Bluetooth: MGMT: Serialize power-on with power-off completion
Subsystems (2)
PATCH v3 1/1 Bluetooth: MGMT: Serialize power-on with power-off completion
mgmt_set_powered_complete() must consume only its own SET_POWERED command. __mgmt_power_off() frees that command before its destroy callback runs, and mgmt_pending_valid() recognizes pending commands only by address.
The power-on handler can reuse the address, but it is not the only allocator. A concurrent set_mesh() can also publish another pending command at that address after close drops hdev->lock:
Close request worker Management socket
-------------------- -----------------
__mgmt_power_off(): free cmd
hci_dev_close_sync(): unlock
set_mesh(): allocate replacement
completion: claim old address
submission fails: remove/free cmd
completion: access replacement
Serialize power-on publication with req_lock, which the request worker holds through its destroy callback. Keep power-off cancellation outside that mutex and keep the req_lock -> hdev->lock ordering.
Also hold hdev->lock while finding the pending SET_POWERED command, checking its identity, and consuming it. An unrelated command at the old address then fails the opcode-specific lookup without being unlinked. Power-on cannot reuse a SET_POWERED address until the old callback drops req_lock, and power-off allocates no command once the device is down.
Dequeue pending power requests before settings_rsp() frees their command data. Their cancellation callback already returns without touching cmd. An already-dequeued request can still be waiting for req_lock. Check its opcode under mgmt_pending_lock before copying parameters as well, so it cannot interpret an unrelated replacement command as power settings.
KASAN report as below:
BUG: KASAN: slab-use-after-free in
mgmt_set_powered_complete+0x450/0x4d0
Read of size 2 at addr ffff88810921ab10 by task kworker/u17:0/500
Workqueue: hci0 hci_cmd_sync_work
Call Trace:
<TASK>
dump_stack_lvl+0x93/0xd0
print_report+0xce/0x630
? mgmt_set_powered_complete+0x450/0x4d0
? srso_alias_return_thunk+0x5/0xfbef5
? __virt_addr_valid+0x20d/0x410
? mgmt_set_powered_complete+0x450/0x4d0
kasan_report+0xe0/0x110
? mgmt_set_powered_complete+0x450/0x4d0
mgmt_set_powered_complete+0x450/0x4d0
? __pfx_mgmt_set_powered_complete+0x10/0x10
hci_cmd_sync_work+0x207/0x450
? __pfx_mgmt_set_powered_complete+0x10/0x10
process_one_work+0x908/0x19c0
? __pfx_process_one_work+0x10/0x10
? srso_alias_return_thunk+0x5/0xfbef5
? lock_is_held_type+0x8f/0x100
? srso_alias_return_thunk+0x5/0xfbef5
worker_thread+0x65c/0xe40
? __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>
Allocated by task 531:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
__kasan_kmalloc+0xaa/0xb0
__kmalloc_cache_noprof+0x251/0x610
mgmt_pending_new+0x60/0x230
mgmt_pending_add+0x3b/0x180
set_powered+0x2a8/0x7e0
hci_sock_sendmsg+0x128b/0x22e0
__sys_sendto+0x425/0x470
__x64_sys_sendto+0xe5/0x1c0
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
Freed by task 531:
kasan_save_stack+0x33/0x60
kasan_save_track+0x14/0x30
kasan_save_free_info+0x3b/0x60
__kasan_slab_free+0x5f/0x80
kfree+0x307/0x580
set_powered+0x336/0x7e0
hci_sock_sendmsg+0x128b/0x22e0
__sys_sendto+0x425/0x470
__x64_sys_sendto+0xe5/0x1c0
do_syscall_64+0x115/0x6a0
entry_SYSCALL_64_after_hwframe+0x77/0x7f
==================================================================---
Changes in v3:
- Address cross-opcode pending-command reuse reported by Sashiko:
https://sashiko.dev/#/patchset/88964
- Validate SET_POWERED identity before claiming it.
- Cancel queued power requests before freeing their command data.
- Check the opcode of an already-dequeued request before reading its parameters.
- Retain the balanced power-on request locking introduced in v2.
Changes in v2:
- Move the common set_powered command path into a helper and acquire req_lock in the explicit power-on branch, fixing Clang context-analysis errors.
- Preserve req_lock -> hdev->lock ordering and keep power-off cancellation outside req_lock.
- Rebase onto current bluetooth-next.
Link to v1: https://lore.kernel.org/r/pm-bluetooth-objects-candidate-0063-v7-35a6680c94a332fea155@gmail.com
Link to v2: https://lore.kernel.org/r/pm-bluetooth-objects-candidate-0063-v11-a70581b6877e5fe118a3@gmail.com
net/bluetooth/mgmt.c +47 -14
@@ -1342,9 +1342,15 @@struct mgmt_pending_cmd *cmd = data;struct mgmt_mode *cp; -/* Make sure cmd still outstanding. */ -if (err == -ECANCELED || !mgmt_pending_valid (hdev, cmd)) +if (err == -ECANCELED )return ; + +hci_dev_lock (hdev); + +/* Other management commands may reuse the old command address. */ +if (pending_find (MGMT_OP_SET_POWERED , hdev) != cmd || + !mgmt_pending_valid (hdev, cmd)) + goto unlock; cp = cmd->param;
@@ -1352,10 +1358,8 @@if (!err) {if (cp->val) { -hci_dev_lock (hdev);restart_le_actions (hdev);hci_update_passive_scan (hdev); -hci_dev_unlock (hdev); }send_settings_rsp (cmd->sk, cmd->opcode, hdev);
@@ -1371,6 +1375,9 @@ }mgmt_pending_free (cmd); + +unlock: +hci_dev_unlock (hdev); }static int set_powered_sync (struct hci_dev *hdev,void *data)
@@ -1380,8 +1387,9 @@mutex_lock (&hdev->mgmt_pending_lock); -/* Make sure cmd still outstanding. */ -if (!__mgmt_pending_listed (hdev, cmd)) { +/* A dequeued request may have waited across a controller reset. */ +if (!__mgmt_pending_listed (hdev, cmd) || + cmd->opcode !=MGMT_OP_SET_POWERED ) {mutex_unlock (&hdev->mgmt_pending_lock);return -ECANCELED ; }
@@ -1395,18 +1403,12 @@return hci_set_powered_sync (hdev, cp.val); } -static int set_powered (struct sock *sk,struct hci_dev *hdev,void *data, -u16 len) +static int __set_powered (struct sock *sk,struct hci_dev *hdev, +void *data,u16 len) {struct mgmt_mode *cp = data;struct mgmt_pending_cmd *cmd;int err; - -bt_dev_dbg (hdev,"sock %p" , sk); - -if (cp->val !=0x00 && cp->val !=0x01 ) -return mgmt_cmd_status (sk, hdev->id,MGMT_OP_SET_POWERED , -MGMT_STATUS_INVALID_PARAMS );hci_dev_lock (hdev);
@@ -1454,6 +1456,32 @@return err; } +static int set_powered (struct sock *sk,struct hci_dev *hdev,void *data, +u16 len) +{ +struct mgmt_mode *cp = data; +int err; + +bt_dev_dbg (hdev,"sock %p" , sk); + +if (cp->val !=0x00 && cp->val !=0x01 ) +return mgmt_cmd_status (sk, hdev->id,MGMT_OP_SET_POWERED , +MGMT_STATUS_INVALID_PARAMS ); + +if (!cp->val) +return __set_powered (sk, hdev, data, len); + +/* +* Wait for the previous power-off completion before publishing a +* new command that could reuse its freed pending-command address. +*/ +hci_req_sync_lock (hdev); + err =__set_powered (sk, hdev, data, len); +hci_req_sync_unlock (hdev); + +return err; +} +int mgmt_new_settings (struct hci_dev *hdev) {return new_settings (hdev,NULL );
@@ -1468,6 +1496,11 @@static void settings_rsp (struct mgmt_pending_cmd *cmd,void *data) {struct cmd_lookup *match = data; + +/* A queued power request must not retain the freed command. */ +if (cmd->opcode ==MGMT_OP_SET_POWERED ) +hci_cmd_sync_dequeue (match->hdev, set_powered_sync, cmd, + mgmt_set_powered_complete);send_settings_rsp (cmd->sk, cmd->opcode, match->hdev);
This is automated email and please do not reply to this email!
Dear submitter,
Thank you for submitting the patches to the linux bluetooth mailing list. This is a CI test results with your patch series:
PW Link:https://patchwork.kernel.org/series/1191273/
---Test result---
Test Summary:
CheckPatch PASS 5.70 seconds VerifyFixes PASS 0.14 seconds VerifySignedoff PASS 0.09 seconds GitLint PASS 0.26 seconds SubjectPrefix PASS 0.11 seconds BuildKernel PASS 28.86 seconds CheckAllWarning PASS 33.45 seconds CheckSparse PASS 35.24 seconds BuildKernel32 PASS 29.26 seconds CheckKernelLLVM PASS 29.80 seconds TestRunnerSetup PASS 679.97 seconds TestRunner_mgmt-tester PASS 87.48 seconds TestRunner_mesh-tester PASS 5.25 seconds IncrementalBuild PASS 27.19 seconds
https://github.com/bluez/bluetooth-next/pull/979
---
Regards,
Linux Bluetooth