IB Architecture Specification vol.1 rel.13. in ch.10.3.1 "QUEUE PAIR
AND EE CONTEXT STATES" suggests the following destroy procedure for
QPs associated with SRQ:
- Put the QP in the Error State;
- wait for the Affiliated Asynchronous Last WQE Reached Event;
- either:
* drain the CQ by invoking the Poll CQ verb and either wait for CQ
to be empty or the number of Poll CQ operations has exceeded CQ
capacity size; or
* post another WR that completes on the same CQ and wait for this WR
to return as a WC;
- and then invoke a Destroy QP or Reset QP.
Without the drain step it is possible that LAST_WQE_REACHED event is
received and QP is destroyed before the last receive WR completion is
polled from the CQ.
In SPDK there is no risk of resource leakage in this case. So, instead
of draining we can destroy QP and then just ignore receive completions
without QP and post receive WRs back to SRQ.
Fixes#903
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465747 (master)
(cherry picked from commit 87ebcb08c1)
Change-Id: Ice6d3d5afc205c489f768e3b51c6cda8809bee9a
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468300
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Eugene Kochetov <evgeniik@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The dependencies between vhost and rte_vhost were not added during
earlier changes. This change moves the rte_vhost directory up to the
level of the other libraries and adds the proper dependencies for when
it is linked.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467700 (master)
(cherry picked from commit 28099e0ed9)
Change-Id: I089de1cd945062b64975a0011887700c0e38bb0f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467947
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>
Start processing ANM events only after the device is fully initialized.
Otherwise some of the structures are partially filled and can be
interpreted incorrectly.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466935 (master)
(cherry picked from commit bd78196c09)
Change-Id: Ia741730cf15d44d76ce8afa7955e6a5bf42ca42b
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468305
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>
Track the number of acquired but not yet submitted write buffer entries
to be able to correctly calculate the required number of entries to be
padded.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466934 (master)
(cherry picked from commit a2714d414f)
Change-Id: Ie201681937ad1d03ec125aa5912311c54a7e35c9
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468304
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>
When recovering the data from the non-volatile cache, the data inside
the volatile cache needs to be flushed before flushing active bands.
Otherwise, if the number of blocks in a band is smaller than the number
of blocks inside the volatile cache, part of the data may not get
flushed.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466883 (master)
(cherry picked from commit cf3d42961b)
Change-Id: I4e99709c8c2a526a928578870d7fbd5fef37db02
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/468303
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 reverts commit 6129e78d26.
When the initiator sends the discovery log page, if the log page
exceeds the size of its data buffer, it will break it up into
multiple log page commands with appropriate offsets. However,
supporting offsets in log pages is an optional feature in NVMe
and reported by the EDLP bit in the identify data.
This commit changed the discovery process to no longer send an
identify command prior to doing the discovery log page command,
so the values in the identify data are always 0. If the discovery
log page exceeds the size of the data buffer (4k), it will then
fail to send the second log page with an offset because it
believes the controller does not support the feature.
Revert this change to fix it. An identify should always be sent
as part of the discovery process. A test case is included in a
follow up patch the demonstrates the bug.
Reported-by: Zahra Khatami <zahra.k.khatami@oracle.com>
Reported-by: Akshay Shah <akshay.shah@oracle.com>
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466819 (master)
(cherry picked from commit 647afdec44)
Change-Id: Iefd512a7521e0fea90541b3eb547671cfa816ea6
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467946
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>
Keeping a global discovery log page was meant to be a time saving
mechanism, but in the current implementation, it doesn't work properly,
and can cause undesirable behavior and potential crashes. There are two
main problems with keeping a global log page.
1. Admin qpairs can be assigned to any SPDK thread. This means that when
multiple initiators connect to the host and request the discovery log,
they can both be running through the spdk_nvmf_ctrlr_get_log_page
function at the same time. In the event that the discovery generation
counter is incremented while these accesses are occurring, it can cause
one or both of the threads to update the log at the same time. This
results in both logs trying to free the old log page (double free) and
set their log as the new one (possible memory leak).
2. The second problem is that each host is supposed to get a unique
discovery log based on the subsystems to which they have access.
Currently the code relies on whether the discovery log page offset in
the request is equal to 0 to determine if it should load a new discovery
log page or use the cached one. This is inherently faulty because it
relies on initiator provided value to determine what information to
provide from the log page. An initiator could easily send a discovery
request with an offset greater than 0 on purpose to procure most of a
log page provided to another host.
Overall, I think it's safest to not cache the log page at all anymore
and rely on a thread local fresh log page each time.
Reported-by: Curt Bruns <curt.e.bruns@intel.com>
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466839 (master)
(cherry picked from commit 20b35d769d)
Change-Id: Ib048e26f139927d888fed7019e0deec346359582
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467594
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>
In the past, memory in spdk could have been unregistered in
different chunks than it was registered, so to account
for that the vtophys code used to register each hugepage
(2MB chunk of memory) separately to the VFIO driver. This
really made the code generally simple.
Now that memory in spdk can only be unregistered in the same
chunks it was registered in, we no longer have to register
each hugepage to VFIO separately. We could register the
entire memory region with just a single VFIO ioctl instead,
so that's we'll do now.
This serves as an optimization as we obviously send less
ioctls now, but most importantly it prevents SPDK from
reaching a VFIO registrations limit that was introduced
in Linux 5.1. [1]
The default limit is 65535, which results in SPDK being able to
make only the first 128GB of memory DMA-able. This is most
problematic for vhost where we need to register the memory
of all the VMs.
Fixes#915
[1] 492855939bdb59c6f947b0b5b44af9ad82b7e38c
("vfio/type1: Limit DMA mappings per container")
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/432442 (master)
(cherry picked from commit be04cfc342)
Change-Id: Ida40306b2684e20daa2fd8d12e0df2eef5a4bff1
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467143
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>
We'll be now able to check contiguity for more than 2MB
regions.
Signed-off-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466073 (master)
(cherry picked from commit 43f4e3932a)
Change-Id: I738ff451d534075c944972918d08e5e0cadea4f5
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467142
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>
One of stop conditions in data WR release function was wrong. This
can cause release of uncompleted data WRs. Release of WRs that are
not yet completed leads to different side-effects, up to data
corruption.
The issue was introduced with send WR batching feature in commit
9d63933b7f.
This patch fixes stop condition and contains some refactoring to
simplify WR release function.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466029 (master)
(cherry picked from commit 01887d3c96)
Change-Id: Ie79f64da345e38038f16a0210bef240f63af325b
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467140
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Every exit path in spdk_bdev_set_qos_rate_limits() should
go through _spdk_bdev_set_qos_limit_done() as soon as
ctx is allocated and qos_mod_in_progress set to true.
This patch fixes one path were it did not occur.
With this change qos_mod_in_progress is set to false,
when failure in allocating qos parameters occurs.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465659 (master)
(cherry picked from commit cfc0fbf11e)
Change-Id: I04a45dfdcde9160fd2701b44f5fde26fb0245177
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467138
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Padding size calculation should take into account
already acquired write buffer entries in case
number of blocks left for closing band is less than
write buffer size.
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463525 (master)
(cherry picked from commit b08337c216)
Change-Id: I0f9c9bc94eb5a4736eede30ec424c81103e1256f
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467136
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
WRITE_TO_RO_PAGE was incorrect and misleading. This
0x82 NVMe status code indicates a write to a read-only
range of LBAs. So modify the constant name and
associated usages to use WRITE_TO_RO_RANGE instead.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465083 (master)
(cherry picked from commit 0aa72ffb74)
Change-Id: I993dbebb5acc2e685a0e99aa14084942ef79d659
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467133
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 case of failure during resource allocation within poll_group_create
there is a lack of return statement which could lead to NULL ptr
dereference.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463195 (master)
(cherry picked from commit db0c7f6a4f)
Change-Id: I84abe64a1843117d76b97e62656bdfc4fe2b35d8
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467127
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: <jacek.kalwas@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Checked code, we do not free the memory allocated
spdk_sock_placement_id_entry.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463725 (master)
(cherry picked from commit a2dcdde2da)
Change-Id: Ie614dc17334b21b8904b16ee7e6e68a24e29d6a3
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467126
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>
The OCF build was broken by some of the recent changes
to the Makefiles. This change aims to fix that by separating out the ocf
environment from the ocf bdev.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465808 (master)
(cherry picked from commit 407e88fd2a)
Change-Id: Id445340033898e9ae70a4bcfc799951110762d55
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467293
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
When building shared libs, we specifically don't set the --as-needed
flag so that we still link against libraries that have constructor
functions. LD reports on its man page that the default behavior is
equivalent to --no-as-needed.
For RHEL based distros like fedora and centos this works fine. While
the LD man page shipped with ubuntu distributions also states that
--no-as-needed is the default, they don't respect that behavior and do
--as-needed linking unless you force them with --no-as-needed.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/464621 (master)
(cherry picked from commit e29c6fe133)
Change-Id: I914ab849323de198af5c5e53fffb1f57fcaff5fe
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/467292
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
By adding this file and modifying the way we define a couple of
variables in the makefiles, we can actually avoid having to redefine
the library dependencies on a per file basis.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466063 (master)
(cherry picked from commit 9d14641dd5)
Change-Id: Ieab4aa1021b0341fc21e3b65677a9ad7f70559c2
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466980
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
These directories fit in with the module concept we are forming inside
of SPDK. Essentially modules are derivative or specialized libraries that
rely on a general or core library.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465455 (master)
(cherry picked from commit f34fa35a52)
Change-Id: Ib40f05422f144ff8fd579f47a3867ef4412b3372
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466976
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is more accurate to what they are, and will make defining library
dependencies much simpler. This change in directory does not affect the
final placement of naming of libraries at the end of time.
Signed-off-by: Seth Howell <seth.howell@intel.com>
reviewed on https://review.gerrithub.io/c/spdk/spdk/+/466975 (master)
(cherry picked from commit 07fe6a43d6)
Change-Id: Ic48a9233dff564e39ce357a9ea0a111ea2b6414b
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466975
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Part of a larger series unifying and simplifying the library directory
structure.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465211 (master)
(cherry picked from commit edc8c91a90)
Change-Id: I2782165aabbea9a31cc466fc7e3bb2b9263142dc
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466974
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This is part of a larger series aimed at simplifying/flattening the
directory structure of the SPDK lib directory. The ultimate goal of this
series is to properly create dynamic linker dependencies between all
spdk shared objects.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465209 (master)
(cherry picked from commit ef6108a7eb)
Change-Id: I6beb7103404ae2c24a3d25dd93a1061680c7176c
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466973
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Part of a larger series simplifying the library directory.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465212 (master)
(cherry picked from commit ff69b368b8)
Change-Id: Ib9c9dc9a0c92ac35a9f0260451f97fc126d10031
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466972
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Combinining these two libraries removes some directory complexity. It
also helps us to align on a common practice for including the rpc code
with a given module. This is how all of our bdev modules already do it.
The nbd, net, and scsi modules also follow this pattern.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465210 (master)
(cherry picked from commit bc83adc2db)
Change-Id: I5e4c99d7f0facacc6dfe30b2274b60e0d151f8d8
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466971
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The log_rpc library has several dependencies which depend upon the log
library. This creates a circular dependency chain that makes single
threaded make unwieldy and makes multi-threaded make impossible.
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465192 (master)
(cherry picked from commit 9d1e4260cb)
Change-Id: I35e6532afcabce0f25974ed97444a56975654904
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/466970
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Will execute short run per patch, longer nightly. This
path only covers the ISAL PMD as there are QAT upgrades
needed in CI. It does so by using the new 'perform_tests'
RPC in both bdevio and bdevperf.
Change-Id: I93afda336d7411b695b2ddd0332cce0050d075a4
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/455113
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Chunyang Hui <chunyang.hui@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We need to compare the group info, then delete the
related entry.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: Ie5ed7f3ce8611fd61448788f54a63f624c813c19
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463176
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.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>
A single I/O may allocate more than one request, since splitting may be
necessary to conform to the device's maximum transfer size, PRP list
compatibility requirements, or driver-assisted striping. Very big
I/O request sent from application may get error due to limited resources
in NVMe driver layer, so here we add an optional parameter to make the
parameter can be configured by users.
Fix issue #745.
Change-Id: I7824232c54865b052dcd0ec6e91484c3837fc2c4
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/461182
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
For aarch64 spdk_mb is defined as "__asm volatile("dsb sy" ::: "memory")"
and spdk_smp_mb is defined as "__asm volatile("dmb ish" ::: "memory")".
"dsb sy" completes when all instructions before this instruction complete.
And "dmb ish" only keeps the order of memory access instructions before
and after it. It doesn't affect the ordering of any other instructions.
Here I think "dmb ish" is enough.
And replace spdk_mb with spdk_smp_mb here will not affect PPC64, X86_64
and i386.
Signed-off-by: Richael Zhuang <richael.zhuang@arm.com>
Change-Id: Idd65f74567c594e5a6514a4ab0665b8f3606c883
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463455
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Was 1024 but w/the performance optimizations we can now use up
to 4 mbufs per outstanding IO so going to 8192 only uses 1MB of
mem and gives us better head room before we have to start queueing
compression operations to wait for mbufs. Note: the mbuf pool is
shared accross all virtual compression bdevs.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I64d3973e7c7d31925b1ce5028ef960f90331c067
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463538
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
NVMf statistics functions use spdk_get_io_channel function to get a
poll group. It increases reference counter in io channel and causes
problems on application exit. spdk_put_io_channel calls were added to
release the channel.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I832d1eae346c3bc3858ed0ed063ff7a7a897a2f5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463389
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
There's a small drive-by in here also, use the max iovec define
from the reduce header instead of manually making sure its the same
here.
The change in the size of the MBUF pool is to accommodate the new
QAT driver being faster (?) than the old such that we could run out
of MBUFS with a fio test using 6 threads. On the + side, when we did
the queueing code in the compression operation worked good.
Also related to QAT and experimenting with fio compression ratios, QAT
can fairly often return an error on compression that we were treating
as an error for printing, change it to a NOTCELOG instead of an error.
Change-Id: Id842064b7d659f7ad7f9d785ffffdc6b827f6989
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463369
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: Jim Harris <james.r.harris@intel.com>
Changed to 17 from 32 per conversation with Jim and moved to
header file so it can be shared with the vbdev module.
Change-Id: Ic516b6466bf21d21d8073e9180deaa9929c6b742
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463368
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Old name was misleading due to earlier changes
Change-Id: Ic04223db983e98c9703a43e6778b0094f99bc81d
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463341
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>
qpairs can be limited by the compression polled mode driver or
system resources and we need one per thread so if we run out we
have to fail the callback.
Change-Id: Ibd07d9c65880e7de321c439119aed479ef00746c
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463339
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>
If a write operation comes in and is not a read-modify-write but
we don't end up compressing the data (because of either error or
we decide the compression ratio isn't good enough) then we not
only need to copy the data to the host buffers but we need to
zero any offset and/or remainder in the decomp scratch buffer.
Change-Id: Ifb2235507826f9ef1110dd9dbaf88045d8979e7c
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463337
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>
When connecting to an iscsi url that effectively does not exist we free
the data structures of the very context we are processing. This results in a
use after free and subsequently in a crash.
Signed-off-by: Jeffry Molanus <jeffry.molanus@gmail.com>
Change-Id: I67cab1efb161bfa23fa1022e150661080d90b556
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/462614
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Jeffry Molanus <Jeffry.molanus@gmail.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>
- New files and updates to existing SPDK files to add the NVMf-FC transport.
- Depends on an existing low level driver library. This driver is not part of SPDK repository.
- Makefile updates to build FC transport (using CONFIG_FC)
- Update configure script for FC build.
- New FC unit test for FC-LS commands.
- Update unittest.sh to run FC unit test (when built).
Signed-off-by: John Barnard <john.barnard@broadcom.com>
Signed-off-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Change-Id: If31d4d25feab76c2dbe90a7faf71d465c2c3a354
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/450077
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>
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>
This patch adds measurement of time request spends from the moment it
was polled till completion.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I1fcda68735f2210c5365dd06f26c10162e4ddf33
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/445291
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 patch adds statistics for pending state in NVMf RDMA subsytem
which may help to detect lack of resources and adjust configuration
correctly.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I9560d931c0dfb469659be42e13b8302c52912420
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452300
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
RDMA polling statistics: number of polls and number of completion
entries returned.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: Iabcf2cb6f6a35f595b89b58cdfcd177a637dda13
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/445289
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 patch adds transport part to nvmf_get_stats RPC method and basic
infrastructure to report NVMf transport specific statistics.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: Ie83b34f4ed932dd5f6d6e37897cf45228114bd88
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/452299
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>
If available, if not use the bdev unique name. This results in
a much friendlier comp_bdev naem for the user. For example,
now it would look something like this: COMP_lvs0/lvs as opposed
to like this COMP_0b149b31-b66b-4cf7-ab39-a55b50788cd1
Change-Id: I319a141221ed8880edcec930ca5be9f256b105e0
Signed-off-by: paul luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/463246
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>
