Currently deleting bdev does not support asynchronous delete
operations. Because of that results are returned before device
is actually deleted and some operation can be peformed on that
device after removal of this device started.
Signed-off-by: Maciej Szwed <maciej.szwed@intel.com>
Change-Id: I305c302d8abd5d7c2c0f947fca70c58396872132
Reviewed-on: https://review.gerrithub.io/383732
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Always start bdev pollers on the calling core.
This removes the lcore concept from the bdev poller abstraction and
simplifies the job of spdk_bdev_initialize() callers providing their own
poller and event implementations.
All callers except the NVMe bdev hotplug poller already used the current
core as the parameter. The NVMe HotplugPollCore option was undocumented
and unused in any of the tests or example configuration files, so it
should be safe to remove.
Change-Id: I93b466e1e58901b8785c40cbe296fa46c157850f
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/382857
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
At very high queue depths, bdev modules may not have enough
internal resources to track all of the incoming I/O. For example,
we allocate a finite number of nvme_request objects per allocated
queue pair. Currently if these resources are exhausted, the
bdev module will return failure (with no indication why) which
gets propagated all the way back to the application.
So instead, add SPDK_BDEV_IO_STATUS_NOMEM to allow bdev modules
to indicate this type of failure. Also add handling for this
status type in the generic bdev layer, involving queuing these
I/O for later retry after other I/O on the failing channel have
completed.
This does place an expectation on the bdev module that these
internal resources are allocated per io_channel. Otherwise we
cannot guarantee forward progress solely on reception of
completions. For example, without this guarantee, a bdev
module could theoretically return ENOMEM even if there were
no I/O oustanding for that io_channel. nvme, aio, rbd,
virtio and null drivers comply with this expectation already.
malloc only complies though when not using copy offload.
This patch will fix malloc w/ copy engine to at least
return ENOMEM when no copy descriptors are available. If the
condition above occurs, I/O waiting for resources will get
failed as part of a subsequent reset which matches the
behavior it has today.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iea7cd51a611af8abe882794d0b2361fdbb74e84e
Reviewed-on: https://review.gerrithub.io/378853
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This is required for upcoming UNMAP
implementation in bdev_virtio.
While here, also added documentation for
spdk_bdev_io_get_buf().
Change-Id: Ia769ee9b8b132f31208ae66598b29a1c9ed37312
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/379721
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
spdk_bdev_unmap_blocks() accepts a 64-bit number of blocks, which can
exceed the NVMe Dataset Management range's 32-bit number of blocks,
which can represent up to 2 TB with 512-byte blocks.
We can support up to 0.5 PB unmap requests by using the maximum number
of descriptors in a single Dataset Management command, which should be
sufficient for now.
Change-Id: I0a4ee77a9be148355991e1a081007ffa020a3ee5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/379202
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Change-Id: I46d564523ea20bc7c79a3896dfb024d3d5172a93
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/379911
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Some of the internal functions used to return the
number of bytes associated with a successfully submitted
IO, but that was changed a while ago to just return 0
for success.
So change some of the callers of these functions to
just look for != 0 for failure rather than < 0. This
preps for some upcoming ENOMEM handling changes.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I0a66dd6bfac9053e0fd6103dee1ea36b20d902df
Reviewed-on: https://review.gerrithub.io/378856
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Since all direct bdev_io types have the same layout,
there is no need to keep them differentiated.
Change-Id: If8bb85e43c9922c0ebfc39837e3a45006e508b56
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/377686
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
I inadvertently misordered the arguments in a call to spdk_bdev_unmap.
This call is unused as of this patch, but is used by future patches.
Change-Id: I6042a3f4e05bded21b0ceb4d3b104856bcbf8c12
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/377961
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
There remove callback function will remove the nvme_bdev from the global
list, we do not need to remove the device in the function bdev_nvme_destruct().
just make sure to remove it from list when the app exit.
Change-Id: I1859bfd696ed9c0ca3ac1cd8ffadfd9488df0fcd
Signed-off-by: Cunyin Chang <cunyin.chang@intel.com>
Reviewed-on: https://review.gerrithub.io/375941
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
The bdev modules now take all read, write, unmap, and flush requests in
terms of blocks rather than bytes.
The public bdev APIs still accept offset and length in bytes for now.
Change-Id: I57f0955d52272f57755f0ff4dbc56721fdc2ef51
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/376037
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
This matches the name to the behavior and prepares for addition of a new
log macro for "info" log level.
Change-Id: I94ccd49face4309d3368e399528776ab140748c4
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/375833
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This makes it consistent it with other parts of this structure.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ie36488813b53ce20663c50a5c9f049c4c9723d3a
Reviewed-on: https://review.gerrithub.io/375494
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Allow passing the NVMe namespace optimal I/O boundary through the bdev
layer.
Change-Id: I27a2d5498df56775d3330e40c31bd7c23bbc77a5
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/374532
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
When users don't enable hotplug option in their configuration
section, SPDK will enable it by default. DPDK will print probing
messages continuously for NVMe devices which don't belong to SPDK.
Change-Id: I8c43335a282ecba206b4b5305bd881d2bd07836e
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/374486
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
In this iteration, we only support write_zeroes in the form of a
deallocate call that returns all zeroes.
Change-Id: Ica837ce70672174df63012719de60463fdb799cf
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/372005
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Change-Id: Ia2202eadfb8140b7a51dd64d4241e85aceff361c
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/373408
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Remove the redundant sector size check; the generic bdev code already
checks for this.
Also use the bdev blocklen field for both offset and size calculations.
The bdev blocklen is the same as the namespace sector size.
Change-Id: Ia8061eb4cfc229d4b6fbe2caabf2dd81656bc697
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/372862
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This is far simpler, although it does limit the bdev
layer to unmapped just one range per command. In practice,
all of our code reports limits of just one range per command
anyway.
Change-Id: I99247ab349fe85b9925769e965833b06708d0d70
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/370382
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Patch afe860ae deferred freeing the io_device. However, for nvme, the
io_device context (spdk_nvme_ctrlr) is still being destructed before
io_channels are destroyed, causing segfaults on hotremove.
This patch defers io_device context destruction and fixes nvme
hotremove.
Fixes: afe860aeb1 ("channel: Correctly defer unregisters if channels exist")
Fixes: 5533c3d208 ("util: defer put_io_channel")
Change-Id: I7af699174cac0c6c6a6faa2cc65418c47347eb9a
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/370459
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Add a new struct spdk_nvme_io_qpair_opts to allow the user to override
controller options on a per-I/O qpair basis.
Existing callers with qprio == 0 can be updated to:
... = spdk_nvme_ctrlr_alloc_io_qpair(ctrlr, NULL, 0);
Callers that need to specify a non-default qprio should be updated to:
struct spdk_nvme_io_qpair_opts opts;
spdk_nvme_ctrlr_get_default_io_qpair_opts(ctrlr, &opts, sizeof(opts));
opts.qprio = SPDK_NVME_QPRIO_...;
... = spdk_nvme_ctrlr_alloc_io_qpair(ctrlr, &opts, sizeof(opts));
Change-Id: I8ac3ea369535cfde759abbe75e1d974b6450a800
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/369676
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Change-Id: Ie25a87c4b3f781299fa744fdcff6c9a63d473935
Signed-off-by: Roman <roman.sudarikov@intel.com>
Reviewed-on: https://review.gerrithub.io/365723
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
We still will sort the bdev_module list so that modules
with an examine() callback are initialized first. This ensures
they have a chance to initialize before later modules start
registering physical block devices.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I792cfb41b0abe030fe2486a2c872cbf329735932
Reviewed-on: https://review.gerrithub.io/369486
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Module initializaiton was made asynchronous recently
to support bdev modules like gpt which need to do
asynchronous I/O. But all modules now do any
asynchronous I/O in their examine() routines, and
init functions only do very basic operations to be
ready to handle examine() callbacks.
So simplify the bdev code and modules to go back to
a synchronous init procedure.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Idb16156796ad7511d00f465d7a2db9acda6315b6
Reviewed-on: https://review.gerrithub.io/369485
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
This includes file names, functions, #defines, etc.
There are still a few uses of "blockdev" outside of
include/ and lib/ - these can be handled later.
This preps for a future patch to consolidate vbdev
modules and bdev modules into just bdev modules.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I70e575709ae1b0a116b08515fd38ae793de05377
Reviewed-on: https://review.gerrithub.io/369325
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
