| Age | Commit message (Collapse) | Author |
|---|
|
https://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux.git
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/modules/linux.git
|
|
https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git
# Conflicts:
# drivers/pci/pci.c
# init/Kconfig
# mm/internal.h
|
|
git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux
Pull block fixes from Jens Axboe:
- NVMe fixes via Keith:
- Fix an out-of-bounds write in nvmet_auth_challenge(), where
sizeof() on a void pointer undercounted the challenge header and
let a short AUTH_RECEIVE buffer pass the check
- nvme-multipath fixes for an ANA log bounds check underflow, the
command effects log lifetime for multipath heads, and only
setting BLK_FEAT_ZONED after the zone info is known.
- nvmet fixes for ns->enabled teardown ordering, rejecting I/O
after the percpu ns reference is killed, device path preservation
on allocation failure, and too-short SGL segments in pci-epf
- nvme-tcp: revert the per-socket dynamic lockdep keys, and delay
the socket reclassification
- A DMA pool alignment quirk for the Micron 4100AT
- Controller state/reset race fixes, and -Wformat-security
workarounds
- blk-mq: set RQF_USE_SCHED when the operation is known, and allow
cached requests to be used for flush operations
- Reject polled dio with user integrity metadata
- Save the IRQ state in blkg_tryget_closest()
- Set the zone write granularity in virtio_blk
- ublk selftest fixes
* tag 'block-7.3-20261002' of git://git.kernel.org/pub/scm/linux/kernel/git/axboe/linux: (23 commits)
virtio_blk: set the zone write granularity
nvme-multipath: set BLK_FEAT_ZONED only after the zone info is known
nvme: fix command effects log lifetime for multipath heads
nvmet: don't allow I/O admission after percpu ns reference is killed
nvmet: defer setting ns->enabled to false in nvmet_ns_disable()
nvmet: copy the hostid into the ctrl before creating PR pc_refs
nvmet-auth: fix out-of-bounds write in nvmet_auth_challenge()
nvmet: pci-epf: reject too-short SGL segments
nvme-multipath: fix underflow in ANA log bounds checks
nvme: work around all -Wformat-security warnings
nvme: work around -Wformat-security warning
nvme: do not reset controllers in NVME_CTRL_NEW state
nvme-tcp: delay nvme_tcp_reclassify_socket()
Revert "nvme-tcp: lockdep: use dynamic lockdep keys per socket instance"
drbd: remove unused drbd_nl_mcgrps[] array
blk-mq: allow cached requests to be used for flush operations
blk-mq: set RQF_USE_SCHED when the operation is known
block: reject polled dio with user integrity metadata
selftests: ublk: fix unused_result error
blk-cgroup: save IRQ state in blkg_tryget_closest()
...
|
|
read_block_state() formats each entry directly into the buffer supplied by
read(). If the remaining buffer is too small for one complete record,
snprintf() returns the full record length and the function stops without
copying data or advancing the file position. A read from this debugfs
file which is smaller than a record therefore returns zero at a non-EOF
position and cannot make progress.
Convert block_state to seq_file so formatted records are buffered
independently of the userspace read size. Keep dev_lock held across each
seq_file iteration and continue to protect individual entries with their
slot locks.
Link: https://lore.kernel.org/20260929071846.24829-1-pooyan.azadparvar@gmail.com
Fixes: c0265342bff4 ("zram: introduce zram memory tracking")
Signed-off-by: Pooyan Azad <pooyan.azadparvar@gmail.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Closes: https://lore.kernel.org/r/CANC3H+LdtoydSp+o2ecErAw7k6R2+gRf9LyxcaoHv_mGhJmyQQ@mail.gmail.com/
Reviewed-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Tested-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Jens Axboe <axboe@kernel.dk>
|
|
Patch series "zsmallc: remove old object read API".
zram remains the only user of old zsmalloc object read API. This series
removes the old API and converts zram to use the new SG-list based API.
This patch (of 2):
zram remains the last user of old zsmalloc object read API, that performed
linearisation on the zsmalloc side. There is a new SG-list API, that has
a bunch of benefits. Switch zram to SG-list zsmalloc object read API.
Link: https://lore.kernel.org/20260907105739.1793316-1-senozhatsky@chromium.org
Link: https://lore.kernel.org/20260907105739.1793316-2-senozhatsky@chromium.org
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Minchan Kim <minchan@kernel.org>
Cc: Nhat Pham <nphamcs@gmail.com>
|
|
Sashiko reported that:
kernel_read_file_from_path() returns negative error for zero-sized
files, so we cannot have "sz == 0" return, remove it and use a
generic error message instead.
Link: https://lore.kernel.org/20260901051335.2202390-1-senozhatsky@chromium.org
Signed-off-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Cc: Haoqin Huang <haoqinhuang7@gmail.com>
|
|
After commit 2e8ff2f51dde ("zram: use u32 for entry ac_time tracking"),
idle_store() computes the idle cutoff as:
cutoff = ktime_sub((u32)ktime_get_boottime_seconds(), age_sec);
Because the left operand is cast to u32, when age_sec exceeds the current
uptime the subtraction wraps modulo 2^32 and the huge result is
zero-extended into the s64 cutoff. mark_idle() then marks every entry as
idle instead of matching nothing. For instance, running
echo 86400 > /sys/block/zramX/idle
on a machine up for only two minutes marks all newly written pages idle
and hands them to idle writeback and recompression.
No slot can have been accessed before the system booted, so an age_sec
that reaches back past uptime cannot match any slot. Return early in that
case, without walking the table or taking any slot locks.
Track the cutoff as time64_t rather than ktime_t. Both cutoff and ac_time
are boot-time values in seconds, so a plain arithmetic comparison against
ac_time in mark_idle() is correct and no ktime helpers are needed.
Link: https://lore.kernel.org/20260828083149.45760-1-jiahao.kernel@gmail.com
Fixes: 2e8ff2f51dde ("zram: use u32 for entry ac_time tracking")
Signed-off-by: Hao Jia <jiahao1@lixiang.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Suggested-by: Sergey Senozhatsky <senozhatsky@chromium.org>
Cc: Brian Geffon <bgeffon@google.com>
Cc: Jens Axboe <axboe@kernel.dk>
Cc: Minchan Kim <minchan@kernel.org>
Cc: <stable@vger.kernel.org>
|
|
* for-7.4/block:
block: amiflop: synchronize flush timer before track flush
block: drop the cached plug time when blk_add_rq_to_plug() flushes
ublk: drop the device reference outside ublk_ctl_mutex in DEL_DEV
ublk: don't use an I/O scheduler by default
|
|
The amiflop track buffer is shared with the timer callback. timer_delete()
does not wait for an in-flight callback, and that callback can rearm itself
when it cannot acquire the floppy controller. A synchronous track flush can
therefore race with the callback.
Stop the timer before synchronous track writes and use timer_delete_sync()
to wait for any callback already running. Keep callback rearming disabled
until the write is complete. Apply this to the FDFMTTRK path as well.
FDFLUSH, FDFMTTRK and close-time flushes run outside the blk-mq
request callback. Freeze their queues so an in-flight request cannot
change the shared track buffer while it is encoded and written. Restart
the timer if a track remains dirty after a failed flush.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: stable@vger.kernel.org
Assisted-by: LLM
Signed-off-by: Runyu Xiao <runyu.xiao@seu.edu.cn>
Link: https://patch.msgid.link/20260930065951.2932739-1-runyu.xiao@seu.edu.cn
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
ublk_ctrl_del_dev() drops its reference to the device while holding the
global ublk_ctl_mutex. When this is the last reference, ublk_cdev_rel()
frees the tag set, and blk_mq_free_tag_set() waits for the pending SRCU
callbacks of the tag set with srcu_barrier(). Every DEL_DEV thus waits
for an SRCU grace period with the mutex held, and deleting devices from
several threads serializes on it.
The release path doesn't need ublk_ctl_mutex. The device number is freed
under ublk_idr_lock, and the last reference is already dropped without
the mutex when the ublk server still has the char device open at
DEL_DEV time, from ublk_ch_release_work_fn().
Drop the reference after unlocking ublk_ctl_mutex.
ublk null target, 4000 single-queue devices, STOP_DEV + DEL_DEV issued
from N threads, 16 vCPU KVM guest:
threads before after
1 77.6 s 77.6 s
4 20.1 s 19.4 s
8 19.3 s 9.0 s
16 19.2 s 4.1 s
Signed-off-by: Qiliang Yuan <odys.yuan@gmail.com>
Reviewed-by: Ming Lei <tom.leiming@gmail.com>
Link: https://patch.msgid.link/20261001-bug-ublk-del-dev-put-unlocked-v1-1-18e9bb9be727@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Requests of a ublk device are handed to the ublk server through
io_uring, and the server does its own queueing and ordering, so an I/O
scheduler in front of the device only adds overhead. A single-queue
ublk device still gets mq-deadline by default, which costs throughput
on every I/O and an elevator setup on every START_DEV.
Set BLK_MQ_F_NO_SCHED_BY_DEFAULT so that add_disk() selects "none", as
loop does since commit 2112f5c1330a ("loop: Select I/O scheduler 'none'
from inside add_disk()"). Doing it in the kernel also avoids switching
the scheduler from user space after the device has been added, which
waits for RCU grace periods. A scheduler can still be selected through
sysfs. Zoned ublk devices don't need mq-deadline either, zone write
plugging serializes writes per zone in the block layer since
commit fde02699c242 ("block: mq-deadline: Remove support for zone write
locking").
ublk null target, single-queue devices with depth 16, fio io_uring 4k
randread at iodepth=16, 16 vCPU KVM guest:
mq-deadline none
1 device 111K IOPS 342K IOPS
4 devices 286K IOPS 1070K IOPS
START_DEV p50 (1000 devices) 8.53ms 0.36ms
Signed-off-by: Qiliang Yuan <odys.yuan@gmail.com>
Reviewed-by: Ming Lei <tom.leiming@gmail.com>
Link: https://patch.msgid.link/20261001-bug-ublk-no-sched-by-default-v1-1-0bc91b6f075e@gmail.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
* block-7.3:
virtio_blk: set the zone write granularity
|
|
virtblk_read_zoned_limits() reads the write granularity that the device
reports in virtio_blk_zoned_characteristics and assigns it to the
physical block size and to io_min, but never to the limit that is named
after it. queue_limits.zone_write_granularity is left at zero, so
blk_validate_zoned_limits() raises it to the logical block size:
if (lim->zone_write_granularity < lim->logical_block_size)
lim->zone_write_granularity = lim->logical_block_size;
A device that reports a granularity coarser than its logical block size,
which is what the field exists to express, therefore has it silently
reduced. A 512e host managed disk passed through to a guest reports a
logical block size of 512 and a write granularity of 4096, and the guest
ends up with a zone write granularity of 512.
bio_split_alignment() returns lim->zone_write_granularity if it is non-zero
and bio_split_io_at() may split a bio with as per bio_split_alignment().
This can real to the write getting rejected by the host drive, as the write
is not aligned to the physical block size.
zonefs also takes its block size from bdev_zone_write_granularity(), so it
would incorrectly use 512 on a disk that requires 4096.
sd_zbc_read_zones() sets the limit from the physical block size for the
same reason. NVMe ZNS and null_blk leave it unset, but the fallback
gives the right answer for them, as their write granularity is the
logical block size. virtio carries a separate value that may exceed it.
Set the zone write granularity from the value that the device reports.
Fixes: 95bfec41bd3d ("virtio-blk: add support for zoned block devices")
Signed-off-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Link: https://patch.msgid.link/20260918140641.2031075-2-cassel@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
* for-7.4/block:
rust: block: require `Sync` for `Operations::QueueData`
rnull: configfs: add power to configfs features
rnull: fix geometry store check-then-act across lock scopes
rust: block: Fix GenDiskBuilder block size documentation
rust: block: gen_disk: set fops.owner from driver module pointer
rust: block: fix `Send` bound for `GenDisk`
rust: block: rnull: use vertical import style
rust: block: mq: remove redundant imports and format
rust: block: mq: use vertical import style
|
|
features displayed by configfs for rnull was inconsistent with the
actual features available. This correctly exposes `power` as a feature,
which is also consistent with the C null_blk driver.
Fixes: d969d504bc13 ("rnull: enable configuration via `configfs`")
Signed-off-by: Malte Wechter <maltewechter@gmail.com>
Link: https://msgid.link/20260924-upstream-v7-3-rc3-v2-1-ca56ae7a73e9@gmail.com
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://patch.msgid.link/20260929-rust-block-for-v7-4-rc1-b4-v1-8-642a4c1aebd7@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The blocksize/rotational/capacity/irqmode stores check powered under one
Mutex acquisition, drop the guard, then update under a second acquisition.
A concurrent power-on can create the live disk from stale geometry in
between, leaving powered==true with config != live disk.
Hold one guard for the powered check and the field update.
Signed-off-by: Qingxiao Xu <qingxiao@tamu.edu>
Link: https://msgid.link/20260908200845.405112-1-qingxiao@tamu.edu
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://patch.msgid.link/20260929-rust-block-for-v7-4-rc1-b4-v1-7-642a4c1aebd7@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Convert `use` imports to vertical layout for better readability and
maintainability.
Signed-off-by: Alvin Sun <alvin.sun@linux.dev>
Link: https://msgid.link/20260521-miscdev-use-format-v3-6-56240ca70d0c@linux.dev
Signed-off-by: Andreas Hindborg <a.hindborg@kernel.org>
Link: https://patch.msgid.link/20260929-rust-block-for-v7-4-rc1-b4-v1-3-642a4c1aebd7@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
* for-7.4/block:
swim3: Fix FDEJECT ioctl for exclusive open
swim: Fix FDEJECT ioctl for exclusive open
swim: Return -EROFS when opened with BLK_OPEN_WRITE flag
|
|
The eject command from the util-linux package does not work with the
swim3 driver: the drive doesn't eject anything and the command fails.
Apparently, ioctl(fd, FDEJECT) returns -EBUSY after fd was opened O_EXCL.
Fix this by checking ref_count for either 1 or -1. Either value means that
the caller is the sole user of the disk.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Link: https://patch.msgid.link/f4cb43726ee6f1cf159a6208f77b4cceea670a3c.1790641175.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The eject command from the util-linux package does not work with the swim
driver: the drive doesn't eject anything and the command fails.
Apparently, ioctl(fd, FDEJECT) returns -EBUSY when fd is opened O_EXCL.
Fix this by checking ref_count for either 1 or -1. Either value means that
the caller is the sole user of the disk.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Link: https://patch.msgid.link/ae3ee8e86833a2739d7857460cbe2addf4146088.1790641175.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Write requests are not supported by the driver and produce IO errors as
shown below. Avoid this by returning -EROFS when opened for writing.
[ 4111.690000] I/O error, dev fd0, sector 2 op 0x1:(WRITE) flags 0x800800 phys_seg 1 prio class 2
[ 4111.700000] Buffer I/O error on dev fd0, logical block 2, lost async page write
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Tested-by: Stan Johnson <userm57@yahoo.com>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Link: https://patch.msgid.link/327be6ca3c92c3411b5dfaeede014ed65a2d0f0f.1790641175.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
* for-7.4/block: (53 commits)
swim: Unexport global symbols
swim: Define symbols for constants
swim: Define macros for constants
swim: Clean up whitespace
swim: Remove unused macro definitions
swim: Add some helpful references
swim: Move swd initialization
swim: Remove pointless specifiers
swim: Don't search beyond the first data mark
swim: Don't needlessly re-read sectors
swim: Remove pointless mode0 register write
swim: Revisit delays
swim: Check drive ready bit
swim: Deduplicate polling loops
swim: Remove redundant RELAX actions
swim: Convert to blocking queue
swim: Fix buffer overflow
swim: Don't use the mark register to read data
swim: Check error register during sector read
swim: Check for CRC errors
...
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
* block-7.3: (22 commits)
nvme-multipath: set BLK_FEAT_ZONED only after the zone info is known
nvme: fix command effects log lifetime for multipath heads
nvmet: don't allow I/O admission after percpu ns reference is killed
nvmet: defer setting ns->enabled to false in nvmet_ns_disable()
nvmet: copy the hostid into the ctrl before creating PR pc_refs
nvmet-auth: fix out-of-bounds write in nvmet_auth_challenge()
nvmet: pci-epf: reject too-short SGL segments
nvme-multipath: fix underflow in ANA log bounds checks
nvme: work around all -Wformat-security warnings
nvme: work around -Wformat-security warning
nvme: do not reset controllers in NVME_CTRL_NEW state
nvme-tcp: delay nvme_tcp_reclassify_socket()
Revert "nvme-tcp: lockdep: use dynamic lockdep keys per socket instance"
drbd: remove unused drbd_nl_mcgrps[] array
blk-mq: allow cached requests to be used for flush operations
blk-mq: set RQF_USE_SCHED when the operation is known
block: reject polled dio with user integrity metadata
selftests: ublk: fix unused_result error
blk-cgroup: save IRQ state in blkg_tryget_closest()
nvmet: preserve device path on allocation failure
...
|
|
After the rework, two files have a copy of drbd_nl_mcgrps[], but one
of them has no references:
drivers/block/drbd/drbd_nl_gen.c:641:42: error: 'drbd_nl_mcgrps' defined but not used [-Werror=unused-const-variable=]
641 | static const struct genl_multicast_group drbd_nl_mcgrps[] = {
| ^~~~~~~~~~~~~~
At the default warning level, -Wunused-const-variables is turned off,
so this has gone unnoticed.
Remove the extra variable.
Fixes: 8098eeb693c4 ("drbd: replace genl_magic with explicit netlink serialization")
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Reviewed-by: Christoph Böhmwalder <christoph.boehmwalder@linbit.com>
Link: https://patch.msgid.link/20260519203057.1340528-1-arnd@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
These symbols aren't used outside of this file so use local ones.
No functional change.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/0810de0df2a670d4d9ccad4e747be2c4b46f6d6e.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Define local symbols to give some meaning to anonymous constants.
No functional change.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/7eee06428905fd7ad60909d9b5726c15cde1b844.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Define a SEL_MASK macro to name the anonymous constant. Define STEPPING
rather than re-use STEP because the latter is a command bit macro (see also
GCR_MODE vs. SETGCR). No functional change, just better readability.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/64ee742adc48218962a989017b9dc5622e2932c7.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
No functional changes.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/d33238b1b2c4a444ed3ae59d4444fafebb550921.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Also remove the horizontal rule at the end of the macro definitions as
it doesn't any add value, IMO.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/d095be631aa91b6fd5704f4c159618d394a82414.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
These documents relate to the IWM, ISM, SWIM 1, 2, 3 and associated
disk drives. No functional change.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/082b79a3f2fa2ff5df9b51c582e01a8e47444613.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
For better readability, initialize the swd backpointer along with the
other floppy_state struct members. No functional change.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/613fa64bc857fecd141b5049eb9188ed72c9ec52.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
If the compiler made these functions "as fast as possible" that wouldn't
actually help because they involve slow mechanical operations. Remove
pointless inline function specifiers.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/e48ea61df817d30b5a50d1ced91c066b13cd47ae.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The ISM chip does an automatic MFM gap/sync search when the Action bit
is first set. That search may stop at any of a) post-index gap, b) address
field gap or c) data field gap. To find the next sector header, the
driver need not search at all. It only has to validate the mark bytes.
Once the sector address mark has been validated, swim_read_sector_data()
is called to read the sector contents. Between the sector address and
data fields lies an intra-sector gap followed by a data field mark.
After this mark is validated, the 512-byte data area is read into the
IO request buffer.
Problem is, if any byte in the data field mark is mis-read, the driver
searches the whole sector and then reaches the data field mark in the
following sector. The wrong sector is then read into the buffer, and
swim_read_sector_data() returns success. The request is silently
corrupted.
The existing limit on polling loop iterations does constrain the search
distance but is inherently tied to CPU speed. This is probably the reason
why corruption was only observed on a 68030 system.
Discontinue the mark search when the mark bytes fail validation.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/f6c97153faa0aceceffd3a61d85c0a82903a6b3e.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
floppy_read_sectors() is confusing because the `track' variable seems to
conflate tracks and cylinders. Rename this variable, eliminate a division
operation and adopt suitable integer types.
For readahead to work effectively, small sequential reads should not
require waiting for spindle rotation. Unfortunately, the present algorithm
is very inefficient and does a lot of unnecessary waiting.
E.g. if the device is asked to read sectors 1 thru 16, and if sector 9
happens to be under the heads, the driver will proceed to read sectors 9
thru 18, but discard the results, while it waits for sector 1 to arrive.
If sector 1 couldn't be read on the first attempt and needs a retry, the
driver will proceed to read sectors 2 thru 18, but discard the results,
while it waits for sector 1 to come around again.
In between reading sector 1 and sector 2, the driver needlessly calls
swim_track() and swim_head() again. But what's worse is re-enabling
interrupts after each sector, because on a 68030 system this can result
in a full rotation between sectors (which would be a 200 ms wait).
Floppy drivers usually implement a track cache that can be filled in a
single rotation to solve such problems. But I think there is a simpler
solution.
After stepping the heads, use a sector bitmap to record sectors that were
successfully read from the present track. Read (or retry, if need be) the
requested sectors in whatever sequence they become available. Keep
interrupts disabled until the whole track has passed under the read head.
swim_read_sector() assumes that it can search a whole track by reading a
fixed number of sector headers (essentially, fs->secpertrack) but this
assumes no false sector headers are matched in the sector contents. To
prevent that, call swim_read_sector_data() unconditionally after any
valid sector header is matched.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/dc4c1944edecb6778534b7b5535e6b6dd8d852ae.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
This write has no effect so remove it. (If side == 0 then no mode bit gets
cleared. If side == 1 then mode bit 0 gets cleared but that's pointless
because that bit is already clear here.)
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/acf12b671bbdbf5b986e41846f62ed8019d8d90b.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
AFAIK, timing requirements for the various FDHD drive mechanisms aren't
well documented. But we do have the UPD72070 spec and secondary sources
like swim3.c and mkLinux source code. This patch is needed to satisfy
the requirements in the UPD72070 spec and follows mkLinux.
Change the LSTRB pulse to 2 microseconds, because this is what mkLinux
does. Inside Macintosh says, "Hold LSTRB high for at least one usec but
not more than one msec".
When a disk is ejected, pause before de-asserting /ENBL. Wait 150 us after
the STEP command for valid signalling. Pause for 1 us after setting the
step direction before sending the STEP command.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/af356248a1e6a4fd956f71acf17feead27ba4918.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The drive provides a readiness signal that has to be tested before
certain commands are issued to the drive. Rename the SEEK_COMPLETE flag
as READY because that's how it's known in the documentation as well as
the mkLinux source code.
Poll for that signal after stepping the heads and also after switching
to MFM mode, as that's what mkLinux does. Check for readiness when
stepping because testing shows that some drives require this.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/8d6fc833c719de2248b64ee521418b1032b1e509.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Replace duplicated polling loops with poll_timeout_us(). Change the
interruptible sleep to uninterruptible because signal delivery shouldn't
be allowed to shorten delays required by the drive hardware.
Change the timeout for the !STEP transition to 20 ms in accordance with
the maximum interval required by the UPD72070 spec. The existing 1 second
timeout is impractical considering the number of steps in a typical seek.
Change the return type of swim_readbit() to bool because that way the
bit names make sense i.e. the reader doesn't have to remember to invert
the active-low logic used for drive signals.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/0ea2d3313dea26f8b4c2abc4d29055ac95aeb74b.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Wherever we have a swim_select() or swim_readbit() call there is an
implicit RELAX. That means the caller doesn't have to do it. Remove the
redundant code.
BTW, Inside Macintosh says, "Be sure [...] that CA0 and CA1 are set high
before changing SEL." Hence the RELAX found in swim_select(). The SwimIII
driver in mkLinux also has that. But the swim3.c driver in Linux is odd:
it scatters RELAX actions around as though SEL was not actually under its
control... In anycase, swim.c really does control SEL so there's no need
for that here.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/08ab67c2780c9107b87a1ece8b5f9cb54a8cea70.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
These drives are slow: completing a request can take hundreds of
milliseconds. Delays are managed by disabling interrupts judiciously and
sleeping opportunistically.
As of commit e3896d77b702 ("swim: convert to blk-mq"), a spinlock is
taken in irq mode as soon as a request is issued. That lock is held for
the duration of the request. Hence the driver sleeps while holding the
lock which is forbidden.
Adopt BLK_MQ_F_BLOCKING and remove the spinlock. Use a mutex to serialize
requests from the two request queues. (The chip cannot simultaneously
process requests on both internal and external drive.)
Cc: Omar Sandoval <osandov@fb.com>
Fixes: e3896d77b702 ("swim: convert to blk-mq")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/d06b19132ec678d6168d158e10cd63e10da45607.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The effect of this bug can be observed as swim_read_sector_data()
inexplicably returning -5, or an error flag indicating that a mark byte
was read from the data register, or other odd behviour.
When copying bytes from the chip FIFO to the read buffer, the driver
keeps count of the remaining buffer space using register %d4. A counter
in register %d2 serves as a timeout. The driver polls (%a2), the handshake
register, until flags indicate that byte(s) have arrived in the FIFO.
movel #sector_size-1, %d4
read_new_data:
movew #max_retry, %d2
read_data_loop:
moveb %a2@, %d5
andb #0xc0, %d5
dbne %d2, read_data_loop
beq data_exit
moveb %a5@, %a4@+
andb #0x40, %d5
dbne %d4, read_new_data
beq exit_loop
Note that the exit_loop branch depends upon a flag in the handshake
register and not on the remaining buffer space. Hence there may be no
branch to exit_loop after %d4 is decremented to -1 (i.e. full buffer).
moveb %a5@, %a4@+
dbra %d4, read_new_data
exit_loop:
Here is a second decrement of %d4 which can now reach -2. But the buffer
bounds check is a comparison with -1, which is now ineffective. Hence the
loop will continue copying until %d2 eventually reaches -1.
Fix this bug by terminating the loop as soon as %d4 or %d2 reach -1.
Reset the timeout whenever a byte is copied.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/d8fceda197b840f892d9b511175e824ca2549f01.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
If an unexpected mark byte were to be read from the data register, an
error would be flagged. But no error gets flagged when such a byte is
read from the mark register, which is misleading. Always use the data
register except when a mark byte is expected.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/fecd77b141bdd8a59ebae5f9efffccb155afdc4c.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Clear the error register only once before a sector read operation. Don't
clear it afterwards -- the caller needs it. Check the error register in
swim_read_sector() and return the appropriate error when necessary. Fully
validate the sector header. Don't terminate the search loop early just
because an erroneous sector header showed up.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/efabd807f4ecf5cbc05b60e91ed69aa4c10691a8.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
After reading either the sector header or sector data, examine that flag
in the handshake register which holds the result of the CRC calculation.
CRC validation has to take place with the last byte still in the FIFO.
This flag can't be checked by the caller because by then all bytes will
have been retrieved from the FIFO. Return an error code when appropriate.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/9bfcd0414192d1b95803c16701d62732ff6e8f3f.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Initialize the error result once only. Update the result only after a
successful read.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/28322f0cbefd865bdfb679bab4352c111996795b.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
When polling the FIFO for a mark byte in the sector header, don't
return zero if the timeout counter has expired, return an error code.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/aadb674d27d6e6d346d9b525597bf2e339c57e04.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The UPD72070 spec indicates that the track zero sensor can take 3 ms
to stabilize following a STEP command so add a call to msleep().
Remove the duplicate swim_readbit() call as there's no need for that
once the sensor signal has stabilized.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/6d17ca20a7286c1d74833b29510c506a66d96e6f.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
Track zero recalibration can be slow and is normally done only once i.e.
during system POST or boot-up. Recalibrate once after the drive is probed
rather than every time the device is opened. Don't register the drive if
recalibration fails. Park the heads before ejecting.
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/119e799afcc00d6bd4a479c31e0c960cb1a1d71b.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
|
The spindle motor should not be running when a disk is to be inserted.
Don't start the motor while the drive is empty.
Reviewed-by: Laurent Vivier <laurent@vivier.eu>
Fixes: 8852ecd97488 ("m68k: mac - Add SWIM floppy support")
Signed-off-by: Finn Thain <fthain@linux-m68k.org>
Tested-by: Stan Johnson <userm57@yahoo.com>
Link: https://patch.msgid.link/7ea60e0bf6135b9bcd4be0991a257d4c5de22bb3.1788513997.git.fthain@linux-m68k.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>
|
