| Age | Commit message (Collapse) | Author |
|---|
|
When an OSS sequencer device is opened, it creates a sequencer port,
then further initializes the internal states like a timer instance,
etc. Since the sequencer port is published, an event can be already
delivered, and this may hit unexpected event handling, which can
eventually cause a NULL dereference (like the timer start/stop event
triggered before the timer initialization).
For avoiding such a race, introduce a simple guard and skips the event
processing until the port is fully initialized.
At releasing the device, in turn, clear this guard, so that any input
events will be discarded during the teardown.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Cc: <stable@vger.kernel.org>
Link: https://patch.msgid.link/20261016161930.482309-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The sync-echo callback needs a write queue before it can wake up a
synchronizing writer. However, snd_seq_oss_open() publishes the session's
port with snd_seq_oss_event_input() as its callback before allocating
and assigning dp->writeq.
A kernel caller using snd_seq_kernel_client_dispatch() with the OSS
client ID can deliver an ECHO containing SEQ_SYNCTIMER to this port
before the write queue is assigned. The helper stamps the supplied
client ID into the event, so the callback's source-client check accepts
it. Regular userspace events carry their own client ID and are rejected.
The failing ordering is:
OSS open Kernel direct dispatch
-------- ----------------------
Zero-allocate dp
Set dp->cseq
Publish callback port
Dispatch ECHO as the OSS client
Look up and pin that port
Enter snd_seq_oss_event_input()
Pass NULL to writeq_wakeup()
Allocate and assign dp->writeq
Dispatch reaches the callback after port publication and before the
queue assignment. OSS register_mutex serializes open and release, but
direct dispatch does not take it. The port use reference keeps dp alive
through the callback; it does not wait for constructor initialization.
The wakeup helper therefore attempts to acquire sync_lock through a
NULL queue pointer, causing a fault.
Sample dp->writeq once in the SEQ_SYNCTIMER branch and ignore the echo
when the queue is absent, matching the adjacent read-queue check. Pass
the checked pointer to the wakeup helper so a NULL sample cannot reach
the lock access. The existing port reference protects the queue's
lifetime through this call, since port deletion drains callbacks before
free_devinfo() frees the queue.
KASAN report as below:
Oops: general protection fault, probably for non-canonical address 0xdffffc000000000d: 0000 [#1] SMP KASAN NOPTI
KASAN: null-ptr-deref in range [0x0000000000000068-0x000000000000006f]
CPU: 2 UID: 0 PID: 69 Comm: kworker/2:1 Tainted: G W 7.2.0-rc6-pmb-bt-functional-v1+ #1 PREEMPT(lazy)
Tainted: [W]=WARN
Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
Workqueue: events pmbd_dispatch_early_sync_echo
RIP: 0010:do_raw_spin_lock+0x71/0x270
Code: d0 60 81 49 8d 14 04 c7 02 f1 f1 f1 f1 c7 42 04 04 f3 f3 f3 65 48 8b 15 05 df 88 09 48 89 54 24 60 31 d2 48 89 fa 48 c1 ea 03 <0f> b6 14 02 48 89 f8 83 e0 07 83 c0 03 38 d0 7c 08 84 d2 0f 85 b9
RSP: 0018:ffff88810450f878 EFLAGS: 00010013
RAX: dffffc0000000000 RBX: 0000000000000068 RCX: 0000000000000000
RDX: 000000000000000d RSI: 0000000000000000 RDI: 000000000000006c
RBP: 0000000000000293 R08: 0000000000000001 R09: 0000000000000000
R10: ffffffff8940fe57 R11: 0000000000000000 R12: 1ffff110208a1f10
R13: 0000000000000068 R14: 0000000000000001 R15: ffff88810450fbee
FS: 0000000000000000(0000) GS:ffff8881fd885000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000006d2890 CR3: 0000000106cbf002 CR4: 0000000000770ef0
PKRU: 55555554
Call Trace:
<TASK>
? pmbd_probe_hit_cookie+0xee/0x1c0
? __pfx_do_raw_spin_lock+0x10/0x10
? lock_acquire+0x2b8/0x2f0
? __pfx_pmbd_probe_hit_cookie+0x10/0x10
_raw_spin_lock_irqsave+0x44/0x60
? snd_seq_oss_writeq_wakeup+0x30/0xf0
snd_seq_oss_writeq_wakeup+0x30/0xf0
snd_seq_oss_event_input+0x1e4/0x2a0
? do_raw_read_unlock+0x49/0xe0
? __pfx_snd_seq_oss_event_input+0x10/0x10
__snd_seq_deliver_single_event+0x105/0x1a0
? snd_seq_port_use_ptr+0x111/0x170
snd_seq_deliver_single_event+0x2fc/0x770
? __pfx_snd_seq_deliver_single_event+0x10/0x10
? __pfx_pmbd_hold_maybe+0x10/0x10
snd_seq_deliver_event+0x45a/0x760
? __pfx_do_raw_spin_lock+0x10/0x10
? lock_acquire+0x2b8/0x2f0
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_snd_seq_deliver_event+0x10/0x10
? _raw_spin_unlock_irqrestore+0x44/0x60
? srso_alias_return_thunk+0x5/0xfbef5
? trace_hardirqs_on+0x18/0x160
? srso_alias_return_thunk+0x5/0xfbef5
? _raw_spin_unlock_irqrestore+0x44/0x60
? srso_alias_return_thunk+0x5/0xfbef5
snd_seq_kernel_client_dispatch+0x19c/0x1e0
pmbd_dispatch_early_sync_echo+0x121/0x1c0
? __pfx_pmbd_dispatch_early_sync_echo+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? finish_task_switch.isra.0+0x1dc/0x980
process_one_work+0x908/0x19c0
? __pfx_process_one_work+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
worker_thread+0x65c/0xe40
? srso_alias_return_thunk+0x5/0xfbef5
? __kthread_parkme+0x175/0x220
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_worker_thread+0x10/0x10
kthread+0x34f/0x460
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_kthread+0x10/0x10
ret_from_fork+0x659/0x940
? __pfx_ret_from_fork+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? __switch_to+0x74f/0xf80
? __pfx_kthread+0x10/0x10
ret_from_fork_asm+0x1a/0x30
</TASK>
Modules linked in:
---[ end trace 0000000000000000 ]---
RIP: 0010:do_raw_spin_lock+0x71/0x270
Code: d0 60 81 49 8d 14 04 c7 02 f1 f1 f1 f1 c7 42 04 04 f3 f3 f3 65 48 8b 15 05 df 88 09 48 89 54 24 60 31 d2 48 89 fa 48 c1 ea 03 <0f> b6 14 02 48 89 f8 83 e0 07 83 c0 03 38 d0 7c 08 84 d2 0f 85 b9
RSP: 0018:ffff88810450f878 EFLAGS: 00010013
RAX: dffffc0000000000 RBX: 0000000000000068 RCX: 0000000000000000
RDX: 000000000000000d RSI: 0000000000000000 RDI: 000000000000006c
RBP: 0000000000000293 R08: 0000000000000001 R09: 0000000000000000
R10: ffffffff8940fe57 R11: 0000000000000000 R12: 1ffff110208a1f10
R13: 0000000000000068 R14: 0000000000000001 R15: ffff88810450fbee
FS: 0000000000000000(0000) GS:ffff8881fd885000(0000) knlGS:0000000000000000
CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033
CR2: 00000000006d2890 CR3: 0000000106cbf002 CR4: 0000000000770ef0
PKRU: 55555554
Kernel panic - not syncing: Fatal exception
Kernel Offset: disabled
Rebooting in 1 seconds..
Assisted-by: LLM
Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
Link: https://patch.msgid.link/pm-pmb-3634251be738461947da-v2-64371e8c8d6cbf6956a3@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Each OSS MIDI alias and its coder must have one final-release owner.
find_slot() pins the alias under register_lock, but
snd_seq_oss_midi_check_exit_port() clears its slot in a separate critical
section without checking whether another callback already removed it.
The use counter protects active users, but draining it does not elect a
single remover.
This requires overlapping raw kernel-client port controls for the same
client. snd_seq_kernel_client_ctl() does not take ioctl_mutex, and
snd_seq_delete_port() releases ports_mutex before returning to the
DELETE_PORT handler's PORT_EXIT broadcast. If the old deletion is delayed
before that broadcast, a new port can reuse the vacant address. Its
PORT_START finds and keeps the old alias. Deleting the replacement then
produces a second PORT_EXIT for the same address. Non-atomic subscriber
delivery holds a read rwsem, so the two receive_announce() callbacks can
run concurrently:
Old-port deletion Replacement-port controls
----------------- -------------------------
Unlink and drain old port
Delay before PORT_EXIT Create port at the same address
PORT_START keeps existing alias
Delete port; PORT_EXIT pins mdev
PORT_EXIT pins the same mdev
Clear slot; drop pin Clear slot; drop pin
Wait for users to drain Wait for users to drain
Free coder; pause before kfree(mdev)
Free coder again: read
freed coder->buf
Both callbacks can select the same alias before either clears its slot.
After both drop their pins, both waits can return. The second call to
snd_midi_event_free() can then read the freed coder's buffer pointer,
as in the report below, and attempt to release the same objects twice.
Check that the pinned alias still occupies its slot under register_lock
before clearing it. A callback that finds it already removed drops its
pin and returns. Only the callback that removes the alias drains users
and frees the coder and record. The pin keeps the record alive through
this check, and the wait remains outside the spinlock.
KASAN report as below:
==================================================================
BUG: KASAN: slab-use-after-free in snd_midi_event_free+0x53/0x60
Read of size 8 at addr ffff888107074518 by task alsa-seq-fixtur/492
CPU: 3 UID: 0 PID: 492 Comm: alsa-seq-fixtur Not tainted 7.2.0-rc6-pmb-bt-functional-v1+ #1 PREEMPT(lazy)
Hardware name: QEMU Ubuntu 24.04 PC v2 (i440FX + PIIX, arch_caps fix, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
Call Trace:
<TASK>
dump_stack_lvl+0x93/0xd0
print_report+0xce/0x630
? snd_midi_event_free+0x53/0x60
? srso_alias_return_thunk+0x5/0xfbef5
? __virt_addr_valid+0x20d/0x410
? snd_midi_event_free+0x53/0x60
kasan_report+0xe0/0x110
? snd_midi_event_free+0x53/0x60
snd_midi_event_free+0x53/0x60
snd_seq_oss_midi_check_exit_port+0x1d8/0x370
receive_announce+0x1b8/0x370
? __pfx_receive_announce+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? __lock_acquire+0x466/0x2260
? snd_seq_port_use_ptr+0x40/0x170
? srso_alias_return_thunk+0x5/0xfbef5
? find_held_lock+0x2b/0x80
? snd_seq_port_use_ptr+0x10c/0x170
? srso_alias_return_thunk+0x5/0xfbef5
? lock_release+0xc8/0x280
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_receive_announce+0x10/0x10
__snd_seq_deliver_single_event+0x105/0x1a0
? snd_seq_port_use_ptr+0x111/0x170
snd_seq_deliver_single_event+0x2fc/0x770
? __pfx_snd_seq_deliver_single_event+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? __pfx_down_read_nested+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? do_raw_read_unlock+0x49/0xe0
? srso_alias_return_thunk+0x5/0xfbef5
? _raw_read_unlock+0x23/0x40
snd_seq_deliver_event+0x351/0x760
? __pfx_snd_seq_deliver_event+0x10/0x10
? _raw_spin_unlock_irqrestore+0x44/0x60
? trace_hardirqs_on+0x18/0x160
? srso_alias_return_thunk+0x5/0xfbef5
? _raw_spin_unlock_irqrestore+0x44/0x60
? srso_alias_return_thunk+0x5/0xfbef5
snd_seq_kernel_client_dispatch+0x13f/0x170
snd_seq_system_broadcast+0x100/0x150
? __pfx_snd_seq_system_broadcast+0x10/0x10
? mark_held_locks+0x40/0x70
? _raw_spin_unlock_irqrestore+0x44/0x60
? srso_alias_return_thunk+0x5/0xfbef5
? lockdep_hardirqs_on_prepare+0xea/0x1a0
? srso_alias_return_thunk+0x5/0xfbef5
snd_seq_ioctl_delete_port+0x179/0x4b0
call_seq_client_ctl+0x90/0x110
snd_seq_kernel_client_ctl+0x8a/0x100
pmbd_case_delete_port+0xa6/0xf0
? __pfx_pmbd_case_delete_port+0x10/0x10
? __might_fault+0xdd/0x190
? srso_alias_return_thunk+0x5/0xfbef5
? __might_fault+0x14e/0x190
? srso_alias_return_thunk+0x5/0xfbef5
? strim+0x19c/0x200
pmbd_case_control_write+0x2c5/0x470
? __pfx_pmbd_case_control_write+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? __debugfs_file_get+0x1bd/0x7f0
? srso_alias_return_thunk+0x5/0xfbef5
? lock_acquire+0x17b/0x2f0
? ksys_write+0x111/0x200
full_proxy_write+0x11f/0x180
vfs_write+0x25a/0x1010
? __pfx_full_proxy_write+0x10/0x10
? __pfx___mutex_lock+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? lock_acquire+0x17b/0x2f0
? __pfx_vfs_write+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? srso_alias_return_thunk+0x5/0xfbef5
? __fget_files+0x1e1/0x370
ksys_write+0x111/0x200
? __pfx_ksys_write+0x10/0x10
? rseq_register.isra.0+0x1f3/0x2e0
? srso_alias_return_thunk+0x5/0xfbef5
? exit_to_user_mode_loop+0x3a5/0x570
? srso_alias_return_thunk+0x5/0xfbef5
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
RIP: 0033:0x7f3532e4e9ee
Code: 08 0f 85 f5 4b ff ff 49 89 fb 48 89 f0 48 89 d7 48 89 ce 4c 89 c2 4d 89 ca 4c 8b 44 24 08 4c 8b 4c 24 10 4c 89 5c 24 08 0f 05 <c3> 66 2e 0f 1f 84 00 00 00 00 00 0f 1f 80 00 00 00 00 48 83 ec 08
RSP: 002b:00007f3532dafe08 EFLAGS: 00000246 ORIG_RAX: 0000000000000001
RAX: ffffffffffffffda RBX: 00007f3532db06c0 RCX: 00007f3532e4e9ee
RDX: 000000000000000a RSI: 000055efbd747156 RDI: 0000000000000003
RBP: 000055efbd747156 R08: 0000000000000000 R09: 0000000000000000
R10: 0000000000000000 R11: 0000000000000246 R12: 000000000000000a
R13: 0000000000000000 R14: 00007ffe7894e7f0 R15: 00007f35325b0000
</TASK>
Allocated by task 490:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
__kasan_kmalloc+0xaa/0xb0
__kmalloc_cache_noprof+0x251/0x610
snd_midi_event_new+0x74/0x210
snd_seq_oss_midi_check_new_port+0x329/0x610
receive_announce+0x277/0x370
__snd_seq_deliver_single_event+0x105/0x1a0
snd_seq_deliver_single_event+0x2fc/0x770
snd_seq_deliver_event+0x351/0x760
snd_seq_kernel_client_dispatch+0x13f/0x170
snd_seq_system_broadcast+0x100/0x150
snd_seq_ioctl_create_port+0x39a/0x9d0
call_seq_client_ctl+0x90/0x110
snd_seq_kernel_client_ctl+0x8a/0x100
pmbd_case_create_port+0xcc/0x160
pmbd_case_control_write+0x21d/0x470
full_proxy_write+0x11f/0x180
vfs_write+0x25a/0x1010
ksys_write+0x111/0x200
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Freed by task 493:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
kasan_save_free_info+0x3b/0x60
__kasan_slab_free+0x5f/0x80
kfree+0x307/0x580
snd_midi_event_free+0x48/0x60
snd_seq_oss_midi_check_exit_port+0x1d8/0x370
receive_announce+0x1b8/0x370
__snd_seq_deliver_single_event+0x105/0x1a0
snd_seq_deliver_single_event+0x2fc/0x770
snd_seq_deliver_event+0x351/0x760
snd_seq_kernel_client_dispatch+0x13f/0x170
snd_seq_system_broadcast+0x100/0x150
snd_seq_ioctl_delete_port+0x179/0x4b0
call_seq_client_ctl+0x90/0x110
snd_seq_kernel_client_ctl+0x8a/0x100
pmbd_case_delete_port+0xa6/0xf0
pmbd_case_control_write+0x398/0x470
full_proxy_write+0x11f/0x180
vfs_write+0x25a/0x1010
ksys_write+0x111/0x200
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
The buggy address belongs to the object at ffff888107074500
which belongs to the cache kmalloc-96 of size 96
The buggy address is located 24 bytes inside of
freed 96-byte region [ffff888107074500, ffff888107074560)
The buggy address belongs to the physical page:
page: refcount:0 mapcount:0 mapping:0000000000000000 index:0x0 pfn:0x107074
flags: 0x200000000000000(node=0|zone=2)
page_type: f5(slab)
raw: 0200000000000000 ffff888100042280 dead000000000122 0000000000000000
raw: 0000000000000000 0000000000200020 00000000f5000000 0000000000000000
page dumped because: kasan: bad access detected
Memory state around the buggy address:
ffff888107074400: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
ffff888107074480: fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc fc
>ffff888107074500: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
^
ffff888107074580: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
ffff888107074600: fa fb fb fb fb fb fb fb fb fb fb fb fc fc fc fc
==================================================================
Assisted-by: LLM
Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
Link: https://patch.msgid.link/pm-pmb-7c5128efe2110a87b872-v2-c718183fd693e73d8e95@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Since the port list was converted to RCU, snd_seq_delete_all_ports()
calls synchronize_rcu() once per port under register_mutex. Closing a
client with 254 ports now takes about 1.6s and every other client
open or close waits for it.
Unlink all ports first, wait for a single grace period and then free
them.
Fixes: 4c252fbc06d6 ("ALSA: seq: Use RCU for the client port list")
Signed-off-by: Ali Ahmet Memiş <aliamemis@disroot.org>
Link: https://patch.msgid.link/20261010222543.1470723-1-aliamemis@disroot.org
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_read() returns an error even if one or more events have been
successfully read before the error (except for -EAGAIN case), but this
rather doesn't follow the POSIX standard that wants the processed
bytes. Change the behavior to return the processed bytes when it's
positive, instead.
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20261007172533.14667-8-tiwai@suse.de
|
|
I forgot to cover the two new fields, direction and ump_group, in
struct snd_seq_port_info for the 32bit compat ioctls of
SNDRV_SEQ_IOCTL_GET_PORT_INFO & co, which ended up with the garbage
copies in those fields. Add the handling of those two fields in the
compat layer.
Also, avoid over-copying the 64bit snd_seq_port_info but only the first
part (before the field kernel); this avoids the copy of kernel-space
garbage value in the reserved field.
Fixes: ff166a9d19fa ("ALSA: seq: Add port direction to snd_seq_port_info")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20261007172533.14667-2-tiwai@suse.de
|
|
In the recent refactoring with RCU, clientptr() takes guard(rcu)()
around the client table lookup, but the RCU read-side section ends as
soon as the function returns, so the returned pointer isn't protected
at all. This gives a false impression as if that the callers were
safe, and confuse reviewers including Sashiko.
Actually, the callers (snd_seq_delete_kernel_client(),
snd_seq_kernel_client_ctl() and snd_seq_kernel_client_write_poll())
never relied on any lock; the caller is the owner of the client (a
kernel client passing its own id, or a user client via its opened
file), hence the client can't be released concurrently by others.
So just drop the confusing and useless RCU guard, read the table via
rcu_dereference_protected(), and document the lifetime rule. Along
with it, fold __clientptr() into its only user client_use_ptr(); the
id range is already checked there, and it's never called with
clients_lock held, so a plain rcu_dereference() suffices.
Fixes: 7a287e4615d6 ("ALSA: seq: Use RCU for the client table")
Reported-by: Sashiko <sashiko-bot@kernel.org>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20261007172051.13240-2-tiwai@suse.de
|
|
Pull 7.3 devel branch for applying the further HD-audio quirks more
cleanly.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
OSS sequencer API emulation has been implemented as a client of ALSA
sequencer, decades ago, at the time where OSS was still use. But the
API itself became very niche, and almost noone using it for a real
purpose any longer -- only exceptions are AI and fuzzers, who have
been causing significant maintenance burden for issues they hit &
found.
Meanwhile, the most features of OSS sequencer API emulation can be
achieved in user-space, either by ALSA-OSS layer (aoss) or oss-seq-emu
program:
https://github.com/alsa-project/alsa-oss
https://github.com/tiwai/oss-seq-emu
where the former is LD_PRELOAD wrapper and the latter is a CUSE-based
daemon. So we can drop the stuff from the kernel sooner or later.
Now, let's mark the OSS sequencer emulation as deprecated and give a
warning when loaded, so that we can drop it later actually.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260912124951.416761-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The native sequencer ioctl path serializes handler calls with
client->ioctl_mutex, but the translated port-info compat path invokes the
same handlers through snd_seq_kernel_client_ctl() without taking that
mutex.
This lets concurrent compat CREATE_PORT requests pass the port-count
check before any request reaches the serialized insertion. The computed
integer port index can then exceed the address field range and wrap to an
existing index. Subsequent subscriber teardown can resolve the duplicate
address to the wrong port and access a freed subscriber.
Take ioctl_mutex while dispatching converted port-info requests, matching
the native ioctl path. All translated port-info commands share this
helper, so their accesses to the client port state are serialized as
well.
Fixes: b3defb791b26 ("ALSA: seq: Make ioctls race-free")
Cc: stable@vger.kernel.org
Assisted-by: Codex:gpt-5.6-sol gpt-6-astra
Signed-off-by: Kyle Zeng <kylebot@openai.com>
Signed-off-by: Bruno Produit <bruno.produit@trailofbits.com>
Link: https://patch.msgid.link/20260917105643.90102-1-bruno.produit@trailofbits.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Correct spelling mistakes in comments. No functional change.
Signed-off-by: Hemanth Selam <hemanth.selam@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260915084424.1007757-2-hemanth.selam@gmail.com
|
|
This is another run of the Coccinelle script for converting kmalloc()
family of allocations to kmalloc_obj() via the existing rules in
scripts/coccinelle/api/kmalloc_objs.cocci
This catches both the set of kmalloc() uses added since the first
kmalloc_obj() conversions in v7.0 and adds a large group missed in the
first pass due to Coccinelle not interacting well with the cleanup.h
scoped_...() family of macros[1]. I worked around this with spatch's
"--macro-file" argument to a file with all the scoped_...() macros mapped
to Coccinelle's YACFE_ITERATOR[2] as that was the closest viable control
flow indicator I could find.
Build tested allmodconfig on x86, arm64, arm, loongarch, mips, powerpc,
riscv, and s390 with no new warnings.
Link: https://lore.kernel.org/lkml/202609021314.8A9C0B8@keescook/ [1]
Link: https://github.com/coccinelle/coccinelle/blob/master/standard.h [2]
Signed-off-by: Kees Cook <kees+treewide@kernel.org>
|
|
snd_midi_input_event() must not be running while a rawmidi substream is
closing, since this can lead to the trigger state becoming out-of-step
through this sequence in snd_rawmidi_input_trigger():
snd_rawmidi_input_trigger(up=0)
snd_midi_input_event()
-> snd_rawmidi_kernel_read()
-> snd_rawmidi_input_trigger(up=1)
-> cancel_work_sync()
which ends with the underlying device being active unexpectedly.
When this is called from close_substream(), further input can re-trigger
the input event leaving it running after rawmidi_release_priv() has set
rfile->rmidi to NULL which leads to:
Unable to handle kernel NULL pointer dereference at virtual address 00000000000000b0
Call trace:
snd_midi_input_event+0x3c/0x134 [snd_seq_midi] (P)
snd_rawmidi_input_event_work+0x1c/0x2c
process_one_work+0x150/0x3a4
worker_thread+0x190/0x318
Apply a similar approach to commit ef7607ab1c8ad ("ALSA: seq: midi:
Serialize output teardown with event_input") which fixed the same issue
in the output direction, but updated to use RCU following Takashi Iwai's
proposed follow-on patch [1].
With this change in place, midisynth_unsubscribe() clears the input file
so snd_midi_input_event() will not re-trigger the stream and will be
quiesced by the cancel_work_sync() in snd_rawmidi_input_trigger().
[1] https://lore.kernel.org/linux-sound/20260813144224.753399-1-tiwai@suse.de/
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: John Keeping <jkeeping@inmusicbrands.com>
Link: https://patch.msgid.link/20260813150810.795393-1-jkeeping@inmusicbrands.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The recent fix for serializing the output teardown introduced a
spinlock invocation at every MIDI output event via event_process_midi.
Since this is a hot path, let's do performance optimization with RCU.
The new output_substream __rcu pointer is published via
rcu_assign_pointer() in midisynth_use() after output_rfile is set, and
cleared in midisynth_unuse() before the resource teardown.
event_process_midi() reads it under rcu_read_lock() and bumps
output_use_lock inside that section, which is necessary to close the
window between the pointer dereference and the refcount increment.
midisynth_unuse() calls synchronize_rcu() before snd_use_lock_sync():
this guarantees that any reader who obtained a non-NULL pointer has
already called atomic_inc (output_use_lock), so the subsequent
snd_use_lock_sync() sees the correct in-flight count.
Fixes: ef7607ab1c8a ("ALSA: seq: midi: Serialize output teardown with event_input")
Link: https://patch.msgid.link/20260813144224.753399-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The struct describes a bounce payload of an error code followed by the
original event and its external data. No kernel has ever sent that.
Before commit efc86691e4d8 ("ALSA: seq: Fix kernel heap address leak in
bounce_error_event()") the kernel emitted no SNDRV_SEQ_EVENT_BOUNCE at
all, and since then it sends the event record alone.
Nothing has ever read it either. Its only accessor,
snd_seq_event_bounce_ext_data(), has had no caller for the whole git
history, and it did not even compile until commit c7e0b5bf9fff ("[ALSA]
Remove xxx_t typedefs: Sequencer") incidentally repaired the type name
it referred to, three years after the git import. Drop the accessor
along with the struct.
This removes a definition from a UAPI header. Since no kernel ever
produced the layout, nothing can have parsed it, but a program that
merely names the type will need to stop.
Suggested-by: Takashi Iwai <tiwai@suse.de>
Assisted-by: Claude:claude-opus-5
Signed-off-by: HyeongJun An <sammiee5311@gmail.com>
Link: https://patch.msgid.link/20260812141506.4016387-1-sammiee5311@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The bounce_error_event() embeds the failed event in the bounce payload
by pointing data.ext.ptr at it. When that event is a queued
variable-length event, its own data.ext.ptr holds the address of its
first extension cell, put there by snd_seq_event_dup(). The payload
goes out verbatim through snd_seq_expand_var_event(), so the address
reaches userspace.
That is the same address commit 705dd6dcbc0e ("ALSA: seq: Clear
variable event pointer on read") removed from the event header. The
read path still clears it there, just above the call that expands the
payload.
Embed a sanitised copy instead, treated exactly as snd_seq_read()
treats the header. A stack copy is enough because delivery is
synchronous and snd_seq_event_dup() copies before returning.
An unprivileged client reaches this by setting SNDRV_SEQ_FILTER_BOUNCE,
queueing a variable-length event to a port that does not exist and
reading the bounce back. Eight bytes on 64-bit, from its own pool.
Fixes: efc86691e4d8 ("ALSA: seq: Fix kernel heap address leak in bounce_error_event()")
Assisted-by: Claude:claude-opus-5
Signed-off-by: HyeongJun An <sammiee5311@gmail.com>
Link: https://patch.msgid.link/20260811131835.3837024-1-sammiee5311@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The UMP sequencer client protects its output rawmidi file (out_rfile)
with an rwlock (output_lock). seq_ump_process_event(), the port's
event_input callback, reads out_rfile.output under read_lock on every
delivered UMP event, while the open/close paths (serialized by
ump->open_mutex) publish and clear out_rfile under write_lock.
Output is opened/closed only on the subscribe/use lifecycle while
delivery happens per event, so this is another read-mostly hot path.
Convert it to RCU and drop the rwlock. out_rfile is an embedded struct
rather than a pointer, so instead of restructuring it, add an
RCU-protected shadow of the substream (out_substream) for the reader;
out_rfile itself becomes writer-only state accessed solely under
open_mutex. The reader now runs lock-free under rcu_read_lock() via
rcu_dereference(), and open publishes the substream with
rcu_assign_pointer().
On close the substream is cleared with rcu_assign_pointer(NULL) and the
rawmidi is released only after synchronize_rcu(), so no reader in the
delivery path can still be writing to the substream when
snd_rawmidi_kernel_release() runs.
Dropping write_lock_irqsave() from the writers is safe: they run in
process context under open_mutex, and the sole atomic reader now uses
RCU, which is IRQ-safe.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260810133711.42483-6-tiwai@suse.de
|
|
Each virmidi device keeps a list of its opened input files (filelist)
protected by both an rwlock (filelist_lock) and a rw_semaphore
(filelist_sem). snd_virmidi_dev_receive_event() walks the list on the
sequencer event input path -- read_lock() when the event is delivered in
atomic context, down_read() otherwise -- decoding each incoming event
into the file's rawmidi buffer. The writers (input open/close) take both
locks to add/remove entries.
This is another typical dual-lock read-mostly pattern as the port
subscriber list: files are opened/closed rarely while the receive
callback runs per event. Let's convert the traversal to RCU and drop
the rwlock; the existing filelist_sem keeps serializing the writers.
The atomic input path now runs lock-free under rcu_read_lock(), and
both readers share a single list_for_each_entry_rcu() (valid under the
rwsem via lockdep_is_held()). The writers switch to
list_add_tail_rcu() / list_del_rcu().
snd_virmidi_input_close() freed the entry (parser and struct)
immediately after list_del. A concurrent lockless reader in the atomic
path may still be dereferencing it, so the close path now waits for an
RCU grace period after list_del_rcu() before freeing; synchronize_rcu()
is used rather than kfree_rcu() because the parser must also be released
after the grace period, not just the struct. Non-atomic readers are
already excluded by the down_write, so only the atomic RCU readers need
the grace period.
Dropping write_lock_irq() from the writers is safe: no writer runs in
atomic/IRQ context, and the sole atomic reader now uses RCU, which is
IRQ-safe.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260810133711.42483-5-tiwai@suse.de
|
|
The sequencer keeps a global table of clients (clienttab[]) indexed by
client id, protected by the global clients_lock spinlock. The lookup
snd_seq_client_use_ptr() reads a slot and takes a use_lock reference on
the client, and this runs on the event delivery hot path: every
dispatched event resolves its destination (and often source) client
through it. The spinlock's only job on the read side is to make the
"pointer is non-NULL" test and the reference increment indivisible with
respect to the writer that nulls the slot and then drains the refcount.
Clients come and go rarely but delivery happens constantly, so this is
yet another read-mostly case as the port and subscriber lists.
Convert the table to RCU: the read side now runs lock-free under
rcu_read_lock() and takes the use_lock reference via
rcu_dereference(), removing contention on the single global spinlock
from the delivery path. The writers keep clients_lock (still needed
to serialize slot allocation) and publish / unpublish via
rcu_assign_pointer(); creation and destruction remain serialized at a
higher level by register_mutex.
As with the ports, the client is not freed via kfree_rcu(): its lifetime
is governed by the use_lock refcount drained in seq_free_client1().
list_del under the old spinlock excluded a concurrent lookup from taking
a new reference once the slot was nulled; rcu_assign_pointer(NULL) offers
no such exclusion, so a reader still holding the old pointer can grab a
reference after the unpublish. seq_free_client1() therefore calls
synchronize_rcu() after nulling the slot and before snd_use_lock_sync():
once the grace period elapses no new reference can appear, and the
existing drain then frees the client safely.
clienttablock[] keeps its slot-reservation role (create/free are
serialized by register_mutex); its read on the lookup path only gates
module autoload, so a lockless read is harmless. Dropping the spinlock
from the read path is safe: clients_lock is now taken only by the
process-context writers, and the sole atomic reader uses RCU, which is
IRQ-safe.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260810133711.42483-4-tiwai@suse.de
|
|
Each sequencer client keeps a list of its ports (ports_list_head)
protected by both an rwlock (ports_lock) and a mutex (ports_mutex).
The rwlock is taken read-side on the event delivery hot path:
snd_seq_port_use_ptr() walks the list to resolve a port on every
dispatched event, while the mutex serializes port creation/deletion.
Ports change rarely but delivery happens constantly, so this is the
another read-mostly case as the port subscriber list. Convert the
port list traversal to RCU and drop the rwlock entirely; the existing
ports_mutex keeps serializing the writers. The atomic delivery path
(snd_seq_port_use_ptr(), snd_seq_port_query_nearest()) now runs
lock-free under rcu_read_lock() instead of contending on the shared
rwlock.
The writers switch to list_add_tail_rcu()/list_del_rcu().
snd_seq_insert_port() now stores the port number and name before
publishing the node so RCU readers only ever observe a fully
initialized port. One drawback is that snd_seq_delete_all_ports()
drops the O(1) splice trick and unlinks each port individually,
though: the splice repointed the last port's ->next away from the list
head, which would send a concurrent lockless reader off the end of the
list.
Unlike the subscriber objects, ports are not freed via kfree_rcu():
port_delete() must drain outstanding use_lock references (and run
private_free()) synchronously. The rwlock previously guaranteed that
no reader could take a new use_lock reference once the port was
unlinked -- list_del under write_lock excluded snd_use_lock_use()
under read_lock. list_del_rcu() offers no such exclusion, so a reader
still traversing the list can grab a reference after the unlink.
port_delete() therefore calls synchronize_rcu() after the port has
been unlinked and before snd_use_lock_sync(): once the grace period
elapses no new reference can appear, and the existing drain then frees
the port safely.
Dropping write_lock_irq() from the writers is safe: no writer runs in
atomic/IRQ context, and the sole atomic reader now uses RCU, which is
IRQ-safe.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260810133711.42483-3-tiwai@suse.de
|
|
Each sequencer port keeps two subscriber groups (c_src and c_dest),
each protected by both an rwlock (list_lock) and a rw_semaphore
(list_mutex). The rwlock is taken read-side in the event delivery hot
path (__deliver_to_subscribers()) for delivering every event to
subscribers, while the mutex serializes subscribe/unsubscribe and
covers the sleepable delivery and query walks.
Subscriptions change rarely but delivery happens constantly, so this is
a textbook read-mostly case. Convert the subscriber list traversal to
RCU and drop the rwlock entirely while keeping the existing list_mutex
for serializing the writers. The atomic delivery path now runs
lock-free under rcu_read_lock() instead of contending on the shared
rwlock.
Along with the conversion to RCU, the subscriber lists are switched
from list_head to hlist so that removal can use hlist_del_init_rcu():
it keeps the ->next pointer intact for concurrent readers while
clearing ->pprev, which lets the double-deletion guard (added in
commit 13d5e5d4725c) keep detecting an already-removed entry via
hlist_unhashed().
Dropping write_lock_irq() from the writers is safe: no writer runs in
atomic/IRQ context, and the sole atomic reader now uses RCU, which is
IRQ-safe. Port lifetime handling (use_lock/closing drain in
port_delete()) is orthogonal and unchanged.
Note that the conversion to RCU has another merit: it automatically
"fixes" the (rather false) lockdep warnings for the doubly read-locks
of the same subscriber list, too.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260810133711.42483-2-tiwai@suse.de
|
|
Pull 7.2 devel branch for applying further patches cleanly.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
A userspace-driven ALSA timer (SND_UTIMER) lets an unprivileged user set
the backing snd_timer's hardware resolution to an arbitrary 64-bit value
via SNDRV_TIMER_IOCTL_CREATE. snd_utimer_create() only rejects zero.
When such a timer is bound to a sequencer queue, initialize_timer()
computes the tick period as
tmr->ticks = 1000000000 / (r * freq);
where r is that user-controlled resolution and freq is the sequencer
update rate in Hz, clamped to MIN_FREQUENCY..MAX_FREQUENCY (10..6250).
A resolution of 2^63 makes the 64-bit product r * freq wrap to zero for
any even freq, including DEFAULT_FREQUENCY (1000), so the division faults
with a divide-by-zero.
The division runs under tmr->lock with interrupts disabled, so the oops
leaves the spinlock held and hangs the CPU. It is reachable by an
unprivileged user with access to /dev/snd/timer and /dev/snd/seq.
Oops: divide error: 0000 [#1] SMP KASAN PTI
CPU: 7 UID: 1000 PID: 456 Comm: alsa_seq_utimer Not tainted 7.2.0-rc4+
RIP: 0010:initialize_timer.constprop.0+0x20a/0x2d0
snd_seq_timer_start+0x15e/0x2b0
snd_seq_control_queue+0x56f/0xba0
snd_seq_write+0x3e0/0x730
Reject an overflowing product with check_mul_overflow() and fall back to
a single tick, which also avoids feeding a wrapped-but-nonzero divisor
(e.g. 2^63 * 1000 mod 2^64 == 0, or other resolutions wrapping to a small
value) into the period computation.
Fixes: 37745918e0e7 ("ALSA: timer: Introduce virtual userspace-driven timers")
Cc: <stable@vger.kernel.org>
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Norbert Szetei <norbert@doyensec.com>
Link: https://patch.msgid.link/DF8A3844-AD5E-4B8A-9CFC-BD83C212BA38@doyensec.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
queue_delete() closes the queue timer, then frees it. snd_seq_timer_close()
clears q->timer->timeri. snd_use_lock_sync() then drains borrowers, and
snd_seq_timer_delete() frees q->timer.
A borrower can re-open the timer inside that window. A SET_QUEUE_CLIENT
that took a queueptr() use_lock reference before the queue was unlinked
runs snd_seq_timer_open() after the close. Open refuses re-open only while
timeri is set, and the close just cleared it, so it re-opens timeri.
snd_seq_timer_delete() does not close that instance. Its snd_seq_timer_stop()
is a no-op, because running was cleared first. So it frees q->timer with the
instance still live. The queue is freed next.
The instance stays on the global timer with callback_data pointing at the
freed queue. A non-owner START on the unlocked queue arms it. The next tick
derefs the freed queue in snd_seq_timer_interrupt().
Reachable by an unprivileged user with access to /dev/snd/seq. No CAP and
no queue ownership required.
Close any lingering instance in the destructor. There, ->timeri can no
longer change: the queue is unlinked and all use_lock borrowers have
drained, so no snd_seq_queue_use() can re-open it. Close it before clearing
q->timer. snd_timer_close() waits for any in-flight snd_seq_timer_interrupt()
to finish, and that callback still reads q->timer (via snd_seq_check_queue()),
so q->timer must stay valid until it drains.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: Norbert Szetei <norbert@doyensec.com>
Link: https://patch.msgid.link/422FDB81-2A68-47C7-A22D-2D3301E2E86D@doyensec.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Align file_operations initializers with tabs to match the standard
kernel coding style and improve readability.
Signed-off-by: bui duc phuc <phucduc.bui@gmail.com>
Link: https://patch.msgid.link/20260708140624.562403-1-phucduc.bui@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The previous conversion from nonseekable_open() to stream_open() retained
the existing error check.
Since stream_open() always returns 0, remove the dead error handling path.
Signed-off-by: bui duc phuc <phucduc.bui@gmail.com>
Reviewed-by: Cezary Rojewski <cezary.rojewski@intel.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260702084445.519669-7-phucduc.bui@gmail.com
|
|
snd_seq_event_dup() copies an incoming event into a pool cell and, in
the UMP-enabled build, clears the trailing cell->ump.raw.extra word that
the memcpy() did not cover. The guard deciding whether to clear it
compares the copied size against sizeof(cell->event):
memcpy(&cell->ump, event, size);
if (size < sizeof(cell->event))
cell->ump.raw.extra = 0;
For a legacy (non-UMP) event, size == sizeof(struct snd_seq_event) ==
sizeof(cell->event), so the condition is false and the extra word keeps
stale data. The cell pool is allocated with kvmalloc() (not zeroed) and
cells are reused via a free list, so that word holds uninitialised heap
or leftover event data.
When such a cell is delivered to a UMP client (client->midi_version > 0)
that set SNDRV_SEQ_FILTER_NO_CONVERT -- so the legacy event reaches it
unconverted -- snd_seq_read() reads it out as the larger struct
snd_seq_ump_event and copies the stale word to user space, a 4-byte
kernel heap infoleak to an unprivileged /dev/snd/seq client.
Compare against sizeof(cell->ump) instead, so the trailing word is zeroed
for every event shorter than the UMP cell.
Fixes: 46397622a3fa ("ALSA: seq: Add UMP support")
Cc: stable@vger.kernel.org
Assisted-by: Claude:claude-opus-4-8
Signed-off-by: HyeongJun An <sammiee5311@gmail.com>
Link: https://patch.msgid.link/20260623233841.853326-1-sammiee5311@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_fifo_resize() still needs to publish the replacement pool
before it waits for FIFO users. A blocking snd_seq_read() holds
f->use_lock while it sleeps, so concurrent senders must be able to
queue to the new pool and wake that reader instead of failing against a
closing old pool.
However, snd_seq_fifo_event_in() duplicates an event before it takes
f->lock, and snd_seq_read() can dequeue a cell and later call
snd_seq_fifo_cell_putback() if copy_to_user() or
snd_seq_expand_var_event() fails. If resize swaps f->pool and detaches
oldhead in between, either path can relink an old-pool cell after the
snapshot. That stale cell sits outside the drained oldhead list, keeps
oldpool->counter elevated, and can leave snd_seq_pool_delete() waiting
for the retired pool to drain.
Keep the existing swap-before-wait ordering in snd_seq_fifo_resize(),
but reject stale cells before any FIFO relink. Revalidate event-in cells
under f->lock and retry them against the published replacement pool, and
free stale putback cells instead of linking them back into the FIFO.
The buggy scenario involves two paths, with each column showing the
order within that path:
resize path: relink path:
1. Allocate newpool. 1. Take f->use_lock.
2. Swap f->pool to newpool and 2. Duplicate or dequeue an old-pool
detach oldhead. cell before oldpool closes.
3. Mark oldpool closing and 3. Reach a later relink point after
wait for FIFO users. resize published newpool.
4. Free oldhead and delete 4. Relink the old-pool cell after
oldpool. resize detached oldhead.
5. Drop f->use_lock.
The reproducer reports a resize ioctl blocked in the expected pool
teardown path:
signal: resize iteration=98 target_pool=4 exceeded 250ms
(elapsed=251ms)
diagnostic: resize_tid=651 wchan=snd_seq_pool_done
diagnostic: resize_tid=651 stack=
snd_seq_pool_done+0x5b/0x140
snd_seq_pool_delete+0x7a/0x90
snd_seq_fifo_resize+0x193/0x1e0
snd_seq_ioctl_set_client_pool+0x214/0x260
snd_seq_ioctl+0x119/0x540
__x64_sys_ioctl+0xd1/0x120
do_syscall_64+0xbb/0x2f0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
A second run with larger pools hit the same target path:
signal: resize iteration=32 target_pool=64 exceeded 250ms
(elapsed=251ms)
diagnostic: resize_tid=663 wchan=snd_seq_pool_done
diagnostic: resize_tid=663 stack=
snd_seq_pool_done+0x5b/0x140
snd_seq_pool_delete+0x7a/0x90
snd_seq_fifo_resize+0x193/0x1e0
snd_seq_ioctl_set_client_pool+0x214/0x260
snd_seq_ioctl+0x119/0x540
__x64_sys_ioctl+0xd1/0x120
do_syscall_64+0xbb/0x2f0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Fixes: 2d7d54002e39 ("ALSA: seq: Fix race during FIFO resize")
Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
Link: https://patch.msgid.link/20260614004801.3507773-2-zzzccc427@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_oss_readq_clear() resets qlen, head, and tail without
q->lock even though the normal reader and producer paths serialize the
same ring state under that spinlock. A reset can therefore race
snd_seq_oss_readq_free() or snd_seq_oss_readq_put_event() and leave
stale records in the queue, drop freshly queued ones, or report the
wrong readiness after wakeup. KCSAN reports a data race between
snd_seq_oss_readq_clear() and snd_seq_oss_readq_free().
Take q->lock while clearing the ring and resetting input_time. Factor
the enqueue logic into a caller-locked helper so
snd_seq_oss_readq_put_timestamp() updates its suppression state under
the same lock instead of racing the reset path.
The buggy scenario involves two paths, with each column showing the
order within that path:
reset path: locked readq updater:
1. snd_seq_oss_reset() or 1. A reader or callback producer
release reaches takes q->lock on the same queue.
snd_seq_oss_readq_clear().
2. snd_seq_oss_readq_clear() 2. The updater tests or modifies
resets qlen, head, tail, qlen, head, and tail.
and input_time.
3. snd_seq_oss_readq_clear() 3. The updater completes its
wakes sleepers on read-modify-write sequence.
q->midi_sleep.
4. Without q->lock, the reset 4. The resulting ring state drives
can overlap the locked later reads and readiness.
update.
KCSAN reports:
BUG: KCSAN: data-race in snd_seq_oss_readq_clear /
snd_seq_oss_readq_free
write to 0xffff8881069fe608 of 4 bytes by task 120516 on cpu 0:
snd_seq_oss_readq_free+0x6c/0x80
snd_seq_oss_read+0xcb/0x250
odev_read+0x38/0x60
vfs_read+0xff/0x600
ksys_read+0xb4/0x140
__x64_sys_read+0x46/0x60
do_syscall_64+0xbb/0x2f0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
read to 0xffff8881069fe608 of 4 bytes by task 120517 on cpu 1:
snd_seq_oss_readq_clear+0x1f/0x90
snd_seq_oss_reset+0xa7/0xf0
snd_seq_oss_ioctl+0x6f6/0x7e0
odev_ioctl+0x56/0xc0
__x64_sys_ioctl+0xd1/0x120
do_syscall_64+0xbb/0x2f0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
value changed: 0x00000001 -> 0x00000000
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Cen Zhang <zzzccc427@gmail.com>
Link: https://patch.msgid.link/20260614004801.3507773-1-zzzccc427@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The error bouncing may fail again, and we have no check for
re-bouncing. For avoiding the loop, add the event type check at
bouncing, and stop re-bouncing if it's already a bounce error.
Link: https://patch.msgid.link/20260612113350.407465-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The comment above bounce_error_event() documents that user clients
should receive SNDRV_SEQ_EVENT_BOUNCE with the original event embedded
as variable-length data, while kernel clients should receive
SNDRV_SEQ_EVENT_KERNEL_ERROR with a quoted kernel pointer.
However, the implementation unconditionally uses
SNDRV_SEQ_EVENT_KERNEL_ERROR with data.quote.event set to the raw
struct snd_seq_event pointer for all clients. When a bounce error
event is delivered to a USER_CLIENT via snd_seq_read(), the kernel
heap address in data.quote.event is exposed to userspace through
copy_to_user() in the fixed-length branch.
This is a distinct leak path from the one addressed by commit
705dd6dcbc0e ("ALSA: seq: Clear variable event pointer on read"),
which sanitizes data.ext.ptr in the variable-length branch of
snd_seq_read(). The bounce_error_event() leak uses fixed-length
events that take the else branch where no sanitization occurs.
Differentiate the bounce event by client type. For USER_CLIENT,
send SNDRV_SEQ_EVENT_BOUNCE with SNDRV_SEQ_EVENT_LENGTH_VARIABLE
and data.ext pointing to the original event. The variable-length
path in snd_seq_event_dup() copies the event data into chained
cells, and snd_seq_expand_var_event() copies only the content --
never the pointer -- to userspace. For KERNEL_CLIENT, keep the
existing SNDRV_SEQ_EVENT_KERNEL_ERROR behavior with the quoted
pointer.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: HanQuan <eilaimemedsnaimel@gmail.com>
Link: https://patch.msgid.link/20260612103222.2528305-1-eilaimemedsnaimel@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
So far we've tried to address UAFs in ALSA timer code by applying the
locks at various places, but the fundamental problem is that the timer
object may be released while the belonging timer instance objects are
still present and accessing to it. This patch is a more proper fix to
address that issue, namely, by refcounting and keeping the timer
object.
The basic implementation is to use kref for the refcount of the timer
object, and take/release the reference at assigning/releasing the
instance, as well as at referring from ioctls or ALSA sequencer code.
The reference from ioctl or ALSA sequencer is abstracted with
snd_timeri_timer auto-cleanup.
Note that this change assumes that the code already took the fix
commit da3039e91d1f ("ALSA: timer: Forcibly close timer instances at
closing"); otherwise the refcount may be unbalanced when the timer is
freed while slave instances are still present.
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260609115100.806869-2-tiwai@suse.de
|
|
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_read() copies a queued variable-length event header to userspace
before expanding the payload. Queued variable-length events use
SNDRV_SEQ_EXT_CHAINED internally, and data.ext.ptr points at the first
extension cell.
The read side strips SNDRV_SEQ_EXT_* bits from data.ext.len before the
copy, but it leaves data.ext.ptr untouched. A userspace sequencer client
can therefore write a direct variable event to itself and read back the
extension-cell kernel address from the returned header.
Clear the temporary header pointer before copy_to_user(). The original
queued event remains unchanged and is still passed to
snd_seq_expand_var_event(), so payload expansion keeps using the
internal chain.
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Kyle Zeng <kylebot@openai.com>
Link: https://patch.msgid.link/20260607004129.61345-1-kylebot@openai.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_expand_var_event_at() clamps the number of bytes to copy to the
remaining variable-event length, but passes the original buffer size to
expand_var_event().
For SNDRV_SEQ_EXT_USRPTR events, expand_var_event() copies exactly the
size argument from userspace. On the final chunk, when the remaining
event data is shorter than the caller's buffer, this can read past the
declared event data and can spuriously fail with -EFAULT if the extra
bytes cross an unmapped page.
Pass the clamped length instead. The chained and kernel-backed paths
already reclamp in dump_var_event(), but the user-pointer path handles
the size directly.
Fixes: ea46f79709b6 ("ALSA: seq: Add snd_seq_expand_var_event_at() helper")
Signed-off-by: HyeongJun An <sammiee5311@gmail.com>
Link: https://patch.msgid.link/20260606040913.230213-1-sammiee5311@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The dummy sequencer port forwards events by copying an incoming
struct snd_seq_event into a stack temporary, rewriting source and
destination, and dispatching the temporary to subscribers. That legacy
event storage is smaller than struct snd_seq_ump_event.
When a UMP event reaches the dummy client, the copy leaves the UMP flag
set but only provides legacy-sized stack storage. The subscriber
delivery path then uses snd_seq_event_packet_size() and copies a
UMP-sized packet from that stack object, reading past the end of the
temporary.
Use the existing union __snd_seq_event storage and copy the packet size
reported for the incoming event before rewriting the common routing
fields. This preserves the full UMP packet for UMP events while keeping
legacy event handling unchanged.
Fixes: 32cb23a0f911 ("ALSA: seq: dummy: Allow UMP conversion")
Signed-off-by: Kyle Zeng <kylebot@openai.com>
Link: https://patch.msgid.link/20260605080204.32045-1-kylebot@openai.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The OSS sequencer write and out-of-band paths may receive a temporary
snd_use_lock_t reference from snd_seq_oss_process_event(). This was added
to keep MIDI device data alive until events with embedded SysEx data are
dispatched.
Use a scoped cleanup helper for that temporary reference. This keeps the
lifetime rule local to the variable declaration and avoids future missing
snd_use_lock_free() paths if these event handling paths gain more exits.
No functional change is intended.
Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
Signed-off-by: Takashi Iwai <tiwai@suse.de>
Link: https://patch.msgid.link/20260604-alsa-scoped-cleanups-v1-3-10c43152a728@gmail.com
|
|
snd_seq_oss_read() checks whether the next queued OSS sequencer event
fits in the remaining userspace buffer before removing it from the read
queue.
The check is inverted. It currently stops when the event is smaller than
the remaining buffer, so a normal 4-byte event is not copied for an
8-byte read buffer. Conversely, an 8-byte event can be copied for a
smaller read count.
Break only when the remaining userspace buffer is smaller than the next
event, and report -EINVAL if no complete event has been copied. This
prevents an undersized read from looking like end-of-file while leaving
the event queued for a later read with a large enough buffer.
Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
Link: https://patch.msgid.link/20260602-alsa-seq-oss-read-size-check-v1-1-10e59b1742e0@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Commit 2ee646353cd5 ("ALSA: seq: Register kernel port with full
information") split sequencer port creation from list insertion so a
port can be filled before it becomes visible.
However, snd_seq_ioctl_create_port() still copies port->addr back to the
ioctl argument before snd_seq_insert_port() assigns the final port
number. A successful SNDRV_SEQ_IOCTL_CREATE_PORT without
SNDRV_SEQ_PORT_FLG_GIVEN_PORT can therefore report port -1 to userspace.
Move the ioctl address copy after successful insertion, and keep the
default "port-%d" name assignment from overwriting a caller-provided port
name. This restores the observable behavior from before the split while
keeping the port populated before publication.
Fixes: 2ee646353cd5 ("ALSA: seq: Register kernel port with full information")
Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
Link: https://patch.msgid.link/20260602-alsa-seq-create-port-info-fix-v1-1-eec0280131e9@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
event_process_midi() borrows msynth->output_rfile.output and then
passes the substream to dump_midi() and snd_rawmidi_kernel_write()
without synchronizing with the output open/close transition.
midisynth_use() also publishes output_rfile before
snd_rawmidi_output_params() has finished.
The last midisynth_unuse() can therefore release the same rawmidi file
and free substream->runtime before snd_rawmidi_kernel_write1() takes
its runtime buffer reference. That leaves the event_input path using a
stale substream or runtime and can end in a NULL-deref or use-after-free.
Fix this with two pieces of synchronization. Keep a short IRQ-safe
spinlock only for publishing or clearing output_rfile and for pairing
the output snapshot with an snd_use_lock_t reference. Once
event_process_midi() has taken that in-flight reference, it drops the
spinlock before calling snd_seq_dump_var_event(), dump_midi(), or
snd_rawmidi_kernel_write(). midisynth_unuse() now detaches the visible
rawmidi file under the same spinlock, waits for the in-flight writers
to drain, and only then drains and releases the saved file.
midisynth_use() likewise opens into a local snd_rawmidi_file and
publishes it only after snd_rawmidi_output_params() succeeds.
The buggy scenario involves two paths, with each column showing the
order within that path:
event_input path: last unuse path:
1. event_process_midi() snapshots 1. midisynth_unuse() starts
output_rfile.output. tearing down output_rfile.
2. dump_midi() reaches 2. snd_rawmidi_kernel_release()
snd_rawmidi_kernel_write() closes the output file.
before runtime is pinned. 3. close_substream() frees
3. The callback keeps using substream->runtime.
the borrowed substream.
Validation reproduced this kernel report:
KASAN null-ptr-deref in snd_rawmidi_kernel_write1+0x56/0x360
RIP: 0033:0x7fde7dd0837f
RIP: 0010:snd_rawmidi_kernel_write1+0x56/0x360
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Assisted-by: Codex:gpt-5.5
Signed-off-by: Zhang Cen <rollkingzzc@gmail.com>
Link: https://patch.msgid.link/20260527062948.3614025-1-rollkingzzc@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The OSS sequencer processes the input MIDI bytes into a sequencer
event to be dispatched later (in snd_seq_oss_midi_putc() called from
snd_seq_oss_process_event()). When it's a SysEx data, the event
record contains data.ext.ptr pointer to the original SysEx bytes, and
the referred data is copied into the pool afterwards at dispatching.
The problem is that, if the sequencer port gets closed concurrently
before the dispatch, the OSS sequencer core also releases the
resources (in snd_seq_oss_midi_check_exit_port()), while the pending
event may hold a stale pointer, eventually leading to a UAF at a later
dispatch.
Fortunately, there is already a refcounting mechanism (snd_use_lock_t)
for the OSS MIDI device access, and for addressing the issue above, we
just need to extend the refcount until the event gets dispatched.
This patch extends snd_seq_oss_process_event() to give back the
refcount object, which is in turn released after calling the sequencer
dispatcher with the given event in the caller side.
According to the original report, KASAN report as below:
KASAN slab-use-after-free in snd_seq_event_dup+0x40c/0x470
RIP: 0033:0x7f2cb66a6340
Read of size 6
Call trace:
dump_stack_lvl+0x73/0xb0 (?:?)
print_report+0xd1/0x650 (?:?)
srso_alias_return_thunk+0x5/0xfbef5 (?:?)
__virt_addr_valid+0x1a7/0x340 (?:?)
kasan_complete_mode_report_info+0x64/0x200 (?:?)
kasan_report+0xf7/0x130 (?:?)
snd_seq_event_dup+0x40c/0x470 (?:?)
kasan_check_range+0x10c/0x1c0 (?:?)
__asan_memcpy+0x27/0x70 (?:?)
snd_seq_event_dup+0x9/0x470 (?:?)
snd_seq_client_enqueue_event+0x139/0x240 (?:?)
_raw_spin_unlock_irqrestore+0x4b/0x60 (?:?)
snd_seq_kernel_client_enqueue+0x102/0x120 (?:?)
snd_seq_oss_write+0x416/0x4e0 (?:?)
apparmor_file_permission+0x20/0x30 (?:?)
odev_write+0x3b/0x60 (?:?)
vfs_write+0x1ce/0x850 (?:?)
lock_release+0xc8/0x2a0 (?:?)
__kasan_check_write+0x18/0x20 (?:?)
__mutex_unlock_slowpath+0x129/0x510 (?:?)
ksys_write+0xe1/0x180 (?:?)
mutex_unlock+0x16/0x20 (?:?)
odev_ioctl+0x65/0xc0 (?:?)
__x64_sys_write+0x46/0x60 (?:?)
x64_sys_call+0x7d/0x20d0 (?:?)
do_syscall_64+0xc1/0x360 (arch/x86/entry/syscall_64.c:87)
entry_SYSCALL_64_after_hwframe+0x77/0x7f (?:?)
Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Reported-and-tested-by: Zhang Cen <rollkingzzc@gmail.com>
Closes: https://lore.kernel.org/20260521233900.478153-1-rollkingzzc@gmail.com
Link: https://patch.msgid.link/20260526152843.617503-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The sequencer priority queue insertion path uses a hardcoded traversal
limit of 10000 entries. The value is intended to catch a corrupted list,
but it also becomes a real limit for valid queues.
The event pool limit is per client, while a sequencer queue can be shared
by multiple clients. A queue can therefore legitimately contain more than
10000 events. In that case, inserting an event that has to be placed past
the arbitrary limit fails with -EINVAL.
Use the queue's own cell count as the traversal bound instead. This keeps
the protection against inconsistent list accounting or cyclic lists without
rejecting valid large queues.
Signed-off-by: Cássio Gabriel <cassiogabrielcontato@gmail.com>
Link: https://patch.msgid.link/20260525-alsa-seq-prioq-limit-v1-1-16c348df5ff7@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Some variables in the 'ALSA/core' are initialized only during the init
phase in the '__init' functions and never changed. So, mark them as
__ro_after_init to reduce the attack surface.
Signed-off-by: Len Bao <len.bao@gmx.us>
Link: https://patch.msgid.link/20260524162914.47764-1-len.bao@gmx.us
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
seq_ump_process_event() borrows client->out_rfile.output without
synchronizing with the first-open and last-close transition in
seq_ump_client_open() and seq_ump_client_close().
The last output unuse can therefore drop opened[STR_OUT] to zero and
release the rawmidi file while an in-flight event_input callback is still
inside snd_rawmidi_kernel_write(). That leaves the rawmidi substream
runtime exposed to teardown before the write path has taken its own
buffer reference.
Add a per-client rwlock for the event_input-visible output file. Publish
a newly opened output file under the write side, and hold the read side
from the output lookup through snd_rawmidi_kernel_write(). The last
output close copies and clears the visible output file under the write
side, then drops the lock and releases the saved rawmidi file. Use
IRQ-safe rwlock guards because event_input can also be reached from
atomic sequencer delivery.
The buggy scenario involves two paths, with each column showing the
order within that path:
path A label: event_input path path B label: last unuse path
1. seq_ump_process_event() reads 1. seq_ump_client_close()
client->out_rfile.output. drops opened[STR_OUT] to zero.
2. snd_rawmidi_kernel_write1() 2. snd_rawmidi_kernel_release()
has not yet pinned runtime. closes the output file.
3. The writer continues using 3. close_substream() frees
the borrowed substream. substream->runtime.
This keeps the output substream and runtime alive for the full
event_input write while keeping rawmidi release outside the rwlock.
KASAN reproduced this as a slab-use-after-free in
snd_rawmidi_kernel_write1(), with allocation through
seq_ump_use()/snd_seq_port_connect() and free through
seq_ump_unuse()/snd_seq_port_disconnect().
Suggested-by: Takashi Iwai <tiwai@suse.de>
Validation reproduced this kernel report:
KASAN slab-use-after-free in snd_rawmidi_kernel_write1+0x9d/0x400
RIP: 0033:0x7f5528af837f
Read of size 8
Call trace:
dump_stack_lvl+0x73/0xb0 (?:?)
print_report+0xd1/0x650 (?:?)
srso_alias_return_thunk+0x5/0xfbef5 (?:?)
__virt_addr_valid+0x1a7/0x340 (?:?)
kasan_complete_mode_report_info+0x64/0x200 (?:?)
kasan_report+0xf7/0x130 (?:?)
snd_rawmidi_kernel_write1+0x9d/0x400 (?:?)
__asan_load8+0x82/0xb0 (?:?)
update_stack_state+0x1ef/0x2d0 (?:?)
snd_rawmidi_kernel_write+0x1a/0x20 (?:?)
seq_ump_process_event+0xd4/0x120 (sound/core/seq/seq_ump_client.c:82)
__snd_seq_deliver_single_event+0x8a/0xe0 (?:?)
snd_seq_deliver_from_ump+0x2b2/0xd60 (?:?)
lock_acquire+0x14e/0x2e0 (?:?)
find_held_lock+0x31/0x90 (?:?)
snd_seq_port_use_ptr+0xa6/0xe0 (?:?)
__kasan_check_write+0x18/0x20 (?:?)
do_raw_read_unlock+0x32/0xa0 (?:?)
_raw_read_unlock+0x26/0x50 (?:?)
snd_seq_deliver_single_event+0x45c/0x4b0 (?:?)
snd_seq_deliver_event+0x10d/0x1b0 (?:?)
snd_seq_client_enqueue_event+0x192/0x240 (?:?)
snd_seq_write+0x2cd/0x450 (?:?)
apparmor_file_permission+0x20/0x30 (?:?)
security_file_permission+0x51/0x60 (?:?)
vfs_write+0x1ce/0x850 (?:?)
__fget_files+0x12b/0x220 (?:?)
lock_release+0xc8/0x2a0 (?:?)
__rcu_read_unlock+0x74/0x2d0 (?:?)
__fget_files+0x135/0x220 (?:?)
ksys_write+0x15a/0x180 (?:?)
rcu_is_watching+0x24/0x60 (?:?)
__x64_sys_write+0x46/0x60 (?:?)
x64_sys_call+0x7d/0x20d0 (?:?)
do_syscall_64+0xc1/0x360 (arch/x86/entry/syscall_64.c:87)
entry_SYSCALL_64_after_hwframe+0x77/0x7f (?:?)
Fixes: 81fd444aa371 ("ALSA: seq: Bind UMP device")
Signed-off-by: Zhang Cen <rollkingzzc@gmail.com>
Link: https://patch.msgid.link/20260520103249.3048345-1-rollkingzzc@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
The current ALSA sequencer core tries to register the new kernel
sequencer port on the list at first, then fill up the port
information. This means that user-space may sneak the wrong
information before the actual data is filled, which isn't ideal.
Although the user-space should try to query the port info after the
port registration notification is sent out, it'd be still better to
have a port available with the full info from the beginning.
This patch changes the sequencer port creation and registration
procedure; now split to two steps, for creation and insertion, and the
port is registered after the information is filled.
Link: https://sashiko.dev/#/patchset/20260518194023.1667857-1-maoyixie.tju%40gmail.com
Link: https://patch.msgid.link/20260519094254.465041-1-tiwai@suse.de
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
snd_seq_create_port() walks client->ports_list_head looking for
the ordered insertion point and on loop fall-through passes
&p->list to list_add_tail():
list_for_each_entry(p, &client->ports_list_head, list) {
if (p->addr.port == port) {
kfree(new_port);
return -EBUSY;
}
if (p->addr.port > num)
break;
...
}
list_add_tail(&new_port->list, &p->list);
When the loop walks all entries without break (e.g., the new
port sorts last), p is past-the-end. &p->list aliases
&client->ports_list_head (the list head) via container_of offset
cancellation, so the insert lands at the list tail. That is the
intended behaviour, but the access is undefined per C11 even
though it works in practice.
Track an explicit insert_before pointer initialised to the list
head and overwritten to &p->list only when the loop breaks
early. The observable behaviour is unchanged.
Fixes: 9244b2c3079f ("[ALSA] alsa core: convert to list_for_each_entry*")
Signed-off-by: Maoyi Xie <maoyixie.tju@gmail.com>
Link: https://patch.msgid.link/20260518194023.1667857-3-maoyixie.tju@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
|
Store MIDI channel entries in the MIDI channel set allocation instead
of allocating them separately.
This ties the channel array lifetime directly to the channel set, removes
a separate allocation failure path, and lets __counted_by() describe the
array bounds. Move the embedded emux channel set to the end of its
containing structure so it can carry the flexible array.
Assisted-by: Codex:GPT-5.5
Signed-off-by: Rosen Penev <rosenp@gmail.com>
Link: https://patch.msgid.link/20260511075447.445350-1-rosenp@gmail.com
Signed-off-by: Takashi Iwai <tiwai@suse.de>
|
