Phenomenon:
Test case: Using the following command to test
./test/nvmf/target/shutdown.sh --iso --transport=tcp
without this patch, it will cause coredump.
The error is that the NVMe/TCP request in data buffer
waiting list has "FREE" state.
We do not need call this function in
spdk_nvmf_tcp_qpair_flush_pdus_internal, it causes the
bug during shutdown test since it will call the function
recursively, and it does not work for the shutdown path.
There are two possible recursive calls:
(1)spdk_nvmf_tcp_qpair_flush_pdus_internal ->
spdk_nvmf_tcp_qpair_process_pending ->
spdk_nvmf_tcp_qpair_flush_pdus_internal ->
>..
(2) spdk_nvmf_tcp_qpair_flush_pdus_internal->
pdu completion (pdu->cb)
->..
-> spdk_nvmf_tcp_qpair_flush_pdus_internal.
And we need to move the processing for NVMe/TCP requests
which are waiting buffer in another function to handle
in order to avoid the complicated possbile recursive
function calls. (Previously, we found the simliar
issue in spdk_nvmf_tcp_qpair_flush_pdus_internal for
pdu sending handling)
But we cannot remove this feature,
otherwise, the initiator will hang for waiting the
I/O. So we add the same functionality in spdk_nvmf_tcp_poll_group_poll
function.
Purpose: To fix the NVMe/TCP shutdown issue.
And this patch also reables the test for shutdown and bdevio.
Change-Id: Ifa193faa3f685429dcba7557df5b311bd566e297
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462658
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When individual tests were run with --iso --transport=tcp, the
check_ip_is_soft_rocefunction incorrectly marked the loopback ip address
as soft_roce due to 'lo' matching with a notice message from the rxe_cfg
config command "module rdma_rxe is not loaded". Instead, grep
specifically to make sure that the argument we are matching against,
which is supposed to be an rxe interface name, contains rxe.
rxe interfaces are numbered sequentially as rxe0 rxe1 . . . rxen
Change-Id: Ibe8afb3e038e104173eb1e2bbf671ce9f417ffeb
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463127
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This test is very unstable on TCP. Until we can get to the bottom of the
latent failures, we should disable it. This will allow us to continue
testing a substantial amount of tcp functionality while avoiding latent
failures in the test pool.
Change-Id: I3c7ec2709ad81a3ec5c1a4d56027b27a9c6ffc7e
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462919
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>
When you run verify on bdev perf, it rejects any user provided core
masks and pins itself to core 0. We were also running the NVMe-oF target
on cores 0-3. Occasionally I was seeing the shutdown tests failing
during the waitforio step. It seemed like the target and bdevperf were
both still online during this failure, so my hypothesis is that the
bdevperf poller was stealing all of the cycles from the NVMe-oF poller
completing I/O so the I/O wasn't completing.
Even if that's not what's happening, it is probably a good idea to split
the NVMe-oF application and bdevperf onto different cores.
Change-Id: Ib3b5b00e639ebd14bd1ed2cfb4b7782076ca364c
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462551
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
In the RDMA transport, the initiator doesn't properly pick up the lost
connection and we have to kill it. However for TCP, the initiator
realizes the closed socket and fails out. Then when we go to kill it, it
fails the test. So add and || true when killing the perf processes.
Change-Id: Ifed5d726946bad2e9396db40b40f1fee72b4597f
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461993
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Kernel versions 5.0.x do not support the -o option from our target.
Currently, kernel versions 5.1.x and 5.2.x are broken for the tcp
transport in loopback, so in order to test against a stock fedora kernel
that works, we will need to use 5.0.x for at least the next few months
until the distributions start shipping 5.3.x which contains the fix we
need.
Change-Id: I41fc5956dd4b53f03879e0adf2427872cd2337ee
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461741
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
For a lot of the tests, we may want to specify different options for
each transport. I believe this will be more prevalent as we add more
transport specific options.
Change-Id: I83a915629460d1d869eaba4bc86822d7563402ac
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461740
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
The file is not meant to be executed, so shebang is
not needed - let's remove it.
Change-Id: I51c32613ffb25b43a2d1b96523a8f25c853594e4
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461355
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>
Currently we test both transports under perf each time we run the nvmf
tests. Once we have created a jenkins job for the tcp transport, we will
only need to run this test against the specified transport for each job.
Change-Id: I053d11f4e6d6f4b86a3f29e6c8f43e1e196bedab
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/460714
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Enables us to test randomized data against the NVMe-oF target interface.
Change-Id: Ie7ab46949ccd89f74b10b79a24256aeae2df89ab
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/431571
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Hailiang Wang <hailiangx.e.wang@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: qun wan <qun.wan@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
By moving the modprobe and modprobe -r calls into nvmftestinit and
nvmftestfini respectively, we can make these calls symmetric in all
cases The previous format where nvmfappstart was adding modules and
nvmfcleanup was being called independently to remove them started to
cause issues when we added the tcp transport.
I think this fixes#846
Change-Id: I68d18e0be5a2d4d2ea83a5a34f0aa2da268209fe
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/460396
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@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>
When we insert the nvme-$TEST_TRANSPORT modules into the kernel, we
allow for it to fail. So, for the sake of symmetry, we should allow for
the opposite to be true.
Thanks to Maciej Wawryk for pointing out the discrepancy.
Change-Id: I3c551950cbe25382c2bbea4f390bead0ba6f2a37
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/460395
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Purpose: Currently, there are many patches related with
NVMe-oF, especially related with big I/O. We need to
add a test case to guard this.
Change-Id: Iddd4070976fe6bf4926ca489d9c97f64a108a422
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/460065
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>
Also, construct this path by grabbing the home directory
of the current user.
Change-Id: Ia8d5cc2be40c0c3ac693c80a3132b970f7124183
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456704
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is much easier.
Change-Id: I4ae5f2f5b9393f65d07f39f03fa30628a40b01cf
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459304
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When we have a really large disk, setting up a nested lvol across the
entire drive can actually be quite time consuming. When trying to run
this test on an 800G p3700 I was actually getting timeouts on the rpcs
when creating the nested lvolstore. By only taking the first 20G of the
drive to set up the stores, we end up with a much quicker test and
conveniently get rid of the timeouts I was facing.
Change-Id: Ifd43c21c66bc77b615da9b0f421e95337b4c7019
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/459849
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
When possible we should make sure that no silent failures or
skipped tests.
Since some of the tests use fio, autotest should fail
in those places when it is missing.
Change-Id: I97581e2df30a02445b560e3b88d03ce92986d0fd
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456316
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
It's always set in autotest_common.sh, there's no need
to set it again in each test script.
Change-Id: Ib14c4189c553dad54a3065c1a1d413a5fc5a5347
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457466
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>
In fio.sh we used `set +e` for bdev deletion RPCs
and for `wait $fio_pid`. Disabling errors from those
RPCs doesn't make sense. We expect FIO to fail due
to bdev hotremoval, but the bdev hotremoval itself
must succeed.
Besides, it's easy to forget to undo `set +e` so
don't use it. For `wait` we can use the following
paradigm instead:
```
rc=0
wait $pid || rc=$?
```
The same applies to nmic.sh, where we expect some
RPC to fail.
Change-Id: I66ce3504a0c6b5da497759f0688b6c0ea6480b61
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457463
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: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
ASAN needs to be LD_PRELOADed before SPDK fio_plugin
in order to analyze its code.
Just adding that will report any issues in fio binary as
well as the fio_plugin.
To prevent known fio leaks from affecting the results,
a suppression list for LeakSanitizer (used in conjunction with ASAN).
At this time the suppression list contains known leaks
for fio 3.3. The list might need adjustments as fio
version is updated.
Side note. Even though it is possible to specify directory
to ignore ("leak:/usr/src/fio/"). Which in theory should
suppress any leaks in fio. It has side effect of hiding
SPDK leaks as well, since the fio_plugins leaks are
seen as coming from /usr/src/fio/ioengines.c.
See below for examples of each suppressed error:
Direct leak of 42 byte(s) in 4 object(s) allocated from:
#0 0x7f9d52f3e320 in strdup (/lib64/libasan.so.5+0x3b320)
#1 0x41f267 in get_new_job /usr/src/fio/init.c:490
Direct leak of 914936 byte(s) in 10397 object(s) allocated from:
#0 0x7f74422e8ea6 in __interceptor_calloc (/lib64/libasan.so.5+0x10dea6)
#1 0x46402e in log_io_piece /usr/src/fio/iolog.c:214
Direct leak of 608 byte(s) in 19 object(s) allocated from:
#0 0x7f74422e8ca8 in __interceptor_malloc (/lib64/libasan.so.5+0x10dca8)
#1 0x44c4e1 in add_to_dump_list /usr/src/fio/parse.c:1039
#2 0x44c4e1 in parse_option /usr/src/fio/parse.c:1098
Direct leak of 173 byte(s) in 20 object(s) allocated from:
#0 0x7f744227153d in strdup (/lib64/libasan.so.5+0x9653d)
#1 0x44b50d in __handle_option /usr/src/fio/parse.c:718
Indirect leak of 111925528 byte(s) in 1271881 object(s) allocated from:
#0 0x7f74422e8ea6 in __interceptor_calloc (/lib64/libasan.so.5+0x10dea6)
#1 0x46402e in log_io_piece /usr/src/fio/iolog.c:214
Indirect leak of 171 byte(s) in 19 object(s) allocated from:
#0 0x7f744227153d in strdup (/lib64/libasan.so.5+0x9653d)
#1 0x44c4ed in add_to_dump_list /usr/src/fio/parse.c:1040
#2 0x44c4ed in parse_option /usr/src/fio/parse.c:1098
Indirect leak of 167 byte(s) in 19 object(s) allocated from:
#0 0x7f744227153d in strdup (/lib64/libasan.so.5+0x9653d)
#1 0x44c502 in add_to_dump_list /usr/src/fio/parse.c:1042
#2 0x44c502 in parse_option /usr/src/fio/parse.c:1098
Change-Id: I9b5811993508421be50b12af160645c77ea93d7e
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456315
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Scripts were using a mix of two approaches, lets unify that so
just dollar-parenthesis syntax is used.
Change-Id: Id093b6c82d1f766ba6af13bed720977eceaa7ffc
Signed-off-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457744
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Pawel Kaminski <pawelx.kaminski@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Partprobe is a program that informs the operating system kernel
of partition table changes, by requesting that the operating system
re-read the partition table. For instance, if you create a new
partition on one of your disks using parted, you should run
partprobe afterwards to make the kernel aware of the new partition
configuration.
This is useful for issue #799.And it's compatible with other systems.
Change-Id: Icd4c85193bd9d9e6c2b32b8463e75c7a6ff06f34
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/457735
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Normal return for qemu checking , this should check before vhost/common.sh.
For qemu not installed on this machine. It may be a VM. Skipping nvmf_vhost test.
Otherwise, code will exit 1 before this normal return.
And in vhost/common.sh should with qemu checking.
Change-Id: I2285fffa6fc4384b4b172e5c443019e9ab391fa0
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456123
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This patch adds functions in autotest_common.sh,
fio_bdev() and fio_nvme() for their respective plugins.
It simplifies its usage throughout the scripts.
Additionally will help with expanding the fio cmd
line with nessecary changes only in single place.
This will be used in next patch in series to
LD_PRELOAD ASAN before fio_plugin.
Note: Did not implement changes for perf scripts,
since they are usualy run separately and didn't want
to affect those.
Similarly didn't change vhost initiator tests,
because the exported directories do not work for
inside the VM. Will need to think of a way to change it
there as well.
Change-Id: Idf750ee7aa9d5e30efc0ce430d15fa45ceccb203
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456314
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Use the new nvmfappstart to simplify this script a
bit and bring it inline with the other nvmf scripts
that have been updated recently.
Also remove the comments about aer with temperature
threshold. We use namespace notifications to test aer
now, so there's no need to try to resurrect the
temperature-based tests.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9c66ed4c9456acb39d6c64d55153304eb6e07100
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456706
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Change autotest.sh to pass --transport=rdma. But this
now allows someone to run nvmf.sh with --transport=tcp
to run all of the nvmf tests with the tcp transport!
Initial testing shows that bdevio.sh fails with tcp,
so only run this test with the rdma transport. Issue
#808 has been filed - once it's resolved this
restriction can be removed.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie50de4d67e504c84d6d8eebfe2b8c68b14f4ba2b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456698
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
While here, remove NVMF_TCP_IP_ADDRESS since we already
have NVMF_FIRST_TARGET_IP which serves the same purpose
when testing the TCP transport.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7cc4712cd9746377937e889127aa5a61566d8846
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456705
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
We also no longer hardcode TEST_TRANSPORT=rdma. Users
of the individual test scripts must now explicitly
pass --transport=rdma. Support for tcp will be coming
in future patches.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I819d69cb0906eb27b692eb2755aca5085298d779
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456685
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Right now this is empty, but we'll add to it later as
part of parameterizing this script for different transports.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Icf0f67ce4dd886031c37986b352129d3c556bd92
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456684
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This prepares for using scripts for tcp testing as well.
Note that this patch just hardcoded TEST_TRANSPORT to
'rdma' for now. An upcoming patch will require the
caller to pass --transport=rdma instead.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I085fdf51910aaf960959c71c73a187be41fd7ee8
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456667
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
A lot of places were still hardcoding 4420 directly.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iff8a549addeb05e441fe3b035a84a0f8a6bf2205
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456666
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This script was developed in parallel to some of the
script infrastructure improvements. So fix it up to
match the more recent changes - including leveraging
the common setup of the NVMF_FIRST_TARGET_IP, and removing
the parameters to nvmftestinit/nvmftestfini.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I6cc4ef30325961822014e2835e575a950c53ecc0
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/456665
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This script was getting a bit unwieldy. adding argparse will make it
easier to add more options in the future or to set default values and
validate arguments going forward.
Change-Id: I1ffdbdf2082287ceb8a88cd3eb6ecf9bbd6c9e11
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455724
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: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Good suggestion from Darek - let's just always
parse common script args from autotest_common.sh.
These arguments follow common arg naming conventions
(i.e. --iso) so there's no harm just doing this for
any test that sources autotest_common.sh. This has
the nice effect of not requiring scripts to
explicitly call this function.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Id89b68c22557a5a771be407873d0e57843f0d05a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455552
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
iscsi test scripts can now take two arguments -
"iso" and then the sock type (posix or vpp). They
need to be in that specific order too. nvmf test
scripts also support "iso" and we want to add
the transport type (rdma or tcp) as well. Even further
out, we may want to use a sock type for nvmf, i.e.
tcp transport with vpp.
We also have the iscsi_tgt fio_remove_nvme.sh test
that does both iscsi and nvmf.
So to make this all work a bit nicer, add a new
function called parse_common_script_args that
will take the command line arguments to a script
and set the appropriate variables, including defaults
when a specific parameter isn't specified. We will
use getopt-like behavior for this also, instead of
enforcing a specific parameter order. Then a script
could be called like this:
test/nvmf/target/shutdown.sh --iso --transport=tcp --sock=vpp
Individual test scripts then just need to do this
after sourcing autotest_common.sh:
parse_common_script_args $@
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ifb8d7666384991482a2d425e26ffa7525b9ac15a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455283
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>
There's a bug in Linux 5.1 (possibly 5.0 too) where
the kernel initiator driver crashes if it can't get
one queue per CPU. This will get fixed eventually,
but for now we need to remove the cases where we
restrict the number of queues per controller so that
we can test on newer kernels.
Even on cases where we're testing the SPDK initiator,
there's no real need to restrict the number of queue
pairs.
The kernel will eventually get fixed, but we should
be testing with default behavior anyways (the kernel
wants lots of queues). We'll also want to add some
regression tests to make sure the kernel doesn't
break again. But that will all come later.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9979e6d94456e075688b822b042936b63e518a4a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454819
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Rather than revert this patch, just modify the path to the common script
to enable it to pass in the test pool.
Change-Id: I33f5d89e1e118df4546f45237b55173ad30fad24
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455277
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 will allow us to verify functionality between the nvmf target and
the vhost target and test extra code paths within the nvmf target
including multi-sgl stuff.
Change-Id: I3f5e9351c11ab896b75cd7bba7a69d95c1d031be
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/451993
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Nested lvolstores take a very long time to create,
since the initial unmap will get broken into
one unmap per cluster for the underlying lvol.
This test is working fine in the test pool, but they
have much smaller SSDs. My system has a 3.7TB SSD,
and this test times out.
So don't bother clearing the nested lvol store - this
saves unmap operations. We still have unmap operations
when deleting the nested lvol, so increase the cluster
size for the main lvolstore to 1GiB to reduce the total
number of unmap operations.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I29a254fff443e963cd620b55e78092d6a96f8ddd
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/453010
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: John Kariuki <John.K.Kariuki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Just name the malloc bdev explicitly. And since there
is only one bdev, remove all of the unnecessary loops.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie3b55161d6ebee64497bc9726edaa3fbf04bb445
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454690
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
1) remove unnecessary disconnect for cnode2 - this
subsystem doesn't exist (and the || true was
hiding the failure
2) remove || true from the cnode1 disconnect
3) remove extra spaces when assigning nvme_model -
maybe these spaces are OK on some versions of
bash, but when I run this locally this doesn't
work at all
4) remove trailing spaces from model returned by
identify data before comparison
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If3dd7ddc2ba44fc880e5b398b5fe0922b6637b80
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454689
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
A few test scripts were creating 128MB malloc bdevs
instead of 64MB - but there's no real reason for the
difference. So make them all the same. A future
patch will make all of these size variables common
instead of duplicated in each file.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I865ef08f384e026d0d3e7b064fb3a2d5958f054c
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454688
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Adding the second bdev doesn't provide any real value since
it's the same kind of bdev.
While here, explicitly name the malloc bdev to simplify
the test some more.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ide6da901d4b90383cc73fa195b5a070a8eda0d01
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454687
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
1) specify malloc bdev name explicitly
2) just use a single loop variable
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7fb4fc28715f7303fc2b39de80985939ab469887
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454686
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is a preparatory patch for specifying the transport
when calling the script. Better to use "rdma" across the
board instead of a mix of "rdma" and "RDMA".
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If7170cff6ba2ec51e92a0423c0b24cc141054b55
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454685
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
All of the checking for NVMF_FIRST_TARGET_IP, starting
the target, waiting on the target and modprobing
nvme-rdma is duplicated in every script. So move
a lot of this either to nvmftestinit() or a new
nvmfappstart() function in common.sh.
Also just kill the nvmf target in nvmftestfini, rather
than the script to explicitly kill it.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I5864404610a4244473f460d48264de92687ed867
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454678
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Pull this parameter when sourcing common.sh
and set a RUN_IN_ISOLATION flag if it was specified.
This reduces the overall code a bit plus avoids having to
pass the "iso" parameter to nvmftestinit and nvmftestfini.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Idbbe27bfb7b62857500176cdb5d90e381fce4ffc
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/454814
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
