- 13 Nov, 2019 8 commits
-
-
Christoph Hellwig authored
Make the instance name match the method name and define the name to NULL instead of providing an inline stub, which is rather pointless for a method call. Signed-off-by:
Christoph Hellwig <hch@lst.de> Signed-off-by:
Damien Le Moal <damien.lemoal@wdc.com> Reviewed-by:
Hannes Reinecke <hare@suse.de> Reviewed-by:
Chaitanya Kulkarni <chaitanya.kulkarni@wdc.com> Signed-off-by:
Jens Axboe <axboe@kernel.dk>
-
Christoph Hellwig authored
Remove the pointless stub open and release methods, give the operations vector a slightly less confusing name, and use normal alignment for the assignment operators. Signed-off-by:
-
Damien Le Moal authored
No known partitioning tool supports zoned block devices, especially the host managed flavor with strong sequential write constraints. Furthermore, there are also no known user nor use cases for partitioned zoned block devices. This patch removes partition device creation for zoned block devices, which allows simplifying the processing of zone commands for zoned block devices. A warning is added if a partition table is found on the device. For report zones operations no zone sector information remapping is necessary anymore, simplifying the code. Of note is that remapping of zone reports for DM targets is still necessary as done by dm_remap_zone_report(). Similarly, remaping of a zone reset bio is not necessary anymore. Testing for the applicability of the zone reset all request also becomes simpler and only needs to check that the number of sectors of the requested zone range is equal to the disk capacity. Reviewed-by:
-
Damien Le Moal authored
All kernel users of blkdev_report_zones() as well as applications use through ioctl(BLKZONEREPORT) expect to potentially get less zone descriptors than requested. As such, the use of the internal report zones command execution loop implemented by blk_report_zones() is not necessary and can even be harmful to performance by causing the execution of inefficient small zones report command to service the reminder of a requested zone array. This patch removes blk_report_zones(), simplifying the code. Also remove a now incorrect comment in dm_blk_report_zones(). Signed-off-by:
Javier Gonzalez <javier@javigon.com> Reviewed-by:
-
Christoph Hellwig authored
blk_revalidate_disk_zones is never called for non-zoned devices. Just return early and warn instead of trying to handle this case. Signed-off-by:
-
Damien Le Moal authored
For ZBC and ZAC zoned devices, the scsi driver revalidation processing implemented by sd_revalidate_disk() includes a call to sd_zbc_read_zones() which executes a full disk zone report used to check that all zones of the disk are the same size. This processing is followed by a call to blk_revalidate_disk_zones(), used to initialize the device request queue zone bitmaps (zone type and zone write lock bitmaps). To do so, blk_revalidate_disk_zones() also executes a full device zone report to obtain zone types. As a result, the entire zoned block device revalidation process includes two full device zone report. By moving the zone size checks into blk_revalidate_disk_zones(), this process can be optimized to a single full device zone report, leading to shorter device scan and revalidation times. This patch implements this optimization, reducing the original full device zone report implemented in sd_zbc_check_zones() to a single, small, report zones command execution to obtain the size of the first zone of the device. Checks whether all zones of the device are the same size as the first zone size are moved to the generic blk_check_zone() function called from blk_revalidate_disk_zones(). This optimization also has the following benefits: 1) fewer memory allocations in the scsi layer during disk revalidation as the potentailly large buffer for zone report execution is not needed. 2) Implement zone checks in a generic manner, reducing the burden on device driver which only need to obtain the zone size and check that this size is a power of 2 number of LBAs. Any new type of zoned block device will benefit from this. Signed-off-by:
-
Jens Axboe authored
* for-5.5/drivers-post: scsi: sd_zbc: add zone open, close, and finish support scsi: core: Handle drivers which set sg_tablesize to zero scsi: qla2xxx: fix NPIV tear down process scsi: sd_zbc: Fix sd_zbc_complete() scsi: qla2xxx: stop timer in shutdown path scsi: sd: define variable dif as unsigned int instead of bool scsi: target: cxgbit: Fix cxgbit_fw4_ack() scsi: qla2xxx: Fix partial flash write of MBI scsi: qla2xxx: Initialized mailbox to prevent driver load failure scsi: lpfc: Honor module parameter lpfc_use_adisc scsi: ufs-bsg: Wake the device before sending raw upiu commands scsi: lpfc: Check queue pointer before use scsi: qla2xxx: fixup incorrect usage of host_byte
-
Jens Axboe authored
* for-5.5/drivers: (38 commits) null_blk: add zone open, close, and finish support dm: add zone open, close and finish support nvme: Fix parsing of ANA log page nvmet: stop using bio_set_op_attrs nvmet: add plugging for read/write when ns is bdev nvmet: clean up command parsing a bit nvme-pci: Spelling s/resdicovered/rediscovered/ nvmet: fill discovery controller sn, fr and mn correctly nvmet: Open code nvmet_req_execute() nvmet: Remove the data_len field from the nvmet_req struct nvmet: Introduce nvmet_dsm_len() helper nvmet: Cleanup discovery execute handlers nvmet: Introduce common execute function for get_log_page and identify nvmet-tcp: Don't set the request's data_len nvmet-tcp: Don't check data_len in nvmet_tcp_map_data() nvme: Introduce nvme_lba_to_sect() nvme: Cleanup and rename nvme_block_nr() nvme: resync include/linux/nvme.h with nvmecli nvme: move common call to nvme_cleanup_cmd to core layer nvme: introduce "Command Aborted By host" status code ...
-
- 08 Nov, 2019 2 commits
-
-
Ming Lei authored
64K PAGE_SIZE is popular on ARM64 or other ARCHs, and 64K has been big enough to break some devices probably, so change the logic to split bio if the only bvec's length is > SZ_4K instead of PAGE_SIZE. Fixes: fa532287 (block: avoid blk_bio_segment_split for small I/O operations) Cc: Christoph Hellwig <hch@lst.de> Signed-off-by:
Ming Lei <ming.lei@redhat.com> Signed-off-by:
-
Ming Lei authored
Some device may set segment boundary as PAGE_SIZE - 1. If the bvec crosses pages, and meantime its length is <= PAGE_SIZE, we still need to split the bvec into 2 segments. Fixes this issue by still splitting bio if the single bvec crosses pages. Reported-by:
kernel test robot <lkp@intel.com> Fixes: fa532287 (block: avoid blk_bio_segment_split for small I/O operations) Cc: Christoph Hellwig <hch@lst.de> Signed-off-by:
-
- 07 Nov, 2019 18 commits
-
-
Tejun Heo authored
blkg_rwstat is now only used by bfq-iosched and blk-throtl when on cgroup1. Let's move it into its own files and gate it behind a config option. Signed-off-by:
Tejun Heo <tj@kernel.org> Signed-off-by:
-
Tejun Heo authored
blk-cgroup has been using blkg_rwstat to track basic IO stats. Unfortunately, reading recursive stats scales badly as itinvolves walking all descendants. On systems with a huge number of cgroups (dead or alive), this can lead to substantial CPU cost when reading IO stats. This patch reimplements basic IO stats using cgroup rstat which uses more memory but makes recursive stat reading O(# descendants which have been active since last reading) instead of O(# descendants). * blk-cgroup core no longer uses sync/async stats. Introduce new stat enums - BLKG_IOSTAT_{READ|WRITE|DISCARD}. * Add blkg_iostat[_set] which encapsulates byte and io stats, last values for propagation delta calculation and u64_stats_sync for correctness on 32bit archs. * Update the new percpu stat counters directly and implement blkcg_rstat_flush() to implement propagation. * blkg_print_stat() can now bring the stats up to date by calling cgroup_rstat_flush() and print them instead of directly summing up all descendants. * It now allocates 96 bytes per cpu. It used to be 40 bytes. Signed-off-by: -
Tejun Heo authored
These don't have users anymore. Remove them. Signed-off-by:
-
Tejun Heo authored
When used on cgroup1, blk-throtl uses the blkg->stat_bytes and ->stat_ios from blk-cgroup core to populate four stat knobs. blk-cgroup core is moving away from blkg_rwstat to improve scalability and won't be able to support this usage. It isn't like the sharing gains all that much. Let's break them out to dedicated rwstat counters which are updated when on cgroup1. Signed-off-by:
-
Tejun Heo authored
When used on cgroup1, bfq uses the blkg->stat_bytes and ->stat_ios from blk-cgroup core to populate six stat knobs. blk-cgroup core is moving away from blkg_rwstat to improve scalability and won't be able to support this usage. It isn't like the sharing gains all that much. Let's break it out to dedicated rwstat counters which are updated when on cgroup1. This makes use of bfqg_*rwstat*() helpers outside of CONFIG_BFQ_CGROUP_DEBUG. Move them out. v2: Compile fix when !CONFIG_BFQ_CGROUP_DEBUG. Signed-off-by:
-
Tejun Heo authored
Collect them right under #ifdef CONFIG_BFQ_CGROUP_DEBUG. The next patch will use them from !DEBUG path and this makes it easy to move them out of the ifdef block. This is pure code reorganization. Signed-off-by:
-
Jens Axboe authored
Pull on for-linus to resolve what otherwise would have been a conflict with the cgroups rstat patchset from Tejun. * for-linus: (942 commits) blkcg: make blkcg_print_stat() print stats only for online blkgs nvme: change nvme_passthru_cmd64 to explicitly mark rsvd nvme-multipath: fix crash in nvme_mpath_clear_ctrl_paths nvme-rdma: fix a segmentation fault during module unload iocost: don't nest spin_lock_irq in ioc_weight_write() io_uring: ensure we clear io_kiocb->result before each issue um-ubd: Entrust re-queue to the upper layers nvme-multipath: remove unused groups_only mode in ana log nvme-multipath: fix possible io hang after ctrl reconnect io_uring: don't touch ctx in setup after ring fd install io_uring: Fix leaked shadow_req Linux 5.4-rc5 riscv: cleanup do_trap_break nbd: verify socket is supported during setup ata: libahci_platform: Fix regulator_get_optional() misuse nbd: handle racing with error'ed out commands nbd: protect cmd->status with cmd->lock io_uring: fix bad inflight accounting for SETUP_IOPOLL|SETUP_SQTHREAD io_uring: used cached copies of sq->dropped and cq->overflow ARM: dts: stm32: relax qspi pins slew-rate for stm32mp157 ...
-
Ajay Joshi authored
Implement REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH support to allow explicit control of zone states. Contains contributions from Matias Bjorling, Hans Holmberg, Keith Busch and Damien Le Moal. Reviewed-by:
Martin K. Petersen <martin.petersen@oracle.com> Reviewed-by:
Ajay Joshi <ajay.joshi@wdc.com> Signed-off-by:
Matias Bjorling <matias.bjorling@wdc.com> Signed-off-by:
Hans Holmberg <hans.holmberg@wdc.com> Signed-off-by:
Keith Busch <kbusch@kernel.org> Signed-off-by:
-
git://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsiJens Axboe authored
Merge branch 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi into for-5.5/drivers-post * 'fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/mkp/scsi: scsi: core: Handle drivers which set sg_tablesize to zero scsi: qla2xxx: fix NPIV tear down process scsi: sd_zbc: Fix sd_zbc_complete()
-
Jens Axboe authored
Merge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi into for-5.5/drivers-post SCSI fixes on 20191101 Nine changes, eight in drivers [ufs, target, lpfc x 2, qla2xxx x 4] and one core change in sd that fixes an I/O failure on DIF type 3 devices. Signed-off-by:
James E.J. Bottomley <jejb@linux.ibm.com> * tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi: (23 commits) scsi: qla2xxx: stop timer in shutdown path scsi: sd: define variable dif as unsigned int instead of bool scsi: target: cxgbit: Fix cxgbit_fw4_ack() scsi: qla2xxx: Fix partial flash write of MBI scsi: qla2xxx: Initialized mailbox to prevent driver load failure scsi: lpfc: Honor module parameter lpfc_use_adisc scsi: ufs-bsg: Wake the device before sending raw upiu commands scsi: lpfc: Check queue pointer before use scsi: qla2xxx: fixup incorrect usage of host_byte scsi: lpfc: remove left-over BUILD_NVME defines scsi: core: try to get module before removing device scsi: hpsa: add missing hunks in reset-patch scsi: target: core: Do not overwrite CDB byte 1 scsi: ch: Make it possible to open a ch device multiple times again scsi: fix kconfig dependency warning related to 53C700_LE_ON_BE scsi: sni_53c710: fix compilation error scsi: scsi_dh_alua: handle RTPG sense code correctly during state transitions scsi: qla2xxx: fix a potential NULL pointer dereference scsi: MAINTAINERS: Update qla2xxx driver scsi: zfcp: fix reaction on bit error threshold notification ...
-
Ajay Joshi authored
Implement REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH support to allow explicit control of zone states. Contains contributions from Matias Bjorling, Hans Holmberg, Keith Busch and Damien Le Moal. Reviewed-by:
-
Ajay Joshi authored
Implement REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH support to allow explicit control of zone states. Contains contributions from Matias Bjorling, Hans Holmberg and Damien Le Moal. Acked-by:
Mike Snitzer <snitzer@redhat.com> Reviewed-by:
-
Jens Axboe authored
Pull in dependencies for the new zoned open/close/finish support. * for-5.5/block: (32 commits) block: add zone open, close and finish ioctl support block: add zone open, close and finish operations block: Simplify REQ_OP_ZONE_RESET_ALL handling block: Remove REQ_OP_ZONE_RESET plugging block: Warn if elevator= parameter is used block: avoid blk_bio_segment_split for small I/O operations blk-mq: make sure that line break can be printed block: sed-opal: Introduce Opal Datastore UID block: sed-opal: Add support to read/write opal tables generically block: sed-opal: Generalizing write data to any opal table bdev: Refresh bdev size for disks without partitioning bdev: Factor out bdev revalidation into a common helper blk-mq: avoid sysfs buffer overflow with too many CPU cores blk-mq: Make blk_mq_run_hw_queue() return void fcntl: fix typo in RWH_WRITE_LIFE_NOT_SET r/w hint name blk-mq: fill header with kernel-doc blk-mq: remove needless goto from blk_mq_get_driver_tag block: reorder bio::__bi_remaining for better packing block: Reduce the amount of memory used for tag sets block: Reduce the amount of memory required per request queue ...
-
Ajay Joshi authored
Introduce three new ioctl commands BLKOPENZONE, BLKCLOSEZONE and BLKFINISHZONE to allow applications to control the condition of zones on a zoned block device through the execution of the REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH operations. Contains contributions from Matias Bjorling, Hans Holmberg, Dmitry Fomichev, Keith Busch, Damien Le Moal and Christoph Hellwig. Reviewed-by:
Dmitry Fomichev <dmitry.fomichev@wdc.com> Signed-off-by:
-
Ajay Joshi authored
Zoned block devices (ZBC and ZAC devices) allow an explicit control over the condition (state) of zones. The operations allowed are: * Open a zone: Transition to open condition to indicate that a zone will actively be written * Close a zone: Transition to closed condition to release the drive resources used for writing to a zone * Finish a zone: Transition an open or closed zone to the full condition to prevent write operations To enable this control for in-kernel zoned block device users, define the new request operations REQ_OP_ZONE_OPEN, REQ_OP_ZONE_CLOSE and REQ_OP_ZONE_FINISH as well as the generic function blkdev_zone_mgmt() for submitting these operations on a range of zones. This results in blkdev_reset_zones() removal and replacement with this new zone magement function. Users of blkdev_reset_zones() (f2fs and dm-zoned) are updated accordingly. Contains contributions from Matias Bjorling, Hans Holmberg, Dmitry Fomichev, Keith Busch, Damien Le Moal and Christoph Hellwig. Reviewed-by:
-
Damien Le Moal authored
There is no need for the function __blkdev_reset_all_zones() as REQ_OP_ZONE_RESET_ALL can be handled directly in blkdev_reset_zones() bio loop with an early break from the loop. This patch removes this function and modifies blkdev_reset_zones(), simplifying the code. Reviewed-by:
-
Damien Le Moal authored
REQ_OP_ZONE_RESET operations cannot be merged as these bios and requests do not have a size and are never sequential due to the zone start sector position required for their execution. As a result, there is no point in using a plug around blkdev_reset_zones() bio issuing loop. This patch removes this unnecessary plugging. Reviewed-by:
-
Tejun Heo authored
blkcg_print_stat() iterates blkgs under RCU and doesn't test whether the blkg is online. This can call into pd_stat_fn() on a pd which is still being initialized leading to an oops. The heaviest operation - recursively summing up rwstat counters - is already done while holding the queue_lock. Expand queue_lock to cover the other operations and skip the blkg if it isn't online yet. The online state is protected by both blkcg and queue locks, so this guarantees that only online blkgs are processed. Signed-off-by:
Roman Gushchin <guro@fb.com> Cc: Josef Bacik <jbacik@fb.com> Fixes: 903d23f0 ("blk-cgroup: allow controllers to output their own stats") Cc: stable@vger.kernel.org # v4.19+ Signed-off-by:
-
- 06 Nov, 2019 4 commits
-
-
Jan Kara authored
With transition to blk-mq, the elevator= kernel argument was removed as it makes less and less sense with the current variety of devices. Since this may surprise some users and there are advices on the Internet that still suggest to use it, let's at least warn if the parameter is used. Reviewed-by:
Jeff Moyer <jmoyer@redhat.com> Signed-off-by:
Jan Kara <jack@suse.cz> Signed-off-by:
-
Michael Schmitz authored
In scsi_mq_setup_tags(), cmd_size is calculated based on zero size for the scatter-gather list in case the low level driver uses SG_NONE in its host template. cmd_size is passed on to the block layer for calculation of the request size, and we've seen NULL pointer dereference errors from the block layer in drivers where SG_NONE is used and a mq IO scheduler is active, apparently as a consequence of this (see commit 68ab2d76 ("scsi: cxlflash: Set sg_tablesize to 1 instead of SG_NONE"), and a recent patch by Finn Thain converting the three m68k NFR5380 drivers to avoid setting SG_NONE). Try to avoid these errors by accounting for at least one sg list entry when calculating cmd_size, regardless of whether the low level driver set a zero sg_tablesize. Tested on 030 m68k with the atari_scsi driver - setting sg_tablesize to SG_NONE no longer results in a crash when loading this driver. CC: Finn Thain <fthain@telegraphics.com.au> Link: https://lore.kernel.org/r/1572922150-4358-1-git-send-email-schmitzmic@gmail.comSigned-off-by:
Michael Schmitz <schmitzmic@gmail.com> Signed-off-by:
-
Martin Wilck authored
Fix two issues with commit f5187b7d ("scsi: qla2xxx: Optimize NPIV tear down process"): a missing negation in a wait_event_timeout() condition, and a missing loop end condition. Fixes: f5187b7d ("scsi: qla2xxx: Optimize NPIV tear down process") Link: https://lore.kernel.org/r/20191105145550.10268-1-martin.wilck@suse.comSigned-off-by:
Martin Wilck <mwilck@suse.com> Acked-by:
Himanshu Madhani <hmadhani@marvell.com> Signed-off-by:
-
Damien Le Moal authored
The ILLEGAL REQUEST/INVALID FIELD IN CDB error generated by an attempt to reset a conventional zone does not apply to the reset write pointer command with the ALL bit set, that is, to REQ_OP_ZONE_RESET_ALL requests. Fix sd_zbc_complete() to be quiet only in the case of REQ_OP_ZONE_RESET, excluding REQ_OP_ZONE_RESET_ALL. Since REQ_OP_ZONE_RESET is the only request handled by sd_zbc_complete(), also simplify the code using a simple if statement. [mkp: applied by hand] Fixes: d81e9d49 ("scsi: implement REQ_OP_ZONE_RESET_ALL") Cc: stable@vger.kernel.org Link: https://lore.kernel.org/r/20191027140549.26272-4-damien.lemoal@wdc.comSigned-off-by:
-
- 05 Nov, 2019 5 commits
-
-
git://git.infradead.org/nvmeJens Axboe authored
Pull NVMe fixes from Keith: "We have a few late nvme fixes for a couple device removal kernel crashes, and a compat fix for a new ioctl introduced during this merge window." * 'nvme-5.4-rc7' of git://git.infradead.org/nvme: nvme: change nvme_passthru_cmd64 to explicitly mark rsvd nvme-multipath: fix crash in nvme_mpath_clear_ctrl_paths nvme-rdma: fix a segmentation fault during module unload
-
Charles Machalow authored
Changing nvme_passthru_cmd64 to add a field: rsvd2. This field is an explicit marker for the padding space added on certain platforms as a result of the enlargement of the result field from 32 bit to 64 bits in size, and fixes differences in struct size when using compat ioctl for 32-bit binaries on 64-bit architecture. Fixes: 65e68edc ("nvme: allow 64-bit results in passthru commands") Reviewed-by:
Charles Machalow <csm10495@gmail.com> [changelog] Signed-off-by:
-
Anton Eidelman authored
nvme_mpath_clear_ctrl_paths() iterates through the ctrl->namespaces list while holding ctrl->scan_lock. This does not seem to be the correct way of protecting from concurrent list modification. Specifically, nvme_scan_work() sorts ctrl->namespaces AFTER unlocking scan_lock. This may result in the following (rare) crash in ctrl disconnect during scan_work: BUG: kernel NULL pointer dereference, address: 0000000000000050 Oops: 0000 [#1] SMP PTI CPU: 0 PID: 3995 Comm: nvme 5.3.5-050305-generic RIP: 0010:nvme_mpath_clear_current_path+0xe/0x90 [nvme_core] ... Call Trace: nvme_mpath_clear_ctrl_paths+0x3c/0x70 [nvme_core] nvme_remove_namespaces+0x35/0xe0 [nvme_core] nvme_do_delete_ctrl+0x47/0x90 [nvme_core] nvme_sysfs_delete+0x49/0x60 [nvme_core] dev_attr_store+0x17/0x30 sysfs_kf_write+0x3e/0x50 kernfs_fop_write+0x11e/0x1a0 __vfs_write+0x1b/0x40 vfs_write+0xb9/0x1a0 ksys_write+0x67/0xe0 __x64_sys_write+0x1a/0x20 do_syscall_64+0x5a/0x130 entry_SYSCALL_64_after_hwframe+0x44/0xa9 RIP: 0033:0x7f8d02bfb154 Fix: After taking scan_lock in nvme_mpath_clear_ctrl_paths() down_read(&ctrl->namespaces_rwsem) as well to make list traversal safe. This will not cause deadlocks because taking scan_lock never happens while holding the namespaces_rwsem. Moreover, scan work downs namespaces_rwsem in the same order. Alternative: sort ctrl->namespaces in nvme_scan_work() while still holding the scan_lock. This would leave nvme_mpath_clear_ctrl_paths() without correct protection against ctrl->namespaces modification by anyone other than scan_work. Reviewed-by:Sagi Grimberg <sagi@grimberg.me> Reviewed-by:
Anton Eidelman <anton@lightbitslabs.com> Signed-off-by:
-
Max Gurtovoy authored
In case there are controllers that are not associated with any RDMA device (e.g. during unsuccessful reconnection) and the user will unload the module, these controllers will not be freed and will access already freed memory. The same logic appears in other fabric drivers as well. Fixes: 87fd1253 ("nvme-rdma: remove redundant reference between ib_device and tagset") Reviewed-by:
Max Gurtovoy <maxg@mellanox.com> Signed-off-by:
-
Christoph Hellwig authored
__blk_queue_split() adds significant overhead for small I/O operations. Add a shortcut to avoid it for cases where we know we never need to split. Based on a patch from Ming Lei. Signed-off-by:
-
- 04 Nov, 2019 3 commits
-
-
Prabhath Sajeepa authored
Check validity of offset into ANA log buffer before accessing nvme_ana_group_desc. This check ensures the size of ANA log buffer >= offset + sizeof(nvme_ana_group_desc) Reviewed-by:
Prabhath Sajeepa <psajeepa@purestorage.com> Signed-off-by:
-
Christoph Hellwig authored
bio_set_op_attrs has been long deprecated, replace it with a direct assignment of the flags to bio->bi_opf. Reviewed-by:
Johannes Thumshirn <jthumshirn@suse.de> Signed-off-by:
-
Christoph Hellwig authored
With reference to the following issue reported on the mailing list :- http://lists.infradead.org/pipermail/linux-nvme/2019-October/027604.html this patch adds plugging for the bdev-ns under nvmet_bdev_execute_rw(). We can see the following performance improvement in random write workload I/Os with the setup described in the link when device_path configured as /dev/md0. Without this patch :- write: IOPS=40.8k, BW=159MiB/s (167MB/s)(4777MiB/30002msec) write: IOPS=41.2k, BW=161MiB/s (169MB/s)(4831MiB/30011msec) slat (usec): min=8, max=10823, avg=15.64, stdev=16.85 slat (usec): min=8, max=401, avg=15.40, stdev= 9.56 clat (usec): min=54, max=2492, avg=759.07, stdev=172.62 clat (usec): min=56, max=1997, avg=768.06, stdev=178.72 With this patch :- write: IOPS=123k, BW=480MiB/s (504MB/s)(14.1GiB/30011msec) write: IOPS=123k, BW=481MiB/s (504MB/s)(14.1GiB/30002msec) slat (usec): min=8, max=9941, avg=13.31, stdev= 8.04 slat (usec): min=8, max=289, avg=13.31, stdev= 3.37 clat (usec): min=43, max=17635, avg=245.46, stdev=171.23 clat (usec): min=44, max=17751, avg=245.25, stdev=183.14 Reviewed-by:
-
