In preparation for msgr2, make the cluster send us maps with addrvecs
including both LEGACY and MSGR2 addrs instead of a single LEGACY addr.
This means advertising support for SERVER_NAUTILUS and also some older
features: SERVER_MIMIC, MONENC and MONNAMES.
MONNAMES and MONENC are actually pre-argonaut, we just never updated
ceph_monmap_decode() for them. Decoding is unconditional, see commit
23c625ce30 ("libceph: assume argonaut on the server side").
SERVER_MIMIC doesn't bear any meaning for the kernel client.
Since ceph_decode_entity_addrvec() is guarded by encoding version
checks (and in msgr2 case it is guarded implicitly by the fact that
server is speaking msgr2), we assume MSG_ADDR2 for it.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
- request service tickets together with auth ticket. Currently we get
auth ticket via CEPHX_GET_AUTH_SESSION_KEY op and then request service
tickets via CEPHX_GET_PRINCIPAL_SESSION_KEY op in a separate message.
Since nautilus, desired service tickets are shared togther with auth
ticket in CEPHX_GET_AUTH_SESSION_KEY reply.
- propagate session key and connection secret, if any. In preparation
for msgr2, update handle_reply() and verify_authorizer_reply() auth
ops to propagate session key and connection secret. Since nautilus,
if secure mode is negotiated, connection secret is shared either in
CEPHX_GET_AUTH_SESSION_KEY reply (for mons) or in a final authorizer
reply (for osds and mdses).
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When we added the btime feature in mainline ceph, we had to extend
struct ceph_mds_request_args so that it could be set. Implement the same
in the kernel client.
Rename ceph_mds_request_head with a _old extension, and a union
ceph_mds_request_args_ext to allow for the extended size of the new
header format.
Add the appropriate code to handle both formats in struct
create_request_message and key the behavior on whether the peer supports
CEPH_FEATURE_FS_BTIME.
The gid_list field in the payload is now populated from the saved
credential. For now, we don't add any support for setting the btime via
setattr, but this does enable us to add that in the future.
[ idryomov: break unnecessarily long lines ]
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
We can always get the mdsc from the session, so there's no need to pass
it in as a separate argument. Pass the session to __prepare_send_request
as well, to prepare for later patches that will need to access it.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Replace req->r_uid/r_gid with an r_cred pointer and take a reference to
that at the point where we previously would sample the two. Use that to
populate the uid and gid in the header and release the reference when
the request is freed.
This should enable us to later add support for sending supplementary
group lists in MDS requests.
[ idryomov: break unnecessarily long lines ]
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When setting the file/dir layout, it may need data pool info. So
in mds server, it needs to check the osdmap. At present, if mds
doesn't find the data pool specified, it will try to get the latest
osdmap. Now if pass the osd epoch for setxattr, the mds server can
only check this epoch of osdmap.
URL: https://tracker.ceph.com/issues/48504
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The variable i is being initialized with a value that is never read
and it is being updated later with a new value in a for-loop. The
initialization is redundant and can be removed.
Addresses-Coverity: ("Unused value")
Signed-off-by: Colin Ian King <colin.king@canonical.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Geng Jichao reported a rather complex deadlock involving several
moving parts:
1) readahead is issued against an inode and some of its pages are locked
while the read is in flight
2) the same inode is evicted from the cache, and this task gets stuck
waiting for the page lock because of the above readahead
3) another task is processing a reply trace, and looks up the inode
being evicted while holding the s_mutex. That ends up waiting for the
eviction to complete
4) a write reply for an unrelated inode is then processed in the
ceph_con_workfn job. It calls ceph_check_caps after putting wrbuffer
caps, and that gets stuck waiting on the s_mutex held by 3.
The reply to "1" is stuck behind the write reply in "4", so we deadlock
at that point.
This patch changes the trace processing to call ceph_get_inode outside
of the s_mutex and snap_rwsem, which should break the cycle above.
[ idryomov: break unnecessarily long lines ]
URL: https://tracker.ceph.com/issues/47998
Reported-by: Geng Jichao <gengjichao@jd.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Luis Henriques <lhenriques@suse.de>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Convert some decodes into unused variables into skips, and fix up some
non-kerneldoc comment headers to not start with "/**".
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Ilya noticed that the first access to a blacklisted mount would often
get back -EACCES, but then subsequent calls would be OK. The problem is
in __do_request. If the session is marked as REJECTED, a hard error is
returned instead of waiting for a new session to come into being.
When the session is REJECTED and the mount was done with
recover_session=clean, queue the request to the waiting_for_map queue,
which will be awoken after tearing down the old session. We can only
do this for sync requests though, so check for async ones first and
just let the callers redrive a sync request.
URL: https://tracker.ceph.com/issues/47385
Reported-by: Ilya Dryomov <idryomov@gmail.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
30 minutes is a long time to wait, and this makes it difficult to test
the feature by manually blocklisting clients. Remove the timeout
infrastructure and just allow the client to reconnect at will.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When recovering a session (a'la recover_session=clean), we want to do
all of the operations that we do on a forced umount, but changing the
mount state to SHUTDOWN is can cause queued MDS requests to fail when
the session comes back. Most of those can idle until the session is
recovered in this situation.
Reserve SHUTDOWN state for forced umount, and make a new RECOVER state
for the forced reconnect situation. Change several tests for equality with
SHUTDOWN to test for that or RECOVER.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
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>
error_string key in the metadata map of MClientSession message
is intended for humans, but unfortunately became part of the on-wire
format with the introduction of recover_session=clean mode in commit
131d7eb4fa ("ceph: auto reconnect after blacklisted").
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Since nautilus, MDS tracks dirfrags whose child inodes have caps in open
file table. When MDS recovers, it prefetches all of these dirfrags. This
avoids using backtrace to load inodes. But dirfrags prefetch may load
lots of useless inodes into cache, and make MDS run out of memory.
Recent MDS adds an option that disables dirfrags prefetch. When dirfrags
prefetch is disabled. Recovering MDS only prefetches corresponding dir
inodes. Including inodes' parent/d_name in cap reconnect message can
help MDS to load inodes into its cache.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Most session messages contain a feature mask, but the MDS will
routinely send a REJECT message with one that is zero-length.
Commit 0fa8263367 ("ceph: fix endianness bug when handling MDS
session feature bits") fixed the decoding of the feature mask,
but failed to account for the MDS sending a zero-length feature
mask. This causes REJECT message decoding to fail.
Skip trying to decode a feature mask if the word count is zero.
Cc: stable@vger.kernel.org
URL: https://tracker.ceph.com/issues/46823
Fixes: 0fa8263367 ("ceph: fix endianness bug when handling MDS session feature bits")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: Ilya Dryomov <idryomov@gmail.com>
Tested-by: Patrick Donnelly <pdonnell@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When doing some tests with multiple mds, we were seeing many mds
forwarding requests between them, causing clients to resend.
If the request is a modification operation and the mode is set to
USE_AUTH_MDS, then the auth mds should be selected to handle the
request. If auth mds for frag is already set, then it should be returned
directly without further processing.
The current logic is wrong because it only returns directly if
mode is USE_AUTH_MDS, but we want to do that for all modes. If we don't,
then when the frag's mds is not equal to cap session's mds, the request
will get sent to the wrong MDS needlessly.
Drop the mode check in this condition.
Signed-off-by: Yanhu Cao <gmayyyha@gmail.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If the ceph_mdsc_init() fails, it will free the mdsc already.
Reported-by: syzbot+b57f46d8d6ea51960b8c@syzkaller.appspotmail.com
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Send metric flags to the MDS, indicating what metrics the client
supports. Currently that consists of cap statistics, and read, write and
metadata latencies.
URL: https://tracker.ceph.com/issues/43435
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This will send the caps/read/write/metadata metrics to any available MDS
once per second, which will be the same as the userland client. It will
skip the MDS sessions which don't support the metric collection, as the
MDSs will close socket connections when they get an unknown type
message.
We can disable the metric sending via the disable_send_metrics module
parameter.
[ jlayton: fix up endianness bug in ceph_mdsc_send_metrics() ]
URL: https://tracker.ceph.com/issues/43215
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If the session is already in closed state, we should skip it.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Make sure the delayed work stopped before releasing the resources.
cancel_delayed_work_sync() will only guarantee that the work finishes
executing if the work is already in the ->worklist. That means after
the cancel_delayed_work_sync() returns, it will leave the work requeued
if it was rearmed at the end. That can lead to a use after free once the
work struct is freed.
Fix it by flushing the delayed work instead of trying to cancel it, and
ensure that the work doesn't rearm if the mdsc is stopping.
URL: https://tracker.ceph.com/issues/46293
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
...and let the errnos bubble up to the callers.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
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>
And remove the unsed mdsc parameter to simplify the code.
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>
send_mds_reconnect takes the s_mutex while the mdsc->mutex is already
held. That inverts the locking order documented in mds_client.h. Drop
the mdsc->mutex, acquire the s_mutex and then reacquire the mdsc->mutex
to prevent a deadlock.
URL: https://tracker.ceph.com/issues/45609
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
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>
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>
Add a new "r_ended" field to struct ceph_mds_request and use that to
maintain the average latency of MDS requests.
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>
For dentry leases, only count the hit/miss info triggered from the vfs
calls. For the cases like request reply handling and ceph_trim_dentries,
ignore them.
For now, these are only viewable using debugfs. Future patches will
allow the client to send the stats to the MDS.
The output looks like:
item total miss hit
-------------------------------------------------
d_lease 11 7 141
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>
Eduard reported a problem mounting cephfs on s390 arch. The feature
mask sent by the MDS is little-endian, so we need to convert it
before storing and testing against it.
Cc: stable@vger.kernel.org
Reported-and-Tested-by: Eduard Shishkin <edward6@linux.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>
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>
Add new request field to hold the delegated inode number. Encode that
into the message when it's set.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Starting in Octopus, the MDS will hand out caps that allow the client
to do asynchronous file creates under certain conditions. As part of
that, the MDS will delegate ranges of inode numbers to the client.
Add the infrastructure to decode these ranges, and stuff them into an
xarray for later consumption by the async creation code.
Because the xarray code currently only handles unsigned long indexes,
and those are 32-bits on 32-bit arches, we only enable the decoding when
running on a 64-bit arch.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
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>
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>
...and ensure that such requests are never queued. The MDS has need to
know that a request is asynchronous so add flags and proper
infrastructure for that.
Also, delegated inode numbers and directory caps are associated with the
session, so ensure that async requests are always transmitted on the
first attempt and are never queued to wait for session reestablishment.
If it does end up looking like we'll need to queue the request, then
have it return -EJUKEBOX so the caller can reattempt with a synchronous
request.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
req->r_timeout is only used during mounting, so this error will
be more accurate.
URL: https://tracker.ceph.com/issues/44215
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
On my machine (x86_64) this struct is 952 bytes, which gets rounded up
to 1024 by kmalloc. Move this to a dedicated slabcache, so we can
allocate them without the extra 72 bytes of overhead per.
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>
When the unsafe reply to a request comes in, the request is put on the
r_unsafe_dir inode's list. In future patches, we're going to need to
wait on requests that may not have gotten an unsafe reply yet.
Change __register_request to put the entry on the dir inode's list when
the pointer is set in the request, and don't check the
CEPH_MDS_R_GOT_UNSAFE flag when unregistering it.
The only place that uses this list today is fsync codepath, and with
the coming changes, we'll want to wait on all operations whether it has
gotten an unsafe reply or not.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
- a set of patches that fixes various corner cases in mount and umount
code (Xiubo Li). This has to do with choosing an MDS, distinguishing
between laggy and down MDSes and parsing the server path.
- inode initialization fixes (Jeff Layton). The one included here
mostly concerns things like open_by_handle() and there is another
one that will come through Al.
- copy_file_range() now uses the new copy-from2 op (Luis Henriques).
The existing copy-from op turned out to be infeasible for generic
filesystem use; we disable the copy offload if OSDs don't support
copy-from2.
- a patch to link "rbd" and "block" devices together in sysfs (Hannes
Reinecke)
And a smattering of cleanups from Xiubo, Jeff and Chengguang.
-----BEGIN PGP SIGNATURE-----
iQFHBAABCAAxFiEEydHwtzie9C7TfviiSn/eOAIR84sFAl47PUcTHGlkcnlvbW92
QGdtYWlsLmNvbQAKCRBKf944AhHzi6LoCACmVli5N6bgnBE4sTixi/jz6aCCbk32
ZPlKiSesHnOGkY6KXHJT58JYy0paITBRik5ypdz06J8aCOtWyPLbn3uCemF9CYn2
g6dId2Lf5vGFrgSm4YSiqp9a86IZmYSDG41LbJD/IJWFDWdMWqNPMDqji6yaIO5O
NJI5N0tk+VFXdV+JyjV9X/FnP1r1D2ReZzz21ZiqTJXSmE8YIkioLjkq36QTMMG7
Gm5qdlc1x2r4qfzA1g+OiWgRQCUMgkuYerFzus4mVbW4hrphsavH2DArbOwFmsXF
46hOq+1uGVVyZILLJfKNiktf1GExBF0icbSREJtmjUHbQvNR8BH0C+fV
=vvIc
-----END PGP SIGNATURE-----
Merge tag 'ceph-for-5.6-rc1' of https://github.com/ceph/ceph-client
Pull ceph fixes from Ilya Dryomov:
- a set of patches that fixes various corner cases in mount and umount
code (Xiubo Li). This has to do with choosing an MDS, distinguishing
between laggy and down MDSes and parsing the server path.
- inode initialization fixes (Jeff Layton). The one included here
mostly concerns things like open_by_handle() and there is another one
that will come through Al.
- copy_file_range() now uses the new copy-from2 op (Luis Henriques).
The existing copy-from op turned out to be infeasible for generic
filesystem use; we disable the copy offload if OSDs don't support
copy-from2.
- a patch to link "rbd" and "block" devices together in sysfs (Hannes
Reinecke)
... and a smattering of cleanups from Xiubo, Jeff and Chengguang.
* tag 'ceph-for-5.6-rc1' of https://github.com/ceph/ceph-client: (25 commits)
rbd: set the 'device' link in sysfs
ceph: move net/ceph/ceph_fs.c to fs/ceph/util.c
ceph: print name of xattr in __ceph_{get,set}xattr() douts
ceph: print r_direct_hash in hex in __choose_mds() dout
ceph: use copy-from2 op in copy_file_range
ceph: close holes in structs ceph_mds_session and ceph_mds_request
rbd: work around -Wuninitialized warning
ceph: allocate the correct amount of extra bytes for the session features
ceph: rename get_session and switch to use ceph_get_mds_session
ceph: remove the extra slashes in the server path
ceph: add possible_max_rank and make the code more readable
ceph: print dentry offset in hex and fix xattr_version type
ceph: only touch the caps which have the subset mask requested
ceph: don't clear I_NEW until inode metadata is fully populated
ceph: retry the same mds later after the new session is opened
ceph: check availability of mds cluster on mount after wait timeout
ceph: keep the session state until it is released
ceph: add __send_request helper
ceph: ensure we have a new cap before continuing in fill_inode
ceph: drop unused ttl_from parameter from fill_inode
...
It's hard to read, especially when it is:
ceph: __choose_mds 00000000b7bc9c15 is_hash=1 (-271041095) mode 0
At the same time, switch to __func__ to get rid of the checkpatch
warning.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The total bytes may potentially be larger than 8.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Just in case the session's refcount reach 0 and is releasing, and
if we get the session without checking it, we may encounter kernel
crash.
Rename get_session to ceph_get_mds_session and make it global.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The m_num_mds here is actually the number for MDSs which are in
up:active status, and it will be duplicated to m_num_active_mds,
so remove it.
Add possible_max_rank to the mdsmap struct and this will be
the correctly possible largest rank boundary.
Remove the special case for one mds in __mdsmap_get_random_mds(),
because the validate mds rank may not always be 0.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If max_mds > 1 and a request is submitted that chooses a random mds
rank, and the relating session is not opened yet, the request will wait
until the session has been opened and resend again.
Every time the request goes through __do_request, it will release the
req->session first and choose a random one again, which may be a
completely different rank than the one it just waited on.
In the worst case, it will open all the mds sessions one by one just
before the request can be successfully sent out.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If all the MDS daemons are down for some reason, then the first mount
attempt will fail with EIO after the mount request times out. A mount
attempt will also fail with EIO if all of the MDS's are laggy.
This patch changes the code to return -EHOSTUNREACH in these situations
and adds a pr_info error message to help the admin determine the cause.
URL: https://tracker.ceph.com/issues/4386
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When reconnecting the session but if it is denied by the MDS due
to client was in blacklist or something else, kclient will receive
a session close reply, and we will never see the important log:
"ceph: mds%d reconnect denied"
And with the confusing log:
"ceph: handle_session mds0 close 0000000085804730 state ??? seq 0"
Let's keep the session state until its memories is released.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
During umount, if there has no any unsafe request in the mdsc and
some requests still in-flight and not got reply yet, and if the
rest requets are all safe ones, after that even all of them in mdsc
are unregistered, the umount must wait until after mount_timeout
seconds anyway.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Even the MDS is in up:active state, but it also maybe laggy. Here
will skip the laggy MDSs.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
We print session's refcount in debug message inside
ceph_put_mds_session() and get_session(), so we don't have to
print it in con_get()/__ceph_lookup_mds_session()/con_put().
Signed-off-by: Chengguang Xu <cgxu519@gmx.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Currently, we just assume that it will stick around by virtue of the
submitter's reference, but later patches will allow the syscall to
return early and we can't rely on that reference at that point.
While I'm not aware of any reports of it, Xiubo pointed out that this
may fix a use-after-free. If the wait for a reply times out or is
canceled via signal, and then the reply comes in after the syscall
returns, the client can end up trying to access r_parent without a
reference.
Take an extra reference to the inode when setting r_parent and release
it when releasing the request.
Cc: stable@vger.kernel.org
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The nr in ceph_reclaim_caps_nr() is very possibly larger than 1,
so we may miss it and the reclaim work couldn't triggered as expected.
Signed-off-by: Xiubo Li <xiubli@redhat.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
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>
Most of these values should never be negative, so convert them to
unsigned values. Add some sanity checking to the parsed values, and
clean up some unneeded casts.
Note that while caps_max should never be negative, this patch leaves
it signed, since this value ends up later being compared to a signed
counter. Just ensure that userland never passes in a negative value
for caps_max.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Since ceph always uses ns granularity, skip the
truncation which is a no-op.
Signed-off-by: Deepa Dinamani <deepa.kernel@gmail.com>
Cc: jlayton@kernel.org
Cc: ceph-devel@vger.kernel.org
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
We currently just pass junk in this field unless we're retransmitting a
create, but in later patches, we'll need a mechanism to pass a delegated
inode number on an initial create request. Prepare for this by ensuring
this field is zeroed out.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When this occurs, it usually means that we raced with a rename, and
there is no need to warn in that case. Only printk if we pass the
rename sequence check but still ended up with pos < 0.
Either way, this doesn't warrant a KERN_ERR message. Change it to
KERN_WARNING.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
In the future, we're going to want to extend the ceph_reply_info_extra
for create replies. Currently though, the kernel code doesn't accept an
extra blob that is larger than the expected data.
Change the code to skip over any unrecognized fields at the end of the
extra blob, rather than returning -EIO.
Cc: stable@vger.kernel.org
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If client mds session is evicted in CEPH_MDS_SESSION_OPENING state,
mds won't send session msg to client, and delayed_work skip
CEPH_MDS_SESSION_OPENING state session, the session hang forever.
Allow ceph_con_keepalive to reconnect a session in OPENING to avoid
session hang. Also, ensure that we skip sessions in RESTARTING and
REJECTED states since those states can't be resurrected by issuing
a keepalive.
Link: https://tracker.ceph.com/issues/41551
Signed-off-by: Erqi Chen chenerqi@gmail.com
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
It's only used to keep count of caps being trimmed, but that requires
that we hold the session->s_mutex to prevent multiple trimming
operations from running concurrently.
We can achieve the same effect using an integer on the stack, which
allows us to (eventually) not need the s_mutex.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Make client use osd reply and session message to infer if itself is
blacklisted. Client reconnect to cluster using new entity addr if it
is blacklisted. Auto reconnect is limited to once every 30 minutes.
Auto reconnect is disabled by default. It can be enabled/disabled by
recover_session=<no|clean> mount option. In 'clean' mode, client drops
any dirty data/metadata, invalidates page caches and invalidates all
writable file handles. After reconnect, file locks become stale because
MDS loses track of them. If an inode contains any stale file locks,
read/write on the indoe are not allowed until applications release all
stale file locks.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
It closes mds sessions, drop all caps and invalidates page caches,
then use new entity address to reconnect to the cluster.
After reconnect, all dirty data/metadata are dropped, file locks
get lost sliently. Open files continue to work because client will
try renewing caps on later read/write.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Use errseq_t to track and report errors of async metadata operations,
similar to how kernel handles errors during writeback.
If any dirty caps or any unsafe request gets dropped during session
eviction, record -EIO in corresponding inode's i_meta_err. The error
will be reported by subsequent fsync,
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
handle_cap_export() may add placeholder caps to session that is in
opening state. These caps' session pointer become wild after session get
unregistered.
The fix is not to unregister session in opening state during mds failovers,
just let client to reconnect later when mds is recovered.
Link: https://tracker.ceph.com/issues/40190
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
MDS InodeStat v3 wire structures include a trailing snapshot creation
time member. Unmarshall this and retain it for a future vxattr.
Signed-off-by: David Disseldorp <ddiss@suse.de>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When ceph_mdsc_build_path is handed a positive dentry, it will return a
zero-length path string with the base set to that dentry. This is not
what we want. Always include at least one path component in the string.
ceph_mdsc_build_path has behaved this way for a long time but it didn't
matter until recent d_name handling rework.
Fixes: 964fff7491 ("ceph: use ceph_mdsc_build_path instead of clone_dentry_name")
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
iput_final() may wait for reahahead pages. The wait can cause deadlock.
For example:
Workqueue: ceph-msgr ceph_con_workfn [libceph]
Call Trace:
schedule+0x36/0x80
io_schedule+0x16/0x40
__lock_page+0x101/0x140
truncate_inode_pages_range+0x556/0x9f0
truncate_inode_pages_final+0x4d/0x60
evict+0x182/0x1a0
iput+0x1d2/0x220
iterate_session_caps+0x82/0x230 [ceph]
dispatch+0x678/0xa80 [ceph]
ceph_con_workfn+0x95b/0x1560 [libceph]
process_one_work+0x14d/0x410
worker_thread+0x4b/0x460
kthread+0x105/0x140
ret_from_fork+0x22/0x40
Workqueue: ceph-msgr ceph_con_workfn [libceph]
Call Trace:
__schedule+0x3d6/0x8b0
schedule+0x36/0x80
schedule_preempt_disabled+0xe/0x10
mutex_lock+0x2f/0x40
ceph_check_caps+0x505/0xa80 [ceph]
ceph_put_wrbuffer_cap_refs+0x1e5/0x2c0 [ceph]
writepages_finish+0x2d3/0x410 [ceph]
__complete_request+0x26/0x60 [libceph]
handle_reply+0x6c8/0xa10 [libceph]
dispatch+0x29a/0xbb0 [libceph]
ceph_con_workfn+0x95b/0x1560 [libceph]
process_one_work+0x14d/0x410
worker_thread+0x4b/0x460
kthread+0x105/0x140
ret_from_fork+0x22/0x40
In above example, truncate_inode_pages_range() waits for readahead pages
while holding s_mutex. ceph_check_caps() waits for s_mutex and blocks
OSD dispatch thread. Later OSD replies (for readahead) can't be handled.
ceph_check_caps() also may lock snap_rwsem for read. So similar deadlock
can happen if iput_final() is called while holding snap_rwsem.
In general, it's not good to call iput_final() inside MDS/OSD dispatch
threads or while holding any mutex.
The fix is introducing ceph_async_iput(), which calls iput_final() in
workqueue.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
I originally thought there was a potential race here, but the fact
that this is called with the mdsc->mutex held, ensures that the
last reference to the session can't be put here.
Still, it's clearer to just return the value from get_session here,
and may prevent a bug later if we ever rework this code to be less
reliant on mutexes.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Nothing calls ceph_mdsc_submit_request today, but in later patches we'll
need to be able to call this separately.
Have the helper return an int so we can check the r_err under the mutex,
and have the caller just check the error code from the submit. Also move
the acquisition of CEPH_CAP_PIN references into the same function.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
No MDS requests use r_callback today, but that will change in the
future. The OSD client always does r_callback and then completes
r_completion. Let's have the MDS client do the same.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
We make copies of the dentry name in set_request_path_attr, but then
create_request_message re-fetches the lengths out of the dentry. While
we don't currently set the *_drop fields unless the parents are locked,
it's still better not to rely on that sort of implicit assumption.
Use the pathlen values that set_request_path_attr returned instead, as
they will always be correct for the returned paths themselves.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Al suggested we get rid of the kmalloc here and just use __getname
and __putname to get a full PATH_MAX pathname buffer.
Since we build the path in reverse, we continue to return a pointer
to the beginning of the string and the length, and add a new helper
to free the thing at the end.
Suggested-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
While it may be slightly more efficient, it's probably not worthwhile to
optimize for the case that clone_dentry_name handles. We can get the
same result by just calling ceph_mdsc_build_path when the parent isn't
locked, with less code duplication.
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
temp is not defined outside of the RCU critical section here. Ensure
we grab that value before we drop the rcu_read_lock.
Reported-by: Al Viro <viro@zeniv.linux.org.uk>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
The CephFS kernel client does not enforce quotas set in a directory that
isn't visible from the mount point. For example, given the path
'/dir1/dir2', if quotas are set in 'dir1' and the filesystem is mounted with
mount -t ceph <server>:<port>:/dir1/ /mnt
then the client won't be able to access 'dir1' inode, even if 'dir2' belongs
to a quota realm that points to it.
This patch fixes this issue by simply doing an MDS LOOKUPINO operation for
unknown inodes. Any inode reference obtained this way will be added to a
list in ceph_mds_client, and will only be released when the filesystem is
umounted.
Link: https://tracker.ceph.com/issues/38482
Reported-by: Hendrik Peyerl <hpeyerl@plusline.net>
Signed-off-by: Luis Henriques <lhenriques@suse.com>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
We missed two places that i_wrbuffer_ref_head, i_wr_ref, i_dirty_caps
and i_flushing_caps may change. When they are all zeros, we should free
i_head_snapc.
Cc: stable@vger.kernel.org
Link: https://tracker.ceph.com/issues/38224
Reported-and-tested-by: Luis Henriques <lhenriques@suse.com>
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Ben reported tripping the BUG_ON in create_request_message during some
performance testing. Analysis of the vmcore showed that the length of
the r_dentry->d_name string changed after we allocated the buffer, but
before we encoded it.
build_dentry_path returns pointers to d_name in the common case of
non-snapped dentries, but this optimization isn't safe unless the parent
directory is locked. When it isn't, have the code make a copy of the
d_name while holding the d_lock.
Cc: stable@vger.kernel.org
Reported-by: Ben England <bengland@redhat.com>
Signed-off-by: Jeff Layton <jlayton@kernel.org>
Reviewed-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
If number of caps exceed the limit, ceph_trim_dentires() also trim
dentries with valid leases. Trimming dentry releases references to
associated inode, which may evict inode and release caps.
By default, there is no limit for caps count.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Previous commit make VFS delete stale dentry when last reference is
dropped. Lease also can become invalid when corresponding dentry has
no reference. This patch make cephfs periodically scan lease list,
delete corresponding dentry if lease is invalid.
There are two types of lease, dentry lease and dir lease. dentry lease
has life time and applies to singe dentry. Dentry lease is added to tail
of a list when it's updated, leases at front of the list will expire
first. Dir lease is CEPH_CAP_FILE_SHARED on directory inode, it applies
to all dentries in the directory. Dentries have dir leases are added to
another list. Dentries in the list are periodically checked in a round
robin manner.
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
When pending cap releases fill up one message, start a work to send
cap release message. (old way is sending cap releases every 5 seconds)
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Reviewed-by: Jeff Layton <jlayton@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
In versioned reply, inodestat, dirstat and lease are encoded with
version, compat_version and struct_len.
Based on a patch from Jos Collin <jcollin@redhat.com>.
Link: http://tracker.ceph.com/issues/26936
Signed-off-by: "Yan, Zheng" <zyan@redhat.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Ilya Dryomov
Jeff Layton
Xiubo Li
Colin Ian King
Yan, Zheng
Yanhu Cao
Linus Torvalds
Chengguang Xu
Deepa Dinamani
Erqi Chen
David Disseldorp
Luis Henriques