When the decision was made to uncouple the number of shared buffers from
the queue depth and allow the user to decide for themselves, the default
was also significantly lowered, which caused some issues when trying
torun performance tests (See https://github.com/spdk/spdk/issues/699).
While this is a user modifiable variable, it is still best to keep the
higher default value.
The original value was equivalent to max_queue_depth *
SPDK_NVMF_MAX_SGL_ENTRIES * 2 with the defaults for max_queue depth and
max_sgl_entries being 128 and 16 respectively. Hence 4096
fixes: 0b20f2e552
Change-Id: I809e97a10973093a2b485b85bca7160091166f70
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/446525 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447465
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>
For devices that support fewer SGE elements than our default values, we
need to adjust the I/O unit size so that we don't ever try to submit
more SGLs than we are allowed to.
Change-Id: I316d88459380f28009cc8a3d9357e9c67b08e871
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442776 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447464
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>
In the error path, we were first decrementing a variable and then
asserting that it must be >0. These operations should occur in the
opposite order.
Change-Id: I6cec544faf17bb75cbfca3d3a3c173dc5db14f99
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/446440 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447463
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 value was not being decremented when we got SEND completions for
write operations because we were using the recv send to indicate when we
had completed all writes associated with the request. I also erroneously
made the assumption that spdk_nvmf_rdma_request_parse_sgl would properly
reset this value to zero for all requests. However, for requests that
return SPDK_NVME_DATA_NONE rom spdk_nvmf_rdma_request_get_xfer, this
funxtion is skipped and the value is never reset. This can cause a
coherency issue on admin queues when we request multiple log files. When
the keep_alive request is resent, it can pick up an old rdma_req which
reports the wrong number of outstanding_wrs and it will permanently
increment the qpairs curr_send_depth.
This change decrements num_outstanding_data_wrs on writes, and also
resets that value when the request is freed to ensure that this problem
doesn't occur again.
Change-Id: I5866af97c946a0a58c30507499b43359fb6d0f64
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443811 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447462
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>
perf application can't generate IO for NVMe namespace with
more than 4G size.
Example of error:
"Attached to NVMe over Fabrics controller at 1.1.75.1:1023:
nqn.2016-06.io.spdk.r-dcs75:rd0
WARNING: controller SPDK bdev Controller (SPDK000DEADBEAF00 ) ns 1 has
invalid ns size 0 / block size 4096 for I/O size 4096
WARNING: Some requested NVMe devices were skipped
No valid NVMe controllers or AIO devices found"
ns_size variable is uint32_t, spdk_nvme_ns_get_size function
returns uint64_t. Result can exceed the maximum size of
uint32_t and ns_size remains 0.
The issue introduced by commit: f2462909
Change-Id: Idc6dd8688d5d6268bda1a1d6b06a611643af6155
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/443996 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447451
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>
On VM these tests takes ages.
Change-Id: Id4799e2d226e59b430e899983a6470080b5c37dc
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443795 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447149
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: Jim Harris <james.r.harris@intel.com>
iter_pci_dev_id abd iter_pci_dev_id functions should
not return BDF for devices that are not ment to be used
in tests.
Note that not all tests are ready for this change as they
discover functions on its own. Lets this changed in
separate patch.
Change-Id: I45a59ec121aa81e9f981acae7ec0379ff68e520a
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/443767 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447148
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: Jim Harris <james.r.harris@intel.com>
Add PCI blacklist so we can skip only some devices.
Change-Id: I8600307dd53f32acb4dfeb3f57845e0b9d29fdb9
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442977 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447145
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: Jim Harris <james.r.harris@intel.com>
Bash interprets everything after command as additional
function arguments. To not confuse user just remove this part
and replace by '!'.
Change-Id: I44228003a1f96324271e726df4f5033f3258523c
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442976 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447143
Tested-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Introduced a new variable to run functional tests.
It's enabled by default, and can be manually disabled
on systems where e.g. only unit tests are run.
SPDK_RUN_FUNCTIONAL_TEST is a supplement to SPDK_UNITTEST.
The two are completely independent - both can be enabled,
disabled, or run in any combination.
The new variable is prefixed SPDK_RUN_ as it aligns nicely
with SPDK_RUN_CHECK_FORMAT, SPDK_RUN_VALGRIND, and
SPDK_RUN_ASAN, all of which control how much is tested.
SPDK_UNITTEST should eventually follow the same pattern
as well.
This gives us 2 layers of configuration:
SPDK_TEST_* <- what is tested
SPDK_RUN_* <- how it is tested
The following would run UT+ASAN for FTL and BlobFS, without
running their functional tests:
```
SPDK_RUN_FUNCTIONAL_TEST=0
SPDK_RUN_ASAN=1
SPDK_TEST_UNITTEST=1
SPDK_TEST_FTL=1
SPDK_TEST_BLOBFS=1
```
Change-Id: I9e592fa41aa2df8e246eca2bb9161b6da6832130
Signed-off-by: Seth Howell <seth.howell@intel.com>
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442327 (master)
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/447261
Tested-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Unmap, discard, write zeros will be sent down from
higher stack. Remove these IOs for the QoS limit.
Change-Id: Ieb3cc19f31c43f8ddf8f8d2fd338f442ef48b679
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442673
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Liang Yan <liang.z.yan@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When a connection goes to close and has no I/O outstanding,
the current_recv_depth was being decremented beyond 0 and rolling over.
If the poll group then finds a successful receive completion on the next
poll (for a command that arrived prior to starting the disconnect but
hadn't been processed yet), it would trip the max queue depth check
added recently and start another disconnect process. If only one command
arrives in this window, everything actually works out ok.
However, if there are two receive completions sitting in the completion
queue after the disconnect process is started, the first one does the
double disconnect and the second one does another disconnect which ends
up dereferencing a null pointer.
Since there is always a special reserved slot for the dummy recv, don't
do decrements or increments of the current_recv_depth for the dummy
recv. This allows the code to still enforce the actual max_queue_depth
on recvs without underflowing or overflowing the counter.
Change-Id: I56c95b2424e956a3b007b25c50cbf47262245b8f
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442642
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The RPC was replaced by set_bdev_qos_limit, but jsonrpc.md
contained declarations of both.
Change-Id: Icd636199b8af93b545d636377926983e62d38d11
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442730
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>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
The stop handler already calls blockdev_heads_destroy(), so we
don't need to call it if bdevperf_construct_targets_tasks()
fails. Calling it twice actually results in double-frees and
other types of memory corruption.
Fixes#592.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iba92b1ae64453036829a67ab6f3dad970a368af0
Reviewed-on: https://review.gerrithub.io/c/442628
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
trace_record is used to poll the spdk trace shm file
and store new entries from it to another specified trace file.
This could help retain the trace_entires from the overlay of
trace circular buffer
Note:
* trace_record reads the input tracefile into a process-local
memory and writes trace entries to the output file only at shutdown.
* trace_record can be shut down on SIGINT or SIGTERM signal.
A usage sample is:
./spdk_trace_record -s bdev_svc -p <spdk app pid> -f trace.tmp -q
Change-Id: If073a05022ec9c1b45923c38ba407a873be8741b
Signed-off-by: Liu Xiaodong <xiaodong.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/433385
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This RPC doesn't really work in some cases - for example,
trying to delete one NVMe namespace bdev from a controller
with multiple namespaces, or just one virtio SCSI device
from a virtio-scsi controller. We've previously kept it
and marked it as "debugging only" - but every bdev module
has its own RPC method now for deleting what it constructed,
so keeping the generic delete_bdev RPC is asking for
trouble in some of the cases mentioned above. We'll remove
it in the 19.04 release.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I639254b32a3e1c840a4e9ae2658c42f4f321b676
Reviewed-on: https://review.gerrithub.io/c/442616
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This was marked deprecated in the v18.10 release, so
remove it now before v19.01 is tagged.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I57673a5ab475b97c812bebcefd77ff90d9305d1c
Reviewed-on: https://review.gerrithub.io/c/442412
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>
nvme-cli in NVMf tests currently fails to initialize
the NVMe driver and exits straight away [1]. It expects
the primary process to initialize the driver first, but
since the primary process doesn't operate on any NVMe
devices, it doesn't initialize the driver at all.
Fix this by running nvme-cli without multi-process mode.
In NVMf tests, nvme-cli is only used to discover, connect
and disconnect from an SPDK NVMf target. It does not need
to access any shared memory resources.
This wasn't an issue before, because we used an outdated
DPDK version for nvme-cli which didn't detect any other
shared memory processes of SPDK.
[1] nvme.c: 337:nvme_driver_init: *ERROR*: primary process
is not started yet
Change-Id: Id56f94c6655049e87ab9d93ee38853faf40a11e5
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442552
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
- Add option to throttle iops in VMs using cgroups
- Add option to measure CPU utilization in VMs using SAR
- Add option to limit kernel vhost CPU cores (not NUMA optimized)
- Add option to do lvol preconditioning using fio bdev plugin
before running IO performance test
Change-Id: I7e0fcf977be96ecf837385c2abc9d5dabbe2f8c5
Signed-off-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/434229
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Paweł Niedźwiecki <pawelx.niedzwiecki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This includes properly detecting when a key's name
extends past the end of the valid data.
Note that the unit tests were using sizeof() instead
of strlen() since some of the strings contain
NULL characters. This means that we should be
subtracting one to account for the implicit null
character at the end of the string. Note that the
iSCSI spec only says that the key/value pair has to
end with a null character - a key/value pair that
is split across two PDUs will not have a NULL character
at the end of the first PDU.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie95d6dd3b9ffa6a3902a31771ac4edb482418cce
Reviewed-on: https://review.gerrithub.io/c/442450
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>
Since params are parsed directly from the PDU's data
buffer, we need to know the end of the valid data. Otherwise
previous PDUs that used this same data buffer may have left
non-zero characters just after the end of the text associated
with a LOGIN or TEXT PDU.
Found this bug while debugging an intermittent Calsoft test
failure. Added a unit test to reproduce the original issue,
which now verifies that it is fixed.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ic3706639ff6c4f8f344fd58c88ec11e247ea654c
Reviewed-on: https://review.gerrithub.io/c/442449
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>
The isa-l directory is generated automatically by
git on checkout, although it's empty. We check for
that directory existence in order to build ISA-L
and without `git submodule update --init` the build
just fails.
To fix, instead of checking the dir existence, we
check for the autogen.sh file inside that dir.
Change-Id: I7b62513afcd7e9f3953f825c41b265efe3484d57
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/442523
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Piotr Pelpliński <piotr.pelplinski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Signed-off-by: Piotr Pelplinski <piotr.pelplinski@intel.com>
Change-Id: Ie042df61e6fbae583cffb5022d3e26ca078f67fb
Reviewed-on: https://review.gerrithub.io/c/442558
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: Jim Harris <james.r.harris@intel.com>
We can use spdk_strtol and spdk_stroll instead, which do a better job of
catching errors.
Change-Id: I09ff75356b932366b6c10cd6953610f609ce4b0e
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/441984
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
