When req->iovcnt is bigger than 1, `memset(req->data, 0, req->length)` is wrong.
Signed-off-by: Ziv Hirsch <zivhirsch13@gmail.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12216 (master)
(cherry picked from commit e749fa9c27)
Change-Id: Ie53eba686b4c5889bbde3b3644d51acbef303b42
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12483
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
OpenSSL 3.0 deprecated the MD5_xxx APIs, so switch
the md5 code in the iscsi library to use the EVP
APIs recommended by OpenSSL instead.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12240 (master)
(cherry picked from commit 92f0be87a0)
Change-Id: Ic5e3cd6e30ebc8b027f0715434cc3be045f1b770
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12482
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The config_bus_number is an offset within the config space reserved for
the devices behind the VMD, while bus_number refers to the actual bus
number assigned by VMD that depend on the VMCAP and VMCONFIG registers.
So, to access the mapped config space we have to use config_bus_number.
We didn't do that when resetting root ports', which could lead to
segfaults if these values were different, as we'd access unmapped
memory.
Fixes#2451
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12208 (master)
(cherry picked from commit 91aee82d74)
Change-Id: I4e7bbb81400462284014565099bec98f6171c8c9
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12481
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The root ports might have been configured by some other driver (e.g.
Linux kernel) prior to loading the SPDK one, so we need to clear it. We
need to before the scanning process, as it's depth-first, so when
scanning the initial root ports, the latter ones might still be using
stale configuration. This can lead to two bridges having the same
secondary/subordinate bus configuration, meaning that their config space
would map to the same memory area, which, of course, isn't correct.
This has manifested in issue #2413, where two root ports were configured
to use the same secondary bus. This caused an endpoint device to be
enumerated twice on two different root ports, with the first instance
being broken once the second port was configured by the SPDK driver.
Fixes#2413
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11863 (master)
(cherry picked from commit 6c3fdade83)
Change-Id: I5ce0931a84c1d23ccadb93fe39e8155ff1281474
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12478
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This will allow to use these functions without having to instantiate an
instance of vmd_pci_device. The following patch will use this to
perform some initial clean up before the scanning process.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11862 (master)
(cherry picked from commit 038f5b2e1b)
Change-Id: Icff92a4a429b259bec13eb6b0c1581aadbaae24d
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12477
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The current error handling of `spdk_mem_map_notify_walk` has off by 2
issue. This issue can split one memory region to multiple smaller
regions when calling the callback to unregister the memory region.
Also, in case of failure, the 1 GB maps of the map failed to be freed.
RDMA doesn't support this behavior and support calling the callback only
once for each previously registered memory region.
Signed-off-by: Aviv Ben-David <aviv.bendavid@vastdata.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11219 (master)
(cherry picked from commit c883771123)
Change-Id: I65b667f2e84533f234a2e330b20e9ad9eef32854
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12476
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
- Reduce the size of initial memory needed by OCF.
Number of allocator buffers equal to 16383 is tested to work
on 24 caches running IO of io_size=512 and io_depth=512, which
should be more than enough for any real life scenario.
This reduces initial OCF memory usage from 726 MiB to 392 MiB.
- Fix string handling for the name of the mempool.
Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10843 (master)
(cherry picked from commit c39647df83)
Change-Id: I40063ab1897c479c25904ae4096c5dae3351f73b
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12474
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The patch
nvme: Set dnr to zero for nvme_qpair_abort_reqs()
1b3172f726
did the change stated in the title.
However,
Revert "nvme/rdma: Correct qpair disconnect process"
c8f986c7ee
destroyed it for RDMA transport.
Additionally, we had still set DNR to 1 in nvme_qpair_init().
This patch fixes both.
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11644 (master)
(cherry picked from commit 7594030409)
Change-Id: Iee60ac24aa7e04cce0f394014c9d9afc9d2b56ec
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12473
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When function returned from the error handling the mempool on
'sess' was not released which lead to a memory leak.
Fixes issue #2393.
Signed-off-by: Weiguo Li <liwg06@foxmail.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11714 (master)
(cherry picked from commit e685574f76)
Change-Id: Ida3651e9369fb5c4948969480d398a723b2cb6a2
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12472
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We do the null check for 'fc_req', but already dereferenced it
before the check. Swap their position to avoid null dereference.
Fixes issue #2395.
Signed-off-by: Weiguo Li <liwg06@foxmail.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11721 (master)
(cherry picked from commit 30af49f77d)
Change-Id: I33b9e6b51b54f6ada9c072cf7ab0acda2622472f
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12471
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Since there is a pthread_mutex_unlock() in normal condition, another
pthread_mutex_unlock() in the "tmp != NULL" branch should be removed,
otherwise will cause a double unlock.
Fixes issue #2378.
Signed-off-by: Weiguo Li <liwg06@foxmail.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11642 (master)
(cherry picked from commit 8be59045ea)
Change-Id: I6c80a9527dd60e0b7c1d3c54b6da371b31118f02
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12470
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
With async connect, we need to avoid the case
where the initiator is sending the icreq, and
meanwhile the application submits enough I/O
such that the request objects are exhausted, leaving
none for the FABRICS/CONNECT command that we need
to send after the icreq is done.
So allocate an extra request, and then use it
when sending the FABRICS/CONNECT command, rather
than trying to pull one from the qpair's STAILQ.
Fixes issue #2371.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11515 (master)
(cherry picked from commit 635d0cbe75)
Change-Id: If42a3fbb3fd9d863ee48cf5cae75a9ba1754c349
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12469
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Group fields such that those not used in the I/O path
are at the end of the structure.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11514 (master)
(cherry picked from commit a97200ad45)
Change-Id: I43eca1faacd29a5bf34be6ee644191d865cd42a9
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12468
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>
This macro will be used in an upcoming patch
that needs to construct an nvme_request structure
outside of the standard nvme_allocate() routines.
Examined x86 optimized assembly with this patch,
and there is no change.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11513 (master)
(cherry picked from commit 56618eacb9)
Change-Id: I0f6b8500e06b56edc33f437f351536cf857d13d3
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12467
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>
if ctrlr->listener was NULL, nvmf_ctrlr_get_ana_state() returned
inaccessible even if ana_reporting was disabled. Then the corresponding
initiator received unexpected ANA error and could not process it
appropriately.
Change nvmf_ctrlr_get_ana_state() to return optimized always if
ana_reporting is disabled.
Additionally, check if ctrlr->listener is not NULL before calling
SPDK_DTRACE_PROBE3().
Fixes#2335
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11402 (master)
(cherry picked from commit c777cfa700)
Change-Id: Ib2376694cf89d85ec5687fba7e87439f494f30b0
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12465
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 <smatsumoto@nvidia.com>
In some scenarios, a split IO can immediately complete. For
example, a very large unmap operation to a newly thin-provisioned
blob has no operations to perform, so the batch for its operation
immediately completes.
But if it immediately completes, we can't recursively submit
the next split IO. So use variables in the context structure
to detect when an operation immediately completes, to allow
it to unwind and submit the next operation without recursing.
Fixes issue #2347.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11388 (master)
(cherry picked from commit c5d80a8bc3)
Change-Id: I8e4c121190c7d08152aa8de20cf6abc55b5edc46
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12464
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>
No functional change here, this only prepares this function for
some functional changes in the next patch. By adding the
do/while loop here we reduce the amount of whitespace changes
in the next patch.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11387 (master)
(cherry picked from commit b6992a90d3)
Change-Id: I09d64fd1fb69ee232af1d298619c762e562fdc79
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12463
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
There are errors occur that uninitialised value created by a stack allocation when running unittest_accel and unittest_nvme_rdma with valgrind.
Signed-off-by: Jaylyn Ren <jaylyn.ren@arm.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10559 (master)
(charry-picked from commit 3e937f07eb)
Change-Id: I4b48b472cc7c189cbcaf8ca772830a23118e7e17
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12267
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>
rte_dmadev was introduced in DPDK 21.11, and rte_vhost
is now dependent on it. So link rte_dmadev if we find
it and if CONFIG_VHOST is enabled.
Fixes issue #2374.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11594 (master)
(cherry picked from commit 662678dfbe)
Change-Id: Iccbf7cb897f51cbc9d545274d4d00a442b2fd353
Signed-off-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11637
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>
The new batching code needs to call the cb_fn for each of the
elements of the batch when a batch that hasn't been submitted
yet needs to be cancelled (due to an error in building it).
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11212 (master)
(cherry picked from commit 21c84e363d)
Change-Id: I6f94b27dd7c64f756193ec3532de98b644b41d7e
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11340
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In several functions. Busy handling also maans paying attention
to the rc when submitting a batch and not clearing chan->batch
unless the call was a success.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11221 (master)
(cherry picked from commit d5e63730ae)
Change-Id: Ic45b10ade2ebdcd845dc33e54dd9c93068ceb98c
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11339
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
With the new added property access API, we can send a internal
property access request to NVMf library, and we can use
it to reset controller.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10952 (master)
(cherry picked from commit 82a95325ce)
Change-Id: Iee8b1146d9eb31bc98a9b297e5c635e43e6fdb12
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11338
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
VFIO in QEMU uses region 9 as the PCI passthrough devices' migration channel.
The format of the region 9 migration region is as follows:
------------------------------------------------------------------
|vfio_device_migration_info| data section |
------------------------------------------------------------------
QEMU will access vfio_device_migration_info to controll the migration
process.
For SPDK vfio-user target, we also implement the BAR9 via libvfio-user,
and we also define the NVMe device specific migration data stored in
data section of BAR9. QEMU doesn't care about the format in data section,
it will help us to gather the NVMe specific migration data in source VM and
then restore the migration date to data section of BAR9 in destination VM.
The core idea to implement live migration will following the device state
change which is controlled by QEMU. First QEMU will try to STOP the device
in the source VM, and set the destination VM to RESUME state, SPDK will save
NVMe devic state data structure to BAR9 in the source VM once the subsystem
is paused, then QEMU will read BAR9 in source VM and restore the content of
BAR9 in destination VM, finally in the destination VM, we will restore the
NVMe device state include BARs/PCI CFG/queue pairs in the destination VM.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7617 (master)
(cherry picked from commit d1c2027d01)
Change-Id: I42e38f28c3ff59831be63290038b50d199d06658
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11337
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
To avoid re-use of descriptors that may have fields set that are
reserved by the one being used now. For example:
If a batch desc is being built and was previously used by a copy
we need to clear out the dst_addr field or things will explode
as this is a reserved field for a batch.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11308 (master)
(cherry picked from commit 057fdedbff)
Change-Id: I6ba50b76589e38a276683291f5ec2970c80e8aa8
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11323
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When subsystem is destroyed, it removes its listeners,
however transport level listeners remain active.
This patch removes all transport listerners when
the transport is being destroyed.
Fixes issue 2353/
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11307 (master)
(cherry picked from commit 29d94b7f01)
Change-Id: Ica7bcb0052b626aa62d0da9049bb8f216027dc49
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11314
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
to a single command
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11209 (master)
(cherry picked from commit 097ccf06a9)
Change-Id: Ic0ca65b7399f3cbc4153327d83de7db69de48709
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11281
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>
When doing live migration the migration BAR region is bytes stream
data, so here we use the helper function to save current controller
state into the stream in source VM and load it as internl data
structure from steam in destination VM.
We will remove the `unused` attrubute in next patch.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10336 (master)
(cherry picked from commit 646fb03fb8)
Change-Id: Ib44adb351c697b50b9220ce6943cc017137a6064
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11279
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>
When doing live migration, the destination VM will construct
ADMIN queue pair at the beginning, but the controller isn't
in READY state, we should not poll the ADMIN queue pair right
now. This is fine for normal controllers, normal controllers
will set ADMIN queue pair state in CC callback.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10621 (master)
(cherry picked from commit d73d43a9cf)
Change-Id: I0db36f75a463fb7476ee62323f9ed0c74c2451dc
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11278
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>
When doing live migration, there are some spdk_nvmf_ctrlr internal
data structures need to be saved/restored.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10058 (master)
(cherry picked from commit 95dd90039f)
Change-Id: Ie39482e8c49765c36fc3700fbac4ce47ef306f29
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11277
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>
When doing live migration we need to restore the AER commands
in the destination VM, so here to provide an API to save
these CIDs and the transport layer can save the value.
After migration in destination VM, we should allocate
new AER requests based on CIDs in vfio-user.
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10040 (master)
(cherry picked from commit 13f7510f1a)
Change-Id: I5881f833bbfacb0f030a2b135b4dd47726240378
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11276
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>
Fix issue: #2320
Only the primary process will do the unmap bar operation as for
the map bar operation.
The DevHandle is process specific and the issue here is the
secondary process's function pointer of DevHandle is not properly
set.
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11216 (master)
(cherry picked from commit 765cf74d07)
Change-Id: I95dddc76c6ce4be8775b6aaf54699002baffd3b9
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11274
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: GangCao <gang.cao@intel.com>
We saw this unexpected behavior by the current SPDK master.
Add the check to clarify this behavior occurs only when we use
Soft RoCE.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I3a5eaa9064a0601c65139e7868898545926d0dbf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11229
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This reverts commit eb09178a59.
Reason for revert:
This caused a degradation for adminq.
For adminq, ctrlr_delete_io_qpair() is not called until ctrlr is destructed.
So necessary delete operations are not done for adminq.
Reverting the patch is practical for now.
Change-Id: Ib55ff81dfe97ee1e2c83876912e851c61f20e354
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11228
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This reverts commit b9518a5540.
Reason for revert: Fix a degradation for adminq
Change-Id: I0e2c5e48a5ca34171fa98fa68216da4354b5d262
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11227
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Using the latest DSA we aren't supposed to (a) touch WQ space that
we aren't configuring and (b) touch WQ config fields that we are
configuring even if we are configuring that WQ. So, this patch
will read in initial values of only the number of desired WQs
and update them accordingly before updating the HW.
Also updates a few vars to use shorter local variables consistently.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I7641cdfc5ccc839e37a1d46d760248799a8fce1f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10981
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
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>
Found via inspection during spec review of latest HW. We were using the
wrong stride for the WQCFG regsiter when configuring but it just so
happened to be the right value for the current DSA version. We were
mixing up the size of the WQCFG register with the stride value used to
configure the next WQCFG regsiter as they are not contiguous in HW, we
need to read another capabilities bit to determine the address of the
next wqcfg to configure..
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I14d1ff95e0131fd30121aa955bfbc8c8fb3fc512
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10968
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Compliant with both current and next gen DSA.
Note: some fields in gencap were mapped incorrectly
previously, but this did not impact the SPDK driver
because the only times those values (max_xfer_shift
and max_batch_shift) were used were in asserts.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I9648184670f661166136e7898d0d8c7e07d8c746
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10966
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
There is a race condition between controller destruction and
subsystem state change, e.g. admin qpair may already be freed
when a namespace is added or removed. As result in function
poll_group_update_subsystem we may get heap-use-after-free error
Another problem is that some qpair's live time may exceed controller's
life time. To avoid it, start controller destruction process when the last
qpair finished the disconnect process (previously controller started
the descruction process before the last qpair starts to disconnect
and it could lead to raise conditions)
Fixes#2055
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: Ibc99b1d840e4796e1588cc217d65834bb556b909
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9995
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Using contructor/destructor to handle g_dpdk_sem will
help later in the series when splitting vhost fini
between vhost.c and virtio abstraction.
Otherwise multiple callbacks would be needed during vhost fini.
Ex. spdk_vhost_fini -> vhost_user_fini to stop the sessions ->
-> back to spdk_vhost_fini to remove vhost devices ->
-> vhost_user_fini to destroy the g_dpdk_sem
g_dpdk_sem will only be used from rte_vhost_user.c.
Until all references are moved, it is placed in vhost_internal.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I0505b906621f0eb0cb1226f96a3b6cf49f66778f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11055
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
There is no need to zero out the g_vhost_core_mask on vhost_fini.
Removing it will help later in the series when splitting vhost fini
between vhost.c and virtio abstraction.
g_vhost_core_mask will only be used in vhost.c and any cpu_mask
shall be passed to virtio abstraction after going through
vhost_parse_core_mask. There is no need to make the
g_vhost_core_mask accessible for virtio transports.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ic936c2a8dd1bb6f93b6f6209ea48e3278b19b54e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11054
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
In later patches rte_vhost functions will be moved
to rte_vhost_user.c. To prepare for this,
iterator is used in place of accessing g_vhost_devices.
While here, followed the same style of iterating in
spdk_vhost_config_json().
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I1b73c00dfe1391f359421d044686e49a8c6c9176
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11022
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
g_vhost_mutex scope is only within vhost.c as
it should. Meanwhile there is an internal vhost API to
use this lock from any of the vhost files.
Later patches in the series move some functions from
vhost.c to rte_vhost_user.c, where using only the
internal vhost API locks will be better suited.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I5916d4dc824ec980fa510fd3cbbd0c8e082d6611
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11021
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Creation of sockets is specific to rte_vhost, so it
functionality responsible for setting path for them.
dev_dirname is renamed to g_vhost_user_dev_dirname
and its definition is moved to rte_vhost_user.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I9bae67667b0f6624f2daf3244a048d10e94e553c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10631
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Changing the vsession coalescing setting is specific
to rte_vhost as such it should be moved the rte_vhost_user
that focues on rte_vhost specific functionality.
Renamed with vhost_user_* prefix to match the file.
Since the rte_vhost functions are still called directly from
vhost.c, temporarily they are added to vhost_internal.h.
Once implementing virtio transport abstraction is complete,
some will be removed and others will be replaced with
a generic callback structure.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I98b3746952cfe09fb724c49e4050efc0c42985a5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10630
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
For some time already the DPDK rte_vhost interface was
accomodating for other types of devices than virtio-net.
rte_vhost_compat.c file contained the use of DPDK rte_vhost,
rather than workarounds. To make that clear it is now renamed
to rte_vhost_user.c.
This patch is first in series that reworks vhost library
with two goals in mind:
1) Refactor vhost and vhost-blk to no longer depend on rte_vhost.
All references to that API will be moved to rte_vhost_user.c.
2) Add a transport abstraction for virtio-blk devices.
vhost-blk will now be able to expose virtio-blk using multiple
implementations of the interface.
First one will be vhost_user that depends on DPDK rte_vhost library.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ib6d4e4a6352069fa76e6b017ec203dab75f887b8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11052
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
A somewhat hidden functionality was present in spdk_vhost_dev_find().
Caller could match a vhost controller by controller name (socket filename)
or by full path to the socket.
This function is used by vhost RPC too.
The functionality of matching by full path was not documented,
nor matches what is presented in spdk_vhost_dev_get_name()
or vhost_get_controllers RPC.
This patch removes this functionality as part of series
to enable non-vhost-user type controllers, which might
not use the path to sockets.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I0e5ce75ac80ed8d1da962eabba86af69f59a43db
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10436
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Previously we didn't post the response for CREATE IO SQ command
until the queue pair is connected finally, but for coming live
migration support, we will connect IO queue pairs in the destination
VM, and this function will also be called for this case, so here
we add a flag to indicate the CREATE IO SQ case.
Change-Id: Iab4c64a7ebb72bcffbfff712dc729c40eead7c7d
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9464
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The miration region data structure is from `vfio_device_migration_info`
defined in `linux/vfio.h`, `vfio_device_migration_info` is in the 0th
offset of the VFIO_REGION_SUBTYPE_MIGRATION region, and in vfio-user,
we reserve first one page of BAR9 for this MMIO accesses.
libvfio-user already helps us to hide some implementation details
based on vfio migration specification, here we just use the two
fields to help the migration process.
Change-Id: I8917ba892bbfdfdf4f135f5d6b4923ab0e4a6250
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7628
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
We will report the live migration region to VM via sparse
mmap, offset after 0x1000 is the NVMe device state data
structure, and offset start from 0 is the structure
vfio_device_migration_info defined by the VFIO driver.
All accesses between 0x0-0x1000 will use the MMIO callbacks,
and accesses to NVMe device state will use shared memory map
way.
Change-Id: Ib456fc61f587c1bffa8b38506b4480a6066abe87
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7627
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
