DPDK 20.11 moved the kernel modules to separate
dpdk-kmod repository. It has to be built separate
from DPDK.
If needed for testing vm_setup.sh script now contains option
to build this driver.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I98a5eb956eb0cc60ec402d88fcdbd66d4854f19a
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/6033
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Michal Berger <michalx.berger@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Broadcom CI
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is done in order to make sure IOMMU extensions, which DPDK is
looking for, are enabled.
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Change-Id: Ia9ff8e5329d14a5eb60721f08ff7063cfb1d6eb7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5259
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Mike Gerdts
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Similarly replace PCI_BLACKLIST with PCI_BLOCKED.
Use of ALLOWED/BLOCKED matches similar changes made
in DPDK.
While here, replace use of term "blacklist" with "blocked"
in one of the nvme perf scripts. The usage there was
associated with how devices are blocked by using the
environment variables that are changed by this patch.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I720d99118ba5e050f436612c9fd415db44294a63
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5275
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
If DPDK was not told to unlink the hugepage backing files during the
hugepage_init, the application will keep them open while holding a
lock on each file. Check if said lock is held, if not, remove the
file.
Change-Id: I738a3d7756c335b63388a2efeb03debd734d9c9e
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5065
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Main changes:
- By default, allocate all hugepages on node0. On NUMA-aware systems,
processes will most often use default policy with local node as the
preffered one for allocations. This usually is node0, thus splitting
hugepages evenly across the nodes, in default setup, would force
allocations from a remote node, impacting overall performance (in
case hugepages on node0 run out). See 68740678e1 as a reference.
- Introduce HUGE_EVEN_ALLOC - force setup.sh to evenly distribute
hugepages across all the nodes.
- Introduce HUGEPGSZ - overrides default page size
- Introduce CLEAR_HUGE - remove all hugepages on demand before
allocation is performed.
- HUGENODE - this now can be a list of nodes to use. NRHUGE will be
set across all the nodes from the list.
Change-Id: I084829edde3c416e7fc6b7b6abe369cc0631fcd7
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5042
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>
This is done so user can easily filter out parts of the output which
are not relevant for additional processing (greping, seding, etc.).
Change-Id: I67eafd1fbeb723224cca6b0026e149cb7f85bfe9
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5152
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>
This is done in order to improve the grepability of the output. Sort
is based on the BDF address.
Hugepages statistics are intact.
Change-Id: I121e9a46bfb1550380f233c1d5c06cefb9d4e7a9
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5022
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>
Each nvme BDF can have multiple namespaces, hence multiple block
devices, return them all.
Change-Id: I93fe8acf5b1904f05514445eb3e970ef4254caed
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5170
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: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: John Kariuki <John.K.Kariuki@intel.com>
lsblk will include all the potential holders|virtual devices given
block device may be part of. This is relevant in case a nvme|virtio
device is part of the lvm|md setup.
Change-Id: I32198ddcf71746c3cbad4a96496a1762a5c201cd
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/5167
Community-CI: Mellanox Build Bot
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: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
If they are not, there's no point to continue setup on FreeBSD.
Change-Id: I6d94712f1507c0597fbd569ff73c52f262ab9df9
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4658
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>
In case contigmem is loaded, but the kernel environment hasn't been
properly updated with hw.contigmem.* options, kenv will fail,
breaking the check to determine if contigmem should be unloaded.
Avoid the above scenario by determining first if given option is
actually set in kernel's environment.
Change-Id: I4118b7622fd876b28f0c07d081c583f5782f0357
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4656
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>
Some 4.18.x kernels are shipped with a broken uio driver due to the
following change:
git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=9421e45f5
Try to detect if we are running against a faulty uio driver and if so
fallback to igb_uio driver if present. The priority of picking up the
drivers has not changed.
Note that this commit may be deemed as not needed since from CI
perspective the https://review.spdk.io/gerrit/c/spdk/spdk/+/4342 may
be simply enough.
Change-Id: I9b12511c203c0be0e8f3f462c9c96babde52dc6e
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4343
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>
Recent refactors removed the behavior introduced with a6edaa9600.
Bring it back and don't try to rebind the VMD devices if they were
not explicitliy allowed in the setup.
Also, shuffle the code a bit and put verification pieces under one
block where $mode is being determined.
Change-Id: Ie2cc41e402f20147b98ab288d623ac76a4472839
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4398
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>
Error log as below:
06:27:53 # nvme_namespace_revert
06:27:53 # /home/storage/workspace/opal-end-to-end-reservation-autotest/spdk/scripts/setup.sh
/home/storage/workspace/opal-end-to-end-reservation-autotest/spdk/scripts/setup.sh: line 109: /sys/bus/pci/devices/0000:80:04.2/driver/remove_id: No such file or directory
home/storage/workspace/opal-end-to-end-reservation-autotest/spdk/scripts/setup.sh: line 110: /sys/bus/pci/devices/0000:80:04.2/driver/unbind: No such file or directory
06:27:54 # trap - ERR
Change-Id: I302e58143f0e165b17a9b5edd14121e6282f55a7
Signed-off-by: wanghailiangx <hailiangx.e.wang@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4094
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Michal Berger <michalx.berger@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Most distros ship these drivers as modules, however, some, like
clearlinux, have them compiled directly in the kernel (e.g. nvme).
In case modalias lookup fails, have a fallback prepared just in case.
Change-Id: Ib9da8ff69edbb0c4ce427aa089fa7d04b399a20b
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4047
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Maciej Wawryk <maciejx.wawryk@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The previous approach didn't take into the account systems where root
partition is representend in a virtual form of /dev/root (see
clearlinux). In case the device was bound to virtio this would fool
setup.sh into thinking nothing is mounted and proceed with unbinding
the entire rootfs.
Fix this by checking the maj:min numbers of the device instead of its
name against the mountinfo list.
Change-Id: I0feb2584869f6bb72df3e9a4e619620240cfce3b
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4046
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Maciej Wawryk <maciejx.wawryk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Make sure that setup.sh waits for block devices during the tests.
This is to make sure that the underlying controllers are ready to
be talked to and minimize the amount of their flakiness.
Change-Id: Id09445de7ac7ccf4c9f679ee2b6c4bdd14c89733
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4029
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>
Change-Id: Ifeadc3fac3f5d09143dd13ef3d510de9cecffb8a
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4032
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
There are some devices for which nvme driver takes a long time to
finalize the unbind stage. With that in mind, each device would
add up a significant amount of time needed for setup.sh to complete.
To mitigate such a scenario, make sure the controllers are unbound
in a parallel fashion.
Examples taken from the system with 19 nvmes on board:
[root@supermicro4 spdk]# time ./scripts/setup.sh &>/dev/null
real 0m36.250s
user 0m1.024s
sys 0m1.990s
[root@supermicro4 spdk]# time ./scripts/setup.sh &>/dev/null
real 0m4.848s
user 0m0.867s
sys 0m17.605s
Also, take note that this is currently done only for the nvme
devices since other, i.e., ioatdma, seem to trigger a BUG in the
kernel when unbound in parallel. Some details here:
https://bugzilla.kernel.org/show_bug.cgi?id=209041
Change-Id: Icaeb2b2ecb306f149587bc5da73743b1519bc5d6
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3893
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: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This is a first patch from a series attempting to merge similar
routines used throughout setup.sh.
Change-Id: I4c71b88f7556b0e9e2f65a37b8175914022486a4
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3889
Community-CI: Mellanox Build Bot
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>
When using vfio a warning for memlock limit is printed.
TARGET_USER is one set by caller when executing the script,
or set to user that called the script (even when using sudo).
Yet ulimit was printed for the one executing the script,
most likely root.
This patch verifies ulimit for a particular TARGET_USER.
Along with more explicit information which user the
information pertains to.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I20ad62109bc316295208e21f959ecfc263307e1c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1964
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The -1 is valid value indicating that the node id for given device
has not been specified. Detect it and map it to "unknown".
Change-Id: Id97d9046be0e4d5ff257bca66ae3e73e906253fe
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2229
Community-CI: Mellanox Build Bot
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>
Use * instead of ? to match all the potential nodes not only those
with a single digit id. The actual node limit is dependent on
kernel's CONFIG_NODES_SHIFT hence in theory, the actual number of
numa nodes may be far greater than 10.
Change-Id: I1d587735b433ed23594725b9d77a44f7cc729494
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2228
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Also, store kernel name in the separate var and use it throughout the
entire script.
Change-Id: Iaa1c4c4aa52fb4aa708fb476478a81e9c6067b58
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/2340
Community-CI: Mellanox Build Bot
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: Aleksey Marchuk <alexeymar@mellanox.com>
Expose a cache of pci devices in form of an assoc array that could be
looked up during the runtime of a script like setup.sh.
In case of setup.sh, caching speeds up execution quite visibly:
config run, no caching:
real 0m4.488s
user 0m1.440s
sys 0m1.260s
config run, caching in use:
real 0m2.876s
user 0m0.365s
sys 0m0.420s
Note that for initial config runs, binding controllers to proper
drivers is the actual bottleneck.
status run, no caching:
real 0m1.877s
user 0m1.252s
sys 0m0.984s
status run, caching in use:
real 0m0.371s
user 0m0.242s
sys 0m0.204s
reset run, no caching:
real 0m2.559s
user 0m1.409s
sys 0m1.322s
reset run, caching in use:
real 0m0.960s
user 0m0.432s
sys 0m0.419s
Additionally, in case common tools, e.g. lspci, are missing, fallback to
sysfs to pick all needed devices from the pci bus. Targeted for Linux
systems only.
Change-Id: Ib69ef724b9f09eca0cbb9b88f1c363edc1efd5dc
Signed-off-by: Michal Berger <michalx.berger@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1845
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Add bin/unbind support between the kernel idxd driver
and VFIO. UIO not tested but tested this extensively
on a system with valid idxd kernel driver and VFIO.
Future patches in this series add the actual idxd
low level library and bdev accel plug-in module. See
those patches for more details on idxd.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I09dd00b1aedcf88577335bde79d31cfd50ac4cde
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/1721
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When overriding with igb_uio you used to have to manually probe uio
before calling setup.sh. I feel this is a common enough use case that we
should check it in setup.sh directly.
Change-Id: I8a31f90bab6960eade468816ffe960820f83d2fe
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/818
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>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Correct shellcheck rule SC2089: Quotes/backslashes will be treated literally. Use an array.
Signed-off-by: Maciej Wawryk <maciejx.wawryk@intel.com>
Change-Id: I40d4eef334cd030294669f8e8494f9e7a563d495
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476466
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>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
It fixes change 474989.
"for blkname in $blknames; do"
was changed to
"for blkname in "${blknames[@]}"; do"
if blknames=" vda" then first blkname="" and
linux_bind_driver is called for vda although it has active mountpoint
Change-Id: I16d4379db274faa5703230072cfcf8d18a8ef5e6
Signed-off-by: Pawel Kaminski <pawelx.kaminski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476061
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Correct shellcheck rule SC2128: Expanding an array without an index only gives the first element.
Signed-off-by: Maciej Wawryk <maciejx.wawryk@intel.com>
Change-Id: I0e7c335af678114dc78dfb12a02369a69158e435
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474989
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Karol Latecki <karol.latecki@intel.com>
Correct shellcheck rule SC2166: Prefer [ p ] && [ q ]
as [ p -a q ] is not well defined.
Signed-off-by: Maciej Wawryk <maciejx.wawryk@intel.com>
Change-Id: I3fb2fb0e790bbdd296b64673c301324632d7c37b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472173
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
