For now we don't provide #ifdef to users, SPDK_CACHE_LINE_SIZE is defined with 64 bytes.
Change-Id: I8ddb01a382008a3e333967650200c5aa8b93e09e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477385
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Operations to queue iSCSI tasks are in iscsi.c and conn.c and cross
references due to this separation makes us difficult to create unit
tests.
This and subsequent patches will try to disentangle cross references
by moving some functions from iscsi.c to conn.c.
This patch moves spdk_iscsi_conn_handle_queued_datain_tasks() from
iscsi.c to conn.c. For unit tests, we don't add anything new in
this patch and just create necessary simple stubs. After code
movement, new unit tests will be added.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: If5b8501a1ef7ea53682a3437c7eb2375aa52ee3b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477416
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Subsequent patches will move a few functions which call
iscsi_queue_task() from iscsi.c to conn.c. This patch is a
preparation to it.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Id136b5d3bf76a9894162115dd0d57d997178c869
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477415
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Move spdk_iscsi_conn_read/readv_data() down closer to the functions
which calls spdk_sock_writev().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ie7ef649e8681efac48adcfb2da1f745660f71782
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477411
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Move iscsi_get_pdu_length() down closer to the caller.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Iacbe43d39f5b23c47a65631ae8004fcfb489e2cc
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477410
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
TCP_REQUEST_STATE_NEW is already set in spdk_nvmf_tcp_req_get.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Ia835f3763cd74ef9b504901c719d9954317f49af
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476164
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This eliminates the flushing logic, simplifying the tcp
transport.
This also happens to greatly improve performance, especially
on random read tests. The batching done in spdk_sock_writev_async seems
to be more effectively than the previous batching logic in the tcp
transport.
Change-Id: Id980ac6073e380dc75f95df3f69cb224f50fb01b
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470532
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add an additional queue for requests that have been sent on the network
but aren't complete yet. As of this patch, the code
is still calling writev with no flags in the POSIX layer, so it completes
synchronously. That means requests pass through this new pending list
only very briefly inside of one function.
Change-Id: Iaab6efc118a6d5fe9589199515eb3a7293db4b8e
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/471768
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Or Gerlitz <gerlitz.or@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add spdk_sock_writev_async for performing asynchronous writes to
sockets. The user of this call is responsible for allocating their own
spdk_sock_request structures to pass to this call.
spdk_sock_writev_async will not return EAGAIN and will instead leave the
requests queued until they are fully sent or aborted due to socket
error.
Change-Id: Idf3239e65d26a3024e578122c23e4fb8f95e241b
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470523
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: 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>
RDMA qpair might be destroyed by defunct timer, so it can have
active recv elements in incoming_queue. This queue is cleaned
incorrectly, so recv element for the destroyed qpair still may
be presented in the queue and be processed later. That leads
to undefined behaviour.
Fixes#1086
Change-Id: Ieae186b2d2dce4ec88ab886b26165f6ef98e8d05
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477957
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
mapping length is initialized with 0 and spdk_vtophys() returns
min(*mapping_length, cur_size) or 0. So length -= mapping_length has no
effect and req will be failed when nseg reaches NVME_MAX_SGL_DESCRIPTORS
Initialize mapping_length = request length
Change-Id: I9082866b7f8055d99fa6930a78335b3b0fdf9b2b
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477575
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Add an API so that the user can enable/disable the bdev IO
timeout. Also, add the bdev io timeout handling callback.
So it means to let the upper user determine how to handle
the IO timeout scenario reset the device or abort the IO.
Change-Id: I9c7138ca46c74c045b687adab59a18d6bccc4996
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/469228
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This function is already declared in the public header which this
function includes.
Change-Id: Iff3f85dc166b7bd4d949b9b099a6bf05dec7dec8
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477868
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
This is a useful utility function.
The end goal of this patch series is to create a python utility that can
be called upon to dump information about DPDK allocated memory in a
human readable way.
Change-Id: I18978732c9decbb39dce5b5151f5eff6b59f6591
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477510
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>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
SPDK has swithed to DPDK rte_vhost library since 19.04 release, the internal
rte_vhost library is not recommended any more, but we still would like to
use it for vhost_nvme until we can switch to the new NVMe virtualization
solution. Fix the compilation issue here, the internal SPDK rte_vhost
library can still work well with vhost blk/scsi/nvme target for now.
Change-Id: I665ccee3e652c318b414dcdc20aab6cdbf7fac3d
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477231
Reviewed-by: Vitaliy Mysak <vitaliy.mysak@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
VIRTIO_F_VERSION_1 is defined in system header file <linux/virtio_config.h> and
SPDK already included a copy of it.
Change-Id: I88d5206c7b0984382b3fed1d062d971a014a85a8
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477224
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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: Vitaliy Mysak <vitaliy.mysak@intel.com>
VIRTIO_BLK_F_* feature bits is defined by virtio_blk specification and
SPDK already has a copy of the system file header <linux/virtio_blk.h>
in the header directory, so just remove it.
Change-Id: I672335218abc8375610a368d84b1a005463a741a
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477223
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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: Vitaliy Mysak <vitaliy.mysak@intel.com>
We defined vhost user protocol in vhost_user.h file, so just
remove it to where it should belong to.
Change-Id: I3e9c6eece1db2a8eb739254608816ef9a4dd9993
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477222
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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>
Request can be freed by transport_req_complete. In such case req
or req->cmd dereference might result in heap-use-after-free.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I2280d3978f1f183a250828aab7d2ca49ef1800ec
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476929
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Some version of firmware on Intel SSDs exhibit poor performance
with an IO queue size of 256 (the SPDK default). Add a quirk to
account for this - and use the same size as the Linux driver (1024)
instead.
If the user specified a controller option to explicitly set the
io_queue_size different than the default, the quirk will not take
effect. Note that users can still override the io_queue_size when
creating the qpair - this only changes the ctrlr's default
io_queue_size.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iff0f7ab9856fc18f226f7171e5109bb8fe430969
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/432227
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>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Previously, invalid io_flags would results in -ENOMEM being
returned to the user which was incorrect.
Change-Id: I53dd0fa8684cb36f3d124baa92244e2ed30e2527
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476938
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
This patch adds tracking of the memory regions allocated from the
hotplug area reserved by a hotplug-capable device.
Fixes issue #1055.
Change-Id: Id4e0939ece181549bdbf1282f3cc77529c935506
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476163
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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>
The following scenario might occur when nvmf_tgt is stopped:
1. nvmf_tgt receives SIGINT, changes state to NVMF_TGT_FINI_STOP_SUBSYSTEMS
2. In this state nvmf_tgt stops all subsystems and disconnects associated qpairs
3. In the case of RDMA qpair, its state will be changed to IBV_QPS_ERR.
Once qpair changes the state to IBV_QPS_ERR, RDMA device generates
LAST_WQE_REACHED event when there are no more WQE that can be sonsumed
from the SRQ by this qpair.
4. When all subsystems are stopped, some of qpair may still be alive since they
haven't received LAST_WQE_REACHED event yet.
5. nvmf_tgt stops all poll groups and forcefully destroyes any qpairs linked to them.
6. At this moment LAST_WQE_REACHED event might be generated and received in another thread.
Handler of this event sends a message with a pointer to qpair. The qpair itself may already
be destroyed.
7. Thread that owned qpair receives a message (LAST_WQE_REACHED) with a pointer to alredy destroyed qpair and
destroyes it for the second time when all pointer are invalid.
ibv events related to qpair should be handled by the thread that
owns this qpair. This commit adds a new structure that describes
ibv event, helper functions for sending the event and a list
of events per rdma qpair; add syncronization for LAST_WQE_REACHED event
Fixes#1075
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Change-Id: I22bff89741708df2518760934ecb4e33fad49473
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476712
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Seth Howell <seth.howell@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
The macro DMIN32/64() are local in iscsi.c. Replace them by the
generic macro spdk_min() will improve the portability.
Replace it in test/unit/lib/iscsi/conn.c/conn_ut.c together.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I5f6992b3dc091cd748b4e138810fb01761a1ab24
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477202
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This is a preparation to replace the private macro DMIN32/64() to the
generic macro spdk_min(). The corresponding input variables of
struct spdk_scsi_task are uint32_t, and so uint32_t are usable.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Icbf8ba1dfa0b170635a2852e6f0e9d841da9741b
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477201
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_iscsi_send_nopin() is called only in a single place of conn.c.
So change it to private in conn.c. Additionally, iscsi_conn_send_nopin()
may be a little more meaningful and rename to it.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1fe70e0468e1dd43468492b8b3d359c094130ed7
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477190
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_iscsi_is_deferred_free_pdu() is called only in a single
place of conn.c. So change it to private in conn.c. Additionally,
iscsi_is_free_pdu()_deferred() may be a little more meaningful
and rename to it.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic1e3d7ff435c454f40e81f9a4f90fe76589ec7b2
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477189
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_iscsi_task_mgmt_cpl()'s function body is in conn.c, and so move
its declaration from iscsi.h to conn.h.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Idd22c360ce1d3d464bf782e21348425ae7debdd8
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477188
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
spdk_iscsi_task_cpl()'s function body is in conn.c, and so move
its declaration from iscsi.h to conn.h.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I7bc6feecd5a2ff698c8e30e7aab547ada398c44f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477187
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
We changed read I/O submission to stop splitting when detecting LUN
hotplug very recently. We can do the same refinement for read I/O
abortion.
In _iscsi_conn_abort_queued_datain_task(), set all remaining length
to the new task and complete it immediately. We keep the code to
process the case that queued_datain_task completed but is still in
queue, but we can change its if condition to assert.
Simplify the corresponding unit tests accordingly, and set
task->scsi.transfer_len in abort_queued_datain_tasks_test() to
exercise the changed paths.
In iscsi_pdu_payload_po_scsi_read(), if task->scsi.transfer_len is not
larger than SPDK_BDEV_LARGE_BUF_MAX_SIZE, no minimum calculation is
necessary and we can substitute task->scsi.transfer_len to
task->scsi.length simply. This change is too small to be an independent
patch and is done together.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Iea93e51b103eae141a007a0abdaf13cbe6d5287f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476984
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>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
The hardware sgl format can describe large contiguous
buffers using just a single element, so it's more
efficient that a prp list even for a single memory
segment. Always use the sgl format.
Change-Id: I9c62582829f0d64dcd1babdbc48930ddb4d9e626
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475542
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: 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>
opts->env_context could have more options specified
than just --legacy-mem. So strcmp() is not a valid
comparison operator - we need to use strstr instead.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie4c8cbcbe7c141693a07a11648d6673ec8c012e5
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477087
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
For TCMalloc regions which we register with spdk at runtime in the MMapHook, we
need to ensure that SPDK doesn't do any allocations in that path otherwise we
will hit a livelock situation. MmapHook is invoked when TCMalloc is out of free
memory and needs to get more memory from the system, for the hugepage case it
gets via mmap.
In the current code, we could end up calling malloc in the spdk_mem_register
call via the following call path.
spdk_mem_register -> spdk_mem_map_set_translation -> spdk_mem_map_get_map_1gb
To avoid this livelock situation we call rte_malloc instead which shouldn't
invoke the system allocator. Note that in try_expand_heap_primary() which is
invoked in the rte_malloc code path, we can still call malloc, so we need to
only use this when dynamic memory allocation is disabled via --legacy-mem.
It is possible in the future we could work around even this limitation,
but for now this implementation will be much simpler.
Have verified this change fixes the livelock condition which I was hitting in
my setup without this fix.
Change-Id: I69d0813a70da1f26f8c4d9d8895e406c026be18b
Signed-off-by: Alok N Kataria <alok.kataria@nutanix.com>
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475943
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>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
We will use this in a future patch to determine whether it's safe
to use DPDK allocated memory when allocating new 1gb page entries.
We could use it in this patch to decide whether or not to register
the memory hotplug handler, but there's really no harm registering
it even when it's not needed.
Ideally DPDK would provide some kind of API to query how DPDK was
configured. In the normal case we know whether legacy-mem was
specified, but if users initialize DPDK themselves and then call
spdk_env_dpdk_post_init(), we won't know if legacy-mem was specified.
So in that case, we will just assume that it wasn't specified.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ied0e5ff777c8ee651043f46a37ce62e44bfcc5fe
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/477086
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>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: SPDK CI Jenkins <sys_sgci@intel.com>
'delay_pcie_doorbel' parameter in 'spdk_nvme_io_qpair_opts' structure
was renamed to 'delay_cmd_submit' to make it suitable for every
transport. Old name is also kept for backward compatibility.
Signed-off-by: Evgeniy Kochetov <evgeniik@mellanox.com>
Signed-off-by: Sasha Kotchubievsky <sashakot@mellanox.com>
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I09ef8028133c4a3d4a5bbc5329ced1f065bcaa46
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475305
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
This patch attempts to solve naming conflict between
CUSE devices created by different SPDK instances.
Each NVMe device is enumerated by SPDK process from 0
up to 127. When process attempts to start cuse device
tries to set exclusive lock on temporary file
"/tmp/spdk_nvme_cuse_lock_<index>" and keep it until
device will be stopped.
If setting lock fails, index is incremented.
It prevents to use the same controller index from
few SPDK instances.
Change-Id: If744ac23f813bd992efb80ae2b61a1acefb5054c
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474829
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This patch removes posibility to set cuse device path. Instead
"/dev/spdk/nvme*" path is used.
Change-Id: I7c3087772a3661eebe03fce21356c35cc8204b49
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474598
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
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: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
To properly return the error case when adding an existing
IP or delete a not existed IP. Proper test case is also
updated.
This is also to fix below issue:
https://github.com/spdk/spdk/issues/992
Change-Id: Ia4d3af8cc86d9bdb66b18a165510cd08f9bfa555
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476543
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>
NVMF_REQ_MAX_BUFFERS is defined in nvmf_internal.h and not used in rdma
source code.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I50f24c8fc4ea773378418f7803361d8592f961ae
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475529
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Alexey Marchuk <alexeymar@mellanox.com>
It is valuable to have more detail status instead
SPDK_NVME_SC_INTERNAL_DEVICE_ERROR.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: Ifd003b490a7ae9af017645c97636ceaf2f93d4b0
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476634
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
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>
The nvme_transport_ctrlr_scan() may return failure while there are
multiple controllers, so the probe context's init_ctrlrs list may
not null for this case, so when free the probe context, let's ensure
there is no controller in the init_ctrlrs list. Also added a UT to
cover this case.
Fix issue #1095.
Change-Id: I4d9a10ad73cf00bbe159edd1f5b919797333feb6
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476969
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
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>
Previously, when additional resets were submitted while a reset was in
progress, those resets were queued and then executed serially as part of
executing the original reset. Doing multiple resets on a bdev in quick
succession is not useful if the first reset succeeds and is very
unlikely to be useful in the negative case. Instead, we should batch
resets and complete them all at once when the current reset succeeds or
fails.
Change-Id: If10e0f37526860eaeeb41a8803d6298a3eff3212
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474599
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>
Community-CI: Broadcom SPDK FC-NVMe CI <spdk-ci.pdl@broadcom.com>
nvme_qpair_get_state fits more closely with the semantics in other
modules.
Change-Id: I6ea8e02abe27253d9b4d779a43ac1963be56356a
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476920
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: Alexey Marchuk <alexeymar@mellanox.com>
The qpair state transport_qpair_is_failed is actually equivalent to
NVME_QPAIR_IS_CONNECTED in the qpair state machine.
There are a couple of places where we check against
transport_qp_is_failed and then immediately check to see if we are in
the connected state. If we are failed, or we are not in the connected
state we return the same value to the calling function.
Since the checks for transport_qpair_is_failed are not necessary, they
can be removed. As a result, there is no need to keep track of it and it
can be removed from the qpair structure.
Change-Id: I4aef5d20eb267bfd6118e5d1d088df05574d9ffd
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475802
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>
The address was padded before the size of requested area was checked,
which for areas bigger than available space resulted in unnecessarily
padding the remaining area.
Change-Id: Ic52a6e30c1cbe2526ac7f3effc20279cd91cadbc
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475832
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
CPU mask parameter was deprecated in v19.10.
If we remove the related code from the portal parser, we will be
able to use the parser for the iSCSI initiator to know the target
portal in iSCSI fuzz testing. So let's remove the parser and its
test code here.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I72ad4364323abda0f0ed10519b56244cd0c7612e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476830
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>
Add the tsc trace for the parent bdev_io.
Change-Id: Idf5c5df3953fa1697c384a4f83b591584dd30a80
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476588
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>
The reset cmd and parent of splited IO would not go into
the _bdev_io_submit() so they don't record the submitted
time. The timeout IO check needs the submitted time so we
add them in this patch.
Change-Id: Ic7217171ec878d4a6dfa80d6635957ca0186928b
Signed-off-by: Jin Yu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476050
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>
Add a link to the bdev channel for linking all the bdev
IOs that were submitted to this channel so that we can
monitor each IO's consuming-time.
Change-Id: I1e425b2059f20fd7b158eb3d6b023ce8629e7a30
Signed-off-by: JinYu <jin.yu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/469227
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>
There are some new features defined in NVMe 1.4 specification also need
the data buffer, but for now we only listed 3 features, we will check
the data length field anyway, so this piece of code can be removed.
Change-Id: I22204cf53077073434bb0ed73693bcf72883f084
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475953
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>
Reservation Acquire Action(rqcqa) is used for reservation acquire command, since
we have replaced cdw10 with specific commands fields, so split the two cases to
make the code looks more clear.
Change-Id: Ib2a2458dd0fe8345f1c33341dab1c4fe3b82829f
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476838
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>
In iSCSI library, macro MATCH_DIGEST_WORD() and MAKE_DIGEST_WORD()
have been used instead of match_digest_word() and make_digest_word().
The latter has not been used for a long time. Hence remove them.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0fdb5c24a120a08fe06f825ee5e6c24ba64c0edf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476416
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>
This log has been observed often after recent refinements but
this log does not mean any error. So use SPDK_DEBUGLOG instead
of SPDK_ERRLOG here.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I3aa39ab3773ae83586c99699fd2473ca02c35eb9
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475182
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: Jim Harris <james.r.harris@intel.com>
This patch series orders login related functions top down in iscsi.c.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I23ea7101030c6d2d3fbccba878bdf77d89b814cf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476415
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>
This patch series orders login related functions top down in iscsi.c.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I39431b213216b1e4e677f80de8c14c40dd7da150
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476414
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>
This patch series orders login related functions top down in iscsi.c.
Other than code movement, fix a typo.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ib55fe86295f47c1c86bb99d5e3a6862f508bfcfa
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476413
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>
This patch series orders login related functions top down in iscsi.c.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I82fe06e3488a1c623c5fc875cf797790ee4ea48f
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476412
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>
This patch series orders login related functions top down in iscsi.c.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I2f7dd681ca430c810dac6fcea122f84a142152d7
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476411
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>
This patch series orders login related functions top down in iscsi.c.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I8e123177124b4526e9bbe9001293c6f668b9c3bb
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476410
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>
conn->data_out_cnt does not control anything now but adding assert
for conn->data_out_cnt will be helpful at least to ensure that the
current SPDK works correctly.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I31ee90769ce0555e64bd41c283e8b437326efebf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476409
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: Ziye Yang <ziye.yang@intel.com>
iscsi_conn_free_task() is used only when exiting connection now.
Hence we can remove the parameter lun and simplify the function
and its unit tests.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I6e7bf09672edca1f70c042ac58f098114d71ec78
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476115
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: Ziye Yang <ziye.yang@intel.com>
We ensure current pending tasks are aborted and no new task or PDU
is pending after LUN is unplugged now. Then we have to ensure all
task completions and aborts are sent to initiator.
However, we had removed pending tasks without any notification to
initiator. For this implementation, we are not surprised if initiator
waits for in-flight commands for a long time.
Due to recent refinements of LUN hotplug, we can wait safely that
all existing tasks complete or abort.
In _iscsi_conn_hotremove_lun(), after aborting R2Ts, start a poller
to wait until all PDUs for the LUN are flushed. Then close the LUN.
We can safely free deferred PDUs for the LUN because they are
already flushed.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I869b79a7c93d2e8a4a1577cc20d3b466548dfaaa
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476033
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>
We want to use poller for LUN hotplug in iSCSI library. According to
the naming used in SCSI library, rename iscsi_conn_remove_lun by
iscsi_conn_hotremove_lun. The next patch will iscsi_conn_remove_lun.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I08bb8c92db23ac3adcde4f39c0e812f3d97430d3
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476114
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: Ziye Yang <ziye.yang@intel.com>
Previously we had allocated temporary context, struct
_iscsi_conn_remove_ctx to process LUN hotplug. However, we
had to consider out of memory during connection is active,
and we could not use any poller to process LUN hotplug because
LUN hotplug may conflict with connection exit, and this
possible conflict made us very difficult to use poller for LUN
hotplug. Introducing struct spdk_iscsi_lun will resolve all
these issues.
Allocate struct spdk_iscsi_lun per LUN and store connection,
LUN, and LUN's descriptor into the struct. Then use the struct
for LUN hotplug and free the struct after LUN is closed when
LUN is removed or connection exits.
struct spdk_iscsi_lun is similar with struct spdk_nvmf_ns in
NVMf library.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ice26330f3948070c96d2fb53b94941be3b467079
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476113
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>
This is a preparation to the subsequent patches.
Subsequent patches will introduce struct spdk_iscsi_lun
to refine LUN hotplug process and fix critical issues of it.
This change will make us easy to introduce struct spdk_iscsi_lun.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I75db59d88bb09ee2ea94e8c02e0e87003352850c
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476112
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: Ziye Yang <ziye.yang@intel.com>
In _iscsi_conn_free_tasks(), we had parsed conn->write_pdu_list
and then parsed conn->queued_datain_tasks. However when we parsed
conn->write_pdu_list, if there was any task in conn->queued_datain_tasks,
some PDUs were inserted conn->write_pdu_list. Hence after parsing
conn->write_pdu_list, new PDUs were in conn->write_pdu_list as orphan.
Then orphaned PDUs were freed later but LUN was already freed and
critical failure occurred.
This patch swaps the order of conn->queued_datain_tasks and
conn->write_pdu_list, and add comment to explain the change.
Additionally, this patch adds unit test which fails if it runs
without this fix.
Fixes issue #1030.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Icb0ffbbbac70792a62939dc55a69df05d2ab9128
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475453
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>
When a iSCSI write is large and split, if LUN is removed between
creating and submitting the last subtask, spdk_clear_all_transfer_task()
completes the primary task and then process_non_read_task_completion()
tries to complete the primary task.
This is the double free case, and the later have to be skipped.
We add a flag is_r2t_active to struct spdk_iscsi_task and use it to
check the duplication. We may be able to use primary's initiator task tag
(ITT) instead but we can not rely on ITT because it is set by the initiator.
We clear is_r2t_active even when primary is removed from
conn->queued_r2t_tasks but it will be no harm.
Fixes the issue #1064.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ia6511bd7adaa8fcb9a07bc40d498e8ee0b7a7ccf
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475044
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>
In process_non_read_task_completion(), when the current I/O is
not split, we have to call only spdk_iscsi_task_response().
The next patch will fix the github issue by changing the path
executed when the current I/O is split.
Hence to make the fix easier, this patch separates split case and
non-split case.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic1603609f760c4bdd41272ba6146e260f668b059
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475043
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: Ben Walker <benjamin.walker@intel.com>
The type of pdu deferred to be free only have
two types, R2T or DATA_IN. And the two types of pdus
are all assoicated a task, so updateing both the code and unit test case.
Also for all pdu free, we should use spdk_iscsi_conn_free function since
for normal pdu free, we all use this function.
PS: I also tested the calsoft local, it does not trigger the assert.
Fixes#1074.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0524965baf5349a100210ef717aedaa5f8ff105e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475657
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This revert commit 6e4e85dcef.
Causing intermittent build pool failures. We had different operations
in remove_acked_pdu() and _iscsi_conn_free_tasks(). The patch tried
to unify them but was wrong. The operation in removed_acked_pdu()
was correct. This patch restores it. The next patch will fix the
operation in _iscsi_conn_free_tasks().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ia398fd295769b786ba4777cc9f7df6e134f15e48
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475791
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Purpose: Simply the code, doing data_cnt_in every where
will make the code diffcult to maintain. If we put the
management in iscsi task get and free related function, then
the code will be easy to be read and easy to maintain.
Change-Id: Ib9af067326630657877a94afc2eb0db28f5d5fd1
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474914
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Purpose: Do not let the primary task do the I/O if it is
splitted into subtasks. This will make the code simplier,
when all the sub I/Os are finished, we can free the
primary task.
Update the corresponding unit test also.
As a result of this change, when read I/O is split into subtasks,
the primary task uses only some of its data. Hence separate
iscsi_pdu_payload_op_scsi_read() into split and non-split
case explicitly.
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I9bbe4b8dd92a2996f35ad810b33676e34670c77e
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/473532
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The code is exactly the same - we can just have
spdk_mem_map_clear_translation call spdk_mem_map_set_translation
with translation = map->default_translation.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I6a2ce39b0397be9d29b1a4c1cdfba15025afba7a
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/476529
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>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
When serializing extents, run-length encoding is supposed to
1) RLE all sequential LBAs
2) RLE zero LBAs (unallocated)
There is one special case, with sequential LBAs that start
with 0 LBA. This is RLE as 1) case, but results in descriptor
matching case 2). Which causes loss of allocated clusters.
This requires following conditions to be met:
- blobstore has just a single cluster reserved for MD
- blob is thin provisioned
- first allocation occurs on cluster_num=1
For last part to be true, very first write for blob has to be
issued to LBA between cluster_size and 2*cluster_size.
Causing allocation of second cluster in blobstore and assiging
it LBA equal to number of LBAs per cluster.
To fix this, case 1) disallows to RLE zeroes.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I136282407966310c882ca97c960e9a71c442c469
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475494
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 prep for storing a clear_method in the blob metadata:
* Set the default to DEFAULT and let the switch statement choose
UNMAP
* Use switch statements to make it clearer which method we are
using and why. (ie previously we set the default to UNMAP and
then had an UNMAP || DEFAULT condition to choose UNMAP.
Later in the patch series it will become clearer why this makes
sense.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: I216cb97fd8eaa772437a36c2c7a47e66618bbfbd
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/472202
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>
Part of a longer series to make clear_method a lot easier to use and
understand. We're going to store this parm now in the blob metadata
instead of lvol xattr. That way it will be usable without having
to remove the option from lvol_create or adding to lvol_delete
(which was what the last version of this patch did for those keeping
score at home)
Note: the current use of this xattr is broken anyway, its missing
plumbing to use it on a subsequent delete so this doesn't break
anything new.
Signed-off-by: paul luse <paul.e.luse@intel.com>
Change-Id: Ie77a25ecdc52fd4aef6c090756c6cafe79486bc8
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/470410
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Now all registered producers should be stopped (unregistered) before
NVMe detach, otherwise NVMe controller cannot be safely detached.
This patch allows to stop all not unregistered io producers before
NVMe detach:
1. Callback to the "struct nvme_io_msg_producer" to stop producer
started on selected controller.
2. On nvme_io_msg_ctrlr_detach() if there's some unregistered producers,
stop all before freeing resources.
This approach also fixes issue with not to stop CUSE device when
NVMe controller is detached without unregistering producer (github
issue #1033).
Fixes#1033
Change-Id: Ia1ffef566bb745edb55c54d6786ea481a35bbefd
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/474273
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>
Destroy in poll_group_add results in heap-use-after-free because
upper layer calls qpair_fini in case poll_group_add returns
error.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I3e921a21b7ab5f7c15c80bc5919cb97cbda0b5d2
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475858
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
If DPDK is using virtual addresses, we should use virtual addresses.
If DPDK is using physical addresses, we should use physical addresses.
This way there can never be a conflict and everything is consistent.
Change-Id: Ie4b0e885e9a52dd6cbc81000a87908102a9771cb
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475928
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>
Avoid the running converting timeout from sec to TSC, thus
make the behaviour same as last_nopin, i.e., initialize
when constructing the connection.
Change-Id: Ibc120fed24d2208cab9ae8a876856e9d05363075
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475711
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: Changpeng Liu <changpeng.liu@intel.com>
DPDK by default guesses that it should be using iova-mode=va
so that it can support running as an unprivileged user. However,
some systems (especially virtual machines) don't have an IOMMU capable
of handling the full virtual address space and DPDK doesn't
currently catch that. Add a check in SPDK and force iova-mode=pa
here.
Change-Id: Ib3a5691a584190feaab4b9064b5a500e361328f2
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475149
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>
When this was originally implemented, it only accessed
*sock once or twice. As more stuff is added, it becomes
worthwhile to dereference the first level of **sock.
Change-Id: Ie31bb0210008f6341b071ba472aaedf897fa459a
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475310
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>
This patch uses lowlevel fuse functions to process messages to
eliminate the need to use signals to interrupt blocking read
operation in fuse_session_loop().
Fixes#1032
Change-Id: Ie9c9ea76cc135c383f5757864aa2d84ac9eb3da3
Signed-off-by: Tomasz Kulasek <tomaszx.kulasek@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/473233
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In DPDK, the ppc iommu support does not currently allow for
iova-mode=va, but DPDK doesn't detect ppc and so still attempts
to guess iova-mode=va in some modes. Force iova-mode=pa from
SPDK to fix this.
Change-Id: I6a1ee25ab74873826ac211c3e0dfdf54afc74502
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/475148
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>
Reviewed-by: JinYu <jin.yu@intel.com>
Added states for keeping track on which reloc
queue band actually is.
Change-Id: Ib05ac4e925002728ddfed3195891f5328eebb0d0
Signed-off-by: Wojciech Malikowski <wojciech.malikowski@intel.com>
Reviewed-on: https://review.gerrithub.io/c/spdk/spdk/+/465072
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
