IB Architecture Specification vol.1 rel.13. in ch.10.3.1 "QUEUE PAIR
AND EE CONTEXT STATES" suggests the following destroy procedure for
QPs associated with SRQ:
- Put the QP in the Error State;
- wait for the Affiliated Asynchronous Last WQE Reached Event;
- either:
* drain the CQ by invoking the Poll CQ verb and either wait for CQ
to be empty or the number of Poll CQ operations has exceeded CQ
capacity size; or
* post another WR that completes on the same CQ and wait for this WR
to return as a WC;
- and then invoke a Destroy QP or Reset QP.
Without the drain step it is possible that LAST_WQE_REACHED event is
received and QP is destroyed before the last receive WR completion is
polled from the CQ.
In SPDK there is no risk of resource leakage in this case. So, instead
of draining we can destroy QP and then just ignore receive completions
without QP and post receive WRs back to SRQ.
Fixes#903
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: Ice6d3d5afc205c489f768e3b51c6cda8809bee9a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465747
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This unifies buffer management among transports further and is a
preparation to make buffer allocation asynchronous.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I8c588eeac4081f50fe32605feb7352f72c628d95
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466847
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Most variables related with I/O buffer are in struct spdk_nvmf_request
now. So we can pass nvmf_request instead of nvmf_rdma_request to
nvmf_rdma_request_fill_buffers and do it in this patch.
Additionally, we use the cached pointer to nvmf_request in
spdk_nvmf_rdma_request_fill_iovs which is the caller to
nvmf_rdma_request_fill_buffers in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ia7664e9688bd9fa157504b4f5075f79759d0e489
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466212
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: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Use spdk_nvmf_request_get_buffers() and spdk_nvmf_request_free_buffers(),
and then remove spdk_nvmf_rdma_request_free_buffers() and
nvmf_rdma_request_get_buffers().
Set rdma_req->data_from_pool to false after
spdk_nvmf_request_free_buffers().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ie1fc4c261c3197c8299761655bf3138eebcea3bc
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465875
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This is a prepration to unify buffer management among transports.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I6b1c208207ae3679619239db4e6e9a77b33291d0
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466002
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>
This is a preparation to unify buffer management among transports.
struct spdk_nvmf_request already has SPDK_NVMF_MAX_SGL_ENTRIES (16) * 2
iovecs. Hence incresing the number of buffers twice will be no problem.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Idb525abbf35dc9f4b8547b785b5dfa77d106d8c9
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465873
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>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
One of stop conditions in data WR release function was wrong. This
can cause release of uncompleted data WRs. Release of WRs that are
not yet completed leads to different side-effects, up to data
corruption.
The issue was introduced with send WR batching feature in commit
9d63933b7f.
This patch fixes stop condition and contains some refactoring to
simplify WR release function.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: Ie79f64da345e38038f16a0210bef240f63af325b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466029
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
From rdma_cma.h "Users must destroy any QP associated with an
rdma_cm_id before destroying the ID."
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I5ed0c25221c5401cdde8b31a4e217b9d79e7caaa
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/464290
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We should be checking directly against the base of the iov when doing
memory map translations. The current behavior is to check against the
starting address of the buffer which is a close address, but not exactly
the same.
Change-Id: I7f65224a6836a814708438f2866d84ae22882216
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463893
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: <jiandong.zheng@broadcom.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case of failure during resource allocation within poll_group_create
there is a lack of return statement which could lead to NULL ptr
dereference.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I84abe64a1843117d76b97e62656bdfc4fe2b35d8
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463195
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
NVMf statistics functions use spdk_get_io_channel function to get a
poll group. It increases reference counter in io channel and causes
problems on application exit. spdk_put_io_channel calls were added to
release the channel.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I832d1eae346c3bc3858ed0ed063ff7a7a897a2f5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463389
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>
This patch adds measurement of time request spends from the moment it
was polled till completion.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I1fcda68735f2210c5365dd06f26c10162e4ddf33
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/445291
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>
This patch adds statistics for pending state in NVMf RDMA subsytem
which may help to detect lack of resources and adjust configuration
correctly.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I9560d931c0dfb469659be42e13b8302c52912420
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452300
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: Ben Walker <benjamin.walker@intel.com>
RDMA polling statistics: number of polls and number of completion
entries returned.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: Iabcf2cb6f6a35f595b89b58cdfcd177a637dda13
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/445289
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>
This patch adds transport part to nvmf_get_stats RPC method and basic
infrastructure to report NVMf transport specific statistics.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: Ie83b34f4ed932dd5f6d6e37897cf45228114bd88
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452299
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>
qpair structure is freed and an error code is returned to the caller in the case of failed qpair initialization in function spdk_nvmf_rdma_qpair_initialize (e.g. bad return value of rdma_create_qp).
The return code is handled by nvmf_tgt_poll_group_add function which destroys the qpair for the second time.
This patch fixes#857
Change-Id: I0773652ecccbbd634ad272106e0a93c1e591d7d2
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Evgenii Kochetov <evgeniik@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462011
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Lorne Li <lorneli@163.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Func spdk_nvmf_rdma_destroy_defunct_qpair is a "last chance option"
to destroy qp manually if some driver/hardware doesn't drain qp's
failed wr as expected.
There's a probability that ibv_poll_cq polls wr of the destoryed qp
after spdk_nvmf_rdma_destroy_defunct_qpair's execution. Although in
practice the risk of this situation is minimal(if not non-existent),
add a log here so that we could detect this situation easily.
Change-Id: Ifa9534397513bcea34c18fbb8168eef8f53599c1
Signed-off-by: lorneli <lorneli@163.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462441
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Currently rqpair will be destroyed directly in ibv_poll_cq path
if it has been drained, regardless of whether there are outstanding
I/Os issued to bdev layer. So after outstanding I/Os completing,
spdk_nvmf_rdma_close_qpair will be called from nvmf layer, accessing
a destroyed qp.
This path defers qp destruction in nvmf_rdma_destroy_drained_qpair
func until nvmf layer closes qp.
Fixes 851
Change-Id: I8bcce66f8053ddb105702ac603d5d73af54bdcfc
Signed-off-by: lorneli <lorneli@163.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461237
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
spdk_dma_*malloc() is about to be deprecated.
Change-Id: I5bcac50baca785255eb068086e67c07d120b042f
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459432
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
No need to have it under lock. Additionally in case of failure
there was a lack of rdma_destroy_id(). This is addresed within this
change as well.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Idbb36d51ad4ef7ef81051463f56efc87ef00c966
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462054
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case of failure during pd or map allocation freeing list of devices
was missing.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: If62f7b072f3894fd1a7e856c19b4ea51646dd20e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462079
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In case of pd allocation by nvmf hooks there is a lack of null
check as oposed to pd allocation by ibv_alloc_pd.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Iead6e0332bdee3da4adb6e657af298215c4e2196
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461576
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Vhost testing crashed from Nightly testing, because a member
access within null pointer of type 'struct ibv_send_wr'.
Change-Id: If8f34f23864883ea73516d2d1fe3b30137c04316
Signed-off-by: Hailiang Wang <hailiangx.e.wang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/458913
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Unsetting this flag will decrease the number of WRs retrieved during CQ polling and will decrease
the oeverall processing time. Since RDMA_WRITE operations are always paired with RDMA_SEND (response),
it is possible to track the number of outstanding WRs relying on the completed response WR.
Completed WRs of type RDMA_WR_TYPE_DATA are now always RDMA_READ operations.
The patch shows %2 better peformance for read operations on x86 machine. The performance was measured using perf with the following parameters:
-q 16 -o 4096 -w read -t 300 -c 2
with nvme null device, each measurement was done 4 times
avg IOPS (with patch): 865861.71
avg IOPS (master): 847958.77
avg latency (with patch): 18.46 [us]
avg latency (master): 18.85 [us]
Change-Id: Ifd3329fbd0e45dd5f27213b36b9444308660fc8b
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Evgenii Kochetov <evgeniik@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456469
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This matches the Linux kernel target. Users can
still decrease this default when creating the
transport (i.e. -p option for nvmf_create_transport
in rpc.py).
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Icad59350a2cd35cfc4ad76d06399345191680c05
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454820
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
We could run into issues with this if we were using an arbitrarily large
amount of cores to run SPDK.
Change-Id: Ia7add027d7e6ef1ccb4a69ac328dbdf4f2751fd8
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452250
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This will help us not iterate through the whole list of connections when
only some of them have pending recvs.
Change-Id: I681bc98befbdda4e77ef333b7a086c08b2708eb3
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449266
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 <shuhei.matsumoto.xt@hitachi.com>
This will help save MMIO overhead. Especially in the SRQ case.
Change-Id: I6fb70cf6de4763450f97961f41ccdce3acec2e63
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449265
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>
By creating a list of qpairs, we can avoid looping over every connected
qpair to process sends each time we poll.
Change-Id: If24bbc363176f52fbfb756d56719edd885a21a11
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449264
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>
By batching ibv sends each time we poll, we can reduce the number of
MMIO writes that we do.
Change-Id: Ia5a07b0037365abfa8732629c34d34a9ed49ac70
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449253
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>
There are cases where srq can be a detriment. Add a flag to allow users
to disable srq even if they have a piece of hardware that supports it.
Change-Id: Ia3be8e8c8e8463964e6ff1c02b07afbf4c3cc8f7
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452271
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This restriction helps reduce the amount of padding when
printing out the event trace, allowing it to fit in a
small number of columns.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ifa31e5a6967c7b9bc7028069effb71533f80596f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452736
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This was not used by any of the trace register descriptions.
Let's remove it rather keeping it around if we don't need it.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Idda809e2911db5be555ff6aa13695484a14bf665
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452734
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
We can use the rpoller->srq to check if a qpair is valid when processing
recv completions.
Change-Id: I6aa360adc48a3312ddcf79f10e2a65b502a7314f
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452247
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
Mempools are based off of a ring structure which allocates its elements
as a power of two. It also only exposes n-1 elements to the user. So
when we create a mempool with 2^n elements in it, we have to allocate a
ring with 2^n+1 entries. By decreasing the number of elements in these
key mempools by 1, we can save a decent amount of memory.
Change-Id: I942c9dd4cf59096969bc2559fb46fd2084a07f09
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448875
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
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>
It used to be that we would get async events very infrequently. However,
with the introduction of SRQ, this number has gone up tremendously.
Change the way we report our these events so that we don't spam/confuse
people running the target.
Change-Id: I33070281fa854cbc17784d61bbbb870196ca8780
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452159
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>
If there are outstanding recvs for a qpair when it is destroyed, we need
to clear the qpair from it before reposting it. Otehrwise, we have a
potential heap-use-after-free of double free (depending on whether the
recv completion is in error state or not).
See github issues #730
Change-Id: Ic2009c761cbcc5e89174f62fbd0872d0489c67ca
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452122
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
We were setting this value in the target from our initiator, but it
turns out the rdma_conn_params struct is responsible for setting the
opposite side so we need to add it in the target side when accepting
connections.
Also, add a test to demonstrate target functionality when we overwhelm
the SRQ. It is useful to note that performance really tanks when you
start overwhelming the srq so it may be useful to use this test case to
check performance gains in edge cases over time.
Change-Id: Iac541bd9fc1d82eca9f21e7abc3f625663a6c460
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451678
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 were incrementing over the end of the descriptor list and assigning
undefined values to the rsp opcode in SEND_WITH_INVAL case. We were only
hitting this error when mixing sgl and inline requests in the same
workload. We were just by chance hitting a four bit value that was set
to all 1s from the in capsule data from the last request.
Change-Id: Ied06356f3d22fa34a2cd869dfad6bdca8720791d
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450873
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>
This was not being properly set in the multi-sgl path.
Also add a verification step to the fio configuration file to prevent
against future regressions.
Change-Id: I510b6acd92bc2fbc9b6fbec1d59945cc53584ad3
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450305
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Rqpair qp and resources maybe not be created, if rqpair fail to
initialise. For example, in function new_qpair, the code run to
spdk_nvmf_qpair_disconnect, but rqpair is initialised in
poll_group_add.
Fix#557 segmentaion fault(core dump)
Change-Id: I1892e6d13e2d53dd5a7c4856d775f9b3b85da961
Signed-off-by: JinYu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450986
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Hailiang Wang <hailiangx.e.wang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
If rsp->status.sc != SUCCESS and xfer == DATA_CONTROLLER_TO_HOST,
We would not send the data WR, so clean the num_outstanding_data_wr.
Fix#728
Change-Id: I32259788e495ed76f8f02a9d871bd56356d93dc4
Signed-off-by: JinYu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450726
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Should speed up operations, and allows us to remove the 16 byte link
object from the request structure.
Change-Id: Ie62df1f44d22580a7a7ae41c498295841d1e3064
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448080
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
We were only using one enum from this whole struct, so there is no need
to store it. Plus the queries we use to update it are so infrequent and
only occur during connect and disconnect so I think we can save quite a
bit of space by removing this without compromising performance.
Change-Id: Icf29977a3c10cb289564fa2760a0059f07a0f8cb
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448072
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There was a lot of duplicated code here between states. I'm trying to
minimize the duplicated code without making it confusing.
Change-Id: I13183431e554c8a9f501b3385bbd7b59e2c83161
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448066
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Catch an edge case where a multi sgl request is longer than the allowed
transfer size.
Change-Id: I79779050fe951d16f1240e2c3d8cf5037e576ea2
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/440766
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This is important to avoid thrash when we don't have enough buffers to
satisfy a request.
Change-Id: Id35fd492078b8e628c2118317f674f07e95d4dba
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449109
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
The buffers are really specific to the request and not the wr or data
object. In the case of multiple wr requests, the maximum number of
buffers per req is equal to the number of SGEs in the NVMe-oF request
*2.
Change-Id: Ic59498bfed461d180adb2fb9a481ac5b11fa9252
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449108
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
Enable parsing an nvmf request that contains an inline
nvme_sgl_last_segment_descriptor element. This is the next step
towards NVMe-oF SGL support in the NVMe-oF target.
Change-Id: Ia2f1f7054e0de8a9e2bfe4dabe6af4085e3f12c4
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/428745
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>
Changing i to iovcnt in all references to the req->iov structure will be
important when we start processing multi-sgl requests.
Change-Id: I90a9b6d872b94f846ae7d29a45dd2703eafa6175
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449201
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This will be used by the multi-sgl version of this function as well.
Change-Id: Iafeba4836a77482fa2a158f86f1c17fe7fdeb510
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/449104
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This event is generated by NICs utilizing the SRQ feature when the last
RECV for that qpair is processed. I have confirmed this feature.
Change-Id: Ib6d6b6d02987f789b4d5dd3daf734e3351ee1974
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448063
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 <shuhei.matsumoto.xt@hitachi.com>
With ASAN to run this cases, it will report issue about heap used after free
in spdk_nvmf_rdma_qpair_destroy. Resources have been released before,
change the order to in this tailq to release resources.
ERROR: AddressSanitizer: heap-use-after-free on address
0x6080000080e0 at pc 0x0000006e1e3f bp 0x7fd48b6c3df0 sp 0x7fd48b6c3de0
READ of size 8 at 0x6080000080e0 thread T3 (reactor_1)
0x6e1e3e in spdk_nvmf_rdma_qpair_destroy spdk/lib/nvmf/rdma.c:813
Change-Id: Ia1c12bca84955a2de60399e6b265c9b8901bb51e
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448534
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Not doing so can cause us to hit asserts during the shutdown path. This
should fix an intermittent failure we are seeing on the test pool where
we hit the assert rdma_req->state != RDMA_REQUEST_STATE_FREE in
spdk_nvmf_rdma_request_process.
Note that this problem doesn't cause any data corruption when debug is
not enabled, it just causes us to probcess a subset of commands through
the state machine one extra time suring qpair shutdown.
Change-Id: Ibc36bfea87ec4089b8e2c7a915f48714fddb0b09
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447843
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This will become important later on.
Change-Id: I94e5af03359e476afbc68664e43f44269ad5974c
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448074
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 we have a shared receive queue, the number of outstanding items
associated with a completion queue is deterministic, and limited by how
many RECVs we have total in the SRQ. So, we can set the total size of
the Completion queue at the beginning of time and never resize it.
Change-Id: I787e4c5bbd52ac8948a323d1301f926f887cd91c
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447492
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>
Consolidating error paths is common practice in SPDK so do that here to
make the function more uniform and save space.
Change-Id: I98c5d5f7feeb688f1d8b24f4d2d3461a43d00c1d
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/448191
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This unifies the clean up path between SRQ and normal
operation.
Change-Id: I396d7e3749579f27b5bb1e89b9d6761a77ba5beb
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/446979
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Depending on whether SRQ is enabled, resources may be allocated
to the rqpair or to the rpoller. Create a struct to hold these
pointers that can be used in both locations to avoid duplicated
code.
Change-Id: I2c8fc59009201d9e41721e6462a81732b529a9e0
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/446978
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Eugene Kochetov <evgeniik@mellanox.com>
This wasn't used anywhere.
Change-Id: I405af3c808be284d19218f3f04c1e90e33e31de8
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/446977
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
We were only using one pd per device anywas, and this is necessary for
shared receive queue support.
Change-Id: I86668d5b7256277fe50836863408af2215b5adf9
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447385
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>
Both Mellanox and Soft-RoCE NICs work with this approach.
Change-Id: I7b05e54037761c4d5e58484e1c55934c47ac1ab9
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/446134
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>
Not all RDMA drivers fail back the dummy recv and send operations that
we send to them when destroying a qpair. We still need to free the
resources from these qpairs to avoid eating up all of the system memory
after multiple connect and disconnect events. Since we won't be getting
any more completions, the best heuristic we can use is waiting a long
time and then freeing the resources.
qpair_fini is only called from the proper polling thread so we can safely
call process_pending to flush the qpair before closing it out.
Change-Id: I61e6931d7316d1e78bad26657bb671aa451e29f4
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/443057
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 <shuhei.matsumoto.xt@hitachi.com>
In the error path, we were first decrementing a variable and then
asserting that it must be >0. These operations should occur in the
opposite order.
Change-Id: I6cec544faf17bb75cbfca3d3a3c173dc5db14f99
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/446440
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: yidong0635 <dongx.yi@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
When the decision was made to uncouple the number of shared buffers from
the queue depth and allow the user to decide for themselves, the default
was also significantly lowered, which caused some issues when trying
torun performance tests (See https://github.com/spdk/spdk/issues/699).
While this is a user modifiable variable, it is still best to keep the
higher default value.
The original value was equivalent to max_queue_depth *
SPDK_NVMF_MAX_SGL_ENTRIES * 2 with the defaults for max_queue depth and
max_sgl_entries being 128 and 16 respectively. Hence 4096
fixes: 0b20f2e552
Change-Id: I809e97a10973093a2b485b85bca7160091166f70
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/446525
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>
I think this simplifies the process a little bit.
Change-Id: Icc87a59c9f6fd965ef35531975b7036d85c4bc95
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/445916
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 were only using one value from this array to tell us if the qpair was
idle or not. Remove this array and all of the functions that are no
longer needed after it is removed.
This series is aimed at reverting
fdec444aa8 which has been tied to
performance decreases on master.
Change-Id: Ia3627c1abd15baee8b16d07e436923d222e17ffe
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/445336
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: Ben Walker <benjamin.walker@intel.com>
Since we no longer rely on the state queues for draining qpairs, we can
get rid of most of them. We cn keep just a few, and since we don't ever
remove arbitrary elements, we can use stailqs to perform those
operations. Operations on Stailqs carry about half the overhead as
operations on tailqs
Change-Id: I8f184e6269db853619a3581d387d97a795034798
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/445332
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>
Error logs in nvmf_rdma_dump_request lead to report error about
address points to the zero page, add judgement to return.
this issue occurs in heavy load fio testing.
Change-Id: I50302be88b3af53f718e3800aa16df7c506ca4e8
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441110
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
We should never be going over these limits in the respective transports,
but add asserts to check this during testing.
Change-Id: Ifcaa82ccf58546a38020b31df54ee5d1d9822b8b
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442777
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: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This intermediate state is unused and meaningless. the qpair transitions
into this state right before calling a synchronous operation and then
transitions to active as soon as that operation completes successfully.
If the operation did not complete successfully, we were leaving qpairs
in this weird intermediate state when for all intents and purposes they
had reverted to an uninitialized state. Keeping qpairs in the
uninitialized state until they have been added to a poll group creates a
meaningful distinction between states that can be actionable from the
transport level.
Change-Id: I6de9bc424b393b6fff221aa2f4212aaa91488629
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443471
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>
Connections in the uninitialized state haven't been added to a poll
group yet, so submitting dummy requests to them will be pointless since
they will never be polled. We need to reject the connection and destroy
the qpair immediately.
Change-Id: Id5dd711882e1ae7c13ae32c06da2285186b00a1b
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443470
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>
Since there are multiple events/conditions that can trigger a qpair
disconnection, we need to funnel them to a single point of entry. If
more than one of these events occurs, we can ignore all but the first
since once a disconnect starts, it can't be stopped.
Change-Id: I749c9087a25779fcd5e3fe6685583a610ad983d3
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443305
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
For devices that support fewer SGE elements than our default values, we
need to adjust the I/O unit size so that we don't ever try to submit
more SGLs than we are allowed to.
Change-Id: I316d88459380f28009cc8a3d9357e9c67b08e871
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442776
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This value was not being decremented when we got SEND completions for
write operations because we were using the recv send to indicate when we
had completed all writes associated with the request. I also erroneously
made the assumption that spdk_nvmf_rdma_request_parse_sgl would properly
reset this value to zero for all requests. However, for requests that
return SPDK_NVME_DATA_NONE rom spdk_nvmf_rdma_request_get_xfer, this
funxtion is skipped and the value is never reset. This can cause a
coherency issue on admin queues when we request multiple log files. When
the keep_alive request is resent, it can pick up an old rdma_req which
reports the wrong number of outstanding_wrs and it will permanently
increment the qpairs curr_send_depth.
This change decrements num_outstanding_data_wrs on writes, and also
resets that value when the request is freed to ensure that this problem
doesn't occur again.
Change-Id: I5866af97c946a0a58c30507499b43359fb6d0f64
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443811
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
max_read_depth should be based on max_qp_init_read_atomic, or the
maximum number of read values that the initiator will accept as
outstanding.
The device attributes object contains values for both the initiator
(remote side) and the target (local side). All attributes with the name
init in them are meant to correspond to the initiator. The
qp_read_atomic value represents the number of reads and atomic
operations that can have this device as the target. qp_init_read_atomic
represents how many read operations the initiator has said that we can
have outstanding that have the initiator's rdma device as the target.
Since this number represents how many outstanding reads we will send to
the initiator at once, we should use the qp_init_read_atomic value.
Change-Id: Iacc044e8321080de8accd9128ac3777bbb948afc
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442409
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>
This is a holdover from before poll groups were introduced.
We just need a per-thread context for a set of connections,
so now that a poll group exists we can use that instead.
Change-Id: I1a91abf52dac6e77ea8505741519332548595c57
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442430
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The READ and ATOMIC in the comment above are capitalized, so
make this all caps too.
Change-Id: I49fae2ceb826b22953d9b26d42b95f17e2dac617
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442427
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
When a connection goes to close and has no I/O outstanding,
the current_recv_depth was being decremented beyond 0 and rolling over.
If the poll group then finds a successful receive completion on the next
poll (for a command that arrived prior to starting the disconnect but
hadn't been processed yet), it would trip the max queue depth check
added recently and start another disconnect process. If only one command
arrives in this window, everything actually works out ok.
However, if there are two receive completions sitting in the completion
queue after the disconnect process is started, the first one does the
double disconnect and the second one does another disconnect which ends
up dereferencing a null pointer.
Since there is always a special reserved slot for the dummy recv, don't
do decrements or increments of the current_recv_depth for the dummy
recv. This allows the code to still enforce the actual max_queue_depth
on recvs without underflowing or overflowing the counter.
Change-Id: I56c95b2424e956a3b007b25c50cbf47262245b8f
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442642
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Since we have different requirements for submitting RDMA read and write
operations, we should track them separately so that we don't block
writes when the device does not have enough resources for read
operations.
Change-Id: I5d6424c0e26f2f5362866d1bb21eb46700c245da
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441794
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Before, the number of WRs and the number of RDMA requests were linked by
a constant multiple. This is no longer the case so we need to make sure
that we don't overshoot the limit of WRs for the qpair.
Change-Id: I0eac75e96c25d78d0656e4b22747f15902acdab7
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/439573
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This gives us more realistic control over the number of requests we can
submit.
Change-Id: Ie717912685eaa56905c32d143c7887b636c1a9e9
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441606
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
rw_depth was a misinterpretation of the spec. It is based on the value
of max_qp_rd_atom which only governs the number of read and atomic
operations. However, we were using rw_depth to block both read and write
operations which is an unnecessary restriction. write operations should
only be governed by the number of Work Requests posted to the send
queue. We currently guarantee that we will never overshoot the queue
depth for Work requests since they are embedded in the requests and
limited to a size of max_queue_depth.
Change-Id: Ib945ade4ef9a63420afce5af7e4852932345a460
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441165
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
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>
This will be necessary later on when we need to throttle send and recv
requests in software.
Change-Id: Ifb25eaabd15e101fbfc2959a08a321f80857b280
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441604
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Assigned CQ size when creating CQ may run over due to
heavy workload with too many qpairs. Enlarge it dynamically
can prevent IBV_EVENT_CQ_ERR caused by CQ's runover.
This patch fixes issue #498:
https://github.com/spdk/spdk/issues/498
Change-Id: I6c2d7194d4147d812d49d4fe787fcba5c6bbede9
Signed-off-by: Xiaodong Liu <xiaodong.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/440853
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.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>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
