This is useful for applications even if they elect not to use the SPDK
event framework.
This doesn't shift everything in one go - just the subsystem
initialization logic. Configuration file loading also needs to move
in a separate patch later.
Change-Id: Id419df1045442d416650ed90e5ee78adfdd623d7
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6641
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
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>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Checked the definition of strncmp. If substring s1
is found in strnmp(s1, s2, len), then it will return 0.
For the len value, it is better to use strcmp. Otherwise,
if s1=cluster1, s2=cluster & len=strlen(s2),
strncmp will return 0. But they are two different strings. For
cluster names, they are different.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I15a06184d834cd1567b329d0322cd6bdea6fee4b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7991
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: <dongx.yi@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Mellanox Build Bot
bdev_nvme_compare_ns() will be used to check if all namespaces of
one nvme_bdev are identical, and this is the convenient location.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I3fa6072c1cceec53268e53bf398fa1e8f069035e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7169
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This will be helpful to simplify the following patches.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I66939f2953c66582bfcb79cfe187814280e89680
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7324
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: Aleksey Marchuk <alexeymar@mellanox.com>
When multipath is supported, a subsystem has multiple controllers and
bdev_nvme_create_cb() will create a channel per nvme_bdev_ctrlr by
iterating the list of nvme_bdev_ctrlrs, and will hold lock while
doing it.
If the code to access nvme_bdev_ctrlr is put in a place, the
following patches will be easier and smaller.
Hence reorder the code of bdev_nvme_create_cb() as a preparation.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I2f2e66758c3374c678cc44bbb0116f4611c6753a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7754
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Factor out spdk_bdev_io_complete() calls into a helper function
bdev_nvme_io_complete().
This simplifies the code a little and will be helpful for the following
patches to retry I/Os. These are not performance critical but we
specify inline explicitly by following bdev_nvme_io_complete_nvme_status().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9fafacfd8571c037c3bc34382c251317309da334
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7497
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Mellanox Build Bot
Factor out spdk_bdev_io_complete_nvme_status() calls into a helper
function bdev_nvme_io_complete_nvme_status().
This simplifies the code a little and will be helpful for the following
patches to retry I/Os. Specify inline explicitly to avoid performance
regression.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1ac451486e1c6a4401842490411e986fac191d59
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7484
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
bdev_nvme_submit_request() calls spdk_bdev_io_complete() with failed
if bdev_nvme_reset() returns negated rc other than -ENOMEM.
So let bdev_nvme_submit_request() process it.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I2569634ff0f18fb433cb685de1366e43abf5a9fe
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7524
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is called only in a place. Inlining this into bdev_nvme_submit_request()
will simplify the following patches by removing unnecessary cast.
Besides, use -ENXIO if I/O path is not found. This will be better than -1.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ib6c38f89db1c1e651941aad18d31dd0891f380de
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7871
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
I/Os will be retried if spdk_bdev_io_complete() is called with
SPDK_BDEV_IO_STATUS_FAILED. To do it easier, consolidate exit paths
of bdev_nvme_get_buf_cb().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0a67b88a107d616c5a5b0fc5ff963ad1402f5651
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7487
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Move from a single flag indicating that the socket is on the
pending_events list to two flags - pipe_has_data and socket_has_data. If
either flag is true, the socket is on the socks_with_data list.
This is necessary to track enough state to avoid doing extra recv()
system calls.
Change-Id: I65e5701dccb0a5bade19f266f164f26706b110d4
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7595
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: Aleksey Marchuk <alexeymar@mellanox.com>
Revise bdev_rbd_create rpc call to add an optional
parameter "--cluster-name", e.g., "--cluster-name Rados".
Then users can create a rbd bdev with registered
Rados Cluster. This shared strategy can be used to
remove the thread creation overhead if multiple rbds
are connected to the same Ceph cluster.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: Ide5800f8fc6b2074805272a59731c666fe279b9a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7584
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
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>
This patch is used to add two rpc calls:
bdev_rbd_register_cluster
bdev_rbd_unregister_cluster
Then in the next patch, rbd bdev constructed on the same cluster object
can share the common Rados_t structure in order to remove the thread creation
overhead and improve the scalability.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I898cc4ffabb8e6721ba5bef099cbf948c64d2c98
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7551
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
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>
The bdev layer can do the unmap split now based on the backend
device. For now we only use 1 unmap descriptor, the bdev layer
can help us to do the split.
Fix issue #1888
Change-Id: Iaf740bafd4f2bb4b108133fee2aafd2f53da9b2b
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7519
Community-CI: Broadcom CI
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Also free scsi_task data structure for the asynchronous
libiscsi APIs.
Change-Id: I0bff706bfb795e51a4b10c357913ae66493dca5d
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7513
Community-CI: Broadcom CI
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
If no ioat devices are registered, we should return.
Change-Id: I03435946716ef653b230515da32e8ccbdf5a188a
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7834
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The coredump info can be viewed in
https://github.com/spdk/spdk/issues/1935
We face this issue because no idxd device is attached, but we
still register the hw engine by spdk_accel_hw_engine_register
in accel_engine_idxd_init.
Fixes#1935
Change-Id: I537f06e2b2923faac7f2cd6a28903e77f1f6aaa5
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7832
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This makes use of newly added spdk_bdev_wait_for_examine(),
to only respond to RPC when bdev was fully examined.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: If82cd913ab6653e8cc0da38c639b384b6c0303ba
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5482
Community-CI: Broadcom CI
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: Aleksey Marchuk <alexeymar@mellanox.com>
This makes use of newly added spdk_bdev_wait_for_examine(),
to only respond to RPC when bdev was fully examined.
Fixes#1760
Issue above was triggered in DD tests where application
finished before the examine had a chance to fully finish.
This patch addresses it by making sure that nvme attach
RPC waits for completion of the examine.
Later patch in series adds the bdev_wait_for_examine RPC
to multiple static configuration files. Making sure similar
issues do no occur for bdev modules which do not have changes
in their RPC as here.
The issue does not occur for JSON configs generated from apps,
see patch:
(e57bb1af)lib/bdev: build bdev_wait_for_examine into subsystem
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ie3ca2933af97a40ae01ecc3eefe2161d2d34c602
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5483
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This patch is design to use the single Rados Cluster
object in the same RBD if there are multiple I/O channels created.
And this patch will be prepared for the next patch to share
the same cluster among different RBD bdevs.
Change-Id: I1509f29a9c1088da308a3f88980f0c7fed26476f
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7601
Community-CI: Broadcom CI
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This module only requires the spdk_bdev_module
APIs, so use this #define to avoid having to
bump so version when spdk_bdev is modified.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I292c38e60cbf21d0d5a6583cb2a2b097c524f3d1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7811
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Michal Berger <michalx.berger@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
- Use consistent cache line size units in KiB across RPC calls
and config files. The KiB units are much easier to use then
the bytes units and are more human readable.
- Properly handle cache start when cache line size is incorrect.
- Add test to check if cache line size value is reported correctly.
- Add cache line size info to JSON RPC documentation.
Fixes#1858
Signed-off-by: Rafal Stefanowski <rafal.stefanowski@intel.com>
Change-Id: Iec9ede85f6884b64605d2d112947b3f175cbd938
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7614
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Recently, checked the spdk_io_device_unregister function,
it will have deferred free behaviour, and the io_device
will possibly be freed in put_io_channel function.
And this means that it is not safe to directly call:
spdk_io_device_unregister (io_device, NULL);
Then free io_device relately resource.
Because there will be channel to use the resources associated
with io_device. Then we will possibly cause a NULL pointer access.
I found this issue in bdev rbd module, and I think that the
same issue could happen in other modules. So it is better to put the
resource free function as the call back function.
Change-Id: Icc1f86d72b672faefb3b7f416030b818a8cf45ce
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7646
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Passing cpu_mask hints that match only single core were
usefull to prevent any accidents when doing round-robin
in case of 'static' scheduler.
In practice this is not required in case of 'static' scheduler,
the threads will be spread out over all reactors anyway.
This hinders other schedulers which try to respect the cpu_mask
hints, as they would not move the thread to any reactor.
Preventing bunching up less used threads on single reactor.
Drawback of this patch is that poll group names will not match
the cores they are on.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I5fb308362dd045228ea9fcca24f988388854c054
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7028
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: <dongx.yi@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Purpose: We will also support the kernel idxd driver, so we do not
need export this feature in the module file.
Change-Id: I965e031497920f527962ba187bccd81de6977b8f
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7336
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
There's no good reason to reduce the capacity by aligning it to the
number of optimal open zones. If such alignment is required by the
users of the zone block bdev, it should be done on their own layer.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ic8614a82715e9f064619aa8fdb75d1a0b851490c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7656
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Niklas Cassel <niklas.cassel@wdc.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
We cannot rely on DSM/DEALLOCATE as a write zeroes
alternative, even if DLFEAT reports that deallocated
blocks will be read as all zeroes. DEALLOCATE is
advisory, meaning that blocks may not actually be
deallocated. In cases where they are not deallocated,
they will not be read back later as zeroes.
QEMU 6.0 started reporting DLFEAT as returning zeroes
for deallocated blocks but for some of our write
zeroes tests, blocks aren't actually deallocated.
We may be able to add quirks in the future if we know
that a controller reliably deallocates blocks, but
for now we need to revert this completely.
Note that since bdev/nvme module now does not support
write zeroes in any cases, we need to disable the
write zeroes call in the unit tests.
Fixes issue #1932.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I79f0673774b621a9ffcc46891728cc7719e34cdb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7723
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
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>
(Note: this patch was previously applied as b32cfc46 and then reverted
as 63642bef.)
Today the in-guest nvme device shows physical_block_size=512 even though
the backend iSCSI bdev supports physical_block_size=4K
iSCSI targets exposes physical block size using
logical_block_per_physical_block_exponent in READ_CAPACITY_16
NPWG is one of the way to let Linux nvme driver set
physical_block_size of the nvme block device.
This patch adds spdk_bdev.phys_blocklen which is updated if the iSCSI
backend exposes physical_block_size.
Later phys_blocklen is used in nvmf to set NPWG and NAWUPF to report
back during NS identity.
Linux driver uses min(nawupf, npwg) to set physical_block_size.
Similarly in scsi_bdev fill lbppbe in READ_CAP16 response
based on spdk_bdev.phys_blocklen.
Fixes#1884
Signed-off-by: Swapnil Ingle <swapnil.ingle@nutanix.com>
Change-Id: I0b6c81f1937e346d448f49c927eda8c79d2d75c0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7739
Community-CI: Broadcom CI
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>
This reverts commit b32cfc467b.
This commit fails the ABI checks and only got through because the checks
were disabled until 21.04 hit.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: Id26b8f8ba551193d99b1ccbd31b35378b4095a20
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7731
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Today the in-guest nvme device shows physical_block_size=512 even though
the backend iSCSI bdev supports physical_block_size=4K
iSCSI targets exposes physical block size using
logical_block_per_physical_block_exponent in READ_CAPACITY_16
NPWG is one of the way to let Linux nvme driver set
physical_block_size of the nvme block device.
This patch adds spdk_bdev.phys_blocklen which is updated if the iSCSI
backend exposes physical_block_size.
Later phys_blocklen is used in nvmf to set NPWG and NAWUPF to report
back during NS identity.
Linux driver uses min(nawupf, npwg) to set physical_block_size.
Similarly in scsi_bdev fill lbppbe in READ_CAP16 response
based on spdk_bdev.phys_blocklen.
Fixes#1884
Signed-off-by: Swapnil Ingle <swapnil.ingle@nutanix.com>
Change-Id: I0b6c81f1937e346d448f49c927eda8c79d2d75cf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7310
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When zcero copy send is enabled and used by initiator,
it could significantly increase latency in some payloads.
To enable more fine graing configuration of zero copy
send feature, add new parameters enable_zerocopy_send_server
and enable_zerocopy_send_client to spdk_sock_impl_opts to
enable/disable zcopy for specific type of sockets.
Exisiting enable_zerocopy_send parameter affects all types
of sockets.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I111c75608f8826980a56e210c076ab8ff16ddbdc
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7457
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
On host side the connections are created and then added to thread's
poll group. Those connections could use different NIC queues underneath.
To route all connections of poll group through single queue a unique
placement id is chosen as group_placement_id and each socket of poll
group is marked with group_placment_id using getsockopt(SO_MARK) option.
The driver could use so_mark value of skb to determine the queue to use.
Change-Id: I06bda777fe07a62133b80b2491fa7772150b3b5d
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6160
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We do not want to do any further work on adding
the sock to the group if the epoll_ctl (or kevent)
fails.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I44b6dc86ce5676aa1b8d6c50b86f22758e4e37fa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7594
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
suppression
When all of the following conditions are met:
- non-blocking socket
- zero copy is enabled
- interrupts are suppressed (i.e. busy polling)
- NIC tx queue is full at the time sendmsg() is called
- epoll_wait sees there is already an EPOLLIN event
then we can get into a situation where data we've sent is queued
up in the kernel network stack, but interrupts have been suppressed
because other traffic is flowing. This makes the kernel miss the
signal to flush the software tx queue. If there wasn't also already
a pending EPOLLIN event, then epoll_wait would have been sufficient
to kick the system out of this state. But when all of this aligns,
it hangs.
We deal with this by detecting the scenario and calling poll(), which
will force the kernel to issue the pending transmits.
Change-Id: Ifb247159b7de16c8fc72a90f0333f5b421c8bd07
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6750
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Community-CI: Mellanox Build Bot
Only passing rbd pointer is enough.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I652581879c9ba228f9af0aa3f5353915a286c6c8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7548
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This patch is used to fix the chained crc32c computing when users pass
a vector. Since we use a union in spdk_accel_task structure to differentiate
the usage on "src" and "the vector info" (iovs and iovcnt). So we cannot
directly write the src field while users pass a vector.
And I verified it in the hardware platform.
Change-Id: I85d6e86fa689b261782f80a2f89d908a5d4db84f
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7471
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The original code does not free the I/O channel
in bdev_rbd_resize.
Change-Id: I8e48616bea37ba7cece5a179c0f1ca03b3f5cb53
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7466
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@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>
Add the missing rados_shutdown call in
bdev_rbd_init function.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I0222d2453d1547cfd735f15237eca124cbc23479
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7430
Community-CI: Broadcom CI
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: <dongx.yi@intel.com>
Busy pollng using recv() is dependent on kernel socket buffer being
empty. Instead poll() function busy polls hw queues with no such dependency.
Signed-off-by: Sudheer Mogilappagari <sudheer.mogilappagari@intel.com>
Change-Id: I1cb101848d51f7778cdf3d4c015d2d03201bdb37
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7014
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>
Since the maps are unique to modules, they can store the group_impls
directly.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I7f11db558e38e940267fdf6eaacbe515334391c2
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7222
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This allows for different policies per module, as well as overlapped
placement_id values.
Change-Id: I0a9c83e68d22733d81f005eb054a4c5f236f88d9
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7221
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
There was a fix for this that went into the posix layer, but the
underlying problem is the logic in the nvme/tcp transport. Attempt to
fix that instead.
Change-Id: I04dd850bb201641d441c8c1f88c7bb8ba1d09e58
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6751
Community-CI: Broadcom CI
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: Aleksey Marchuk <alexeymar@mellanox.com>
This list will hold any socket that has some event pending and needs to
be part of the set returned during polling of the group.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I5acf01677e59c1026f93671c7b7b3dc458075bf7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6748
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Mellanox Build Bot
Instead of iterating the list, we can just manipulate the list
in a single step.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I0172cdbce9af35a62d62dbccfac573e5d723f43a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6747
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Mellanox Build Bot
Change-Id: I4ed583d91ae9e820be1ee6f4553f29d6650c4922
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5791
Community-CI: Broadcom CI
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: Jim Harris <james.r.harris@intel.com>
It can divide to two parts:
1, UIO driver - sigbus error handling and uevent
process.
2, VFIO - request notify handling.
sigbus error process is in previous patch.
Change-Id: Idc09754b83ae9ddcaea1f2afcbc13e528ead9863
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5768
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
