this is a prework for further changes - with lock on generic layer
lock on specific transport (e.g. tcp, rdma) layer becomes optional
possibly it won't be required if some contract introduced on public
interfaces (to be considered)
- spdk_nvmf_poll_group_[create|destroy]
- spdk_nvmf_tgt_listen_ext, spdk_nvmf_tgt_stop_listen
- spdk_nvmf_get_optimal_poll_group
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Ib132babf9e7022342129fe795991cdad834e7f53
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13665
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@nvidia.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
When there are not enought transport buffers for
multi SGL request in state NEED_BUFFER, WRs
received from the data_wr_pool are returned back
to the pool. However rdma_req->data.wr.next pointer
still points to the first WR from the pool. Usually
it doesn't cause any problems since rdma_req will
try to fill buffers again, but when qpair is being
destroyed, all requests are completed forcefully.
When the request is completed and data.wr.next
pointer is not NULL, we'll try to put already
released WRs into the pool one more time.
That corrupts the pool and leads to undefined
behavior.
Fixes#2541
Signed-off-by: Alexey Marchuk <alexeymar@nvidia.com>
Change-Id: I238b92eec132d8d845330362af6f335421177454
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13760
Reviewed-by: Michael Haeuptle <michaelhaeuptle@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The function `nvme_ctrlr_init_ana_log_page` is exactly
same with `nvme_ctrlr_update_ana_log_page`, so remove it.
Change-Id: I1ad51635f47cf95cfa6de217e3b9144885c3b74e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13652
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
allow DSA device to async offload crc32 calculation in nvmf-TCP
This patch can use DSA to accelerate crc32 computation, making
the io performance of TCP paths using crc32 approach the io
performance of TCP paths that do not use crc32.
Using SLIST to minimize the performance drop. SLIST has less
operation compared to TAILQ.
Thinking about memory thrashing, we should use the same memory as
possible to receive new PDUs. So, insert newly freed PDU in to head
is better.
The performance drop is within 1% compared to the TCP path without
crc32.
Signed-off-by: MengjinWu <mengjin.wu@intel.com>
Change-Id: I480eb8db25f0e730cb198ca5ec19dbe3b4d38440
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11708
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We have to process whole QoS queue on each QoS poll. It may contain
IOs that still have quota or not affected by QoS rules at all. If we
stop on the first queued IO, all IOs will be limited by the minimum
QoS rule even if they're not affected by this rule.
Here is an example and simple test. We have a NVMf target with Null
bdev and QoS configured with read bandwidth limited to 10 MB/s and
write bandwidth limited to 100 MB/s. First we start nvme_perf with
only write IOs and we see that reported bandwidth is 100 MB/s. Then we
start another instance of nvme_perf with only read IOs. We see that
reported read bandwidth is 10 MB/s but we also see that write
bandwidth also drops to 10 MB/s.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I1edf09d038e65f873deef19ecb0f4bf9725a5ca5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13767
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Original implementation creates pollers and CQs for all discovered
devices at poll group creation. Device (ibv_context) that has no
references, i.e. has no QPs, may be removed from the system and
ibv_context may be closed by rdma_cm. In this case we will have a CQ
that refers to closed ibv_context and it may crash in ibv_poll_cq.
With this patch pollers are created on demand when we create the first
QP for a device. When there are no more QPs on the poller, we destroy
the poller. This also helps to avoid polling CQs that don't have any
QPs attached.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: I46dd2c8b9b2902168dba24e139c904f51bd1b101
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13692
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: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Both PCIE and VFIO-USER can use the same APIs to get IO queue
pair statistic data, so merge them here.
Change-Id: Iadf9ead2bd5abaf11d2ef5d1884acb67369f85bb
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13538
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: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
When a bdev is registered, it is examined by the bdev modules before the
bdev register even is notified.
Examination may be asychronous, e.g. when the bdev module has to perform
I/O on the new bdev.
This causes a race condition where the bdev might be destroyed while
examination is not finished. Then, once all modules have signaled that
examination is done, `bdev_register_finished` makes an invalid access to
the freed bdev pointer.
To fix this, defer the unregistration until the examine is completed by
opening a descriptor on the bdev.
Change-Id: I79a2faa96c1c893fc1cee645fbe31f689b03ea4a
Signed-off-by: Nathan Claudel <nclaudel@kalray.eu>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13630
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
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 a secondary process exit without deleting allocated IO
queue pair, then a new secondary process will do cleanup for
previous allocated queue pair, then segment fault will happen
due to `stat` inside IO queue pair data strucutre can't be
accessed in this cleanup process.
Fix issue #2565.
Change-Id: I01a037642683901941b5268ac20d17b78b6c6350
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13537
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: Aleksey Marchuk <alexeymar@nvidia.com>
This avoids conflict with public vhost_user.h header
file which can cause problems with abidiff.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ia258b4621eda9f6855d46bbf67d8369a053a7116
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13732
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This avoids conflict with public vmd.h header which
can cause problems with abidiff.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I2f00c07226dec273516868f5fa9d7aa384378308
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13731
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Avoids conflict with public tree.h that can cause
problems with abidiff.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I3ccf4c0198f7975d8ebbee57f50c52f9f2e96fc0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13730
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This avoids confusion with the public idxd.h
header file which causes problems with abidiff.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7910c93d9d95b99c82f4dfdba845e6804e1b6568
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13729
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When doing DIF insert and strip, we will reserve extra
buffer in block device layer to save DIF information,
so when attaching one device to Namespace, we will
check the value first so that the reserved buffer
size isn't smaller than metadata size.
Change-Id: Id9272886ce8a7c01271279686730af4e5b24f35a
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12188
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
When `dif_insert_or_strip` is enabled, NVMf library will do
DIF insert and strip automatically, client isn't aware of
it, when `dif_insert_or_strip` is disabled, we will report
Namespace E2E Protection Capabilities to client, but we
don't process PRACT and PRCHK flags in NVMf library, so
here we don't report the capabilities to client and leave
the use of extended LBA buffer to users.
Change-Id: Ic610dc65fef210a7799c6ab693d89138b99e1193
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12165
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: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Some of the options in sock_impl_opts could be different for different
sockets (even if they're using the same impl). However, outside of a
few selected options (recv_buf_size, send_buf_size), there was no
interface to change them.
This change will allow users to change impl_opts on a per-socket basis
when creating a socket. Sockets created through accept() inherit
impl_opts from the listening socket.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I7628ae19def25cef6ffa62aa54bd34e446632579
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13661
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>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Now that spdk_sock has impl_opts, we no longer need to store a copy of
impl_opts.zerocopy_threshold in spdk_sock.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I96377e330351b1afb57811578acfadf05d53f49c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13660
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Also remove ftl_mngt_get_status() because it won't be necessary now.
Signed-off-by: Artur Paszkiewicz <artur.paszkiewicz@intel.com>
Change-Id: I335831cb1c506379e9afeb0bf87f1f873033073d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13668
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
it is impl and used only in nvmf.c source file
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I1236f9ede28c5da313d118ce73e1da64381379c5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13664
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: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
That is possible to get/set registers from any thread,
during regs processing we are polling admin qpair to
get a completion. At the same time, another thread
can also poll admin qpair and that can lead to
undefined behavior.
This patch fixes an issue when bdev_nvme is configured
with io_timeout. If remote target becomes unresponsive
(e.g. due to link down), IO timeout occurs and bdev_nvme
tries to get csts registers in timeout_cb. At the same
time another thread can process adminq, so we may have
2 simultaneous adminq polls. If admin qpair is disconnecting
at that time (RDMA transport) we may destroy resources
twice from different threads.
We don't see a problem with set_regs function but it
won't be redundant to lock mutex in set_regs as well.
Signed-off-by: Alexey Marchuk <alexeymar@nvidia.com>
Change-Id: I7ec3984d25d0249061005533d13b22315b44ddf2
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13687
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>
We cannot count AERs as outstanding IO for purposes
of subsystem pause, because we cannot expect them
to be completed. Previously we would account for this
in nvmf_ctrlr_async_event_request() by decrementing
the counter, but this did not consider cases in the
calling function (nvmf_ctrlr_process_admin_cmd) where
an AER might complete with error before this function,
resulting in the counter getting stuck indefinitely
with a >0 value.
Rather than adding a decrement in all of those
error cases, do a single check at the beginning
of nvmf_ctrlr_process_admin_cmd, and remove the
one from nvmf_ctrlr_async_event_request.
Fixes issue #2215.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ica969f116d80dfba0168369ff2fba9a4a42fc076
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13678
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
This patch defines a new function, spdk_sock_readv_async(), which allows
the user to send a readv request and receive a callback once the
supplied buffer is filled with data from the socket. It works simiarly
to asynchronous writes, but there can only be a single outstanding read
request at a time.
For now, the interface isn't implemented and any calls will return
-ENOTSUP. Subsequent patches will add support for it in the uring
module and as well as emulation in the posix module.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I924e2cdade49ffa18be6390109dc7e65c2728087
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12170
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: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
In NVMe/TCP target, the socket low water mark is set to
sizeof(struct spdk_nvme_tcp_common_pdu_hdr), which is 8 bytes.
In corner test, there might be 4 bytes data packet sent to
NVMe/TCP target, after that, if there is no more data sent to
the same socket, the 4 bytes won't be read by NVMe/TCP target
qpair thread. Because of this, there is a IO request didn't
complete in initiator. Then, if manual call the readv function to
read the 4 bytes for the pdu in target, the io request complete
normally in initiator. It seems like the pdu might be split,
and in the situation, the IO request will not complete until
new IO request reach.
After set low water mark in NVMe/TCP target to 1 byte, just
like iscsi target done, the issue disappear immediately.
Signed-off-by: BinYang0 <bin.yang@jaguarmicro.com>
Change-Id: I59d3d900f0b25632d786ef25ab096eabe43476bd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13633
Reviewed-by: <chuanwei.ji@jaguarmicro.com>
Reviewed-by: Qingmin Liu <qingmin.liu@jaguarmicro.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
It is a nicer API to allow users to use an
md-related API such as spdk_bdev_read_blocks_with_md
passing md_buf as NULL to mean "don't read metadata".
This avoids the need for an if-statement in the users
code to check if the md buffer is NULL before deciding
which API needs to be called.
This basically requires two changes:
1) only check if the metadata is separate for the bdev
if the md_buf != NULL
2) do not fail if the buffer is specified but the
md buffer is not (we only need to fail the case where
the md buffer is specified but the data buffer is not)
Note that spdk_bdev_readv/writev_blocks_ext was already
allowing the metadata buffer to be NULL, but change
those functions too to match the others on how we check
if the data buffer isn't allocated.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I764cf49b9f573fccb19e73876a376fd231cc3580
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13612
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This will allow the code calling this RPC to interpret the error and
check whether the transport already exists (-EEXIST) or some other error
occurred.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I8c4af84763ddba908c59ff881b09834a439186a8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13577
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: Aleksey Marchuk <alexeymar@nvidia.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
In the multi-process case, a process may call `spdk_nvme_ctrlr_free_io_qpair` on
a foreign I/O qpair (i.e. one that this process did not create) when that qpairs
process exits unexpectedly.
The variable `qpair->poll_group` isn't multi-process safe, we can't use it
in `spdk_nvme_ctrlr_free_io_qpair` and related transport poll group APIs.
Change-Id: Ic13a6a2c7d760477be5be5a56a45caa2b5518717
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13573
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@nvidia.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
It will not find the h2c related reqs in the tailq now.
We can get it from tqpair->reqs directly.
Signed-off-by: MengjinWu <mengjin.wu@intel.com>
Change-Id: I25f0900e875b054d7617450477e9719e7a59aa18
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/12861
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-by: Aleksey Marchuk <alexeymar@nvidia.com>
Currently we initialize pending_bytes only in pre-copy state. This is
pointless since we don't generate any migration data at this state, so
if the vfio-user client reads migration data it will be garbage. Even
worse, we don't re-initialize pending_bytes in stop-and-copy state, so
if the vfio-user client reads the entire migration data in pre-copy state
then there will be nothing left to read in the stop-and-copy state,
which is where we actually produce the migration data. This results in
corruption of the controller's state (e.g. queues).
This patch ensures that migration data are available in the
stop-and-copy state, by setting pending_bytes accordingly only in that
state.
Signed-off-by: Thanos Makatos <thanos.makatos@nutanix.com>
Change-Id: I0b215e64cd1f58f254e1079f06402d196f984099
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11718
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>
The read_data, write_data, and data_written migration callbacks assume
that the migration data are accessed in one go. Until this is fixed,
with this patch we ensure we don't ignore unsupported ranges.
Change-Id: I640415858b8c374ffc9e487cd20f5130e0be9305
Signed-off-by: Thanos Makatos <thanos.makatos@nutanix.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11717
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: Ben Walker <benjamin.walker@intel.com>
All of the callers immediately put the req right
after the nvme_rdma_req_complete call, so just move
the put into that function instead.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ic370cf689850924e0c902a6071af8b3a7ed58c0b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13527
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
This follows similar logic in the pcie and tcp
completion paths, including omitting error
messages when aborting aers by adding a print_on_error
parameter to the completion function.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Id558d0af2cdd705dfb60abb842bd567a0949ccce
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/13525
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
