QIDs need to be validated for both I/O queues and admin
queues, so move the checking to a common path.
Change-Id: Icb2009b3fe4d70f146b741c046812d9df62cc2fb
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/413859
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This one is actually a bit tough to deduce in the specification.
The NVMe-oF spec says that QID errors detected in the RDMA
transport shall return an RDMA-specific error indicating
the problem. However, our code doesn't detect the error in
RDMA-specific code, and it isn't clear if the language is
a "must" or a "should".
The NVMe specification does clearly indicate what error
to return on invalid QID in response to a Create
I/O Queue Pair command. For now, return that while
we game plan whether we need to call into the RDMA
transport to correctly report this error.
Change-Id: I7faf37bad9b9202bc50a906214a51c17e4808fc0
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/413858
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When the shadow doorbell feature is enabled, no actual MMIO takes place
when a shadow update is sufficient; tighten the bounds of the updates to
g_thread_mmio_ctrlr in the two doorbell update locations so that we only
need to touch the thread-local variable when actual MMIO access is
required.
Change-Id: Ida974bec33f56cbb9f7d3611f483c6975ec773ab
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/413856
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This patch does a little cleanup of SPDK SCSI layer APIs.
spdk_scsi_task_alloc_data() is only used in spdk_scsi_task_scatter_data().
spck_scsi_task_free_data() is only used in spdk_scsi_task_put().
The latter was called in UT code but this can be removed without any
degradation.
SPDK SCSI layer is relatively matured and these will not be used
out of lib/scsi/task.c.
Additionally memory leak detected by ASAN is fixed in this patch.
Change-Id: I8eff7b4dbfc307c211087734649a9b9b10555f8d
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/413872
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Method "remove_vhost_scsi_dev" changes to "remove_vhost_scsi_target".
Change-Id: I1e368c7cf2b16accd3431100e384c1d868e04a47
Signed-off-by: Chen Wang <chenx.wang@intel.com>
Reviewed-on: https://review.gerrithub.io/413873
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Suggested-by: Ziye Yang
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I6ca7f08aed2e12c59518269c7b277895b4817bd8
Reviewed-on: https://review.gerrithub.io/413844
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
This is consistent with the Open Channel naming we are using elsewhere.
Change-Id: Ib088359bed29a958f8b50e41cf34143a23429f54
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/413840
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
int16_t is not used in SPDK yet. Hence add decode function only
for uint16_t.
Change-Id: I86f86a6768d95896d1840370f2d176c770c41418
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/413711
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Suggested-by: Jonas Pfefferle
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I33cf41d39a4546fbb5f44fccc1380c731d0224af
Reviewed-on: https://review.gerrithub.io/413705
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jonas Pfefferle <pepperjo@japf.ch>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Also, as we are here, switch to new spdk_json_write_named_* API in
bdev_iscsi_dump_info_json()
Change-Id: Ib29466eb5c6d6496dd1d8efc1f2064577bf56272
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/406635
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Also convert bdev_iscsi_set_module_init() into new callback.
Change-Id: I439369180e1e10eab5dee6aba27ffb0489ea2550
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/410095
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Fio commit d3b07186b1d4c7c1d9adc1306407458ce41ad048 changed return
type of the ioengine->queue method to a new enumeration type
fio_q_status. Add preprocessor checks to ensure SPDK fio_plugin builds
successfully with both current and earlier fio ioengine API version.
Change-Id: Ie9eb818473c6453624d6cfe51a19c6bf0d17e334
Signed-off-by: Andrey Kuzmin <andrey.v.kuzmin@gmail.com>
Reviewed-on: https://review.gerrithub.io/413749
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Adds option to disable pci access in spdk_vhost_run function.
This is particularly required in vhost live migration tc2 & tc3
Change-Id: I5ec215d316b97dbabd1297856d8f2dfbf7b9d3cb
Signed-off-by: Pawel Niedzwiecki <pawelx.niedzwiecki@intel.com>
Reviewed-on: https://review.gerrithub.io/413658
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
NVMf cnx acceptor poller is changed to check the asynchronous events
from the RDMA devices.
RDMA async events are polled together with RDMA CM events; the file
descriptors are combined into a poll fd array and processed in a single
poll syscall.
The errors handler is an empty placeholder for this patch, it just
prints the kind of event read from the IB device context.
The work for implementing event handling is left for later.
Signed-off-by: Philipp Skadorov <philipp.skadorov@wdc.com>
Change-Id: Ib167990651b585090aceef1404a88d431a910226
Reviewed-on: https://review.gerrithub.io/412540
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I1cac468c58d84ed03bb646b8485e92fdb533049e
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/413254
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
This begins the process of hiding data members in
spdk_bdev_io that don't need to be accessed from
within bdev modules.
One strategy would be to implement accessors for
every data member in the structure. However, that
may have negative performance effects. Instead,
create a new internal structure within the old
structure. This new structure will still be visible
for now, but at least makes clear which members
are accessible and which are not.
This patch shifts one data member to the new structure
as an example.
Change-Id: I68525db995325732fe9f5fc3f45b06920948309b
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/412298
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I841d7b47bc85498abb608944587e7b7585138263
Signed-off-by: Xiaodong Liu <xiaodong.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/411588
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I862a7eb89d6357c85285d2a6cd90ad9f633d2435
Reviewed-on: https://review.gerrithub.io/413285
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
This reduces submission+completion time by 10-15
core clocks per IO on an Intel Xeon Platinum
processor. Similar improvements should be seen
on other processors as well.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I3241ba53ef5f21a8eef930b523a951525922e6b8
Reviewed-on: https://review.gerrithub.io/413284
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
This variable will no longer be static in a future patch.
So make the name a bit more verbose since it will now
be visible outside of the module that defines it.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I5e720ef44aa5c5f38e0fe91de091a89b9970fcb7
Reviewed-on: https://review.gerrithub.io/413283
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
The subsystems created in this test only have one namespace, not 10;
wait for just nvme0n1, nvme1n1, etc. instead of trying to wait for
nvme0n1, nvme0n2, nvme0n3..., which will time out and fail.
Change-Id: I2d131e9b92c9c89a5dc8fe0cf9138b980f810233
Signed-off-by: xuhuagen <huagenx.xu@intel.com>
Reviewed-on: https://review.gerrithub.io/413056
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
The memset was zeroing a lot of bytes that get
initialized either later in this function or elsewhere
in the submission code path. Eliminating these
extra memsets saves a few nanoseconds of CPU overhead
in the NVMe submission path.
Note: one use of the cpl data member depended on
the nvme_allocate_request memset. Since this use
case is not in the primary I/O path, just memset it
in that specific location before using it.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ife483a4d9c24c033cc7d26d94ec1700905a936f4
Reviewed-on: https://review.gerrithub.io/413153
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Rather than storing nvme_payload::type explicitly, use the SGL reset
function pointer as an indicator: if reset_sgl_fn is non-NULL, then the
payload is an SGL type; otherwise it is a contiguous buffer type.
This eliminates the one-byte type member from struct nvme_payload,
making it an even 32 bytes instead of 33, allowing the removal of the
awkward packing inside struct nvme_request.
Change-Id: If2a32437a23fe14eb5287e096ac060067296f1dd
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/413175
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The definitions of these macros will change in an upcoming patch that
modifies the way nvme_payload is laid out.
Change-Id: Ic6edc18928542b07be7519a72bdbf6babbeb0131
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/413174
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This will simplify upcoming patches that change the way nvme_payload
stores its type.
Change-Id: Idf0a5b8dfd7d66a10f89254d2c5c54fee2968a43
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/413173
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The check_io function always calls spdk_get_ticks() -
so return that value to the work_fn for checking
when the test is complete, rather than have work_fn
call spdk_get_ticks() again. This eliminates an
extra rdtsc call that counts against the completion
path.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I5a4f52eb2bec70a119c9ea7368bc7f2e8f17155c
Reviewed-on: https://review.gerrithub.io/413152
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There is an existing unmap "workload" but it is not a
real workload - it's just used as a type of verify
function to check if an unmapped region later returns
zeroes. This has typically been true for Intel NVMe SSDs
but is not universal, so this functionality is not really
useful. It's not used anywhere in any of the tests in the
SPDK tree either.
So replace it with a real "unmap" workload that will
do sequential unmap operations on the underlying
bdev. A future enhancement could make these unmap
operations random across the bdev - but that can be
added later.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I99d7c2d455fb5c3c87712a5bbfd890d257630f88
Reviewed-on: https://review.gerrithub.io/413151
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
The Intel SDM clearly states that only an lfence (not
an mfence) is needed to ensure rdtsc is executed before
subsequent instructions. So change spdk_mb calls
to spdk_rmb to avoid overhead related to the unneeded
part of the fencing operation. This provides a more
accurate view of the overhead of both the SPDK and
kernel NVMe drivers.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Idaeb59db241cc1630e438e300559a02c4d59cd49
Reviewed-on: https://review.gerrithub.io/413148
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
These are needed in some cases on x86 where a compiler
barrier is not sufficient - for example, ensuring an
rdtsc instruction has executed before subsequent
instructions.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I339e2dd6138ccb11b1492e70f7c724976ef3038b
Reviewed-on: https://review.gerrithub.io/413145
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
This patch is a pair to the following
- subsystem/iscsi: Add set_iscsi_options RPC to set global params
Now options can be loaded from JSON config file.
Change-Id: Ifb68cddbb045d51fbaf8161ad59ede9d399e70cb
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/410874
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
An new RPC set_iscsi_options allocates and set options dynamically.
Initialization of iSCSI subsystem skips initialization of options
if it is already allocated.
To use and test this RPC easily, add python script too.
Change-Id: I71e252da6495a194ae9a1a9e3aaae4feb543487a
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/403624
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This got broken in 93cb4a31 [1]
[1] 93cb4a31: event/app: Refactor initialization of app environment
in spdk_app_start()
Change-Id: I9dca9694d50b9132761b57da192ecd1a774337ab
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/413168
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Although contiguous, memory regions may appear as separate
/proc/self/maps entries.
This patch brings support for DPDK 18.05.
Change-Id: I91eb8adc2c073a103f687320ec7b9dabe1e066b3
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/413167
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
These are trivial to abort: since we just hold on to aer_req until an
event is triggered, we just need to set aer_req back to NULL and
complete the request.
Change-Id: I74ffe7a227fcaf816c0c584fe5c82a940475687f
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/412881
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There is some overhead notified for the message processing,
instead to always send a message to the QoS thread, check
whether the current thread is QoS thread or not. If yes,
queue the IO immediately base on the QoS logic.
Change-Id: I9c1f93aeaf68c9b1a0282c3b690614413949d901
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/413028
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Combine request lookup and abort into a single operation. Keeping them
separate would result in duplicating a lot of logic for finding the
proper list from which to remove aborted requests.
This is still a no-op for now, but it paves the way for aborting
requests that are still queued in software.
Change-Id: If8f268521f2c9f93b413261d87e9f39e539813aa
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/412880
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reason: If we trigger the spdk_iscsi_conn_migration,
we may face the issue that the io channel of lun
is NULL(which means not allocated). So I think
that if we trigger the migration, we need to
stop the executing the further pdu of this conn
by the current polling group.
Also since the connection is triggered migration,
we should stop execution on the current core, and
let this connection be handled by next round.
Change-Id: I0ab89d79c976f3233890ae25cb7eac98de5e30ac
Signed-off-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-on: https://review.gerrithub.io/409984
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
