That prevents nvmf target from starting to destroy poll
groups prematurely
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I833f6198ef0e3083fdadf70dd3b62844c905aceb
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7881
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: Jim Harris <james.r.harris@intel.com>
This patch changes the order of IDENTIFY_ACTIVE_NS and CONSTRUCT_NS
controller states. It is required to further improve memory management
for namespaces by allocating memory only for active ones.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: Ie540442b1bd9e897afcbaa4319c139109dd0c515
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6503
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Previous implementation allocated memory just once at the beginning of
active NS list retrieval procedure. It allocated memory for maximum
possible number of active namespaces, i.e. 'cdata.nn'.
This patch changes allocation logic. One page is allocated at the
beginning. If more is needed, reallocation is done with one more
page.
This patch also removes SPDK_MALLOC_DMA flag from allocation since we
don't do RDMA directly into this buffer.
Signed-off-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Change-Id: Iaa80c4d70c54daaf71dcbf755c63a01a1d83b772
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6502
Community-CI: Broadcom CI
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: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Declare single trap from nvmftestinit() instead overriding main trap
depending on the use-case (veth vs phy).
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Change-Id: I23c66fc5728510d06eee9837843f0734b2d03dcf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7850
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Additionally, include the first resolved alias of the module.
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Change-Id: I71c7c15398ea36f3dea3340f5c98568ff6b6a68a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7848
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>
As these tests were limited to SPDK_TEST_BLOCKBDEV the spdk_dd was
not being tested against liburing. Make sure it is the case now.
Also, fail the test if spdk_dd is not linked to liburing in case
where such tests were requested.
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Change-Id: I5704577530048ed4775ea15a6cdacf23380b5072
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7653
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>
Added 'rate_iops' FIO option to the performance scripts.
This will be used during tests where limiting factor
has to be artifically imposed, rather than come from
the target/initiator system or enviroment.
One case for this are scheduler tests, to limit
each connection and balance the threads accordingly.
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I10bb09be9205a8b7ba90b36cd4e0bf922b0ecbc3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7840
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
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>
Use the macros for red black tree provided by Free BSD to manage
timed pollers efficiently.
Allow RB_INSERT() to insert elements with duplicated keys by changing
the compare function to return 1 if two keys are equal.
Check the return code of RB_INSERT() because this is the first use case
for RB tree macros in SPDK. We did the same for RB_REMOVE() by
adding another temporary variable but we remove it from this patch
because it is not so important compared with RB_INSERT().
When a timed poller is inserted, update the cache for the closest (leftmost)
timed poller only if the tree was empty before or the closest (leftmost)
timed poller was actually changed. We do not have to use RB_MIN()
because all duplicated entries are inserted on the right side.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ibe253ca8eecc10116548b5eedbcdba8fb961b88d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7722
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
We already hold thehe next closest timed poller in tmp. Inlining
poller_remove_timer() into thread_poll() makes the cache update
more efficient.
After this patch, poller_remove_timer() is called only in a single case
and the case is compiled only on Linux. So add it inside of a temporary
block is much clearner. However it will be used by spdk_poller_reschedule()
in the end of this patch series. So keep
the current position.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I2e6858223713eed84f5d70b160da6122edae6d03
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7910
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This enable us to optimize the cache update when RB tree is supported.
Call poller_remove_timer() after getting the next element because
as TAILQ_FOREACH_SAFE() and RB_FOREACH_SAFE() do, TAILQ_NEXT() may
not be valid after the current element is removed.
Previously, the patch had called poller_remove_timer() before getting
the next element. However, thanks to the nice testing, this bug was
found.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I18afb4412115dc1696cc568610cbe3dc618c2357
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7909
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This change is a preparation to first dequeue the closest timed poller
always when it is expired. Previously the poller_remove_timer() calls
were not consistent and difficult to follow.
spdk_poller_pause() sets poller to PAUSING even when it in RUNNING
and move it to PAUSED after returning from its context.
If spdk_poller_pause() and spdk_poller_resume() are called while poller
runs, it is moved to WAITING. Hence thread_execute_poller() and
thread_execute_timed_poller() ignore such cases.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I29340613a2ec0c3529d0886f4d81c0a0fdf8745d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7908
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Add a little complex test cases to avoid regression by the following
patches.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1c94edd913f0a582868e355be085a1faf9bd1a94
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7913
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
libvfio-user now uses cmake3 if available, so check for that as well as cmake.
Signed-off-by: John Levon <john.levon@nutanix.com>
Change-Id: Idf7a496d0d3fec21140afeede56337e53ad475fd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7954
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This API was removed previously, so remove remaining
references in map file and unit tests.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iba2f6a5f5ba590d3996dc133c8181083a33d7405
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7963
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
There are logs like below in existing CI tests:
"nvmf_subsystem_create is deprecated, use nvmf_create_subsystem instead."
Change-Id: Id257ab13120a71ad3d0ae7ed22ed6e0a4f005cf0
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7670
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Monica Kenguva <monica.kenguva@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.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 eliminates the thread-local seed variable. But
we're also adding zipf distributions in an upcoming
patch, and we'll want to store that context in the
ns_entry rather than making it thread local.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Icc4a8b7bdbc9cd35525f2d35c9ada8e3ec0ba76c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7791
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>
Community-CI: Mellanox Build Bot
Currently this will only support a global setting
that must be set on the command line. We can make
it per-job later, but will require adding float
support to the conf library.
Tested by running bdevperf with some malloc
bdevs. Performance with low theta values (i.e. 0.2)
are almost identical to random w/o zipf. But
higher zipf values start to show better performance,
because we get more hits on data that is already
in the CPU cache.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I55b1587cdec2919973b488786d361042ca210606
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7790
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>
Community-CI: Mellanox Build Bot
This eliminates the thread-local seed variable. But
we're also adding zipf distributions in an upcoming
patch, and we'll want to store that context in the
job rather than making it thread local.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If8079682e7d3da8f989ee6b880edc8d3fcb4fdd8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7789
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>
Community-CI: Mellanox Build Bot
This app shows how to use the SPDK zipf functionality,
and uses spdk_histogram to show the cumulative
distributions.
Usage:
build/examples/zipf 1.2 100000000 10000
This will generate 10000 random numbers from a range
[0, 100000000) using zipf theta parameter = 1.2.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I36486399b54f57e771cbc5530ae4905afca60b64
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7780
Community-CI: Broadcom CI
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: Ben Walker <benjamin.walker@intel.com>
__builtin_clzll(0) is technically undefined, but
returns 64 on all currently tested architectures,
which is the desired value. So remove the
assert(datapoint != 0) and instead just set clz=64
for that case so that we aren't depending on
undefined behavior.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ibb05f756e07f20a250d24f0c5adecc4dfbc5a056
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7939
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>
Community-CI: Mellanox Build Bot
Some Linux distributions reports kernel panic using the same
command line as we suggested in this document, that's because
number of IO queues is bigger than number of CPUs, so here
just fix it and added a comment on the number of IO queues
parameter.
See issues #1295 and #1737.
Change-Id: Ie1e18d5e83a80523f71d98b8761d13a8d57cc9ab
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7852
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In json_config test, when there are multiple nvme disks, we see errors like:
ERROR: expected event 'bdev_register:Nvme0n1p1' but got 'bdev_register:Nvme1n1' (whole event line: bdev_register:Nvme1n1:1)
The reason is when there are multiple nvme disks in the system, when
start nvmf_tgt, it will register all the disks in bdev, so there would
be multiple bdev_register events for these disks before we run
create_bdev_subsystem_config. So in create_bdev_subsystem_config we need
add these events to expected events list first.
Signed-off-by: Rui Chang <rui.chang@arm.com>
Change-Id: Ice8f5464975e68ae63fec891ef18dbc7590a9ab5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7858
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Michal Berger <michalx.berger@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Currently we allocate buffers perf each SGL descriptor.
That can lead to a problem when we use NVME bdev with
PRP controller and length of the 1st SGL descriptor is
not multiple of block size, i.e. the initiator may send
PRP1 (which is SGL[0]) which end address is page aligned
while start address is not aligned. This is allowed by
the spec. But when we read such a data to a local buffer,
start of the buffer is page aligned when its end is not.
That violates PRP requirements and we can't handle such
request. However if we use contig buffer to write both
PRP1 and PRP2 (SGL[0] and SGL[1]) then we won't meet
this problem.
Some existing unit tests were updated, 1 new was added.
Fixes github issue #1853
Change-Id: Ib2d56112b7b25e235d17bbc6df8dce4dc556e12d
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7259
Community-CI: Broadcom CI
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: Ben Walker <benjamin.walker@intel.com>
The definition is that:
max active zones >= max open zones.
The REAME.md incorrectly writes max active twice.
Fix this typo in the README.md.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I4b7d24d7bd1efcbe1608d3ee04d8900f83857f5e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7912
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: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Implement support for the recently added fio .get_max_open_zones callback.
If our ioengine does not implement this callback, fio will always result
in an error when using --zonemode=zbd, on platforms which does not have a
fio oslib implementation for this callback, e.g. FreeBSD.
On Linux, fio will by default try to parse sysfs, which will of course not
work on SPDK.
Implement this callback so that our ioengine will be able to provide fio
with the proper max open zones limit.
This will ensure that fio will be able to fetch the proper max open zones
limit, regardless of OS.
While our SPDK nvme ioengine did overwrite the max_open_zones option if it
was set to zero, this is a bit of a hack. The new fio callback is the
proper way to inform fio about the max open zones limit, so that fio itself
can have access to the actual device limit.
(Just overwriting the requested max_open_zones option will not allow fio
to know if the requested max_open_zones option exceeds the device limit.)
Remove the SPDK specific hack and update our README.md accordingly.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I532a0fa065b9e215ee6229b9100135e5403f198e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7898
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: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Move get_fio_qpair() out of ifdef and reuse it in spdk_fio_queue(),
since the code is next to identical.
The only difference is that get_fio_qpair() returns NULL if qpair->ns
is not set.
Since we know that get_fio_qpair() returns NULL if qpair>ns is not set,
we can also remove the equivalent check in spdk_fio_queue().
Also use ns pointer instead of qpair->ns pointer in two other places
in spdk_fio_queue().
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: Ie647a814ff0b7a6f2bb81cbc9d7e95537fbe5256
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7897
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Implement support for the recently added fio .get_max_open_zones callback.
If our ioengine does not implement this callback, fio will always result
in an error when using --zonemode=zbd, on platforms which does not have a
fio oslib implementation for this callback, e.g. FreeBSD.
On Linux, fio will by default try to parse sysfs, which will of course not
work on SPDK.
Implement this callback so that our ioengine will be able to provide fio
with the proper max open zones limit.
This will ensure that fio will be able to fetch the proper max open zones
limit, regardless of OS.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: Ia9c281290e11e4204d270ba4090edb73212ce20f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7896
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: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
zipf is a power law probability distribution. When
applied to performance testing of block devices, it
will select blocks over the full range of LBAs, but
will more frequently select lower-numbered LBAs.
The theta parameter governs the distribution - higher
values of theta will concentrate the distribution on
a smaller number of LBAs.
Note that fio supports zipf, so adding it to SPDK
will enable our perf tools (bdevperf, nvme-perf) to
provide similar functionality.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7df129c9d61996a2070188c6cd9f1fde631ac208
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7779
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
We cannot solely rely on the qpair_ctx->count reaching
0, because qpairs that are in process of being
disconnected will immediately invoke the qpair
disconnect cb.
Instead, we need to wait until the poll group
no longer has any qpairs remaining on the subsystem.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I977747d367d14a4bf60f66a1147b3d75679e5179
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7870
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: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: <dongx.yi@intel.com>
When running scheduler/interrupt.sh, the core_freq displayed from
refresh_cores_tab() and show_core() mismatch for lcore 1. We should
get core_freq from cores[] array which has been qsort.
Change-Id: I4eba9b023cfa00700ab82c679d42fff22b85be46
Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7854
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@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>
Copy Free BSD's sys/sys/tree.h to SPDK's include/spdk/tree.h by
changing as little as possible.
A few changes are as follows:
- Remove a few lines located at the head of the file.
- Change the name of ifdef from _SYS_TREE_H_ to SPDK_TREE_H.
- Change the type __uintptr_t to uintptr_t.
- Change the attribute __unused to __attribute__((unused))
- Insert a space to the front and the end of CONSTCOND.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I26488d8fd61f80508cf7d96fbeb6007f12080a01
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7721
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Monica Kenguva <monica.kenguva@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
When we introduce RB tree, getting the closest timed poller is not
O(1) but O(log N). To mitigate such delay, cache the closest timed
poller into thread, and update the cache when its content is changed.
Add unit test cases for this change. They will also clarify the current
behavior of spdk_poller_unregister() and spdk_poller_pause() for
timed pollers.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ibb98a54c261859a3210034038d3953e5c93ef8aa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7720
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Mellanox Build Bot
Current `test -f` never succeeds, because `/dev/cpu/0/msr` is a
character device file.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Iaa28f04866ab24140668ad43351a37f0ccccef65
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7166
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: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Requests that still reside on retry queue should be
submitted to disk before shutdown.
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Change-Id: Id2d020fcaef6443d01cfd8628686e9b0f34a1cfa
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6771
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This patch reduces admount of changes in the next patch,
no functional changes added.
The next patch will add usage of contig IO buffers for
multi SGL payload. To support it we need to pass an
offset to fill_wr_sgl function. Also in the current
version we assume that for 1 iteration we fill 1 IO
buffer, the next patch will change it and we'll need
to swtich to the next IO buffer in special case. That
can't be done easily if we use fill_wr_sge function
Change-Id: Iee8209634637697f700f8fa9fe61ead156b6d622
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7258
Community-CI: Broadcom CI
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The header is small enough that it likely won't ever make sense
to offload the digest computation.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: Ib6baa201a76d769d978f498f5c65985d5ab06ffd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7766
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
