RDMA qpair might be destroyed by defunct timer, so it can have
active recv elements in incoming_queue. This queue is cleaned
incorrectly, so recv element for the destroyed qpair still may
be presented in the queue and be processed later. That leads
to undefined behaviour.
Fixes#1086
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477957 (master)
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit 4af2b9bfb9)
Change-Id: Ieae186b2d2dce4ec88ab886b26165f6ef98e8d05
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478357
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
The following scenario might occur when nvmf_tgt is stopped:
1. nvmf_tgt receives SIGINT, changes state to NVMF_TGT_FINI_STOP_SUBSYSTEMS
2. In this state nvmf_tgt stops all subsystems and disconnects associated qpairs
3. In the case of RDMA qpair, its state will be changed to IBV_QPS_ERR.
Once qpair changes the state to IBV_QPS_ERR, RDMA device generates
LAST_WQE_REACHED event when there are no more WQE that can be sonsumed
from the SRQ by this qpair.
4. When all subsystems are stopped, some of qpair may still be alive since they
haven't received LAST_WQE_REACHED event yet.
5. nvmf_tgt stops all poll groups and forcefully destroyes any qpairs linked to them.
6. At this moment LAST_WQE_REACHED event might be generated and received in another thread.
Handler of this event sends a message with a pointer to qpair. The qpair itself may already
be destroyed.
7. Thread that owned qpair receives a message (LAST_WQE_REACHED) with a pointer to alredy destroyed qpair and
destroyes it for the second time when all pointer are invalid.
ibv events related to qpair should be handled by the thread that
owns this qpair. This commit adds a new structure that describes
ibv event, helper functions for sending the event and a list
of events per rdma qpair; add syncronization for LAST_WQE_REACHED event
Fixes#1075
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476712 (master)
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit dc84fbaaa1)
Change-Id: I22bff89741708df2518760934ecb4e33fad49473
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478355
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The hardware sgl format can describe large contiguous
buffers using just a single element, so it's more
efficient that a prp list even for a single memory
segment. Always use the sgl format.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475542 (master)
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
(cherry picked from commit bed4cdf6c7)
Change-Id: I9c62582829f0d64dcd1babdbc48930ddb4d9e626
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478354
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
This patch attempts to solve naming conflict between
CUSE devices created by different SPDK instances.
Each NVMe device is enumerated by SPDK process from 0
up to 127. When process attempts to start cuse device
tries to set exclusive lock on temporary file
"/tmp/spdk_nvme_cuse_lock_<index>" and keep it until
device will be stopped.
If setting lock fails, index is incremented.
It prevents to use the same controller index from
few SPDK instances.
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474829 (master)
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
(cherry picked from commit 46316bb5db)
Change-Id: If744ac23f813bd992efb80ae2b61a1acefb5054c
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478353
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When serializing extents, run-length encoding is supposed to
1) RLE all sequential LBAs
2) RLE zero LBAs (unallocated)
There is one special case, with sequential LBAs that start
with 0 LBA. This is RLE as 1) case, but results in descriptor
matching case 2). Which causes loss of allocated clusters.
This requires following conditions to be met:
- blobstore has just a single cluster reserved for MD
- blob is thin provisioned
- first allocation occurs on cluster_num=1
For last part to be true, very first write for blob has to be
issued to LBA between cluster_size and 2*cluster_size.
Causing allocation of second cluster in blobstore and assiging
it LBA equal to number of LBAs per cluster.
To fix this, case 1) disallows to RLE zeroes.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475494 (master)
(cherry picked from commit 0d1aa0252d)
Change-Id: I136282407966310c882ca97c960e9a71c442c469
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478351
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Now all registered producers should be stopped (unregistered) before
NVMe detach, otherwise NVMe controller cannot be safely detached.
This patch allows to stop all not unregistered io producers before
NVMe detach:
1. Callback to the "struct nvme_io_msg_producer" to stop producer
started on selected controller.
2. On nvme_io_msg_ctrlr_detach() if there's some unregistered producers,
stop all before freeing resources.
This approach also fixes issue with not to stop CUSE device when
NVMe controller is detached without unregistering producer (github
issue #1033).
Fixes#1033
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474273 (master)
(cherry picked from commit fd2af7afa9)
Change-Id: Ia1ffef566bb745edb55c54d6786ea481a35bbefd
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478350
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
DPDK by default guesses that it should be using iova-mode=va
so that it can support running as an unprivileged user. However,
some systems (especially virtual machines) don't have an IOMMU capable
of handling the full virtual address space and DPDK doesn't
currently catch that. Add a check in SPDK and force iova-mode=pa
here.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475149 (master)
(cherry picked from commit 97b0f7733f)
Change-Id: Ib3a5691a584190feaab4b9064b5a500e361328f2
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478349
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This patch uses lowlevel fuse functions to process messages to
eliminate the need to use signals to interrupt blocking read
operation in fuse_session_loop().
Fixes#1032
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/473233 (master)
(cherry picked from commit 88808c5ab7)
Change-Id: Ie9c9ea76cc135c383f5757864aa2d84ac9eb3da3
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478348
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
In DPDK, the ppc iommu support does not currently allow for
iova-mode=va, but DPDK doesn't detect ppc and so still attempts
to guess iova-mode=va in some modes. Force iova-mode=pa from
SPDK to fix this.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475148 (master)
(cherry picked from commit 07ca02210a)
Change-Id: I6a1ee25ab74873826ac211c3e0dfdf54afc74502
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478347
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
NVMe hotplug must be monitored in the primary process -
DPDK doesn't support trying to handle it in the
secondary process.
This issue was somewhat masked previously in secondary
processes, since usually it would just probe(NULL) which
meant probe all attached NVMe controllers. So in the
secondary process, we would probe just once, and create
the hotplug fd - it would never actually try to monitor
it.
But when explicitly specifying multiple trids in a
secondary process, probe would get called multiple
times. First time would be fine since it only creates
the hotplug fd. But second time would segfault since
monitoring for hotplug requires checking the DPDK-allocated
context which doesn't exist in the secondary process.
Fixes issue #1063.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475015 (master)
(cherry picked from commit c3aaaa0181)
Change-Id: I2a9a91e222c206034293d90e30e3f598c8d7baa8
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478344
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The jsonrpc client has a limit of 1024 JSON values per
request which is hardly enough for any meaningful config.
For example, calling getbdevs for 24 NVMe drives require
~2300 JSON values.
I kept the original 1024 limit for the RPC server where
it makes sense to have a smaller limit and introduced
a seperate limit for the client.
Signed-off-by: Michael Haeuptle <michael.haeuptle@hpe.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/473568 (master)
(cherry picked from commit 92df995526)
Change-Id: Id0300991b76151e4003e323f5ea29bc5fc0d2d11
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478342
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Michael Haeuptle <michaelhaeuptle@gmail.com>
In release builds, the assert() is compiled out, making
it look like the rc value is never referenced after it's
set.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/473472 (master)
(cherry picked from commit 750f2b4b3d)
Change-Id: I59305b0e928f2044146e30b7addc86f81e7a1d3f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478584
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
This at least allows the caller to know there was a
problem, and that the messages wasn't actually sent.
SPDK by default creates huge rings so this problem
should never occur, but out-of-tree use cases may
send messages much more often and require at least
a notification when it fails.
While here, change the thread check to an assert.
There's no need to work around someone calling
this function with a null thread parameter.
Fixes issue #811.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472438 (master)
(cherry picked from commit 4036f95bf8)
Change-Id: Ie6d432d616be45c7a4232aff1548cef198702bc0
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/478442
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Previously iSCSI task was created after allocating data buffer
and reading all data, and hence creating iSCSI task and processing
iSCSI task were not separated.
However, the recent refactoring separate PDU header handling and
PDU payload handling, and then inserted allocating data buffer and
reading data segment in the middle.
If any critical error occurs during allocating data buffer or
reading data segment, PDU payload handling is not done, and hence
created iSCSI task is left in PDU receive process.
If any critical error occurs, the current connection starts exiting
and there is no way to continue PDU receive process.
The task left in PDU receive process is never freed, and hence
LUN hotplug or exiting connection never complete.
This patch do the following:
- Consolidate freeing pre-allocated PDU to spdk_iscsi_conn_destruct()
because this is the only path to exit connection.
- Abort SCSI task of the task left in PDU receive process if found
when freeing pre-allocated PDU. If the task is not SCSI or Data Out,
remove it simply.
Fix issues #1018.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472896 (master)
(cherry picked from commit cd654cc512)
Change-Id: I8a2464c446c43bf4cfb5afbc0cd78b5bdef7d080
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472988
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This reverts commit Iad6ecdc37493fa9f2d7ccab262a2c75dac2fcd48.
Both estimated cause and code change were wrong and didn't fix
the issue.
The next patch will fix the issue.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472895 (master)
(cherry picked from commit 832d90c1e2)
Change-Id: I00c8bb515ee39522c0e744dccfb839af15e946c4
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472987
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>
Users already have to poll the admin queue, so embed the io_msg
queue polling there to simplify the API.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472833 (master)
(cherry picked from commit 11739f3cb1)
Change-Id: I4d4d3be100be0798bee4096e0bbda96e20d2405e
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472963
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Disconnecting qpairs from the admin thread during a reset led to an
inevitable race with the data thread. QP related memory is freed during
the disconnect and cannot be touched from the other threads.
The only way to fix this is to force the qpair disconnect onto the
data thread.
This requires a small change in the way that resets are handled for
pcie. Please see the code in reset.c for that change.
fixes: bb01a089
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472749 (master)
(cherry picked from commit 13f30a254e)
Change-Id: I8a39e444c7cbbe85fafca42ffd040e929721ce95
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472960
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
_iscsi_conn_remove_lun() which is the callback to LUN hot-removal
returns immediately without closing the LUN if the connecion is
already in exiting, then expects that the LUN will be closed by
after the connection moves to the exited state.
LUN hot removal process doesn't check any R2T task if it is not
pending in SCSI layer but connection close process checks any R2T
task even if it is not pending in SCSI layer.
LUN hot removal will not complete until all LUN accesses are closed.
iscsi_conn_close_lun() checks if the LUN is already closed or not,
and so it will be no harm even if _iscsi_conn_remove_lun() calls
iscsi_conn_close_lun(). If the connection is in exited state,
all LUNs are already closed.
This patch changes _iscsi_conn_remove_lun() to return immediately
if the connection is in exited state.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472507 (master)
(cherry picked from commit 1ef8449feb)
Change-Id: Iad6ecdc37493fa9f2d7ccab262a2c75dac2fcd48
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472776
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Recent patches refactored iSCSI target to separate PDU header
and payload handling. However for SCSI Data-Out PDU, the division
of roles done by refactoring was wrong. Before refactoring, LUN
hotplug was checked after sending R2T, but after refactoring LUN
hotplug is checked before sending R2T. This change stopped PDU
exchange between iSCSI initiator and target and caused timeout of
LUN removal.
This patch restores the original ordering of checking LUN hotplug
and sending R2T by changing the division of roles.
SCSI Write Command PDU handling don't have any issue related with
this.
Fixes#1004
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472308 (master)
(cherry picked from commit 84f59335c2)
Change-Id: I7b2866d8394b522fb5420d2936de2fbddc7d1daa
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472775
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
If we disconnect qpairs without taking the lock, we run the risk of
trying to double free qpair resources before they have been marked as
NULL.
For example, polling on one thread and calling
nvme_rdma_qpair_disconnect from one thread while doing an
nvme_ctrlr_reset on another thread. nvme_ctrlr_reset will call down to
nvme_rdma_qpair_disconnect on the same qpair and without any locking it
can result in trying to destroy the qpair resources multiple times.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472413 (master)
(cherry picked from commit a4925ba744)
Change-Id: I9eef6f2f92961ef8e3f8ece0e4a3d54f3434cff8
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472711
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>
Modifying the target_disconnect.sh test to include an example of
transport_id failover for an NVMe-oF controller.
Change-Id: I746ed737ab56c7dec6ee99e840c631ba46ee359e
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472230
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Added RPC commands to register/unregister CUSE devices
to NVMe controllers:
- bdev_nvme_cuse_register
- bdev_nvme_cuse_unregister
Additionally two RPC now return CUSE device names:
- bdev_get_bdevs for namespaces
- bdev_nvme_get_controllers for controllers
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I69c4bf41ec8f78a7522894268a67dd733881712f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472211
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
IO request handling implementation for CUSE namespaces:
- NVME_IOCTL_SUBMIT_IO,
a. SPDK_NVME_OPC_READ,
b. SPDK_NVME_OPC_WRITE,
Other ioctls related to the namespace:
- NVME_IOCTL_ID,
- BLKPBSZGET,
- BLKGETSIZE,
- BLKGETSIZE64,
Additionally NVME_IOCTL_ADMIN_CMD has been implemented
for namespace to send admin cmd to its controller.
Change-Id: Iaa2a5ee409b35c682ae8b2bd2309e8b67eb51981
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/469691
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>
NVMe character device implementation. This patch adds implementation
of IO producer using CUSE library. It allows to create nvme device
nodes in linux kernel for controller as well as for namespace and
process ioctl requests as usual from linux environment.
Both devices (controller and namespaces) are exposed as character
devices.
To compile NVMe CUSE module use "./configure --with-nvme-cuse".
Names for created CUSE devices can be retrieved using
spdk_nvme_cuse_get_ctrlr_name() and spdk_nvme_cuse_get_ns_name().
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I0fc9a9a1ef3c9c2b3112d07c2b4b1f8d49665ee1
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466917
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This patch allows to send IO requests from external module to the nvme
device.
External module should call nvme_io_msg_ctrlr_start() to start IO message
producer on the controller and enable sending messages.
nvme_io_msg_send() is used to send IO to NVMe driver thread context,
where passed function will be called. Allowing the external module to
issue IO as needed.
NVMe driver users should poll spdk_nvme_io_msg_process() to move forward,
sending IO from external module and process their completions.
Change-Id: Ie59abac69870c4e4daa50120c747f3b620395921
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471386
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This change allows setting of the NVMe completion queue
CDW0 in spdk_bdev_io_complete_nvme_status.
Before that change, handling of vendor specific NVMe IO
commands was limited since there wasn't a way to return
command specific info back to the initiator.
Change-Id: I250d5df3bd1e62ddb89a011503d42bd4c8390f9b
Signed-off-by: Michael Haeuptle <michael.haeuptle@hpe.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470678
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>
Since we use big buffer to read the data, so the incoming
data may already be read when the req is waiting for the buffer.
So if we use the orginalstatement machine, there will be no
read event will be generated again.
The quick solution is to restore the original code, since
for req which has incapsule data, we not need to wait for the
buffer from the shared buffer pool.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: Ib195d57cc2969235203c34664115c3322d1c9eae
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472047
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This can be useful when trying to perform multipath failover at the
application level. However, the controller must be in the failed state
before calling this function.
Change-Id: I5403c0036fed5dd3600ee20592925297494ba8aa
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470699
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
These will be useful helper functions for the trid modification code
that gets introduced later.
Change-Id: Ief73e3045710bf35c511794c19b4dfefb93018f1
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471780
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
While it is unlikely that a single qpair will be failed, it is important
to make it possible to reconnect a single qpair.
This function is also handy at the application layer when going through
a reconnect workflow. If we get -ENXIO from a qpair when we poll, we
will turn around and call this function. If we get -ENXIO from this
function, then we know the whole controller is failed and we need to do
a reset.
Change-Id: I6a8ea0ce27fce2f5fc0a5b3db05834acd68e6a39
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471417
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Ideally we'd have a way to query if DPDK is already
initialized but we don't have that yet. We want that
for the case where we have an SPDK application that's part
of a framework that may (or may not) have already initialized
DPDK. If it's already been initialized, let's print an
error message that isn't quite as inflammatory.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ifc095245dcdef24cdeeaab2dbe791ca4e840870e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471422
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
The header file already says it returns negative errnos,
but the env_dpdk implementation was just returning -1
on failure.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie2236f83094672548327dba945b33e3f28fee338
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471421
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Previously, get locking range info will allocate
memory everytime. Thus calling this function multiple
times will cause memory leak.
Now, we allocate only if it is NULL. If it is not,
then we just memset to zero.
Change-Id: If048416a2056176f86206a33e2b5db210288fe4f
Signed-off-by: Chunyang Hui <chunyang.hui@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472112
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Some nvme drives might take 6-7mins for
this operation. Thus, introduce async function
to avoid waiting.
Change-Id: Id48478aec653d3fb75a3c5ce75d4997284ed016c
Signed-off-by: Chunyang Hui <chunyang.hui@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468916
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Initialize the hotplug structures if the device is hotplug capable.
Change-Id: I2bab4ec820f37f352d79203e3ad37aef31bdc3c6
Signed-off-by: orden smith <orden.e.smith@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470650
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Check PCIe Capability slot_implemented field before hotplug structures
instantiation.
Change-Id: Id40c10839e51b238a6ffd64592052604d50e2850
Signed-off-by: orden smith <orden.e.smith@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470649
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Changed the way the devices are enumerated, so that only the first one
performs active enumeration and stores the result on the config bars of
the devices, while subsequent ones use those saved values. It prepares
the code for hotplug support, as it'll make it easier to detect it.
Change-Id: I38c8eada91f573a13d1739a73bff3614ae7e1f89
Signed-off-by: orden smith <orden.e.smith@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/469936
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Moved the base / limit register reset to a separate function. It'll
make it easier to perform the reset only under specific conditions are
met (e.g. only during the initial setup).
Change-Id: I5231e6d5a1e1673885bf549b1b6761092b4b62be
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472248
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This patch adds support for LED management of the devices behind VMD.
It's possible to set four different states: off, on (fault), 1Hz
blinking (rebuild), and 4Hz blinking (identification).
Change-Id: Icce25d0b46c1c4bed803f18201606203980ccc38
Signed-off-by: orden smith <orden.e.smith@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470651
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
