When destroy qos, spdk_bdev_qos_destroy() allocates a new qos, and swap
old one. After spdk_bdev_unregister() frees the new qos, the old qos poller
may still reference new qos via bdev->internal.qos. Fix this error by
using old qos in _spdk_bdev_qos_io_submit().
Reported in 72aac51430.1539054028/ubuntu16.04/build.log
Change-Id: Id1bce6c8b1cefae604dd2c69e8f3482ec34b1b54
Signed-off-by: wuzhouhui <wuzhouhui@kingsoft.com>
Reviewed-on: https://review.gerrithub.io/428444
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.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 patch added the support of RPC method to enable,
adjust, disable the bandwidth rate limit on the bdev.
And it can work together with the existing IOPS rate limit.
The RPC method has been consolidated to support both IOPS
and bandwidth rate limits as below:
usage:
rpc.py set_bdev_qos_limit [-h]
[--rw_ios_per_sec RW_IOS_PER_SEC]
[--rw_mbytes_per_sec RW_MBYTES_PER_SEC]
name
positional arguments:
name Blockdev name to set QoS. Example: Malloc0
optional arguments:
-h, --help show this help message and exit
--rw_ios_per_sec RW_IOS_PER_SEC
R/W IOs per second limit (>=10000, example: 20000).
0 means unlimited.
--rw_mbytes_per_sec RW_MBYTES_PER_SEC
R/W megabytes per second limit (>=10, example: 100).
0 means unlimited.
Change-Id: I9c03cd635280add01801a81c6a6c02f0cf85bee1
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/416511
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
scan-build requests to check the size of parent iovec by using
artificially large LBA in unit tests.
Fix the error by using not pointer but position and checking if
position is less than count of parent iovec.
Change-Id: I74c4f6d1b68ecfca93e9247acc5ac6bd5412a960
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/427965
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Each file that need to check SPDK_CONFIG_* options need to include
spdk/config.h explicitly.
Change-Id: If9f2a91ac4c2b1a300dcf88ec3e2a12714ad344a
Signed-off-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-on: https://review.gerrithub.io/427221
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>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
This patch is to introduce the specific QoS related structure
and the enumeration for types of QoS rate limits. Later new
types of QoS rate limits can be supported easily.
Change-Id: Idb8d2e7627fd145bf2b0ddb296c968b6b068f48c
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/424459
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.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>
Currently write to bdev fails in _spdk_bdev_write_zero_buffer_next
due to other than -ENOMEM, assert is called.
RAID bdev using this feature is generally availale now and it will be
OK to remove this assert and return error instead.
Additionally, applying the factored function to _bdev_write_zero_buffer_next
will improve readability slightly.
These two changes are done in this patch.
Change-Id: I462630a71e57e2e5146b085b215d62a378ea9402
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/427186
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
This patch factors out that operation into a function and adds
error handling to that operation to improve readability slightly.
Change-Id: Ic24df0c0a9abbebc38d30fc17779dc5a5f6138a6
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/427026
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Currently when read/write to bdev fails in _spdk_bdev_io_split_with_payload
due to other than -ENOMEM, assert is called.
RAID bdev that utilizes the split IO feature is generally availale now
and it will be OK to remove this assert and return error instead.
Change-Id: I6ea6fd45b94bff0ea84e498e0c4dfd1dd31e0260
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/427025
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
When a bdev IO is split, if iovec size in a strip is more than 32,
the IO will fail.
Remove the limitation by spliting the split IO further.
Change-Id: I962ad86dfe63ea1fcd86ffa52ead7452fb80e53d
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/425876
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Darek Stojaczyk <dariusz.stojaczyk@intel.com>
To achieve its goal, this patch changes the order
in which bdev modules are finished. All modules
that examine bdevs (e.g. lvol,split,...) will be now
finished last. It should not cause any issues though,
since all bdevs are already removed at the time when
any module finish is called
Fixes#387
Change-Id: Id60c375eb5c3d7306b69cdce86bded77354868d8
Signed-off-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-on: https://review.gerrithub.io/421158
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
The type of the second parameter `success' of spdk_bdev_io_completion_cb
is bool. Hence change the code to use bool type success or failure.
Change-Id: I9e93f4ccbb085e8e184f209e706915dcd34aa966
Signed-off-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-on: https://review.gerrithub.io/426648
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>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
We cannot split an iov if a buffer hasn't been
allocated yet. So always call spdk_bdev_io_get_buf
on reads before trying to split.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I2c26efb9dc6cb2c7c3e3b7ae5bab2c37844b9113
Reviewed-on: https://review.gerrithub.io/424879
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
spdk_bdev_io_put_buf() is responsible for reclaiming
bdev-allocated buffers from a bdev_io. If there are
bdev_ios waiting for one of these buffers, it calls
spdk_bdev_io_set_buf() on the next bdev_io in the queue.
This will set the iov_base and iov_len on the bdev_io
to point to the bdev-allocated buffer.
But spdk_bdev_io_put_buf() was calling spdk_bdev_io_set_buf()
on the just completed bdev_io, not the next bdev_io in the
queue. So fix that.
Fixes: 844aedf8 ("bdev: Simplify get/set/put buf functions")
Reported-by: Alan Tu
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ibbcad6e35a3db6991bd7deb3516229572f021638
Reviewed-on: https://review.gerrithub.io/424880
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When new bdev was created, the struct spdk_bdev_module::examine_disk()
may open and close bdev. On the other hand, if something goes wrong,
the creation procedure may unregister new created bdev, so race
condition appeared between _remove_notify() and spdk_bdev_close().
Add the new field "closed" and "remove_notified" in struct spdk_bdev_desc,
so _remove_notify() and spdk_bdev_close() knows how to deal with this
situation.
Change-Id: Ibfe915a4d76096796b039a13a4f49f26669eba2c
Signed-off-by: wuzhouhui <wuzhouhui@kingsoft.com>
Reviewed-on: https://review.gerrithub.io/423369
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>
spdk_bdev_close() must be called on the same thread as
spdk_bdev_open(). Further, the remove callback on the
descriptor will also be run on the same thread as
spdk_bdev_open().
Change-Id: I949d6dd67de1e63d39f06944d473e4aa7134111b
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/424738
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
When emulating write_zeroes commands on device that
don't natively support it, we submit a write with
a zeroed buffer. We used to just reuse the original
bdev_io, but that was recently changed due to other
splitting code added for iovs. But when making those
changes, we forgot to free the bdev_io for the
write that was sent down to the device.
Fixes: 183f37e8 (bdev: do not reuse bdev_io when...)
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If08782c65f6305c0a9f9d15d74fd8823e1158e9b
Reviewed-on: https://review.gerrithub.io/424733
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Xiaodong Liu <xiaodong.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
This RPC does not work for a lot of bdev types. For
example, NVMe namespaces and virtio scsi LUNs are not
explicitly constructed by an RPC - they are indirectly
constructed by an RPC associated with an NVMe controller
or virtio-scsi controller.
While here, remove spdk_bdev_config_json. It was
only created to facilitate this get_bdevs_config RPC.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I537166d8f91ab458bd2000859d74f7254bfc9c0a
Reviewed-on: https://review.gerrithub.io/424584
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This isn't valid RPC so it needs to be removed. Bdev modules were
working around this issue by defining empty write_config_json
methods.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I3c4c20249eadfcfb4103430f5801190b14897249
Reviewed-on: https://review.gerrithub.io/424582
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
This was added a long time back for tracking an rte_mbuf
whose buffer was a different rte_mbuf - all related to
a userspace TCP stack that is no longer in development.
The concept isn't useful now, so remove it to reduce
the complexity of the tracing code.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I310e492eba7f55df242bb29d82fb19f6daee1f51
Reviewed-on: https://review.gerrithub.io/424565
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
This is a string name used for debugging only.
Change-Id: I9827f0e6c83be7bc13951c7b5f0951ce6c2a1ece
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/424127
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>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
There could be cases (especially in virtualized and/or test
environments) where we could accumulate significant skew in
the timeslice frequency. Rather than depend on the application
framework to try to guarantee the rate of timeslice poller
callbacks, keep track internally of the last time the poller
was invoked. If/when we accumulate and detect skew equivalent
to one or more timeslices, increase the allowed IO and bandwidth
of the next timeslice to accomodate.
Since bdev poller now calls spdk_get_ticks() to do accounting,
this patch also fixes up the increment_time() unit test function
and the test env layer to properly increment the fake TSC.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Iba301ddc0fb3d02042106a8bf6e4a6a9a84dc263
Reviewed-on: https://review.gerrithub.io/423580
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Splitting a 1TB unmap into individual 64KB unmap commands
(for a RAID volume with 64KB strip size) would be awful -
the RAID module can be much smarter about this.
So back out the changes for splitting I/O without payload.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I24fe6d911f4e3c9db4b2cb5d66c7236a5596e0d9
Reviewed-on: https://review.gerrithub.io/424103
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Now that we split on I/O boundaries, that code needs to
be able to use the bdev_io split* members to track
what is left to submit. This means that the write_zeroes
code cannot submit the parent bdev_io as the child bdev_io,
since the I/O boundary code will overwrite the write_zeroes
split accounting.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9316b59267508f60799766fc4f1ea05a4b3e5d9e
Reviewed-on: https://review.gerrithub.io/423404
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
This will simplify some future patches which will
account for missed timeslice timers by allowing
additional IO/BW in the following timeslice.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I9dd46a768c98ce267c733a9f9719a2d3d2c3c915
Reviewed-on: https://review.gerrithub.io/423579
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This was defined in two places, so consolidate
the definitions.
Change-Id: I0bbb262b97e90d1064bcc50ee201928f6ca9518a
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/423182
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
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: Jim Harris <james.r.harris@intel.com>
A number of modules (RAID, logical volumes) have logical
"stripes" that require splitting an I/O into several
child I/O. For example, on a RAID-0 with 128KB strip size,
an I/O that spans a 128KB boundary will require sending
one I/O for the portion that comes before the boundary to
one member disk, and another I/O for the portion that comes
after the boundary to another member disk. Logical volumes
are similar - data is allocated in clusters, so an I/O that
spans a cluster boundary may need to be split since the
clusters may not be contiguous on disk.
Putting the splitting logic in the common bdev layer ensures
bdev module authors don't have to always do this themselves.
This is especially helpful for cases like splitting an I/O
described by many iovs - we can simplify this a lot by
handling it in the common bdev layer.
Note that currently we will only submit one child I/O
at a time. This could be improved later to submit multiple
child I/O in parallel, but the complexity in the iov splitting
code also increases a lot.
Note: Some Intel NVMe SSDs have a similar characteristic.
We will not use this bdev stripe feature for NVMe though -
we want to primarily use the splitting functionality inside
of the NVMe driver itself to ensure it remains fully
functional. Many SPDK users use the NVMe driver without
the bdev layer.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Ife804ecc56f6b2b55345a0d0ae9fda9e68632b3b
Reviewed-on: https://review.gerrithub.io/423024
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
This value is used to calculate the disk utilization of a given bdev.
Change-Id: I4bf101c524b92bdd21573941e17f61db59c5c6b8
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/423017
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This will be needed for using this same descriptor when
splitting an I/O.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: Idec759df7ab27f8de567d3c8a4214e25dbe173f5
Reviewed-on: https://review.gerrithub.io/423022
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
So we don't need to allocate memory (maybe failed) just for free other
memory.
Change-Id: I2c83f6acc2aa6ed79455bff90f952a2e70b44d59
Signed-off-by: wuzhouhui <wuzhouhui@kingsoft.com>
Reviewed-on: https://review.gerrithub.io/422203
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@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: Changpeng Liu <changpeng.liu@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
The intents of these arrays was to keep track in the
bdev layer of all base<->virtual bdev relationships -
i.e. which member disk bdevs make up a RAID bdev,
which logical volume bdevs are associated with a
bdev that contains an lvolstore, etc.
Currently none of this is used however. And trying
to keep track in the bdev layer instead of asking
the bdev modules for the relationships has a number
of complications. Early one, we tried to do this
with TAILQs - but that doesn't work since this can't
be done with a single TAILQ_ENTRY in the bdev
structures. So we moved to arrays - that works a bit
better, but then the pointer arrays have to be
realloc'd which isn't ideal.
The biggest problem though with these arrays is that
they held bdev pointers - not bdev descriptor pointers.
It's not really valid to access bdevs without a
descriptor - the descriptors are what make sure active
references are accounted for when a bdev is hotplugged.
Of course the bdev layer knows when a bdev is getting
removed and could go and do the updates to these
arrays separately - but that just seems very convoluted.
So for now just remove these arrays completely. If
there is a future need for the bdev layer to
understand relationships between bdevs, we can add
module APIs so that the generic layer can ask
the modules about the relationships.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I99ef1068240bff1262f64f234260cf2fb44df51d
Reviewed-on: https://review.gerrithub.io/420932
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
This ensures (for example) that a RAID volume is
unregistered before its member disks.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I7a7c16acc351f2d5d4218b64b370e2c77c6e2b5e
Reviewed-on: https://review.gerrithub.io/420812
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
When an SPDK application shuts down, the bdev layer will
automatically unregister all of the bdevs to ensure they
are properly quiesced and cleaned up.
Some modules may want to perform different operations when
a bdev is destructed during normal runtime vs. shutdown.
For example, for lvol, when the last lvol is cleaned up,
it should unload the lvolstore, release and close the bdev
that contains the lvolstore. You never want to do this
during normal runtime though - it is perfectly valid to
have an lvolstore that contains no lvols. RAID and future
bdev modules such as multipath have similar use cases.
So add a new bdev module callback named "fini_start".
If a module specifies a function pointer for this callback,
the bdev layer will call it before it starts the bdev
unregistrations.
This enables some future patches to the bdev layer such
that it will always unregister block devices that are not
claimed (i.e. logical volumes) before block devices that
are claimed (i.e. the bdev containing an lvolstore).
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I6e87f5c2b27f16731ea5def858f26e882a29495a
Reviewed-on: https://review.gerrithub.io/421175
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Claim the block device when adding it to a new Namespace,
and prevent the block device to be added twice for other
modules and Namespaces. Also remove the test that using
same block device over different Namespaces.
Fix issue #371.
Change-Id: Ib7ce18e9fde4a15c0f19ce9e28e69145e54570e0
Signed-off-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-on: https://review.gerrithub.io/420472
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Currently, the bdev layer iterates over all of the
existing channels of a bdev to collect I/O
statistics. But this ignores statistics for
channels that are deleted.
Fix that by keeping an io_stat structure in the
bdev which accumulates statistics for deleted
channels. Use the bdev mutex to protect these
accumulations.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I3103c0b8b55973c827d977765f47e5b9e7f58e5f
Reviewed-on: https://review.gerrithub.io/421029
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
This will be used further in an additional patch.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: I1126e5adfc24f17e5cdf33b0d3e04c78470771c8
Reviewed-on: https://review.gerrithub.io/421027
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
These were left out previously.
Change-Id: I4e97068d2f13ca1dd1cfae1b25564641cee794ef
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/419601
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@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>
Two extra fields are added to the iostat rpc.
1. io_time. The amount of time since queue depth tracking was
enabled that has been spent on I/O processing.
2. weighted_io_time. Incremented each time this bdev's queue depth is
polled by the amount of time spent processing I/O since the last polling
event times the measured queue depth.
Change-Id: Ie70489ec24dee83f3eeac8f4f813ec7074ff458f
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/419031
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Daisuke Aoyama originally contributed to istgt, the
iSCSI target in FreeBSD. The SPDK iSCSI code is
originally derived from that. Due to copy and paste,
some incorrect copyright attributions have been added
to other files that do not derive from istgt, so
this patch removes those.
It is doubtful, at this point, that there is any code
whatsoever that remains from the original istgt, but
we can revisit that at a later time.
Change-Id: I207e1e6b99d271e2f12690be90a96f7d0c113af7
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/420679
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
If a module fails during init, github issue #363 reported ASAN
errors due to leaks from not cleaning up modules that may have
completed init before the failure. An earlier partial fix was
added, this is a more correct fix that addresses the specific
scenario reported as well as the simplified one used during
early debug...
Change-Id: Ia43476faf33d8e31b581529de3b6d75caed09096
Signed-off-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/420118
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Also make the iov_len always set to the used length, not
the total length of the buffer.
Change-Id: I7ebb5b63c6ca7570369f4af0131a23c520c1f7b0
Signed-off-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-on: https://review.gerrithub.io/419025
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Maciej Szwed <maciej.szwed@intel.com>
Reviewed-by: Dariusz Stojaczyk <dariuszx.stojaczyk@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
In the event that one bdev module failed, we'd leak a bunch
of stuff from any that init'd correctly beforehand. Also
added a guard around the calling of modile init_done routines
so that it's not done if module init didn't work.
Change-Id: I4e6170e1eee67b131252ed30d0d20124d2c5ff35
Signed-off-by: Paul Luse <paul.e.luse@intel.com>
Reviewed-on: https://review.gerrithub.io/419446
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
This function always return 0, so change the type of function
to void
Change-Id: I715cf82c74619a2707b8e5a453710a992489f2c8
Signed-off-by: Ziye Yang <optimistyzy@gmail.com>
Reviewed-on: https://review.gerrithub.io/419045
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
When spdk_put_io_channel is called, if its the last reference,
we defer actual destruction of the channel, so that code
in the same context which may be referring to the channel
doesn't crash.
But it is possible that an io_channel for that same io_device
could be requested before the deferred message is processed.
This would result in a second io_channel being created for
that device on the same thread.
To avoid this case, don't immediately remove the channel from
the list when the last reference is put. When the deferred
message is processed, if additional references were allocated
in the meantime, don't destroy the channel.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Signed-off-by: Ziye Yang <ziye.yang@intel.com>
Change-Id: Idb8d4705fda0eb9c338e4960430e04edbe537e05
Reviewed-on: https://review.gerrithub.io/418878
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Chandler-Test-Pool: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: GangCao <gang.cao@intel.com>
Reviewed-by: Pawel Wodkowski <pawelx.wodkowski@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
Reviewed-by: Changpeng Liu <changpeng.liu@intel.com>
Considering the I/O size is variant from small in byte to
large in megabyte, need to consider the compensation of less
allowed bytes in next timeslice if the current timeslice
sends more bytes down.
Change-Id: I885f0bb21001bd90879aa8622e2b34e3bf78cf6e
Signed-off-by: GangCao <gang.cao@intel.com>
Reviewed-on: https://review.gerrithub.io/417829
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Chandler-Test-Pool: 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>
This change includes a function to enable this feature on a per-bdev
basis. The new information stored in the bdev includes the following:
measured_queue_depth: The aggregate of the outstanding oparations from
each channel associated with this bdev.
period: The period at which this bdev's measured_queue_depth is being
updated.
With this information, one could calculate the average queue depth and
the disk utilization of the device
Change-Id: Ie0623ee4796e33b125504fb0965d5ef348cbff7d
Signed-off-by: Seth Howell <seth.howell@intel.com>
Reviewed-on: https://review.gerrithub.io/418102
Chandler-Test-Pool: 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>
Tested-by: SPDK CI Jenkins <sys_sgci@intel.com>
Rather than adding the full alignment size (512), add one less than the
alignment so that already-aligned buffers don't get rounded up again.
Change-Id: I96323b848bfb90f2aa1774b869e2b8a81d253077
Reported-by: Shuhei Matsumoto
Signed-off-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-on: https://review.gerrithub.io/418879
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>
Reviewed-by: Shuhei Matsumoto <shuhei.matsumoto.xt@hitachi.com>
During spdk_bdev_init, examine_config is called.
This call can claim bdev synchronously, based on
configuration. On spdk_bdev_start if none module
claimed bdev, examine_disk is called and can
perform I/O before claiming bdev.
Signed-off-by: Piotr Pelplinski <piotr.pelplinski@intel.com>
Change-Id: I1448dd368cf3a24a5daccab387d7af7c3d231127
Reviewed-on: https://review.gerrithub.io/413913
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Maciej Szwed <maciej.szwed@intel.com>
Reviewed-by: Jim Harris <james.r.harris@intel.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
There's no reason to abort IO that have been queued
due to QoS limits, when QoS is switched from enabled
to disabled. Submit them to the bdev instead.
Fixes issue #357.
Signed-off-by: Jim Harris <james.r.harris@intel.com>
Change-Id: If5eafc53418ac686120e1d6a1da884b42cef845e
Reviewed-on: https://review.gerrithub.io/418128
Tested-by: SPDK Automated Test System <sys_sgsw@intel.com>
Reviewed-by: Seth Howell <seth.howell5141@gmail.com>
Reviewed-by: Daniel Verkamp <daniel.verkamp@intel.com>
Reviewed-by: Ben Walker <benjamin.walker@intel.com>
