We recently improved qpair disconnect process and added assert
if we get a completion without any error when a qpair is disconnected.
However unexpectedly we saw this case very often when we ran the test
test/nvmf/host/multipath.sh for the real hardware in the test pool.
So we remove the assert and change the ERRLOG to INFOLOG.
Fixes one of the issues in #2300
Signed-off-by: Shuhei Matsumoto <shuheimatsumoto@gmail.com>
Change-Id: Iedbf7e0afa5025da6a810043ba95348ba5b856b3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10901
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: Ben Walker <benjamin.walker@intel.com>
If nvme ctrlr is resetting or initializing, free_io_qids
bitmap is already freed or not created yet. In that case
an attempt to create IO qpair leads to segmentation fault.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I6a97bf81d5a568db20d23b3f88cf01e994ba42e3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10827
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: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuheimatsumoto@gmail.com>
In current implementation RDMA qpair is destroyed right after
disconnect. That is not graceful qpair shutdown process since
there can be requests submitted to HW and we may receive
completions for already destroyed/freed qpair.
To avoid this, only disconnect qpair in ctrlr_disconnect_qpair
transport callback, all other resources will be released in
ctrlr_delete_io_qpair cb.
This patch is useful when nvme poll groups are used since in
that case we use shared CQ, if the disconnected qpair has WRs
submitted to HW then qpair's destruction will be deferred to
poll group.
When nvme poll groups are not used, this patch doesn't change
anything, in that case destruction flow is still ungraceful.
However since CQ is destroyed immediately after qpair,
we shouldn't receive any requests which point to released
resources. A correct solution for non-poll group case
requires async diconnect API which may lead to significant
rework.
There is a bug when Soft Roce is used - we may receive
a completion with "normal" status when qpair is already
disconnected and all nvme requests are aborted. Added
a workaround for it.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I0680d9ef9aaa8737d7a6d1454cd70a384bb8efac
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10327
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 <shuheimatsumoto@gmail.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This API is a helper for getting the full discovery
log page from a discovery controller. It will read the
log page header to get the total number of entries,
allocate a buffer for all of the entries, and then
issue a series of get_log_page commands to read each
4KiB worth of entries.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I02666ef5adcb9fc8825a221655811ace708f97b8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10564
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.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>
nvme_ctrlr_construct_namespaces
We can just reallocate here to be more efficient.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I8cfc87da23aee6c05ff83aea2165683dddba1dbd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10688
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In the one place this was called, we can call nvme_ns_construct
instead. There's no harm in re-fetching the identify pages.
Change-Id: I91292ff9650bdc7edd5588a05837b671dcac1922
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10102
Community-CI: Mellanox Build Bot
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The host may have specified a hostnqn to use to connect to
a discovery ctrlr, so we can't just use the default ctrlr
opts to connect - we need to call the probe_cb (if there is
one) to get any options that the host may have specified.
Tested by using discovery_aer tool, creating a subsystem and
listener, and then adding a host on the target side that matches
the hostnqn specified to the discovery_aer tool.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I07266e984e0094d3a768e6a0d5ea3a3bd71e32ba
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10547
Community-CI: Mellanox Build Bot
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The NVMe bdev module enables asynchronous IO QP creation by default, after
calling `spdk_nvme_ctrlr_alloc_io_qpair` and `spdk_nvme_ctrlr_connect_io_qpair`,
the queue pair is in connecting state at the beginning, then users may call
`spdk_nvme_ctrlr_free_io_qpair` immediately, and the common layer will
change queue state to NVME_QPAIR_DISCONNECTING and NVME_QPAIR_DESTROYING,
so in function `nvme_pcie_ctrlr_delete_io_qpair` the workaround to wait
for create cq/sq callbacks will not be called, instead of using the common
layer queue state here, we should use the internal `pcie_state`.
Fix#2245.
Change-Id: I801caf26563464b135035bf7fa2f63def13de9f4
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10445
Reviewed-by: Dong Yi <dongx.yi@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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The NVMe bdev module will support two features, delayed reconnect and
delete after multiple failures of reconnect to improve error recovery.
The recently added two APIs, spdk_nvme_ctrlr_reset_async() and
spdk_nvme_ctrlr_reset_poll_async(), were not good enough.
spdk_nvme_ctrlr_reset_ctx was not necessary. It had only a pointer to ctrlr.
Using a pointer to ctrlr directly saves us from undesirable malloc error
processing.
Separate spdk_nvme_ctrlr_reset_async() into spdk_nvme_ctrlr_disconnect()
and spdk_nvme_ctrlr_reconnect_async(). spdk_nvme_ctrlr_disconnect()
disconnects ctrlr including disconnecting adminq.
spdk_nvme_ctrlr_reconnect_async() moves the ctrlr state to INIT.
Then rename spdk_nvme_ctrlr_reset_poll_async() by
spdk_nvme_ctrlr_reconnect_poll_async().
Finally deprecate spdk_nvme_ctrlr_reset_async() and
spdk_nvme_ctrlr_reset_poll_async().
The following patches will change the NVMe bdev module to use these new APIs.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Id1d6858dcdc5fc2e9db0a6ebf3f79cab4f9bbcb7
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10091
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>
Also we don't treat exceptions when getting INTEL log pages
as a fatal error, the initialization will still contine.
Change-Id: Ic2fd2be510fde2679c1546482934d0a180266936
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10341
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>
Move the CONFIGURE_AER state before SET_KEEP_ALIVE to
make sure that we run the CONFIGURE_AER state for
discovery controllers.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ia4e24f6507c43e3fece06b9161ff8e0b8fa0e97d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10332
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: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
We will actually run the CONFIGURE_AER state for
discovery controllers in a future patch.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ib114beb886ab4b9214e4525479eb5ec7e038e5d9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10331
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: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
These can produce a lot of output, which doesn't really give any
additional information.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I572cd203d61c717ce6400f67ef27ec1d7bb54c0c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10414
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: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
For error case, just set ctrlr->num_aers to 0, and
then the loop won't execute at all. This avoids an
extra call to nvme_ctrlr_set_state() and simplifies
the code a bit.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iff7bbf6e03d18b5f553b9e8527b4c803db583917
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10330
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>
Discovery services using the SPDK nvme driver may
use long-lasting connections that detect AER completions
to determine when there are changes in the discovery
log. This means that we still need to send keep alives
on discovery controller admin queues. So move the
SET_KEEP_ALIVE_TIMEOUT state immediately after
IDENTIFY, and run the SET_KEEP_ALIVE_TIMEOUT state
even for discovery controllers.
Note, we need the IDENTIFY's KAS value to properly
set the keep alive timeout, so we have to keep the
IDENTIFY state before SET_KEEP_ALIVE_TIMEOUT.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I5c6403c28fb72d42629c5f9009a89c4bfd44d162
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10329
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>
Keep alive is valid for discovery controllers, so don't overwrite
the value requested with zero in nvme_fabric_ctrlr_scan().
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7dcda6ebf4ab1c8a9085e4e3a02b814d8e586a97
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10328
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
There is a bad memory corruption where the code for CUSE attempts to write
one byte (with value 0x1) after the memory is freed.
Context:
When the CUSE device is unregistered, the poller thread is signaled with
fuse_session_exit(), which writes the value 1 to fuse_session::exited.
The poller thread then detects with fuse_session_exited() that it must exit
the routing and finally destroys its own fuse session with
cuse_lowlevel_teardown() before it exits.
However, FUSE may also call fuse_session_exit() for its internal purposes.
I'm not sure exactly under what conditions that happens, but I added
some trace messages and I could clearly see that the CUSE thread exits
before it was requested to exit in cuse_nvme_ns_stop().
If the poller thread early-exits, it would destroy its own FUSE session
(and free the memory) before fuse_session_exit() gets executed, causing
the memory to be corrupted with a single byte of value 0x1.
Reproducer:
The bug can be reproduced by resetting the FUSE session to NULL after it
is destroyed. This will cuse_nvme_ns_stop() to crash with a segmentation
fault in fuse_session_exit() because it tries to access a NULL pointer.
static void *
cuse_thread(void *arg)
{
[...]
free(buf.mem);
fuse_session_reset(cuse_device->session);
+ cuse_device->session = NULL;
pthread_exit(NULL);
}
This fix:
The fix I suggest is to destroy the FUSE session with
cuse_lowlevel_teardown() after the thread is joined.
Signed-off-by: Sylvain Didelot <sdidelot@ddn.com>
Change-Id: I47202891a358f139506845110b012f840974b6fc
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9931
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: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
spdk_nvme_ctrlr_free_io_qpair can be called when
qpair is already disconnected. In that case qpair's
state is changed to NVME_QPAIR_DESTROYING and
transport's ctrlr_delete_io_qpair callback is
called. RDMA and TCP transports call
nvme_transport_ctrlr_disconnect_qpair in
the callback and since qpair's state is
not DISCONNECTED or DISCONNECTING, qpair
is disconnected for the second time.
If spdk_nvme_ctrlr_free_io_qpair is called
when qpair is in ENABLED state than nothing
changes, qpair will be disconnected before destroy.
PCIE/vfio_user don't implement transport disconnect
callback, so they are not affected.
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I23e11856ecafb51669acf4a3118be049c11eecda
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10326
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: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
For DSM command, the NVMe drive may take a long time to finish it,
if we set a small timeout value for DSM command, the bdev/nvme module
will try to reset the IO queue pair when timeout happens,
in `spdk_nvme_ctrlr_free_io_qpair`, we will abort the outstanding
IO requests first, then in the `nvme_pcie_ctrlr_delete_io_qpair`,
we will poll the CQ for any requests that have been completed by
the NVMe controller, if there are NVMe completions in the CQ,
we will finish them again, thus double completions happened.
Here we rename `nvme_qpair_abort_reqs` to `nvme_qpair_abort_all_queued_reqs`,
so the common layer will just abort queued request, and let each
transport to abort outstanding requests case by case.
Fix#2233.
Change-Id: Icae6214239160c615418cb514fc51cfe77b59211
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10233
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>
Add a parameter which determines the owner of the
map - target or initiator. It allows to set different
access flags when creating Memory Regions
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I0016847fe116e193d0954db1c8e65066b4ff82bf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10283
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
The unit test test_nvme_cuse_stop() manually creates 2 cuse devices
and executes nvme_cuse_stop(). Problem is that the Fuse session is
never initialized for those 2 cuse devices, causing cuse_nvme_ns_stop()
to access 'ns_device->session', which is a NULL pointer.
This bug is detected by ASAN as follows:
==77298==ERROR: AddressSanitizer: SEGV on unknown address 0x000000000180 (pc 0x7fdac6d7d40e bp 0x000000000000 sp 0x7fff74768320 T0)
==77298==The signal is caused by a READ memory access.
==77298==Hint: address points to the zero page.
0 0x7fdac6d7d40e in fuse_session_destroy (/usr/lib64/libfuse3.so.3+0x1640e)
1 0x40dc7a in cuse_nvme_ns_stop /home/vagrant/spdk_repo/spdk/lib/nvme/nvme_cuse.c:851
2 0x40df59 in cuse_nvme_ctrlr_stop /home/vagrant/spdk_repo/spdk/lib/nvme/nvme_cuse.c:923
3 0x40f103 in nvme_cuse_stop /home/vagrant/spdk_repo/spdk/lib/nvme/nvme_cuse.c:1094
4 0x415803 in test_nvme_cuse_stop /home/vagrant/spdk_repo/spdk/test/unit/lib/nvme/nvme_cuse.c/nvme_cuse_ut.c:393
5 0x7fdac724c1a6 (/usr/lib64/libcunit.so.1+0x41a6)
6 0x7fdac724c528 (/usr/lib64/libcunit.so.1+0x4528)
7 0x7fdac724d456 in CU_run_all_tests (/usr/lib64/libcunit.so.1+0x5456)
8 0x415a4e in main /home/vagrant/spdk_repo/spdk/test/unit/lib/nvme/nvme_cuse.c/nvme_cuse_ut.c:415
9 0x7fdac62351e1 in __libc_start_main (/usr/lib64/libc.so.6+0x281e1)
10 0x403ddd in _start (/home/vagrant/spdk_repo/spdk/test/unit/lib/nvme/nvme_cuse.c/nvme_cuse_ut+0x403ddd)
The fix is to call fuse_session_destroy() only if the fuse session is != NULL.
Signed-off-by: Sylvain Didelot <sdidelot@ddn.com>
Change-Id: I41881243227d83e8d1e6b90e72c1b6d62ccd98d3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10225
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: Ben Walker <benjamin.walker@intel.com>
Wait until the namespace is attached, where it does this operation
again. As of this commit it doesn't really matter because it is just
filling in some values in a structure and if it does it twice it's not a
problem. But later when we only allocate active namespaces, we do not
want to allocate the namespace twice.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: Ie28653b178975d1ca80bf71ca6b5095224f1c5d1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10026
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: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Eugene Kochetov <ekochetov@yandex.ru>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This function destructs a single namespace and removes it from the
controller.
Change-Id: I4b7b3576beda85c9ddad4e0f2db6d1964fa72b82
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10024
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: Eugene Kochetov <ekochetov@yandex.ru>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This was sometimes used as the maximum array index and sometimes as the
maximum count. Make it consistent everywhere and give it a better name.
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: I518efd99a7d36584624490b0b3497bb6e81ce9ac
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10101
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 <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
list
The list should always be null terminated, but add an additional layer
of buffer overrun protection.
Change-Id: Iee31057fdca5ec4a6177615dff5171e5cb07984e
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10027
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: Eugene Kochetov <ekochetov@yandex.ru>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
This quirk was already applied to the 0x0A53 SSD, but is
likely needed on 0x0A54 as well.
Possible fix for issue #2231.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I36a48ab92d1698a411472f714b5108413bbc3c56
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10162
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Community-CI: Mellanox Build Bot
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
We already provides the API `spdk_nvme_ctrlr_is_fabrics` to return
input controller is fabrics controller or not, but it needs a controller
data structure as the input, so here we add another API to do the same
thing and it takes the transport type as the input, with this change,
both nvme and nvmf library can use the API.
Also we should treat UINT8_MAX(255) as valid fabrics transport type.
Change-Id: Ib62e7d3eca3da1ddb1a4cc55b0b62e274522f1ce
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/10059
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>
The qpair's state member is only 3 bits of a uint8_t,
and the in_completion_context bit is another bit in that
same uint8_t.
We know that the qpair's state is only ever updated by
one thread, but it is possible that the state could
be modified by one thread, while another thread
is modifying in_completion_context.
in_completion_context is only modified by the thread
that is polling the qpair (or the qpair's poll group).
But with async mode, another thread that has a qpair
on the same PCIe controller could poll its adminq and
reap the SQ completion for the qpair that's owned by
the other thread.
So do *not* set the generic qpair state to CONNECTED
from the SQ completion callback. Instead just set
the pcie_state to READY, and let the thread that owns
the qpair detect the qpair is READY and set the state
to CONNECTED itself.
Fixes issue #2157.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9efc0c954504f1841e1c3890ae78211ad0d1990e
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9975
Reviewed-by: Konrad Sztyber <konrad.sztyber@intel.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>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
In some cases a single virtually contriguos memory
buffer can be translated to several chunks of memory.
To make such translation possible, update structure
spdk_memory_domain_translation_result to use a pointer
to iovec.
Add a single iov structure or cases where translation
is always 1:1, it will make easier translation callback
implementation. For RDMA transport translation of address
is always 1:1, so treat iovcnt other than 1 as an
error.
Change-Id: I65605575d43a490490eba72c1eb19f3a09d55ec6
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9779
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>
Instead of a union with domain type specific
parameters, store an opaque pointer to user
context. Depending on the memory domain type,
this context can be cast to a specific struct,
e.g. to spdk_memory_domain_rdma_ctx for RDMA
memory domains.
This change provides more flexibility to
applications to create and manage custom
memory domains
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Max Gurtovoy <mgurtovoy@nvidia.com>
Change-Id: Ib0a8297de80773d86edc9849beb4cbc693ef5414
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9778
Community-CI: Broadcom 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: Ben Walker <benjamin.walker@intel.com>
This is to help with binding trace objects together and
for the convenience (all trace definitions are in one place
instad of being scattered accross multiple files).
Change-Id: Ib15bc9c2eeee9c4d0816bcee509ab69f3f558e19
Signed-off-by: Krzysztof Karas <krzysztof.karas@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9574
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
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: Konrad Sztyber <konrad.sztyber@gmail.com>
Detaching a controller with `no_shn_notification` flag set will follow
the regular detach path making it asynchronous too.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I0b9c6c23626b4cc1cfaedb3268024776a07b9195
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9003
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.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>
The controller detach had asynchronous API (with async/poll), but the
register operations were synchronous, so they would block on fabrics
controllers. In this patch, they're changed to their non-blocking
counterparts, making the detach fully asynchronous.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I74df12ab40a54f1d675639672e03755c89768bef
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8726
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: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
When data digest is enabled for a nvme tcp qpair, we can use accel_fw
to calculate the data crc32c. Then if there are multiple
c2h pdus are coming, we can use both CPU resource directly
and accel_fw framework to caculate the checksum. Then the datao value compare
will not match since we will not update "datao" in the pdu coming order.
For example, if we receive 4 pdus, named as A, B, C, D.
offset data_len (in bytes)
A: 0 8192
B: 8192 4096
C: 12288 8192
D: 20480 4096
For receving the pdu, we hope that we can continue exeution even if
we use the offloading engine in accel_fw. Then in this situation,
if Pdu(C) is offloaded by accel_fw. Then our logic will continue receving
PDU(D). And according to the logic in our code, this time we leverage CPU
to calculate crc32c (Because we only have one active pdu to receive data).
Then we find the expected data offset is still 12288. Because "datao" in tcp_req will
only be updated after calling nvme_tcp_c2h_data_payload_handle function. So
while we enter nvme_tcp_c2h_data_hdr_handle function, we will find the
expected datao value is not as expected compared with the data offset value
contained in Pdu(D).
So the solution is that we create a new variable "expected_datao"
in tcp_req to do the comparation because we want to comply with the tp8000 spec
and do the offset check.
We still need use "datao" to count whether we receive the whole data or not.
So we cannot reuse "datao" variable in an early way. Otherwise, we will
release tcp_req structure early and cause another bug.
PS: This bug was not found early because previously the sw path in accel_fw
directly calculated the crc32c and called the user callback. Now we use a list and the
poller to handle, then it triggers this issue. Definitely, it will be much easier to
trigger this issue if we use real hardware engine.
Fixes#2098
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: I10f5938a6342028d08d90820b2c14e4260134d77
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9612
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>
Reviewed-by: GangCao <gang.cao@intel.com>
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Abort any queued admin requests once admin queue gets enabled. A request
can get queued if a controller is being reset and it gets submitted
while admin qpair is being reconnected. If these requests aren't
aborted, the init process will stall, as requests don't get resubmitted
while controller is resetting and subsequent admin commands required for
the initialization would be queued too.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: If456a297d2d434b3cc741816cbfb13b01d37e963
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9324
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jim Harris <james.r.harris@intel.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>
Enable dump of transport stats in functional test.
Update unit tests to support the new statistics
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I815aeea7d07bd33a915f19537d60611ba7101361
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8885
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: 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 function may return an error and we should handle it
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I996ceb6e300bd9527385aded9068b2841e94a20d
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9278
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>
nvme_transport_poll_group_remove() clears qpair->poll_group. Hence
we should not use it after that.
Fixes#2170
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I57ceee8c66684e2d02b51b8a0f3d66aacbcb9915
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9560
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: GangCao <gang.cao@intel.com>
Reviewed-by: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Dong Yi <dongx.yi@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Renamed nvme_qpair_abort_reqs() to nvme_qpair_abort_reqs_with_cbarg() to
highlight the fact that it only aborts requests with specified cb_arg
and to distinguish it from _nvme_qpair_abort_reqs() which aborts all
requests immediately.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I32fec5ab0501b1beb8605689d73ec42a6424fba5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9323
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>
When a controller is reset, it goes through the whole init process,
including establishing keep-alive, so there's no point in sending it
during that time. Additionally, it can stall the initialization if it
gets queued when adminq is connecting.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ib9fee42f6097972e8ba336958f81e1b6b1a5f006
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9310
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>
Additionally, this patch removes reading the CC and CSTS registers from
`nvme_ctrlr_process_init()`, as it's no longer needed.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: If4f9e57dbf249fbce87e90018cff389f59906e38
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8621
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>
When the controller initialization is failed due to not being able to
read the CSTS register, an ERRLOG is now printed instead of DEBUGLOG.
It should make it easier to debug initialization failures.
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: I42602bd08e263faf2e87fe4a68abcc82ddec11ba
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/9556
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: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
The CC register is now re-read again when disabling the controller as
preparation for subsequent patches, in which the synchronous CC register
read will be removed from nvme_ctrlr_process_init().
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Konrad Sztyber <konrad.sztyber@intel.com>
Change-Id: Ibfc8ed85bab188c3938451fbdfb771b969157807
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/8619
Community-CI: Broadcom CI <spdk-ci.pdl@broadcom.com>
Community-CI: Mellanox Build Bot
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
