If ctrlr_loss_timeout_sec is set to -1, reconnect is tried repeatedly
indefinitely, and I/Os continue to be queued.
This patch adds another option fast_io_fail_timeout_sec, a flag
fast_io_fail_timedout to nvme_ctrlr.
If the time fast_io_fail_timeout_sec passed after starting reset,
set fast_io_fail_timedout to true not to use the path for I/O submission.
fast_io_fail_timeout_sec is initialized to zero as same as
ctrlr_loss_timeout_sec and reconnect_delay_sec.
The name of the parameter follows the famous DM-multipath, its fast_io_fail_tmo.
Change-Id: Ib870cf8e2fd29300c47f1df69617776f4e67bd8c
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10301
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Previously reconnect retry was not controlled and was repeated indefinitely.
This patch adds two options, ctrlr_loss_timeout_sec and reconnect_delay_sec,
to nvme_ctrlr and add reset_start_tsc, reconnect_is_delayed, and
reconnect_delay_timer to nvme_ctrlr to control reconnect retry.
Both of ctrlr_loss_timeout_sec and reconnect_delay_sec are initialized to
zero. This means reconnect is not throttled as we did before this patch.
A few more changes are added.
Change nvme_io_path_is_failed() to return false if reset is throttled
even if nvme_ctrlr is reseting or is to be reconnected.
spdk_nvme_ctrlr_reconnect_poll_async() may continue returning -EAGAIN
infinitely. To check out such exceptional case, use ctrlr_loss_timeout_sec.
Not only ctrlr reset but also non-multipath ctrlr failover is controlled.
So we need to include path failover into ctrlr reconnect.
When the active path is removed and switched to one of the alternative paths,
if ctrlr reconnect is scheduled, connecting to the alternative path is left
to the scheduled reconnect.
If reset or reconnect ctrlr is failed and the retry is scheduled,
switch the active path to one of alternative paths.
Restore unit test cases removed in the previous patches.
Change-Id: Idec636c4eced39eb47ff4ef6fde72d6fd9fe4f85
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10128
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Monica Kenguva <monica.kenguva@intel.com>
We noticed the difference between the SPDK 21.10 and the latest master
in a test.
The simplified scenario is as follows:
1. Start SPDK NVMe-oF target
2. Run bdevperf for the target with -f parameter to suppress exit
on failure.
3. Kill the target after I/O started.
With the SPDK 21.10, bdevperf retries failed I/Os and exits after
the test time is over.
With the latest SPDK master, bdevperf hungs and does not exit even
after the test time is over.
The cause was as follows:
reset ctrlr is repeated very quickly (once per 10ms by default) and hence
I/Os were queued infinitely because nvme_io_path_is_failed() returned
false if nvme_ctrlr is resetting.
We should queue I/O when nvme_ctrlr is resetting only if reset is throttoled
and fail-fast for the repeated failures is supported.
Hence in this patch, fix the degradation and remove the related unit
test cases.
Reported-by: Evgeniy Kochetov <evgeniik@nvidia.com>
Signed-off-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
Change-Id: I4047d42dc44488a05264c6a841d101a7c371358b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/11062
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We sort secondary trids to avoid using disconnected trids for failover.
However the sort had a bug.
This bug was found by running test/nvmf/host/multipath.sh in a loop.
Verify the fix by adding unit test.
Fixes#2300
Signed-off-by: Shuhei Matsumoto <shuheimatsumoto@gmail.com>
Change-Id: I22b0ede4d2ef98b786c3e0d1f5337a2d568ba56d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10921
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This patch adds the framework for a discovery
service in the bdev/nvme module.
Users can specify an IP/port of a discovery service.
The bdev/nvme module will connect to a discovery
controller, get the discovery log page, and then
register for AERs. It will connect to each
subsystem specified in the initial log page.
AER completions will trigger fetching the log
page again, at which point new subsystems will
be connected to, or removed subsystems will be
detached.
This patch does the following:
* Adds the new start_discovery RPC
* Connects to the discovery controller
* Gets the discovery log page
* Registers for AERs
* Detach from discovery controllers at shutdown
Subsequent patches in this series will:
* Connect to subsystems listed in discovery log page
* Detach from subsystems that were listed in earlier
discovery log pages but subsequently removed
* Add a stop_discovery RPC
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I54bfa896a48c5619676f156b5ea9f2d1f886c72f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10694
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <smatsumoto@nvidia.com>
The failover_in_progress flag is used to decide the return value of
bdev_nvme_failover().
bdev_nvme_delete() calls bdev_nvme_failover() with remove=true to remove
nvme_ctrlr->active_path_id. However bdev_nvme_failover() returns zero
if nvme_ctrlr->failover_in_progress is true. bdev_nvme_failover() may
return zero even if it does not remove nvme_ctrlr->active_path_id.
The following will be better.
bdev_nvme_failover() returns -EBUSY if nvme_ctrlr->resetting is true,
and the caller repeats calling bdev_nvme_failover() until the target trid
becomes alternative path or bdev_nvme_failover() returns zero.
To do that, the failover_in_progress flag is not necessary any more.
Removing the failover_in_progress will also simplify the following
patches to unify ctrlr reset and failover.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I57ab944beb1d06ea4def144c81c69705860de35f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10441
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
We had not checked the bit 0 of the Namespace Multipath I/O and
Namespace Sharing Capabilities (NMIC) field in the Identify Namespace
data structure.
If the bit 0 of the NMIC is zero, it is likely that namespaces are not
identical.
We should check if the value of the NMIC first, and do it in this patch.
Additionally, it is not usual if the bit 0 of the CMIC and the bit 0 of
the NMIC do not match. So in unit tests rename the parameter multi_ctrlr
by multipath for ut_attach_ctrlr() and use it for the value of the NMIC.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I6aa7cbcc99be2507dbf18930f7b585a9ea7d0f90
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10380
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.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: Aleksey Marchuk <alexeymar@mellanox.com>
Replace the spdk_nvme_ctrlr_reset_async() and spdk_nvme_reset_poll_async()
calls by the spdk_nvme_ctrlr_disconnect(), spdk_nvme_ctrlr_reconnect_async(),
and spdk_nvme_ctrlr_reconnect_poll_async() calls in a reset ctrlr
sequence.
spdk_nvme_ctrlr_disconnect() can fail if ctrlr is already resetting or
removed. But both cases are not possible. reset is controlled and the callback
to the hot remove is called when the ctrlr is hot removed. So we assume
spdk_nvme_ctrlr_disconnect() always succeed.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1299e198597b2a2110f80b9a868e2dae015682ee
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10092
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In the following patches, we want to retry reconnect if reconnect failed
in a reset ctrlr sequence but we want to delay the retry. While
we wait the delayed retry, we want to quiesce ctrlr completely.
As part of quiesce ctrlr operations, we want to pause adminq poller but
we need to do it on the nvme_ctrlr->thread.
If a reset ctrlr sequence runs on the nvme_ctrlr->thread, we can avoid
redirecting the pending destruct request at completion too.
So we redirect the reset ctrlr sequence into the nvme_ctrlr->thread.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I538b962e2a7b5cf00ebbac2a1e888482ddeeee61
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10075
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.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: Aleksey Marchuk <alexeymar@mellanox.com>
In the following patches, bdev_nvme_reset() will execute the reset ctrlr
operation on the nvme_ctrlr->thread until completion as bdev_nvme_admin_passthru()
does. Hence change the callback bdev_nvme_reset_io_continue() to
redirect to the orig_thread by using bio. Furthermore, use bio->cpl.cdw0
to store the completion status of the reset processing. bdev_nvme_reset()
does not use bio->cpl.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I361cc44494190ba83ad6e360788d78851416c46c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10074
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This patch supports admin passthrough retry when we get any error
with DNR=0 but ABORTED_BY_REQUEST up to retry_count times.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I1bf29570791fdbe8651fa70c4c8685bb740fb86b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9944
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When resetting ctrlr, adminq is disconnected first. If adminq is disconnected,
admin passthrough request is rejected with -ENXIO.
But resetting ctrlr may succeed. If resetting ctrlr succeeds, adminq is
connected again, and admin passthrough request will be
submitted successfully.
On the other hand, if ctrlr is failed, admin passthrough request is
rejected with -ENXIO. But when resetting ctrlr, ctrlr is set to unfailed.
Hence bdev_nvme_admin_passthru() skips any ctrlr which is resetting
or failed, and calls bdev_nvme_admin_passthru_complete() with -ENXIO
if no available ctrlr is found.
bdev_nvme_admin_passthru_complete() queues admin passthrough request
and retry it one second later if ctrlr is resetting.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic748dc4faf29ebf717ae5c29dcf7c55fe2ea9243
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9942
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
The NVMe bdev module queues retried I/Os itself now.
bdev_nvme_abort() needs to check and abort the target I/O if it
is queued for retry.
This change will cover admin passthrough requests too because they
will be queued on the same thread as their callers and the public
API spdk_bdev_reset() requires to be submitted on the same thread
as the target I/O or admin passthrough requests.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: If37e8188bd3875805cef436437439220698124b9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9913
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
After multipath feature is supported, one bdev will have more than one
nvme ctrlr. Fore ease of view, display each ctrlr's trid info.
Moreover, rename nvme_bdev_ctrlr_get as nvme_bdev_ctrlr_get_by_name here
to keep consistent with nvme_ctrlr_get_by_name.
Signed-off-by: Kai Li <lik271@chinatelecom.cn>
Change-Id: I417506699bbea6ed13dac0fee942749757d2ae47
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10129
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
If I/O got ANA error, ANA state may be out of date. So in this case
read ANA log page and update ANA states. Mark nvme_ns to be updating
to avoid using while updating ANA state.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ia43d38b3a589c84d6d0479dedcced033e76fb194
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9458
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
If an I/O failed by ANA error, the corresponding ANA state might be
out of date. In the following patches, for this case, read the latest
ANA log page and update the ANA state. Such reading ANA log page may be
done on multiple threads concurrently including AER ANA change.
Hence protect ANA log page by adding an new flag ana_log_page_updating
to struct nvme_ctrlr and using it.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I8bb84091d50a5fdc0d9893b585be972dfd31c0f1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9526
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Add bdev_retry_count to spdk_bdev_nvme_opts and retry_count to
nvme_bdev_io, respectively.
Set type of both to int because we want use -1 for infinite retry.
Set the default value of bdev_retry_count to zero for the backward
compatibility.
bdev_retry_count is configurable by the RPC bdev_nvme_set_options.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9bc746fcea54aa8722c76f79c70c2ae2b375aa53
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9864
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
The previous patch supported I/O retry when no available io_path
was found at submission.
This patch supports I/O retry when we get I/O path error at completion.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I93a1664944b15ab0a826a321e2ea7a2574263afe
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9850
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
If ANA state is inaccessible or qpair is disconnected, I/O cannot
be submitted.
But if qpair is connected, ANA state may become accessible, or if
qpair is disconnected, it may become connected via resetting.
Hence even if find_io_path() returned NULL, queue I/O and retry it
one second later if qpair is connected or ctrlr is resetting.
Sort retried I/Os by expiration values in ticks, and activate a timed
poller per nvme_bdev_channel only if there is any retried I/O. So
the poller function bdev_nvme_retry_ios() always returns BUSY because
if the poller runs earlier than the closest retried I/O or runs when
there is no retried I/O, it is more like a bug of the framework.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Id28110a0d63ebc1c5772814e2ff8a47934df1644
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9830
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
If NN is very large this saves a lot of memory. This lookup is
not generally used in the I/O path anyway.
Change-Id: I98e190006843ad5d0bac8483bf9feb800d4a665a
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9884
Community-CI: Broadcom 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: GangCao <gang.cao@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
In the SPDK NVMe driver, spdk_nvme_ctrlr_reset_async() sets
ctrlr->is_failed to false and spdk_nvme_ctrlr_reset_poll_async() sets
ctrlr->is_failed to true if it fails.
On the other hand, in the unit test for the NVMe bdev module,
the stub for spdk_nvme_ctrlr_reset_async() does nothing and
the stub for spdk_nvme_ctrlr_reset_poll_async() sets ctrlr->is_failed
to false if it succeeds.
This bug made us very difficult to write unit test for I/O retry.
Hence fix this bug.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ic0dcf1109ce543a53fca74708fc86c8c74a17692
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9829
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
bdev_nvme_find_io_path() selects an io_path whose qpair is connected
and ANA state is optimized or non-optimized.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I79c978795562b606ee27aa43020684d8bcbf50c5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9405
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reset all controllers of a bdev controller sequentially. When resetting
a controller is completed, check if there is next controller, and
start resetting the controller.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I169a84b931c6b03b36bb971d73d5a05caabf8e65
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/7274
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Previously the NVMe bdev module had completed the outstanding reset and
then canceled pending resets. This was complex.
On the other hand, the generic bdev layer cancels pending resets
and then completes the outstanding reset.
Following the generic bdev layer simplifies the code and makes us easier
to control retry reset, delay retry reset by a few seconds, or stop retry
after repeated failures and then delete ctrlr.
Update unit tests accordingly.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9a68422918ebcb052b3a281316ffba9b3450ecd4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9816
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
bdev_nvme_admin_passthru() chooses the first ctrlr which is not failed.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: If41a1d1e1bde4bddfa92e5a385509daa3f0ce4de
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9525
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
We have io_path structure now and returning io_path rather than
ns and qpair match the function name. The following patches will
cache the returned io_path into nvme_bdev_io.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I5d773da18591fc324667f6b5c489a38f497bf3d8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9295
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
This patch removes the critical limitation that ctrlrs which are
aggregated need to have no namespace. After this patch, we can
add multiple namespaces into a single nvme_bdev.
The conditions that such namespaces satisfy are,
- they are in the same NVM subsystem,
- they are in different ctrlrs,
- they are identical.
Additionally, if we add one or more namespaces to an existing
nvme_bdev and there are active nvme_bdev_channels, the corresponding
I/O paths are added to these nvme_bdev_channels.
Even after this patch, ANA state is not utilized in I/O paths yet.
It will be done in the following patches.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I15db35451e640d4beb99b138a4762243bee0d0f4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8131
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Specifying only a transport id is not enough. We need to be able to
describe the host parameters too.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: Iadbea553aee4b38e7cacab0b486e7e5746d0d1ab
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9825
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
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: Aleksey Marchuk <alexeymar@mellanox.com>
This is the currently active path identifier in a failover scenario. The
path is defined by more than just the transport identifier, so fix the
name.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I682c6f4c54f75307e2615bf80e70358180d99fe2
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9576
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This defines a unique path between a host and a target.
Change-Id: Ia3d24c1b34199a8b596aaf17900ca9694a9da77d
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9505
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This patch replaces the synchronous `spdk_nvme_detach()` calls with its
asynchronous counterparts in the controller unregister path.
An additional poller is introduced to periodically poll the NVMe driver
for detach completion. Once the detach is completed, the poller is
unregistered and the nvme_ctrlr is destroyed. The poller uses the same
period (1ms) as the async probe poller.
Since reset and detach cannot happen at the same time, reset_poller was
renamed to reset_detach_poller and it can now store the pointer either
to the reset or detach poller, depending on the circumstances.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I5eb2dd6383d98d25d1f9748af08c1a13d18acb0e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8729
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
It can match by any provided parameter to remove paths.
Change-Id: I5e7a87342bbb90943dc97fb52f142814fcf0acfa
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9453
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Instead of storing an spdk_nvme_transport_id, store the object that
contains it. This will make a few later patches easier.
Change-Id: I36b74889fe39af3b7ab2b900fb3ea4b3f39e1f83
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9484
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Allow to return more than one memory domain.
This change aligns bdev and nvme API and provides
more flexibility for custom transports.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: Ica9b12ad8463c361be6cb62ee2c0513eec0b486d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9546
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This patch enables us to aggrete multiple ctrlrs in the same NVM
subsystem into a single bdev ctrlr to create multipath.
This patch has a critical limitation that ctrlrs which are aggregated
need to have no namespace. Hence any nvme bdev is not created.
However it will be removed in the next patch.
The design is as follows.
A nvme_bdev_ctrlr is created to aggregate multiple nvme_ctrlrs in
the same NVM subsystem. The name of the nvme_ctrlr is changed to be
the name of the nvme_bdev_ctrlr.
NVMe bdev module has both the failover feature and the multipath
feature now. To choose which of failover or multipath to use, add an new
parameter multipath to the RPC bdev_nvme_attach_controller.
When we attach a new trid to the existing nvme_bdev_ctrlr, we use the failover
feature if multipath is false, we use the multipath feature if multipath is
false.
nvme_bdev_ctrlr has a list for nvme_ctrlr and it is guarded by the
global mutex. Callers can query nvme_ctrlrs from a nvme_bdev_ctrlr via
trid as a key. nvme_bdev_ctrlr is not registered as io_device.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I20571bf89a65d53a00fb77236ad1b193e88b8153
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8119
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Previously, if an I/O qpair is disconnected, we tried reconnecting
the qpair. However, this reconnect operation was very likely to fail
and will not match the upcoming asynchronous connect/reconnect
operation. We need an extra callback to make this reconnect operation
asynchronous, but we do not want to have it.
Hence if an I/O qpair is disconnected, we free the I/O qpair and then
reset the corresponding nvme_ctrlr immediately. If the admin qpair is
also disconnected, the nvme_ctrlr is reset immediately. However this
event may never happen. So we do not wait for the error of the admin
qpair.
The NVMf host may disconnect connections by itself intentionally.
In this case, resetting the nvme_ctrlr will surely fail. But resetting
the nvme_ctrlr frees all I/O qpairs of the nvme_ctrlr and these I/O
qpairs are not created again until resetting the nvme_ctrlr succeeds.
Resetting the nvme_ctrlr once at most is more efficient than repeating
reconnecting the I/O qpair. So this change is valuable even for such
intentional disconnection. However, it is helpful to know the event that
I/O qpair is disconnected. Hence change DEBUGLOG to NOTICELOG in the
disconnected callback. The disconnected callback is not repeated, and
we do not need to worry about NOTICELOG flooding.
Refine the unit test case to verify this change.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I376b749c2f55d010692bf916370e8bb4249b795f
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9515
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
In unit tests, spdk_nvme_ctrlr had opts but did not use it.
Hostnqn will be checked to determine if multipath can be created.
Hence we implement the stub spdk_nvme_ctrlr_get_default_ctrlr_opts()
and copy probe_ctx->opts to ctrlr->opts as we do in lib/nvme.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I13980424d5f463877eae7f7cd1e5ffcae888aebe
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9333
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@gmail.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Previously bdev_nvme_reset() returned -EBUSY if ctrlr is being
destructed and returned -EAGAIN if ctrlr is being reset.
These did not match what spdk_nvme_ctrlr_reset() returned.
Reset operation will be more important than current when multipath
is supported and reset operation is made asynchronous.
Hence change bdev_nvme_reset() to follow spdk_nvme_ctrlr_reset().
bdev_nvme_reset() returns -ENXIO if ctrlr is being destructed and
returns -EBUSY if ctrlr is being reset.
Additionally change the return value of bdev_nvme_failover()
accordingly. After the change bdev_nvme_failover() returns -ENXIO
if being destructed and returns -EBUSY if ctrlr is being reset.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ie2c6f8601050b1043d83de9cf01490751784e4e5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8859
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Konrad Sztyber <konrad.sztyber@gmail.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Following the last patch, include hostid into ctrlr_opts rather than
passing it as a parameter for bdev_nvme_create().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I0d04db1c5767ec76a9a7cd255c3a8d56b0b8f583
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9344
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@gmail.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
This only existed to share code between OCSSD and regular NVM
namespaces. Now OCSSD is gone, so just merge the files into bdev_nvme.
Change-Id: Idb73cc05d67144de5dd20af8db24c8f6974d10a7
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9337
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.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>
nvme_ctrlr_populate_namespaces
Avoid relying on this number. Different targets have interpreted its
meaning in different ways and it cannot be used anymore in practice. It
may also be very, very large.
Change-Id: I94e8eae49d6ccdbd8be302b30a120d89242b6d39
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9316
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Try to use these accessors instead of directly using the namespaces
array. This will make changing the data structure easier later on.
Change-Id: I3367d0e0065894f3aa199ed1698d27976b4cbbb5
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9315
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
If the number of namespaces is very large, this can cause excessive
memory allocation. This is especially true because when the number of
namespaces is large, it is almost always very sparsely populated.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I27d94956c222ae3c49c6a7422164ae3a8ec8d963
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9302
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
When preparing for a reset, use this new call to tell
the driver to avoid sending DELETE_CQ/SQ commands to a
PCIe controller when they aren't needed.
Fixes issue #2073.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9ebb7d5c3f7cbb1c3192f162f32edbbea41acde1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9250
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Matt Dumm <matt.dumm@hpe.com>
Reviewed-by: Michael Haeuptle <michaelhaeuptle@gmail.com>
As far as we're aware, this is not in use by anyone. OCSSD has largely
been replaced by ZNS and no OCSSD drives made it to the market.
Change-Id: I020ee277da5292f8c4777f224acafd87586f8238
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9328
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
bdev_nvme_create() is called only by a single caller and hostnqn is
just copied to ctrlr_opts even if it is passed separately.
Hence include hostnqn into ctrlr_opts rather than passing it as a
parameter for bdev_nvme_create().
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I75b640bcecefa94950b0c19936fab0571c428125
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9332
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@gmail.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
