With this patch each target has separate suite,
tests are ran just once for all setup suites.
Suite init/fini functions were added that move
current g_current_io_target ahead for next suite.
Results what works for particular bdev are of more use
rather than combination for all bdevs.
This will come useful when we want to perform RPC for only
select targets.
In case of CUnit library error, application is stopped without
returning number of failures. This is because at that point
its numbers are not reliable.
Change-Id: Ie82b6e33374841bf829da09a16a401fddd1fd81c
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455199
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>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
We flush a cache buffer once it's filled. When the write
for that cache buffer has completed, we look to see if
there's more data to flush. Currently if there's *any*
more data to flush, we will flush it, even if it's not
a full buffer.
That can hurt performance though. Ideally we only want
to flush a partial buffer if there's been an explicit
sync operation that requires that partial buffer to be
flushed. Otherwise we will end up writing the partial
buffer to disk, and then come back and write that data
again later when the buffer is full.
Add a new unit test to test for this condition. This
patch breaks one of the existing unit tests which was
designed specifically around a RocksDB failure condition.
Change that file_length unit test to now write exactly
one CACHE_BUFFER, which still tests the general logic
making sure that we don't confuse the amount of data
flushed with the value written to the file's length
xattr.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I83795fb45afe854b38648d0e0c1a7928219307a2
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455698
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Data can get implicitly flushed as cache buffers are filled. But
the length xattr is only written in response to a sync or close
operation. So we cannot just look at the amount of data flushed,
and ignore the sync operation if all of the data written has been
flushed - we still need to write the length xattr.
This also adds a unit test which reproduces the original problem.
Fixes issue #297.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Icca6ef4d1544f72e9bc31c4ee77d26b4b7f0cce4
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455692
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Interpret volume->open() options as our vbdev_ocf_base struct.
This is used during metadata probe procedure, when there is no
vbdev configurations created, so we need to pass base structure
as option.
This patch is a prepartion for persistant metadata support,
and it does not change current behavior.
Change-Id: I0b7435df1692d8b3028931c6c9fc50d2d84b2557
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455415
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>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
We don't provide windows executable for this test.
Change-Id: Iedaafba06662f776c102528f8073b6c3d56bb323
Signed-off-by: Pawel Kaminski <pawelx.kaminski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450406
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Hotremoving a bdev exposed as a vhost-scsi LUN which
has an allocated io_channel is broken now. By detaching
LUNs from their SCSI devices at the start of LUN
hotremoval, we can end up with the LUN's descriptor
completely leaked.
Upon hotremove, each LUN fires a hotremove_cb to the
upper layer (iscsi/vhost) and starts a poller to
periodically check if lun->io_channel has been closed
already. While iSCSI allocates and frees io_channels
for particular LUNs separately using
spdk_scsi_lun_allocate_io_channel(), vhost allocates
and frees io_channels for the entire SCSI device using
spdk_scsi_dev_free_io_channels(), as it knows there's
only one LUN per SCSI device.
Since LUNs are removed from SCSI devices at the start
of LUN hotremoval, the spdk_scsi_dev_free_io_channels()
is not able to close the io_channel of the child LUN.
That LUN is no longer referenced there, and the function
returns immediately without any error.
By reverting this patch, we practically defer removing
LUN from SCSI device until the bdev descriptor is closed,
that is, until that spdk_scsi_dev_free_io_channels()
and the subsequent spdk_scsi_dev_destruct() in vhost
is called.
Historically, spdk_scsi_lun_free_io_channel() was
introduced after vhost-scsi already supported hotremove,
so before we can introduce a change like this one we
have to refactor vhost to use the new APIs first.
This reverts commit 497997bc60.
Change-Id: I58a9988a22694f5b5b927173098ac8218270462e
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455371
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Pawel Kaminski <pawelx.kaminski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Recently, we moved the check for running an installed version of DPDK
above the point where we export the SPDK_* test flags. This resulted in
SPDK_RUN_INSTALLED_DPDK being undefined when trying to check fi we
should run against an installed DPDK. This just caused autotest_common
to print an error message every time it was loaded. I don't think that
it actually caused any errant behavior but this gets rid of the error
message. It's also probably a good idea to keep these definitions as
close to the top of the file as possible.
Change-Id: I5aadbe5c925ecf1ac92926b75c8c043aab73b36b
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455456
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: Jim Harris <james.r.harris@intel.com>
Padding requests are issued in multiple batches at the same time
when shutdown occurs. This allows for faster removal.
Change-Id: Iea40d2418bedbd7cf3c6865e5eb8f85871db13cd
Signed-off-by: Kozlowski Mateusz <mateusz.kozlowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454578
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
For the latest TLC NAND, one write buffer unit (rwb batch)
needs to be spread over three PUs instead of being allocated
to a single PU for better sequential read performance
since the optimal write size(ws_opt) of 3D TLC NAND is
3 times bigger than the optimal read size(rs_opt).
I added num_interleave_units in 'struct spdk_ftl_conf'
to configure the number of interleaving units per ws_opt.
If num_interleave_units is set as 1, the whole of the ws_opt
blocks are placed sequentially around each PU.
If num_interleave_units is set as N, the 1/N of the ws_opt
blocks are staggered. So consecutively numbered blocks
are separated by ws_opt / num_interleave_units.
The sequential read performance is improved from 1.9GiB/s
up to 2.97GiB/S with this patch on our system. No performance
degradation is observed on sequential writes or
4KB random reads/writes.
Please refer to the Trello card for more details.
https://trello.com/c/Osol93ZU
Change-Id: I371e72067b278ef43c3ac87a3d9ce9010d3fcb15
Signed-off-by: Claire J. In <claire.in@circuitblvd.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450976
Reviewed-by: Young Tack Jin <youngtack.jin@circuitblvd.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is the same intention as the patch for iSCSI in this series.
This change will be helpful to extract common part into a specific
helper library if necessary in future.
Change-Id: I1ce36b424ff2afb85f998149f4ef0d7153290802
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455621
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>
Recently DIF library refined SGL create operation by unifying
size and used count into unused count. This patch applies the
good practice in DIF library to create SGL in NVMe/TCP.
The next patch refines names of related function and variables
to be consistent in NVMe/TCP.
Change-Id: I1e73310c0e3650ede53672d76071a6c37dba82c1
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455473
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>
Recently DIF library refined _iov_ctx to support multiple iovecs
for DIF insert and strip operations. iSCSI library is the first
user of DIF insert and strip, and so rename _iov_ctx to _iscsi_sgl
to match DIF library. Names of related helper functions and
a member variable are changed together.
This change will be helpful to extract common part into a specific
helper library if necessary in future.
Change-Id: I3be3b6c3c476b0309ffafdecf3181fb2bb19abc6
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455620
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 patch adopts the good practice in DIF library, and unifies
array size and used count into unused count.
Change-Id: Ifde15d1fa66dbd2578da771b5500e45a84664f63
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455475
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>
In some cases user may want to flag blob for removal
then do some operations (before removing it) and while
it happens there might be power failure. In such cases
we should remove this blob on next blobstore load.
Example of such usage is delete snapshot functionality
that will be introduced in upcoming patch.
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I85f396b73762d2665ba8aec62528bb224acace74
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/453835
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 patch moves _spdk_blob_set_thin_provision function
higher in the file as it will be later used during
blobstore load.
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: Ife37ef8c69b88903646b2002b3561101c1eb5135
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455488
Reviewed-by: Piotr Pelpliński <piotr.pelplinski@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>
A PERSISTENT RESERVE OUT command with PREEMPT service action action is used to:
a) preempt (i.e., replace) the persistent reservation and remove registrations; or
b) remove registrations.
Change-Id: I906b09057e5ddef0ea2fe180bb5fff6e9e0a04f6
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/436092
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Any application client can release the persistent reservation and remove all
registrations from a device server by issuing a PERSISTENT RESERVE OUT command
with CLEAR service action through a registered I_T nexus.
Change-Id: Ie21885f7f9632b9da77e877fbce7c51b2e0b47a7
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/436091
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>
An application client releases the persistent reservation by issuing a
PERSISTENT RESERVE OUT command with RELEASE service action through an
I_T nexus that is a persistent reservation holder. The RELEASE service
action will release the persistent reservation and do noting to
existing registrants. Unit Attention isn't supported by SPDK so here
we will not establish any unit attention condition.
Change-Id: If0df3b3fc4e89ffa78d827e2b31cf5c61aa149db
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/436090
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
An application client creates a persistent reservation by issuing a
PERSISTENT RESERVE OUT command with RESERVE service action through
a registered I_T nexus with RESERVATION KEY and TYPE, and persistent
reservation has a scope of LUN, only one persistent reservation is
allowed at a time per LUN.
Change-Id: I052d0ddd439dd9994c0793218dd5c5d6ed176d2c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/436089
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>
To establish a persistent reservation the application client shall first
register an I_T nexus with the device server. An application client registers
with a logical unit by issuing a PERSISTENT RESERVE OUT command with REGISTER
service action or REGISTER AND IGNORE EXISTING KEY service action.
Specify Initiator Ports (SPEC_I_PT) bit and All Target Ports (ALL_TG_PT) bit
are not supported for now, the registrants belong to the I_T nexus who sends
the command.
Also Activate Persist Through Power Loss (APTPL) bit will be supported in
following patches.
Change-Id: If057a764a4bfa73017f98048c94b697dbfa4b4a1
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/436088
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Good suggestion from Darek - let's just always
parse common script args from autotest_common.sh.
These arguments follow common arg naming conventions
(i.e. --iso) so there's no harm just doing this for
any test that sources autotest_common.sh. This has
the nice effect of not requiring scripts to
explicitly call this function.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Id89b68c22557a5a771be407873d0e57843f0d05a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455552
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
We throttle the number of data_in operations per
connection. Currently after a read is completed,
we try to send more data_in operations since one
has just been completed.
But we are trying to send more too early. The data_in_cnt
doesn't actually get decremented until after the PDU is
written on the socket. So this results in a case
where data_in_cnt == 64, and all 64 read operations
complete before any of those 64 are actually transmitted
onto the TCP socket. There are no more read operations
waiting, so we won't try to handle the data_in list
again, and if none of these 64 resulted in a SCSI
command completing, then the initiator may not send us
any more read I/O which would have also kicked the data_in
list.
So the solution is to kick the data_in list after the
PDU has been written - not after a read I/O is completed
back from the SCSI layer.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ia01cf96e8eb6e08ddcaaeff449386e78de7c5bc5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455454
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>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
The datain handling code path will set the LUN to
NULL if it finds a task's LUN has been hotremoved.
This could happen before the iscsi hotplug routine
actually gets a chance to run. If this happens,
one of these tasks doesn't actually get freed, and
then will be freed after the lun is closed -
causing a segfault in the bdev layer since it may
have a bdev_io associated with it.
Found by running the iscsi_tgt/fio test after
applying the next patch in this series.
There's more work needed in this hot remove clean up
path - currently we are just freeing a lot of PDUs
rather than completing them with error status when
a LUN is hot removed. But let's tackle that
separately.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I8d27f0c7a79ae91cb6504e5ff6ffc8e346c9e54c
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455460
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>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
when qemu connect to vhost, but don't send msg to vhost. We use
kill -15 to destroy vhost process. it will lead to deadlock.
(A)
* rte_vhost_driver_unregister()
* pthread_mutex_lock hold vhost_user.mutex (1)
* wait TAILQ_FIRST(&vsocket->conn_list) is NULL
(B)
* fdset_event_dispatch()
* vhost_user_read_cb() start
* vhost_user_msg_handler() start
* dev->notify_ops is NULL because qemu just connect, no message recv.
* vhost_driver_callback_get()
* pthread_mutex_lock hold vhost_user.mutex (2)
(A) & (B) deadlock
To avoid this scenes, when qemu connect in vhost_new_device()
initialize dev->notify_ops
Change-Id: Iaf699da41dfa3088cfc0f09688b50fada6b2c8d6
Signed-off-by: Tianyu yang <yangtianyu2@huawei.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454832
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
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>
Use OCF per-volume UUID field to serialize bdev names.
This is going to be used during cache load to find out
which bdevs we want to attach.
Note that there is in fact "user_metadata" buffor that is
also stored in cache metadata, but we cannot use it
because its size is limited to 64 bytes.
UUID in OCF terms is not standarized
and is meant to be used to store custom immutable data.
This change preparation for persistent metadata support.
Functionality is not changed. (everything works the same way)
Change-Id: Ia9204fae29106f5b816d93a6771425a223d6c028
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455414
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Change shutdown behavior such that now first bdev destruct() call
stops OCF cache instance.
Previusly, cache was stopped when single vbdev was referencing it.
This patch is related to persistent metadata support.
Without this change, every time a SPDK application is stopped,
only the last core is remembered in metadata,
because vbdevs detach by 1 by 1, each time updating the metadata.
Change-Id: I57db3f77db525177c024ee85e660a85aff2f8c31
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455413
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Fix instance function for OCF ctx interface.
This function is used for persistent metadata only.
Change-Id: I9583ea8eb21f07a3e9072a9552bed1c077cb7114
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455412
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Make init_vbdev() accept NULL as cache_mode if loadq flag
is set.
This is needed for load path because we don't know cache mode
before the actual load (happens later).
This change is related to persistant metadata support.
This patch does not change current behavior because loadq flag
is always false.
Change-Id: I42727eb841d87903c81bd5e6d51b6d4869ed9be3
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455411
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Implement load path for OCF cache.
During load OCF will read metadata from cache disk and restore its state.
management_channel initialization is moved before cache load/attach part
because ocf_mngt_cache_load() needs it to read metadata.
This patch is a preparation for persistant metadata support.
Load path is never used yet because loadq flag is always false.
Change-Id: Iff6c1c52eae7b9f52812a8bd3d5ae6d6facedd60
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455410
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add new loadq flag that indicates if vbdev
should load cache instance from disk or start a new one.
This change is a preparation for persistent metadata support.
Functionality is not changed in this patch as the loadq flag is always false.
Change-Id: I1baac7f988e6eeadb4f365ba7bfab8019d55a753
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455409
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Modify FIO test to use all cache modes that we support,
including WriteBack
New test config uses Nvme bdevs instead of mallocs because
memory is an issue when testing OCF
Change-Id: I3abec9605b61791f8ebaaaf08b88a011a50d3f26
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451022
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Do management flush during OCF shutdown to write all dirty requests to cores.
Dirty requests are relevant to WriteBack mode only.
Change-Id: I778a73ed8ab5659921f192f638027d513c239814
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450683
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Implement cleaner that is used in WriteBack mode.
Cleaner is a background agent that does synchronization
of data between cache and its cores.
Cleaner usually runs every ~20 seconds to perform cleaning.
The synchronization is a simmilar operation to OCF management flushes.
We need cleaner for WriteBack because only WriteBack mode
produces dirty data that cleaner needs to deal with.
Cleaner requires adopting trylock() because in current version
cleaner uses management lock when performs cleaning,
which may lead to deadlocks if cleaner runs on
the same thread as management operations.
WriteBack mode is fully functional after this change,
but persistent metadata support is required to use it for production.
Cleaner will run on management thread for now.
We plan to implement functionality of
chosing a CPU core where cleaner should run.
Cleaning policy is not configurable yet.
The default is ALRU with 20 sec interval.
Change-Id: I35aa7e00c44e0d7a77e64e60df1f66f20be03f55
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448537
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Create new management channel and handle sharing between vbdevs.
This channel is going to be used for management operations
that produce IOs (such as flush) and also for cleaner.
Change-Id: Ieeed8454a7ab7459c86ac06ec6c0ece038bc928e
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455272
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Tests for checking read-only option no longer fail.
Removing rootdir variable and sourcing commands as
*_vm.sh scripts are designed to be run as standalone
inside spawned VMs.
Change-Id: I491ebf8e2b305539e3fabdc6219ea53693ad2883
Signed-off-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455203
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Pawel Kaminski <pawelx.kaminski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Declaractions for those functions already have dst
and src as param names, but definitions use set1 and
set2, which doesn't really tell which one is which.
My IDE actually gives me tooltips with param names
from definitions and it would be much easier to see
"dst" and "src" in there.
Change-Id: I45013cf27fc95f69cb2c776fb1b6701c5e60e373
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455163
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>
iscsi test scripts can now take two arguments -
"iso" and then the sock type (posix or vpp). They
need to be in that specific order too. nvmf test
scripts also support "iso" and we want to add
the transport type (rdma or tcp) as well. Even further
out, we may want to use a sock type for nvmf, i.e.
tcp transport with vpp.
We also have the iscsi_tgt fio_remove_nvme.sh test
that does both iscsi and nvmf.
So to make this all work a bit nicer, add a new
function called parse_common_script_args that
will take the command line arguments to a script
and set the appropriate variables, including defaults
when a specific parameter isn't specified. We will
use getopt-like behavior for this also, instead of
enforcing a specific parameter order. Then a script
could be called like this:
test/nvmf/target/shutdown.sh --iso --transport=tcp --sock=vpp
Individual test scripts then just need to do this
after sourcing autotest_common.sh:
parse_common_script_args $@
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ifb8d7666384991482a2d425e26ffa7525b9ac15a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455283
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>
Implement management queue for asynchronous management that is provided
in new OCF API. This is a neccessery step to be able to implement
metadata support.
OCF submodule was updated to get management queue functionality
It has to be done in this patch because OCF module will not work
otherwise.
Change-Id: Id19c2e5bd6a5d26fee41752b62720e408dc082e8
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448619
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
OCF queue list needs to be managed synchronously.
This patch uses our own mutex to achieve that because we cannot rely on
ocf_mngt_cache_lock() as it may produce deadlocks when using
cleaner.
Alternative way would be to use trylock, but
we need to register a poller for it and do locking concurently
which doesn't seem to be possible in callbacks of io channel.
We agreed with OCF team that we are going to change this part
when OCF will deliver safe ocf_mngt_cache_lock() function.
This patch fixes a very rare failure on our CI that looked like this:
```
04:33:02 vbdev_ocf.c: 134:stop_vbdev: *NOTICE*: Not stopping cache instance 'Malloc0' because it is referenced by other OCF bdev
04:33:03 MalCache1: Core core2 successfully removed
04:33:03 MalCache1: Stopping cache
04:33:03 MalCache1: Done saving cache state!
04:33:03 src/ocf/mngt/ocf_mngt_cache.c:1576:2: runtime error: pointer index expression with base 0x000000000000 overflowed to 0xffffffffffffffa8
04:33:03 #0 0x7f3c52d54c26 in _ocf_mngt_cache_stop src/ocf/mngt/ocf_mngt_cache.c:1576
04:33:03 #1 0x7f3c52d5579f in ocf_mngt_cache_stop src/ocf/mngt/ocf_mngt_cache.c:1657
04:33:03 #2 0x7f3c52cbe9f4 in stop_vbdev /var/jenkins/workspace/NVMe_tests/nvme_phy_autotest/spdk/lib/bdev/ocf/vbdev_ocf.c:147
04:33:03 #3 0x7f3c52cbf4a0 in vbdev_ocf_destruct /var/jenkins/workspace/NVMe_tests/nvme_phy_autotest/spdk/lib/bdev/ocf/vbdev_ocf.c:216
```
Change-Id: Id6fafb444958f3becdc480e44762074c6c081e1f
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450682
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Introduce cache context structure that is going to be used
for sharing per cache info. For example: management queue,
cleaner_channel, cleaner_thread.
Lifetime of this structure ends with last vbdev that gets unregistered
and NOT when cache stops because cache does not have to be freed
there.
Change-Id: I66252084d7efda92edd10fb737c8e9c8169b4f6d
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451403
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
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>
Adapt to new async API which was changed recently
This is a neccessery step to be able to implement persistent metadata support
OCF submodule was updated to get new API
Change-Id: I6bf9941ab0557981235c5be27686594a1b8ac3a0
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448397
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>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Move last step of register chain to the top because it is arguably more readable.
This should not be done in previus patch because it spoils the gerrithub diff.
Change-Id: If51642a32c5cf2a757ca20485d2347410653ca2a
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454574
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Use new trylock API to prevent lockups when using OCF cleaner.
Starting and stoping OCF bdev becomes asynchronous with this patch.
Using trylock means that we have to poll function that does locking each
time we want to initiate some operation on cache instance.
This is the reason why management functions become asynchronous.
Each management operation now has a _poll() operation associated with it.
_poll() uses trylock and continues when cache is locked.
Change-Id: I83b9fbe87c27433e178583411b87a68b8efaf58e
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447888
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>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This patch is a preparation for adopting ocf_cache_trylock() function.
Register OCF bdev path uses vbdev_ocf_mngt_ interface now
Register stays synchronous blocking operation in this patch,
but with adoption of ocf_cache_trylock() function,
register will be asynchronous, in which case we
want to use mngt_ interface.
This series of OCF patches will enable WriteBack support for OCF bdev.
There is a lot of preparation before we will be able
to actually implement WriteBack.
Dependencies look like this:
- WriteBack
- Cleaner
- trylock
- Persistent metadata
- asynchronous management API
Cleaner is a background agent that does synchronization
of data between cache and its cores.
Cleaner usually runs every ~30 seconds to perform cleaning.
The synchronization is a simmilar operation to OCF management flushes.
We need cleaner for WriteBack because only WriteBack mode
produces dirty data that cleaner needs to deal with.
Cleaner requires adopting trylock() because in current version
cleaner uses management lock when performs cleaning,
which may lead to deadlocks if cleaner runs on
the same thread as management operations.
Peristent metadata is functionality of OCF
that allows to restore cache state after shutdown
We need persistent metadata in context of shutdown
with some data being dirty which may only happen
when using WriteBack mode
Support for persistent metadata requieres
asynchronous OCF API because in current version
we will have a deadlock during metadata initialization
because it is required to be done on different thread
than cache initialization.
Change-Id: I45d84a5fa0c96581d522050dad186b06d489226e
Signed-off-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455225
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
