Merge branch 'for-6.11/block' into for-next

* for-6.11/block: (26 commits)
  block: move the bounce flag into the features field
  block: move the skip_tagset_quiesce flag to queue_limits
  block: move the pci_p2pdma flag to queue_limits
  block: move the zone_resetall flag to queue_limits
  block: move the zoned flag into the features field
  block: move the poll flag to queue_limits
  block: move the dax flag to queue_limits
  block: move the nowait flag to queue_limits
  block: move the synchronous flag to queue_limits
  block: move the stable_writes flag to queue_limits
  block: move the io_stat flag setting to queue_limits
  block: move the add_random flag to queue_limits
  block: move the nonrot flag to queue_limits
  block: move cache control settings out of queue->flags
  block: remove blk_flush_policy
  block: freeze the queue in queue_attr_store
  nbd: move setting the cache control flags to __nbd_set_size
  virtio_blk: remove virtblk_update_cache_mode
  loop: fold loop_update_rotational into loop_reconfigure_limits
  loop: also use the default block size from an underlying block device
  ...

Merge branch 'for-6.11/block-limits' into for-6.11/block

Merge in last round of queue limits changes from Christoph.

* for-6.11/block-limits: (26 commits)
  block: move the bounce flag into the features field
  block: move the skip_tagset_quiesce flag to queue_limits
  block: move the pci_p2pdma flag to queue_limits
  block: move the zone_resetall flag to queue_limits
  block: move the zoned flag into the features field
  block: move the poll flag to queue_limits
  block: move the dax flag to queue_limits
  block: move the nowait flag to queue_limits
  block: move the synchronous flag to queue_limits
  block: move the stable_writes flag to queue_limits
  block: move the io_stat flag setting to queue_limits
  block: move the add_random flag to queue_limits
  block: move the nonrot flag to queue_limits
  block: move cache control settings out of queue->flags
  block: remove blk_flush_policy
  block: freeze the queue in queue_attr_store
  nbd: move setting the cache control flags to __nbd_set_size
  virtio_blk: remove virtblk_update_cache_mode
  loop: fold loop_update_rotational into loop_reconfigure_limits
  loop: also use the default block size from an underlying block device
  ...

Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the bounce flag into the features field

Move the bounce flag into the features field to reclaim a little bit of
space.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-27-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the skip_tagset_quiesce flag to queue_limits

Move the skip_tagset_quiesce flag into the queue_limits feature field so
that it can be set atomically with the queue frozen.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-26-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the pci_p2pdma flag to queue_limits

Move the pci_p2pdma flag into the queue_limits feature field so that it
can be set atomically with the queue frozen.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-25-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the zone_resetall flag to queue_limits

Move the zone_resetall flag into the queue_limits feature field so that
it can be set atomically with the queue frozen.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-24-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the zoned flag into the features field

Move the zoned flags into the features field to reclaim a little
bit of space.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-23-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the poll flag to queue_limits

Move the poll flag into the queue_limits feature field so that it can
be set atomically with the queue frozen.

Stacking drivers are simplified in that they now can simply set the
flag, and blk_stack_limits will clear it when the features is not
supported by any of the underlying devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-22-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the dax flag to queue_limits

Move the dax flag into the queue_limits feature field so that it can be
set atomically with the queue frozen.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-21-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the nowait flag to queue_limits

Move the nowait flag into the queue_limits feature field so that it can
be set atomically with the queue frozen.

Stacking drivers are simplified in that they now can simply set the
flag, and blk_stack_limits will clear it when the features is not
supported by any of the underlying devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-20-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the synchronous flag to queue_limits

Move the synchronous flag into the queue_limits feature field so that it
can be set atomically with the queue frozen.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-19-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the stable_writes flag to queue_limits

Move the stable_writes flag into the queue_limits feature field so that
it can be set atomically with the queue frozen.

The flag is now inherited by blk_stack_limits, which greatly simplifies
the code in dm, and fixed md which previously did not pass on the flag
set on lower devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-18-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the io_stat flag setting to queue_limits

Move the io_stat flag into the queue_limits feature field so that it can
be set atomically with the queue frozen.

Simplify md and dm to set the flag unconditionally instead of avoiding
setting a simple flag for cases where it already is set by other means,
which is a bit pointless.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-17-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the add_random flag to queue_limits

Move the add_random flag into the queue_limits feature field so that it
can be set atomically with the queue frozen.

Note that this also removes code from dm to clear the flag based on
the underlying devices, which can't be reached as dm devices will
always start out without the flag set.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-16-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move the nonrot flag to queue_limits

Move the nonrot flag into the queue_limits feature field so that it can
be set atomically with the queue frozen.

Use the chance to switch to defaulting to non-rotational and require
the driver to opt into rotational, which matches the polarity of the
sysfs interface.

For the z2ram, ps3vram, 2x memstick, ubiblock and dcssblk the new
rotational flag is not set as they clearly are not rotational despite
this being a behavior change.  There are some other drivers that
unconditionally set the rotational flag to keep the existing behavior
as they arguably can be used on rotational devices even if that is
probably not their main use today (e.g. virtio_blk and drbd).

The flag is automatically inherited in blk_stack_limits matching the
existing behavior in dm and md.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-15-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: move cache control settings out of queue->flags

Move the cache control settings into the queue_limits so that the flags
can be set atomically with the device queue frozen.

Add new features and flags field for the driver set flags, and internal
(usually sysfs-controlled) flags in the block layer.  Note that we'll
eventually remove enough field from queue_limits to bring it back to the
previous size.

The disable flag is inverted compared to the previous meaning, which
means it now survives a rescan, similar to the max_sectors and
max_discard_sectors user limits.

The FLUSH and FUA flags are now inherited by blk_stack_limits, which
simplified the code in dm a lot, but also causes a slight behavior
change in that dm-switch and dm-unstripe now advertise a write cache
despite setting num_flush_bios to 0.  The I/O path will handle this
gracefully, but as far as I can tell the lack of num_flush_bios
and thus flush support is a pre-existing data integrity bug in those
targets that really needs fixing, after which a non-zero num_flush_bios
should be required in dm for targets that map to underlying devices.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Ulf Hansson <ulf.hansson@linaro.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-14-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: remove blk_flush_policy

Fold blk_flush_policy into the only caller to prepare for pending changes
to it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617060532.127975-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: freeze the queue in queue_attr_store

queue_attr_store updates attributes used to control generating I/O, and
can cause malformed bios if changed with I/O in flight.  Freeze the queue
in common code instead of adding it to almost every attribute.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617060532.127975-12-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

nbd: move setting the cache control flags to __nbd_set_size

Move setting the cache control flags in nbd in preparation for moving
these flags into the queue_limits structure.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-11-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

virtio_blk: remove virtblk_update_cache_mode

virtblk_update_cache_mode boils down to a single call to
blk_queue_write_cache.  Remove it in preparation for moving the cache
control flags into the queue_limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Stefan Hajnoczi <stefanha@redhat.com>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617060532.127975-10-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

loop: fold loop_update_rotational into loop_reconfigure_limits

This prepares for moving the rotational flag into the queue_limits and
also fixes it for the case where the loop device is backed by a block
device.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240617060532.127975-9-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

loop: also use the default block size from an underlying block device

Fix the code in loop_reconfigure_limits to pick a default block size for
O_DIRECT file descriptors to also work when the loop device sits on top
of a block device and not just on a regular file on a block device based
file system.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240617060532.127975-8-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

loop: regularize upgrading the block size for direct I/O

The LOOP_CONFIGURE path automatically upgrades the block size to that
of the underlying file for O_DIRECT file descriptors, but the
LOOP_SET_BLOCK_SIZE path does not.  Fix this by lifting the code to
pick the block size into common code.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Link: https://lore.kernel.org/r/20240617060532.127975-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

loop: always update discard settings in loop_reconfigure_limits

Simplify loop_reconfigure_limits by always updating the discard limits.
This adds a little more work to loop_set_block_size, but doesn't change
the outcome as the discard flag won't change.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240617060532.127975-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

loop: stop using loop_reconfigure_limits in __loop_clr_fd

__loop_clr_fd wants to clear all settings on the device.  Prepare for
moving more settings into the block limits by open coding
loop_reconfigure_limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240617060532.127975-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: move zone limits setup out of sd_read_block_characteristics

Move a bit of code that sets up the zone flag and the write granularity
into sd_zbc_read_zones to be with the rest of the zoned limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: remove sd_is_zoned

Since commit 7437bb73f087 ("block: remove support for the host aware zone
model"), only ZBC devices expose a zoned access model.  sd_is_zoned is
used to check for that and thus return false for host aware devices.

Replace the helper with the simple open coded TYPE_ZBC check to fix this.

Fixes: 7437bb73f087 ("block: remove support for the host aware zone model")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617060532.127975-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

xen-blkfront: don't disable cache flushes when they fail

blkfront always had a robust negotiation protocol for detecting a write
cache.  Stop simply disabling cache flushes in the block layer as the
flags handling is moving to the atomic queue limits API that needs
user context to freeze the queue for that.  Instead handle the case
of the feature flags cleared inside of blkfront.  This removes old
debug code to check for such a mismatch which was previously impossible
to hit, including the check for passthrough requests that blkfront
never used to start with.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Roger Pau Monné <roger.pau@citrix.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240617060532.127975-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  z2ram: add missing MODULE_DESCRIPTION() macro
  ataflop: add missing MODULE_DESCRIPTION() macro
  amiflop: add missing MODULE_DESCRIPTION() macro

z2ram: add missing MODULE_DESCRIPTION() macro

With ARCH=m68k, make allmodconfig && make W=1 C=1 reports:
WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/z2ram.o

Add the missing invocation of the MODULE_DESCRIPTION() macro.

Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-3-b200599a315e@quicinc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

ataflop: add missing MODULE_DESCRIPTION() macro

With ARCH=m68k, make allmodconfig && make W=1 C=1 reports:
WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/ataflop.o

Add the missing invocation of the MODULE_DESCRIPTION() macro.

Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-2-b200599a315e@quicinc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

amiflop: add missing MODULE_DESCRIPTION() macro

With ARCH=m68k, make allmodconfig && make W=1 C=1 reports:
WARNING: modpost: missing MODULE_DESCRIPTION() in drivers/block/amiflop.o

Add the missing invocation of the MODULE_DESCRIPTION() macro.

Signed-off-by: Jeff Johnson <quic_jjohnson@quicinc.com>
Reviewed-by: Geert Uytterhoeven <geert@linux-m68k.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240617-md-m68k-drivers-block-v1-1-b200599a315e@quicinc.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  block, bfq: remove blkg_path()

block, bfq: remove blkg_path()

After commit 35fe6d763229 ("block: use standard blktrace API to output
cgroup info for debug notes"), the field 'bfqg->blkg_path' is not used
and hence can be removed, and therefor blkg_path() is not used anymore
and can be removed.

Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20240618032753.3502528-1-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  block: cleanup flag_{show,store}

block: cleanup flag_{show,store}

Remove a superfluous argument that flag_show and flag_store currently
take.

Signed-off-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240617044918.374608-1-joshi.k@samsung.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  block: BFQ: Refactor bfq_exit_icq() to silence sparse warning
  block: Drop locking annotation for limits_lock
  bdev: make blockdev_mnt static

block: BFQ: Refactor bfq_exit_icq() to silence sparse warning

Currently building for C=1 generates the following warning:
block/bfq-iosched.c:5498:9: warning: context imbalance in 'bfq_exit_icq' - different lock contexts for basic block

Refactor bfq_exit_icq() into a core part which loops for the actuators,
and only lock calling this routine when necessary.

Signed-off-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Link: https://lore.kernel.org/r/20240614090345.655716-4-john.g.garry@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: Drop locking annotation for limits_lock

Currently compiling block/blk-settings.c with C=1 gives the following
warning:
block/blk-settings.c:262:9: warning: context imbalance in 'queue_limits_commit_update' - wrong count at exit

request_queue.limits_lock is a mutex. Sparse locking annotation for
mutexes are currently not supported - see [0] - so drop that locking
annotation.

[0] https://lore.kernel.org/lkml/cover.1579893447.git.jbi.octave@gmail.com/T/#mbb8bda6c0a7ca7ce19f46df976a8e3b489745488

Fixes: d690cb8ae14bd ("block: add an API to atomically update queue limits")
Signed-off-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20240614090345.655716-3-john.g.garry@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

bdev: make blockdev_mnt static

The blockdev_mnt are not used outside the file bdev.c, so the modification
is defined as static.

block/bdev.c:377:17: warning: symbol 'blockdev_mnt' was not declared. Should it be static?

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
jpg: Remove closes bugzilla link
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Jiapeng Chong <jiapeng.chong@linux.alibaba.com>
Signed-off-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Fixes: 8f3a608827d1 ("bdev: open block device as files")
Tested-by: John Garry <john.g.garry@oracle.com>
Link: https://lore.kernel.org/r/20240614090345.655716-2-john.g.garry@oracle.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  dm: Remove unused macro DM_ZONE_INVALID_WP_OFST
  dm: Improve zone resource limits handling
  dm: Call dm_revalidate_zones() after setting the queue limits
  block: Improve checks on zone resource limits

dm: Remove unused macro DM_ZONE_INVALID_WP_OFST

With the switch to using the zone append emulation of the block layer
zone write plugging, the macro DM_ZONE_INVALID_WP_OFST is no longer used
in dm-zone.c. Remove its definition.

Fixes: f211268ed1f9 ("dm: Use the block layer zone append emulation")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Link: https://lore.kernel.org/r/20240611023639.89277-5-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

dm: Improve zone resource limits handling

The generic stacking of limits implemented in the block layer cannot
correctly handle stacking of zone resource limits (max open zones and
max active zones) because these limits are for an entire device but the
stacking may be for a portion of that device (e.g. a dm-linear target
that does not cover an entire block device). As a result, when DM
devices are created on top of zoned block devices, the DM device never
has any zone resource limits advertized, which is only correct if all
underlying target devices also have no zone resource limits.
If at least one target device has resource limits, the user may see
either performance issues (if the max open zone limit of the device is
exceeded) or write I/O errors if the max active zone limit of one of
the underlying target devices is exceeded.

While it is very difficult to correctly and reliably stack zone resource
limits in general, cases where targets are not sharing zone resources of
the same device can be dealt with relatively easily. Such situation
happens when a target maps all sequential zones of a zoned block device:
for such mapping, other targets mapping other parts of the same zoned
block device can only contain conventional zones and thus will not
require any zone resource to correctly handle write operations.

For a mapped device constructed with such targets, which includes mapped
devices constructed with targets mapping entire zoned block devices, the
zone resource limits can be reliably determined using the non-zero
minimum of the zone resource limits of all targets.

For mapped devices that include targets partially mapping the set of
sequential write required zones of zoned block devices, instead of
advertizing no zone resource limits, it is also better to set the mapped
device limits to the non-zero minimum of the limits of all targets. In
this case the limits for a target depend on the number of sequential
zones being mapped: if this number of zone is larger than the limits,
then the limits of the device apply and can be used. If on the other
hand the target maps a number of zones smaller than the limits, then no
limits is needed and we can assume that the target has no limits (limits
set to 0).

This commit improves zone resource limits handling as described above
by modifying dm_set_zones_restrictions() to iterate the targets of a
mapped device to evaluate the max open and max active zone limits. This
relies on an internal "stacking" of the limits of the target devices
combined with a direct counting of the number of sequential zones
mapped by the targets.
1) For a target mapping an entire zoned block device, the limits for the
   target are set to the limits of the device.
2) For a target partially mapping a zoned block device, the number of
   mapped sequential zones is used to determine the limits: if the
   target maps more sequential write required zones than the device
   limits, then the limits of the device are used as-is. If the number
   of mapped sequential zones is lower than the limits, then we assume
   that the target has no limits (limits set to 0).
As this evaluation is done for each target, the zone resource limits
for the mapped device are evaluated as the non-zero minimum of the
limits of all the targets.

For configurations resulting in unreliable limits, i.e. a table
containing a target partially mapping a zoned device, a warning message
is issued.

The counting of mapped sequential zones for the target is done using the
new function dm_device_count_zones() which performs a report zones on
the entire block device with the callback dm_device_count_zones_cb().
This count of mapped sequential zones is also used to determine if the
mapped device contains only conventional zones. This allows simplifying
dm_set_zones_restrictions() to not do a report zones just for this.
For mapped devices mapping only conventional zones, as before, the
mapped device is changed to a regular device by setting its zoned limit
to false and clearing all its zone related limits.

Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Link: https://lore.kernel.org/r/20240611023639.89277-4-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

dm: Call dm_revalidate_zones() after setting the queue limits

dm_revalidate_zones() is called from dm_set_zone_restrictions() when the
mapped device queue limits are not yet set. However,
dm_revalidate_zones() calls blk_revalidate_disk_zones() and this
function consults and modifies the mapped device queue limits. Thus,
currently, blk_revalidate_disk_zones() operates on limits that are not
yet initialized.

Fix this by moving the call to dm_revalidate_zones() out of
dm_set_zone_restrictions() and into dm_table_set_restrictions() after
executing queue_limits_set().

To further cleanup dm_set_zones_restrictions(), the message about the
type of zone append (native or emulated) is also moved inside
dm_revalidate_zones().

Fixes: 1c0e720228ad ("dm: use queue_limits_set")
Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Link: https://lore.kernel.org/r/20240611023639.89277-3-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: Improve checks on zone resource limits

Make sure that the zone resource limits of a zoned block device are
correct by checking that:
(a) If the device has a max active zones limit, make sure that the max
    open zones limit is lower than the max active zones limit.
(b) If the device has zone resource limits, check that the limits
    values are lower than the number of sequential zones of the device.
    If it is not, assume that the zoned device has no limits by setting
    the limits to 0.

For (a), a check is added to blk_validate_zoned_limits() and an error
returned if the max open zones limit exceeds the value of the max active
zone limit (if there is one).

For (b), given that we need the number of sequential zones of the zoned
device, this check is added to disk_update_zone_resources(). This is
safe to do as that function is executed with the disk queue frozen and
the check executed after queue_limits_start_update() which takes the
queue limits lock. Of note is that the early return in this function
for zoned devices that do not use zone write plugging (e.g. DM devices
using native zone append) is moved to after the new check and adjustment
of the zone resource limits so that the check applies to any zoned
device.

Signed-off-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Niklas Cassel <cassel@kernel.org>
Reviewed-by: Benjamin Marzinski <bmarzins@redhat.com>
Link: https://lore.kernel.org/r/20240611023639.89277-2-dlemoal@kernel.org
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block:
  rust: block: do not use removed queue limit API

rust: block: do not use removed queue limit API

The Rust block layer API was using the old queue limit API, which was just
removed. Use the new API instead.

Reported-by: Boqun Feng <boqun.feng@gmail.com>
Fixes: 3253aba3408a ("rust: block: introduce `kernel::block::mq` module")
Signed-off-by: Andreas Hindborg <a.hindborg@samsung.com>
Link: https://lore.kernel.org/r/20240614235350.621121-1-nmi@metaspace.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block' into for-next

* for-6.11/block: (48 commits)
  block: move integrity information into queue_limits
  block: invert the BLK_INTEGRITY_{GENERATE,VERIFY} flags
  block: bypass the STABLE_WRITES flag for protection information
  block: don't require stable pages for non-PI metadata
  block: use kstrtoul in flag_store
  block: factor out flag_{store,show} helper for integrity
  block: remove the blk_flush_integrity call in blk_integrity_unregister
  block: remove the blk_integrity_profile structure
  dm-integrity: use the nop integrity profile
  md/raid1: don't free conf on raid0_run failure
  md/raid0: don't free conf on raid0_run failure
  block: initialize integrity buffer to zero before writing it to media
  block: add special APIs for run-time disabling of discard and friends
  block: remove unused queue limits API
  sr: convert to the atomic queue limits API
  sd: convert to the atomic queue limits API
  sd: cleanup zoned queue limits initialization
  sd: factor out a sd_discard_mode helper
  sd: simplify the disable case in sd_config_discard
  sd: add a sd_disable_write_same helper
  ...

Merge branch 'for-6.11/io_uring' into for-next

* for-6.11/io_uring:
  io_uring/io-wq: make io_wq_work flags atomic
  io_uring: use 'state' consistently
  io_uring/eventfd: move eventfd handling to separate file
  io_uring/eventfd: move to more idiomatic RCU free usage
  io_uring/rsrc: Drop io_copy_iov in favor of iovec API
  io_uring: Drop per-ctx dummy_ubuf

io_uring/io-wq: make io_wq_work flags atomic

The work flags can be set/accessed from different tasks, both the
originator of the request, and the io-wq workers. While modifications
aren't concurrent, it still makes KMSAN unhappy. There's no real
downside to just making the flag reading/manipulation use proper
atomics here.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: use 'state' consistently

__io_submit_flush_completions() assigns ctx->submit_state to a local
variable and uses it in all but one spot, switch that forgotten
statement to using 'state' as well.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring/eventfd: move eventfd handling to separate file

This is pretty nicely abstracted already, but let's move it to a separate
file rather than have it in the main io_uring file. With that, we can
also move the io_ev_fd struct and enum out of global scope.

Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring/eventfd: move to more idiomatic RCU free usage

In some ways, it just "happens to work" currently with using the ops
field for both the free and signaling bit. But it depends on ordering
of operations in terms of freeing and signaling. Clean it up and use the
usual refs == 0 under RCU read side lock to determine if the ev_fd is
still valid, and use the reference to gate the freeing as well.

Fixes: 21a091b970cd ("io_uring: signal registered eventfd to process deferred task work")
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring/rsrc: Drop io_copy_iov in favor of iovec API

Instead of open coding an io_uring function to copy iovs from userspace,
rely on the existing iovec_from_user function.  While there, avoid
repeatedly zeroing the iov in the !arg case for io_sqe_buffer_register.

tested with liburing testsuite.

Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
Link: https://lore.kernel.org/r/20240523214535.31890-1-krisman@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

io_uring: Drop per-ctx dummy_ubuf

Commit 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf")
replaced it with a global static object but this stayed behind.

Fixes: 19a63c402170 ("io_uring/rsrc: keep one global dummy_ubuf")
Signed-off-by: Gabriel Krisman Bertazi <krisman@suse.de>
Link: https://lore.kernel.org/r/20240523214517.31803-1-krisman@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge branch 'for-6.11/block-limits' into for-6.11/block

Pull in block limits branch, which exists as a shared branch for both
the block and SCSI tree.

* for-6.11/block-limits: (26 commits)
  block: move integrity information into queue_limits
  block: invert the BLK_INTEGRITY_{GENERATE,VERIFY} flags
  block: bypass the STABLE_WRITES flag for protection information
  block: don't require stable pages for non-PI metadata
  block: use kstrtoul in flag_store
  block: factor out flag_{store,show} helper for integrity
  block: remove the blk_flush_integrity call in blk_integrity_unregister
  block: remove the blk_integrity_profile structure
  dm-integrity: use the nop integrity profile
  md/raid1: don't free conf on raid0_run failure
  md/raid0: don't free conf on raid0_run failure
  block: initialize integrity buffer to zero before writing it to media
  block: add special APIs for run-time disabling of discard and friends
  block: remove unused queue limits API
  sr: convert to the atomic queue limits API
  sd: convert to the atomic queue limits API
  sd: cleanup zoned queue limits initialization
  sd: factor out a sd_discard_mode helper
  sd: simplify the disable case in sd_config_discard
  sd: add a sd_disable_write_same helper
  ...

block: move integrity information into queue_limits

Move the integrity information into the queue limits so that it can be
set atomically with other queue limits, and that the sysfs changes to
the read_verify and write_generate flags are properly synchronized.
This also allows to provide a more useful helper to stack the integrity
fields, although it still is separate from the main stacking function
as not all stackable devices want to inherit the integrity settings.
Even with that it greatly simplifies the code in md and dm.

Note that the integrity field is moved as-is into the queue limits.
While there are good arguments for removing the separate blk_integrity
structure, this would cause a lot of churn and might better be done at a
later time if desired.  However the integrity field in the queue_limits
structure is now unconditional so that various ifdefs can be avoided or
replaced with IS_ENABLED().  Given that tiny size of it that seems like
a worthwhile trade off.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: invert the BLK_INTEGRITY_{GENERATE,VERIFY} flags

Invert the flags so that user set values will be able to persist
revalidating the integrity information once we switch the integrity
information to queue_limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-12-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: bypass the STABLE_WRITES flag for protection information

Currently registering a checksum-enabled (aka PI) integrity profile sets
the QUEUE_FLAG_STABLE_WRITE flag, and unregistering it clears the flag.
This can incorrectly clear the flag when the driver requires stable
writes even without PI, e.g. in case of iSCSI or NVMe/TCP with data
digest enabled.

Fix this by looking at the csum_type directly in bdev_stable_writes and
not setting the queue flag.  Also remove the blk_queue_stable_writes
helper as the only user in nvme wants to only look at the actual
QUEUE_FLAG_STABLE_WRITE flag as it inherits the integrity configuration
by other means.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240613084839.1044015-11-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: don't require stable pages for non-PI metadata

Non-PI metadata doesn't contain checksums and thus doesn't require
stable pages.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240613084839.1044015-10-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: use kstrtoul in flag_store

Use the text to integer helper that has error handling and doesn't modify
the input pointer.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-9-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: factor out flag_{store,show} helper for integrity

Factor the duplicate code for the generate and verify attributes into
common helpers.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240613084839.1044015-8-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: remove the blk_flush_integrity call in blk_integrity_unregister

Now that there are no indirect calls for PI processing there is no
way to dereference a NULL pointer here.  Additionally drivers now always
freeze the queue (or in case of stacking drivers use their internal
equivalent) around changing the integrity profile.

This is effectively a revert of commit 3df49967f6f1 ("block: flush the
integrity workqueue in blk_integrity_unregister").

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Link: https://lore.kernel.org/r/20240613084839.1044015-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: remove the blk_integrity_profile structure

Block layer integrity configuration is a bit complex right now, as it
indirects through operation vectors for a simple two-dimensional
configuration:

 a) the checksum type of none, ip checksum, crc, crc64
 b) the presence or absence of a reference tag

Remove the integrity profile, and instead add a separate csum_type flag
which replaces the existing ip-checksum field and a new flag that
indicates the presence of the reference tag.

This removes up to two layers of indirect calls, remove the need to
offload the no-op verification of non-PI metadata to a workqueue and
generally simplifies the code. The downside is that block/t10-pi.c now
has to be built into the kernel when CONFIG_BLK_DEV_INTEGRITY is
supported.  Given that both nvme and SCSI require t10-pi.ko, it is loaded
for all usual configurations that enabled CONFIG_BLK_DEV_INTEGRITY
already, though.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

dm-integrity: use the nop integrity profile

Use the block layer built-in nop profile instead of duplicating it.

Tested by:

$ dd if=/dev/urandom of=key.bin bs=512 count=1

$ cryptsetup luksFormat -q --type luks2 --integrity hmac-sha256 \
  --integrity-no-wipe /dev/nvme0n1 key.bin
$ cryptsetup luksOpen /dev/nvme0n1 luks-integrity --key-file key.bin

and then doing mkfs.xfs and simple I/O on the mount file system.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Milan Broz <gmazyland@gmail.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Reviewed-by: Hannes Reinecke <hare@suse.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

md/raid1: don't free conf on raid0_run failure

The core md code calls the ->free method which already frees conf.

Fixes: 07f1a6850c5d ("md/raid1: fail run raid1 array when active disk less than one")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

md/raid0: don't free conf on raid0_run failure

The core md code calls the ->free method which already frees conf.

Fixes: 0c031fd37f69 ("md: Move alloc/free acct bioset in to personality")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Yu Kuai <yukuai3@huawei.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: initialize integrity buffer to zero before writing it to media

Metadata added by bio_integrity_prep is using plain kmalloc, which leads
to random kernel memory being written media.  For PI metadata this is
limited to the app tag that isn't used by kernel generated metadata,
but for non-PI metadata the entire buffer leaks kernel memory.

Fix this by adding the __GFP_ZERO flag to allocations for writes.

Fixes: 7ba1ba12eeef ("block: Block layer data integrity support")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Reviewed-by: Kanchan Joshi <joshi.k@samsung.com>
Reviewed-by: Chaitanya Kulkarni <kch@nvidia.com>
Link: https://lore.kernel.org/r/20240613084839.1044015-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: add special APIs for run-time disabling of discard and friends

A few drivers optimistically try to support discard, write zeroes and
secure erase and disable the features from the I/O completion handler
if the hardware can't support them.  This disable can't be done using
the atomic queue limits API because the I/O completion handlers can't
take sleeping locks or freeze the queue.  Keep the existing clearing
of the relevant field to zero, but replace the old blk_queue_max_*
APIs with new disable APIs that force the value to 0.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Nitesh Shetty <nj.shetty@samsung.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-15-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: remove unused queue limits API

Remove all APIs that are unused now that sd and sr have been converted
to the atomic queue limits API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Nitesh Shetty <nj.shetty@samsung.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-14-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sr: convert to the atomic queue limits API

Assign all queue limits through a local queue_limits variable and
queue_limits_commit_update so that we can't race updating them from
multiple places, and free the queue when updating them so that
in-progress I/O submissions don't see half-updated limits.

Also use the chance to clean up variable names to standard ones.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-13-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: convert to the atomic queue limits API

Assign all queue limits through a local queue_limits variable and
queue_limits_commit_update so that we can't race updating them from
multiple places, and freeze the queue when updating them so that
in-progress I/O submissions don't see half-updated limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: John Garry <john.g.garry@oracle.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-12-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: cleanup zoned queue limits initialization

Consolidate setting zone-related queue limits in sd_zbc_read_zones
instead of splitting them between sd_zbc_revalidate_zones and
sd_zbc_read_zones, and move the early_zone_information initialization
in sd_zbc_read_zones above setting up the queue limits.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-11-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: factor out a sd_discard_mode helper

Split the logic to pick the right discard mode into a little helper
to prepare for further changes.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-10-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: simplify the disable case in sd_config_discard

Fall through to the main call to blk_queue_max_discard_sectors given that
max_blocks has been initialized to zero above instead of duplicating the
call.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-9-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: add a sd_disable_write_same helper

Add helper to disable WRITE SAME when it is not supported and use it
instead of sd_config_write_same in the I/O completion handler.  This
avoids touching more fields than required in the I/O completion handler
and  prepares for converting sd to use the atomic queue limits API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-8-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: add a sd_disable_discard helper

Add helper to disable discard when it is not supported and use it
instead of sd_config_discard in the I/O completion handler.  This avoids
touching more fields than required in the I/O completion handler and
prepares for converting sd to use the atomic queue limits API.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-7-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

sd: simplify the ZBC case in provisioning_mode_store

Don't reset the discard settings to no-op over and over when a user
writes to the provisioning attribute as that is already the default
mode for ZBC devices.  In hindsight we should have made writing to
the attribute fail for ZBC devices, but the code has probably been
around for far too long to change this now.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-6-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

block: take io_opt and io_min into account for max_sectors

The soft max_sectors limit is normally capped by the hardware limits and
an arbitrary upper limit enforced by the kernel, but can be modified by
the user.  A few drivers want to increase this limit (nbd, rbd) or
adjust it up or down based on hardware capabilities (sd).

Change blk_validate_limits to default max_sectors to the optimal I/O
size, or upgrade it to the preferred minimal I/O size if that is
larger than the kernel default if no optimal I/O size is provided based
on the logic in the SD driver.

This keeps the existing kernel default for drivers that do not provide
an io_opt or very big io_min value, but picks a much more useful
default for those who provide these hints, and allows to remove the
hacks to set the user max_sectors limit in nbd, rbd and sd.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Acked-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-5-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

rbd: increase io_opt again

Commit 16d80c54ad42 ("rbd: set io_min, io_opt and discard_granularity to
alloc_size") lowered the io_opt size for rbd from objset_bytes which is
4MB for typical setup to alloc_size which is typically 64KB.

The commit mostly talks about discard behavior and does mention io_min
in passing.  Reducing io_opt means reducing the readahead size, which
seems counter-intuitive given that rbd currently abuses the user
max_sectors setting to actually increase the I/O size.  Switch back
to the old setting to allow larger reads (the readahead size despite it's
name actually limits the size of any buffered read) and to prepare
for using io_opt in the max_sectors calculation and getting drivers out
of the business of overriding the max_user_sectors value.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Acked-by: Ilya Dryomov <idryomov@gmail.com>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-4-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

ubd: untagle discard vs write zeroes not support handling

Discard and Write Zeroes are different operation and implemented
by different fallocate opcodes for ubd.  If one fails the other one
can work and vice versa.

Split the code to disable the operations in ubd_handler to only
disable the operation that actually failed.

Fixes: 50109b5a03b4 ("um: Add support for DISCARD in the UBD Driver")
Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Bart Van Assche <bvanassche@acm.org>
Reviewed-by: Damien Le Moal <dlemoal@kernel.org>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-By: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-3-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

ubd: refactor the interrupt handler

Instead of a separate handler function that leaves no work in the
interrupt hanler itself, split out a per-request end I/O helper and
clean up the coding style and variable naming while we're at it.

Signed-off-by: Christoph Hellwig <hch@lst.de>
Reviewed-by: Martin K. Petersen <martin.petersen@oracle.com>
Acked-By: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Link: https://lore.kernel.org/r/20240531074837.1648501-2-hch@lst.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>

MAINTAINERS: add entry for Rust block device driver API

Add an entry for the Rust block device driver abstractions.

Signed-off-by: Andreas Hindborg <a.hindborg@samsung.com>
Link: https://lore.kernel.org/r/20240611114551.228679-4-nmi@metaspace.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>

rust: block: add rnull, Rust null_blk implementation

This patch adds an initial version of the Rust null block driver.

Signed-off-by: Andreas Hindborg <a.hindborg@samsung.com>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Link: https://lore.kernel.org/r/20240611114551.228679-3-nmi@metaspace.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>

rust: block: introduce `kernel::block::mq` module

Add initial abstractions for working with blk-mq.

This patch is a maintained, refactored subset of code originally published
by Wedson Almeida Filho <wedsonaf@gmail.com> [1].

[1] https://github.com/wedsonaf/linux/tree/f2cfd2fe0e2ca4e90994f96afe268bbd4382a891/rust/kernel/blk/mq.rs

Cc: Wedson Almeida Filho <wedsonaf@gmail.com>
Signed-off-by: Andreas Hindborg <a.hindborg@samsung.com>
Reviewed-by: Benno Lossin <benno.lossin@proton.me>
Link: https://lore.kernel.org/r/20240611114551.228679-2-nmi@metaspace.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>

Merge tag 'net-6.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from bluetooth and netfilter.

  Slim pickings this time, probably a combination of summer, DevConf.cz,
  and the end of first half of the year at corporations.

  Current release - regressions:

   - Revert "igc: fix a log entry using uninitialized netdev", it traded
     lack of netdev name in a printk() for a crash

  Previous releases - regressions:

   - Bluetooth: L2CAP: fix rejecting L2CAP_CONN_PARAM_UPDATE_REQ

   - geneve: fix incorrectly setting lengths of inner headers in the
     skb, confusing the drivers and causing mangled packets

   - sched: initialize noop_qdisc owner to avoid false-positive
     recursion detection (recursing on CPU 0), which bubbles up to user
     space as a sendmsg() error, while noop_qdisc should silently drop

   - netdevsim: fix backwards compatibility in nsim_get_iflink()

  Previous releases - always broken:

   - netfilter: ipset: fix race between namespace cleanup and gc in the
     list:set type"

* tag 'net-6.10-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (35 commits)
  bnxt_en: Adjust logging of firmware messages in case of released token in __hwrm_send()
  af_unix: Read with MSG_PEEK loops if the first unread byte is OOB
  bnxt_en: Cap the size of HWRM_PORT_PHY_QCFG forwarded response
  gve: Clear napi->skb before dev_kfree_skb_any()
  ionic: fix use after netif_napi_del()
  Revert "igc: fix a log entry using uninitialized netdev"
  net: bridge: mst: fix suspicious rcu usage in br_mst_set_state
  net: bridge: mst: pass vlan group directly to br_mst_vlan_set_state
  net/ipv6: Fix the RT cache flush via sysctl using a previous delay
  net: stmmac: replace priv->speed with the portTransmitRate from the tc-cbs parameters
  gve: ignore nonrelevant GSO type bits when processing TSO headers
  net: pse-pd: Use EOPNOTSUPP error code instead of ENOTSUPP
  netfilter: Use flowlabel flow key when re-routing mangled packets
  netfilter: ipset: Fix race between namespace cleanup and gc in the list:set type
  netfilter: nft_inner: validate mandatory meta and payload
  tcp: use signed arithmetic in tcp_rtx_probe0_timed_out()
  mailmap: map Geliang's new email address
  mptcp: pm: update add_addr counters after connect
  mptcp: pm: inc RmAddr MIB counter once per RM_ADDR ID
  mptcp: ensure snd_una is properly initialized on connect
  ...

Merge tag 'nfs-for-6.10-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs

Pull NFS client fixes from Trond Myklebust:
 "Bugfixes:
   - NFSv4.2: Fix a memory leak in nfs4_set_security_label
   - NFSv2/v3: abort nfs_atomic_open_v23 if the name is too long.
   - NFS: Add appropriate memory barriers to the sillyrename code
   - Propagate readlink errors in nfs_symlink_filler
   - NFS: don't invalidate dentries on transient errors
   - NFS: fix unnecessary synchronous writes in random write workloads
   - NFSv4.1: enforce rootpath check when deciding whether or not to trunk

  Other:
   - Change email address for Trond Myklebust due to email server concerns"

* tag 'nfs-for-6.10-2' of git://git.linux-nfs.org/projects/trondmy/linux-nfs:
  NFS: add barriers when testing for NFS_FSDATA_BLOCKED
  SUNRPC: return proper error from gss_wrap_req_priv
  NFSv4.1 enforce rootpath check in fs_location query
  NFS: abort nfs_atomic_open_v23 if name is too long.
  nfs: don't invalidate dentries on transient errors
  nfs: Avoid flushing many pages with NFS_FILE_SYNC
  nfs: propagate readlink errors in nfs_symlink_filler
  MAINTAINERS: Change email address for Trond Myklebust
  NFSv4: Fix memory leak in nfs4_set_security_label

Merge tag 'fixes-2024-06-13' of git://git.kernel.org/pub/scm/linux/kernel/git/rppt/memblock

Pull memblock fixes from Mike Rapoport:
 "Fix validation of NUMA coverage.

  memblock_validate_numa_coverage() was checking for a unset node ID
  using NUMA_NO_NODE, but x86 used MAX_NUMNODES when no node ID was
  specified by buggy firmware.

  Update memblock to substitute MAX_NUMNODES with NUMA_NO_NODE in
  memblock_set_node() and use NUMA_NO_NODE in x86::numa_init()"

* tag 'fixes-2024-06-13' of git://git.kernel.org/pub/scm/linux/kernel/git/rppt/memblock:
  x86/mm/numa: Use NUMA_NO_NODE when calling memblock_set_node()
  memblock: make memblock_set_node() also warn about use of MAX_NUMNODES

bnxt_en: Adjust logging of firmware messages in case of released token in __hwrm_send()

In case of token is released due to token->state == BNXT_HWRM_DEFERRED,
released token (set to NULL) is used in log messages. This issue is
expected to be prevented by HWRM_ERR_CODE_PF_UNAVAILABLE error code. But
this error code is returned by recent firmware. So some firmware may not
return it. This may lead to NULL pointer dereference.
Adjust this issue by adding token pointer check.

Found by Linux Verification Center (linuxtesting.org) with SVACE.

Fixes: 8fa4219dba8e ("bnxt_en: add dynamic debug support for HWRM messages")
Suggested-by: Michael Chan <michael.chan@broadcom.com>
Signed-off-by: Aleksandr Mishin <amishin@t-argos.ru>
Reviewed-by: Wojciech Drewek <wojciech.drewek@intel.com>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20240611082547.12178-1-amishin@t-argos.ru
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

af_unix: Read with MSG_PEEK loops if the first unread byte is OOB

Read with MSG_PEEK flag loops if the first byte to read is an OOB byte.
commit 22dd70eb2c3d ("af_unix: Don't peek OOB data without MSG_OOB.")
addresses the loop issue but does not address the issue that no data
beyond OOB byte can be read.

>>> from socket import *
>>> c1, c2 = socketpair(AF_UNIX, SOCK_STREAM)
>>> c1.send(b'a', MSG_OOB)
1
>>> c1.send(b'b')
1
>>> c2.recv(1, MSG_PEEK | MSG_DONTWAIT)
b'b'

>>> from socket import *
>>> c1, c2 = socketpair(AF_UNIX, SOCK_STREAM)
>>> c2.setsockopt(SOL_SOCKET, SO_OOBINLINE, 1)
>>> c1.send(b'a', MSG_OOB)
1
>>> c1.send(b'b')
1
>>> c2.recv(1, MSG_PEEK | MSG_DONTWAIT)
b'a'
>>> c2.recv(1, MSG_PEEK | MSG_DONTWAIT)
b'a'
>>> c2.recv(1, MSG_DONTWAIT)
b'a'
>>> c2.recv(1, MSG_PEEK | MSG_DONTWAIT)
b'b'
>>>

Fixes: 314001f0bf92 ("af_unix: Add OOB support")
Signed-off-by: Rao Shoaib <Rao.Shoaib@oracle.com>
Reviewed-by: Kuniyuki Iwashima <kuniyu@amazon.com>
Link: https://lore.kernel.org/r/20240611084639.2248934-1-Rao.Shoaib@oracle.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

bnxt_en: Cap the size of HWRM_PORT_PHY_QCFG forwarded response

Firmware interface 1.10.2.118 has increased the size of
HWRM_PORT_PHY_QCFG response beyond the maximum size that can be
forwarded.  When the VF's link state is not the default auto state,
the PF will need to forward the response back to the VF to indicate
the forced state.  This regression may cause the VF to fail to
initialize.

Fix it by capping the HWRM_PORT_PHY_QCFG response to the maximum
96 bytes.  The SPEEDS2_SUPPORTED flag needs to be cleared because the
new speeds2 fields are beyond the legacy structure.  Also modify
bnxt_hwrm_fwd_resp() to print a warning if the message size exceeds 96
bytes to make this failure more obvious.

Fixes: 84a911db8305 ("bnxt_en: Update firmware interface to 1.10.2.118")
Reviewed-by: Somnath Kotur <somnath.kotur@broadcom.com>
Reviewed-by: Pavan Chebbi <pavan.chebbi@broadcom.com>
Signed-off-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/20240612231736.57823-1-michael.chan@broadcom.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

gve: Clear napi->skb before dev_kfree_skb_any()

gve_rx_free_skb incorrectly leaves napi->skb referencing an skb after it
is freed with dev_kfree_skb_any(). This can result in a subsequent call
to napi_get_frags returning a dangling pointer.

Fix this by clearing napi->skb before the skb is freed.

Fixes: 9b8dd5e5ea48 ("gve: DQO: Add RX path")
Cc: stable@vger.kernel.org
Reported-by: Shailend Chand <shailend@google.com>
Signed-off-by: Ziwei Xiao <ziweixiao@google.com>
Reviewed-by: Harshitha Ramamurthy <hramamurthy@google.com>
Reviewed-by: Shailend Chand <shailend@google.com>
Reviewed-by: Praveen Kaligineedi <pkaligineedi@google.com>
Link: https://lore.kernel.org/r/20240612001654.923887-1-ziweixiao@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

ionic: fix use after netif_napi_del()

When queues are started, netif_napi_add() and napi_enable() are called.
If there are 4 queues and only 3 queues are used for the current
configuration, only 3 queues' napi should be registered and enabled.
The ionic_qcq_enable() checks whether the .poll pointer is not NULL for
enabling only the using queue' napi. Unused queues' napi will not be
registered by netif_napi_add(), so the .poll pointer indicates NULL.
But it couldn't distinguish whether the napi was unregistered or not
because netif_napi_del() doesn't reset the .poll pointer to NULL.
So, ionic_qcq_enable() calls napi_enable() for the queue, which was
unregistered by netif_napi_del().

Reproducer:
   ethtool -L <interface name> rx 1 tx 1 combined 0
   ethtool -L <interface name> rx 0 tx 0 combined 1
   ethtool -L <interface name> rx 0 tx 0 combined 4

Splat looks like:
kernel BUG at net/core/dev.c:6666!
Oops: invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
CPU: 3 PID: 1057 Comm: kworker/3:3 Not tainted 6.10.0-rc2+ #16
Workqueue: events ionic_lif_deferred_work [ionic]
RIP: 0010:napi_enable+0x3b/0x40
Code: 48 89 c2 48 83 e2 f6 80 b9 61 09 00 00 00 74 0d 48 83 bf 60 01 00 00 00 74 03 80 ce 01 f0 4f
RSP: 0018:ffffb6ed83227d48 EFLAGS: 00010246
RAX: 0000000000000000 RBX: ffff97560cda0828 RCX: 0000000000000029
RDX: 0000000000000001 RSI: 0000000000000000 RDI: ffff97560cda0a28
RBP: ffffb6ed83227d50 R08: 0000000000000400 R09: 0000000000000001
R10: 0000000000000001 R11: 0000000000000001 R12: 0000000000000000
R13: ffff97560ce3c1a0 R14: 0000000000000000 R15: ffff975613ba0a20
FS:  0000000000000000(0000) GS:ffff975d5f780000(0000) knlGS:0000000000000000
CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f8f734ee200 CR3: 0000000103e50000 CR4: 00000000007506f0
PKRU: 55555554
Call Trace:
 <TASK>
 ? die+0x33/0x90
 ? do_trap+0xd9/0x100
 ? napi_enable+0x3b/0x40
 ? do_error_trap+0x83/0xb0
 ? napi_enable+0x3b/0x40
 ? napi_enable+0x3b/0x40
 ? exc_invalid_op+0x4e/0x70
 ? napi_enable+0x3b/0x40
 ? asm_exc_invalid_op+0x16/0x20
 ? napi_enable+0x3b/0x40
 ionic_qcq_enable+0xb7/0x180 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8]
 ionic_start_queues+0xc4/0x290 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8]
 ionic_link_status_check+0x11c/0x170 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8]
 ionic_lif_deferred_work+0x129/0x280 [ionic 59bdfc8a035436e1c4224ff7d10789e3f14643f8]
 process_one_work+0x145/0x360
 worker_thread+0x2bb/0x3d0
 ? __pfx_worker_thread+0x10/0x10
 kthread+0xcc/0x100
 ? __pfx_kthread+0x10/0x10
 ret_from_fork+0x2d/0x50
 ? __pfx_kthread+0x10/0x10
 ret_from_fork_asm+0x1a/0x30

Fixes: 0f3154e6bcb3 ("ionic: Add Tx and Rx handling")
Signed-off-by: Taehee Yoo <ap420073@gmail.com>
Reviewed-by: Brett Creeley <brett.creeley@amd.com>
Reviewed-by: Shannon Nelson <shannon.nelson@amd.com>
Link: https://lore.kernel.org/r/20240612060446.1754392-1-ap420073@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Revert "igc: fix a log entry using uninitialized netdev"

This reverts commit 86167183a17e03ec77198897975e9fdfbd53cb0b.

igc_ptp_init() needs to be called before igc_reset(), otherwise kernel
crash could be observed. Following the corresponding discussion [1] and
[2] revert this commit.

Link: https://lore.kernel.org/all/8fb634f8-7330-4cf4-a8ce-485af9c0a61a@intel.com/
Link: https://lore.kernel.org/all/87o78rmkhu.fsf@intel.com/
Fixes: 86167183a17e ("igc: fix a log entry using uninitialized netdev")
Signed-off-by: Sasha Neftin <sasha.neftin@intel.com>
Tested-by: Naama Meir <naamax.meir@linux.intel.com>
Signed-off-by: Tony Nguyen <anthony.l.nguyen@intel.com>
Link: https://lore.kernel.org/r/20240611162456.961631-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge branch 'net-bridge-mst-fix-suspicious-rcu-usage-warning'

Nikolay Aleksandrov says:

====================
net: bridge: mst: fix suspicious rcu usage warning

This set fixes a suspicious RCU usage warning triggered by syzbot[1] in
the bridge's MST code. After I converted br_mst_set_state to RCU, I
forgot to update the vlan group dereference helper. Fix it by using
the proper helper, in order to do that we need to pass the vlan group
which is already obtained correctly by the callers for their respective
context. Patch 01 is a requirement for the fix in patch 02.

Note I did consider rcu_dereference_rtnl() but the churn is much bigger
and in every part of the bridge. We can do that as a cleanup in
net-next.

[1] https://syzkaller.appspot.com/bug?extid=9bbe2de1bc9d470eb5fe
 =============================
 WARNING: suspicious RCU usage
 6.10.0-rc2-syzkaller-00235-g8a92980606e3 #0 Not tainted
 -----------------------------
 net/bridge/br_private.h:1599 suspicious rcu_dereference_protected() usage!

 other info that might help us debug this:

 rcu_scheduler_active = 2, debug_locks = 1
 4 locks held by syz-executor.1/5374:
  #0: ffff888022d50b18 (&mm->mmap_lock){++++}-{3:3}, at: mmap_read_lock include/linux/mmap_lock.h:144 [inline]
  #0: ffff888022d50b18 (&mm->mmap_lock){++++}-{3:3}, at: __mm_populate+0x1b0/0x460 mm/gup.c:2111
  #1: ffffc90000a18c00 ((&p->forward_delay_timer)){+.-.}-{0:0}, at: call_timer_fn+0xc0/0x650 kernel/time/timer.c:1789
  #2: ffff88805fb2ccb8 (&br->lock){+.-.}-{2:2}, at: spin_lock include/linux/spinlock.h:351 [inline]
  #2: ffff88805fb2ccb8 (&br->lock){+.-.}-{2:2}, at: br_forward_delay_timer_expired+0x50/0x440 net/bridge/br_stp_timer.c:86
  #3: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: rcu_lock_acquire include/linux/rcupdate.h:329 [inline]
  #3: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: rcu_read_lock include/linux/rcupdate.h:781 [inline]
  #3: ffffffff8e333fa0 (rcu_read_lock){....}-{1:2}, at: br_mst_set_state+0x171/0x7a0 net/bridge/br_mst.c:105

 stack backtrace:
 CPU: 1 PID: 5374 Comm: syz-executor.1 Not tainted 6.10.0-rc2-syzkaller-00235-g8a92980606e3 #0
 Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 04/02/2024
 Call Trace:
  <IRQ>
  __dump_stack lib/dump_stack.c:88 [inline]
  dump_stack_lvl+0x241/0x360 lib/dump_stack.c:114
  lockdep_rcu_suspicious+0x221/0x340 kernel/locking/lockdep.c:6712
  nbp_vlan_group net/bridge/br_private.h:1599 [inline]
  br_mst_set_state+0x29e/0x7a0 net/bridge/br_mst.c:106
  br_set_state+0x28a/0x7b0 net/bridge/br_stp.c:47
  br_forward_delay_timer_expired+0x176/0x440 net/bridge/br_stp_timer.c:88
  call_timer_fn+0x18e/0x650 kernel/time/timer.c:1792
  expire_timers kernel/time/timer.c:1843 [inline]
  __run_timers kernel/time/timer.c:2417 [inline]
  __run_timer_base+0x66a/0x8e0 kernel/time/timer.c:2428
  run_timer_base kernel/time/timer.c:2437 [inline]
  run_timer_softirq+0xb7/0x170 kernel/time/timer.c:2447
  handle_softirqs+0x2c4/0x970 kernel/softirq.c:554
  __do_softirq kernel/softirq.c:588 [inline]
  invoke_softirq kernel/softirq.c:428 [inline]
  __irq_exit_rcu+0xf4/0x1c0 kernel/softirq.c:637
  irq_exit_rcu+0x9/0x30 kernel/softirq.c:649
  instr_sysvec_apic_timer_interrupt arch/x86/kernel/apic/apic.c:1043 [inline]
  sysvec_apic_timer_interrupt+0xa6/0xc0 arch/x86/kernel/apic/apic.c:1043
  </IRQ>
  <TASK>
====================

Link: https://lore.kernel.org/r/20240609103654.914987-1-razor@blackwall.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: bridge: mst: fix suspicious rcu usage in br_mst_set_state

I converted br_mst_set_state to RCU to avoid a vlan use-after-free
but forgot to change the vlan group dereference helper. Switch to vlan
group RCU deref helper to fix the suspicious rcu usage warning.

Fixes: 3a7c1661ae13 ("net: bridge: mst: fix vlan use-after-free")
Reported-by: syzbot+9bbe2de1bc9d470eb5fe@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=9bbe2de1bc9d470eb5fe
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/r/20240609103654.914987-3-razor@blackwall.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net: bridge: mst: pass vlan group directly to br_mst_vlan_set_state

Pass the already obtained vlan group pointer to br_mst_vlan_set_state()
instead of dereferencing it again. Each caller has already correctly
dereferenced it for their context. This change is required for the
following suspicious RCU dereference fix. No functional changes
intended.

Fixes: 3a7c1661ae13 ("net: bridge: mst: fix vlan use-after-free")
Reported-by: syzbot+9bbe2de1bc9d470eb5fe@syzkaller.appspotmail.com
Closes: https://syzkaller.appspot.com/bug?extid=9bbe2de1bc9d470eb5fe
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
Link: https://lore.kernel.org/r/20240609103654.914987-2-razor@blackwall.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

net/ipv6: Fix the RT cache flush via sysctl using a previous delay

The net.ipv6.route.flush system parameter takes a value which specifies
a delay used during the flush operation for aging exception routes. The
written value is however not used in the currently requested flush and
instead utilized only in the next one.

A problem is that ipv6_sysctl_rtcache_flush() first reads the old value
of net->ipv6.sysctl.flush_delay into a local delay variable and then
calls proc_dointvec() which actually updates the sysctl based on the
provided input.

Fix the problem by switching the order of the two operations.

Fixes: 4990509f19e8 ("[NETNS][IPV6]: Make sysctls route per namespace.")
Signed-off-by: Petr Pavlu <petr.pavlu@suse.com>
Reviewed-by: David Ahern <dsahern@kernel.org>
Link: https://lore.kernel.org/r/20240607112828.30285-1-petr.pavlu@suse.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rmk/linux

Pull ARM and clkdev fixes from Russell King:

 - Fix clkdev - erroring out on long strings causes boot failures, so
   don't do this. Still warn about the over-sized strings (which will
   never match and thus their registration with clkdev is useless)

 - Fix for ftrace with frame pointer unwinder with recent GCC changing
   the way frames are stacked.

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rmk/linux:
  ARM: 9405/1: ftrace: Don't assume stack frames are contiguous in memory
  clkdev: don't fail clkdev_alloc() if over-sized

Merge tag 'nf-24-06-11' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf

Pablo Neira Ayuso says:

====================
Netfilter fixes for net

The following patchset contains Netfilter fixes for net:

Patch #1 fixes insufficient sanitization of netlink attributes for the
 inner expression which can trigger nul-pointer dereference,
 from Davide Ornaghi.

Patch #2 address a report that there is a race condition between
         namespace cleanup and the garbage collection of the list:set
         type. This patch resolves this issue with other minor issues
 as well, from Jozsef Kadlecsik.

Patch #3 ip6_route_me_harder() ignores flowlabel/dsfield when ip dscp
 has been mangled, this unbreaks ip6 dscp set $v,
 from Florian Westphal.

All of these patches address issues that are present in several releases.

* tag 'nf-24-06-11' of git://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf:
  netfilter: Use flowlabel flow key when re-routing mangled packets
  netfilter: ipset: Fix race between namespace cleanup and gc in the list:set type
  netfilter: nft_inner: validate mandatory meta and payload
====================

Link: https://lore.kernel.org/r/20240611220323.413713-1-pablo@netfilter.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>