We do technically support initiators without eventq or
controlq, but the lun hotplug/hotremove path expected the
eventq to be always present.
This was causing vhost to randomly crash in the fuzz tests.
Specifically, the crash happened if lun hotplug was handled
while a VM was in the middle of switching from BIOS to OS.
We fix it by checking if eventq is set before putting
any event there. LUN hotplug and hotremove won't work
without an eventq, but the entire session will be restarted
after new queues are initialized. This will make the VM
retrieve all up-to-date luns after OS initialization is
complete.
Change-Id: I5d28cbedad8fb2a35ede5a491aeb7fdc52faad06
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451789
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_dma_malloc() is not required here, as the session
array is nether DMA-able nor shared between processes.
Change-Id: Ia9218d00eea8a3207abf07c1d78eaa585ffcad2c
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450552
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
logical_block_size and geometry_logical_block_size are uint32_t,
so correct conversion for endianness is used now.
Change-Id: Iaa5ff576da8bc4ca65fa21f28eb3afd30507b00a
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451867
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
We were setting this value in the target from our initiator, but it
turns out the rdma_conn_params struct is responsible for setting the
opposite side so we need to add it in the target side when accepting
connections.
Also, add a test to demonstrate target functionality when we overwhelm
the SRQ. It is useful to note that performance really tanks when you
start overwhelming the srq so it may be useful to use this test case to
check performance gains in edge cases over time.
Change-Id: Iac541bd9fc1d82eca9f21e7abc3f625663a6c460
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451678
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This will allow us to test a greater number of connections without
having to add more subsystems when doing NVMe-oF testing.
Change-Id: I33203d6db79b30abb065f098c16840096478c5de
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451677
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Before enabling SPDK thread scheduler,
each POSIX thread name was reactor_% (% was the core number).
After enabling SPDK thread scheduler, the name of the master POSIX
thread is reactor_% (% is the max core number), and the name of the
slave posix threads are lcore-slave-% (% is the current core number).
SPDK threads are light-weight threads - sometimes also called
green threads or fibers, and so are independent from posix
threads.
But reactor is tied to the POSIX thread in the SPDK event library.
So SPDK thread doesn't rename the POSIX thread at its creation
but reactor renames the POSIX thread at start instead.
This change makes POSIX thread name compatible between before and
after enabling SPDK thread scheduler.
Change-Id: I26e8dabc73e163c9f74e18b3640cf54954603b1f
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451712
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This signals which RocksDB commit should be checked
out for the SPDK RocksDB tests.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I1ba0be00747a2642b359b1e0e0c8c2c6d99cc4f1
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451772
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This will be used in next patch to ensure synchronous
threads wait until any background operations are done
before freeing its request memory, and also to print
errors if internal channels have outstanding
requests when freed.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I24bc8096f84b70270f87b051f9034ed08e639500
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451780
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
All error paths must free opts->nodebase and opts->authfile.
Change-Id: I655f112dd36bbd0dca6050bc5cc3ade1a5b05b7d
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451770
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Allow users to optionally specify an affinity mask when
creating a thread. This isn't currently used, but we want
the API to be in its final form for the next release.
Change-Id: I7bd05e921ece6d8d5f61775bd14286f6a58f267f
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451683
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
For bdevs that don't support zero copy, emulate the
API using regular read and write operations.
Change-Id: Iabd7ff474bf740b096f38bd47196987cbd89e915
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/416465
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add a ZCOPY operation to obtain buffers that represent
data regions on the backing block device.
Change-Id: Ie941c16ee051d0009e3888b52b8f41773bba47b3
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/386166
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Bdev descriptors could be closed only from the same thread
that opened them. This restriction was suddenly introduced
at one point without making sure all the SPDK code respects
it. Vhost can still close descriptors from any arbitrary
thread and fixing that would require some more effort.
With this patch we remove the thread-specific assert from
spdk_bdev_close() and hence allow vhost to work properly
in debug builds. Vhost can still have a possible data
race with bdev hotremove notification, but let's get rid
of the abort() from the usual code path first.
Change-Id: I6fac66a5ebc907b1c5418fff618f0b64cd9b69f4
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451561
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
We used to wait only for those descriptors which
specified the hotremove notification callback. The
bdev could've been removed before the descriptor
was closed and the subsequent spdk_bdev_close would
simply segfault.
This patch modifies spdk_bdev_unregister to always
wait for all descriptors to be closed before actually
unregistering the bdev. This consolidates the bdev
unregister behavior for descriptors with and without
the hotremove callback.
Change-Id: I9b358209c6abd301b6fe8660e27bc6fa4ef485d6
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450175
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Make the use of spdk_uuid_compare() to be consistent in the file,
also change the SPDK_INFOLOG to SPDK_DEBUGLOG to avoid the
repeated log messages for RESERVATION CONFLICT response.
Change-Id: I72fefbd520cefcaf25182c3ca3d21e3d87d17e94
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450884
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Now Host can get an asynchronous event notification when
registrants were unregistered/preempted or reservation was
released from the associate namespace, Host can send
get log page to clear related log pages and reservation
report to get the full overview of current reservation
configuration.
Change-Id: Idc57c19812490c7536503308989871515e9f2361
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/439935
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The mode of dmb oshld can guarantees cpu sequential execution,
which has less impact on performance.
Change-Id: If30b6a682a2216eecd1da039267ed4f5471afc38
Signed-off-by: h00448672 <heyang18@huawei.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/446827
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Previously, when creating a snapshot in blobstore the snapshot's cluster map
was copied from the "original" blob, with the original's map zeroed. These
operations are both O(num_clusters*cluster_size/page_size) while io
operations are frozen. This change replaces the linear operation with an
O(1) pointer swap at the critical moment that io is frozen, while
doing the zeroing before the freeze when preparing the snapshot to
minimize freeze time.
Change-Id: I1e468bc97623f5da161a8ddba1393c271acd3aed
Signed-off-by: Amir More <habeanf@gmail.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451486
Reviewed-by: Maciej Szwed <maciej.szwed@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is the end of the patch series. After this patch,
delete_target_node RPC will wait for the completion of
removal of the SCSI device and then free the iSCSI target.
SCSI device holds passed callback and calls it in free_dev().
free_dev() is ensured to be called after all iSCSI sessions
are closed. So iSCSI target resource can be freed safely
after that.
Change-Id: I25921b4014207092b7b3845dfeae58bcdffa2edc
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450607
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This commit adds functionality for installing most of the packages,
which were previously installed for Fedora only,
for Ubuntu as well
Some packages are not going to be installed by the script:
For ubuntu 16
* libpmempool
* open-isns-utils
* perl-open
For ubuntu 18
* perl-open
Remaining packages deserve their own patches since they are not so easy to
install on Ubuntu
Change-Id: Ic5b744193ff4df46ab4e4fb3783e0515ccdbfecb
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/425388
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Paweł Niedźwiecki <pawelx.niedzwiecki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
spdk_dma_malloc() is not required here, as the device
object is neither DMA-able nor shared between processes.
The device structures used to be aligned to cache line
size, but that's just a leftover from before sessions
were introduced. The device object is just a generic
device information that can be accessed from any thread
holding the proper mutex. The hot data used in the I/O
path sits in the session structure, which is now allocated
with posix_memalloc() to ensure proper alignment.
Vhost NVMe is an exception, as the device struct is used
as hot I/O data for the one and only session it supports,
so it's also allocated with posix_memalloc().
While here, also allocate various vhost buffers using
spdk_zmalloc() instead of spdk_dma_zmalloc(), as
spdk_dma_*malloc() is about to be deprecated.
Change-Id: Ic7f63185639b7b98dc1ef756166c826a0af87b44
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450551
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
They used to be allocated with spdk_dma_zmalloc() which did
provide zeroed memory, but we unintentionally changed that
when switching to posix_memalign.
The structure might have some unitialized memory, so with
this patch we just memset it right after allocating it.
Change-Id: Id5a5685e09419901513925abaeed605c36f5199a
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451546
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
blob_delete_cb() takes spdk_fs_request as the input
parameter.
Change-Id: Ie6150e7d31d187296a448e82784e2ac2fecfe52c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450895
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
fs_truncate_complete_cb() takes spdk_fs_request as
the input parameter.
Change-Id: I0413fecd30a1f7cb6528b502848fa999454ba638
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450891
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
For existing filesystem APIs, SPDK can only support
1 data buffer, we add the iovec data structure here
so that it can support multiple buffer vectors in
following patches.
Change-Id: I26984f3ea985f349a5016060e0801e3989ce2fc6
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450722
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We did this a while ago for login notifications - do
it for logout as well.
While here, just use SPDK_DEBUGLOG instead of printing
to a buffer first.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I77f566d139cd818428371ec887efeca6eee08898
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451062
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Without this, spdkcli crashes when refreshing if any iSCSI
connections are established.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I82cd7a98e8c7d48e4730a7bd228ed483bdac28ef
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451060
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
The next patch will add the function ponter typedef
spdk_scsi_dev_destruct_cb for SCSI device destruction.
Hence add lun to the names of descriptors and callback for SCSI
LUN for clarification.
This patch doesn't change any behavior.
Change-Id: I73f2bce9129f7a6f16770ab6ed18428b16589108
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450883
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
This patch changes iscsi_tgt_node_destruct to be asynchronous.
spdk_iscsi_shutdown_tgt_node_by_name() uses the callback
provided by iscsi_tgt_node_destruct(), but
spdk_iscsi_shutdown_tgt_nodes() doesn't use it.
The reason is that during shutdown all sessions are terminated
already, and spdk_iscsi_shutdown_tgt_nodes() doesn't need to
wait for the completion of iscsi_tgt_node_destruct() by using
the callback. iscsi_tgt_node_destruct() will complete immediately
in this case.
spdk_scsi_dev_destruct() is not asynchronized yet and so
iscsi_tgt_node_destruct() calls its callback directly in this patch.
The next patch will replace the call.
Change-Id: I86911c5297a93560551e0404d6ea85f49d647c22
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450606
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This patch changes spdk_rpc_delete_target_node() to pass
rpc_delete_target_node_done() and its context to
spdk_iscsi_shutdown_tgt_node_by_name().
iscsi_tgt_node_destruct() is not asynchronized yet and so
spdk_iscsi_shutdown_tgt_node_by_name() calls the callback
passed from the caller directly for now.
The next patch will replace the call.
Change-Id: Ide2d9fcc6738e02de19f91249c21ac2d0d37138a
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450605
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
delete_target_node RPC had not waited for completion of
spdk_iscsi_shutdown_tgt_node_by_name. This implementation
is not the direct cause of the deadlock issue in
spdk_iscsi_conn_stop() but changing delete_target_node RPC
to wait for completion of spdk_iscsi_shutdown_tgt_node_by_name
will prevent future errors.
spdk_iscsi_shutdown_tgt_node_by_name() is not asynchronized yet
and so rpc_delete_target_node calls rpc_delete_target_node_done
directly in this patch.
The next patch will replace the call.
Change-Id: I6a042028b93c36ca9f68e3b212141e101b33d394
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450604
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
When any iSCSI target is destructed, if the target exits all corresponding
connections first, destructing SCSI device will be easier.
Hence, iscsi_tgt_node_destruct() starts exiting all corresponding
connections. Then it destructs SCSI device immediately if no active
active connections, or waits for the completion if there is any
active connection for the target.
Change-Id: Ibd4a29789faecfefccefa1153a519c43d040a00d
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450737
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
In iscsi_conns_cleanup(), g_conns_array was unmapped but was not
invalidated by setting MAP_FAILED.
So, find_iscsi_connection_by_id() caused segmentation fault if
it is called after iscsi_conns_cleanup().
Change-Id: Ib91c9240c62c2aaa32713dd4aa382d31e5ea2eed
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450901
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Disabled temporarily earlier to get to basic functionality, circling
back now to begin work on UT again.
Change-Id: Ie7606f91072257f392727bdecc5f1eac26380453
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451063
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
In case of a DSM Deallocate (unmap) with multiple ranges, individual
bdev IOs are submitted for each range. If the bdev IO cannot be
allocated, the request is queued on io_wait_queue; however previously
submitted ranges may complete before memory is available for the next
range. In such a case, the completion callback will free unmap_ctx,
while the request is still queued for memory - causing a segfault
when the request is dequeued. To fix, introduce a new field tracking
the unmap ranges, and make sure the count is nonzero when the request
is queued for memory.
Signed-off-by: Yair Elharrar <yair@excelero.com>
Change-Id: Ifcac018f14af5ca408c7793ca9543c1e2d63b777
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447542
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The previous patches described as optimizations also
fixed some issues. They seem sufficient to cover all
the error cases, but the real source of the problem
lies in foreach_session() initiated by the device backend,
which can use sessions that were never seen by the
backend.
The backends are only notified when a session is
*started*, but foreach_session() iterates through
all the sessions - even those that were never started.
Vhost SCSI, for example, in the foreach_session() callbacks
used to expect svsession->svdev to be always set, but
that field is only set when the session gets started.
A perfect solution would to introduce a new backend
callback to be called on new connection. Vhost SCSI
could set e.g. svsession->svdev inside. For now we go
with much easier solution that prevents sessions from
being used in foreach-session() unless they were
started at least once. (...and e.g. got their ->svdev set)
Change-Id: Ida30a1f27f99977360d08a71a64fc92931b25b75
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449394
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Before SCSI target is removed, all vhost sessions need
to drain their pending I/O and put their I/O channels.
After a session puts it channel, it sends an async
notification to the entire vhost device. The device
will check if there are any other sessions still
referencing the SCSI target and if not - it will
continue removing the spdk_scsi_dev object. There may
be multiple sessions sending those async events at the
same time, and while we do protect from removing the
same spdk_scsi_dev twice, we can still remove
a different spdk_scsi_dev that was hot-attached in the
meantime with the same target ID.
1. SCSI target hotremove (e.g. via RPC or bdev hotremove)
/ \
/ \
session A session B
drain I/O drain I/O
| |
v |
done v
send event done
\ send event*
\
All sessions have detached the SCSI target, remove
it from the entire vhost device. From this point
a new target can be hot-attached (e.g. via RPC).
2. Attach a SCSI target with with same target ID.
3. Hotremove event* from the previous SCSI target gets
finally executed. SCSI target with that ID is
occupied (again) and may be hotremoved by mistake.
The role of that hotremove event is just to kick the
vhost device and make it remove any scsi targets that
can be removed, so add a check preventing it from
removing devices in states other than REMOVING.
Change-Id: Ia1cc7cae797fd8859d485e63f0ef37aeac2945d0
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449990
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Always unset the VHOST_SCSI_DEV_REMOVED status on
session stop, so that we won't send hotremove SCSI
sense codes after e.g. a VM gets rebooted. The VM
should generally enumerate the SCSI devices again
in such case. We already unset the REMOVED status
for devices which were still attached at the time
of the session stop, but the devices hotremoved
before the session stop retained their REMOVED
status, giving us inconsistent behavior.
Change-Id: I7c5876e29f4bdc99cc060f1d891e24ac57051f37
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449709
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
