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>
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>
We will use the NVMe device state data structure to save/restore
a NVMe controller in source/destination VM.
NVMe device migration region is defined as below:
----------------------------------------------------------------------
| nvme_migr_device_state | private controller data | queue pairs | BARs |
----------------------------------------------------------------------
Change-Id: Idc73976e1de7f6da2da58e71db86df8cbb0d314d
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7626
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>
For CREATE IO SQ command, we will defer to post completion
until the SQ was connected, we may call post_completion()
in different threads, so here we will send a message
to CQ thread when necessary.
Change-Id: I87a0f8982811c76ce8eb49db6a136f4cbe6e0a93
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11078
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
SPDK RDMA target reports msdbd=16, these addtitional
SGL descriptors are located in capsule. The user can
set ICD size lower than required for msdbd=16. This
patch verifies that ICD can hold all additional SGLs.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I875d40e14e6506c39169d084e56df7ca5d761209
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10686
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>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
We will assign each SQ with different poll group in round
robin way by default, this may cause race condition to
post completions to one CQ in different threads, so here
we will assign the SQs which share one CQ into same poll
group.
Also enable multiple cores NVMe compliance tests so that
to cover shared IO CQ case.
Change-Id: I9d7cc78aaedceed23986d9f89ed945e0eb337e09
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11115
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: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Previously we mixed SQ/CQ definition together, one queue pair
data structure may contain CQ,SQ or both CQ and SQ separately,
while here, we split the queue pair definition into SQ and CQ
respectively as code cleanup.
The NVMf library uses queue pair concept, but for vfio-user
case, each SQ created by VM is mapped to NVMf queue pair, so
we also change `connected_qps` to `connected_sqs` to reflect
the fact.
No actual code logic change in this commit.
Change-Id: I293ccbfbf054fe864d348fc56793dd1ccd366f6d
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11036
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>
libvfio-user will call quiesce callback when there are
memory region add/remove and device state change requests
from client, and in the quiesce callback, we will pause
the subsystem so that it's safe to do everything after
it, then after quiesce callback, we will resume the
subsystem. The quiesce callback is also used in
live migration, each device state change will quiesce
the device first.
Change-Id: I3a6a0320ad76c6b2d1d65c754b9f79cce5c9c683
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10620
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 use the controller state to implement the coming
device quiesce feature, it's safe to do anyting when
a subsystem is in PAUSED state.
Change-Id: I3b466ed01848e668a1ffcea1d4f1466e971afa23
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10619
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>
Users may remove the listener while VM is connected, the endpoint is
associated with Unix Domain socket file, we should destroy the endpoint,
however, the controller maybe still active for now, because nvmf
library will help us to disconnect all queue pairs in asynchronous
way. Here we use the same way as the NVMf library to destroy the
controller when there is no connected queue pairs.
Fix#2246.
Change-Id: I0775d5294269d848d859968edafc8eaa1d89a32c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10379
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>
The controller may be freed eailer than endpoint, so we still
need to unregister the memory region from SPDK. The case
can happen when removing the listener while VM is connected.
Change-Id: I95d49cefdbff3e0bdea316fac824ef8b218fcd2c
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10378
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>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
We should hold the transport lock to iterate endpoints.
Fix issue #2313.
Change-Id: I8e0539a51e843a3299908d9da7749fe9becb5e7e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11037
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: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: John Levon <levon@movementarian.org>
We met an issue that client got a NVMe completion with old SC
bit, so we add a memory barrier here to ensure the NVMe completion
is fully populated.
Fix issue #2323.
Change-Id: I7887d789a0acd3634a10aa7dc8de81a153137ae7
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11076
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: John Levon <levon@movementarian.org>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There is an error case that the block device didn't complete
outstanding IOs during the controller reset or shutdown, so
the NVMf library will wait until all the IOs returned from
the backend, however, so here we added a timeout timer, when
the time expired, we will try to reset the block device which
hold the outstanding IOs.
Fix#2194.
Change-Id: I8d0746335e1f20a09e6a9ea87730551808a898d1
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9909
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: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
For the purpose to support shared IO CQ feature, we will construct
the queue pair data structure at the beginning, and setup SQ/CQ
separately in CREATE IO SQ/CQ routine.
Previously we will disconnect queue pair when got a DELETE IO CQ
command, now we disconnect queue pair when got a DELETE IO SQ command,
and in the disconnect completion callback, we will release the IO SQ
resources, there is a case that the VM will just RESET/SHUTDOWN
controller when IO queue pairs are connected, for this case, we
will also try to release CQ resources in the disconnect completion
callback.
`free_qp` function now is only called when destroying a controller.
Change-Id: I45ec679ddb63bdf1feeba5dc2bd39cae3ba4aa89
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10532
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Currently we only use round robin way to assign queue
pair to each poll group.
Change-Id: I8efaf3ef25402102dd1eaa7f7aa8bd8bbe071c25
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11114
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>
Reviewed-by: John Levon <levon@movementarian.org>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Acceptor poller is registered using rate value
from transport opts structure, but this structure is
initialized on generic transport layer when create()
function completes, so at this time acceptor poll rate
is 0.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I2138825f3ff9dd3cc0ccaa65e8d5c23aab338ad4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11095
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: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Set the SQ/CQ size to 0 so that we will not try to remmap
the ADMIN queue pair in the memory region callback before
the ADMIN queue pair was enabled.
Change-Id: I739a2ec3abcb54b17f31f2bc120312cd02ffeef1
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10531
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>
When deleting a CQ, we will use its reference count to check
how many SQs associate with it.
Change-Id: Ic82e50de0fa92d2f03119ac2cc90ef86a0ea375e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10530
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: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This is a preparation to support shared IO CQ case, and we will
create/delete SQ/CQ separately, so define the queue state as the
first step.
Change-Id: Ie7b5807dc4aa5a2c117e15f61f3a9baa60135653
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10529
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: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add dtrace probes aroung qpair/controller/subsystem management
to help with debugging issue #2055.
Change-Id: I0b981bffadee3fe4172ad6916c059bf357959dde
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10237
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: Aleksey Marchuk <alexeymar@mellanox.com>
It makes it easier to read the logs, as the state values are printed as
integers.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I70a9e8860401c18e9305a5fc5771df0bc564d337
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10800
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This patch adds support for using zero-copy operations to execute IO
requests in the TCP transport. Of course, they're only used if the
underlying bdev supports them. Additionally, only requests with no
in-capsule-data can be executed using this mechanism.
Added several new states to accommodate for the difference in a way
zero-copy is handled. Also, these flows very depending on the type of a
request (read or write). It stems from zero-copy semantics: to perform
a write we need to wait for zcopy_end completion, while for reads
zcopy_end can only be submitted once we send all of the requested data
to the host.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ie02b494c24bc1acc98557cb4b02e867abf9064e4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10796
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Zero-copy requests are kept on the outstanding queue for the whole
duration of the request - from the initial zcopy_start submission to the
completion of zcopy_end. This means, that there's a period in which a
request doesn't wait for a completion from the bdev layer, but is still
on the oustanding queue (after zcopy_start callback, before zcopy_end
submit). If a qpair gets disconnected while a request is in this state,
we need to manually force its completion, as otherwise it might hang
indefinitely (e.g. waiting for host data).
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I53731b8e363b725efa564ca3c7d89b46f5fb2a24
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10793
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
The zero-copy requests can also be queued when a subsystem is paused, so
we need to properly resume and submit them by using zcopy_start.
Since only requests that haven't received the zero-copy buffer (i.e.
before zcopy_start was called) can be queued, we don't need to bother
with checking zcopy_phase.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ie629688f6961eb2ae05741df496720b91be4d80d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10792
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
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>
It ensures that we decrement io_outstanding counter for requests for
which zcopy_start failed. Also, removed a note stating that such
requests are reverted to regular IO path, as this is not the case.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I8eaee88abfd94b73614b367fef9bb938c9962617
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10791
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Since spdk_bdev_zcopy_end() cannot really fail (it only fails if we pass
a bad bdev_io), we can simplify the nvmf zcopy_end functions by making
them void and always expect asynchronous completion.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I6e88ac28aba13acadea88489ac0dd20d1f52f999
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10790
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>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Since this path now supports sending zero-copy, use it for zcopy_start.
Additionally, it makes it possible make zcopy_start void, as it reports all errors
asynchronously via request_complete(), and remove some of the duplicated
error checks.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I41f43ce1651432d9a7d74e3680d4a3f780128a1d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10789
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>
If a request gets resubmitted and is completed immediately (i.e. the
processing function returns SPDK_NVMF_REQUEST_EXEC_STATUS_COMPLETE), the
upper layer needs to be notified via spdk_nvmf_request_complete().
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I236d6e68d1d7d83afdaa30d8bb07e2b133f43155
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10788
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Additionally, the NVMe completion status is now updated and the IOs are
queued if the bdev layer doesn't have enough IO descriptors. It makes
the zcopy operations behave similarly to the other IO operations.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I455ae781e32aa6e60d144d2c91f109bd8be46664
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10787
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>
It makes their names consistent with the bdev API.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I314051f0980b46959d6560aa25885f13b4c28f2a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10786
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: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The zcopy_start requests are now executed through
nvmf_ctrlr_process_io_cmd. It makes the zero-copy share checks with the
regular IO path.
Note, that zcopy_end doesn't utilize this path and is directly submitted
to the bdev layer, as it doesn't need to perform these checks (they were
already verified in the accompanying zcopy_start).
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ic9ffecb15e7c351a9d60e731cc711d6500b845db
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10785
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This will allow the zero-copy requests to share more code with the
regular IO path.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Idb57447694903a56d984f95aa2f1a3bddc3e4e82
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10784
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>
It will make it possible to submit zero-copy requests through
spdk_nvmf_request_exec().
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ibc14fe77cd477b11ed55d1350a7486caaad81add
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10783
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The code should never reach these functions for requests using
zero-copy.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: If9f30e05a43b340a982604d5b985242d63ce252b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10782
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>
It's more descriptive that way, as it's clear the function works on a
single request. Also, passing a request instead of zcopy_phase makes it
more convient to use.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: If4d7b087511e128f3590ac7b3b5adcb8ace12003
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10781
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Dong Yi <dongx.yi@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>
It makes it possible for the user to specify whether a transport should
try to use zero-copy to execute requests when possible.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I40a92b0d7a6707f4c9292795f380846acb227200
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10780
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>
we don't remove the socket fd from socket group when
nvmf_tcp_poll_group_add() return error, and when
closing the socket there is an assertion.
This was found via llvm_nvme_fuzz via TCP transport.
Change-Id: Ib4ab6fc3fc5e2bc6a9545f6ce854bae8f1157fd5
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10849
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We have very frequent failures when we run test/nvmf/host/multipath.sh
in the test pool.
Call stack showed nvmf_stop_listen_disconnect_qpairs() accessed
qpair->ctrlr even if qpair->ctrlr was NULL.
nvmf_stop_listen_disconnect_qpairs() did not check if qpair->ctrlr is
not NULL before accessing qpair->ctrlr->subsys.
When a qpair is added to a poll group, qpair->ctrlr is cleared to NULL.
The test code test/nvmf/host/multipath.sh executes multiple
reconnects for path error.
So a conflict might occur between adding a qpair to a poll group and
disconnecting a qpair in a poll group.
In this case, it may be acceptable even if we disconnect a qpair whose
qpair->ctrlr is NULL. It will be better than SIGSEGV.
Fixes one of the issues in #2300
Signed-off-by: Shuhei Matsumoto <shuheimatsumoto@gmail.com>
Change-Id: I308fcb886dd410d01e3361c1850dec9a8eacbccf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10860
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
These definitions will be used in the next patch to check if
device is rxe
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: Icc073344103991ff24fc3bb88a1ceb9867de6f6a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10727
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: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Some devices may support SRQ depth lower than defaulut
value 4096
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I58da0ac268a6d4c4a7e3b500ae37b8fad4810e17
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10654
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
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>
