_run+0xb9/0xe0
[ 293.903410] exit_to_usermode_loop+0xd3/0xe0
[ 293.903413] do_syscall_64+0x1a0/0x1c0
[ 293.903417] entry_SYSCALL_64_after_hwframe+0x44/0xa9
This happens because __ceph_remove_cap() may queue a cap release
(__ceph_queue_cap_release) which can be scheduled before that cap is
removed from the inode list with
rb_erase(&cap->ci_node, &ci->i_caps);
And, when this finally happens, the use-after-free will occur.
This can be fixed by removing the cap from the inode list before being
removed from the session list, and thus eliminating the risk of an UAF.
Cc: stable@vger.kernel.org
Signed-off-by: Luis Henriques <lhenriques@suse.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
It doesn't do anything to invalidate the cache when dropping RD caps.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
cap->session is always non-NULL, so we can just do a single test for
equality w/o testing explicitly for a NULL pointer.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
