The current approach checks "rc == 0". It worked before adding
polling of poll group since a single qpair should return 1
completion for its own icreq while poll group can return
several completions for all qpairs attached to this poll
group (but .e.g not for those qpair who is waiting for the
completion).
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Change-Id: I60d05d8d6640e4e2bbaf3cd533d2f5a3637adea1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4768
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>
Add support for the ZNS zone management receive command.
An internal nvme_zns_zone_mgmt_recv() function is created
that matches the parameters of the zone management receive
function in the ZNS specification.
Convenience functions are provided for the following
Zone Receive Action: Report Zones.
Zone Receive Actions not implemented: Extended Report
Zones.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I23589a602336da5dffccec7230d07026a868e81b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4793
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>
Add a function to get the number of zones for a zoned namespace.
Since the ZNS specification does not allow zone size == 0,
divide by zero should not be possible on a spec compliant drive.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I15e0ac8a72e244e248b7fb44f7156b182ecd98b4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4792
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 CI
Add a function to get the zone size of a zoned namespace.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I89429fa2fd36307a8b34bdf91e3251c68596a48c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4791
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>
Community-CI: Broadcom CI
Add two new public APIs, spdk_nvme_detach_async() and
spdk_nvme_detach_poll_async() to detach multiple controllers in parallel
as a simple manner to users.
Hold the target controller to nvme_ctrlr_detach_ctx because users
will free any object which held it after returning spdk_nvme_detach_async().
spdk_nvme_detach_ctx holds all nvme_ctrlr_detach_ctx in a sequence
by linked list.
spdk_nvme_detach_ctx has a boolean variable polling_started to prevent
the user from calling spdk_nvme_detach_async() while
spdk_nvme_detach_poll_async() is called repeatedly.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>i
Change-Id: Ib049c19f7ef24410b963fd5c777a21184f3012d1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4758
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Add two new helper functions, nvme_ctrlr_detach_async() and
nvme_ctrlr_detach_poll_async() to make the internal of
spdk_nvme_detach() asynchronous.
Use callback function to remove controller from the attached list after
completing shutdown and before freeing to avoid conflict between
attach and detach.
Update MOCKs in the corresponding unit test cases.
The next patch will add two public APIs spdk_nvme_detach_async()
and spdk_nvme_detach_poll_async() based on this patch.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ifbdfec2a1facde9354007c6248f280e245a36eed
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4416
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Following the last patch, separate nvme_ctrlr_destruct()
into nvme_ctrlr_destruct_async() and nvme_ctrlr_destruct_poll_async(),
but keep nvme_ctrlr_destruct() by replacing the internal by
nvme_ctrlr_destruct_async() and nvme_ctrlr_destruct_poll_async().
Add shutdown_complete to nvme_ctrlr_detach_ctx. If shutdown_complete is true,
we can destruct the controller. The case that nvme_ctrlr_shutdown_async()
failed sets shutdown_complete to true. The case that nvme_ctrlr_disable()
is called sets shutdown_complete to true unconditionally.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I3994e259f9d3ccf8fede3ac03aadef911eefb9dd
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4415
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jacek Kalwas <jacek.kalwas@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>
This patch is the first of the patch series to make spdk_nvme_detach()
asynchronous.
We have lengthy shutdown notification, i.e., we have to wait a long time
until shutdown processing is completed, in some SSDs. If the running system
has many such SSDs, we see large intolerable delay.
SPDK provides a controller option, no_shn_notification as a workaround.
We can use the workaround if the use case of the detach is to switch to
the next application without system reboot. However, we cannot use the
workaround if we want to do system reboot after detach.
To mitigate such lengthy shutdown notification, we need to parallelize
detachment among SSDs.
Hence the patch series will introduce an asynchronous detach API and
will use the API to parallelize detachment.
This patch adds the following changes.
Introduce a context structure and separate nvme_ctrlr_shutdown()
itno nvme_ctrlr_shutdown_async() and nvme_ctrlr_shutdown_poll_async()
using the context structure.
Name the context structure as nvme_ctrlr_detach_ctx because it will be
used only in internal APIs. The upcoming public APIs will support
multiple detachment and will have the contest structure named as
spdk_nvme_detach_ctx.
Use TSC instead of counter because polling interval will be controlled
by the caller.
Use the convenient macro, SPDK_CEIL_DIV(), to round off the time
value in milliseconds.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I9e2355fd24b6d6a4d6c1813577d53822304d4f33
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4414
Reviewed-by: Jacek Kalwas <jacek.kalwas@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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add support for the ZNS zone management send command.
An internal nvme_zns_zone_mgmt_send() function is created
that matches the parameters of the zone management send
function in the ZNS specification.
Convenience functions are provided for the following
Zone Send Actions: Close Zone, Finish Zone, Open Zone,
Reset Zone, and Offline Zone.
Zone Send Actions not implemented: Set Zone Descriptor
Extension.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I9cc04776c249d2b9c515438b81a0f06aea4198fe
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4682
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>
Move ZNS specific functions to their own header file:
include/spdk/nvme_zns.h.
This is similar to how it's done for e.g. NVMe-oF with
include/spdk/nvmf.h.
This avoids cluttering the nvme.h header with ZNS specific functions.
A SPDK program that wants to use the spdk_nvme_zns_* functions will
need to include include/spdk/nvme_zns.h, however, this is similar
to how e.g. spdk_nvmf namespaced functions work.
This change is safe to do now, since there is no SPDK release with
ZNS so far.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: If39fab4e7a6924cc3bd67aa0c03e9c6de5f8a324
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4681
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>
Since namespace types were introduced in NVMe, the CC.CSS register
has a new value (SPDK_NVME_CC_CSS_IOCS) which indicates that all
supported command sets should be selected/enabled. This possibly
includes command sets other than NVM and ADMIN only.
Therefore, if a SPDK application wants enable all the command sets
that the controller supports, it has to explicitly set
opts->command_set to SPDK_NVME_CC_CSS_IOCS.
To avoid possibly a lot of SPDK applications having to set this
parameter, check if the user requested a command set explicitly,
if not, make SPDK automatically use the most reasonable default,
based on the supported bits set by the controller.
The most common case is that you want to enable (all) the command
sets that the controller supports.
A user will still be able to restrict the controller to only use
the NVM command set (or ADMIN only), by setting opts->command_set
to a specific value.
Since the current default command set value specified by
spdk_nvme_ctrlr_get_default_ctrlr_opts() is SPDK_NVME_CC_CSS_NVM,
which is defined as 0, we cannot know if the user specified a
command set explicitly or not.
To solve this, change the default command set value specified by
spdk_nvme_ctrlr_get_default_ctrlr_opts() to CHAR_BIT (0x8), which
is larger than the largest value that can be set in CS.CSS (which
is only 3 bits wide, thus 0x7).
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I45ec148d3667ab87c41fbfb6d6612a1e0e5c9d9c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4701
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>
cuse defines the ioctl handler to take an int for
the command, but the kernel uses unsigned int. So
cast the int parameter to unsigned int when doing
the switch.
Fixes issue #1613.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7ef13afc755ee79d8a79b3c7a09dccee7bb7aa1c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4716
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
There is an error when do following sequences:
1. Allocate an I/O queue pair
2. Do controller reset via spdk_nvme_ctrlr_reset
3. Allocate an I/O queue pair
becaues the free_io_qids was reset and didn't
restore.
Fix issue #1621.
Change-Id: Icd533f171079c12fe03be07e659e8eed9b082384
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4698
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
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 file included from nvme_rdma_ut.c:36:
/home/clear/spdk/lib/nvme/nvme_rdma.c:651:22: note: ‘bad_send_wr’ was declared here
651 | struct ibv_send_wr *bad_send_wr;
| ^~~~~~~~~~~
In file included from /home/clear/spdk/lib/nvme/nvme_rdma.c:41,
from nvme_rdma_ut.c:36:
/home/clear/spdk/lib/nvme/nvme_rdma.c: In function ‘nvme_rdma_poll_group_process_completions’:
/home/clear/spdk/include/spdk/log.h:132:2: error: ‘bad_send_wr’ may be used uninitialized in
this function [-Werror=maybe-uninitialized]
132 | spdk_log(SPDK_LOG_ERROR, __FILE__, __LINE__, __func__, __VA_ARGS__)
| ^~~~~~~~
cc1: all warnings being treated as errors.
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Change-Id: I38ae36756b4bacef7e89f0f1737684c8b8981b12
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4696
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
There is nothing left here, so remove it.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: Ib947d42bc577dbebb4650b1be885e05a80f8f8cf
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4541
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Community-CI: Broadcom CI
This patch removes the string from register component.
Removed are all instances in libs or hardcoded in apps.
Starting with this patch literal passed to register,
serves as name for the flag.
All instances of SPDK_LOG_* were replaced with just *
in lowercase.
No actual name change for flags occur in this patch.
Affected are SPDK_LOG_REGISTER_COMPONENT() and
SPDK_*LOG() macros.
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
Change-Id: I002b232fde57ecf9c6777726b181fc0341f1bb17
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4495
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Mellanox Build Bot
Reviewed-by: Anil Veerabhadrappa <anil.veerabhadrappa@broadcom.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Community-CI: Broadcom CI
We can receive buffer reclaim notifications only when a qpair is
attached to a poll group (so qpair's socket is connected to a socket
poll group).
The previous assumption that we enable zcopy only for IO qpairs was
wrong since IO qpair might not use poll groups too (e.g. abort
application).
Fixes issue #1607
Change-Id: I67329d755d81da6606e65eddfeceb20839346d87
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4476
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>
When we support spdk_nvme_detach_async(), any controller may be
destructed asynchronously. We will be able to know the case by
ctrlr->is_destructed is true and ctrlr is queued in the attached list.
nvme_ctrlr_probe() should fail if the found ctrlr satisfies these
conditions.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I299c2e5ea3c16cc1239899c163bb9e0eb921ade5
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4413
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Add getters to the ZNS specific data structures, so that an
SPDK application, e.g. examples/nvme/identify/identify.c,
has the ability to get and utilize the information in them.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I26056161093cc811acb6840ff7e2068e5f6058f6
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4412
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>
Add a new state in the SPDK NVMe state machine in order to fetch
I/O Command Set Specific Namespace data structures.
Right now there is only support for the Zoned Namespace Command Set
Specific Identify Namespace data structure.
The NVM Command Set Specific Identify Namespace data structure is
all zeroes right now, reserved for future use.
The Key Value Command Set Identify Namespace data structure is not
all zeroes, however, adding support for Key Value is outside the
scope of this patch.
The new NVME_CTRLR_STATE_IDENTIFY_NS_IOCS_SPECIFIC state is added
after the NVME_CTRLR_STATE_IDENTIFY_ID_DESCS state. This is because
we need to have fetched the identifiers in the desc list in order
to know which command set a namespace belongs to.
A slightly nicer design might have been to refactor the NVMe state
machine to first fetch the id desc list, then the identify namespace
struct, and finally the identify IOCS specific namespace struct.
However, since this would have required a lot of changes, it didn't
really seem justified.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I62cbc533c2c3eec1ccf0ba9b1c414d5a70919cff
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4368
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>
Add a new state in the SPDK NVMe state machine in order to fetch
I/O Command Set Specific Controller data structures.
Right now there is only support for the Zoned Namespace Command Set
Specific Identify Controller data structure.
The NVM Command Set Specific Identify Controller data structure is
all zeroes right now, reserved for future use.
The Key Value Command Set Identify Controller data structure is also
all zeroes right now, reserved for future use.
The new NVME_CTRLR_STATE_IDENTIFY_IOCS_SPECIFIC state is added
after the NVME_CTRLR_STATE_IDENTIFY state. That way, if support for
the Zoned Namespace Command Set is enabled during probing, we will
fetch the Zoned Namespace Command Set Specific Identify Controller data
structure, regardless if any Zoned Namespaces are attached or not, and
no additional steps will be needed once a Zoned Namespace is attached.
Since we only have one command set to fetch, avoid creating
NVME_CTRLR_STATE_IDENTIFY_IOCS_SPECIFIC substates, although that will
probably be needed when support for another command set is added.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I95535b09b03b7ef2ee9a11eebdbd28aad66d65ba
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4367
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>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
When adding an additional state to enum nvme_ctrlr_state, abidiff (1.6.0)
will report that almost every public interface in the nvme library has
been impacted, causing test/make/check_so_deps.sh to fail.
While it is possible that by adding another state, the compiler decides
to use a larger data type for representing enum nvme_ctrlr_state, abidiff
shouldn't complain in the first place, since spdk_nvme_ctrlr is only
ever exposed as an opaque handle. It can never be accessed directly.
Jim Harris suggested to workaround this abidiff bug by changing the type
of spdk_nvme_ctrlr::state from enum to int.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I8b85446580043e95cf791249d643907587e2f982
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4427
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: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This seems to be causing some CI test failures. So
disable zero copy in all cases for now for client
sockets.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iceea09fe65fb90c7df15f500878a473f1ad4152c
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4473
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
LOCAL_SYS_LIBS is meant to define *direct* system
library dependencies for a given library. libuuid
is directly used by the SPDK util library and then
other SPDK libraries use uuid indirectly through
util.
So only the util library should include uuid in
LOCAL_SYS_LIBS.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ia0d2d63f48e6f89891164cf2f9dc4c7a6476d4e3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4366
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
In NVME TCP initiator zero copy is enabled for IO qpairs
and disabled for admin qpairs
Change-Id: Ibdf521dccde9b95ec5dd15a5eb2baed8fcf8b88e
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4211
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>
A preparation step for enabling zero copy in NVMEoF TCP initiator.
With zero copy enabled, some requests might be completed out
of "process_completions" call and we should take them into
account to return the correct number of completions.
Change-Id: Iba7973f6da815645bbfad0334619d46b66379226
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4209
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>
A preparation step for enabling zero copy in NVMEoF TCP initiator.
We should wait for both events to occur before continue qpair
initialization.
Add a new bit to nvme_tcp_qpair::flags to track receiving of icreq ack
since icreq is sent without tcp_req and there is no way to apply
existing synchronization mechanisms.
Move tcp qpair to initializing state if we receive icresp before icreq ack,
this state will be checked during handling of icreq ack to continue
qpair initialization
Change-Id: I7f1ec710d49fb1322eb0a7f133190220b9f585ab
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4207
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>
A preparation step for enabling zero copy in NVMEoF TCP initiator.
Since nvme_tcp_qpair_process_completions doesn't process poll
group, we can't get asycn notification from kernel.
1. Add a qpair to poll group before we send icreq in order to be able
to process buffer reclaim notification.
2. Check if qpair is connected to a poll group and call
nvme_tcp_poll_group_process_completions instead of
nvme_tcp_qpair_process_completions when waiting for icresp
3. Add processing of poll group to nvme_wait_for_completion_timeout
and nvme_wait_for_completion_robust_lock since they are used to
process FABRIC_CONNECT command
Change-Id: I38d2d9496bca8d0cd72e44883df2df802e31a87d
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4208
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>
Currently host/data digest are bool members of nvme_tcp_qpair
structure. Change the type of this members to bitfield, reserved
bits will be used in the next patches to support zero copy.
Change-Id: If0659bf2445901e45fe0816af5f4fca5f494b154
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4206
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>
A preparation step for enabling zero copy in NVMEoF TCP initiator.
Make sure that we complete a request (call user's callback)
when all acknowledgements are received. For write operation -
when we received send cmd ack, h2c ack and response from target.
For read operation - when we received send cmd ack and c2h
completed
Since we can receive send ack after resp command, store
nvme completion received in resp command in a new field
added to tcp_req structure
Change-Id: Id10d506a346738c7a641a979e1c8f86bc07465a4
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4204
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>
A preparation step for enabling zero copy in NVMEoF TCP initiator.
Some NVMEoF TCP targets can send several R2T requests. We should
check that we finished the previous H2C (received buffer reclaim
notification from kernel) before sending the next H2C.
This patch adds a new ordering bit indicating the described case
and 2 fields to nvme_tcp_req to store the values from the last R2T
request which will be applied when send ack is received.
Change-Id: Iaa5ad846712ca18a8382680baa02413c18c4eb37
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4203
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: Ben Walker <benjamin.walker@intel.com>
Meanwhile, to verify an issue about git push unittest failure.
Signed-off-by: yidong0635 <dongx.yi@intel.com>
Change-Id: Idac60e5832390eb8bdce68aee639be2e9ac6cff6
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4373
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Add spdk_nvme_ns_get_ana_group_id() and spdk_nvme_ns_get_ana_state()
to getthe ANA group ID and the ANA state of the given namespace,
respectively.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Id5f1f7ee488a1eb2a7a77f9986a3bb89146628e0
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4354
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Monica Kenguva <monica.kenguva@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Add ana_state and ana_group_id to struct spdk_nvme_ns and keep
them up-to-date by updating when spdk_nvme_ctrlr is created or
ANA change notice is received asynchronously. For both cases,
struct spdk_nvme_ctrlr holds the latest ANA state.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I311fe1c8015c8b8ac9659c38661244706c04b3e3
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4287
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Add an internal API nvme_ctrlr_parse_ana_log_page() to parse an ANA
log page and execute the specified callback function for each
ANA group descriptor in the ANA log page.
We will be able to copy the ANA group descriptor to the caller instead.
To do that, we will need to inform the size of the descriptor first,
but the size will not be constant.
Passing parser to the API will be more convenient.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: Ifd8fda30a83965948017fb8ad992c0d889197cde
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4279
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
When creating a controller, allocate a buffer to the controller
and read ANA log page into the buffer.
When receiving ANA change notice, read ANA log page into the buffer
to keep the contents up to date.
The next patch will provide a public API to get the contents of
ANA log page the controller holds.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: If5c653f4e80d157e5120bb754e6660250b2b8fa1
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4233
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Add an internal API nvme_wait_for_completion_robust_mutex_lock_timeout()
and related internal APIs just call it with adjusting parameters.
nvme_wait_for_completion_robust_lock_timeout() will be usable for
the current use cases of nvme_wait_for_completion_robust_lock() and
future use cases.
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Change-Id: I2b499643930256a39ebe279f56a399f20a7a2fde
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4217
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>
Fix some spelling and make the message clearer
Change-Id: Ib291542a9735d6409db84f16c530e78567123f67
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4249
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Aleksey Marchuk <alexeymar@mellanox.com>
Reviewed-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
With the introduction of namespace types, the identify command has
gained an additional parameter: Command Set Identifier (CSI).
This parameter is similar to the existing parameters NSID and CNTID,
and is not used by all CNS values.
Most notably, the CSI parameter is not used for the existing CNS
values 00h (ID NS) and 01h (ID CTRL).
There are new CNS values, e.g. 05h (ID IOCS specific NS), and
06h (ID IOCS specific CTRL), which do take the new CSI parameter.
The new CNS values instead return Command Set Specific data structures,
which is basically an additional data structure. Therefore, the CNS
values 00h and 01h are very much still in use.
(Even the NVM Command Set has a Command Set Specific data structure,
even though all fields in that data structure are currently reserved.)
Since the CSI parameter is unused by all the existing calls to
nvme_ctrlr_cmd_identify() (since none of the calls send in a CNS value
that uses CSI), simply send in 0 for all existing calls.
No functional change intended.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: Ia2b2324393a0707152b2f8511f0a22ad4a12bd46
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4309
Community-CI: Broadcom CI
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>
There are two different is_active() functions.
spdk_nvme_ctrlr_is_active_ns() which iterates through the active_ns_list,
and spdk_nvme_ns_is_active(), which simply checks the nsdata.
There is an event callback that refreshes active_ns_list when a relevant
events has occured.
In nvme_ns_construct(), nvme_ctrlr_identify_ns() has just been called,
so we know that nsdata is as fresh as possible.
Hence, there is no reason to iterate through a less fresh active_ns_list.
Since we know that the nvme_ctrlr_identify_ns() call was done through the
same controller, we also know that the active/inactive is from the
perspective of the correct controller, so that is not a reason to use the
less efficient is_active() function.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I185f59b53e16e70163e33a3909f4b55ebf631cc4
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4293
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Since the command set identifier might be accessed at several
different states in the nvme state machine, cache it so that
we don't need to loop through the ns id desc list every time.
This is similar to how other identify fields are cached using
nvme_ns_set_identify_data().
None of the identifiers in the desc list (including the new CSI)
can change over the life time of a namespace, so caching them
should be safe.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: Ie06180a4b3750dfa1a42f47afe0f7f9e3ec04ba9
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4266
Community-CI: Broadcom CI
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
If the nvme completion was an error, the function will return,
so there is no reason for an else statement.
In fact, the else statement in nvme_ctrlr_identify_ns_async_done()
differs from the coding style used in other nvme_ctrlr_identify_*
functions, and arguably makes the code harder to read.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: If76b823b7ca04ab98abb2912927c344ee9f12314
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4265
Community-CI: Broadcom CI
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Clear the ns id desc list in nvme_ns_destruct().
Without this, someone can get stale data by calling e.g.
spdk_nvme_ns_get_uuid() on a destructed namespace.
Signed-off-by: Niklas Cassel <niklas.cassel@wdc.com>
Change-Id: I965dd4cd6101d3a77eddbd582b9618b3436d39c8
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4263
Community-CI: Broadcom CI
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
When we disconnect a qpair, part of the code path is
calling _nvme_qpair_abort_queued_reqs. This takes
care of aborting any requests that were queued waiting
for slots to open on the submission queue.
It walks the STAILQ one by one and manually completes
them with ABORT status back to the caller.
But if the callback path submits another request, this
request may also get queued to the end of the queued_req
TAILQ. This can result in an infinite loop.
The solution is to use an STAILQ_SWAP to a local, empty
STAILQ. Then we ensure we only abort the requests that
were queued when _nvme_qpair_abort_queued_reqs() started
executing.
Fixes issue #1588.
I used the multipath.sh test to reproduce this on my local
system. If it ever dropped into the STAILQ loop in this
function, we would hit the infinite loop. With this patch,
I confirmed locally that now we safely avoid the infinite
loop and the test passes.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I657db23efe5983bd8613c870ad62695a7fc7f689
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4284
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: Ziye Yang <ziye.yang@intel.com>
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Rename ordering bit r2t_recv to h2c_send_waiting_ack, that is more
descriptive name.
Change-Id: I6d6143ff4c1cccc74e11226b7974706808092f9a
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4202
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>
This makes it easier to zerofy ordering bits.
Change-Id: If5696bfedfff1bf75e41c1449eac7fccb469e98b
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Signed-off-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4201
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>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
The issue happens when SPDK RDMA initiator is connected to a remote
target and this target reports rather small (or zero) ICD and we try
to send several SGL descriptors.
Since SGL descriptors are located in ICD, we should check that their
total length fits into ICD. In other case sending such a command
will cause RDMA errors (local length error)
Change-Id: I8c0e8375dae799bc442ed2fab249cad2c4ccce51
Signed-off-by: Alexey Marchuk <alexeymar@mellanox.com>
Reported-by: Or Gerlitz <ogerlitz@mellanox.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4131
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
In cases where the SPDK nvme driver is being used as a validation/test
vehicle, users may need to allocate a currently unused qid that can be
used for creating queues using the raw interfaces. One example would be
testing N:1 SQ:CQ mappings which are supported by PCIe controllers but
not through the standard SPDK nvme driver APIs.
These new functions fulfill this purpose, and ensure that the allocated
qid will not be used by the SPDK driver for any future queues allocated
through the spdk_nvme_ctrlr_alloc_io_qpair API.
Signed-off-by: Jacek Kalwas <jacek.kalwas@intel.com>
Change-Id: I21c33596ec415c2816728a600972b242da9d971b
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/3896
Community-CI: Broadcom CI
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>
This may happen when resetting a controller, if the ADMIN queue failed
to reconnect, the controller is set to failed state, so for this case
we don't need to loop until timeout, just exit.
Change-Id: I2b37af5453086cd64f3609c41eb8f6475da55fd4
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.spdk.io/gerrit/c/spdk/spdk/+/4143
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Community-CI: Broadcom CI
Reviewed-by: <dongx.yi@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: JinYu <jin.yu@intel.com>
