When tested on Linux 5.8 kernel and configure spdk
with debug mode (--enable-debug), and test SPDK NVMe-oF
tcp transport, and we see the coredump in sock_map_release
with the following statements:
assert(entry->ref > 0);
After debug, I can confirm that the placement_id value got
from the following function (sock->net_impl->get_placement_id)
changes.
It means that: When the sock is added into the poll group
(spdk_sock_group_add_sock), we get the placement_id (named as
Value(begin)); and when the sock is removed from the poll group
(spdk_sock_group_remove_sock), we get the plaemednt_id on
the same sock (named as Vaule(end)). I found that
Value(begin) ! = Value(end).
So our solution is for a socket, we will get placement_id once,
then we can solve this issue.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: Ia1d0cf39247b53410260561aca5af38130cc0abb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3983
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Patch (922d90c8) removed backtrace printing.
Some of API was leftover in header files,
so this patch removes them.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ieb2069e707dee2a196f0cecc283064661a0ae9a9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3953
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The one large global mempool was a waste of memory for apps that
don't use the accel framework as its always allocated a pool sized
to handle a heavy load with multiple threads.
Instead move to a per channel list of just 1024 tasks greatly
decreasing the memory footprint but still able to scale as more
threads are added.
Also renamed all accel_req to acccel_taak and simply task to
accel_task as this was being touched anyways and not consistent.
fixes issue #1510
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I0e93ca6270323e2df4b739711c5d9b667a52e1eb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3740
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>
Because of the Linux kernel has limitation, IORING_SETUP_IOPOLL is only
used for local devices (e.g., local files, pcie NVMe SSDs etc.). However,
it does not work for devices atttached from the remote. So in order to
make bdev uring generic, Let's do not use IORING_SETUP_IOPOLL to create the
uring.
Change-Id: I6aea1ff222a8a0d67ab040ada75aa0ef6730e725
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3587
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Added to the framework as well as all 3 engines. Needed by apps
in the event that they have to fail following the creation of a
batch, allows them to tell the framework to forget about the batch
as they have no intent to send it.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Id94754ab1350e5a969a5fd2306bd59c38f0a0120
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3389
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>
This was sort of a clunky interface requiring a couple of inline
functions in every app that wants to use the accel_fw moving
forward. By having the accel_fw public API accept a callback arg
instead of an accel_task combined with adding a pool of accel_tasks
in the accel_fw engine we can eliminate this.
After changing the parm to a cb_arg, changes were made to all accel_fw
interfaces to put cb_fn and cb_arg as the last parms in public and
private function calls.
Related bdev_malloc changes need to be in this patch in order to pass CI.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I2b75764e534562d91484a094c3352266156d8425
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3209
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>
Also update accel_perf to support it.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ida7b863de02c64239ec4cfbdc3b0235d4e0521f9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2954
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Also as a minor drive-by, added verify capability for fill that
wasn't there before, useful in making sure the prep function
was working and really should have been there anyway.
idxd support for prep fill will follow.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ib54311c1fb98abd2fb61df6603cf3c5300b71161
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2952
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Adds support for both the accel fw and the sw implementation.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ib8748f272a688a074ca475a86dea14179acc020d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2928
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>
1 Change the default factor from 4 to 8, which can be used
to improve the performance.
2 Change the base buffer size in nvme_tcp.c,
we should not use sizeof(struct spdk_nvme_tcp_cmd),
it is 72 bytes. Normally, the initiator will receive
C2h pdus and R2T Pdus by most, so set the size of using
sizeof(struct spdk_nvme_tcp_c2h_data_hdr) is enough.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I384f4cb026cb8d83e75b639f7256ee8cb8ed1df1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3283
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Includes a few related changes:
* slight refactor to have common prep function for batch prep
commands in the sw engine
* added support for dualcast batching in accel_perf example app
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: If9f078a0f8eb509d296e60a94331c92a72f8cb10
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2913
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>
This patch only includes the basic framework for batching and the
ability to batch one type of command, copy. Follow-on patches will
add the ability to batch other commands and include an example of
how to do so via the accel perf tool. SW engine support for batching
will also come in a future patch. Documentation will also be coming.
Batching allows the application to submit a list of independent
descriptors to DSA with one single "batch" descriptor. This is beneficial
when the application is in a position to have several operations ready
at once; batching saves the overhead of submitting each one separately.
The way batching works in SPDK is as follows:
1) The app gets a handle to a new batch with spdk_accel_batch_create()
2) The app uses that handle to prepare a command to be included in the
batch. For copy the command is spdk_accel_batch_prep_copy(). The
app many continue to prep commands for the batch up to the max via
calling spdk_accel_batch_get_max()
3) The app then submits the batch with spdk_accel_batch_submit()
4) The callback provided for each command in the batch will be called as
they complete, the callback provided to the batch submit itself will be
called then the entire batch is done.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I4102e9291fe59a245cedde6888f42a923b6dbafd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2248
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This API is a wrapper for rdma_accept which allows
to remove spdk_rdma_qp_init_attr::initiator_side.
Change-Id: Iba2be5e74e537c498fb11c939c922b2bbda95309
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2908
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Dual-cast copies the same source to two separate destination buffers.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Icadae34a75c35e2db672a193287b147416012a5a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2129
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>
Community-CI: Mellanox Build Bot
Thus, we can make sure that when read data is larger than
the pipe size, it will not read the data into the pipe.
Change-Id: I87f3b03fd9b81eb693e9eae0fea9eef7d1b9eaa8
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2450
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: Aleksey Marchuk <alexeymar@mellanox.com>
spdk_sock_impl_get/set_opts functions allow to set different socket layer
configuration options. Options can be set independently for each
socket layer implementation.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I617e58366a153fae2cf0de1b271cc4f4f19ec451
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/607
Community-CI: Broadcom CI
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>
Along with the sw engine back end implementation for it. IDXD
back end will follow, the use in the accel perf tool as part
of verify.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ia1013cd884ff71990f898320d6a22e96e16ad2a4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2107
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Verbs implementation: doesn't differ from the original:
- send_wr: saves WRs in the qpair internal structure
- flush_wr: calls ibv_post_send
Direct Verbs implementation:
- send_wr: calls ibv_wr_start for the first WR;
Calls opcode-specific function to add a WQE to SQ
Tracks queued WRs to return a correct pointer to 'bad_wr'
- flush_wr: calls ibv_wr_complete - it flushes to the NIC
all WRs added between ibv_wr_start and ibv_wr_complete
Change-Id: I65a26914db688af20589f3b69a994d214d860726
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1659
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The new RDMA provider can be enabled by passing
--with-rdma=mlx5_dv parameter to configure script
This provider uses "externally created qpair"
functionality of rdma cm - it must move a qpair
to RTS state manually
Change-Id: I72484f6edd1f4dad15430e2c8d36b65d1975e8a2
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1658
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This is a wrapper over RDMA CM rdma_disconnect function
The wrapper is needed since in Mellanox Direct Verbs
(aka DV) we must move qpair to error state manually
before calling rdma_disconnect
Change-Id: Ia8623c6989e7679591f2da56bafa7f4262eeebf9
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1975
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
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>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
RDMA provider is used as an abstraction layer and allows
to add vendor specific verbs implementation. This patch adds
an interface to create and destroy qpairs and its implementation
using regular verbs API.
Future patches will add functions to queue/flush Work Requests,
enable usage of RDMA providers API in NVMEoF target/initiator and
add Mellanox Direct Verbs implementation. This abstraction layer
is flexible enough to add other vendor's implementations.
The choice of a specific RDMA provider will be made at
the configuration step, it doesn't add any runtime overhead.
Change-Id: If8eec5bd45c82c71f73a14f6e84121cd02a1d65d
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1655
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Add the CRC function at the framework level and implement the
software engine back end to use ISAL. The patch series will continue
to include an option for accel_perf to test CRC as well as IDXD
implementation.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I4eff3bbcf98c0bc2928a48272a57031c8b96394e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2072
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This file isn't exclusive to the nvme lib. As such, it shouldn't
use the internal SPDK_LOG_NVME flag.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: Ib6d239ad2e45e58fb97a5ea70b01ce72afa938c0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2246
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add both the plumbing in the engine to call module entry
points if they exist as well as the json write config
for idxd (the only module with config info at this time).
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I91376d3fc60227cd79fae17b164722619eafb9e5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2052
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This was added before the usage of having a SW engine and 2 HW
engines was fully thought out. The current rules are:
* if no HW engine specific enable RPC is sent, use SW
* if a HW engine specific enable RPC is sent, use it
* If a 2nd HW engine specific enable RPC is sent, ignore
In this scheme there's no need for an RPC that lets the user
choose which engine to use because they already do so when
they enable an engine.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I006ffb3b417f1e93bb061b29535d157ba66f03b4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2033
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>
reduce library uses unlink, but the unit tests need to
override it in a specific way.
But linking unit tests with LTO requires the wrapper
definitions be in objects/libraries listed *after*
the object/library that refers to it. So we need to
make the unlink wrapper somewhat generic. We do this
by exporting a string and callback function that the
user can set to enable a user-defined function to be
called when unlink() is called with a specific file
name.
Also revert 3ef6d06 as part of this patch, since we
no longer require the workaround that it implemented.
Fixes issue #1357.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I1ee4c424ad31fe7d91d7b524ed47aedd279e5b5c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1948
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This patch also implements the new API for the 3 existing engines.
There was also some minor clean in one file, moving a function to
eliminate multiple forward declarations (there would have been
another one with this new API).
The next patch will use this API in the accel perf tool.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I4ebc9cb3d1c588919235b5080cbeec29189efa21
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2025
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Purpose: This is used to make users can specify
some options on the socket, e.g., the different priority for the socket.
While creating sockets, the priority needs to be set before connect()
and listen system calls, so better to add one parameter in spdk_sock_opts
which can contain options (e.g., priority) in spdk_sock_listen_ext and
spdk_sock_connect_ext functions.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Change-Id: I406238e9da7abd69f937b7072535a19124ed0169
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1874
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
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>
Last and only usage of struct lvol_task was removed in
previous patch.
Since it is no longer used, remove the structure itself.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I3612106de2ab3a9197245de078c73c53beeff443
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1971
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Docs, RPC, unit tests, etc., will follow. Notes:
* The current implementation will only work with VFIO.
* The current implementation supports only the existing accel
framework API. The API will be expanded for DSA exclusive features
in a subsequent patch.
* SW is required to manage flow control, to not over-run the work queues.
This is provided in the accel plug-in module. The upper layers use public
API to manage this.
* As we need to support any number of channels (we can't limit ourselves
to the number of work queues) we need to dynamically size/resize our
per channel descriptor rings based on the number of current channels. This
is done from upper layers via public API into the lib.
* As channels are created, the total number of work queue slots is divided
across the channels evenly. Same thing when they are destroyed, remaining
channels will see the ring sizes increase. This is done from upper layers
via public API into the lib.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ifaa39935107206a2d990cec992854675e5502057
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1722
Community-CI: Mellanox Build Bot
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>
Module, etc., will follow. Notes:
* IDXD is an Intel silicon feature available in future Intel CPUs.
Initial development is being done on a simulator. Once HW is
available and the code fully tested the experimental label will be
lifted. Spec can be found here: https://software.intel.com/en-us/download/intel-data-streaming-accelerator-preliminary-architecture-specification
* The current implementation will only work with VFIO.
* DSA has a number of engines that can be grouped based on application
need such as type of memory being served or QoS. Engines are processing
units and are assigned to groups. Work queues are on device structures
that act as front-end groups for queueing descriptors. Full details on
what is configurable & how will come in later doc patches.
* There is a finite number of work queue slots that are divided amongst
the number of desired work queues in some fashion (ie evenly).
* SW (outside of the idxd lib) is required to manage flow control, to not
over-run the work queues.This is provided in the accel plug-in module.
The upper layers use public API to manage this.
* Work queue submissions are done with a 64 byte atomic instruction
* The design here creates a set of descriptor rings per channel that match
the size of the work queues. Then, an spdk_bit_array is used to make sure
we don't overrun a queue. If there are not slots available, the operation
is put on a linked list to be retried later from the poller.
* As we need to support any number of channels (we can't limit ourselves
to the number of work queues) we need to dynamically size/resize our
per channel descriptor rings based on the number of current channels. This
is done from upper layers via public API into the lib.
* As channels are created, the total number of work queue slots is divided
across the channels evenly. Same thing when they are destroyed, remaining
channels with see the ring sizes increase. This is done from upper layers
via public API into the lib.
* The sim has 64 total work queue entries (WQE) that get dolled out to the
work queues (WQ) evenly.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I899bbeda3cef3db05bea4197b8757e89dddb579d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1809
Community-CI: Mellanox Build Bot
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>
Reviewed-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This allows us to hide the implementations of
g_subsystems and s_subsystems_deps within the
libraries themselves.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: I60716b9465b58d6cdb3a43262a7ded844bd80eda
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1786
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This is part of a larger effort to remove the g_subsystems and
g_subsystems_deps variables from the spdk_event map file. The
implementation of those variables should be internal to the
library.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: I3c2c04933859c6c484a903e666df10f810b26709
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1785
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Since the related feature is already contained in
spdk_sock_listen and spdk_sock_connect functions,
we no longer need this function.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I1eafff0d139fa266a355fbee2bf0fc3947db69fc
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1876
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI
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>
Remove unused io_device_delete_count field
Reorganize the fields so that ones used in the data path are
located in the beggining of the structure and occupy 2 cache lines
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I98e22dacb84153e6a10a6ff6d77c1c9dfb14e02b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1803
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>
Currently run count of poller has been incremented per execution.
It will be helpful for us to know how poller is busy by adding busy
count which is incremented only when some work is done.
spdk_thread_poll() has used the same timestamp in it, and so this is
the maximum we can do for now.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0adfbf9a62c959499978124ecc97d377c96c3769
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1713
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Set 5 seconds timeout to wait until exiting thread is exited into
spdk_thread_poll(). After the timeout, collect error log and then
move the thread to exited forcefully.
Add necessary unit test case accordingly.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ied8f58a2023a3bbe098530810fd3288bef93c3e7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1644
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add enum spdk_thread_state made of RUNNING, EXITING, and EXITED, and
the current state to struct spdk_thread.
The state EXITING is not actually used in this patch yet.
Replace the flag exit simply by the state EXITED.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I6e5dc7184d50ae6d00e6ba00f5e2cf6045e5d48d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1630
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>
Following the idea of thread CPU stats, add reactor CPU stats.
Reactor CPU stats accumulates run time of spdk_thread_poll() calls
to idle TSC or busy TSC according to their return codes.
Add necessary unit tests.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I1a1391e79d74387c68f1651a61c8900e4c6faf66
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1501
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
We have been intermittently hitting the assert where
we check sock->cb_fn != NULL in spdk_sock_group_impl_poll_count.
The only way we could be hitting this specific error is if we
wereremoving a socket from a sock group within after receiving
an event for it.
Specifically, we are seeing this error on the NVMe-oF TCP target
which relies on posix sockets using epoll.
The man page for epoll states the following:
If you use an event cache or store all the file descriptors
returned from epoll_wait(2), then make sure to provide
a way to mark its closure dynamically (i.e., caused by
a previous event's processing). Suppose you receive 100 events
from epoll_wait(2), and in event #47 a condition causes event
#13 to be closed. If you remove the structure and close(2)
the file descriptor for event #13, then your event cache might
still say there are events waiting for that file descriptor
causing confusion.
One solution for this is to call, during the processing
of event 47, epoll_ctl(EPOLL_CTL_DEL) to delete file
descriptor 13 and close(2), then mark its associated data
structure as removed and link it to a cleanup list. If
you find another event for file descriptor 13 in your batch
processing, you will discover the file descriptor had
been previously removed and there will be no confusion.
Since we do store all of the file descriptors returned from
epoll_wait, we need to implement the tracking mentioned above.
fixes issue #1294
Signed-off-by: Seth Howell <seth.howell@intel.com>
Change-Id: Ib592ce19e3f0b691e3a825d02ebb42d7338e3ceb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1589
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: Aleksey Marchuk <alexeymar@mellanox.com>
Collect elapsed time of each SPDK thread and add it to output of
framework_get_reactors RPC.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9e2f0487de81720327428cda5738284a4ce2c557
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1278
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
This patch updates reactor to count thread run time correctly
on multiple SPDK threads per CPU core configuration by using
the refined spdk_thread_poll().
Add tsc_last to struct spdk_reactor to use the end time of the
last thread as the start time of the next thread.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I21042867885d289ff0c23bf2a9ba6a8076a59673
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1256
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
There's no reason not to publish those. Especially if
they're needed in other public headers.
Change-Id: I7dfc6922fcc0dfc46822ad8a16a375f997b98e84
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1041
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This reverts commit ea5ad0b286.
This code is moving from the nvmf target to the posix sock
layer in this series.
Change-Id: I333bdf325848e726ab82a9e6916e1bbdcd34009c
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/446
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 will be used by upcoming spdk_top application.
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I9ffcc3f2e36b8044bbc394938fc7a1dca1dc6892
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1211
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Add thread_count to struct spdk_reactor to count number of threads
per reactor. This number will be used in the next patch to know
if all threads are idle or not for each reactor to support CPU
power saving.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I4f7cc5a6b78d85e9f8d0b539c60058c13e282759
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1169
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
This is a preparation to the next patch which factors out the main
polling loop of _spdk_reactor_run() into a helper function reactor_run().
One of the subsequent patches will support CPU power saving by
adding sleep into reactor_run(). We should not insert sleep between
the main polling loop and getrusage() because now is got before
entering the main polling loop.
To put getrusage() into reactor_run(), we need to maintain last_rusage
in struct spdk_reactor and maintain g_rusage_period as a global
variable. This patch does these changes.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I61bf50de6a170ac73c8fe17e85077b90171dd9c0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1185
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
