xfs: fix xfs_inodegc_stop racing with mod_delayed_work

commit 2254a7396a0ca6309854948ee1c0a33fa4268cec upstream.

syzbot reported this warning from the faux inodegc shrinker that tries
to kick off inodegc work:

------------[ cut here ]------------
WARNING: CPU: 1 PID: 102 at kernel/workqueue.c:1445 __queue_work+0xd44/0x1120 kernel/workqueue.c:1444
RIP: 0010:__queue_work+0xd44/0x1120 kernel/workqueue.c:1444
Call Trace:
 __queue_delayed_work+0x1c8/0x270 kernel/workqueue.c:1672
 mod_delayed_work_on+0xe1/0x220 kernel/workqueue.c:1746
 xfs_inodegc_shrinker_scan fs/xfs/xfs_icache.c:2212 [inline]
 xfs_inodegc_shrinker_scan+0x250/0x4f0 fs/xfs/xfs_icache.c:2191
 do_shrink_slab+0x428/0xaa0 mm/vmscan.c:853
 shrink_slab+0x175/0x660 mm/vmscan.c:1013
 shrink_one+0x502/0x810 mm/vmscan.c:5343
 shrink_many mm/vmscan.c:5394 [inline]
 lru_gen_shrink_node mm/vmscan.c:5511 [inline]
 shrink_node+0x2064/0x35f0 mm/vmscan.c:6459
 kswapd_shrink_node mm/vmscan.c:7262 [inline]
 balance_pgdat+0xa02/0x1ac0 mm/vmscan.c:7452
 kswapd+0x677/0xd60 mm/vmscan.c:7712
 kthread+0x2e8/0x3a0 kernel/kthread.c:376
 ret_from_fork+0x1f/0x30 arch/x86/entry/entry_64.S:308

This warning corresponds to this code in __queue_work:

/*
 * For a draining wq, only works from the same workqueue are
 * allowed. The __WQ_DESTROYING helps to spot the issue that
 * queues a new work item to a wq after destroy_workqueue(wq).
 */
if (unlikely(wq->flags & (__WQ_DESTROYING | __WQ_DRAINING) &&
     WARN_ON_ONCE(!is_chained_work(wq))))
return;

For this to trip, we must have a thread draining the inodedgc workqueue
and a second thread trying to queue inodegc work to that workqueue.
This can happen if freezing or a ro remount race with reclaim poking our
faux inodegc shrinker and another thread dropping an unlinked O_RDONLY
file:

Thread 0 Thread 1 Thread 2

xfs_inodegc_stop

xfs_inodegc_shrinker_scan
xfs_is_inodegc_enabled
<yes, will continue>

xfs_clear_inodegc_enabled
xfs_inodegc_queue_all
<list empty, do not queue inodegc worker>

xfs_inodegc_queue
<add to list>
xfs_is_inodegc_enabled
<no, returns>

drain_workqueue
<set WQ_DRAINING>

llist_empty
<no, will queue list>
mod_delayed_work_on(..., 0)
__queue_work
<sees WQ_DRAINING, kaboom>

In other words, everything between the access to inodegc_enabled state
and the decision to poke the inodegc workqueue requires some kind of
coordination to avoid the WQ_DRAINING state.  We could perhaps introduce
a lock here, but we could also try to eliminate WQ_DRAINING from the
picture.

We could replace the drain_workqueue call with a loop that flushes the
workqueue and queues workers as long as there is at least one inode
present in the per-cpu inodegc llists.  We've disabled inodegc at this
point, so we know that the number of queued inodes will eventually hit
zero as long as xfs_inodegc_start cannot reactivate the workers.

There are four callers of xfs_inodegc_start.  Three of them come from the
VFS with s_umount held: filesystem thawing, failed filesystem freezing,
and the rw remount transition.  The fourth caller is mounting rw (no
remount or freezing possible).

There are three callers ofs xfs_inodegc_stop.  One is unmounting (no
remount or thaw possible).  Two of them come from the VFS with s_umount
held: fs freezing and ro remount transition.

Hence, it is correct to replace the drain_workqueue call with a loop
that drains the inodegc llists.

Fixes: 6191cf3ad59f ("xfs: flush inodegc workqueue tasks before cancel")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: disable reaping in fscounters scrub

commit 2d5f38a31980d7090f5bf91021488dc61a0ba8ee upstream.

The fscounters scrub code doesn't work properly because it cannot
quiesce updates to the percpu counters in the filesystem, hence it
returns false corruption reports.  This has been fixed properly in
one of the online repair patchsets that are under review by replacing
the xchk_disable_reaping calls with an exclusive filesystem freeze.
Disabling background gc isn't sufficient to fix the problem.

In other words, scrub doesn't need to call xfs_inodegc_stop, which is
just as well since it wasn't correct to allow scrub to call
xfs_inodegc_start when something else could be calling xfs_inodegc_stop
(e.g. trying to freeze the filesystem).

Neuter the scrubber for now, and remove the xchk_*_reaping functions.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: check that per-cpu inodegc workers actually run on that cpu

commit b37c4c8339cd394ea6b8b415026603320a185651 upstream.

Now that we've allegedly worked out the problem of the per-cpu inodegc
workers being scheduled on the wrong cpu, let's put in a debugging knob
to let us know if a worker ever gets mis-scheduled again.

Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

xfs: explicitly specify cpu when forcing inodegc delayed work to run immediately

commit 03e0add80f4cf3f7393edb574eeb3a89a1db7758 upstream.

I've been noticing odd racing behavior in the inodegc code that could
only be explained by one cpu adding an inode to its inactivation llist
at the same time that another cpu is processing that cpu's llist.
Preemption is disabled between get/put_cpu_ptr, so the only explanation
is scheduler mayhem.  I inserted the following debug code into
xfs_inodegc_worker (see the next patch):

ASSERT(gc->cpu == smp_processor_id());

This assertion tripped during overnight tests on the arm64 machines, but
curiously not on x86_64.  I think we haven't observed any resource leaks
here because the lockfree list code can handle simultaneous llist_add
and llist_del_all functions operating on the same list.  However, the
whole point of having percpu inodegc lists is to take advantage of warm
memory caches by inactivating inodes on the last processor to touch the
inode.

The incorrect scheduling seems to occur after an inodegc worker is
subjected to mod_delayed_work().  This wraps mod_delayed_work_on with
WORK_CPU_UNBOUND specified as the cpu number.  Unbound allows for
scheduling on any cpu, not necessarily the same one that scheduled the
work.

Because preemption is disabled for as long as we have the gc pointer, I
think it's safe to use current_cpu() (aka smp_processor_id) to queue the
delayed work item on the correct cpu.

Fixes: 7cf2b0f9611b ("xfs: bound maximum wait time for inodegc work")
Signed-off-by: Darrick J. Wong <djwong@kernel.org>
Reviewed-by: Dave Chinner <dchinner@redhat.com>
Signed-off-by: Dave Chinner <david@fromorbit.com>
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Acked-by: Darrick J. Wong <djwong@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs: no need to check source

commit 66d8fc0539b0d49941f313c9509a8384e4245ac1 upstream.

The @source inode must be valid. It is even checked via IS_SWAPFILE()
above making it pretty clear. So no need to check it when we unlock.

What doesn't need to exist is the @target inode. The lock_two_inodes()
helper currently swaps the @inode1 and @inode2 arguments if @inode1 is
NULL to have consistent lock class usage. However, we know that at least
for vfs_rename() that @inode1 is @source and thus is never NULL as per
above. We also know that @source is a different inode than @target as
that is checked right at the beginning of vfs_rename(). So we know that
@source is valid and locked and that @target is locked. So drop the
check whether @source is non-NULL.

Fixes: 28eceeda130f ("fs: Lock moved directories")
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Closes: https://lore.kernel.org/r/202307030026.9sE2pk2x-lkp@intel.com
Message-Id: <20230703-vfs-rename-source-v1-1-37eebb29b65b@kernel.org>
[brauner: use commit message from patch I sent concurrently]
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

blktrace: use inline function for blk_trace_remove() while blktrace is disabled

commit cbe7cff4a76bc749dd70264ca5cf924e2adf9296 upstream.

If config is disabled, call blk_trace_remove() directly will trigger
build warning, hence use inline function instead, prepare to fix
blktrace debugfs entries leakage.

Signed-off-by: Yu Kuai <yukuai3@huawei.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20230610022003.2557284-2-yukuai1@huaweicloud.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

leds: trigger: netdev: Recheck NETDEV_LED_MODE_LINKUP on dev rename

commit cee4bd16c3195a701be683f7da9e88c6e11acb73 upstream.

Dev can be renamed also while up for supported device. We currently
wrongly clear the NETDEV_LED_MODE_LINKUP flag on NETDEV_CHANGENAME
event.

Fix this by rechecking if the carrier is ok on NETDEV_CHANGENAME and
correctly set the NETDEV_LED_MODE_LINKUP bit.

Fixes: 5f820ed52371 ("leds: trigger: netdev: fix handling on interface rename")
Cc: stable@vger.kernel.org # v5.5+
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Lee Jones <lee@kernel.org>
Link: https://lore.kernel.org/r/20230419210743.3594-2-ansuelsmth@gmail.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: orion5x: fix d2net gpio initialization

commit f8ef1233939495c405a9faa4bd1ae7d3f581bae4 upstream.

The DT version of this board has a custom file with the gpio
device. However, it does nothing because the d2net_init()
has no caller or prototype:

arch/arm/mach-orion5x/board-d2net.c:101:13: error: no previous prototype for 'd2net_init'

Call it from the board-dt file as intended.

Fixes: 94b0bd366e36 ("ARM: orion5x: convert d2net to Device Tree")
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230516153109.514251-10-arnd@kernel.org
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ARM: dts: qcom: ipq4019: fix broken NAND controller properties override

commit edcbdd57de499305e2a3737d4a73fe387f71d84c upstream.

After renaming NAND controller node name from "qpic-nand" to
"nand-controller", the board DTS/DTSI also have to be updated:

  Warning (unit_address_vs_reg): /soc/qpic-nand@79b0000: node has a unit name, but no reg or ranges property

Cc: <stable@vger.kernel.org>
Fixes: 9e1e00f18afc ("ARM: dts: qcom: Fix node name for NAND controller node")
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org>
Reviewed-by: Konrad Dybcio <konrad.dybcio@linaro.org>
Reviewed-by: Manivannan Sadhasivam <mani@kernel.org>
Signed-off-by: Bjorn Andersson <andersson@kernel.org>
Link: https://lore.kernel.org/r/20230420072811.36947-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

regulator: tps65219: Fix matching interrupts for their regulators

commit f050e56de80591fee55bedbdf5b6b998c740cd0c upstream.

The driver's probe() first registers regulators in a loop and then in a
second loop passes them as irq data to the interrupt handlers.  However
the function to get the regulator for given name
tps65219_get_rdev_by_name() was a no-op due to argument passed by value,
not pointer, thus the second loop assigned always same value - from
previous loop.  The interrupts, when fired, where executed with wrong
data.  Compiler also noticed it:

  drivers/regulator/tps65219-regulator.c: In function ‘tps65219_get_rdev_by_name’:
  drivers/regulator/tps65219-regulator.c:292:60: error: parameter ‘dev’ set but not used [-Werror=unused-but-set-parameter]

Fixes: c12ac5fc3e0a ("regulator: drivers: Add TI TPS65219 PMIC regulators support")
Cc: <stable@vger.kernel.org
Signed-off-by: Krzysztof Kozlowski <krzysztof.kozlowski@linaro.org
Reviewed-by: Markus Schneider-Pargmann <msp@baylibre.com
Link: https://lore.kernel.org/r/20230507144656.192800-1-krzysztof.kozlowski@linaro.org
Signed-off-by: Mark Brown <broonie@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ASoC: mediatek: mt8173: Fix snd_soc_component_initialize error path

commit a46d37012a5be1737393b8f82fd35665e4556eee upstream.

If the second component fails to initialize, cleanup the first on.

Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: stable@kernel.org
Fixes: f1b5bf07365d ("ASoC: mt2701/mt8173: replace platform to component")
Signed-off-by: Ricardo Ribalda Delgado <ribalda@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20230612-mt8173-fixup-v2-1-432aa99ce24d@chromium.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ASoC: mediatek: mt8173: Fix irq error path

commit f9c058d14f4fe23ef523a7ff73734d51c151683c upstream.

After reordering the irq probe, the error path was not properly done.
Lets fix it.

Reported-by: Dan Carpenter <dan.carpenter@linaro.org>
Cc: stable@kernel.org
Fixes: 4cbb264d4e91 ("ASoC: mediatek: mt8173: Enable IRQ when pdata is ready")
Signed-off-by: Ricardo Ribalda Delgado <ribalda@chromium.org>
Reviewed-by: AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>
Link: https://lore.kernel.org/r/20230612-mt8173-fixup-v2-2-432aa99ce24d@chromium.org
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: do not BUG_ON() on tree mod log failure at __btrfs_cow_block()

commit 40b0a749388517de244643c09bdbb98f7dcb6ef1 upstream.

At __btrfs_cow_block(), instead of doing a BUG_ON() in case we fail to
record a tree mod log root insertion operation, do a transaction abort
instead. There's really no need for the BUG_ON(), we can properly
release all resources in this context and turn the filesystem to RO mode
and in an error state instead.

CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: fix extent buffer leak after tree mod log failure at split_node()

commit ede600e497b1461d06d22a7d17703d9096868bc3 upstream.

At split_node(), if we fail to log the tree mod log copy operation, we
return without unlocking the split extent buffer we just allocated and
without decrementing the reference we own on it. Fix this by unlocking
it and decrementing the ref count before returning.

Fixes: 5de865eebb83 ("Btrfs: fix tree mod logging")
CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: fix race when deleting quota root from the dirty cow roots list

commit b31cb5a6eb7a48b0a7bfdf06832b1fd5088d8c79 upstream.

When disabling quotas we are deleting the quota root from the list
fs_info->dirty_cowonly_roots without taking the lock that protects it,
which is struct btrfs_fs_info::trans_lock. This unsynchronized list
manipulation may cause chaos if there's another concurrent manipulation
of this list, such as when adding a root to it with
ctree.c:add_root_to_dirty_list().

This can result in all sorts of weird failures caused by a race, such as
the following crash:

  [337571.278245] general protection fault, probably for non-canonical address 0xdead000000000108: 0000 [#1] PREEMPT SMP PTI
  [337571.278933] CPU: 1 PID: 115447 Comm: btrfs Tainted: G        W          6.4.0-rc6-btrfs-next-134+ #1
  [337571.279153] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
  [337571.279572] RIP: 0010:commit_cowonly_roots+0x11f/0x250 [btrfs]
  [337571.279928] Code: 85 38 06 00 (...)
  [337571.280363] RSP: 0018:ffff9f63446efba0 EFLAGS: 00010206
  [337571.280582] RAX: ffff942d98ec2638 RBX: ffff9430b82b4c30 RCX: 0000000449e1c000
  [337571.280798] RDX: dead000000000100 RSI: ffff9430021e4900 RDI: 0000000000036070
  [337571.281015] RBP: ffff942d98ec2000 R08: ffff942d98ec2000 R09: 000000000000015b
  [337571.281254] R10: 0000000000000009 R11: 0000000000000001 R12: ffff942fe8fbf600
  [337571.281476] R13: ffff942dabe23040 R14: ffff942dabe20800 R15: ffff942d92cf3b48
  [337571.281723] FS:  00007f478adb7340(0000) GS:ffff94349fa40000(0000) knlGS:0000000000000000
  [337571.281950] CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  [337571.282184] CR2: 00007f478ab9a3d5 CR3: 000000001e02c001 CR4: 0000000000370ee0
  [337571.282416] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  [337571.282647] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  [337571.282874] Call Trace:
  [337571.283101]  <TASK>
  [337571.283327]  ? __die_body+0x1b/0x60
  [337571.283570]  ? die_addr+0x39/0x60
  [337571.283796]  ? exc_general_protection+0x22e/0x430
  [337571.284022]  ? asm_exc_general_protection+0x22/0x30
  [337571.284251]  ? commit_cowonly_roots+0x11f/0x250 [btrfs]
  [337571.284531]  btrfs_commit_transaction+0x42e/0xf90 [btrfs]
  [337571.284803]  ? _raw_spin_unlock+0x15/0x30
  [337571.285031]  ? release_extent_buffer+0x103/0x130 [btrfs]
  [337571.285305]  reset_balance_state+0x152/0x1b0 [btrfs]
  [337571.285578]  btrfs_balance+0xa50/0x11e0 [btrfs]
  [337571.285864]  ? __kmem_cache_alloc_node+0x14a/0x410
  [337571.286086]  btrfs_ioctl+0x249a/0x3320 [btrfs]
  [337571.286358]  ? mod_objcg_state+0xd2/0x360
  [337571.286577]  ? refill_obj_stock+0xb0/0x160
  [337571.286798]  ? seq_release+0x25/0x30
  [337571.287016]  ? __rseq_handle_notify_resume+0x3ba/0x4b0
  [337571.287235]  ? percpu_counter_add_batch+0x2e/0xa0
  [337571.287455]  ? __x64_sys_ioctl+0x88/0xc0
  [337571.287675]  __x64_sys_ioctl+0x88/0xc0
  [337571.287901]  do_syscall_64+0x38/0x90
  [337571.288126]  entry_SYSCALL_64_after_hwframe+0x72/0xdc
  [337571.288352] RIP: 0033:0x7f478aaffe9b

So fix this by locking struct btrfs_fs_info::trans_lock before deleting
the quota root from that list.

Fixes: bed92eae26cc ("Btrfs: qgroup implementation and prototypes")
CC: stable@vger.kernel.org # 4.14+
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: reinsert BGs failed to reclaim

commit 7e27180994383b7c741ad87749db01e4989a02ba upstream.

The reclaim process can temporarily fail. For example, if the space is
getting tight, it fails to make the block group read-only. If there are no
further writes on that block group, the block group will never get back to
the reclaim list, and the BG never gets reclaimed. In a certain workload,
we can leave many such block groups never reclaimed.

So, let's get it back to the list and give it a chance to be reclaimed.

Fixes: 18bb8bbf13c1 ("btrfs: zoned: automatically reclaim zones")
CC: stable@vger.kernel.org # 5.15+
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: add block-group tree to lockdep classes

commit 1a1b0e729d227f9f758f7b5f1c997e874e94156e upstream.

The block group tree was not present among the lockdep classes. We could
get potentially lockdep warnings but so far none has been seen, also
because block-group-tree is a relatively new feature.

CC: stable@vger.kernel.org # 6.1+
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: bail out reclaim process if filesystem is read-only

commit 93463ff7b54626f8276c0bd3d3f968fbf8d5d380 upstream.

When a filesystem is read-only, we cannot reclaim a block group as it
cannot rewrite the data. Just bail out in that case.

Note that it can drop block groups in this case. As we did
sb_start_write(), read-only filesystem means we got a fatal error and
forced read-only. There is no chance to reclaim them again.

Fixes: 18bb8bbf13c1 ("btrfs: zoned: automatically reclaim zones")
CC: stable@vger.kernel.org # 5.15+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: delete unused BGs while reclaiming BGs

commit 3ed01616bad6c7e3de196676b542ae3df8058592 upstream.

The reclaiming process only starts after the filesystem volumes are
allocated to a certain level (75% by default). Thus, the list of
reclaiming target block groups can build up so huge at the time the
reclaim process kicks in. On a test run, there were over 1000 BGs in the
reclaim list.

As the reclaim involves rewriting the data, it takes really long time to
reclaim the BGs. While the reclaim is running, btrfs_delete_unused_bgs()
won't proceed because the reclaim side is holding
fs_info->reclaim_bgs_lock. As a result, we will have a large number of
unused BGs kept in the unused list. On my test run, I got 1057 unused BGs.

Since deleting a block group is relatively easy and fast work, we can call
btrfs_delete_unused_bgs() while it reclaims BGs, to avoid building up
unused BGs.

Fixes: 18bb8bbf13c1 ("btrfs: zoned: automatically reclaim zones")
CC: stable@vger.kernel.org # 5.15+
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: add handling for RAID1C23/DUP to btrfs_reduce_alloc_profile

commit 160fe8f6fdb13da6111677be6263e5d65e875987 upstream.

Callers of `btrfs_reduce_alloc_profile` expect it to return exactly
one allocation profile flag, and failing to do so may ultimately
result in a WARN_ON and remount-ro when allocating new blocks, like
the below transaction abort on 6.1.

`btrfs_reduce_alloc_profile` has two ways of determining the profile,
first it checks if a conversion balance is currently running and
uses the profile we're converting to. If no balance is currently
running, it returns the max-redundancy profile which at least one
block in the selected block group has.

This works by simply checking each known allocation profile bit in
redundancy order. However, `btrfs_reduce_alloc_profile` has not been
updated as new flags have been added - first with the `DUP` profile
and later with the RAID1C34 profiles.

Because of the way it checks, if we have blocks with different
profiles and at least one is known, that profile will be selected.
However, if none are known we may return a flag set with multiple
allocation profiles set.

This is currently only possible when a balance from one of the three
unhandled profiles to another of the unhandled profiles is canceled
after allocating at least one block using the new profile.

In that case, a transaction abort like the below will occur and the
filesystem will need to be mounted with -o skip_balance to get it
mounted rw again (but the balance cannot be resumed without a
similar abort).

  [770.648] ------------[ cut here ]------------
  [770.648] BTRFS: Transaction aborted (error -22)
  [770.648] WARNING: CPU: 43 PID: 1159593 at fs/btrfs/extent-tree.c:4122 find_free_extent+0x1d94/0x1e00 [btrfs]
  [770.648] CPU: 43 PID: 1159593 Comm: btrfs Tainted: G        W 6.1.0-0.deb11.7-powerpc64le #1  Debian 6.1.20-2~bpo11+1a~test
  [770.648] Hardware name: T2P9D01 REV 1.00 POWER9 0x4e1202 opal:skiboot-bc106a0 PowerNV
  [770.648] NIP:  c00800000f6784fc LR: c00800000f6784f8 CTR: c000000000d746c0
  [770.648] REGS: c000200089afe9a0 TRAP: 0700   Tainted: G        W (6.1.0-0.deb11.7-powerpc64le Debian 6.1.20-2~bpo11+1a~test)
  [770.648] MSR:  9000000002029033 <SF,HV,VEC,EE,ME,IR,DR,RI,LE>  CR: 28848282  XER: 20040000
  [770.648] CFAR: c000000000135110 IRQMASK: 0
    GPR00: c00800000f6784f8 c000200089afec40 c00800000f7ea800 0000000000000026
    GPR04: 00000001004820c2 c000200089afea00 c000200089afe9f8 0000000000000027
    GPR08: c000200ffbfe7f98 c000000002127f90 ffffffffffffffd8 0000000026d6a6e8
    GPR12: 0000000028848282 c000200fff7f3800 5deadbeef0000122 c00000002269d000
    GPR16: c0002008c7797c40 c000200089afef17 0000000000000000 0000000000000000
    GPR20: 0000000000000000 0000000000000001 c000200008bc5a98 0000000000000001
    GPR24: 0000000000000000 c0000003c73088d0 c000200089afef17 c000000016d3a800
    GPR28: c0000003c7308800 c00000002269d000 ffffffffffffffea 0000000000000001
  [770.648] NIP [c00800000f6784fc] find_free_extent+0x1d94/0x1e00 [btrfs]
  [770.648] LR [c00800000f6784f8] find_free_extent+0x1d90/0x1e00 [btrfs]
  [770.648] Call Trace:
  [770.648] [c000200089afec40] [c00800000f6784f8] find_free_extent+0x1d90/0x1e00 [btrfs] (unreliable)
  [770.648] [c000200089afed30] [c00800000f681398] btrfs_reserve_extent+0x1a0/0x2f0 [btrfs]
  [770.648] [c000200089afeea0] [c00800000f681bf0] btrfs_alloc_tree_block+0x108/0x670 [btrfs]
  [770.648] [c000200089afeff0] [c00800000f66bd68] __btrfs_cow_block+0x170/0x850 [btrfs]
  [770.648] [c000200089aff100] [c00800000f66c58c] btrfs_cow_block+0x144/0x288 [btrfs]
  [770.648] [c000200089aff1b0] [c00800000f67113c] btrfs_search_slot+0x6b4/0xcb0 [btrfs]
  [770.648] [c000200089aff2a0] [c00800000f679f60] lookup_inline_extent_backref+0x128/0x7c0 [btrfs]
  [770.648] [c000200089aff3b0] [c00800000f67b338] lookup_extent_backref+0x70/0x190 [btrfs]
  [770.648] [c000200089aff470] [c00800000f67b54c] __btrfs_free_extent+0xf4/0x1490 [btrfs]
  [770.648] [c000200089aff5a0] [c00800000f67d770] __btrfs_run_delayed_refs+0x328/0x1530 [btrfs]
  [770.648] [c000200089aff740] [c00800000f67ea2c] btrfs_run_delayed_refs+0xb4/0x3e0 [btrfs]
  [770.648] [c000200089aff800] [c00800000f699aa4] btrfs_commit_transaction+0x8c/0x12b0 [btrfs]
  [770.648] [c000200089aff8f0] [c00800000f6dc628] reset_balance_state+0x1c0/0x290 [btrfs]
  [770.648] [c000200089aff9a0] [c00800000f6e2f7c] btrfs_balance+0x1164/0x1500 [btrfs]
  [770.648] [c000200089affb40] [c00800000f6f8e4c] btrfs_ioctl+0x2b54/0x3100 [btrfs]
  [770.648] [c000200089affc80] [c00000000053be14] sys_ioctl+0x794/0x1310
  [770.648] [c000200089affd70] [c00000000002af98] system_call_exception+0x138/0x250
  [770.648] [c000200089affe10] [c00000000000c654] system_call_common+0xf4/0x258
  [770.648] --- interrupt: c00 at 0x7fff94126800
  [770.648] NIP:  00007fff94126800 LR: 0000000107e0b594 CTR: 0000000000000000
  [770.648] REGS: c000200089affe80 TRAP: 0c00   Tainted: G        W (6.1.0-0.deb11.7-powerpc64le Debian 6.1.20-2~bpo11+1a~test)
  [770.648] MSR:  900000000000d033 <SF,HV,EE,PR,ME,IR,DR,RI,LE>  CR: 24002848  XER: 00000000
  [770.648] IRQMASK: 0
    GPR00: 0000000000000036 00007fffc9439da0 00007fff94217100 0000000000000003
    GPR04: 00000000c4009420 00007fffc9439ee8 0000000000000000 0000000000000000
    GPR08: 00000000803c7416 0000000000000000 0000000000000000 0000000000000000
    GPR12: 0000000000000000 00007fff9467d120 0000000107e64c9c 0000000107e64d0a
    GPR16: 0000000107e64d06 0000000107e64cf1 0000000107e64cc4 0000000107e64c73
    GPR20: 0000000107e64c31 0000000107e64bf1 0000000107e64be7 0000000000000000
    GPR24: 0000000000000000 00007fffc9439ee0 0000000000000003 0000000000000001
    GPR28: 00007fffc943f713 0000000000000000 00007fffc9439ee8 0000000000000000
  [770.648] NIP [00007fff94126800] 0x7fff94126800
  [770.648] LR [0000000107e0b594] 0x107e0b594
  [770.648] --- interrupt: c00
  [770.648] Instruction dump:
  [770.648] 3b00ffe4 e8898828 481175f5 60000000 4bfff4fc 3be00000 4bfff570 3d220000
  [770.648] 7fc4f378 e8698830 4811cd95 e8410018 <0fe00000> f9c10060 f9e10068 fa010070
  [770.648] ---[ end trace 0000000000000000 ]---
  [770.648] BTRFS: error (device dm-2: state A) in find_free_extent_update_loop:4122: errno=-22 unknown
  [770.648] BTRFS info (device dm-2: state EA): forced readonly
  [770.648] BTRFS: error (device dm-2: state EA) in __btrfs_free_extent:3070: errno=-22 unknown
  [770.648] BTRFS error (device dm-2: state EA): failed to run delayed ref for logical 17838685708288 num_bytes 24576 type 184 action 2 ref_mod 1: -22
  [770.648] BTRFS: error (device dm-2: state EA) in btrfs_run_delayed_refs:2144: errno=-22 unknown
  [770.648] BTRFS: error (device dm-2: state EA) in reset_balance_state:3599: errno=-22 unknown

Fixes: 47e6f7423b91 ("btrfs: add support for 3-copy replication (raid1c3)")
Fixes: 8d6fac0087e5 ("btrfs: add support for 4-copy replication (raid1c4)")
CC: stable@vger.kernel.org # 5.10+
Signed-off-by: Matt Corallo <blnxfsl@bluematt.me>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ipvs: increase ip_vs_conn_tab_bits range for 64BIT

commit 04292c695f82b6cf0d25dd5ae494f16ddbb621f6 upstream.

Current range [8, 20] is set purely due to historical reasons
because at the time, ~1M (2^20) was considered sufficient.
With this change, 27 is the upper limit for 64-bit, 20 otherwise.

Previous change regarding this limit is here.

Link: https://lore.kernel.org/all/86eabeb9dd62aebf1e2533926fdd13fed48bab1f.1631289960.git.aclaudi@redhat.com/T/#u
Signed-off-by: Abhijeet Rastogi <abhijeet.1989@gmail.com>
Acked-by: Julian Anastasov <ja@ssi.bg>
Acked-by: Simon Horman <horms@kernel.org>
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
Cc: Allen Pais <apais@linux.microsoft.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

usb: typec: ucsi: Mark dGPUs as DEVICE scope

commit a7fbfd44c0204f0629288edfd0d77829edb4a2f8 upstream.

power_supply_is_system_supplied() checks whether any power
supplies are present that aren't batteries to decide whether
the system is running on DC or AC.  Downstream drivers use
this to make performance decisions.

Navi dGPUs include an UCSI function that has been exported
since commit 17631e8ca2d3 ("i2c: designware: Add driver
support for AMD NAVI GPU").

This UCSI function registers a power supply since commit
992a60ed0d5e ("usb: typec: ucsi: register with power_supply class")
but this is not a system power supply.

As the power supply for a dGPU is only for powering devices connected
to dGPU, create a device property to indicate that the UCSI endpoint
is only for the scope of `POWER_SUPPLY_SCOPE_DEVICE`.

Link: https://lore.kernel.org/lkml/20230516182541.5836-2-mario.limonciello@amd.com/
Reviewed-by: Evan Quan <evan.quan@amd.com>
Tested-by: Evan Quan <evan.quan@amd.com>
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Reviewed-by: Sebastian Reichel <sebastian.reichel@collabora.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i2c: nvidia-gpu: Remove ccgx,firmware-build property

commit 430b38764fbb931c6dbd1af13c8b2e4508994662 upstream.

Now the Cypress CCG driver has been updated to support the
'firmware-name' property to align with device-tree, remove the
'ccgx,firmware-build' property as this is no longer needed.

Signed-off-by: Wayne Chang <waynec@nvidia.com>
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Acked-by: Ajay Gupta <ajayg@nvidia.com>
Acked-by: Wolfram Sang <wsa@kernel.org>
Link: https://lore.kernel.org/r/20230131175748.256423-5-jonathanh@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i2c: nvidia-gpu: Add ACPI property to align with device-tree

commit f510b0a3565b9231e828e23a7e0f9790b97edf96 upstream.

Device-tree uses the 'firmware-name' string property to pass a name of
the firmware build to the Cypress CCGx driver. Add a new ACPI string
property to the NVIDIA GPU I2C driver to align with device-tree so that
we can migrate to using a common property name for both ACPI and
device-tree.

Signed-off-by: Wayne Chang <waynec@nvidia.com>
Co-developed-by: Jon Hunter <jonathanh@nvidia.com>
Signed-off-by: Jon Hunter <jonathanh@nvidia.com>
Reviewed-by: Heikki Krogerus <heikki.krogerus@linux.intel.com>
Acked-by: Ajay Gupta <ajayg@nvidia.com>
Acked-by: Wolfram Sang <wsa@kernel.org>
Link: https://lore.kernel.org/r/20230131175748.256423-3-jonathanh@nvidia.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs: Lock moved directories

commit 28eceeda130f5058074dd007d9c59d2e8bc5af2e upstream.

When a directory is moved to a different directory, some filesystems
(udf, ext4, ocfs2, f2fs, and likely gfs2, reiserfs, and others) need to
update their pointer to the parent and this must not race with other
operations on the directory. Lock the directories when they are moved.
Although not all filesystems need this locking, we perform it in
vfs_rename() because getting the lock ordering right is really difficult
and we don't want to expose these locking details to filesystems.

CC: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20230601105830.13168-5-jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs: Establish locking order for unrelated directories

commit f23ce757185319886ca80c4864ce5f81ac6cc9e9 upstream.

Currently the locking order of inode locks for directories that are not
in ancestor relationship is not defined because all operations that
needed to lock two directories like this were serialized by
sb->s_vfs_rename_mutex. However some filesystems need to lock two
subdirectories for RENAME_EXCHANGE operations and for this we need the
locking order established even for two tree-unrelated directories.
Provide a helper function lock_two_inodes() that establishes lock
ordering for any two inodes and use it in lock_two_directories().

CC: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20230601105830.13168-4-jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

Revert "f2fs: fix potential corruption when moving a directory"

commit cde3c9d7e2a359e337216855dcb333a19daaa436 upstream.

This reverts commit d94772154e524b329a168678836745d2773a6e02. The
locking is going to be provided by VFS.

CC: Jaegeuk Kim <jaegeuk@kernel.org>
CC: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20230601105830.13168-3-jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

ext4: Remove ext4 locking of moved directory

commit 3658840cd363f2be094f5dfd2f0b174a9055dd0f upstream.

Remove locking of moved directory in ext4_rename2(). We will take care
of it in VFS instead. This effectively reverts commit 0813299c586b
("ext4: Fix possible corruption when moving a directory") and followup
fixes.

CC: Ted Tso <tytso@mit.edu>
CC: stable@vger.kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20230601105830.13168-1-jack@suse.cz>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

fs: avoid empty option when generating legacy mount string

commit 62176420274db5b5127cd7a0083a9aeb461756ee upstream.

As each option string fragment is always prepended with a comma it would
happen that the whole string always starts with a comma. This could be
interpreted by filesystem drivers as an empty option and may produce
errors.

For example the NTFS driver from ntfs.ko behaves like this and fails
when mounted via the new API.

Link: https://github.com/util-linux/util-linux/issues/2298
Signed-off-by: Thomas Weißschuh <linux@weissschuh.net>
Fixes: 3e1aeb00e6d1 ("vfs: Implement a filesystem superblock creation/configuration context")
Cc: stable@vger.kernel.org
Message-Id: <20230607-fs-empty-option-v1-1-20c8dbf4671b@weissschuh.net>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

jffs2: reduce stack usage in jffs2_build_xattr_subsystem()

commit 1168f095417643f663caa341211e117db552989f upstream.

Use kcalloc() for allocation/flush of 128 pointers table to
reduce stack usage.

Function now returns -ENOMEM or 0 on success.

stackusage
Before:
./fs/jffs2/xattr.c:775  jffs2_build_xattr_subsystem     1208
dynamic,bounded

After:
./fs/jffs2/xattr.c:775  jffs2_build_xattr_subsystem     192
dynamic,bounded

Also update definition when CONFIG_JFFS2_FS_XATTR is not enabled

Tested with an MTD mount point and some user set/getfattr.

Many current target on OpenWRT also suffer from a compilation warning
(that become an error with CONFIG_WERROR) with the following output:

fs/jffs2/xattr.c: In function 'jffs2_build_xattr_subsystem':
fs/jffs2/xattr.c:887:1: error: the frame size of 1088 bytes is larger than 1024 bytes [-Werror=frame-larger-than=]
  887 | }
      | ^

Using dynamic allocation fix this compilation warning.

Fixes: c9f700f840bd ("[JFFS2][XATTR] using 'delete marker' for xdatum/xref deletion")
Reported-by: Tim Gardner <tim.gardner@canonical.com>
Reported-by: kernel test robot <lkp@intel.com>
Reported-by: Ron Economos <re@w6rz.net>
Reported-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Fabian Frederick <fabf@skynet.be>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Cc: stable@vger.kernel.org
Message-Id: <20230506045612.16616-1-ansuelsmth@gmail.com>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

shmem: use ramfs_kill_sb() for kill_sb method of ramfs-based tmpfs

commit 36ce9d76b0a93bae799e27e4f5ac35478c676592 upstream.

As the ramfs-based tmpfs uses ramfs_init_fs_context() for the
init_fs_context method, which allocates fc->s_fs_info, use ramfs_kill_sb()
to free it and avoid a memory leak.

Link: https://lkml.kernel.org/r/20230607161523.2876433-1-roberto.sassu@huaweicloud.com
Fixes: c3b1b1cbf002 ("ramfs: add support for "mode=" mount option")
Signed-off-by: Roberto Sassu <roberto.sassu@huawei.com>
Cc: Hugh Dickins <hughd@google.com>
Cc: David Howells <dhowells@redhat.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mm/damon/ops-common: atomically test and clear young on ptes and pmds

commit c11d34fa139e4b0fb4249a30f37b178353533fa1 upstream.

It is racy to non-atomically read a pte, then clear the young bit, then
write it back as this could discard dirty information.  Further, it is bad
practice to directly set a pte entry within a table.  Instead clearing
young must go through the arch-provided helper,
ptep_test_and_clear_young() to ensure it is modified atomically and to
give the arch code visibility and allow it to check (and potentially
modify) the operation.

Link: https://lkml.kernel.org/r/20230602092949.545577-3-ryan.roberts@arm.com
Fixes: 3f49584b262c ("mm/damon: implement primitives for the virtual memory address spaces").
Signed-off-by: Ryan Roberts <ryan.roberts@arm.com>
Reviewed-by: Zi Yan <ziy@nvidia.com>
Reviewed-by: SeongJae Park <sj@kernel.org>
Reviewed-by: Mike Rapoport (IBM) <rppt@kernel.org>
Cc: Christoph Hellwig <hch@lst.de>
Cc: Kirill A. Shutemov <kirill.shutemov@linux.intel.com>
Cc: Lorenzo Stoakes <lstoakes@gmail.com>
Cc: Matthew Wilcox (Oracle) <willy@infradead.org>
Cc: Uladzislau Rezki (Sony) <urezki@gmail.com>
Cc: Yu Zhao <yuzhao@google.com>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

autofs: use flexible array in ioctl structure

commit e910c8e3aa02dc456e2f4c32cb479523c326b534 upstream.

Commit df8fc4e934c1 ("kbuild: Enable -fstrict-flex-arrays=3") introduced a warning
for the autofs_dev_ioctl structure:

In function 'check_name',
    inlined from 'validate_dev_ioctl' at fs/autofs/dev-ioctl.c:131:9,
    inlined from '_autofs_dev_ioctl' at fs/autofs/dev-ioctl.c:624:8:
fs/autofs/dev-ioctl.c:33:14: error: 'strchr' reading 1 or more bytes from a region of size 0 [-Werror=stringop-overread]
   33 |         if (!strchr(name, '/'))
      |              ^~~~~~~~~~~~~~~~~
In file included from include/linux/auto_dev-ioctl.h:10,
                 from fs/autofs/autofs_i.h:10,
                 from fs/autofs/dev-ioctl.c:14:
include/uapi/linux/auto_dev-ioctl.h: In function '_autofs_dev_ioctl':
include/uapi/linux/auto_dev-ioctl.h:112:14: note: source object 'path' of size 0
  112 |         char path[0];
      |              ^~~~

This is easily fixed by changing the gnu 0-length array into a c99
flexible array. Since this is a uapi structure, we have to be careful
about possible regressions but this one should be fine as they are
equivalent here. While it would break building with ancient gcc versions
that predate c99, it helps building with --std=c99 and -Wpedantic builds
in user space, as well as non-gnu compilers. This means we probably
also want it fixed in stable kernels.

Cc: stable@vger.kernel.org
Cc: Kees Cook <keescook@chromium.org>
Cc: "Gustavo A. R. Silva" <gustavoars@kernel.org>
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20230523081944.581710-1-arnd@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

integrity: Fix possible multiple allocation in integrity_inode_get()

commit 9df6a4870dc371136e90330cfbbc51464ee66993 upstream.

When integrity_inode_get() is querying and inserting the cache, there
is a conditional race in the concurrent environment.

The race condition is the result of not properly implementing
"double-checked locking". In this case, it first checks to see if the
iint cache record exists before taking the lock, but doesn't check
again after taking the integrity_iint_lock.

Fixes: bf2276d10ce5 ("ima: allocating iint improvements")
Signed-off-by: Tianjia Zhang <tianjia.zhang@linux.alibaba.com>
Cc: Dmitry Kasatkin <dmitry.kasatkin@gmail.com>
Cc: <stable@vger.kernel.org> # v3.10+
Signed-off-by: Mimi Zohar <zohar@linux.ibm.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

um: Use HOST_DIR for mrproper

commit a5a319ec2c2236bb96d147c16196d2f1f3799301 upstream.

When HEADER_ARCH was introduced, the MRPROPER_FILES (then MRPROPER_DIRS)
list wasn't adjusted, leaving SUBARCH as part of the path argument.
This resulted in the "mrproper" target not cleaning up arch/x86/... when
SUBARCH was specified. Since HOST_DIR is arch/$(HEADER_ARCH), use it
instead to get the correct path.

Cc: Richard Weinberger <richard@nod.at>
Cc: Anton Ivanov <anton.ivanov@cambridgegreys.com>
Cc: Johannes Berg <johannes@sipsolutions.net>
Cc: Azeem Shaikh <azeemshaikh38@gmail.com>
Cc: linux-um@lists.infradead.org
Fixes: 7bbe7204e937 ("um: merge Makefile-{i386,x86_64}")
Cc: stable@vger.kernel.org
Signed-off-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/20230606222442.never.807-kees@kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

watch_queue: prevent dangling pipe pointer

commit 943211c87427f25bd22e0e63849fb486bb5f87fa upstream.

NULL the dangling pipe reference while clearing watch_queue.

If not done, a reference to a freed pipe remains in the watch_queue,
as this function is called before freeing a pipe in free_pipe_info()
(see line 834 of fs/pipe.c).

The sole use of wqueue->defunct is for checking if the watch queue has
been cleared, but wqueue->pipe is also NULLed while clearing.

Thus, wqueue->defunct is superfluous, as wqueue->pipe can be checked
for NULL. Hence, the former can be removed.

Tested with keyutils testsuite.

Cc: stable@vger.kernel.org # 6.1
Signed-off-by: Siddh Raman Pant <code@siddh.me>
Acked-by: David Howells <dhowells@redhat.com>
Message-Id: <20230605143616.640517-1-code@siddh.me>
Signed-off-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bcache: Fix __bch_btree_node_alloc to make the failure behavior consistent

commit 80fca8a10b604afad6c14213fdfd816c4eda3ee4 upstream.

In some specific situations, the return value of __bch_btree_node_alloc
may be NULL. This may lead to a potential NULL pointer dereference in
caller function like a calling chain :
btree_split->bch_btree_node_alloc->__bch_btree_node_alloc.

Fix it by initializing the return value in __bch_btree_node_alloc.

Fixes: cafe56359144 ("bcache: A block layer cache")
Cc: stable@vger.kernel.org
Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
Signed-off-by: Coly Li <colyli@suse.de>
Link: https://lore.kernel.org/r/20230615121223.22502-6-colyli@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bcache: Remove unnecessary NULL point check in node allocations

commit 028ddcac477b691dd9205c92f991cc15259d033e upstream.

Due to the previous fix of __bch_btree_node_alloc, the return value will
never be a NULL pointer. So IS_ERR is enough to handle the failure
situation. Fix it by replacing IS_ERR_OR_NULL check by an IS_ERR check.

Fixes: cafe56359144 ("bcache: A block layer cache")
Cc: stable@vger.kernel.org
Signed-off-by: Zheng Wang <zyytlz.wz@163.com>
Signed-off-by: Coly Li <colyli@suse.de>
Link: https://lore.kernel.org/r/20230615121223.22502-5-colyli@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

bcache: fixup btree_cache_wait list damage

commit f0854489fc07d2456f7cc71a63f4faf9c716ffbe upstream.

We get a kernel crash about "list_add corruption. next->prev should be
prev (ffff9c801bc01210), but was ffff9c77b688237c.
(next=ffffae586d8afe68)."

crash> struct list_head 0xffff9c801bc01210
struct list_head {
  next = 0xffffae586d8afe68,
  prev = 0xffffae586d8afe68
}
crash> struct list_head 0xffff9c77b688237c
struct list_head {
  next = 0x0,
  prev = 0x0
}
crash> struct list_head 0xffffae586d8afe68
struct list_head struct: invalid kernel virtual address: ffffae586d8afe68  type: "gdb_readmem_callback"
Cannot access memory at address 0xffffae586d8afe68

[230469.019492] Call Trace:
[230469.032041]  prepare_to_wait+0x8a/0xb0
[230469.044363]  ? bch_btree_keys_free+0x6c/0xc0 [escache]
[230469.056533]  mca_cannibalize_lock+0x72/0x90 [escache]
[230469.068788]  mca_alloc+0x2ae/0x450 [escache]
[230469.080790]  bch_btree_node_get+0x136/0x2d0 [escache]
[230469.092681]  bch_btree_check_thread+0x1e1/0x260 [escache]
[230469.104382]  ? finish_wait+0x80/0x80
[230469.115884]  ? bch_btree_check_recurse+0x1a0/0x1a0 [escache]
[230469.127259]  kthread+0x112/0x130
[230469.138448]  ? kthread_flush_work_fn+0x10/0x10
[230469.149477]  ret_from_fork+0x35/0x40

bch_btree_check_thread() and bch_dirty_init_thread() may call
mca_cannibalize() to cannibalize other cached btree nodes. Only one thread
can do it at a time, so the op of other threads will be added to the
btree_cache_wait list.

We must call finish_wait() to remove op from btree_cache_wait before free
it's memory address. Otherwise, the list will be damaged. Also should call
bch_cannibalize_unlock() to release the btree_cache_alloc_lock and wake_up
other waiters.

Fixes: 8e7102273f59 ("bcache: make bch_btree_check() to be multithreaded")
Fixes: b144e45fc576 ("bcache: make bch_sectors_dirty_init() to be multithreaded")
Cc: stable@vger.kernel.org
Signed-off-by: Mingzhe Zou <mingzhe.zou@easystack.cn>
Signed-off-by: Coly Li <colyli@suse.de>
Link: https://lore.kernel.org/r/20230615121223.22502-7-colyli@suse.de
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: mt76: mt7921e: fix init command fail with enabled device

commit 525c469e5de9bf7e53574396196e80fc716ac9eb upstream.

For some cases as below, we may encounter the unpreditable chip stats
in driver probe()
* The system reboot flow do not work properly, such as kernel oops while
  rebooting, and then the driver do not go back to default status at
  this moment.
* Similar to the flow above. If the device was enabled in BIOS or UEFI,
  the system may switch to Linux without driver fully shutdown.

To avoid the problem, force push the device back to default in probe()
* mt7921e_mcu_fw_pmctrl() : return control privilege to chip side.
* mt7921_wfsys_reset()    : cleanup chip config before resource init.

Error log
[59007.600714] mt7921e 0000:02:00.0: ASIC revision: 79220010
[59010.889773] mt7921e 0000:02:00.0: Message 00000010 (seq 1) timeout
[59010.889786] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59014.217839] mt7921e 0000:02:00.0: Message 00000010 (seq 2) timeout
[59014.217852] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59017.545880] mt7921e 0000:02:00.0: Message 00000010 (seq 3) timeout
[59017.545893] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59020.874086] mt7921e 0000:02:00.0: Message 00000010 (seq 4) timeout
[59020.874099] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59024.202019] mt7921e 0000:02:00.0: Message 00000010 (seq 5) timeout
[59024.202033] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59027.530082] mt7921e 0000:02:00.0: Message 00000010 (seq 6) timeout
[59027.530096] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59030.857888] mt7921e 0000:02:00.0: Message 00000010 (seq 7) timeout
[59030.857904] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59034.185946] mt7921e 0000:02:00.0: Message 00000010 (seq 8) timeout
[59034.185961] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59037.514249] mt7921e 0000:02:00.0: Message 00000010 (seq 9) timeout
[59037.514262] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59040.842362] mt7921e 0000:02:00.0: Message 00000010 (seq 10) timeout
[59040.842375] mt7921e 0000:02:00.0: Failed to get patch semaphore
[59040.923845] mt7921e 0000:02:00.0: hardware init failed

Cc: stable@vger.kernel.org
Fixes: 5c14a5f944b9 ("mt76: mt7921: introduce mt7921e support")
Tested-by: Kai-Heng Feng <kai.heng.feng@canonical.com>
Tested-by: Juan Martinez <juan.martinez@amd.com>
Co-developed-by: Leon Yen <leon.yen@mediatek.com>
Signed-off-by: Leon Yen <leon.yen@mediatek.com>
Signed-off-by: Quan Zhou <quan.zhou@mediatek.com>
Signed-off-by: Deren Wu <deren.wu@mediatek.com>
Message-ID: <39fcb7cee08d4ab940d38d82f21897483212483f.1688569385.git.deren.wu@mediatek.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: ath10k: Serialize wake_tx_queue ops

commit b719ebc37a1eacd4fd4f1264f731b016e5ec0c6e upstream.

Serialize the ath10k implementation of the wake_tx_queue ops.
ath10k_mac_op_wake_tx_queue() must not run concurrent since it's using
ieee80211_txq_schedule_start().

The intend of this patch is to sort out an issue discovered in the discussion
referred to by the Link tag.

I can't test it with real hardware and thus just implemented the per-ac queue
lock Felix suggested. One obvious alternative to the per-ac lock would be to
bring back the txqs_lock commit bb2edb733586 ("ath10k: migrate to mac80211 txq
scheduling") dropped.

Fixes: bb2edb733586 ("ath10k: migrate to mac80211 txq scheduling")
Reported-by: Felix Fietkau <nbd@nbd.name>
Link: https://lore.kernel.org/r/519b5bb9-8899-ae7c-4eff-f3116cdfdb56@nbd.name
CC: <stable@vger.kernel.org>
Signed-off-by: Alexander Wetzel <alexander@wetzel-home.de>
Signed-off-by: Kalle Valo <quic_kvalo@quicinc.com>
Link: https://lore.kernel.org/r/20230323165527.156414-1-alexander@wetzel-home.de
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

wifi: cfg80211: fix regulatory disconnect for non-MLO

commit b22552fcaf1970360005c805d7fba4046cf2ab4a upstream.

The multi-link loop here broke disconnect when multi-link
operation (MLO) isn't active for a given interface, since
in that case valid_links is 0 (indicating no links, i.e.
no MLO.)

Fix this by taking that into account properly and skipping
the link only if there are valid_links in the first place.

Cc: stable@vger.kernel.org
Fixes: 7b0a0e3c3a88 ("wifi: cfg80211: do some rework towards MLO link APIs")
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Link: https://lore.kernel.org/r/20230616222844.eb073d650c75.I72739923ef80919889ea9b50de9e4ba4baa836ae@changeid
Signed-off-by: Johannes Berg <johannes.berg@intel.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: sdhci: fix DMA configure compatibility issue when 64bit DMA mode is used.

commit 20dbd07ef0a8bc29eb03d6a95258ac8934cbe52d upstream.

Bayhub SD host has hardware limitation:
1.The upper 32bit address is inhibited to be written at SD Host Register
  [03E][13]=0 (32bits addressing) mode, is admitted to be written only at
  SD Host Register [03E][13]=1 (64bits addressing) mode.
2.Because of above item#1, need to configure SD Host Register [03E][13] to
  1(64bits addressing mode) before set 64bit ADMA system address's higher
  32bits SD Host Register [05F~05C] if 64 bits addressing mode is used.

The hardware limitation is reasonable for below reasons:
1.Normal flow should set DMA working mode first, then do
  DMA-transfer-related configuration, such as system address.
2.The hardware limitation may avoid the software to configure wrong higher
  32bit address at 32bits addressing mode although it is redundant.

The change that set 32bits/64bits addressing mode before set ADMA address,
  has no side-effect to other host IPs for below reason:
The setting order is reasonable and standard: DMA Mode setting first and
  then DMA address setting. It meets all DMA setting sequence.

Signed-off-by: Chevron Li <chevron.li@bayhubtech.com>
Acked-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230523111114.18124-1-chevron_li@126.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: mmci: Set PROBE_PREFER_ASYNCHRONOUS

commit 3108eb2e8aa7e955a9dd3a4c1bf19a7898961822 upstream.

All mmc host drivers should have the asynchronous probe option enabled, but
it seems like we failed to set it for mmci, so let's do that now.

Fixes: 21b2cec61c04 ("mmc: Set PROBE_PREFER_ASYNCHRONOUS for drivers that existed in v4.4")
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Tested-by: Linus Walleij <linus.walleij@linaro.org>
Tested-by: Yann Gautier <yann.gautier@foss.st.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230612143730.210390-1-ulf.hansson@linaro.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: core: disable TRIM on Micron MTFC4GACAJCN-1M

commit dbfbddcddcebc9ce8a08757708d4e4a99d238e44 upstream.

It seems that Micron MTFC4GACAJCN-1M despite advertising TRIM support does
not work when the core is trying to use REQ_OP_WRITE_ZEROES.

We are seeing the following errors in OpenWrt under 6.1 on Qnap Qhora 301W
that we did not previously have and tracked it down to REQ_OP_WRITE_ZEROES:
[   18.085950] I/O error, dev loop0, sector 596 op 0x9:(WRITE_ZEROES) flags 0x800 phys_seg 0 prio class 2

Disabling TRIM makes the error go away, so lets add a quirk for this eMMC
to disable TRIM.

Signed-off-by: Robert Marko <robimarko@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230530213259.1776512-1-robimarko@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

mmc: core: disable TRIM on Kingston EMMC04G-M627

commit f1738a1f816233e6dfc2407f24a31d596643fd90 upstream.

It seems that Kingston EMMC04G-M627 despite advertising TRIM support does
not work when the core is trying to use REQ_OP_WRITE_ZEROES.

We are seeing I/O errors in OpenWrt under 6.1 on Zyxel NBG7815 that we did
not previously have and tracked it down to REQ_OP_WRITE_ZEROES.

Trying to use fstrim seems to also throw errors like:
[93010.835112] I/O error, dev loop0, sector 16902 op 0x3:(DISCARD) flags 0x800 phys_seg 1 prio class 2

Disabling TRIM makes the error go away, so lets add a quirk for this eMMC
to disable TRIM.

Signed-off-by: Robert Marko <robimarko@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20230619193621.437358-1-robimarko@gmail.com
Signed-off-by: Ulf Hansson <ulf.hansson@linaro.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

io_uring: wait interruptibly for request completions on exit

commit 4826c59453b3b4677d6bf72814e7ababdea86949 upstream.

WHen the ring exits, cleanup is done and the final cancelation and
waiting on completions is done by io_ring_exit_work. That function is
invoked by kworker, which doesn't take any signals. Because of that, it
doesn't really matter if we wait for completions in TASK_INTERRUPTIBLE
or TASK_UNINTERRUPTIBLE state. However, it does matter to the hung task
detection checker!

Normally we expect cancelations and completions to happen rather
quickly. Some test cases, however, will exit the ring and park the
owning task stopped (eg via SIGSTOP). If the owning task needs to run
task_work to complete requests, then io_ring_exit_work won't make any
progress until the task is runnable again. Hence io_ring_exit_work can
trigger the hung task detection, which is particularly problematic if
panic-on-hung-task is enabled.

As the ring exit doesn't take signals to begin with, have it wait
interruptibly rather than uninterruptibly. io_uring has a separate
stuck-exit warning that triggers independently anyway, so we're not
really missing anything by making this switch.

Cc: stable@vger.kernel.org # 5.10+
Link: https://lore.kernel.org/r/b0e4aaef-7088-56ce-244c-976edeac0e66@kernel.dk
Signed-off-by: Jens Axboe <axboe@kernel.dk>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

irqchip/loongson-pch-pic: Fix initialization of HT vector register

commit f679616565f1cf1a4acb245dbc0032dafcd40637 upstream.

In an ACPI-based dual-bridge system, IRQ of each bridge's
PCH PIC sent to CPU is always a zero-based number, which
means that the IRQ on PCH PIC of each bridge is mapped into
vector range from 0 to 63 of upstream irqchip(e.g. EIOINTC).

      EIOINTC N: [0 ... 63 | 64 ... 255]
                  --------   ----------
                      ^          ^
                      |          |
                  PCH PIC N      |
                             PCH MSI N

For example, the IRQ vector number of sata controller on
PCH PIC of each bridge is 16, which is sent to upstream
irqchip of EIOINTC when an interrupt occurs, which will set
bit 16 of EIOINTC. Since hwirq of 16 on EIOINTC has been
mapped to a irq_desc for sata controller during hierarchy
irq allocation, the related mapped IRQ will be found through
irq_resolve_mapping() in the IRQ domain of EIOINTC.

So, the IRQ number set in HT vector register should be fixed
to be a zero-based number.

Cc: stable@vger.kernel.org
Reviewed-by: Huacai Chen <chenhuacai@loongson.cn>
Co-developed-by: liuyun <liuyun@loongson.cn>
Signed-off-by: liuyun <liuyun@loongson.cn>
Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230614115936.5950-2-lvjianmin@loongson.cn
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

NFSD: add encoding of op_recall flag for write delegation

commit 58f5d894006d82ed7335e1c37182fbc5f08c2f51 upstream.

Modified nfsd4_encode_open to encode the op_recall flag properly
for OPEN result with write delegation granted.

Signed-off-by: Dai Ngo <dai.ngo@oracle.com>
Reviewed-by: Jeff Layton <jlayton@kernel.org>
Signed-off-by: Chuck Lever <chuck.lever@oracle.com>
Cc: stable@vger.kernel.org
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

irqchip/loongson-pch-pic: Fix potential incorrect hwirq assignment

commit 783422e704ca0fa41cb2fe9ed79e46b6fe7eae29 upstream.

In DeviceTree path, when ht_vec_base is not zero, the hwirq of PCH PIC
will be assigned incorrectly. Because when pch_pic_domain_translate()
adds the ht_vec_base to hwirq, the hwirq does not have the ht_vec_base
subtracted when calling irq_domain_set_info().

The ht_vec_base is designed for the parent irq chip/domain of the PCH PIC.
It seems not proper to deal this in callbacks of the PCH PIC domain and
let's put this back like the initial commit ef8c01eb64ca ("irqchip: Add
Loongson PCH PIC controller").

Fixes: bcdd75c596c8 ("irqchip/loongson-pch-pic: Add ACPI init support")
Cc: stable@vger.kernel.org
Reviewed-by: Huacai Chen <chenhuacai@loongson.cn>
Signed-off-by: Liu Peibao <liupeibao@loongson.cn>
Signed-off-by: Jianmin Lv <lvjianmin@loongson.cn>
Signed-off-by: Marc Zyngier <maz@kernel.org>
Link: https://lore.kernel.org/r/20230614115936.5950-3-lvjianmin@loongson.cn
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

i2c: qup: Add missing unwind goto in qup_i2c_probe()

commit cd9489623c29aa2f8cc07088168afb6e0d5ef06d upstream.

Smatch Warns:
drivers/i2c/busses/i2c-qup.c:1784 qup_i2c_probe()
warn: missing unwind goto?

The goto label "fail_runtime" and "fail" will disable qup->pclk,
but here qup->pclk failed to obtain, in order to be consistent,
change the direct return to goto label "fail_dma".

Fixes: 9cedf3b2f099 ("i2c: qup: Add bam dma capabilities")
Signed-off-by: Shuai Jiang <d202180596@hust.edu.cn>
Reviewed-by: Dongliang Mu <dzm91@hust.edu.cn>
Reviewed-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Cc: <stable@vger.kernel.org> # v4.6+
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

btrfs: do not BUG_ON() on tree mod log failure at balance_level()

[ Upstream commit 39020d8abc7ec62c4de9b260e3d10d4a1c2478ce ]

At balance_level(), instead of doing a BUG_ON() in case we fail to record
tree mod log operations, do a transaction abort and return the error to
the callers. There's really no need for the BUG_ON() as we can release
all resources in this context, and we have to abort because other future
tree searches that use the tree mod log (btrfs_search_old_slot()) may get
inconsistent results if other operations modify the tree after that
failure and before the tree mod log based search.

CC: stable@vger.kernel.org # 5.4+
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

extcon: usbc-tusb320: Unregister typec port on driver removal

[ Upstream commit 3adbaa30d973093a4f37927baf9596cca51b593d ]

The driver can register a typec port if suitable firmware properties are
present. But if the driver is removed through sysfs unbind, rmmod or
similar, then it does not clean up after itself and the typec port
device remains registered. This can be seen in sysfs, where stale typec
ports get left over in /sys/class/typec.

In order to fix this we have to add an i2c_driver remove function and
call typec_unregister_port(), which is a no-op in the case where no
typec port is created and the pointer remains NULL.

In the process we should also put the fwnode_handle when the typec port
isn't registered anymore, including if an error occurs during probe. The
typec subsystem does not increase or decrease the reference counter for
us, so we track it in the driver's private data.

Note that the conditional check on TYPEC_PWR_MODE_PD was removed in the
probe path because a call to tusb320_set_adv_pwr_mode() will perform an
even more robust validation immediately after, hence there is no
functional change here.

Fixes: bf7571c00dca ("extcon: usbc-tusb320: Add USB TYPE-C support")
Cc: stable@vger.kernel.org
Signed-off-by: Alvin Šipraga <alsi@bang-olufsen.dk>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

extcon: usbc-tusb320: Convert to i2c's .probe_new()

[ Upstream commit 5313121b22fd11db0d14f305c110168b8176efdc ]

The probe function doesn't make use of the i2c_device_id * parameter so it
can be trivially converted.

Signed-off-by: Uwe Kleine-König <u.kleine-koenig@pengutronix.de>
Signed-off-by: Chanwoo Choi <cw00.choi@samsung.com>
Stable-dep-of: 3adbaa30d973 ("extcon: usbc-tusb320: Unregister typec port on driver removal")
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm ioctl: Avoid double-fetch of version

[ Upstream commit 249bed821b4db6d95a99160f7d6d236ea5fe6362 ]

The version is fetched once in check_version(), which then does some
validation and then overwrites the version in userspace with the API
version supported by the kernel.  copy_params() then fetches the version
from userspace *again*, and this time no validation is done.  The result
is that the kernel's version number is completely controllable by
userspace, provided that userspace can win a race condition.

Fix this flaw by not copying the version back to the kernel the second
time.  This is not exploitable as the version is not further used in the
kernel.  However, it could become a problem if future patches start
relying on the version field.

Cc: stable@vger.kernel.org
Signed-off-by: Demi Marie Obenour <demi@invisiblethingslab.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm ioctl: have constant on the right side of the test

[ Upstream commit 5cae0aa77397015f530aeb34f3ced32db6ac2875 ]

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Stable-dep-of: 249bed821b4d ("dm ioctl: Avoid double-fetch of version")
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm: avoid split of quoted strings where possible

[ Upstream commit 2e84fecf19e1694338deec8bf6c90ff84f8f31fb ]

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Stable-dep-of: 249bed821b4d ("dm ioctl: Avoid double-fetch of version")
Signed-off-by: Sasha Levin <sashal@kernel.org>

dm: fix undue/missing spaces

[ Upstream commit 43be9c743c2553519c2093d1798b542f28095a51 ]

Signed-off-by: Heinz Mauelshagen <heinzm@redhat.com>
Signed-off-by: Mike Snitzer <snitzer@kernel.org>
Stable-dep-of: 249bed821b4d ("dm ioctl: Avoid double-fetch of version")
Signed-off-by: Sasha Levin <sashal@kernel.org>

i2c: xiic: Don't try to handle more interrupt events after error

[ Upstream commit cb6e45c9a0ad9e0f8664fd06db0227d185dc76ab ]

In xiic_process, it is possible that error events such as arbitration
lost or TX error can be raised in conjunction with other interrupt flags
such as TX FIFO empty or bus not busy. Error events result in the
controller being reset and the error returned to the calling request,
but the function could potentially try to keep handling the other
events, such as by writing more messages into the TX FIFO. Since the
transaction has already failed, this is not helpful and will just cause
issues.

This problem has been present ever since:

commit 7f9906bd7f72 ("i2c: xiic: Service all interrupts in isr")

which allowed non-error events to be handled after errors, but became
more obvious after:

commit 743e227a8959 ("i2c: xiic: Defer xiic_wakeup() and
__xiic_start_xfer() in xiic_process()")

which reworked the code to add a WARN_ON which triggers if both the
xfer_more and wakeup_req flags were set, since this combination is
not supposed to happen, but was occurring in this scenario.

Skip further interrupt handling after error flags are detected to avoid
this problem.

Fixes: 7f9906bd7f72 ("i2c: xiic: Service all interrupts in isr")
Signed-off-by: Robert Hancock <robert.hancock@calian.com>
Acked-by: Andi Shyti <andi.shyti@kernel.org>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

apparmor: fix missing error check for rhashtable_insert_fast

[ Upstream commit 000518bc5aef25d3f703592a0296d578c98b1517 ]

 rhashtable_insert_fast() could return err value when memory allocation is
 failed. but unpack_profile() do not check values and this always returns
 success value. This patch just adds error check code.

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

Fixes: e025be0f26d5 ("apparmor: support querying extended trusted helper extra data")
Signed-off-by: Danila Chernetsov <listdansp@mail.ru>
Signed-off-by: John Johansen <john.johansen@canonical.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

sh: dma: Fix DMA channel offset calculation

[ Upstream commit e82e47584847129a20b8c9f4a1dcde09374fb0e0 ]

Various SoCs of the SH3, SH4 and SH4A family, which use this driver,
feature a differing number of DMA channels, which can be distributed
between up to two DMAC modules. The existing implementation fails to
correctly accommodate for all those variations, resulting in wrong
channel offset calculations and leading to kernel panics.

Rewrite dma_base_addr() in order to properly calculate channel offsets
in a DMAC module. Fix dmaor_read_reg() and dmaor_write_reg(), so that
the correct DMAC module base is selected for the DMAOR register.

Fixes: 7f47c7189b3e8f19 ("sh: dma: More legacy cpu dma chainsawing.")
Signed-off-by: Artur Rojek <contact@artur-rojek.eu>
Reviewed-by: Geert Uytterhoeven <geert+renesas@glider.be>
Reviewed-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Link: https://lore.kernel.org/r/20230527164452.64797-2-contact@artur-rojek.eu
Signed-off-by: John Paul Adrian Glaubitz <glaubitz@physik.fu-berlin.de>
Signed-off-by: Sasha Levin <sashal@kernel.org>

s390/qeth: Fix vipa deletion

[ Upstream commit 80de809bd35e2a8999edf9f5aaa2d8de18921f11 ]

Change boolean parameter of function "qeth_l3_vipa_store" inside the
"qeth_l3_dev_vipa_del4_store" function from "true" to "false" because
"true" is used for adding a virtual ip address and "false" for deleting.

Fixes: 2390166a6b45 ("s390/qeth: clean up L3 sysfs code")
Reviewed-by: Alexandra Winter <wintera@linux.ibm.com>
Reviewed-by: Wenjia Zhang <wenjia@linux.ibm.com>
Signed-off-by: Thorsten Winkler <twinkler@linux.ibm.com>
Signed-off-by: Alexandra Winter <wintera@linux.ibm.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

afs: Fix accidental truncation when storing data

[ Upstream commit 03275585cabd0240944f19f33d7584a1b099a3a8 ]

When an AFS FS.StoreData RPC call is made, amongst other things it is
given the resultant file size to be.  On the server, this is processed
by truncating the file to new size and then writing the data.

Now, kafs has a lock (vnode->io_lock) that serves to serialise
operations against a specific vnode (ie.  inode), but the parameters for
the op are set before the lock is taken.  This allows two writebacks
(say sync and kswapd) to race - and if writes are ongoing the writeback
for a later write could occur before the writeback for an earlier one if
the latter gets interrupted.

Note that afs_writepages() cannot take i_mutex and only takes a shared
lock on vnode->validate_lock.

Also note that the server does the truncation and the write inside a
lock, so there's no problem at that end.

Fix this by moving the calculation for the proposed new i_size inside
the vnode->io_lock.  Also reset the iterator (which we might have read
from) and update the mtime setting there.

Fixes: bd80d8a80e12 ("afs: Use ITER_XARRAY for writing")
Reported-by: Marc Dionne <marc.dionne@auristor.com>
Signed-off-by: David Howells <dhowells@redhat.com>
Reviewed-by: Jeffrey Altman <jaltman@auristor.com>
Reviewed-by: Marc Dionne <marc.dionne@auristor.com>
cc: linux-afs@lists.infradead.org
cc: linux-fsdevel@vger.kernel.org
Link: https://lore.kernel.org/r/3526895.1687960024@warthog.procyon.org.uk/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx-af: fix hardware timestamp configuration

[ Upstream commit 14bb236b29922c4f57d8c05bfdbcb82677f917c9 ]

MAC block on CN10K (RPM) supports hardware timestamp configuration. The
previous patch which added timestamp configuration support has a bug.
Though the netdev driver requests to disable timestamp configuration,
the driver is always enabling it.

This patch fixes the same.

Fixes: d1489208681d ("octeontx2-af: cn10k: RPM hardware timestamp configuration")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: sja1105: always enable the send_meta options

[ Upstream commit a372d66af48506d9f7aaae2a474cd18f14d98cb8 ]

incl_srcpt has the limitation, mentioned in commit b4638af8885a ("net:
dsa: sja1105: always enable the INCL_SRCPT option"), that frames with a
MAC DA of 01:80:c2:xx:yy:zz will be received as 01:80:c2:00:00:zz unless
PTP RX timestamping is enabled.

The incl_srcpt option was initially unconditionally enabled, then that
changed with commit 42824463d38d ("net: dsa: sja1105: Limit use of
incl_srcpt to bridge+vlan mode"), then again with b4638af8885a ("net:
dsa: sja1105: always enable the INCL_SRCPT option"). Bottom line is that
it now needs to be always enabled, otherwise the driver does not have a
reliable source of information regarding source_port and switch_id for
link-local traffic (tag_8021q VLANs may be imprecise since now they
identify an entire bridging domain when ports are not standalone).

If we accept that PTP RX timestamping (and therefore, meta frame
generation) is always enabled in hardware, then that limitation could be
avoided and packets with any MAC DA can be properly received, because
meta frames do contain the original bytes from the MAC DA of their
associated link-local packet.

This change enables meta frame generation unconditionally, which also
has the nice side effects of simplifying the switch control path
(a switch reset is no longer required on hwtstamping settings change)
and the tagger data path (it no longer needs to be informed whether to
expect meta frames or not - it always does).

Fixes: 227d07a07ef1 ("net: dsa: sja1105: Add support for traffic through standalone ports")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: tag_sja1105: fix MAC DA patching from meta frames

[ Upstream commit 1dcf6efd5f0c1f4496b3ef7ec5a7db104a53b38c ]

The SJA1105 manual says that at offset 4 into the meta frame payload we
have "MAC destination byte 2" and at offset 5 we have "MAC destination
byte 1". These are counted from the LSB, so byte 1 is h_dest[ETH_HLEN-2]
aka h_dest[4] and byte 2 is h_dest[ETH_HLEN-3] aka h_dest[3].

The sja1105_meta_unpack() function decodes these the other way around,
so a frame with MAC DA 01:80:c2:11:22:33 is received by the network
stack as having 01:80:c2:22:11:33.

Fixes: e53e18a6fe4d ("net: dsa: sja1105: Receive and decode meta frames")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Florian Fainelli <florian.fainelli@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

pptp: Fix fib lookup calls.

[ Upstream commit 84bef5b6037c15180ef88ac4216dc621d16df1a6 ]

PPTP uses pppox sockets (struct pppox_sock). These sockets don't embed
an inet_sock structure, so it's invalid to call inet_sk() on them.

Therefore, the ip_route_output_ports() call in pptp_connect() has two
problems:

  * The tos variable is set with RT_CONN_FLAGS(sk), which calls
    inet_sk() on the pppox socket.

  * ip_route_output_ports() tries to retrieve routing flags using
    inet_sk_flowi_flags(), which is also going to call inet_sk() on the
    pppox socket.

While PPTP doesn't use inet sockets, it's actually really layered on
top of IP and therefore needs a proper way to do fib lookups. So let's
define pptp_route_output() to get a struct rtable from a pptp socket.
Let's also replace the ip_route_output_ports() call of pptp_xmit() for
consistency.

In practice, this means that:

  * pptp_connect() sets ->flowi4_tos and ->flowi4_flags to zero instead
    of using bits of unrelated struct pppox_sock fields.

  * pptp_xmit() now respects ->sk_mark and ->sk_uid.

  * pptp_xmit() now calls the security_sk_classify_flow() security
    hook, thus allowing to set ->flowic_secid.

  * pptp_xmit() now passes the pppox socket to xfrm_lookup_route().

Found by code inspection.

Fixes: 00959ade36ac ("PPTP: PPP over IPv4 (Point-to-Point Tunneling Protocol)")
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

riscv: move memblock_allow_resize() after linear mapping is ready

[ Upstream commit 85fadc0d04119c2fe4a20287767ab904c6d21ba1 ]

The initial memblock metadata is accessed from kernel image mapping. The
regions arrays need to "reallocated" from memblock and accessed through
linear mapping to cover more memblock regions. So the resizing should
not be allowed until linear mapping is ready. Note that there are
memblock allocations when building linear mapping.

This patch is similar to 24cc61d8cb5a ("arm64: memblock: don't permit
memblock resizing until linear mapping is up").

In following log, many memblock regions are reserved before
create_linear_mapping_page_table(). And then it triggered reallocation
of memblock.reserved.regions and memcpy the old array in kernel image
mapping to the new array in linear mapping which caused a page fault.

[    0.000000] memblock_reserve: [0x00000000bf01f000-0x00000000bf01ffff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf021000-0x00000000bf021fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf023000-0x00000000bf023fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf025000-0x00000000bf025fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf027000-0x00000000bf027fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf029000-0x00000000bf029fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf02b000-0x00000000bf02bfff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf02d000-0x00000000bf02dfff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf02f000-0x00000000bf02ffff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] memblock_reserve: [0x00000000bf030000-0x00000000bf030fff] early_init_fdt_scan_reserved_mem+0x28c/0x2c6
[    0.000000] OF: reserved mem: 0x0000000080000000..0x000000008007ffff (512 KiB) map non-reusable mmode_resv0@80000000
[    0.000000] memblock_reserve: [0x00000000bf000000-0x00000000bf001fed] paging_init+0x19a/0x5ae
[    0.000000] memblock_phys_alloc_range: 4096 bytes align=0x1000 from=0x0000000000000000 max_addr=0x0000000000000000 alloc_pmd_fixmap+0x14/0x1c
[    0.000000] memblock_reserve: [0x000000017ffff000-0x000000017fffffff] memblock_alloc_range_nid+0xb8/0x128
[    0.000000] memblock: reserved is doubled to 256 at [0x000000017fffd000-0x000000017fffe7ff]
[    0.000000] Unable to handle kernel paging request at virtual address ff600000ffffd000
[    0.000000] Oops [#1]
[    0.000000] Modules linked in:
[    0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 6.4.0-rc1-00011-g99a670b2069c #66
[    0.000000] Hardware name: riscv-virtio,qemu (DT)
[    0.000000] epc : __memcpy+0x60/0xf8
[    0.000000]  ra : memblock_double_array+0x192/0x248
[    0.000000] epc : ffffffff8081d214 ra : ffffffff80a3dfc0 sp : ffffffff81403bd0
[    0.000000]  gp : ffffffff814fbb38 tp : ffffffff8140dac0 t0 : 0000000001600000
[    0.000000]  t1 : 0000000000000000 t2 : 000000008f001000 s0 : ffffffff81403c60
[    0.000000]  s1 : ffffffff80c0bc98 a0 : ff600000ffffd000 a1 : ffffffff80c0bcd8
[    0.000000]  a2 : 0000000000000c00 a3 : ffffffff80c0c8d8 a4 : 0000000080000000
[    0.000000]  a5 : 0000000000080000 a6 : 0000000000000000 a7 : 0000000080200000
[    0.000000]  s2 : ff600000ffffd000 s3 : 0000000000002000 s4 : 0000000000000c00
[    0.000000]  s5 : ffffffff80c0bc60 s6 : ffffffff80c0bcc8 s7 : 0000000000000000
[    0.000000]  s8 : ffffffff814fd0a8 s9 : 000000017fffe7ff s10: 0000000000000000
[    0.000000]  s11: 0000000000001000 t3 : 0000000000001000 t4 : 0000000000000000
[    0.000000]  t5 : 000000008f003000 t6 : ff600000ffffd000
[    0.000000] status: 0000000200000100 badaddr: ff600000ffffd000 cause: 000000000000000f
[    0.000000] [<ffffffff8081d214>] __memcpy+0x60/0xf8
[    0.000000] [<ffffffff80a3e1a2>] memblock_add_range.isra.14+0x12c/0x162
[    0.000000] [<ffffffff80a3e36a>] memblock_reserve+0x6e/0x8c
[    0.000000] [<ffffffff80a123fc>] memblock_alloc_range_nid+0xb8/0x128
[    0.000000] [<ffffffff80a1256a>] memblock_phys_alloc_range+0x5e/0x6a
[    0.000000] [<ffffffff80a04732>] alloc_pmd_fixmap+0x14/0x1c
[    0.000000] [<ffffffff80a0475a>] alloc_p4d_fixmap+0xc/0x14
[    0.000000] [<ffffffff80a04a36>] create_pgd_mapping+0x98/0x17c
[    0.000000] [<ffffffff80a04e9e>] create_linear_mapping_range.constprop.10+0xe4/0x112
[    0.000000] [<ffffffff80a05bb8>] paging_init+0x3ec/0x5ae
[    0.000000] [<ffffffff80a03354>] setup_arch+0xb2/0x576
[    0.000000] [<ffffffff80a00726>] start_kernel+0x72/0x57e
[    0.000000] Code: b303 0285 b383 0305 be03 0385 be83 0405 bf03 0485 (b023) 00ef
[    0.000000] ---[ end trace 0000000000000000 ]---
[    0.000000] Kernel panic - not syncing: Attempted to kill the idle task!
[    0.000000] ---[ end Kernel panic - not syncing: Attempted to kill the idle task! ]---

Fixes: 671f9a3e2e24 ("RISC-V: Setup initial page tables in two stages")
Signed-off-by: Woody Zhang <woodylab@foxmail.com>
Tested-by: Song Shuai <songshuaishuai@tinylab.org>
Link: https://lore.kernel.org/r/tencent_FBB94CE615C5CCE7701CD39C15CCE0EE9706@qq.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

fanotify: disallow mount/sb marks on kernel internal pseudo fs

[ Upstream commit 69562eb0bd3e6bb8e522a7b254334e0fb30dff0c ]

Hopefully, nobody is trying to abuse mount/sb marks for watching all
anonymous pipes/inodes.

I cannot think of a good reason to allow this - it looks like an
oversight that dated back to the original fanotify API.

Link: https://lore.kernel.org/linux-fsdevel/20230628101132.kvchg544mczxv2pm@quack3/
Fixes: 0ff21db9fcc3 ("fanotify: hooks the fanotify_mark syscall to the vfsmount code")
Signed-off-by: Amir Goldstein <amir73il@gmail.com>
Reviewed-by: Christian Brauner <brauner@kernel.org>
Signed-off-by: Jan Kara <jack@suse.cz>
Message-Id: <20230629042044.25723-1-amir73il@gmail.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net/sched: act_pedit: Add size check for TCA_PEDIT_PARMS_EX

[ Upstream commit 30c45b5361d39b4b793780ffac5538090b9e2eb1 ]

The attribute TCA_PEDIT_PARMS_EX is not be included in pedit_policy and
one malicious user could fake a TCA_PEDIT_PARMS_EX whose length is
smaller than the intended sizeof(struct tc_pedit). Hence, the
dereference in tcf_pedit_init() could access dirty heap data.

static int tcf_pedit_init(...)
{
  // ...
  pattr = tb[TCA_PEDIT_PARMS]; // TCA_PEDIT_PARMS is included
  if (!pattr)
    pattr = tb[TCA_PEDIT_PARMS_EX]; // but this is not

  // ...
  parm = nla_data(pattr);

  index = parm->index; // parm is able to be smaller than 4 bytes
                       // and this dereference gets dirty skb_buff
                       // data created in netlink_sendmsg
}

This commit adds TCA_PEDIT_PARMS_EX length in pedit_policy which avoid
the above case, just like the TCA_PEDIT_PARMS.

Fixes: 71d0ed7079df ("net/act_pedit: Support using offset relative to the conventional network headers")
Signed-off-by: Lin Ma <linma@zju.edu.cn>
Reviewed-by: Pedro Tammela <pctammela@mojatatu.com>
Link: https://lore.kernel.org/r/20230703110842.590282-1-linma@zju.edu.cn
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

xsk: Honor SO_BINDTODEVICE on bind

[ Upstream commit f7306acec9aae9893d15e745c8791124d42ab10a ]

Initial creation of an AF_XDP socket requires CAP_NET_RAW capability. A
privileged process might create the socket and pass it to a non-privileged
process for later use. However, that process will be able to bind the socket
to any network interface. Even though it will not be able to receive any
traffic without modification of the BPF map, the situation is not ideal.

Sockets already have a mechanism that can be used to restrict what interface
they can be attached to. That is SO_BINDTODEVICE.

To change the SO_BINDTODEVICE binding the process will need CAP_NET_RAW.

Make xsk_bind() honor the SO_BINDTODEVICE in order to allow safer workflow
when non-privileged process is using AF_XDP.

The intended workflow is following:

  1. First process creates a bare socket with socket(AF_XDP, ...).
  2. First process loads the XSK program to the interface.
  3. First process adds the socket fd to a BPF map.
  4. First process ties socket fd to a particular interface using
     SO_BINDTODEVICE.
  5. First process sends socket fd to a second process.
  6. Second process allocates UMEM.
  7. Second process binds socket to the interface with bind(...).
  8. Second process sends/receives the traffic.

All the steps above are possible today if the first process is privileged
and the second one has sufficient RLIMIT_MEMLOCK and no capabilities.
However, the second process will be able to bind the socket to any interface
it wants on step 7 and send traffic from it. With the proposed change, the
second process will be able to bind the socket only to a specific interface
chosen by the first process at step 4.

Fixes: 965a99098443 ("xsk: add support for bind for Rx")
Signed-off-by: Ilya Maximets <i.maximets@ovn.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Acked-by: John Fastabend <john.fastabend@gmail.com>
Acked-by: Jason Wang <jasowang@redhat.com>
Link: https://lore.kernel.org/bpf/20230703175329.3259672-1-i.maximets@ovn.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

bpf, btf: Warn but return no error for NULL btf from __register_btf_kfunc_id_set()

[ Upstream commit 3de4d22cc9ac7c9f38e10edcf54f9a8891a9c2aa ]

__register_btf_kfunc_id_set() assumes .BTF to be part of the module's .ko
file if CONFIG_DEBUG_INFO_BTF is enabled. If that's not the case, the
function prints an error message and return an error. As a result, such
modules cannot be loaded.

However, the section could be stripped out during a build process. It would
be better to let the modules loaded, because their basic functionalities
have no problem [0], though the BTF functionalities will not be supported.
Make the function to lower the level of the message from error to warn, and
return no error.

  [0] https://lore.kernel.org/bpf/20220219082037.ow2kbq5brktf4f2u@apollo.legion

Fixes: c446fdacb10d ("bpf: fix register_btf_kfunc_id_set for !CONFIG_DEBUG_INFO_BTF")
Reported-by: Alexander Egorenkov <Alexander.Egorenkov@ibm.com>
Suggested-by: Kumar Kartikeya Dwivedi <memxor@gmail.com>
Signed-off-by: SeongJae Park <sj@kernel.org>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>
Acked-by: Jiri Olsa <jolsa@kernel.org>
Link: https://lore.kernel.org/bpf/87y228q66f.fsf@oc8242746057.ibm.com
Link: https://lore.kernel.org/bpf/20220219082037.ow2kbq5brktf4f2u@apollo.legion
Link: https://lore.kernel.org/bpf/20230701171447.56464-1-sj@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

tcp: annotate data races in __tcp_oow_rate_limited()

[ Upstream commit 998127cdb4699b9d470a9348ffe9f1154346be5f ]

request sockets are lockless, __tcp_oow_rate_limited() could be called
on the same object from different cpus. This is harmless.

Add READ_ONCE()/WRITE_ONCE() annotations to avoid a KCSAN report.

Fixes: 4ce7e93cb3fe ("tcp: rate limit ACK sent by SYN_RECV request sockets")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: fix net_dev_start_xmit trace event vs skb_transport_offset()

[ Upstream commit f88fcb1d7d961b4b402d675109726f94db87571c ]

After blamed commit, we must be more careful about using
skb_transport_offset(), as reminded us by syzbot:

WARNING: CPU: 0 PID: 10 at include/linux/skbuff.h:2868 skb_transport_offset include/linux/skbuff.h:2977 [inline]
WARNING: CPU: 0 PID: 10 at include/linux/skbuff.h:2868 perf_trace_net_dev_start_xmit+0x89a/0xce0 include/trace/events/net.h:14
Modules linked in:
CPU: 0 PID: 10 Comm: kworker/u4:1 Not tainted 6.1.30-syzkaller #0
Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 05/27/2023
Workqueue: bat_events batadv_iv_send_outstanding_bat_ogm_packet
RIP: 0010:skb_transport_header include/linux/skbuff.h:2868 [inline]
RIP: 0010:skb_transport_offset include/linux/skbuff.h:2977 [inline]
RIP: 0010:perf_trace_net_dev_start_xmit+0x89a/0xce0 include/trace/events/net.h:14
Code: 8b 04 25 28 00 00 00 48 3b 84 24 c0 00 00 00 0f 85 4e 04 00 00 48 8d 65 d8 5b 41 5c 41 5d 41 5e 41 5f 5d c3 cc e8 56 22 01 fd <0f> 0b e9 f6 fc ff ff 89 f9 80 e1 07 80 c1 03 38 c1 0f 8c 86 f9 ff
RSP: 0018:ffffc900002bf700 EFLAGS: 00010293
RAX: ffffffff8485d8ca RBX: 000000000000ffff RCX: ffff888100914280
RDX: 0000000000000000 RSI: 000000000000ffff RDI: 000000000000ffff
RBP: ffffc900002bf818 R08: ffffffff8485d5b6 R09: fffffbfff0f8fb5e
R10: 0000000000000000 R11: dffffc0000000001 R12: 1ffff110217d8f67
R13: ffff88810bec7b3a R14: dffffc0000000000 R15: dffffc0000000000
FS: 0000000000000000(0000) GS:ffff8881f6a00000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00007f96cf6d52f0 CR3: 000000012224c000 CR4: 0000000000350ef0
Call Trace:
<TASK>
[<ffffffff84715e35>] trace_net_dev_start_xmit include/trace/events/net.h:14 [inline]
[<ffffffff84715e35>] xmit_one net/core/dev.c:3643 [inline]
[<ffffffff84715e35>] dev_hard_start_xmit+0x705/0x980 net/core/dev.c:3660
[<ffffffff8471a232>] __dev_queue_xmit+0x16b2/0x3370 net/core/dev.c:4324
[<ffffffff85416493>] dev_queue_xmit include/linux/netdevice.h:3030 [inline]
[<ffffffff85416493>] batadv_send_skb_packet+0x3f3/0x680 net/batman-adv/send.c:108
[<ffffffff85416744>] batadv_send_broadcast_skb+0x24/0x30 net/batman-adv/send.c:127
[<ffffffff853bc52a>] batadv_iv_ogm_send_to_if net/batman-adv/bat_iv_ogm.c:393 [inline]
[<ffffffff853bc52a>] batadv_iv_ogm_emit net/batman-adv/bat_iv_ogm.c:421 [inline]
[<ffffffff853bc52a>] batadv_iv_send_outstanding_bat_ogm_packet+0x69a/0x840 net/batman-adv/bat_iv_ogm.c:1701
[<ffffffff8151023c>] process_one_work+0x8ac/0x1170 kernel/workqueue.c:2289
[<ffffffff81511938>] worker_thread+0xaa8/0x12d0 kernel/workqueue.c:2436

Fixes: 66e4c8d95008 ("net: warn if transport header was not set")
Signed-off-by: Eric Dumazet <edumazet@google.com>
Reported-by: syzbot <syzkaller@googlegroups.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: tag_sja1105: fix source port decoding in vlan_filtering=0 bridge mode

[ Upstream commit a398b9ea0c3b791b7a0f4c6029a62cf628f97f22 ]

There was a regression introduced by the blamed commit, where pinging to
a VLAN-unaware bridge would fail with the repeated message "Couldn't
decode source port" coming from the tagging protocol driver.

When receiving packets with a bridge_vid as determined by
dsa_tag_8021q_bridge_join(), dsa_8021q_rcv() will decode:
- source_port = 0 (which isn't really valid, more like "don't know")
- switch_id = 0 (which isn't really valid, more like "don't know")
- vbid = value in range 1-7

Since the blamed patch has reversed the order of the checks, we are now
going to believe that source_port != -1 and switch_id != -1, so they're
valid, but they aren't.

The minimal solution to the problem is to only populate source_port and
switch_id with what dsa_8021q_rcv() came up with, if the vbid is zero,
i.e. the source port information is trustworthy.

Fixes: c1ae02d87689 ("net: dsa: tag_sja1105: always prefer source port information from INCL_SRCPT")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: bridge: keep ports without IFF_UNICAST_FLT in BR_PROMISC mode

[ Upstream commit 6ca3c005d0604e8d2b439366e3923ea58db99641 ]

According to the synchronization rules for .ndo_get_stats() as seen in
Documentation/networking/netdevices.rst, acquiring a plain spin_lock()
should not be illegal, but the bridge driver implementation makes it so.

After running these commands, I am being faced with the following
lockdep splat:

$ ip link add link swp0 name macsec0 type macsec encrypt on && ip link set swp0 up
$ ip link add dev br0 type bridge vlan_filtering 1 && ip link set br0 up
$ ip link set macsec0 master br0 && ip link set macsec0 up

  ========================================================
  WARNING: possible irq lock inversion dependency detected
  6.4.0-04295-g31b577b4bd4a #603 Not tainted
  --------------------------------------------------------
  swapper/1/0 just changed the state of lock:
  ffff6bd348724cd8 (&br->lock){+.-.}-{3:3}, at: br_forward_delay_timer_expired+0x34/0x198
  but this lock took another, SOFTIRQ-unsafe lock in the past:
   (&ocelot->stats_lock){+.+.}-{3:3}

  and interrupts could create inverse lock ordering between them.

  other info that might help us debug this:
  Chain exists of:
    &br->lock --> &br->hash_lock --> &ocelot->stats_lock

   Possible interrupt unsafe locking scenario:

         CPU0                    CPU1
         ----                    ----
    lock(&ocelot->stats_lock);
                                 local_irq_disable();
                                 lock(&br->lock);
                                 lock(&br->hash_lock);
    <Interrupt>
      lock(&br->lock);

   *** DEADLOCK ***

(details about the 3 locks skipped)

swp0 is instantiated by drivers/net/dsa/ocelot/felix.c, and this
only matters to the extent that its .ndo_get_stats64() method calls
spin_lock(&ocelot->stats_lock).

Documentation/locking/lockdep-design.rst says:

| A lock is irq-safe means it was ever used in an irq context, while a lock
| is irq-unsafe means it was ever acquired with irq enabled.

(...)

| Furthermore, the following usage based lock dependencies are not allowed
| between any two lock-classes::
|
|    <hardirq-safe>   ->  <hardirq-unsafe>
|    <softirq-safe>   ->  <softirq-unsafe>

Lockdep marks br->hash_lock as softirq-safe, because it is sometimes
taken in softirq context (for example br_fdb_update() which runs in
NET_RX softirq), and when it's not in softirq context it blocks softirqs
by using spin_lock_bh().

Lockdep marks ocelot->stats_lock as softirq-unsafe, because it never
blocks softirqs from running, and it is never taken from softirq
context. So it can always be interrupted by softirqs.

There is a call path through which a function that holds br->hash_lock:
fdb_add_hw_addr() will call a function that acquires ocelot->stats_lock:
ocelot_port_get_stats64(). This can be seen below:

ocelot_port_get_stats64+0x3c/0x1e0
felix_get_stats64+0x20/0x38
dsa_slave_get_stats64+0x3c/0x60
dev_get_stats+0x74/0x2c8
rtnl_fill_stats+0x4c/0x150
rtnl_fill_ifinfo+0x5cc/0x7b8
rtmsg_ifinfo_build_skb+0xe4/0x150
rtmsg_ifinfo+0x5c/0xb0
__dev_notify_flags+0x58/0x200
__dev_set_promiscuity+0xa0/0x1f8
dev_set_promiscuity+0x30/0x70
macsec_dev_change_rx_flags+0x68/0x88
__dev_set_promiscuity+0x1a8/0x1f8
__dev_set_rx_mode+0x74/0xa8
dev_uc_add+0x74/0xa0
fdb_add_hw_addr+0x68/0xd8
fdb_add_local+0xc4/0x110
br_fdb_add_local+0x54/0x88
br_add_if+0x338/0x4a0
br_add_slave+0x20/0x38
do_setlink+0x3a4/0xcb8
rtnl_newlink+0x758/0x9d0
rtnetlink_rcv_msg+0x2f0/0x550
netlink_rcv_skb+0x128/0x148
rtnetlink_rcv+0x24/0x38

the plain English explanation for it is:

The macsec0 bridge port is created without p->flags & BR_PROMISC,
because it is what br_manage_promisc() decides for a VLAN filtering
bridge with a single auto port.

As part of the br_add_if() procedure, br_fdb_add_local() is called for
the MAC address of the device, and this results in a call to
dev_uc_add() for macsec0 while the softirq-safe br->hash_lock is taken.

Because macsec0 does not have IFF_UNICAST_FLT, dev_uc_add() ends up
calling __dev_set_promiscuity() for macsec0, which is propagated by its
implementation, macsec_dev_change_rx_flags(), to the lower device: swp0.
This triggers the call path:

dev_set_promiscuity(swp0)
-> rtmsg_ifinfo()
   -> dev_get_stats()
      -> ocelot_port_get_stats64()

with a calling context that lockdep doesn't like (br->hash_lock held).

Normally we don't see this, because even though many drivers that can be
bridge ports don't support IFF_UNICAST_FLT, we need a driver that

(a) doesn't support IFF_UNICAST_FLT, *and*
(b) it forwards the IFF_PROMISC flag to another driver, and
(c) *that* driver implements ndo_get_stats64() using a softirq-unsafe
    spinlock.

Condition (b) is necessary because the first __dev_set_rx_mode() calls
__dev_set_promiscuity() with "bool notify=false", and thus, the
rtmsg_ifinfo() code path won't be entered.

The same criteria also hold true for DSA switches which don't report
IFF_UNICAST_FLT. When the DSA master uses a spin_lock() in its
ndo_get_stats64() method, the same lockdep splat can be seen.

I think the deadlock possibility is real, even though I didn't reproduce
it, and I'm thinking of the following situation to support that claim:

fdb_add_hw_addr() runs on a CPU A, in a context with softirqs locally
disabled and br->hash_lock held, and may end up attempting to acquire
ocelot->stats_lock.

In parallel, ocelot->stats_lock is currently held by a thread B (say,
ocelot_check_stats_work()), which is interrupted while holding it by a
softirq which attempts to lock br->hash_lock.

Thread B cannot make progress because br->hash_lock is held by A. Whereas
thread A cannot make progress because ocelot->stats_lock is held by B.

When taking the issue at face value, the bridge can avoid that problem
by simply making the ports promiscuous from a code path with a saner
calling context (br->hash_lock not held). A bridge port without
IFF_UNICAST_FLT is going to become promiscuous as soon as we call
dev_uc_add() on it (which we do unconditionally), so why not be
preemptive and make it promiscuous right from the beginning, so as to
not be taken by surprise.

With this, we've broken the links between code that holds br->hash_lock
or br->lock and code that calls into the ndo_change_rx_flags() or
ndo_get_stats64() ops of the bridge port.

Fixes: 2796d0c648c9 ("bridge: Automatically manage port promiscuous mode.")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc: dts: turris1x.dts: Fix PCIe MEM size for pci2 node

[ Upstream commit abaa02fc944f2f9f2c2e1925ddaceaf35c48528c ]

Freescale PCIe controllers on their PCIe Root Ports do not have any
mappable PCI BAR allocate from PCIe MEM.

Information about 1MB window on BAR0 of PCIe Root Port was misleading
because Freescale PCIe controllers have at BAR0 position different register
PEXCSRBAR, and kernel correctly skipts BAR0 for these Freescale PCIe Root
Ports.

So update comment about P2020 PCIe Root Port and decrease PCIe MEM size
required for PCIe controller (pci2 node) on which is on-board xHCI
controller.

lspci confirms that on P2020 PCIe Root Port is no PCI BAR and /proc/iomem
sees that only c0000000-c000ffff and c0010000-c0011fff ranges are used.

Fixes: 54c15ec3b738 ("powerpc: dts: Add DTS file for CZ.NIC Turris 1.x routers")
Signed-off-by: Pali Rohár <pali@kernel.org>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20230505172818.18416-1-pali@kernel.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

powerpc: allow PPC_EARLY_DEBUG_CPM only when SERIAL_CPM=y

[ Upstream commit 39f49684036d24af800ff194c33c7b2653c591d7 ]

In a randconfig with CONFIG_SERIAL_CPM=m and
CONFIG_PPC_EARLY_DEBUG_CPM=y, there is a build error:
ERROR: modpost: "udbg_putc" [drivers/tty/serial/cpm_uart/cpm_uart.ko] undefined!

Prevent the build error by allowing PPC_EARLY_DEBUG_CPM only when
SERIAL_CPM=y.

Fixes: c374e00e17f1 ("[POWERPC] Add early debug console for CPM serial ports.")
Signed-off-by: Randy Dunlap <rdunlap@infradead.org>
Reviewed-by: Pali Rohár <pali@kernel.org>
Reviewed-by: Christophe Leroy <christophe.leroy@csgroup.eu>
Signed-off-by: Michael Ellerman <mpe@ellerman.id.au>
Link: https://msgid.link/20230701054714.30512-1-rdunlap@infradead.org
Signed-off-by: Sasha Levin <sashal@kernel.org>

ntfs: Fix panic about slab-out-of-bounds caused by ntfs_listxattr()

[ Upstream commit 3c675ddffb17a8b1e32efad5c983254af18b12c2 ]

Here is a BUG report from syzbot:

BUG: KASAN: slab-out-of-bounds in ntfs_list_ea fs/ntfs3/xattr.c:191 [inline]
BUG: KASAN: slab-out-of-bounds in ntfs_listxattr+0x401/0x570 fs/ntfs3/xattr.c:710
Read of size 1 at addr ffff888021acaf3d by task syz-executor128/3632

Call Trace:
 ntfs_list_ea fs/ntfs3/xattr.c:191 [inline]
 ntfs_listxattr+0x401/0x570 fs/ntfs3/xattr.c:710
 vfs_listxattr fs/xattr.c:457 [inline]
 listxattr+0x293/0x2d0 fs/xattr.c:804

Fix the logic of ea_all iteration. When the ea->name_len is 0,
return immediately, or Add2Ptr() would visit invalid memory
in the next loop.

Fixes: be71b5cba2e6 ("fs/ntfs3: Add attrib operations")
Reported-by: syzbot+9fcea5ef6dc4dc72d334@syzkaller.appspotmail.com
Signed-off-by: Zeng Heng <zengheng4@huawei.com>
[almaz.alexandrovich@paragon-software.com: lines of the patch have changed]
Signed-off-by: Konstantin Komarov <almaz.alexandrovich@paragon-software.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: Add validation before accessing cgx and lmac

[ Upstream commit 79ebb53772c95d3a6ae51b3c65f9985fdd430df6 ]

with the addition of new MAC blocks like CN10K RPM and CN10KB
RPM_USX, LMACs are noncontiguous and CGX blocks are also
noncontiguous. But during RVU driver initialization, the driver
is assuming they are contiguous and trying to access
cgx or lmac with their id which is resulting in kernel panic.

This patch fixes the issue by adding proper checks.

[   23.219150] pc : cgx_lmac_read+0x38/0x70
[   23.219154] lr : rvu_program_channels+0x3f0/0x498
[   23.223852] sp : ffff000100d6fc80
[   23.227158] x29: ffff000100d6fc80 x28: ffff00010009f880 x27:
000000000000005a
[   23.234288] x26: ffff000102586768 x25: 0000000000002500 x24:
fffffffffff0f000

Fixes: 91c6945ea1f9 ("octeontx2-af: cn10k: Add RPM MAC support")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

octeontx2-af: Fix mapping for NIX block from CGX connection

[ Upstream commit 2e7bc57b976bb016c6569a54d95c1b8d88f9450a ]

Firmware configures NIX block mapping for all MAC blocks.
The current implementation reads the configuration and
creates the mapping between RVU PF  and NIX blocks. But
this configuration is only valid for silicons that support
multiple blocks. For all other silicons, all MAC blocks
map to NIX0.

This patch corrects the mapping by adding a check for the same.

Fixes: c5a73b632b90 ("octeontx2-af: Map NIX block from CGX connection")
Signed-off-by: Hariprasad Kelam <hkelam@marvell.com>
Signed-off-by: Sunil Goutham <sgoutham@marvell.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Sasha Levin <sashal@kernel.org>

f2fs: fix error path handling in truncate_dnode()

[ Upstream commit 0135c482fa97e2fd8245cb462784112a00ed1211 ]

If truncate_node() fails in truncate_dnode(), it missed to call
f2fs_put_page(), fix it.

Fixes: 7735730d39d7 ("f2fs: fix to propagate error from __get_meta_page()")
Signed-off-by: Chao Yu <chao@kernel.org>
Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mailbox: ti-msgmgr: Fill non-message tx data fields with 0x0

[ Upstream commit 1b712f18c461bd75f018033a15cf381e712806b5 ]

Sec proxy/message manager data buffer is 60 bytes with the last of the
registers indicating transmission completion. This however poses a bit
of a challenge.

The backing memory for sec_proxy / message manager is regular memory,
and all sec proxy does is to trigger a burst of all 60 bytes of data
over to the target thread backing ring accelerator. It doesn't do a
memory scrub when it moves data out in the burst. When we transmit
multiple messages, remnants of previous message is also transmitted
which results in some random data being set in TISCI fields of
messages that have been expanded forward.

The entire concept of backward compatibility hinges on the fact that
the unused message fields remain 0x0 allowing for 0x0 value to be
specially considered when backward compatibility of message extension
is done.

So, instead of just writing the completion register, we continue
to fill the message buffer up with 0x0 (note: for partial message
involving completion, we already do this).

This allows us to scale and introduce ABI changes back also work with
other boot stages that may have left data in the internal memory.

While at this, be consistent and explicit with the data_reg pointer
increment.

Fixes: aace66b170ce ("mailbox: Introduce TI message manager driver")
Signed-off-by: Nishanth Menon <nm@ti.com>
Signed-off-by: Jassi Brar <jaswinder.singh@linaro.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amd: Don't try to enable secure display TA multiple times

[ Upstream commit 5c6d52ff4b61e5267b25be714eb5a9ba2a338199 ]

If the securedisplay TA failed to load the first time, it's unlikely
to work again after a suspend/resume cycle or reset cycle and it appears
to be causing problems in futher attempts.

Fixes: e42dfa66d592 ("drm/amdgpu: Add secure display TA load for Renoir")
Reported-by: Filip Hejsek <filip.hejsek@gmail.com>
Closes: https://gitlab.freedesktop.org/drm/amd/-/issues/2633
Signed-off-by: Mario Limonciello <mario.limonciello@amd.com>
Acked-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

drm/amdgpu: fix number of fence calculations

[ Upstream commit 570b295248b00c3cf4cf59e397de5cb2361e10c2 ]

Since adding gang submit we need to take the gang size into account
while reserving fences.

Signed-off-by: Christian König <christian.koenig@amd.com>
Fixes: 4624459c84d7 ("drm/amdgpu: add gang submit frontend v6")
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

spi: bcm-qspi: return error if neither hif_mspi nor mspi is available

[ Upstream commit 7c1f23ad34fcdace50275a6aa1e1969b41c6233f ]

If neither a "hif_mspi" nor "mspi" resource is present, the driver will
just early exit in probe but still return success. Apart from not doing
anything meaningful, this would then also lead to a null pointer access
on removal, as platform_get_drvdata() would return NULL, which it would
then try to dereference when trying to unregister the spi master.

Fix this by unconditionally calling devm_ioremap_resource(), as it can
handle a NULL res and will then return a viable ERR_PTR() if we get one.

The "return 0;" was previously a "goto qspi_resource_err;" where then
ret was returned, but since ret was still initialized to 0 at this place
this was a valid conversion in 63c5395bb7a9 ("spi: bcm-qspi: Fix
use-after-free on unbind"). The issue was not introduced by this commit,
only made more obvious.

Fixes: fa236a7ef240 ("spi: bcm-qspi: Add Broadcom MSPI driver")
Signed-off-by: Jonas Gorski <jonas.gorski@gmail.com>
Reviewed-by: Kamal Dasu <kamal.dasu@broadcom.com>
Link: https://lore.kernel.org/r/20230629134306.95823-1-jonas.gorski@gmail.com
Signed-off-by: Mark Brown <broonie@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

mlxsw: minimal: fix potential memory leak in mlxsw_m_linecards_init

[ Upstream commit 08fc75735fda3be97194bfbf3c899c87abb3d0fe ]

The line cards array is not freed in the error path of
mlxsw_m_linecards_init(), which can lead to a memory leak. Fix by
freeing the array in the error path, thereby making the error path
identical to mlxsw_m_linecards_fini().

Fixes: 01328e23a476 ("mlxsw: minimal: Extend module to port mapping with slot index")
Signed-off-by: Zhengchao Shao <shaozhengchao@huawei.com>
Reviewed-by: Petr Machata <petrm@nvidia.com>
Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Link: https://lore.kernel.org/r/20230630012647.1078002-1-shaozhengchao@huawei.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: vsc73xx: fix MTU configuration

[ Upstream commit 3cf62c8177adb0db9e15c8b898c44f997acf3ebf ]

Switch in MAXLEN register stores the maximum size of a data frame.
The MTU size is 18 bytes smaller than the frame size.

The current settings are causing problems with packet forwarding.
This patch fixes the MTU settings to proper values.

Fixes: fb77ffc6ec86 ("net: dsa: vsc73xx: make the MTU configurable")
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Pawel Dembicki <paweldembicki@gmail.com>
Reviewed-by: Vladimir Oltean <olteanv@gmail.com>
Link: https://lore.kernel.org/r/20230628194327.1765644-1-paweldembicki@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

ibmvnic: Do not reset dql stats on NON_FATAL err

[ Upstream commit 48538ccb825b05544ec308a509e2cc9c013402db ]

All ibmvnic resets, make a call to netdev_tx_reset_queue() when
re-opening the device. netdev_tx_reset_queue() resets the num_queued
and num_completed byte counters. These stats are used in Byte Queue
Limit (BQL) algorithms. The difference between these two stats tracks
the number of bytes currently sitting on the physical NIC. ibmvnic
increases the number of queued bytes though calls to
netdev_tx_sent_queue() in the drivers xmit function. When, VIOS reports
that it is done transmitting bytes, the ibmvnic device increases the
number of completed bytes through calls to netdev_tx_completed_queue().
It is important to note that the driver batches its transmit calls and
num_queued is increased every time that an skb is added to the next
batch, not necessarily when the batch is sent to VIOS for transmission.

Unlike other reset types, a NON FATAL reset will not flush the sub crq
tx buffers. Therefore, it is possible for the batched skb array to be
partially full. So if there is call to netdev_tx_reset_queue() when
re-opening the device, the value of num_queued (0) would not account
for the skb's that are currently batched. Eventually, when the batch
is sent to VIOS, the call to netdev_tx_completed_queue() would increase
num_completed to a value greater than the num_queued. This causes a
BUG_ON crash:

ibmvnic 30000002: Firmware reports error, cause: adapter problem.
Starting recovery...
ibmvnic 30000002: tx error 600
ibmvnic 30000002: tx error 600
ibmvnic 30000002: tx error 600
ibmvnic 30000002: tx error 600
------------[ cut here ]------------
kernel BUG at lib/dynamic_queue_limits.c:27!
Oops: Exception in kernel mode, sig: 5
[....]
NIP dql_completed+0x28/0x1c0
LR ibmvnic_complete_tx.isra.0+0x23c/0x420 [ibmvnic]
Call Trace:
ibmvnic_complete_tx.isra.0+0x3f8/0x420 [ibmvnic] (unreliable)
ibmvnic_interrupt_tx+0x40/0x70 [ibmvnic]
__handle_irq_event_percpu+0x98/0x270
---[ end trace ]---

Therefore, do not reset the dql stats when performing a NON_FATAL reset.

Fixes: 0d973388185d ("ibmvnic: Introduce xmit_more support using batched subCRQ hcalls")
Signed-off-by: Nick Child <nnac123@linux.ibm.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Bluetooth: MGMT: Fix marking SCAN_RSP as not connectable

[ Upstream commit 73f55453ea5236a586a7f1b3d5e2ee051d655351 ]

When receiving a scan response there is no way to know if the remote
device is connectable or not, so when it cannot be merged don't
make any assumption and instead just mark it with a new flag defined as
MGMT_DEV_FOUND_SCAN_RSP so userspace can tell it is a standalone
SCAN_RSP.

Link: https://lore.kernel.org/linux-bluetooth/CABBYNZ+CYMsDSPTxBn09Js3BcdC-x7vZFfyLJ3ppZGGwJKmUTw@mail.gmail.com/
Fixes: c70a7e4cc8d2 ("Bluetooth: Add support for Not Connectable flag for Device Found events")
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Bluetooth: MGMT: Use BIT macro when defining bitfields

[ Upstream commit a80d2c545ded86d0350b9a870735565d8b749786 ]

This makes use of BIT macro when defining bitfields which makes it
clearer what bit it is toggling.

Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Stable-dep-of: 73f55453ea52 ("Bluetooth: MGMT: Fix marking SCAN_RSP as not connectable")
Signed-off-by: Sasha Levin <sashal@kernel.org>

Bluetooth: MGMT: add CIS feature bits to controller information

[ Upstream commit 2394186a2cefb9a45a029281a55749804dd8c556 ]

Userspace needs to know whether the adapter has feature support for
Connected Isochronous Stream - Central/Peripheral, so it can set up
LE Audio features accordingly.

Expose these feature bits as settings in MGMT controller info.

Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Stable-dep-of: 73f55453ea52 ("Bluetooth: MGMT: Fix marking SCAN_RSP as not connectable")
Signed-off-by: Sasha Levin <sashal@kernel.org>

Bluetooth: ISO: use hci_sync for setting CIG parameters

[ Upstream commit 6b9545dc9f8ff01d8bc1229103960d9cd265343f ]

When reconfiguring CIG after disconnection of the last CIS, LE Remove
CIG shall be sent before LE Set CIG Parameters.  Otherwise, it fails
because CIG is in the inactive state and not configurable (Core v5.3
Vol 6 Part B Sec. 4.5.14.3). This ordering is currently wrong under
suitable timing conditions, because LE Remove CIG is sent via the
hci_sync queue and may be delayed, but Set CIG Parameters is via
hci_send_cmd.

Make the ordering well-defined by sending also Set CIG Parameters via
hci_sync.

Fixes: 26afbd826ee3 ("Bluetooth: Add initial implementation of CIS connections")
Signed-off-by: Pauli Virtanen <pav@iki.fi>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Bluetooth: fix invalid-bdaddr quirk for non-persistent setup

[ Upstream commit 0cb7365850bacb8c2a9975cae672d65714d8daa1 ]

Devices that lack persistent storage for the device address can indicate
this by setting the HCI_QUIRK_INVALID_BDADDR which causes the controller
to be marked as unconfigured until user space has set a valid address.

Once configured, the device address must be set on every setup for
controllers with HCI_QUIRK_NON_PERSISTENT_SETUP to avoid marking the
controller as unconfigured and requiring the address to be set again.

Fixes: 740011cfe948 ("Bluetooth: Add new quirk for non-persistent setup settings")
Signed-off-by: Johan Hovold <johan+linaro@kernel.org>
Signed-off-by: Luiz Augusto von Dentz <luiz.von.dentz@intel.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

Add MODULE_FIRMWARE() for FIRMWARE_TG357766.

[ Upstream commit 046f753da6143ee16452966915087ec8b0de3c70 ]

Fixes a bug where on the M1 mac mini initramfs-tools fails to
include the necessary firmware into the initrd.

Fixes: c4dab50697ff ("tg3: Download 57766 EEE service patch firmware")
Signed-off-by: Tobias Heider <me@tobhe.de>
Reviewed-by: Michael Chan <michael.chan@broadcom.com>
Link: https://lore.kernel.org/r/ZJt7LKzjdz8+dClx@tobhe.de
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: tag_sja1105: always prefer source port information from INCL_SRCPT

[ Upstream commit c1ae02d876898b1b8ca1e12c6f84d7b406263800 ]

Currently the sja1105 tagging protocol prefers using the source port
information from the VLAN header if that is available, falling back to
the INCL_SRCPT option if it isn't. The VLAN header is available for all
frames except for META frames initiated by the switch (containing RX
timestamps), and thus, the "if (is_link_local)" branch is practically
dead.

The tag_8021q source port identification has become more loose
("imprecise") and will report a plausible rather than exact bridge port,
when under a bridge (be it VLAN-aware or VLAN-unaware). But link-local
traffic always needs to know the precise source port. With incorrect
source port reporting, for example PTP traffic over 2 bridged ports will
all be seen on sockets opened on the first such port, which is incorrect.

Now that the tagging protocol has been changed to make link-local frames
always contain source port information, we can reverse the order of the
checks so that we always give precedence to that information (which is
always precise) in lieu of the tag_8021q VID which is only precise for a
standalone port.

Fixes: d7f9787a763f ("net: dsa: tag_8021q: add support for imprecise RX based on the VBID")
Fixes: 91495f21fcec ("net: dsa: tag_8021q: replace the SVL bridging with VLAN-unaware IVL bridging")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: sja1105: always enable the INCL_SRCPT option

[ Upstream commit b4638af8885af93cd70351081da1909c59342440 ]

Link-local traffic on bridged SJA1105 ports is sometimes tagged by the
hardware with source port information (when the port is under a VLAN
aware bridge).

The tag_8021q source port identification has become more loose
("imprecise") and will report a plausible rather than exact bridge port,
when under a bridge (be it VLAN-aware or VLAN-unaware). But link-local
traffic always needs to know the precise source port.

Modify the driver logic (and therefore: the tagging protocol itself) to
always include the source port information with link-local packets,
regardless of whether the port is standalone, under a VLAN-aware or
VLAN-unaware bridge. This makes it possible for the tagging driver to
give priority to that information over the tag_8021q VLAN header.

The big drawback with INCL_SRCPT is that it makes it impossible to
distinguish between an original MAC DA of 01:80:C2:XX:YY:ZZ and
01:80:C2:AA:BB:ZZ, because the tagger just patches MAC DA bytes 3 and 4
with zeroes. Only if PTP RX timestamping is enabled, the switch will
generate a META follow-up frame containing the RX timestamp and the
original bytes 3 and 4 of the MAC DA. Those will be used to patch up the
original packet. Nonetheless, in the absence of PTP RX timestamping, we
have to live with this limitation, since it is more important to have
the more precise source port information for link-local traffic.

Fixes: d7f9787a763f ("net: dsa: tag_8021q: add support for imprecise RX based on the VBID")
Fixes: 91495f21fcec ("net: dsa: tag_8021q: replace the SVL bridging with VLAN-unaware IVL bridging")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: dsa: felix: don't drop PTP frames with tag_8021q when RX timestamping is disabled

[ Upstream commit 2edcfcbb3c5946609be1d8875473a240b170673b ]

The driver implements a workaround for the fact that it doesn't have an
IRQ source to tell it whether PTP frames are available through the
extraction registers, for those frames to be processed and passed
towards the network stack. That workaround is to configure the switch,
through felix_hwtstamp_set() -> felix_update_trapping_destinations(),
to create two copies of PTP packets: one sent over Ethernet to the DSA
master, and one to be consumed through the aforementioned CPU extraction
queue registers.

The reason why we want PTP packets to be consumed through the CPU
extraction registers in the first place is because we want to see their
hardware RX timestamp. With tag_8021q, that is only visible that way,
and it isn't visible with the copy of the packet that's transmitted over
Ethernet.

The problem with the workaround implementation is that it drops the
packet received over Ethernet, in expectation of its copy being present
in the CPU extraction registers. However, if felix_hwtstamp_set() hasn't
run (aka PTP RX timestamping is disabled), the driver will drop the
original PTP frame and there will be no copy of it in the CPU extraction
registers. So, the network stack will simply not see any PTP frame.

Look at the port's trapping configuration to see whether the driver has
previously enabled the CPU extraction registers. If it hasn't, just
don't RX timestamp the frame and let it be passed up the stack by DSA,
which is perfectly fine.

Fixes: 0a6f17c6ae21 ("net: dsa: tag_ocelot_8021q: add support for PTP timestamping")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: mscc: ocelot: don't keep PTP configuration of all ports in single structure

[ Upstream commit 45d0fcb5bc9558d0bf3d2fa7fabc5d8a88d35439 ]

In a future change, the driver will need to determine whether PTP RX
timestamping is enabled on a port (including whether traps were set up
on that port in particular) and that is currently not possible.

The driver supports different RX filters (L2, L4) and kinds of TX
timestamping (one-step, two-step) on its ports, but it saves all
configuration in a single struct hwtstamp_config that is global to the
switch. So, the latest timestamping configuration on one port
(including a request to disable timestamping) affects what gets reported
for all ports, even though the configuration itself is still individual
to each port.

The port timestamping configurations are only coupled because of the
common structure, so replace the hwtstamp_config with a mask of trapped
protocols saved per port. We also have the ptp_cmd to distinguish
between one-step and two-step PTP timestamping, so with those 2 bits of
information we can fully reconstruct a descriptive struct
hwtstamp_config for each port, during the SIOCGHWTSTAMP ioctl.

Fixes: 4e3b0468e6d7 ("net: mscc: PTP Hardware Clock (PHC) support")
Fixes: 96ca08c05838 ("net: mscc: ocelot: set up traps for PTP packets")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>

net: mscc: ocelot: don't report that RX timestamping is enabled by default

[ Upstream commit 4fd44b82b7aceaa35c2901c6546d2c4198e0799d ]

PTP RX timestamping should be enabled when the user requests it, not by
default. If it is enabled by default, it can be problematic when the
ocelot driver is a DSA master, and it sidesteps what DSA tries to avoid
through __dsa_master_hwtstamp_validate().

Additionally, after the change which made ocelot trap PTP packets only
to the CPU at ocelot_hwtstamp_set() time, it is no longer even true that
RX timestamping is enabled by default, because until ocelot_hwtstamp_set()
is called, the PTP traps are actually not set up. So the rx_filter field
of ocelot->hwtstamp_config reflects an incorrect reality.

Fixes: 96ca08c05838 ("net: mscc: ocelot: set up traps for PTP packets")
Fixes: 4e3b0468e6d7 ("net: mscc: PTP Hardware Clock (PHC) support")
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>