Ensure that info->sector_size and info->physical_sector_size are set
before the call to blkif_set_queue_limits by doing away with the
local variables and arguments that propagate them.
Thanks to Marek Marczykowski-Górecki and Jürgen Groß for root causing
the issue.
Fixes: ba3f67c116 ("xen-blkfront: atomically update queue limits")
Reported-by: Rusty Bird <rustybird@net-c.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Roger Pau Monné <roger.pau@citrix.com>
Link: https://lore.kernel.org/r/20240625055238.7934-1-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The current tag reservation code is based on a misunderstanding of the
meaning of data->shallow_depth. Fix the tag reservation code as follows:
* By default, do not reserve any tags for synchronous requests because
for certain use cases reserving tags reduces performance. See also
Harshit Mogalapalli, [bug-report] Performance regression with fio
sequential-write on a multipath setup, 2024-03-07
(https://lore.kernel.org/linux-block/5ce2ae5d-61e2-4ede-ad55-551112602401@oracle.com/)
* Reduce min_shallow_depth to one because min_shallow_depth must be less
than or equal any shallow_depth value.
* Scale dd->async_depth from the range [1, nr_requests] to [1,
bits_per_sbitmap_word].
Cc: Christoph Hellwig <hch@lst.de>
Cc: Damien Le Moal <dlemoal@kernel.org>
Cc: Zhiguo Niu <zhiguo.niu@unisoc.com>
Fixes: 07757588e5 ("block/mq-deadline: Reserve 25% of scheduler tags for synchronous requests")
Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240509170149.7639-3-bvanassche@acm.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
NOWS is one of the annoying "0's based values" in NVMe, where 0 means one
and we thus can't detect if it isn't set. Thus a NOWS value of 0 means
that the Namespace Optimal Write Size is a single LBA, which is clearly
bogus. Ignore the value in that case and don't propagate an io_opt
value to the block layer.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Sagi Grimberg <sagi@grimberg.me>
Reviewed-by: Nitesh Shetty <nj.shetty@samsung.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20240701051800.1245240-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Now we avoid throttling swap writes by determining whether the current
process is kswapd (aka current_is_kswapd()), but swap writes can come
from either kswapd or direct reclaim, so the swap writes from direct
reclaim will still be throttled.
When a process holds a lock to allocate a free page, and enters direct
reclaim because there is no free memory, then it might trigger a hung
due to the wbt throttling that causes other processes to fail to get
the lock.
Both kswapd and direct reclaim set the REQ_SWAP flag, so use REQ_SWAP
instead of current_is_kswapd() to avoid throttling swap writes. Also
renamed WBT_KSWAPD to WBT_SWAP and WBT_RWQ_KSWAPD to WBT_RWQ_SWAP.
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240604030522.3686177-1-libaokun@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
User will configure allowed iops limit in 1s, and calculate_io_allowed()
will calculate allowed iops in the slice by:
limit * HZ / throtl_slice
However, if limit is quite low, the result can be 0, then
allowed IO in the slice is 0, this will cause missing dispatch and
control will be lower than limit.
For example, set iops_limit to 5 with HD disk, and test will found that
iops will be 3.
This is usually not a big deal, because user will unlikely to configure
such low iops limit, however, this is still a problem in the extreme
scene.
Fix the problem by making sure the wait time calculated by
tg_within_iops_limit() should allow at least one IO to be dispatched.
Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Acked-by: Tejun Heo <tj@kernel.org>
Link: https://lore.kernel.org/r/20240618062108.3680835-1-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
QUEUE_FLAG_NOMERGES isn't really a driver interface, but a user tunable.
There also isn't any good reason to set it in the loop driver.
The original commit adding it (5b5e20f421 "block: loop: set
QUEUE_FLAG_NOMERGES for request queue of loop") claims that "It doesn't
make sense to enable merge because the I/O submitted to backing file is
handled page by page." which of course isn't true for multi-page bvec
now, and it never has been for direct I/O, for which commit 40326d8a33
("block/loop: allow request merge for directio mode") alredy disabled
the nomerges flag.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240627124926.512662-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
IO logical block size is one fundamental queue limit, and every IO has
to be aligned with logical block size because our bio split can't deal
with unaligned bio.
The check has to be done with queue usage counter grabbed because device
reconfiguration may change logical block size, and we can prevent the
reconfiguration from happening by holding queue usage counter.
logical_block_size stays in the 1st cache line of queue_limits, and this
cache line is always fetched in fast path via bio_may_exceed_limits(),
so IO perf won't be affected by this check.
Cc: Yi Zhang <yi.zhang@redhat.com>
Cc: Christoph Hellwig <hch@infradead.org>
Cc: Ye Bin <yebin10@huawei.com>
Cc: stable@vger.kernel.org
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240620030631.3114026-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Sometimes we need to track the processing order of requests with
ioprio set. So the ioprio of request can be useful information.
Example：
block_rq_insert: 8,0 RA 16384 () 6500840 + 32 be,0,6 [binder:815_3]
block_rq_issue: 8,0 RA 16384 () 6500840 + 32 be,0,6 [binder:815_3]
block_rq_complete: 8,0 RA () 6500840 + 32 be,0,6 [0]
Signed-off-by: Dongliang Cui <dongliang.cui@unisoc.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240614074936.113659-1-dongliang.cui@unisoc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
1. Userspace sends the command "losetup -d" which uses the open() call
to open the device
2. Kernel receives the ioctl command "LOOP_CLR_FD" which calls the
function loop_clr_fd()
3. If LOOP_CLR_FD is the first command received at the time, then the
AUTOCLEAR flag is not set and deletion of the
loop device proceeds ahead and scans the partitions (drop/add
partitions)
if (disk_openers(lo->lo_disk) > 1) {
lo->lo_flags |= LO_FLAGS_AUTOCLEAR;
loop_global_unlock(lo, true);
return 0;
}
4. Before scanning partitions, it will check to see if any partition of
the loop device is currently opened
5. If any partition is opened, then it will return EBUSY:
if (disk->open_partitions)
return -EBUSY;
6. So, after receiving the "LOOP_CLR_FD" command and just before the above
check for open_partitions, if any other command
(like blkid) opens any partition of the loop device, then the partition
scan will not proceed and EBUSY is returned as shown in above code
7. But in "__loop_clr_fd()", this EBUSY error is not propagated
8. We have noticed that this is causing the partitions of the loop to
remain stale even after the loop device is detached resulting in the
IO errors on the partitions
Fix:
Defer the detach of loop device to release function, which is called when
the last close happens, by setting the lo_flags to LO_FLAGS_AUTOCLEAR at
the time of detach i.e in loop_clr_fd() function.
Test case involves the following two scripts:
script1.sh:
while [ 1 ];
do
losetup -P -f /home/opt/looptest/test10.img
blkid /dev/loop0p1
done
script2.sh:
while [ 1 ];
do
losetup -d /dev/loop0
done
Without fix, the following IO errors have been observed:
kernel: __loop_clr_fd: partition scan of loop0 failed (rc=-16)
kernel: I/O error, dev loop0, sector 20971392 op 0x0:(READ) flags 0x80700
phys_seg 1 prio class 0
kernel: I/O error, dev loop0, sector 108868 op 0x0:(READ) flags 0x0
phys_seg 1 prio class 0
kernel: Buffer I/O error on dev loop0p1, logical block 27201, async page
read
Signed-off-by: Gulam Mohamed <gulam.mohamed@oracle.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240618164042.343777-1-gulam.mohamed@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
dma_pad_mask is a queue_limits by all ways of looking at it, so move it
there and set it through the atomic queue limits APIs.
Add a little helper that takes the alignment and pad into account to
simplify the code that is touched a bit.
Note that there never was any need for the > check in
blk_queue_update_dma_pad, this probably was just copy and paste from
dma_update_dma_alignment.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240626142637.300624-9-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
The md driver wants to enforce a number of flags for all devices, even
when not inheriting them from the underlying devices. To make sure these
flags survive the queue_limits_set calls that md uses to update the
queue limits without deriving them form the previous limits add a new
md_init_stacking_limits helper that calls blk_set_stacking_limits and sets
these flags.
Fixes: 1122c0c1cc ("block: move cache control settings out of queue->flags")
Reported-by: kernel test robot <oliver.sang@intel.com>
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240626142637.300624-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
If we allocate a bio that is larger than NVMe maximum request size,
attach integrity metadata to it and send it to the NVMe subsystem, the
integrity metadata will be corrupted.
Splitting the bio works correctly. The function bio_split will clone the
bio, trim the iterator of the first bio and advance the iterator of the
second bio.
However, the function rq_integrity_vec has a bug - it returns the first
vector of the bio's metadata and completely disregards the metadata
iterator that was advanced when the bio was split. Thus, the second bio
uses the same metadata as the first bio and this leads to metadata
corruption.
This commit changes rq_integrity_vec, so that it calls mp_bvec_iter_bvec
instead of returning the first vector. mp_bvec_iter_bvec reads the
iterator and uses it to build a bvec for the current position in the
iterator.
The "queue_max_integrity_segments(rq->q) > 1" check was removed, because
the updated rq_integrity_vec function works correctly with multiple
segments.
Signed-off-by: Mikulas Patocka <mpatocka@redhat.com>
Reviewed-by: Anuj Gupta <anuj20.g@samsung.com>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/49d1afaa-f934-6ed2-a678-e0d428c63a65@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is no need to conditionally define on CONFIG_BLK_DEV_ZONED the
inline helper functions bdev_nr_zones(), bdev_max_open_zones(),
bdev_max_active_zones() and disk_zone_no() as these function will return
the correct valu in all cases (zoned device or not, including when
CONFIG_BLK_DEV_ZONED is not set). Furthermore, disk_nr_zones()
definition can be simplified as disk->nr_zones is always 0 for regular
block devices.
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20240621031506.759397-4-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
There is no need for bdev_nr_zones() to be an exported function
calculating the number of zones of a block device. Instead, given that
all callers use this helper with a fully initialized block device that
has a gendisk, we can redefine this function as an inline helper in
blkdev.h.
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Link: https://lore.kernel.org/r/20240621031506.759397-3-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add support to set block layer request_queue atomic write limits. The
limits will be derived from either the namespace or controller atomic
parameters.
NVMe atomic-related parameters are grouped into "normal" and "power-fail"
(or PF) class of parameter. For atomic write support, only PF parameters
are of interest. The "normal" parameters are concerned with racing reads
and writes (which also applies to PF). See NVM Command Set Specification
Revision 1.0d section 2.1.4 for reference.
Whether to use per namespace or controller atomic parameters is decided by
NSFEAT bit 1 - see Figure 97: Identify – Identify Namespace Data
Structure, NVM Command Set.
NVMe namespaces may define an atomic boundary, whereby no atomic guarantees
are provided for a write which straddles this per-lba space boundary. The
block layer merging policy is such that no merges may occur in which the
resultant request would straddle such a boundary.
Unlike SCSI, NVMe specifies no granularity or alignment rules, apart from
atomic boundary rule. In addition, again unlike SCSI, there is no
dedicated atomic write command - a write which adheres to the atomic size
limit and boundary is implicitly atomic.
If NSFEAT bit 1 is set, the following parameters are of interest:
- NAWUPF (Namespace Atomic Write Unit Power Fail)
- NABSPF (Namespace Atomic Boundary Size Power Fail)
- NABO (Namespace Atomic Boundary Offset)
and we set request_queue limits as follows:
- atomic_write_unit_max = rounddown_pow_of_two(NAWUPF)
- atomic_write_max_bytes = NAWUPF
- atomic_write_boundary = NABSPF
If in the unlikely scenario that NABO is non-zero, then atomic writes will
not be supported at all as dealing with this adds extra complexity. This
policy may change in future.
In all cases, atomic_write_unit_min is set to the logical block size.
If NSFEAT bit 1 is unset, the following parameter is of interest:
- AWUPF (Atomic Write Unit Power Fail)
and we set request_queue limits as follows:
- atomic_write_unit_max = rounddown_pow_of_two(AWUPF)
- atomic_write_max_bytes = AWUPF
- atomic_write_boundary = 0
A new function, nvme_valid_atomic_write(), is also called from submission
path to verify that a request has been submitted to the driver will
actually be executed atomically. As mentioned, there is no dedicated NVMe
atomic write command (which may error for a command which exceeds the
controller atomic write limits).
Note on NABSPF:
There seems to be some vagueness in the spec as to whether NABSPF applies
for NSFEAT bit 1 being unset. Figure 97 does not explicitly mention NABSPF
and how it is affected by bit 1. However Figure 4 does tell to check Figure
97 for info about per-namespace parameters, which NABSPF is, so it is
implied. However currently nvme_update_disk_info() does check namespace
parameter NABO regardless of this bit.
Signed-off-by: Alan Adamson <alan.adamson@oracle.com>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
jpg: total rewrite
Signed-off-by: John Garry <john.g.garry@oracle.com>
Link: https://lore.kernel.org/r/20240620125359.2684798-11-john.g.garry@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Add initial support for atomic writes.
As is standard method, feed device properties via modules param, those
being:
- atomic_max_size_blks
- atomic_alignment_blks
- atomic_granularity_blks
- atomic_max_size_with_boundary_blks
- atomic_max_boundary_blks
These just match sbc4r22 section 6.6.4 - Block limits VPD page.
We just support ATOMIC WRITE (16).
The major change in the driver is how we lock the device for RW accesses.
Currently the driver uses a per-device lock for accessing device metadata
and "media" data (calls to do_device_access()) atomically for the duration
of the whole read/write command.
This should not suit verifying atomic writes. Reason being that currently
all reads/writes are atomic, so using atomic writes does not prove
anything.
Change device access model to basis that regular writes only atomic on a
per-sector basis, while reads and atomic writes are fully atomic.
As mentioned, since accessing metadata and device media is atomic,
continue to have regular writes involving metadata - like discard or PI -
as atomic. We can improve this later.
Currently we only support model where overlapping going reads or writes
wait for current access to complete before commencing an atomic write.
This is described in 4.29.3.2 section of the SBC. However, we simplify,
things and wait for all accesses to complete (when issuing an atomic
write).
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: John Garry <john.g.garry@oracle.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Luis Chamberlain <mcgrof@kernel.org>
Link: https://lore.kernel.org/r/20240620125359.2684798-10-john.g.garry@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
