Used (void) on cmd and removed increment to fix
clang 15 werror.
vmd.c:368:11: error: variable 'cmd' set but not used [-Werror,-Wunused-but-set-variable]
uint16_t cmd = dev->header->zero.command;
^
1 error generated.
Signed-off-by: Kamil Godzwon <kamilx.godzwon@intel.com>
Change-Id: I4e383ac41b46d13df0210bf90f11f6130290f243
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16127
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Patch below started checking development version of DPDK
using rte_version_release():
(32e6ffb) env_dpdk: add support for DPDK main branch for 23.03
rte_version_release() is present starting with DPDK 21.11,
so it broke earlier versions like DPDK 20.11 packaged on Fedora 35.
SPDK supports only last two DPDK LTS versions, which does not include DPDK 20.11.
Yet there is no need to break older versions unnecessarily.
Another aspect is that rte_version_release() is marked as experimental,
so it could change in the future. Only using stable rte_version(),
helps with forwards compatibility too.
Change-Id: Id17d643a12dcfc03c2d4688d1bc5030dc339f428
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reported-by: Michal Berger <michal.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16017
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Michal Berger <michal.berger@intel.com>
Added spdk_nvme_qpair_get_num_outstanding_reqs to get the number
of outstanding reqs for a specific qpair.
Change-Id: I55d75a7363ac63bd26db76594e70e8b17b3e5830
Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15916
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Added num_outstanding_reqs in struct spdk_nvme_qpair to record outstanding
req number in each qpair. This can be used by multipath to select I/O
path.
Increment num_outstaning_reqs when req is removed from free_req queue and
decrement it when req is put back in free_req queue.
Change-Id: I31148fc7d0a9a85bec4c56d1f6e3047b021c2f48
Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15875
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
bdev_crypto uses memset() to zero secrets passed
by the user (cleanup/error path) which is not safe -
compiler may detect that the buffer being zeroed
is not accessed any more and may "optimize" (drop)
zerofying.
C11 standard introduces memset_s which guarantess to
change the buffer content, but this function is optional,
gcc may not support it. As alternative, add not optimal
from performance point of view default implementation.
Add unit test to math_ut.c to avoid creating new .c file
for 1 simple test
Signed-off-by: Alexey Marchuk <alexeymar@nvidia.com>
Change-Id: I11c7d15610df02e4a3761a88c85f6f8c54fb4b0a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16038
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The FIELD_OK macro in blob_open_opts_copy() should consider offsets in
struct spdk_blob_open_opts, not struct spdk_blob_opts.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I62e22acbe7dfb994453a379c92f78b7e9bc7fc13
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14962
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>
Blob IDs are sequentially assigned starting at 0x100000000. When
debugging with a small number of blob IDs, it is much more intuitive to
see blob ID 0x100000000 rather than blob ID 4294967296.
In commit 76a577b082 a similar change was
made to blobcli.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: Ic5321a83b57cf8c9f8df48cd424a926b6fec4ba8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14963
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>
It's the same as spdk_iovcpy(), but the dst/src buffers can overlap.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I6daa0a846d7d1deac2c01d1a1be09171fa8bf796
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15747
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: Aleksey Marchuk <alexeymar@nvidia.com>
The data buffers backed by these accel buffers aren't allocated
immediately, but only when they're necessary to execute a given
operation. It allows users to append operations to a sequence, without
actually reserving large space for the data. That way, if some of these
buffers aren't needed to execute a sequence, they won't be allocated.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ieeea8a011b40c7f2f33e9a6f03fe34264e9316f3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15746
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: Aleksey Marchuk <alexeymar@nvidia.com>
It will be used for allocating buffers from accel domain and
allocating bounce buffers to push/pull the data from memory domains for
modules that don't support memory domains.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Idbe4d2129d0aff87d9e517214e9f81e8470c5088
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15745
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 domain is meant to represent data being transformed by accel
engine. Users will be able to allocate buffers from that memory domain
and use them when appending operations to an accel sequence.
Since these buffers are only meant to be used as placeholders for actual
buffers, none of the push/pull/translate callbacks are implemented. To
access the data after it was transformed by accel, users should make
sure that the final command's destination buffer isn't allocated from
accel memory domain.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ia031c7b205e98792d0a93f01513101b86afa9faa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15744
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>
Reversing a sequence means that the order of its operations is reversed,
i.e. the first operation becomes last and vice versa. It's especially
useful in read paths, as it makes it possible to build the sequence
during submission, then, once the data is read from storage, reverse the
sequence and execute it.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I93d617c1e6d251f8c59b94c50dc4300e51908096
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15636
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: Aleksey Marchuk <alexeymar@nvidia.com>
Operation sequence should always be treated as a whole, meaning that
users cannot rely on the contents of any intermediate buffers and should
only care about the buffer that's the destination of the whole
operation. This allows us to remove some of those copy operations by
changing source / destination buffer of a preceding / following
operation.
If a sequence is using buffers from non-local memory domain, users can
append a copy operation to a sequence to specify a local destination
buffer. If the module executing the operations is aware of memory
domains, this can avoid doing an extra spdk_memory_domain_pull_data().
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I93b94d46ee32700819e9e6f1c55350692db8a67a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15530
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: Aleksey Marchuk <alexeymar@nvidia.com>
This patch introduces the concept of chaining multiple accel operations
and executing them all at once in a single step. This means that it
will be possible to schedule accel operations at different layers of the
stack (e.g. copy in NVMe-oF transport, crypto in bdev_crypto), but
execute them all in a single place. Thanks to this, we can take
advantage of hardware accelerators that supports executing multiple
operations as a single operation (e.g. copy + crypto).
This operation group is called spdk_accel_sequence and operations can be
appended to that object via one of the spdk_accel_append_* functions.
New operations are always added at the end of a sequence. Users can
specify a callback to be notified when a particular operation in a
sequence is completed, but they don't receive the status of whether it
was successful or not. This is by design, as they shouldn't care about
the status of an individual operation and should rely on other means to
receive the status of the whole sequence. It's also important to note
that any intermediate steps within a sequence may not produce observable
results. For instance, appending a copy from A to B and then a copy
from B to C, it's indeterminate whether A's data will be in B after a
sequence is executed. It is only guaranteed that A's data will be in C.
A sequence can also be reversed using spdk_accel_sequence_reverse(),
meaning that the first operation becomes last and vice versa. It's
especially useful in read paths, as it makes it possible to build the
sequence during submission, then, once the data is read from storage,
reverse the sequence and execute it.
Finally, there are two ways to terminate a sequence: aborting or
executing. It can be aborted via spdk_accel_sequence_abort() which will
execute individual operations' callbacks and free any allocated
resources. To execute it, one must use spdk_accel_sequence_finish().
For now, each operation is executed one by one and is submitted to the
appropriate accel module. Executing multiple operations as a single one
will be added in the future.
Also, currently, only fill and copy operations can be appended to a
sequence. Support for more operations will be added in subsequent
patches.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Id35d093e14feb59b996f780ef77e000e10bfcd20
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15529
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: Aleksey Marchuk <alexeymar@nvidia.com>
Existing `vhost_user_session_send_event` is only used to
stop vhost user device's session now, so we rename it to
`vhost_user_wait_for_session_stop` and also rename the
whole function calls when stopping the device with
more apposite names.
Change-Id: Ib8ea48273e85f7856ca2dfca57b5fd933ac4cf7a
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15296
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
For vhost-user device, the variable `active_session_num` is used
to count number of sessions of a vhost-user device, we don't use
it anywhere, and the assertion of this variable is already
guaranteed by `vsessions_num`, so just remove it.
Change-Id: I335a75d17583b3744a41152b35cd5a1a8762a687
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15189
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
If we kill the vhost process while VM is connected, the `g_fini_cb`
will not be called due to active session is in the vhost-user device,
but we're sure that this VM is stopped for this case, because
`vhost_driver_unregister` is called in the shutdown thread, so here
we reuse `g_vhost_user_started` flag for this case and free the sessions,
the following call to `vhost_driver_unregister` can also handle this
case, because the Unix Domain socket is already unregistered.
Fixes commit 327d1c98 ("vhost: defer vhost_dev_unregister until scsi tgts removed")
Change-Id: I4f368ac8c304dd9525d15abdce8fd5b2ed79b96e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15623
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
`rte_vhost_driver_unregister` API for removing socket is not
asynchronous, it may call SPDK ops for adding a new connection
or removing a connection, so we can't hold the user device lock
when calling this function, and reject to add a new connection
while calling `rte_vhost_driver_unregister`.
Fix issue #2748.
Change-Id: I5594224f26374b2336d64175ecd5e5ec3d545a58
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15483
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
`spdk_vhost_dev` is created|deleted via RPC or APIs, and
we use a global `spdk_vhost_lock` to protect it, but for
some other places such as: vhost-user message processing,
we also use the global lock for now, actually we don't
need to use this lock, because these vhost-user messages
processing will not delete nor add vhost devices.
While here, we add a `spdk_vhost_user_dev` access lock to
protect vhost-user message processing as an optimization.
Change-Id: Ia9c45b056cebb7b65f458d56ed775a15e386f905
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15184
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Feng Li <lifeng1519@gmail.com>
nvmf_ctrlr_cmd_connect() can only handle a request in one buffer
(req->data); sanity check it's not split across IOVs.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: I595d8542ce71e56cf2b074f4cf41bce440f6dc26
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16123
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This code has a similar potential problem as the identify
and log page commands did: stop using req->data in favour of IOVs.
We also need to fix the unit tests to initialize the iovs.
We don't change the existing "set" behaviour of requiring a single IOV
here.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: I257567a7abd5fc3ed9ee21b432c7da7d70fbbde0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16122
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In the previous fix:
adc2942ad nvmf: nvmf_ctrlr_get_log_page use iovs to store the log page
a data corruption bug in the log page code was fixed. Previously, it
used req->data, which may be too short a buffer in the case that the
buffer is split across more than one IOV. req->data is never safe to use
in this situation. The code was changed to use the provided iovs instead
of req->data.
However, the identify command handling was still vulnerable to this
problem, and has been seen in real life at least with a CentOS guest VM.
The fix is basically the same: use the IOV utility functions to write
out the response instead of directly trying to use req->data.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: I00445895af20e43be73189629576eee0667f86dd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16121
Reviewed-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Move the IOV handling code in ctrlr.c to the top of the file, for
subsequent use.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: Ibddde1cb964d8aaecf4673ffa6d4147d0a48020c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16120
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Add a define for the Identify command buffer instead of using a raw
value.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: I9073ff84e2fa2ef9268051b898fe1027d8e97baa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16119
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
In preparation for supporting additional claim types, create a claim
type that represents the current claim type. Everything that sticks to
the public APIs should continue to work as before.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I0d02e4b3f4bbf4eb5a7391028aa31e999f9da915
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15286
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In preparation for an updated claims API, refactor
bdev->internal.claim_module into a union that will eventually hold
different information based on the the type of claim.
Change-Id: I7ade6f03128bdb0f8375a95ae953cb63d6aa686d
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15285
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Mellanox Build Bot
This calls bdev_ok_to_examine() once per bdev_examine(). Prior to this
commit, bdev_ok_to_examine() may be called up to twice per bdev module.
The results returned by bdev_ok_to_examine() could be affected by:
1. g_bdev_opts.bdev_auto_examime changing
2. spdk_bdev_examine() being called on a particular bdev
3. An alias being added for an existing bdev
It's not clear that anything good comes from racing in conditions 1 and
3. In condition 2, spdk_bdev_examine() calls bdev_examine(), so any
required examine_config() and examine_disk() calls are still made, just
now with less of a race with the previous invocation of
spdk_examine_confg().
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I496fc44fd74693837d6b449d7fa60f58f9dbf36f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15284
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
This closes races between concurrent spdk_bdev_module_claim_bdev()
and/or spdk_bdev_module_release_bdev() calls affecting the same bdev by
holding bdev->internal.spinlock while claiming and releasing a bdev. It
also closes a potential TOCTOU bug in that optimizing compilers probably
already eliminate in bdev_finish_unregister_bdevs_iter() and documents
that bdev->internal.claim_module is protected by
bdev->internal.spinlock.
This can be removed when the bdev_register_examine_thread deprecation
is removed.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: Ib48552df065d5172139a61bbc00b391f36552c0c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15282
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Mellanox Build Bot
Since bdev_examine() can happen on any thread and it happens without any
other lock being held on the spdk_bdev_module, it is possible for
multiple threads to try to simultaneously increment
module->internal.action_in_progress. Decrements may also race.
This commit adds bdev_module->internal.spinlock and holds it while
modifying module->internal.action_in_progress.
This can be removed when the bdev_register_examine_thread deprecation
is removed.
Change-Id: I9c401eeb3c7c97c484e16fa9cfd82668b32e508b
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15281
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Mellanox Build Bot
This introduces a deprecation for calling spdk_bdev_register() and
spdk_bdev_examine() on a thread other than the app thread. The
deprecation period starts in SPDK 23.01 and removal is expected in SPDK
23.05.
The intent of this deprecation is to ensure that bdev modules'
examine_config() and examine_disk() callbacks are only ever called on
the app thread. This largely a formalization of what has long happened
due to the RPC poller running on the first thread started by
spdk_app_start().
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: Ic9d7b87b6522be20357d2eab2d0c77cd5753452f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15690
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Mateusz Kozlowski <mateusz.kozlowski@intel.com>
Some memory alloc in nvme_allocate_request_user_copy, and submit
through nvme_qpair_submit_request, if nvme ctrlr is failed or
qpair state not meet the requirements, submit will return -ENXIO,
and call nvme_free_request(), but it will not free
req->payload.contig_or_cb_arg, those memory only gets freed when the
request is actually completed, through nvme_user_copy_cmd_complete().
Let's fix this by add check when submit failed.
Fixes issue #2832
Change-Id: I54f0fc60dbb53ced9f52da7d89017be13db2eee1
Signed-off-by: Fengnan Chang <changfengnan@bytedance.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15985
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
rc to -ENXIO and goto error, make all error handle in one place,
so it's easy to add more check in later patch.
Change-Id: I13edeef75bbf6c52e18d6b94b78c2e560012bfee
Signed-off-by: Fengnan Chang <changfengnan@bytedance.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16004
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The spdk_nvme_ctrlr_opts now supports a transport_tos option
that allows setting of the 'type of service' value in the IPv4 header.
This is needed to support lossless RoCE setups.
Note: Only RDMA is supported at this point.
Change-Id: I21825fc197c60f539a7d2d651a970ea380d8b56d
Signed-off-by: Michael Haeuptle <michael.haeuptle@hpe.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15908
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add spdk_nvme_cpl_get_status_type_string() to return ASCII
string for the type of an error.
Append a dummy entry to return "RESERVED" for unknown types.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ibc07132ee067f146ac149884c6344f313bfcbfff
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15835
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
spdk_nvme_cpl_get_status_string() will be used to count and display
NVMe specific errors via JSON-RPC. This patch is a preparation.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ia96890172d752d2906549e3033c0b26eef9c20bd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15834
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
However, when querying or resetting module specific statistics,
the generic bdev layer have to access it.
For this purpose, add functions pointers.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ie86d0a4a406cec7e0f1e9a62de5982cd3d877eae
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14839
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Define struct spdk_bdev_io_error_stat privately in lib/bdev/bdev.c.
Add a pointer to struct spdk_bdev_io_error_stat to struct
spdk_bdev_io_stat.
Allocate spdk_bdev_io_error_stat for bdev and RPC, but do not allocate
spdk_bdev_io_error_stat for I/O channel.
Dump the contents of spdk_bdev_io_error_stat only if its total is
non-zero.
As a result of these, only spdk_bdev_get_device_stat() can query
spdk_bdev_io_error_stat for the bdev_get_iostat RPC. This will be
acceptable.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Idae868afe65347a96529eedc3dcc692101de4a29
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14826
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
The following patches will make some of io_stat helper functions
public APIs. Then, for consistency, bdev_ + verb + _io_stat will
be better naming rules.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: If36d4ed29253e87954c23c270e8414731d083f03
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15896
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Commit 41f59559e added code to skip adding EXITING connections
to the new poll group in the full_feature_migrate message
callback. The problem is that since the connection is in
EXITING state and is not in a poll group, it will never move
to EXITED state, nor get removed from g_active_conns, and
hence will block the iscsi subsystem from being able to
shutdown.
So instead, assert that the connection is not in EXITED
state. If it is in EXITING state, we will add it to the
poll group, and then when the poll group is next polled,
it will destroy the connection, moving it to EXITED
state and removing it from the g_active_conns STAILQ.
This fix is related to issue #2416.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie8e64c811a5602ba4b28871bc535f5fa49dffc18
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/16019
Reviewed-by: Michal Berger <michal.berger@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Many parts of the blobstore.c seem to have gone with the assumption that
blob creation, deletion, etc. all happen on the md thread. This
assumption would allow modification of the bs->used_md_pages and
bs->used_clusters bit arrays without holding a lock. Placing
"assert(spdk_get_thread() == bs->md_thread)" in bs_claim_md_page() and
bs_claim_cluster() show that each of these functions are called on other
threads due writes to thin provisioned volumes.
This problem was first seen in the wild with this failed assertion:
bs_claim_md_page: Assertion
`spdk_bit_array_get(bs->used_md_pages, page) == false' failed.
This commit adds "assert(spdk_spin_held(&bs->used_lock))" in those
places where bs->used_md_pages and bs->used_lock are modified, then
holds bs->used_lock in the places needed to satisfy these assertions.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I0523dd343ec490d994352932b2a73379a80e36f4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15953
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
A future commit will add to the complexity when returning with a
non-zero value. Rather than further complicating the several error
return locations, all affected error returns are handled after the error
label.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I56e8e338b0560f849399c085d0bb07efb7df26fb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15983
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
A future commit may need to release a lock before returning. This
refactors blob_resize() to always return at end of the function using an
out label and goto.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I671fbdbe0e3b766c264c45589dad3a864ba1f192
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15982
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
Convert bs->used_lock to a spinlock. This is being done to help with
the debugging and fixing of a race that has led to a failed assertion in
bs_claim_md_page.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I11b80096de022f79a217c65d787ee57ca54240f9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15952
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
The bs->used_clusters_mutex protects used_md_pages, used_clusters, and
num_free_clusters. A more generic name is appropraite. The next patch in
this series will convert it from a mutex to a spinlock and having
"mutex" or "spin" in the name is of little help to maintainers, so a
more generic name is used.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I5ce7b85b84fdec2a0c5d2ac959e0109e1d80c7f5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15981
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
Basic IO completion counting can be done at the common
layer, to enable some level of stat tracking even for
transports that don't have transport-specific tracking
yet.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If04f854b97440089b8ad149b64cb59173c73975c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15912
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Mellanox Build Bot
For validation of upcoming DPDK releases, pci_dpdk needs
to initialize and work.
This patch adds support for testing DPDK main branch,
with appropriate notice given when that DPDK version is used.
Change-Id: I5257beac3a3926bd432d9c00e50858facd21e6f5
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15891
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Since SPDK holds copies of local DPDK headers for DPDK PCI API,
the same headers will now be used as includes.
It was already the case for DPDK 22.11, but not for DPDK 22.07.
Change-Id: I5859a630d1fb20b4ebf8628adb962f5e46c23788
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15969
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Shortly after DPDK 22.11 release it was amended with single
patch, which bumped the minor version.
No changes have occurred to the DPDK PCI API.
Change-Id: I94dadb23b3ad79cfbb21e848d718d909493137d1
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15890
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Michal Berger <michal.berger@intel.com>
As spdk_lvs_init() validates arguments, it uses o->cluster_sz in a
comparison but misleadingly prints opts.cluster_sz in the error message.
This changes the error message to print cluster_sz from the proper
structure.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I810bf9ad4a24ed7cc844c2835e0edda988cb2cbe
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15970
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
The internal mempools were replaced with the newly added iobuf
interface.
To make sure we respect spdk_bdev_opts's (small|large)_buf_pool_size, we
call spdk_iobuf_set_opts() from spdk_bdev_set_opts(). These two options
are now deprecated and users should switch to spdk_iobuf_set_opts().
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ib1424dc5446796230d103104e272100fac649b42
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15328
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
This is done in a couple of places, so it makes sense to extract it to a
separate function.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Id34b2545d9912c2b7b65b1277711e9683db92658
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15327
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Users can now specify a number of small/large buffers to be cached on
each iobuf channel. Previously, we relied on the cache of the
underlying spdk_mempool, which has per-core caches. However, since iobuf
channels are tied to a module and an SPDK thread, each module and each
thread is now guaranteed to have a number of buffers available, so it
won't be starved by other modules/threads.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I1e29fe29f78a13de371ab21d3e40bf55fbc9c639
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15634
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
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>
Community-CI: Mellanox Build Bot
The idea behind "iobuf" is to have a single place for allocating data
buffers across different libraries. That way, each library won't need
to allocate its own mempools, therefore decreasing the memory footprint
of the whole application.
There are two reasons for putting these kind of functions in the thread
library. Firstly, the code is pretty small, so it doesn't make sense to
create a new library. Secondly, it relies on the IO channel abstraction,
so users will need to pull in the thread library anyway.
It's very much inspired by the way bdev layer handles data buffers (much
of the code was directly copied over). There are two global mempools,
one for small and one for large buffers, and per-thread queues that hold
requests waiting for a buffer. The main difference is that we also need
to track which module requested a buffer in order to allow users to
iterate over its pending requests.
The usage is fairly simple:
```
/* Embed spdk_iobuf_channel into an existing IO channel */
struct foo_channel {
...
struct spdk_iobuf_channel iobuf;
};
/* Embed spdk_iobuf_entry into objects that will request buffers */
struct foo_object {
...
struct spdk_iobuf_entry entry;
};
/* Register the module as iobuf user */
spdk_iobuf_register_module("foo");
/* Initialize iobuf channel in foo_channel's create cb */
spdk_iobuf_channel_init(&foo_channel->iobuf, "foo", 0, 0);
/* Finally, request a buffer... */
buf = spdk_iobuf_get(&foo_channel->iobuf, length,
&foo_objet.entry, buf_get_cb);
...
/* ...and release it */
spdk_iobuf_put(&foo_channel->iobuf, buf, length);
```
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ifaa6934c03ed6587ddba972198e606921bd85008
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15326
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Recently while updating the copyright notices
throughout SPDK the headers, env_dpdk copies of
DPDK headers were modified too.
This patch brings them to the exact version as
in DPDK upstream.
Change-Id: If30b8556386a539d81d2fc1a5e42293522ed91f5
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15856
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Michal Berger <michal.berger@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Copies of headers for DPDK PCI API were created before the
actual DPDK 22.11 release. The rte_bus_pci.h was
modified slightly with addition of rte_compat.h
include.
Please see relevant DPDK patch:
(1094dd9)cleanup compat header inclusions
This patch only makes the two align.
Change-Id: Ieb0397c6cf2d9027cf600bd0e064863b3782b846
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15855
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Michal Berger <michal.berger@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This patch adds "--msg-mempool-size" option for spdk app to allow
reactors' msg_mempool_size being configurable via commond line.
We tested the rbd_bdev performance for Ceph CTX sharing with high RBD volume count via bdevperf.
When testing with 256 volumes and limited Ceph CTX (e.g., 2 Ceph ctx for 256 volumes,
which are created though bdev_rbd_register_cluster),
error message "the *ERROR*: msg could not be allocated error message"
keeps showing and the bdev_perf program hangs.
We found the issue from the limited msg_mempool_size size, which is hardcoded
by SPDK_DEFAULT_MSG_MEMPOOL_SIZE in thread.h.
Therefore, we enable the "--msg-mempool-size" option to allow configurable msg_mempool_size.
Signed-off-by: Yue-Zhu <yue.zhu@ibm.com>
Change-Id: I54db7fd46247b2f18112bb994ecce6f4b7e5bf9c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15552
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
The description was swapped with removal release, causing the logs to
look like this:
foo_bar: deprecation 'v23.05' scheduled for removal in foo.bar hit 1 times
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I422a35c5ec20c8a817bed0dd5d565dfc53ef6dc9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15874
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Mike Gerdts <mgerdts@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Mellanox Build Bot
The kernel vfio_pci driver module introduced vf_token checking
mechanism since kernel version 5.7, and has been supported by
DPDK. So add support for it to deal with the scenario of VF.
Signed-off-by: Jun Zeng <jun1.zeng@intel.com>
Change-Id: Ie9700fa395327da4e847c6213167284c148a64e3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14424
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In _free_ctrlr(), ->endpoint can never be NULL, and the code was
self-contradictory; assume it's not NULL.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: I81a449123ca05f64460380dc3a8ad8af2143d166
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15831
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Use standard "sqid" naming for a log message.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: Icca8415cd17272ca7bd82667721c4131dd1df7f1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15828
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This fixes the behavior of spdk_vhost_(set|get)_coalescing() on
non-vhost-user devices.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ia17cd4c0ed4bad262090e05f83727c1516c21f92
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15772
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>
The current code for setting/getting coalescing setting only works with
vhost-user devices, while users can create virtio-blk devices with
non-vhost-user transport. Calling spdk_vhost_(set|get)coalescing() on
such device results in a segfault.
So, spdk_vhost_dev_backend interface is extended with methods to
set / get coalescing parameters. In the following patch, the virtio_blk
interface will be also extended with similar callbacks allowing us to
pipe coalescing settings to the appropriate transport.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ide5d5f633b17dcdbedb4b7804d5e45bf41373eca
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15771
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>
Both the length of a request and the number of ranges to copy are
controlled by the user, so we should check them and return an error
instead of asserting that they're correct.
This fixes the `test/nvmf/target/fabrics_fuzz.sh` test.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I3481c4bb1f2c7676df81f41dfc95ef063924222e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15805
Reviewed-by: Pawel Piatek <pawelx.piatek@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Found with misspell-fixer.
Signed-off-by: Michal Berger <michal.berger@intel.com>
Change-Id: If062df0189d92e4fb2da3f055fb981909780dc04
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15207
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Use the deprecation API to annotate and log the deprecation of
spdk_nvme_ctrlr_prepare_for_reset() using the tag
"nvme_ctrlr_prepare_for_reset".
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: I98fd840aa9acc028a49bb47daf4ab7e88f1eb818
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15756
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
srandomdev is *only* used to emulate arc4random, so only
bother defining it on Linux when it's needed. This avoids
unused errors on newer distros packaging glibc versions
that now defined arc4random.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I6e64a697d9633709cedd0198f75cf094d514562d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15814
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This patch fixes issue # 2809, by changing the max
completions processed per poll. A new parameter called
IDXD_MAX_COMPLETIONS is used to set maximum completions
processed per poll to 128 because we observed performance
degradation on a system with 16 NVMe SSDs at a queue depth
of 64 per SSD. When using DSA to compute the data digest,
the target application can issue upto 1024(16x64) request
to compute data digest concurrently to DSA. Limiting the
maximum completions processed per poll to 32 using
DESC_PER_BATCH cause up to 43% IOPS degradation.
Use IDXD_MAX_COMPLETIONS to control the number of
completions proccessed per poll in spdk_idxd_process_event
based on your workload. For example, if your application
is issuing 1000s of concurrent request to DSA you might
want to set IDXD_MAX_COMPLETIONS to a value higher than
128.
Change-Id: I2a1db993283a83a20266f40dac851728d63e6127
Signed-off-by: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15801
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 is in prep for adding a new compressDev accel_fw
module that will contain all of the DPDK compressDev specifics
on it, the vbdev will make calls to the accel_fw instead.
As the accel_fw has SW based compression, we want the configure
option to apply to building the vbdev module but not the accel_sw
software implementation or the upcoming compressdev module.
Renamed to "compress" as reduce is a term specific to the vbdev
implementation of the compression to be provided by the accel_fw
and thus the same reason why we leave the test flag called REDUCE
because it's controlling tests for the reduce library as well as
the vbdev module that is using reduce. The flag does not apply
to the SW implementation of compression.
This does not affect upcoming accel_fw compressdev module, that
will have its own configure option.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: If8ed3e48e1e3dabcaad1cd161289e78122cd9d58
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15179
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>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Per specification.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ic93349c7d3ed50fa6e502e39db0347141804d4c4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15673
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>
While lvs->destruct is set in a few places, it is never read. Since it
is not used, it is removed.
Signed-off-by: Mike Gerdts <mgerdts@nvidia.com>
Change-Id: Iee21e92c9049d143fca13930b4b5f328f9ec38f0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15716
Community-CI: Mellanox Build Bot
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>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Copy-on-write happens when cluster is written for the first time for
thin provisioned volume. Currently it is implemented as two separate
requests to underlying bdev: read of the whole cluster to bounce
buffer and then write of this buffer to the new location on the same
underlying bdev.
This patch improves copy-on-write flow by utilizing copy command of
underlying bdev if it is supported. In this case we have just one
request to bdev and don't need the bounce buffer.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I92552e0f18f7a41820d589e7bb1e86160c69183f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14351
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
New `translate_lba` operation allows to translate blob lba to lba on
the underlying bdev. It recurses down the whole chain of bs_dev's. The
operation may fail to do the translation when blob lba is not backed
by the real bdev. For example, when we eventually hit zeroes device in
the chain.
This operation is used in the next commit to get source LBA for copy
operation.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I89c2d03d1982d66b9137a3a3653a98c361984fab
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14528
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
In the following patches, nvme_rdma_poll_group_set_cq() will
touch not only CQ but also SRQ and receive WR objects.
All these resources are of a poller.
Hence for clarification, rename nvme_rdma_poll_group_set_cq()
by nvme_rdma_qpair_set_poller().
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ic59ba5a45833e39b1b2647c000c8b953f1031d6b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14910
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Factor out reset failed recvs operation into a helper function
nvme_rdma_reset_failed_recvs(). This will make the following
patches simpler.
For send operation, this change is not required yet, but in future
we may support something like shared SQ. Hence, we do this change
for send operation too.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Signed-off-by: Denis Nagorny <denisn@nvidia.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: Ib44acebe63e97e5a60ea6fa701b49278c7f44b45
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14171
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
In the following patches, poll group will have rsps objects and to share
the code between poll group and qpair, option for creation will be used.
As a preparation, merge nvme_rdma_alloc_rsps() and
nvme_rdma_register_rsps() into nvme_rdma_create_rsps(). For consistency,
merge nvme_rdma_alloc_reqs() and nvme_rdma_register_reqs() into
nvme_rdma_create_reqs().
Update unit tests accordingly.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Signed-off-by: Denis Nagorny <denisn@nvidia.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I92ec9e642043da601b38b890089eaa96c3ad870a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14170
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
When SRQ is supported, recv objects will be allocated by poll group
and qpair will associated and use them. In this case, we do not want
qpair to allocate and free recv objects. When connection is established,
it will be decided if SRQ is used or not. Hence, defer recv objects
allocation until connection is established.
Send objects are not affected directly by SRQ, but
nvme_rdma_register_reqs() no longer does any registration and deferring
send objects allocation makes the code more consistent. Hence, defer
send objects allocation until connection is established too.
Even after this patch, we rely on nvme_rdma_ctrlr_delete_io_qpair()
to free resources completely.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Signed-off-by: Denis Nagorny <denisn@nvidia.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: Ic151fad01009d92a7fc809a730e6e9dff1a365f3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14169
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Response objects will be in poll group when SRQ is enabled. But we want
to share the code to allocate and register response objects between SRQ
is enabled or disabled. To do it cleanly, move
nvme_rdma_qpair_submit_recvs() from nvme_rdma_register_rsps() to
nvme_rdma_connect_established(). A few clean up of error handling are
done in this patch. Unregistration will be done when qpair is
disconnected.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Signed-off-by: Denis Nagorny <denisn@nvidia.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I38dc5a6cb84a6bf56c01d5fb7f2cf3d3b63918e0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14168
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
This will make the following patches simpler.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Id3d7c025525b35c1c2b96027430789a8d8f2697b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14422
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Inline nvme_rdma_post_recv() into the callers.
We do not have any similar helper function for posting send WR.
This will make the following patches simpler and will be reasonable.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ia95a4b350942d20bdb65e84f7575c2dcf67c149b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14421
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Extract and inline the conditional nvme_rdma_qpair_submit_sends()
and nvme_rdma_qpair_submit_recvs() calls.
This will cralify the logic and make the following patches simpler.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Ibe217c6f4fb2880af1add8c0429f92b4de107da8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14420
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
When SRQ is supported, rsp array will be in either qpair or poller.
To make this difference transparent, rdma_req caches rdma_rsp and
rdma_rsp caches recv_wr directly instead of caching indecies.
Additionally, do a very small clean up together.
spdk_rdma_get_translation() gets a translation for a single entry
of a rsps array. It is more intuitive to use rsp.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Signed-off-by: Denis Nagorny <denisn@nvidia.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I61c9d6981227dc69d3e306cf51e08ea1318fac4b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13602
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Factor out processing recv completion and send completion into helper
functions to make the following patches simpler.
Additionally, invert if condition to check if both send and recv are
completed to make the following patches simpler.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: Idcd951adc7b42594e33e195e82122f6fe55bc4aa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14419
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Both max and min should be reset periodically. We can use the queue
depth sampling poller to reset these but the queue depth sampling poller
is optional. We extend the bdev_reset_iostat RPC to support mode to
reset all or only max/min fields.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I9ce54892f6e808f6a82754b6930092f3a16d51ff
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15444
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Add max/min_read/write/unmap_latency_ticks into the struct
spdk_bdev_io_stat.
When initializing or resetting the instance of the struct
spdk_bdev_io_stat, initialize max to 0 and min to UINT64_MAX.
Then update max if a new value is larger than the current max,
and update min if a new value is smaller than the current min.
For the bdev_get_iostat RPC, it prints max and prints min if min is not
UINT64_MAX or 0 if min is UINT64_MAX.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I1b30b3825c15e37e9f0cf20104b866186de788a2
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/14825
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Add a helper function bdev_reset_device_stat() to reset I/O statistics.
This funciton is used for the bdev_reset_iostat RPC.
We do not have any plan to use bdev_reset_device_stat() outside
lib/bdev. Hence, we do not add this as a public API.
Then, add a new RPC bdev_reset_iostat to reset I/O statistics of a
single bdev or all bdevs.
Resetting I/O statistics affects all consumers. Add a note to CHANGELOG
and doc/jsonrpc.md.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I97af09107b5c3ad1f9c19bf3cbf027457c4fbae7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15350
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Hold not only io_stat pointer but also num_blocks and blocklen in local
variables.
This will shorten and simplify bdev_io_update_io_stat(), and improve
readability and changeability.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I527b72538a169a1faafd32863ff539306a8763a9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15732
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
The following patches will add max/min latencies and more optional
counters. This factorization will improve the readability.
In addition to factorization, add spdk_likely to check if completed
successfully or not.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I57581ece2b73d486aa138f8d26a5afaf6953a322
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/15480
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Krzysztof Karas <krzysztof.karas@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
