It is currently hard-coded in the mon_client that mdsmap and monmap
subs are continuous, while osdmap sub is always "onetime". To better
handle full clusters/pools in the osd_client, we need to be able to
issue continuous osdmap subs. Revamp subs code to allow us to specify
for each sub whether it should be continuous or not.
Although not strictly required for the above, switch to SUBSCRIBE2
protocol while at it, eliminating the ambiguity between a request for
"every map since X" and a request for "just the latest" when we don't
have a map yet (i.e. have epoch 0). SUBSCRIBE2 feature bit is now
required - it's been supported since pre-argonaut (2010).
Move "got mdsmap" call to the end of ceph_mdsc_handle_map() - calling
in before we validate the epoch and successfully install the new map
can mess up mon_client sub state.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Commit d15f9d694b ("libceph: check data_len in ->alloc_msg()")
mistakenly bumped the log level on the "tid %llu unknown, skipping"
message. Turn it back into a dout() - stray replies are perfectly
normal when OSDs flap, crash, get killed for testing purposes, etc.
Cc: stable@vger.kernel.org # 4.3+
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Empty request_redirect_t (struct ceph_request_redirect in the kernel
client) is now encoded with a bool. NEW_OSDOPREPLY_ENCODING feature
bit overlaps with already supported CRUSH_TUNABLES5.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
The following bit in ceph_msg_revoke_incoming() is unsafe:
struct ceph_connection *con = msg->con;
if (!con)
return;
mutex_lock(&con->mutex);
<more msg->con use>
There is nothing preventing con from getting destroyed right after
msg->con test. One easy way to reproduce this is to disable message
signing only on the server side and try to map an image. The system
will go into a
libceph: read_partial_message ffff880073f0ab68 signature check failed
libceph: osd0 192.168.255.155:6801 bad crc/signature
libceph: read_partial_message ffff880073f0ab68 signature check failed
libceph: osd0 192.168.255.155:6801 bad crc/signature
loop which has to be interrupted with Ctrl-C. Hit Ctrl-C and you are
likely to end up with a random GP fault if the reset handler executes
"within" ceph_msg_revoke_incoming():
<yet another reply w/o a signature>
...
<Ctrl-C>
rbd_obj_request_end
ceph_osdc_cancel_request
__unregister_request
ceph_osdc_put_request
ceph_msg_revoke_incoming
...
osd_reset
__kick_osd_requests
__reset_osd
remove_osd
ceph_con_close
reset_connection
<clear con->in_msg->con>
<put con ref>
put_osd
<free osd/con>
<msg->con use> <-- !!!
If ceph_msg_revoke_incoming() executes "before" the reset handler,
osd/con will be leaked because ceph_msg_revoke_incoming() clears
con->in_msg but doesn't put con ref, while reset_connection() only puts
con ref if con->in_msg != NULL.
The current msg->con scheme was introduced by commits 38941f8031
("libceph: have messages point to their connection") and 92ce034b5a
("libceph: have messages take a connection reference"), which defined
when messages get associated with a connection and when that
association goes away. Part of the problem is that this association is
supposed to go away in much too many places; closing this race entirely
requires either a rework of the existing or an addition of a new layer
of synchronization.
In lieu of that, we can make it *much* less likely to hit by
disassociating messages only on their destruction and resend through
a different connection. This makes the code simpler and is probably
a good thing to do regardless - this patch adds a msg_con_set() helper
which is is called from only three places: ceph_con_send() and
ceph_con_in_msg_alloc() to set msg->con and ceph_msg_release() to clear
it.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
We can use msg->con instead - at the point we sign an outgoing message
or check the signature on the incoming one, msg->con is always set. We
wouldn't know how to sign a message without an associated session (i.e.
msg->con == NULL) and being able to sign a message using an explicitly
provided authorizer is of no use.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This patch changes the osd_req_op_data() macro to not evaluate
arguments more than once in order to follow the kernel coding style.
Signed-off-by: Ioana Ciornei <ciorneiioana@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
[idryomov@gmail.com: changelog, formatting]
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Since handle_reply() does not use its con argument, remove it.
Signed-off-by: Shraddha Barke <shraddha.6596@gmail.com>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
This covers only the simplest case - an object size sized write, but
it's still useful in tiering setups when EC is used for the base tier
as writefull op can be proxied, saving an object promotion.
Even though updating ceph_osdc_new_request() to allow writefull should
just be a matter of fixing an assert, I didn't do it because its only
user is cephfs. All other sites were updated.
Reflects ceph.git commit 7bfb7f9025a8ee0d2305f49bf0336d2424da5b5b.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Only ->alloc_msg() should check data_len of the incoming message
against the preallocated ceph_msg, doing it in the messenger is not
right. The contract is that either ->alloc_msg() returns a ceph_msg
which will fit all of the portions of the incoming message, or it
returns NULL and possibly sets skip, signaling whether NULL is due to
an -ENOMEM. ->alloc_msg() should be the only place where we make the
skip/no-skip decision.
I stumbled upon this while looking at con/osd ref counting. Right now,
if we get a non-extent message with a larger data portion than we are
prepared for, ->alloc_msg() returns a ceph_msg, and then, when we skip
it in the messenger, we don't put the con/osd ref acquired in
ceph_con_in_msg_alloc() (which is normally put in process_message()),
so this also fixes a memory leak.
An existing BUG_ON in ceph_msg_data_cursor_init() ensures we don't
corrupt random memory should a buggy ->alloc_msg() return an unfit
ceph_msg.
While at it, I changed the "unknown tid" dout() to a pr_warn() to make
sure all skips are seen and unified format strings.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
There are currently three libceph-level timeouts that the user can
specify on mount: mount_timeout, osd_idle_ttl and osdkeepalive. All of
these are in seconds and no checking is done on user input: negative
values are accepted, we multiply them all by HZ which may or may not
overflow, arbitrarily large jiffies then get added together, etc.
There is also a bug in the way mount_timeout=0 is handled. It's
supposed to mean "infinite timeout", but that's not how wait.h APIs
treat it and so __ceph_open_session() for example will busy loop
without much chance of being interrupted if none of ceph-mons are
there.
Fix all this by verifying user input, storing timeouts capped by
msecs_to_jiffies() in jiffies and using the new ceph_timeout_jiffies()
helper for all user-specified waits to handle infinite timeouts
correctly.
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Alex Elder <elder@linaro.org>
This reverts commit ba9d114ec5.
.. which introduced a regression that prevented all lingering requests
requeued in kick_requests() from ever being sent to the OSDs, resulting
in a lot of missed notifies. In retrospect it's pretty obvious that
r_req_lru_item item in the case of lingering requests can be used not
only for notarget, but also for unsent linkage due to how tightly
actual map and enqueue operations are coupled in __map_request().
The assertion that was being silenced is taken care of in the previous
("libceph: request a new osdmap if lingering request maps to no osd")
commit: by always kicking homeless lingering requests we ensure that
none of them ends up on the notarget list outside of the critical
section guarded by request_mutex.
Cc: stable@vger.kernel.org # 3.18+, needs b049453221 "libceph: request a new osdmap if lingering request maps to no osd"
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
This commit does two things. First, if there are any homeless
lingering requests, we now request a new osdmap even if the osdmap that
is being processed brought no changes, i.e. if a given lingering
request turned homeless in one of the previous epochs and remained
homeless in the current epoch. Not doing so leaves us with a stale
osdmap and as a result we may miss our window for reestablishing the
watch and lose notifies.
MON=1 OSD=1:
# cat linger-needmap.sh
#!/bin/bash
rbd create --size 1 test
DEV=$(rbd map test)
ceph osd out 0
rbd map dne/dne # obtain a new osdmap as a side effect (!)
sleep 1
ceph osd in 0
rbd resize --size 2 test
# rbd info test | grep size -> 2M
# blockdev --getsize $DEV -> 1M
N.B.: Not obtaining a new osdmap in between "osd out" and "osd in"
above is enough to make it miss that resize notify, but that is a
bug^Wlimitation of ceph watch/notify v1.
Second, homeless lingering requests are now kicked just like those
lingering requests whose mapping has changed. This is mainly to
recognize that a homeless lingering request makes no sense and to
preserve the invariant that a registered lingering request is not
sitting on any of r_req_lru_item lists. This spares us a WARN_ON,
which commit ba9d114ec5 ("libceph: clear r_req_lru_item in
__unregister_linger_request()") tried to fix the _wrong_ way.
Cc: stable@vger.kernel.org # 3.10+
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
a255651d4c ("ceph: ensure auth ops are defined before use") made
kfree() in put_osd() conditional on the authorizer. A mechanical
mistake most likely - fix it.
Cc: Alex Elder <elder@linaro.org>
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
It turns out it's possible to get __remove_osd() called twice on the
same OSD. That doesn't sit well with rb_erase() - depending on the
shape of the tree we can get a NULL dereference, a soft lockup or
a random crash at some point in the future as we end up touching freed
memory. One scenario that I was able to reproduce is as follows:
<osd3 is idle, on the osd lru list>
<con reset - osd3>
con_fault_finish()
osd_reset()
<osdmap - osd3 down>
ceph_osdc_handle_map()
<takes map_sem>
kick_requests()
<takes request_mutex>
reset_changed_osds()
__reset_osd()
__remove_osd()
<releases request_mutex>
<releases map_sem>
<takes map_sem>
<takes request_mutex>
__kick_osd_requests()
__reset_osd()
__remove_osd() <-- !!!
A case can be made that osd refcounting is imperfect and reworking it
would be a proper resolution, but for now Sage and I decided to fix
this by adding a safe guard around __remove_osd().
Fixes: http://tracker.ceph.com/issues/8087
Cc: Sage Weil <sage@redhat.com>
Cc: stable@vger.kernel.org # 3.9+: 7c6e6fc53e73: libceph: assert both regular and lingering lists in __remove_osd()
Cc: stable@vger.kernel.org # 3.9+: cc9f1f518cec: libceph: change from BUG to WARN for __remove_osd() asserts
Cc: stable@vger.kernel.org # 3.9+
Signed-off-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Sage Weil <sage@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
allow specifying position of extent operation in multi-operations
osd request. This is required for cephfs to convert inline data to
normal data (compare xattr, then write object).
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>
Add CEPH_OSD_OP_CREATE support. Also change libceph to not treat
CEPH_OSD_OP_DELETE as an extent op and add an assert to that end.
Signed-off-by: Yan, Zheng <zyan@redhat.com>
Reviewed-by: Ilya Dryomov <idryomov@redhat.com>
kick_requests() can put linger requests on the notarget list. This
means we need to clear the much-overloaded req->r_req_lru_item in
__unregister_linger_request() as well, or we get an assertion failure
in ceph_osdc_release_request() - !list_empty(&req->r_req_lru_item).
AFAICT the assumption was that registered linger requests cannot be on
any of req->r_req_lru_item lists, but that's clearly not the case.
Signed-off-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Requests have to be unlinked from both osd->o_requests (normal
requests) and osd->o_linger_requests (linger requests) lists when
clearing req->r_osd. Otherwise __unregister_linger_request() gets
confused and we trip over a !list_empty(&osd->o_linger_requests)
assert in __remove_osd().
MON=1 OSD=1:
# cat remove-osd.sh
#!/bin/bash
rbd create --size 1 test
DEV=$(rbd map test)
ceph osd out 0
sleep 3
rbd map dne/dne # obtain a new osdmap as a side effect
rbd unmap $DEV & # will block
sleep 3
ceph osd in 0
Signed-off-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Bring in missing osd ops and strings, use macros to eliminate multiple
points of maintenance.
Signed-off-by: Ilya Dryomov <idryomov@redhat.com>
Reviewed-by: Sage Weil <sage@redhat.com>
queue_work() doesn't "fail to queue", it returns false if work was
already on a queue, which can't happen here since we allocate
event_work right before we queue it. So don't bother at all.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Use the more common pr_warn.
Other miscellanea:
o Coalesce formats
o Realign arguments
Signed-off-by: Joe Perches <joe@perches.com>
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Both not yet registered (r_linger && list_empty(&r_linger_item)) and
registered linger requests should use the new tid on resend to avoid
the dup op detection logic on the OSDs, yet we were doing this only for
"registered" case. Factor out and simplify the "registered" logic and
use the new helper for "not registered" case as well.
Fixes: http://tracker.ceph.com/issues/8806
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Introduce ceph_osdc_cancel_request() intended for canceling requests
from the higher layers (rbd and cephfs). Because higher layers are in
charge and are supposed to know what and when they are canceling, the
request is not completed, only unref'ed and removed from the libceph
data structures.
__cancel_request() is no longer called before __unregister_request(),
because __unregister_request() unconditionally revokes r_request and
there is no point in trying to do it twice.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
We should check if request is on the linger request list of any of the
OSDs, not whether request is registered or not.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Linger requests that have not yet been registered should not be
unregistered by __unregister_linger_request(). This messes up ref
count and leads to use-after-free.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
It is important that both regular and lingering requests lists are
empty when the OSD is removed.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Add some WARN_ONs to alert us when we try to destroy requests that are
still registered.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Add dout()s to ceph_osdc_request_{get,put}(). Also move them to .c and
turn kref release callback into a static function.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Abstract out __move_osd_to_lru() logic from __unregister_request() and
__unregister_linger_request().
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
So that:
req->r_osd_item --> osd->o_requests list
req->r_linger_osd_item --> osd->o_linger_requests list
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Sparse complained about this bogus extern on definition of
a function.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Pull Ceph updates from Sage Weil:
"The biggest chunk is a series of patches from Ilya that add support
for new Ceph osd and crush map features, including some new tunables,
primary affinity, and the new encoding that is needed for erasure
coding support. This brings things into parity with the server side
and the looming firefly release. There is also support for allocation
hints in RBD that help limit fragmentation on the server side.
There is also a series of patches from Zheng fixing NFS reexport,
directory fragmentation support, flock vs fnctl behavior, and some
issues with clustered MDS.
Finally, there are some miscellaneous fixes from Yunchuan Wen for
fscache, Fabian Frederick for ACLs, and from me for fsync(dirfd)
behavior"
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (79 commits)
ceph: skip invalid dentry during dcache readdir
libceph: dump pool {read,write}_tier to debugfs
libceph: output primary affinity values on osdmap updates
ceph: flush cap release queue when trimming session caps
ceph: don't grabs open file reference for aborted request
ceph: drop extra open file reference in ceph_atomic_open()
ceph: preallocate buffer for readdir reply
libceph: enable PRIMARY_AFFINITY feature bit
libceph: redo ceph_calc_pg_primary() in terms of ceph_calc_pg_acting()
libceph: add support for osd primary affinity
libceph: add support for primary_temp mappings
libceph: return primary from ceph_calc_pg_acting()
libceph: switch ceph_calc_pg_acting() to new helpers
libceph: introduce apply_temps() helper
libceph: introduce pg_to_raw_osds() and raw_to_up_osds() helpers
libceph: ceph_can_shift_osds(pool) and pool type defines
libceph: ceph_osd_{exists,is_up,is_down}(osd) definitions
libceph: enable OSDMAP_ENC feature bit
libceph: primary_affinity decode bits
libceph: primary_affinity infrastructure
...
In preparation for adding support for primary_temp, stop assuming
primaryness: add a primary out parameter to ceph_calc_pg_acting() and
change call sites accordingly. Primary is now specified separately
from the order of osds in the set.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Split osdmap allocation and initialization into a separate function,
ceph_osdmap_decode().
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
This is primarily for rbd's benefit and is supposed to combat
fragmentation:
"... knowing that rbd images have a 4m size, librbd can pass a hint
that will let the osd do the xfs allocation size ioctl on new files so
that they are allocated in 1m or 4m chunks. We've seen cases where
users with rbd workloads have very high levels of fragmentation in xfs
and this would mitigate that and probably have a pretty nice
performance benefit."
SETALLOCHINT is considered advisory, so our backwards compatibility
mechanism here is to set FAILOK flag for all SETALLOCHINT ops.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
Encode ceph_osd_op::flags field so that it gets sent over the wire.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
One of my pet coding style peeves is the practice of
adding extra return; at the end of function.
Kill several instances of this in network code.
I suppose some coccinelle wizardy could do this automatically.
Signed-off-by: Stephen Hemminger <stephen@networkplumber.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
Handling redirect replies requires both map_sem and request_mutex.
Taking map_sem unconditionally near the top of handle_reply() avoids
possible race conditions that arise from releasing request_mutex to be
able to acquire map_sem in redirect reply case. (Lock ordering is:
map_sem, request_mutex, crush_mutex.)
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Factor out logic from ceph_osdc_start_request() into a new helper,
__ceph_osdc_start_request(). ceph_osdc_start_request() now amounts to
taking locks and calling __ceph_osdc_start_request().
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
msgpool_op_reply message pool isn't destroyed if workqueue construction
fails. Fix it.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Follow redirect replies from osds, for details see ceph.git commit
fbbe3ad1220799b7bb00ea30fce581c5eadaf034.
v1 (current) version of redirect reply consists of oloc and oid, which
expands to pool, key, nspace, hash and oid. However, server-side code
that would populate anything other than pool doesn't exist yet, and
hence this commit adds support for pool redirects only. To make sure
that future server-side updates don't break us, we decode all fields
and, if any of key, nspace, hash or oid have a non-default value, error
out with "corrupt osd_op_reply ..." message.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Rename ceph_osd_request::r_{oloc,oid} to r_base_{oloc,oid} before
introducing r_target_{oloc,oid} needed for redirects.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Overwrite ceph_osd_request::r_oloc.pool with read_tier for read ops and
write_tier for write and read+write ops (aka basic tiering support).
{read,write}_tier are part of pg_pool_t since v9. This commit bumps
our pg_pool_t decode compat version from v7 to v9, all new fields
except for {read,write}_tier are ignored.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Switch ceph_calc_ceph_pg() to new oloc and oid abstractions and rename
it to ceph_oloc_oid_to_pg() to make its purpose more clear.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
In preparation for tiering support, which would require having two
(base and target) object names for each osd request and also copying
those names around, introduce struct ceph_object_id (oid) and a couple
helpers to facilitate those copies and encapsulate the fact that object
name is not necessarily a NUL-terminated string.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
In preparation for adding oid abstraction, rename MAX_OBJ_NAME_SIZE to
CEPH_MAX_OID_NAME_LEN.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Instead of relying on pool fields in ceph_file_layout (for mapping) and
ceph_pg (for enconding), start using ceph_object_locator (oloc)
abstraction. Note that userspace oloc currently consists of pool, key,
nspace and hash fields, while this one contains only a pool. This is
OK, because at this point we only send (i.e. encode) olocs and never
have to receive (i.e. decode) them.
This makes keeping a copy of ceph_file_layout in every osd request
unnecessary, so ceph_osd_request::r_file_layout field is nuked.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The check that makes sure that we have enough memory allocated to read
in the entire header of the message in question is currently busted.
It compares front_len of the incoming message with iov_len field of
ceph_msg::front structure, which is used primarily to indicate the
amount of data already read in, and not the size of the allocated
buffer. Under certain conditions (e.g. a short read from a socket
followed by that socket's shutdown and owning ceph_connection reset)
this results in a warning similar to
[85688.975866] libceph: get_reply front 198 > preallocated 122 (4#0)
and, through another bug, leads to forever hung tasks and forced
reboots. Fix this by comparing front_len with front_alloc_len field of
struct ceph_msg, which stores the actual size of the buffer.
Fixes: http://tracker.ceph.com/issues/5425
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Rename front local variable to front_len in get_reply() to make its
purpose more clear.
Signed-off-by: Ilya Dryomov <ilya.dryomov@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
With the current full handling, there is a race between osds and
clients getting the first map marked full. If the osd wins, it will
return -ENOSPC to any writes, but the client may already have writes
in flight. This results in the client getting the error and
propagating it up the stack. For rbd, the block layer turns this into
EIO, which can cause corruption in filesystems above it.
To avoid this race, osds are being changed to drop writes that came
from clients with an osdmap older than the last osdmap marked full.
In order for this to work, clients must resend all writes after they
encounter a full -> not full transition in the osdmap. osds will wait
for an updated map instead of processing a request from a client with
a newer map, so resent writes will not be dropped by the osd unless
there is another not full -> full transition.
This approach requires both osds and clients to be fixed to avoid the
race. Old clients talking to osds with this fix may hang instead of
returning EIO and potentially corrupting an fs. New clients talking to
old osds have the same behavior as before if they encounter this race.
Fixes: http://tracker.ceph.com/issues/6938
Reviewed-by: Sage Weil <sage@inktank.com>
Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
The PAUSEWR and PAUSERD flags are meant to stop the cluster from
processing writes and reads, respectively. The FULL flag is set when
the cluster determines that it is out of space, and will no longer
process writes. PAUSEWR and PAUSERD are purely client-side settings
already implemented in userspace clients. The osd does nothing special
with these flags.
When the FULL flag is set, however, the osd responds to all writes
with -ENOSPC. For cephfs, this makes sense, but for rbd the block
layer translates this into EIO. If a cluster goes from full to
non-full quickly, a filesystem on top of rbd will not behave well,
since some writes succeed while others get EIO.
Fix this by blocking any writes when the FULL flag is set in the osd
client. This is the same strategy used by userspace, so apply it by
default. A follow-on patch makes this configurable.
__map_request() is called to re-target osd requests in case the
available osds changed. Add a paused field to a ceph_osd_request, and
set it whenever an appropriate osd map flag is set. Avoid queueing
paused requests in __map_request(), but force them to be resent if
they become unpaused.
Also subscribe to the next osd map from the monitor if any of these
flags are set, so paused requests can be unblocked as soon as
possible.
Fixes: http://tracker.ceph.com/issues/6079
Reviewed-by: Sage Weil <sage@inktank.com>
Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
Wake up possible waiters, invoke the call back if any, unregister the request
Signed-off-by: Li Wang <liwang@ubuntukylin.com>
Signed-off-by: Yunchuan Wen <yunchuanwen@ubuntukylin.com>
Signed-off-by: Sage Weil <sage@inktank.com>
Without a way to flush the osd client's notify workqueue, a watch
event that is unregistered could continue receiving callbacks
indefinitely.
Unregistering the event simply means no new notifies are added to the
queue, but there may still be events in the queue that will call the
watch callback for the event. If the queue is flushed after the event
is unregistered, the caller can be sure no more watch callbacks will
occur for the canceled watch.
Signed-off-by: Josh Durgin <josh.durgin@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@linaro.org>
create_singlethread_workqueue() returns NULL on error, and it doesn't
return ERR_PTRs.
I tweaked the error handling a little to be consistent with earlier in
the function.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Sage Weil <sage@inktank.com>
There are two places where we read "nr_maps" if both of them are set to
zero then we would hit a NULL dereference here.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Sage Weil <sage@inktank.com>
We've tried to fix the error paths in this function before, but there
is still a hidden goto in the ceph_decode_need() macro which goes to the
wrong place. We need to release the "req" and unlock a mutex before
returning.
Signed-off-by: Dan Carpenter <dan.carpenter@oracle.com>
Reviewed-by: Sage Weil <sage@inktank.com>
This patch implements fallocate and punch hole support for Ceph kernel client.
Signed-off-by: Li Wang <liwang@ubuntukylin.com>
Signed-off-by: Yunchuan Wen <yunchuanwen@ubuntukylin.com>
For nofail == false request, if __map_request failed, the caller does
cleanup work, like releasing the relative pages. It doesn't make any sense
to retry this request.
CC: stable@vger.kernel.org
Signed-off-by: Jianpeng Ma <majianpeng@gmail.com>
Reviewed-by: Sage Weil <sage@inktank.com>
We can't use !req->r_sent to check if OSD request is sent for the
first time, this is because __cancel_request() zeros req->r_sent
when OSD map changes. Rather than adding a new variable to struct
ceph_osd_request to indicate if it's sent for the first time, We
can call the unsafe callback only when unsafe OSD reply is received.
If OSD's first reply is safe, just skip calling the unsafe callback.
The purpose of unsafe callback is adding unsafe request to a list,
so that fsync(2) can wait for the safe reply. fsync(2) doesn't need
to wait for a write(2) that hasn't returned yet. So it's OK to add
request to the unsafe list when the first OSD reply is received.
(ceph_sync_write() returns after receiving the first OSD reply)
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
handle_reply() calls complete_request() only if the first OSD reply
has ONDISK flag.
Signed-off-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
If an osd client response message arrives that has a front section
that's too big for the buffer set aside to receive it, a warning
gets reported and a new buffer is allocated.
The warning says nothing about which connection had the problem.
Add the peer type and number to what gets reported, to be a bit more
informative.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
When an osd request is set to linger, the osd client holds onto the
request so it can be re-submitted following certain osd map changes.
The osd client holds a reference to the request until it is
unregistered. This is used by rbd for watch requests.
Currently, the reference is taken when the request is marked with
the linger flag. This means that if an error occurs after that
time but before the the request completes successfully, that
reference is leaked.
There's really no reason to take the reference until the request is
registered in the the osd client's list of lingering requests, and
that only happens when the lingering (watch) request completes
successfully.
So take that reference only when it gets registered following
succesful completion, and drop it (as before) when the request
gets unregistered. This avoids the reference problem on error
in rbd.
Rearrange ceph_osdc_unregister_linger_request() to avoid using
the request pointer after it may have been freed.
And hold an extra reference in kick_requests() while handling
a linger request that has not yet been registered, to ensure
it doesn't go away.
This resolves:
http://tracker.ceph.com/issues/3859
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd client has a red-black tree describing its osds, and
occasionally we would get crashes due to one of these trees tree
becoming corrupt somehow.
The problem turned out to be that reset_changed_osds() was being
called without protection of the osd client request mutex. That
function would call __reset_osd() for any osd that had changed, and
__reset_osd() would call __remove_osd() for any osd with no
outstanding requests, and finally __remove_osd() would remove the
corresponding entry from the red-black tree. Thus, the tree was
getting modified without having any lock protection, and was
vulnerable to problems due to concurrent updates.
This appears to be the only osd tree updating path that has this
problem. It can be fairly easily fixed by moving the call up
a few lines, to just before the request mutex gets dropped
in kick_requests().
This resolves:
http://tracker.ceph.com/issues/5043
Cc: stable@vger.kernel.org # 3.4+
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The rbd code has a need to be able to restart an osd request that
has already been started and completed once before. This currently
wouldn't work right because the osd client code assumes an osd
request will be started exactly once Certain fields in a request
are never cleared and this leads to trouble if you try to reuse it.
Specifically, the r_sent, r_got_reply, and r_completed fields are
never cleared. The r_sent field records the osd incarnation at the
time the request was sent to that osd. If that's non-zero, the
message won't get re-mapped to a target osd properly, and won't be
put on the unsafe requests list the first time it's sent as it
should. The r_got_reply field is used in handle_reply() to ensure
the reply to a request is processed only once. And the r_completed
field is used for lingering requests to avoid calling the callback
function every time the osd client re-sends the request on behalf of
its initiator.
Each osd request passes through ceph_osdc_start_request() when
responsibility for the request is handed over to the osd client for
completion. We can safely zero these three fields there each time a
request gets started.
One last related change--clear the r_linger flag when a request
is no longer registered as a linger request.
This resolves:
http://tracker.ceph.com/issues/5026
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Create a slab cache to manage allocation of ceph_osdc_request
structures.
This resolves:
http://tracker.ceph.com/issues/3926
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
A WATCH op includes an object version. The version that's supplied
is incorrectly byte-swapped osd_req_op_watch_init() where it's first
assigned (it's been this way since that code was first added).
The result is that the version sent to the osd is wrong, because
that value gets byte-swapped again in osd_req_encode_op(). This
is the source of a sparse warning related to improper byte order in
the assignment.
The approach of using the version to avoid a race is deprecated
(see http://tracker.ceph.com/issues/3871), and the watch parameter
is no longer even examined by the osd. So fix the assignment in
osd_req_op_watch_init() so it no longer does the byte swap.
This resolves:
http://tracker.ceph.com/issues/3847
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Add the ability to provide an array of pages as outbound request
data for object class method calls.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Allow osd request ops that aren't otherwise structured (not class,
extent, or watch ops) to specify "raw" data to be used to hold
incoming data for the op. Make use of this capability for the osd
STAT op.
Prefix the name of the private function osd_req_op_init() with "_",
and expose a new function by that (earlier) name whose purpose is to
initialize osd ops with (only) implied data.
For now we'll just support the use of a page array for an osd op
with incoming raw data.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There are a bunch of functions defined to encapsulate getting the
address of a data field for a particular op in an osd request.
They're all defined the same way, so create a macro to take the
place of all of them.
Two of these are used outside the osd client code, so preserve them
(but convert them to use the new macro internally). Stop exporting
the ones that aren't used elsewhere.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
In the incremental move toward supporting distinct data items in an
osd request some of the functions had "write_request" parameters to
indicate, basically, whether the data belonged to in_data or the
out_data. Now that we maintain the data fields in the op structure
there is no need to indicate the direction, so get rid of the
"write_request" parameters.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request currently has two callbacks. They inform the
initiator of the request when we've received confirmation for the
target osd that a request was received, and when the osd indicates
all changes described by the request are durable.
The only time the second callback is used is in the ceph file system
for a synchronous write. There's a race that makes some handling of
this case unsafe. This patch addresses this problem. The error
handling for this callback is also kind of gross, and this patch
changes that as well.
In ceph_sync_write(), if a safe callback is requested we want to add
the request on the ceph inode's unsafe items list. Because items on
this list must have their tid set (by ceph_osd_start_request()), the
request added *after* the call to that function returns. The
problem with this is that there's a race between starting the
request and adding it to the unsafe items list; the request may
already be complete before ceph_sync_write() even begins to put it
on the list.
To address this, we change the way the "safe" callback is used.
Rather than just calling it when the request is "safe", we use it to
notify the initiator the bounds (start and end) of the period during
which the request is *unsafe*. So the initiator gets notified just
before the request gets sent to the osd (when it is "unsafe"), and
again when it's known the results are durable (it's no longer
unsafe). The first call will get made in __send_request(), just
before the request message gets sent to the messenger for the first
time. That function is only called by __send_queued(), which is
always called with the osd client's request mutex held.
We then have this callback function insert the request on the ceph
inode's unsafe list when we're told the request is unsafe. This
will avoid the race because this call will be made under protection
of the osd client's request mutex. It also nicely groups the setup
and cleanup of the state associated with managing unsafe requests.
The name of the "safe" callback field is changed to "unsafe" to
better reflect its new purpose. It has a Boolean "unsafe" parameter
to indicate whether the request is becoming unsafe or is now safe.
Because the "msg" parameter wasn't used, we drop that.
This resolves the original problem reportedin:
http://tracker.ceph.com/issues/4706
Reported-by: Yan, Zheng <zheng.z.yan@intel.com>
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Yan, Zheng <zheng.z.yan@intel.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Right now the data for a method call is specified via a pointer and
length, and it's copied--along with the class and method name--into
a pagelist data item to be sent to the osd. Instead, encode the
data in a data item separate from the class and method names.
This will allow large amounts of data to be supplied to methods
without copying. Only rbd uses the class functionality right now,
and when it really needs this it will probably need to use a page
array rather than a page list. But this simple implementation
demonstrates the functionality on the osd client, and that's enough
for now.
This resolves:
http://tracker.ceph.com/issues/4104
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Change the names of the functions that put data on a pagelist to
reflect that we're adding to whatever's already there rather than
just setting it to the one thing. Currently only one data item is
ever added to a message, but that's about to change.
This resolves:
http://tracker.ceph.com/issues/2770
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Finally! Convert the osd op data pointers into real structures, and
make the switch over to using them instead of having all ops share
the in and/or out data structures in the osd request.
Set up a new function to traverse the set of ops and release any
data associated with them (pages).
This and the patches leading up to it resolve:
http://tracker.ceph.com/issues/4657
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Still using the osd request r_data_in and r_data_out pointer, but
we're basically only referring to it via the data pointers in the
osd ops. And we're transferring that information to the request
or reply message only when the op indicates it's needed, in
osd_req_encode_op().
To avoid a forward reference, ceph_osdc_msg_data_set() was moved up
in the file.
Don't bother calling ceph_osd_data_init(), in ceph_osd_alloc(),
because the ops array will already be zeroed anyway.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
This ends up being a rather large patch but what it's doing is
somewhat straightforward.
Basically, this is replacing two calls with one. The first of the
two calls is initializing a struct ceph_osd_data with data (either a
page array, a page list, or a bio list); the second is setting an
osd request op so it associates that data with one of the op's
parameters. In place of those two will be a single function that
initializes the op directly.
That means we sort of fan out a set of the needed functions:
- extent ops with pages data
- extent ops with pagelist data
- extent ops with bio list data
and
- class ops with page data for receiving a response
We also have define another one, but it's only used internally:
- class ops with pagelist data for request parameters
Note that we *still* haven't gotten rid of the osd request's
r_data_in and r_data_out fields. All the osd ops refer to them for
their data. For now, these data fields are pointers assigned to the
appropriate r_data_* field when these new functions are called.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
All calls of ceph_osdc_start_request() are preceded (in the case of
rbd, almost) immediately by a call to ceph_osdc_build_request().
Move the build calls at the top of ceph_osdc_start_request() out of
there and into the ceph_osdc_build_request(). Nothing prevents
moving these calls to the top of ceph_osdc_build_request(), either
(and we're going to want them there in the next patch) so put them
at the top.
This and the next patch are related to:
http://tracker.ceph.com/issues/4657
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
This simply moves ceph_osdc_build_request() later in its source
file without any change. Done as a separate patch to facilitate
review of the change in the next patch.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An object class method is formatted using a pagelist which contains
the class name, the method name, and the data concatenated into an
osd request's outbound data.
Currently when a class op is initialized in osd_req_op_cls_init(),
the lengths of and pointers to these three items are recorded.
Later, when the op is getting formatted into the request message, a
new pagelist is created and that is when these items get copied into
the pagelist.
This patch makes it so the pagelist to hold these items is created
when the op is initialized instead.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request now holds all of its source op structures, and every
place that initializes one of these is in fact initializing one
of the entries in the the osd request's array.
So rather than supplying the address of the op to initialize, have
caller specify the osd request and an indication of which op it
would like to initialize. This better hides the details the
op structure (and faciltates moving the data pointers they use).
Since osd_req_op_init() is a common routine, and it's not used
outside the osd client code, give it static scope. Also make
it return the address of the specified op (so all the other
init routines don't have to repeat that code).
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An extent type osd operation currently implies that there will
be corresponding data supplied in the data portion of the request
(for write) or response (for read) message. Similarly, an osd class
method operation implies a data item will be supplied to receive
the response data from the operation.
Add a ceph_osd_data pointer to each of those structures, and assign
it to point to eithre the incoming or the outgoing data structure in
the osd message. The data is not always available when an op is
initially set up, so add two new functions to allow setting them
after the op has been initialized.
Begin to make use of the data item pointer available in the osd
operation rather than the request data in or out structure in
places where it's convenient. Add some assertions to verify
pointers are always set the way they're expected to be.
This is a sort of stepping stone toward really moving the data
into the osd request ops, to allow for some validation before
making that jump.
This is the first in a series of patches that resolve:
http://tracker.ceph.com/issues/4657
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There are fields "indata" and "indata_len" defined the ceph osd
request op structure. The "in" part is with from the point of view
of the osd server, but is a little confusing here on the client
side. Change their names to use "request" instead of "in" to
indicate that it defines data provided with the request (as opposed
the data returned in the response).
Rename the local variable in osd_req_encode_op() to match.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request keeps a pointer to the osd operations (ops) array
that it builds in its request message.
In order to allow each op in the array to have its own distinct
data, we will need to keep track of each op's data, and that
information does not go over the wire.
As long as we're tracking the data we might as well just track the
entire (source) op definition for each of the ops. And if we're
doing that, we'll have no more need to keep a pointer to the
wire-encoded version.
This patch makes the array of source ops be kept with the osd
request structure, and uses that instead of the version encoded in
the message in places where that was previously used. The array
will be embedded in the request structure, and the maximum number of
ops we ever actually use is currently 2. So reduce CEPH_OSD_MAX_OP
to 2 to reduce the size of the structure.
The result of doing this sort of ripples back up, and as a result
various function parameters and local variables become unnecessary.
Make r_num_ops be unsigned, and move the definition of struct
ceph_osd_req_op earlier to ensure it's defined where needed.
It does not yet add per-op data, that's coming soon.
This resolves:
http://tracker.ceph.com/issues/4656
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
One more osd data helper, which returns the length of the
data item, regardless of its type.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Define ceph_osd_data_init() and ceph_osd_data_release() to clean up
a little code.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Define and use functions that encapsulate the initializion of a
ceph_osd_data structure.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
This is a simple change, extracting the number of incoming data
bytes just once in handle_reply().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Hold off building the osd request message in ceph_writepages_start()
until just before it will be submitted to the osd client for
execution.
We'll still create the request and allocate the page pointer array
after we learn we have at least one page to write. A local variable
will be used to keep track of the allocated array of pages. Wait
until just before submitting the request for assigning that page
array pointer to the request message.
Create ands use a new function osd_req_op_extent_update() whose
purpose is to serve this one spot where the length value supplied
when an osd request's op was initially formatted might need to get
changed (reduced, never increased) before submitting the request.
Previously, ceph_writepages_start() assigned the message header's
data length because of this update. That's no longer necessary,
because ceph_osdc_build_request() will recalculate the right
value to use based on the content of the ops in the request.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Defer building the osd request until just before submitting it in
all callers except ceph_writepages_start(). (That caller will be
handed in the next patch.)
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
This patch moves the call to ceph_osdc_build_request() out of
ceph_osdc_new_request() and into its caller.
This is in order to defer formatting osd operation information into
the request message until just before request is started.
The only unusual (ab)user of ceph_osdc_build_request() is
ceph_writepages_start(), where the final length of write request may
change (downward) based on the current inode size or the oldest
snapshot context with dirty data for the inode.
The remaining callers don't change anything in the request after has
been built.
This means the ops array is now supplied by the caller. It also
means there is no need to pass the mtime to ceph_osdc_new_request()
(it gets provided to ceph_osdc_build_request()). And rather than
passing a do_sync flag, have the number of ops in the ops array
supplied imply adding a second STARTSYNC operation after the READ or
WRITE requested.
This and some of the patches that follow are related to having the
messenger (only) be responsible for filling the content of the
message header, as described here:
http://tracker.ceph.com/issues/4589
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Keep track of the length of the data portion for a message in a
separate field in the ceph_msg structure. This information has
been maintained in wire byte order in the message header, but
that's going to change soon.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
A field in an osd request keeps track of whether a connection is
currently filling the request's reply message. This patch gets rid
of that field.
An osd request includes two messages--a request and a reply--and
they're both associated with the connection that existed to its
the target osd at the time the request was created.
An osd request can be dropped early, even when it's in flight.
And at that time both messages are released. It's possible the
reply message has been supplied to its connection to receive
an incoming response message at the time the osd request gets
dropped. So ceph_osdc_release_request() revokes that message
from the connection before releasing it so things get cleaned up
properly.
Previously this may have caused a problem, because the connection
that a message was associated with might have gone away before the
revoke request. And to avoid any problems using that connection,
the osd client held a reference to it when it supplies its response
message.
However since this commit:
38941f80 libceph: have messages point to their connection
all messages hold a reference to the connection they are associated
with whenever the connection is actively operating on the message
(i.e. while the message is queued to send or sending, and when it
data is being received into it). And if a message has no connection
associated with it, ceph_msg_revoke_incoming() won't do anything
when asked to revoke it.
As a result, there is no need to keep an additional reference to the
connection associated with a message when we hand the message to the
messenger when it calls our alloc_msg() method to receive something.
If the connection *were* operating on it, it would have its own
reference, and if not, there's no work to be done when we need to
revoke it.
So get rid of the osd request's r_con_filling_msg field.
This resolves:
http://tracker.ceph.com/issues/4647
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There are two basically identical definitions of __decode_pgid()
in libceph, one in "net/ceph/osdmap.c" and the other in
"net/ceph/osd_client.c". Get rid of both, and instead define
a single inline version in "include/linux/ceph/osdmap.h".
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The osd client mutex is acquired just before getting a reference to
a request in handle_reply(). However the error paths after that
don't drop the mutex before returning as they should.
Drop the mutex after dropping the request reference. Also add a
bad_mutex label at that point and use it so the failed request
lookup case can be handled with the rest.
This resolves:
http://tracker.ceph.com/issues/4615
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Use osd_req_op_extent_init() in ceph_osdc_new_request() to
initialize the one or two ops built in that function.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
All callers of ceph_osd_new_request() pass either CEPH_OSD_OP_READ
or CEPH_OSD_OP_WRITE as the opcode value. The function assumes it
by filling in the extent fields in the ops array it builds. So just
assert that is the case, and don't bother calling op_has_extent()
before filling in the first osd operation in the array.
Define some local variables to gather the information to fill into
the first op, and then fill in the op array all in one place.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The ceph_osdc_new_request() an array of osd operations is built up
and filled in partially within that function and partially in the
called function calc_layout(). Move the latter part back out to
ceph_osdc_new_request() so it's all done in one place. This makes
it unnecessary to pass the op pointer to calc_layout(), so get rid
of that parameter.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The purpose of calc_layout() is to determine, given a file offset
and length and a layout describing the placement of file data across
objects, where in "object space" that data resides.
Specifically, it determines which object should hold the first part
of the specified range of file data, and the offset and length of
data within that object. The length will not exceed the bounds
of the object, and the caller is informed of that maximum length.
Add two parameters to calc_layout() to allow the object-relative
offset and length to be passed back to the caller.
This is the first steps toward having ceph_osdc_new_request() build
its osd op structure using osd_req_op_extent_init().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The rbd code has a function that allocates and populates a
ceph_osd_req_op structure (the in-core version of an osd request
operation). When reviewed, Josh suggested two things: that the
big varargs function might be better split into type-specific
functions; and that this functionality really belongs in the osd
client rather than rbd.
This patch implements both of Josh's suggestions. It breaks
up the rbd function into separate functions and defines them
in the osd client module as exported interfaces. Unlike the
rbd version, however, the functions don't allocate an osd_req_op
structure; they are provided the address of one and that is
initialized instead.
The rbd function has been eliminated and calls to it have been
replaced by calls to the new routines. The rbd code now now use a
stack (struct) variable to hold the op rather than allocating and
freeing it each time.
For now only the capabilities used by rbd are implemented.
Implementing all the other osd op types, and making the rest of the
code use it will be done separately, in the next few patches.
Note that only the extent, cls, and watch portions of the
ceph_osd_req_op structure are currently used. Delete the others
(xattr, pgls, and snap) from its definition so nobody thinks it's
actually implemented or needed. We can add it back again later
if needed, when we know it's been tested.
This (and a few follow-on patches) resolves:
http://tracker.ceph.com/issues/3861
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Define a separate function to determine the validity of an opcode,
and use it inside osd_req_encode_op() in order to unclutter that
function.
Don't update the destination op at all--and return zero--if an
unsupported or unrecognized opcode is seen in osd_req_encode_op().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
In ceph_osdc_build_request() there is a call to cpu_to_le16() which
provides a 64-bit value as its argument. Because of the implied
byte swapping going on it looked pretty suspect to me.
At the moment it turns out the behavior is well defined, but masking
off those bottom bits explicitly eliminates this distraction, and is
in fact more directly related to the purpose of the message header's
data_off field.
This resolves:
http://tracker.ceph.com/issues/4125
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd expects the transaction ids of arriving request messages from
a given client to a given osd to increase monotonically. So the osd
client needs to send its requests in ascending tid order.
The transaction id for a request is set at the time it is
registered, in __register_request(). This is also where the request
gets placed at the end of the osd client's unsent messages list.
At the end of ceph_osdc_start_request(), the request message for a
newly-mapped osd request is supplied to the messenger to be sent
(via __send_request()). If any other messages were present in the
osd client's unsent list at that point they would be sent *after*
this new request message.
Because those unsent messages have already been registered, their
tids would be lower than the newly-mapped request message, and
sending that message first can violate the tid ordering rule.
Rather than sending the new request only, send all queued requests
(including the new one) at that point in ceph_osdc_start_request().
This ensures the tid ordering property is preserved.
With this in place, all messages should now be sent in tid order
regardless of whether they're being sent for the first time or
re-sent as a result of a call to osd_reset().
This resolves:
http://tracker.ceph.com/issues/4392
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
In __map_request(), when adding a request to an osd client's unsent
list, add it to the tail rather than the head. That way the newest
entries (with the highest tid value) will be last.
Maintain an osd's request list in order of increasing tid also.
Finally--to be consistent--maintain an osd client's "notarget" list
in that order as well.
This partially resolves:
http://tracker.ceph.com/issues/4392
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
The osd expects incoming requests for a given object from a given
client to arrive in order, with the tid for each request being
greater than the tid for requests that have already arrived. This
patch fixes two places the osd client might not maintain that
ordering.
For the osd client, the connection fault method is osd_reset().
That function calls __reset_osd() to close and re-open the
connection, then calls __kick_osd_requests() to cause all
outstanding requests for the affected osd to be re-sent after
the connection has been re-established.
When an osd is reset, any in-flight messages will need to be
re-sent. An osd client maintains distinct lists for unsent and
in-flight messages. Meanwhile, an osd maintains a single list of
all its requests (both sent and un-sent). (Each message is linked
into two lists--one for the osd client and one list for the osd.)
To process an osd "kick" operation, the request list for the *osd*
is traversed, and each request is moved off whichever osd *client*
list it was on (unsent or sent) and placed onto the osd client's
unsent list. (It remains where it is on the osd's request list.)
When that is done, osd_reset() calls __send_queued() to cause each
of the osd client's unsent messages to be sent.
OK, with that background...
As the osd request list is traversed each request is prepended to
the osd client's unsent list in the order they're seen. The effect
of this is to reverse the order of these requests as they are put
(back) onto the unsent list.
Instead, build up a list of only the requests for an osd that have
already been sent (by checking their r_sent flag values). Once an
unsent request is found, stop examining requests and prepend the
requests that need re-sending to the osd client's unsent list.
Preserve the original order of requests in the process (previously
re-queued requests were reversed in this process). Because they
have already been sent, they will have lower tids than any request
already present on the unsent list.
Just below that, traverse the linger list in forward order as
before, but add them to the *tail* of the list rather than the head.
These requests get re-registered, and in the process are give a new
(higher) tid, so the should go at the end.
This partially resolves:
http://tracker.ceph.com/issues/4392
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
Since we no longer drop the request mutex between registering and
mapping an osd request in ceph_osdc_start_request(), there is no
chance of a race with kick_requests().
We can now therefore map and send the new request unconditionally
(but we'll issue a warning should it ever occur).
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
One of the first things ceph_osdc_start_request() does is register
the request. It then acquires the osd client's map semaphore and
request mutex and proceeds to map and send the request.
There is no reason the request has to be registered before acquiring
the map semaphore. So hold off doing so until after the map
semaphore is held.
Since register_request() is nothing more than a wrapper around
__register_request(), call the latter function instead, after
acquiring the request mutex.
That leaves register_request() unused, so get rid of it.
This partially resolves:
http://tracker.ceph.com/issues/4392
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
Use wrapper functions that check whether the auth op exists so that callers
do not need a bunch of conditional checks. Simplifies the external
interface.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Currently the messenger calls out to a get_authorizer con op, which will
create a new authorizer if it doesn't yet have one. In the meantime, when
we rotate our service keys, the authorizer doesn't get updated. Eventually
it will be rejected by the server on a new connection attempt and get
invalidated, and we will then rebuild a new authorizer, but this is not
ideal.
Instead, if we do have an authorizer, call a new update_authorizer op that
will verify that the current authorizer is using the latest secret. If it
is not, we will build a new one that does. This avoids the transient
failure.
This fixes one of the sorry sequence of events for bug
http://tracker.ceph.com/issues/4282
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
The osd trail is a pagelist, used only for a CALL osd operation
to hold the class and method names, along with any input data for
the call.
It is only currently used by the rbd client, and when it's used it
is the only bit of outbound data in the osd request. Since we
already support (non-trail) pagelist data in a message, we can
just save this outbound CALL data in the "normal" pagelist rather
than the trail, and get rid of the trail entirely.
The existing pagelist support depends on the pagelist being
dynamically allocated, and ownership of it is passed to the
messenger once it's been attached to a message. (That is to say,
the messenger releases and frees the pagelist when it's done with
it). That means we need to dynamically allocate the pagelist also.
Note that we simply assert that the allocation of a pagelist
structure succeeds. Appending to a pagelist might require a dynamic
allocation, so we're already assuming we won't run into trouble
doing so (we're just ignore any failures--and that should be fixed
at some point).
This resolves:
http://tracker.ceph.com/issues/4407
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Add support for recording a ceph pagelist as data associated with an
osd request.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The length of outgoing data in an osd request is dependent on the
osd ops that are embedded in that request. Each op is encoded into
a request message using osd_req_encode_op(), so that should be used
to determine the amount of outgoing data implied by the op as it
is encoded.
Have osd_req_encode_op() return the number of bytes of outgoing data
implied by the op being encoded, and accumulate and use that in
ceph_osdc_build_request().
As a result, ceph_osdc_build_request() no longer requires its "len"
parameter, so get rid of it.
Using the sum of the op lengths rather than the length provided is
a valid change because:
- The only callers of osd ceph_osdc_build_request() are
rbd and the osd client (in ceph_osdc_new_request() on
behalf of the file system).
- When rbd calls it, the length provided is only non-zero for
write requests, and in that case the single op has the
same length value as what was passed here.
- When called from ceph_osdc_new_request(), (it's not all that
easy to see, but) the length passed is also always the same
as the extent length encoded in its (single) write op if
present.
This resolves:
http://tracker.ceph.com/issues/4406
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
When an incoming message is destined for the osd client, the
messenger calls the osd client's alloc_msg method. That function
looks up which request has the tid matching the incoming message,
and returns the request message that was preallocated to receive the
response. The response message is therefore known before the
request is even started.
Between the start of the request and the receipt of the response,
the request and its data fields will not change, so there's no
reason we need to hold off setting them. In fact it's preferable
to set them just once because it's more obvious that they're
unchanging.
So set up the fields describing where incoming data is to land in a
response message at the beginning of ceph_osdc_start_request().
Define a helper function that sets these fields, and use it to
set the fields for both outgoing data in the request message and
incoming data in the response.
This resolves:
http://tracker.ceph.com/issues/4284
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Change it so we only assign outgoing data information for messages
if there is outgoing data to send.
This then allows us to add a few more (currently commented-out)
assertions.
This is related to:
http://tracker.ceph.com/issues/4284
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Greg Farnum <greg@inktank.com>
Define ceph_msg_data_set_pagelist(), ceph_msg_data_set_bio(), and
ceph_msg_data_set_trail() to clearly abstract the assignment of the
remaining data-related fields in a ceph message structure. Use the
new functions in the osd client and mds client.
This partially resolves:
http://tracker.ceph.com/issues/4263
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
When setting page array information for message data, provide the
byte length rather than the page count ceph_msg_data_set_pages().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Define a function ceph_msg_data_set_pages(), which more clearly
abstracts the assignment page-related fields for data in a ceph
message structure. Use this new function in the osd client and mds
client.
Ideally, these fields would never be set more than once (with
BUG_ON() calls to guarantee that). At the moment though the osd
client sets these every time it receives a message, and in the event
of a communication problem this can happen more than once. (This
will be resolved shortly, but setting up these helpers first makes
it all a bit easier to work with.)
Rearrange the field order in a ceph_msg structure to group those
that are used to define the possible data payloads.
This partially resolves:
http://tracker.ceph.com/issues/4263
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Record the byte count for an osd request rather than the page count.
The number of pages can always be derived from the byte count (and
alignment/offset) but the reverse is not true.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request defines information about where data to be read
should be placed as well as where data to write comes from.
Currently these are represented by common fields.
Keep information about data for writing separate from data to be
read by splitting these into data_in and data_out fields.
This is the key patch in this whole series, in that it actually
identifies which osd requests generate outgoing data and which
generate incoming data. It's less obvious (currently) that an osd
CALL op generates both outgoing and incoming data; that's the focus
of some upcoming work.
This resolves:
http://tracker.ceph.com/issues/4127
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request uses either pages or a bio list for its data. Use a
union to record information about the two, and add a data type
tag to select between them.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Pull the fields in an osd request structure that define the data for
the request out into a separate structure.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Currently ceph_osdc_new_request() assigns an osd request's
r_num_pages and r_alignment fields. The only thing it does
after that is call ceph_osdc_build_request(), and that doesn't
need those fields to be assigned.
Move the assignment of those fields out of ceph_osdc_new_request()
and into its caller. As a result, the page_align parameter is no
longer used, so get rid of it.
Note that in ceph_sync_write(), the value for req->r_num_pages had
already been calculated earlier (as num_pages, and fortunately
it was computed the same way). So don't bother recomputing it,
but because it's not needed earlier, move that calculation after the
call to ceph_osdc_new_request(). Hold off making the assignment to
r_alignment, doing it instead r_pages and r_num_pages are
getting set.
Similarly, in start_read(), nr_pages already holds the number of
pages in the array (and is calculated the same way), so there's no
need to recompute it. Move the assignment of the page alignment
down with the others there as well.
This and the next few patches are preparation work for:
http://tracker.ceph.com/issues/4127
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The purpose of ceph_calc_object_layout() is to fill in the pool
number and seed for a ceph_pg structure provided, based on a given
osd map and target object id.
Currently that function takes a file layout parameter, but the only
thing used out of that is its pool number.
Change the function so it takes a pool number rather than the full
file layout structure. Only update the ceph_pg if the pool is found
in the osd map. Get rid of few useless lines of code from the
function while there.
Since the function now very clearly just fills in the ceph_pg
structure it's provided, rename it ceph_calc_ceph_pg().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The new cases added to osd_req_encode_op() caused a new sparse
error, which highlighted an existing problem that had been
overlooked since it was originally checked in. When an unsupported
opcode is found the destination rather than the source opcode was
being used in the error message. The two differ in their byte
order, and we want to be using the one in the source.
Fix the problem in both spots.
Reported-by: Fengguang Wu <fengguang.wu@intel.com>
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request marked to linger will be re-submitted in the event
a connection to the target osd gets dropped. Currently, if there
is a callback function associated with a request it will be called
each time a request is submitted--which for lingering requests can
be more than once.
Change it so a request--including lingering ones--will get completed
(from the perspective of the user of the osd client) exactly once.
This resolves:
http://tracker.ceph.com/issues/3967
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The page alignment field for a request is currently set in
ceph_osdc_build_request(). It's not needed at that point
nor do either of its callers need that value assigned at
any point before they call ceph_osdc_start_request().
So move that assignment into ceph_osdc_start_request().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Use distinct fields for tracking the number of pages in a message's
page array and in a message's page list. Currently only one or the
other is used at a time, but that will be changing soon.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The only remaining reason to pass the osd request to calc_layout()
is to fill in its r_num_pages and r_page_alignment fields. Once it
fills those in, it doesn't do anything more with them.
We can therefore move those assignments into the caller, and get rid
of the "req" parameter entirely.
Note, however, that the only caller is ceph_osdc_new_request(),
and that immediately overwrites those fields with values based on
its passed-in page offset. So the assignment inside calc_layout()
was redundant anyway.
This resolves:
http://tracker.ceph.com/issues/4262
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Move the formatting of the object name (oid) to use for an object
request into the caller of calc_layout(). This makes the "vino"
parameter no longer necessary, so get rid of it.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Have calc_layout() pass the computed object number back to its
caller. (This is a small step to simplify review.)
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
If an invalid layout is provided to ceph_osdc_new_request(), its
call to calc_layout() might return an error. At that point in the
function we've already allocated an osd request structure, so we
need to free it (drop a reference) in the event such an error
occurs.
The only other value calc_layout() will return is 0, so make that
explicit in the successful case.
This resolves:
http://tracker.ceph.com/issues/4240
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Use the new version of the encoding for osd requests and replies. In the
process, update the way we are tracking request ops and reply lengths and
results in the struct ceph_osd_request. Update the rbd and fs/ceph users
appropriately.
The main changes are:
- we keep pointers into the request memory for fields we need to update
each time the request is sent out over the wire
- we keep information about the result in an array in the request struct
where the users can easily get at it.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Instead of using the old ceph_object_layout struct, update our internal
ceph_calc_object_layout method to use the ceph_pg type. This allows us to
pass the full 32-bit precision of the pgid.seed to the callers. It also
allows some callers to avoid reaching into the request structures for the
struct ceph_object_layout fields.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Always decode data into our cpu-native ceph_pg type that has the correct
field widths. Limit any remaining uses of ceph_pg_v1 to dealing with the
legacy protocol.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Rename the old version this type to distinguish it from the new version.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Add support for CEPH_OSD_OP_STAT operations in the osd client
and in rbd.
This operation sends no data to the osd; everything required is
encoded in identity of the target object.
The result will be ENOENT if the object doesn't exist. If it does
exist and no other error occurs the server returns the size and last
modification time of the target object as output data (in little
endian format). The size is a 64 bit unsigned and the time is
ceph_timespec structure (two unsigned 32-bit integers, representing
a seconds and nanoseconds value).
This resolves:
http://tracker.ceph.com/issues/4007
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Simplify the way the data length recorded in a message header is
calculated in ceph_osdc_build_request().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
In osd_req_encode_op() there are a few cases that handle osd
opcodes that are never used in the kernel. The presence of
this code gives the impression it's correct (which really can't
be assumed), and may impose some unnecessary restrictions on
some upcoming refactoring of this code.
So delete this effectively dead code, and report uses of the
previously handled cases as unsupported.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
If osd_req_encode_op() is given any opcode it doesn't recognize
it reports an error.
This patch fleshes out that routine to distinguish between
well-defined but unsupported values and values that are simply
bogus.
This and the next commit are related to:
http://tracker.ceph.com/issues/4126
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There are no actual users of ceph_osdc_wait_event(). This would
have been one-shot events, but we no longer support those so just
get rid of this function.
Since this leaves nothing else that waits for the completion of an
event, we can get rid of the completion in a struct ceph_osd_event.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is only one caller of ceph_osdc_create_event(), and it
provides 0 as its "one_shot" argument. Get rid of that argument and
just use 0 in its place.
Replace the code in handle_watch_notify() that executes if one_shot
is nonzero in the event with a BUG_ON() call.
While modifying "osd_client.c", give handle_watch_notify() static
scope.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is no caller of ceph_calc_raw_layout() outside of libceph, so
there's no need to export from the module.
Furthermore, there is only one caller, in calc_layout(), and it
is not much more than a simple wrapper for that function.
So get rid of ceph_calc_raw_layout() and embed it instead within
calc_layout().
While touching "osd_client.c", get rid of the unnecessary forward
declaration of __send_request().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The only callers of ceph_osdc_init() and ceph_osdc_stop()
ceph_create_client() and ceph_destroy_client() (respectively)
and they are in the same kernel module as those two functions.
There's therefore no need to export those interfaces, so don't.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Two of the three callers of the osd client's send_queued() function
already hold the osd client mutex and drop it before the call.
Change send_queued() so it assumes the caller holds the mutex, and
update all callers accordingly. Rename it __send_queued() to match
the convention used elsewhere in the file with respect to the lock.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The "num_reply" parameter to ceph_osdc_new_request() is never
used inside that function, so get rid of it.
Note that ceph_sync_write() passes 2 for that argument, while all
other callers pass 1. It doesn't matter, but perhaps someone should
verify this doesn't indicate a problem.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is only one caller of ceph_osdc_writepages(), and it always
passes 0 as its "flags" argument. Get rid of that argument and
replace its use in ceph_osdc_writepages() with 0.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is only one caller of ceph_osdc_writepages(), and it always
passes 0 as its "dosync" argument. Get rid of that argument and
replace its use in ceph_osdc_writepages() with 0.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is only one caller of ceph_osdc_writepages(), and it always
passes the value true as its "nofail" argument. Get rid of that
argument and replace its use in ceph_osdc_writepages() with the
constant value true.
This and a number of cleanup patches that follow resolve:
http://tracker.ceph.com/issues/4126
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
There is a check in the completion path for osd requests that
ensures the number of pages allocated is enough to hold the amount
of incoming data expected.
For bio requests coming from rbd the "number of pages" is not really
meaningful (although total length would be). So stop requiring that
nr_pages be supplied for bio requests. This is done by checking
whether the pages pointer is null before checking the value of
nr_pages.
Note that this value is passed on to the messenger, but there it's
only used for debugging--it's never used for validation.
While here, change another spot that used r_pages in a debug message
inappropriately, and also invalidate the r_con_filling_msg pointer
after dropping a reference to it.
This resolves:
http://tracker.ceph.com/issues/3875
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Currently, if the OSD client finds an osd request has had a bio list
attached to it, it drops a reference to it (or rather, to the first
entry on that list) when the request is released.
The code that added that reference (i.e., the rbd client) is
therefore required to take an extra reference to that first bio
structure.
The osd client doesn't really do anything with the bio pointer other
than transfer it from the osd request structure to outgoing (for
writes) and ingoing (for reads) messages. So it really isn't the
right place to be taking or dropping references.
Furthermore, the rbd client already holds references to all bio
structures it passes to the osd client, and holds them until the
request is completed. So there's no need for this extra reference
whatsoever.
So remove the bio_put() call in ceph_osdc_release_request(), as
well as its matching bio_get() call in rbd_osd_req_create().
This change could lead to a crash if old libceph.ko was used with
new rbd.ko. Add a compatibility check at rbd initialization time to
avoid this possibilty.
This resolves:
http://tracker.ceph.com/issues/3798 and
http://tracker.ceph.com/issues/3799
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Both ceph_osdc_alloc_request() and ceph_osdc_build_request() are
provided an array of ceph osd request operations. Rather than just
passing the number of operations in the array, the caller is
required append an additional zeroed operation structure to signal
the end of the array.
All callers know the number of operations at the time these
functions are called, so drop the silly zero entry and supply that
number directly. As a result, get_num_ops() is no longer needed.
This also means that ceph_osdc_alloc_request() never uses its ops
argument, so that can be dropped.
Also rbd_create_rw_ops() no longer needs to add one to reserve room
for the additional op.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Only one of the two callers of ceph_osdc_alloc_request() provides
page or bio data for its payload. And essentially all that function
was doing with those arguments was assigning them to fields in the
osd request structure.
Simplify ceph_osdc_alloc_request() by having the caller take care of
making those assignments
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The only thing ceph_osdc_alloc_request() really does with the
flags value it is passed is assign it to the newly-created
osd request structure. Do that in the caller instead.
Both callers subsequently call ceph_osdc_build_request(), so have
that function (instead of ceph_osdc_alloc_request()) issue a warning
if a request comes through with neither the read nor write flags set.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The osdc parameter to ceph_calc_raw_layout() is not used, so get rid
of it. Consequently, the corresponding parameter in calc_layout()
becomes unused, so get rid of that as well.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
A snapshot id must be provided to ceph_calc_raw_layout() even though
it is not needed at all for calculating the layout.
Where the snapshot id *is* needed is when building the request
message for an osd operation.
Drop the snapid parameter from ceph_calc_raw_layout() and pass
that value instead in ceph_osdc_build_request().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
ceph_calc_file_object_mapping() takes (among other things) a "file"
offset and length, and based on the layout, determines the object
number ("bno") backing the affected portion of the file's data and
the offset into that object where the desired range begins. It also
computes the size that should be used for the request--either the
amount requested or something less if that would exceed the end of
the object.
This patch changes the input length parameter in this function so it
is used only for input. That is, the argument will be passed by
value rather than by address, so the value provided won't get
updated by the function.
The value would only get updated if the length would surpass the
current object, and in that case the value it got updated to would
be exactly that returned in *oxlen.
Only one of the two callers is affected by this change. Update
ceph_calc_raw_layout() so it records any updated value.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The len argument to ceph_osdc_build_request() is set up to be
passed by address, but that function never updates its value
so there's no need to do this. Tighten up the interface by
passing the length directly.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Since every osd message is now prepared to include trailing data,
there's no need to check ahead of time whether any operations will
make use of the trail portion of the message.
We can drop the second argument to get_num_ops(), and as a result we
can also get rid of op_needs_trail() which is no longer used.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
An osd request structure contains an optional trail portion, which
if present will contain data to be passed in the payload portion of
the message containing the request. The trail field is a
ceph_pagelist pointer, and if null it indicates there is no trail.
A ceph_pagelist structure contains a length field, and it can
legitimately hold value 0. Make use of this to change the
interpretation of the "trail" of an osd request so that every osd
request has trailing data, it just might have length 0.
This means we change the r_trail field in a ceph_osd_request
structure from a pointer to a structure that is always initialized.
Note that in ceph_osdc_start_request(), the trail pointer (or now
address of that structure) is assigned to a ceph message's trail
field. Here's why that's still OK (looking at net/ceph/messenger.c):
- What would have resulted in a null pointer previously will now
refer to a 0-length page list. That message trail pointer
is used in two functions, write_partial_msg_pages() and
out_msg_pos_next().
- In write_partial_msg_pages(), a null page list pointer is
handled the same as a message with 0-length trail, and both
result in a "in_trail" variable set to false. The trail
pointer is only used if in_trail is true.
- The only other place the message trail pointer is used is
out_msg_pos_next(). That function is only called by
write_partial_msg_pages() and only touches the trail pointer
if the in_trail value it is passed is true.
Therefore a null ceph_msg->trail pointer is equivalent to a non-null
pointer referring to a 0-length page list structure.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
The last two parameters to ceph_osd_build_request() describe the
object id, but the values passed always come from the osd request
structure whose address is also provided. Get rid of those last
two parameters.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
Reformat __reset_osd() into three distinct blocks of code
handling the three return cases.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Josh Durgin <josh.durgin@inktank.com>
When ceph_osdc_handle_map() is called to process a new osd map,
kick_requests() is called to ensure all affected requests are
updated if necessary to reflect changes in the osd map. This
happens in two cases: whenever an incremental map update is
processed; and when a full map update (or the last one if there is
more than one) gets processed.
In the former case, the kick_requests() call is followed immediately
by a call to reset_changed_osds() to ensure any connections to osds
affected by the map change are reset. But for full map updates
this isn't done.
Both cases should be doing this osd reset.
Rather than duplicating the reset_changed_osds() call, move it into
the end of kick_requests().
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The kick_requests() function is called by ceph_osdc_handle_map()
when an osd map change has been indicated. Its purpose is to
re-queue any request whose target osd is different from what it
was when it was originally sent.
It is structured as two loops, one for incomplete but registered
requests, and a second for handling completed linger requests.
As a special case, in the first loop if a request marked to linger
has not yet completed, it is moved from the request list to the
linger list. This is as a quick and dirty way to have the second
loop handle sending the request along with all the other linger
requests.
Because of the way it's done now, however, this quick and dirty
solution can result in these incomplete linger requests never
getting re-sent as desired. The problem lies in the fact that
the second loop only arranges for a linger request to be sent
if it appears its target osd has changed. This is the proper
handling for *completed* linger requests (it avoids issuing
the same linger request twice to the same osd).
But although the linger requests added to the list in the first loop
may have been sent, they have not yet completed, so they need to be
re-sent regardless of whether their target osd has changed.
The first required fix is we need to avoid calling __map_request()
on any incomplete linger request. Otherwise the subsequent
__map_request() call in the second loop will find the target osd
has not changed and will therefore not re-send the request.
Second, we need to be sure that a sent but incomplete linger request
gets re-sent. If the target osd is the same with the new osd map as
it was when the request was originally sent, this won't happen.
This can be fixed through careful handling when we move these
requests from the request list to the linger list, by unregistering
the request *before* it is registered as a linger request. This
works because a side-effect of unregistering the request is to make
the request's r_osd pointer be NULL, and *that* will ensure the
second loop actually re-sends the linger request.
Processing of such a request is done at that point, so continue with
the next one once it's been moved.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
In kick_requests(), we need to register the request before we
unregister the linger request. Otherwise the unregister will
reset the request's osd pointer to NULL.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The red-black node in the ceph osd request structure is initialized
in ceph_osdc_alloc_request() using rbd_init_node(). We do need to
initialize this, because in __unregister_request() we call
RB_EMPTY_NODE(), which expects the node it's checking to have
been initialized. But rb_init_node() is apparently overkill, and
may in fact be on its way out. So use RB_CLEAR_NODE() instead.
For a little more background, see this commit:
4c199a93 rbtree: empty nodes have no color"
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The red-black node node in the ceph osd event structure is not
initialized in create_osdc_create_event(). Because this node can
be the subject of a RB_EMPTY_NODE() call later on, we should ensure
the node is initialized properly for that.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The red-black node node in the ceph osd structure is not initialized
in create_osd(). Because this node can be the subject of a
RB_EMPTY_NODE() call later on, we should ensure the node is
initialized properly for that. Add a call to RB_CLEAR_NODE()
initialize it.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
In __unregister_linger_request(), the request is being removed
from the osd client's req_linger list only when the request
has a non-null osd pointer. It should be done whether or not
the request currently has an osd.
This is most likely a non-issue because I believe the request
will always have an osd when this function is called.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
If an osd has no requests and no linger requests, __reset_osd()
will just remove it with a call to __remove_osd(). That drops
a reference to the osd, and therefore the osd may have been free
by the time __reset_osd() returns. That function offers no
indication this may have occurred, and as a result the osd will
continue to be used even when it's no longer valid.
Change__reset_osd() so it returns an error (ENODEV) when it
deletes the osd being reset. And change __kick_osd_requests() so it
returns immediately (before referencing osd again) if __reset_osd()
returns *any* error.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
In __unregister_request(), there is a call to list_del_init()
referencing a request that was the subject of a call to
ceph_osdc_put_request() on the previous line. This is not
safe, because the request structure could have been freed
by the time we reach the list_del_init().
Fix this by reversing the order of these lines.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-off-by: Sage Weil <sage@inktank.com>
This would reset a connection with any OSD that had an outstanding
request that was taking more than N seconds. The idea was that if the
OSD was buggy, the client could compensate by resending the request.
In reality, this only served to hide server bugs, and we haven't
actually seen such a bug in quite a while. Moreover, the userspace
client code never did this.
More importantly, often the request is taking a long time because the
OSD is trying to recover, or overloaded, and killing the connection
and retrying would only make the situation worse by giving the OSD
more work to do.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
If we are creating an osd request and get an invalid layout, return
an EINVAL to the caller. We switch up the return to have an error
code instead of NULL implying -ENOMEM.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
If we encounter an invalid (e.g., zeroed) mapping, return an error
and avoid a divide by zero.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
There are many (normal) conditions that can lead to us getting
unexpected replies, include cluster topology changes, osd failures,
and timeouts. There's no need to spam the console about it.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
This is a trivial fix for the debug output, as it is inconsistent
with the function name so may confuse people when debugging.
[elder@inktank.com: switched to use __func__]
Signed-off-by: Jiaju Zhang <jjzhang@suse.de>
Reviewed-by: Alex Elder <elder@inktank.com>
The linger op registration (i.e., watch) modifies the object state. As
such, the OSD will reply with success if it has already applied without
doing the associated side-effects (setting up the watch session state).
If we lose the ACK and resubmit, we will see success but the watch will not
be correctly registered and we won't get notifies.
To fix this, always resubmit the linger op with a new tid. We accomplish
this by re-registering as a linger (i.e., 'registered') if we are not yet
registered. Then the second loop will treat this just like a normal
case of re-registering.
This mirrors a similar fix on the userland ceph.git, commit 5dd68b95, and
ceph bug #2796.
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Reviewed-by: Yehuda Sadeh <yehuda@inktank.com>
These don't strictly need to be initialized based on how they are used, but
it is good practice to do so.
Reported-by: Alex Elder <elder@inktank.com>
Signed-off-by: Sage Weil <sage@inktank.com>
Initialize the type field for messages in a msgpool. The caller was doing
this for osd ops, but not for the reply messages.
Reported-by: Alex Elder <elder@inktank.com>
Signed-off-by: Sage Weil <sage@inktank.com>
ceph_con_revoke_message() is passed both a message and a ceph
connection. A ceph_msg allocated for incoming messages on a
connection always has a pointer to that connection, so there's no
need to provide the connection when revoking such a message.
Note that the existing logic does not preclude the message supplied
being a null/bogus message pointer. The only user of this interface
is the OSD client, and the only value an osd client passes is a
request's r_reply field. That is always non-null (except briefly in
an error path in ceph_osdc_alloc_request(), and that drops the
only reference so the request won't ever have a reply to revoke).
So we can safely assume the passed-in message is non-null, but add a
BUG_ON() to make it very obvious we are imposing this restriction.
Rename the function ceph_msg_revoke_incoming() to reflect that it is
really an operation on an incoming message.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
ceph_con_revoke() is passed both a message and a ceph connection.
Now that any message associated with a connection holds a pointer
to that connection, there's no need to provide the connection when
revoking a message.
This has the added benefit of precluding the possibility of the
providing the wrong connection pointer. If the message's connection
pointer is null, it is not being tracked by any connection, so
revoking it is a no-op. This is supported as a convenience for
upper layers, so they can revoke a message that is not actually
"in flight."
Rename the function ceph_msg_revoke() to reflect that it is really
an operation on a message, not a connection.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
The function ceph_alloc_msg() is only used to allocate a message
that will be assigned to a connection's in_msg pointer. Rename the
function so this implied usage is more clear.
In addition, make that assignment inside the function (again, since
that's precisely what it's intended to be used for). This allows us
to return what is now provided via the passed-in address of a "skip"
variable. The return type is now Boolean to be explicit that there
are only two possible outcomes.
Make sure the result of an ->alloc_msg method call always sets the
value of *skip properly.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Move the initialization of a ceph connection's private pointer,
operations vector pointer, and peer name information into
ceph_con_init(). Rearrange the arguments so the connection pointer
is first. Hide the byte-swapping of the peer entity number inside
ceph_con_init()
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
There were a few direct calls to ceph_con_{get,put}() instead of the con
ops from osd_client.c. This is a bug since those ops aren't defined to
be ceph_con_get/put.
This breaks refcounting on the ceph_osd structs that contain the
ceph_connections, and could lead to all manner of strangeness.
The purpose of the ->get and ->put methods in a ceph connection are
to allow the connection to indicate it has a reference to something
external to the messaging system, *not* to indicate something
external has a reference to the connection.
[elder@inktank.com: added that last sentence]
Signed-off-by: Sage Weil <sage@newdream.net>
Reviewed-by: Alex Elder <elder@inktank.com>
In ceph_osdc_release_request(), a reference to the r_reply message
is dropped. But just after that, that same message is revoked if it
was in use to receive an incoming reply. Reorder these so we are
sure we hold a reference until we're actually done with the message.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Pass the osd number to the create_osd() routine, and move the
initialization of fields that depend on it therein.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
A ceph client has a pointer to a ceph messenger structure in it.
There is always exactly one ceph messenger for a ceph client, so
there is no need to allocate it separate from the ceph client
structure.
Switch the ceph_client structure to embed its ceph_messenger
structure.
Signed-off-by: Alex Elder <elder@inktank.com>
Reviewed-by: Yehuda Sadeh <yehuda@inktank.com>
Reviewed-by: Sage Weil <sage@inktank.com>
Pull ceph updates from Sage Weil:
"There are some updates and cleanups to the CRUSH placement code, a bug
fix with incremental maps, several cleanups and fixes from Josh Durgin
in the RBD block device code, a series of cleanups and bug fixes from
Alex Elder in the messenger code, and some miscellaneous bounds
checking and gfp cleanups/fixes."
Fix up trivial conflicts in net/ceph/{messenger.c,osdmap.c} due to the
networking people preferring "unsigned int" over just "unsigned".
* git://git.kernel.org/pub/scm/linux/kernel/git/sage/ceph-client: (45 commits)
libceph: fix pg_temp updates
libceph: avoid unregistering osd request when not registered
ceph: add auth buf in prepare_write_connect()
ceph: rename prepare_connect_authorizer()
ceph: return pointer from prepare_connect_authorizer()
ceph: use info returned by get_authorizer
ceph: have get_authorizer methods return pointers
ceph: ensure auth ops are defined before use
ceph: messenger: reduce args to create_authorizer
ceph: define ceph_auth_handshake type
ceph: messenger: check return from get_authorizer
ceph: messenger: rework prepare_connect_authorizer()
ceph: messenger: check prepare_write_connect() result
ceph: don't set WRITE_PENDING too early
ceph: drop msgr argument from prepare_write_connect()
ceph: messenger: send banner in process_connect()
ceph: messenger: reset connection kvec caller
libceph: don't reset kvec in prepare_write_banner()
ceph: ignore preferred_osd field
ceph: fully initialize new layout
...
There is a race between two __unregister_request() callers: the
reply path and the ceph_osdc_wait_request(). If we get a reply
*and* the timeout expires at roughly the same time, both callers
will try to unregister the request, and the second one will do bad
things.
Simply check if the request is still already unregistered; if so,
return immediately and do nothing.
Fixes http://tracker.newdream.net/issues/2420
Signed-off-by: Sage Weil <sage@inktank.com>
Reviewed-by: Alex Elder <elder@inktank.com>
Ilya Dryomov