tcp_req->orig_length had been set just before I/O submission but
the value is already fixed in spdk_nvmf_tcp_req_parse_sgl().
Hence move setting tcp_req->orig_length accordingly.
This follows the good practice of RDMA transport.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I99f6e266d8f7027bce810864314f3ee24a1af10c
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468910
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
used for creating a new spdk_nvmf_tgt structure in the application.
Change-Id: Ib0182ea6d935b84b4fe4fcad79e173cb46859669
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468387
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Functions added in this patch:
spdk_nvmf_tgt_get_name - get human readable name from target.
spdk_nvmf_get_first_tgt - start iterating over global list of targets
spdk_nvmf_get_next_tgt - get next target in iteration
These functions will facilitate the following RPC
nvmf_get_targets - get the names of all active NVMe-oF targets.
In this series, I will also add two more RPCs, nvmf_create_target, and
nvmf_destroy_target, as wrappers around the create and destroy
functions. Since all of these changes are pretty minor and closely
related, I will just do one big changelog entry at the end.
Change-Id: Ia9f1248fbf9726fa3889998a169211fb25e724f2
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468386
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This issue can be reproduced on fedora30.
Add assert here is enough to fix this kind of warning.
Error log:
rdma.c:3070:20: warning: Access to field 'data_buf_pool' results in a
dereference of a null pointer (loaded from field 'transport')
spdk_mempool_put(group->transport->data_buf_pool, buf);
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
1 warning generated.
This is to fix issue #965.
Change-Id: Ifb742ab914ee9a0381dca0bb769ba8aa564c816f
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468908
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Update transaction length wrt to medata size
Change buffers handling in the case of enabled DIF - add function nvmf_rdma_fill_buffer_with_md_interleave to split SGL into several parts with metadata blocks between them in order to perform RDMA operation with appropriate offsets
Add DIF generation before executing bdev IO operation
Add parsing of DifInsertOrStrip config parameter.
Since there is a limitation on the number of entries in SG list (16), the current approach has a limitation on the max transaction size which depends on the data block size. E.g. if data block size is 512 bytes then the maximum transaction size will be 512 * 16 = 8192 bytes.
In adiition, the size of IO buffer (IOUnitSize conf param) must be aligned to metadata size for better perfromance since metadata is treated as part of this buffer. E.g. if the initiator uses transaction size = 4096, data block size on nvme disk is 512 then IO buffer size should be aligned to (512 + 8) which is 4160. In other case an extra IO buffer will be consumed which will increase the number of entries in SGL and in iov.
Change-Id: I7ad2270fe9dcceb114ece34675eac44e5783a0d5
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/+/465248
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Now code always return 0 , do this like nvme_rdma_mr_map_notify.
That callback can get the right return.
Change-Id: Ief2924e14321b2062f6001e7ae3f50d507206594
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468663
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: Changpeng Liu <changpeng.liu@intel.com>
It is a very rare thing for a buffer to be split over two memory
regions. In fact, it is only possible in dpdk versions where
--match-allocations is not passed as a startup parameter to dpdk but
dynamic memory allocation is enabled.
By adding a small helper function, we avoid failing an I/O because it
was assigned one of these improperly aligned buffers. Also, we try to
remove the buffer from circulation so that it doesn't get picked up
again by another request.
Also, add a unit test to catch this case.
Change-Id: Ia09865c2f77160a960571665b29c4533b11758ae
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467446
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: Paul Luse <paul.e.luse@intel.com>
Just cleaning up a few things like variable names and ordering to make
the whole function more readable.
Change-Id: I1503cdb43ddd73e063d6e57e9ff0cf2a06e79728
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467445
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
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>
The default behavior is to set it to 2MB, so this isn't
required anymore.
Change-Id: I62d7605cd4d5bc41347128f32f9a1aa373a15744
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466993
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.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: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Since we use pdu->data_iovcnt to
build the iov in nvme_tcp_build_iovs, so
send out pdu has the maximal iov number
equals to: 2 + pdu->data_iovcnt,
so we change the comparison.
This makes sure that we can handle all the data
owned by one pdu.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I2b9258cc5716d706c0fa38af609726c439708768
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467207
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: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.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>
I/O buffer cache is per transport_poll_group now. Hence moving
pending_data_buf_queue from struct spdk_nvmf_fc_conn to struct
spdk_nvmf_fc_poll_group is reasonable and do it in this patch.
This change is based on RDMA and TCP transport.
Further unification among transports will be done in subsequent
patches.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic857046be8da238cb3ff9e89b83cdac5f6349bcf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466844
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The pointer to transport is set to struct nvmf_transport_poll_group
in nvmf_transport_poll_group_create() after returning
nvmf_fc_poll_group_create(). Hence use it and remove ftransport pointer
from struct nvmf_fc_poll_group.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9f2b2ade77afa18d0e97949fc0c2403eb000cdad
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467060
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
RDMA transport have used rtransport and TCP transport have used
ttransport, respectively. So FC transport changes to use ftransport
instead of fc_transport.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I7d98eb2f6efbae7e2b4784f31b9de5e1a81bc2ac
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467059
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Both RDMA and TCP transport have uesd group for such case. Hence
FC transport changes to use group instead of tp_poll_group.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic4b401179da506bb204c3ec48650db87f91fe72a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466843
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The pointer to nvmf_poll_group is set in nvmf_transport_poll_group_create()
after returning nvmf_fc_poll_group_create(). Hence holding it into
struct spdk_nvmf_fc_poll_group is duplicated and can be removed.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I7087c5cdb94b0b0c5f51b0b63b631c08266c90d0
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466842
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
RDMA transport have used rgroup and TCP transport have used tgroup
for such case. Hence FC transport changes to use fgroup instead of
fc_poll_group.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I91b7ad6a1c6e45caf92801b0635b18d48b3c9810
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466841
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: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Keeping a global discovery log page was meant to be a time saving
mechanism, but in the current implementation, it doesn't work properly,
and can cause undesirable behavior and potential crashes. There are two
main problems with keeping a global log page.
1. Admin qpairs can be assigned to any SPDK thread. This means that when
multiple initiators connect to the host and request the discovery log,
they can both be running through the spdk_nvmf_ctrlr_get_log_page
function at the same time. In the event that the discovery generation
counter is incremented while these accesses are occurring, it can cause
one or both of the threads to update the log at the same time. This
results in both logs trying to free the old log page (double free) and
set their log as the new one (possible memory leak).
2. The second problem is that each host is supposed to get a unique
discovery log based on the subsystems to which they have access.
Currently the code relies on whether the discovery log page offset in
the request is equal to 0 to determine if it should load a new discovery
log page or use the cached one. This is inherently faulty because it
relies on initiator provided value to determine what information to
provide from the log page. An initiator could easily send a discovery
request with an offset greater than 0 on purpose to procure most of a
log page provided to another host.
Overall, I think it's safest to not cache the log page at all anymore
and rely on a thread local fresh log page each time.
Reported-by: Curt Bruns <curt.e.bruns@intel.com>
Change-Id: Ib048e26f139927d888fed7019e0deec346359582
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466839
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>
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>
Most variables related with I/O buffer are in struct spdk_nvmf_request
now. So we can pass nvmf_request instead of nvmf_tcp_req to
nvmf_tcp_req_fill_buffers and do it in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I00eff578a98891e99fcb9a3aafa3d99126d6f1c1
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466089
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>
Most variables related with I/O buffer are in struct spdk_nvmf_request
now. So we can pass nvmf_request instead of nvmf_fc_request to
nvmf_fc_request_fill_buffers and do it in this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ibe87e7641e5c364b20a6d877ce7928c612b0b83a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466088
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: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This is a small performance optimization and an effor to unify I/O
buffer management further among transports.
it is ensured that the request is the first of STAILQ when
nvmf_fc_request_execute() completes successfully.
Hence change TAILQ_REMOVE to STAILQ_REMOVE_HEAD for the case.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: If982842bf53ba00426a854a18eaadf8a1b8d642d
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466676
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This is an effort to unify I/O buffer management further among
transports. RDMA and TCP transport have named pending_queue
pending_data_buf_queue. So FC transport follows RDMA and TCP transport.
The next patch will change pending_data_buf_queue to use STAILQ
instead of TAILQ.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I57c3c678a1e92ec262eb8940418529a62b6768c3
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466675
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: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This is a small performance optimization and an effort to unify
I/O buffer management further among transports.
It is ensured that the request is the first of STAILQ when
spdk_nvmf_tcp_send_c2h_data() is called or the case
TCP_REQUEST_STATE_NEED_BUFFER is executed in spdk_nvmf_tcp_req_process().
Hence change TAILQ_REMOVE to STAILQ_REMOVE_HEAD for these two cases.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0b195874ac22a8d5ecfb283a9865d2615b7d5912
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466637
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In this patch, we directly point the hdr_p
to the memory owned by the pdu_recv_buf to avoid
memory copy.
Change-Id: Iee0dd98058928f429bf7ad22103cd4826226400f
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465158
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
RDMA transport executes spdk_nvmf_rdma_request_parse_sgl() only if
the request is the first of the pending requests in the case
RDMA_REQUEST_STATE_NEED_BUFFER in the state machine
spdk_nvmf_rdma_requests_process().
This made RDMA transport possible to use STAILQ for pending requests
because STAILQ_REMOVE parses from head and is slow when the target is in
the middle of STAILQ.
On the other hand, TCP transport executes spdk_nvmf_tcp_req_parse_sgl()
even if the request is in the middle of the pending request in the case
TCP_REQUEST_STATE_NEED_BUFFER in the state machine
spdk_nvmf_tcp_req_process() if the request has in-capsule data.
Hence TCP transport have used TAILQ for pending requests.
This patch removes the condition if the request has in-capsule data
from the case TCP_REQUEST_STATE_NEED_BUFFER.
The purpose of this patch is to unify I/O buffer management further.
Performance degradation was not observed even after this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Idc97fe20f7013ca66fd58587773edb81ef7cbbfc
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466636
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>
Use spdk_nvmf_request_get_buffers() and spdk_nvmf_request_free_buffers(),
and then remove nvmf_fc_request_free_buffers() and nvmf_fc_request_get_buffers().
Set fc_req->data_from_pool to false after spdk_nvmf_request_free_buffers().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I046a642156411da3935bc2fa2c2816fc2e025147
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465877
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.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>
Use spdk_nvmf_request_get_buffers() and spdk_nvmf_request_free_buffers(),
and then remove spdk_nvmf_tcp_request_free_buffers() and
spdk_nvmf_tcp_request_get_buffers().
Set tcp_req->data_from_pool to false after spdk_nvmf_request_free_buffers().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I286b48149530c93784a4865b7215b5a33a4dd3c3
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465876
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>
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 patch adds new APIs spdk_nvmf_request_get_buffers() and
spdk_nvmf_request_free_buffers() to be used among transports.
Subsequent patches will replace transport specific APIs by them.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ib153e2c5806b7276915a0aa91179fe9dbcb2a1f0
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465874
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.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>
Purpose: Reduce the recv/readv system call.
Method: Use a big recv buffer to conduct the read.
Though it will introduce addtional buffer copy,
we hope that the overhead introduced by buffer copy will
be smaller compared with frequent recv/readv system call overhead.
And the design is to make a trade off between them.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I9286fd9cec0b512cea8e3f2c335c5bf862b98573
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/464842
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Purpose: Prepare the further optimnization in the
target side whening receving pdu headers, we expect
to use zero copy.
Change-Id: Iae7f9106844736d7160d39d0af1f5941084422ec
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465380
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: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This follows the practice of RDMA transport and is a preparation to
unify buffer allocation among transports.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ib85625f2a0eca01ef4028685dd838d6c41faad7b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465872
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This follows the practice of RDMA transport and a preparation to
unify buffer management among transports.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I4e9b81b2bec813935064a6d49109b6a0365cb950
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465871
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@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 preparation to the next patch to use spdk_mempool_get_bulk.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I28a5ad941004f139c9032d85c2ef92680081f1ce
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465870
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This follows the practice of RDMA transport and is a preparation to
unify buffer allocation among transports.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I3cd4377ae31e47bbde697837be2d9bc1b1b582f1
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465869
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
FC transport can use buffer cache as same as RDMA and TCP transport
now. The next patch will factor out getting buffers and filling
buffers to iovs in nvmf_fc_request_alloc_buffers().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0d7b4552f6ba053ba8fb5b3ca8fe7657b86f9984
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465868
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is a preparation to the next patch to use buffer cache in
FC transport.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I116b064ea0b0a437f9a3293a6f3d46a0e5fc8ecf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465867
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This follows the practice of RDMA transport and a preparation to
unify buffer management among transport types.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic7dc8e6b826baf7f471d192630e8a048a35056ac
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465866
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
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>
NVMe-oF FC transport have used its own buffer pool and have not used
common buffer pool yet.
It looks that there is no particular reason to prevent FC transport
from using the common buffer pool.
This patch removes FC transport specific buffer pool and changes
FC transport to use common buffer pool instead. Add transport
as a parameter to nvmf_fc_request_free_buffers() because similar
APIs of RDMA and TCP transport do that.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Iae3a117466c21eaddbe78a8e8023d80ef37bb3e9
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465865
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
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>
