lkdtm: Add folio_lock deadlock scenarios
Subsystems (4)
PATCH 1/1 lkdtm: Add folio_lock deadlock scenarios
Introduces four new crash types to LKDTM to reproduce deadlock patterns involving folio_lock(), which operates on a wait-on-bit mechanism:
1. FOLIO_LOCK_AA:
Triggers a self-deadlock (AA) by attempting to acquire the same folio lock twice in the same execution context.
2. FOLIO_LOCK_ABBA:
Triggers a classic ABBA deadlock between two threads trying to folio lock two different folios in reverse order.
3. FOLIO_MUTEX_LOCK_ABBA:
Reproduces an ABBA deadlock involving a folio_lock() and a mutex.
4. FOLIO_DEFERRED_EVENT_ABBA:
Creates a deferred deadlock where a thread holding a folio_lock() waits on a wait queue. By deferring its lock acquisition to a workqueue, the waker forms a circular dependency that blocks both the waiter and the kworker.
These tests allow developers to validate the kernel's behavior (e.g., hung task detection, DEPT[1][2][3] report) under wait/event-based deadlock conditions.
[1] https://lwn.net/Articles/1036222/
[2] https://youtu.be/pfWxBuMzxks?si=mW699Yz6dp38diiM
[3] https://lore.kernel.org/lkml/[email protected]/
Below are the call traces from hungtaskd (for 1. FOLIO_LOCK_AA and
4. FOLIO_DEFERRED_EVENT_ABBA) and DEPT (for 2. FOLIO_LOCK_ABBA and
3. FOLIO_MUTEX_LOCK_ABBA):
# echo FOLIO_LOCK_AA > /sys/kernel/debug/provoke-crash/DIRECT
[ 26.841460] lkdtm: Performing direct entry FOLIO_LOCK_AA
[ 61.151664] INFO: task bash:915 blocked for more than 30 seconds.
[ 61.152107] Not tainted 6.19.0-virtme #20
[ 61.152482] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 61.152843] task:bash state:D stack:12536 pid:915 tgid:915 ppid:909 task_flags:0x400100 flags:0x00080000
[ 61.153440] Call Trace:
[ 61.153585] <TASK>
[ 61.153732] ? __schedule+0x5e9/0x11b0
[ 61.153941] __schedule+0x61c/0x11b0
[ 61.154157] schedule+0x3a/0x130
[ 61.154447] io_schedule+0x46/0x70
[ 61.154649] folio_wait_bit_common+0x1ab/0x440
[ 61.154918] ? __pfx_wake_page_function+0x10/0x10
[ 61.155167] lkdtm_FOLIO_LOCK_AA+0x10c/0x1b0
[ 61.155562] lkdtm_do_action+0x18/0x30
[ 61.155754] direct_entry+0x8d/0xe0
[ 61.155955] full_proxy_write+0x69/0xa0
[ 61.156157] vfs_write+0xea/0x600
[ 61.156448] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.156706] ? find_held_lock+0x2b/0x80
[ 61.156902] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157141] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157469] ? from_pool+0x7d/0x190
[ 61.157664] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157900] ? dept_enter+0x68/0xa0
[ 61.158098] ksys_write+0x76/0xf0
[ 61.158396] do_syscall_64+0xc2/0xf80
[ 61.158599] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 61.158851] RIP: 0033:0x7fe2e7d58340
[ 61.159063] RSP: 002b:00007ffc370df3a8 EFLAGS: 00000202 ORIG_RAX: 0000000000000001
[ 61.159526] RAX: ffffffffffffffda RBX: 000000000000000e RCX: 00007fe2e7d58340
[ 61.159863] RDX: 000000000000000e RSI: 0000559c6fb13ed0 RDI: 0000000000000001
[ 61.160353] RBP: 0000559c6fb13ed0 R08: 0000000000000007 R09: 0000000000000073
[ 61.160695] R10: 0000000000001000 R11: 0000000000000202 R12: 000000000000000e
[ 61.161032] R13: 00007fe2e7e34760 R14: 000000000000000e R15: 00007fe2e7e2f9e0
[ 61.161490] </TASK>
[ 61.161639]
[ 61.161639] Showing all locks held in the system:
[ 61.161942] 1 lock held by khungtaskd/116:
[ 61.162142] #0: ffffffffa2380620 (rcu_read_lock){....}-{1:3}, at: debug_show_all_locks+0x36/0x1c0
[ 61.162884] 1 lock held by bash/915:
[ 61.163422] #0: ffff89af81beb570 (sb_writers#8){.+.+}-{0:0}, at: ksys_write+0x76/0xf0
[ 61.164077]
[ 61.164367] =============================================# echo FOLIO_LOCK_ABBA > /sys/kernel/debug/provoke-crash/DIRECT [ 182.473798] lkdtm: Performing direct entry FOLIO_LOCK_ABBA [ 182.475397] =================================================== [ 182.475637] DEPT: Circular dependency has been detected. [ 182.475774] 6.19.0-virtme #20 Not tainted [ 182.475885] --------------------------------------------------- [ 182.476054] summary [ 182.476112] --------------------------------------------------- [ 182.476259] *** AA DEADLOCK *** [ 182.476259] [ 182.476379] context A [ 182.476463] [S] (unknown)(pg_locked_map:0) [ 182.476604] [W] dept_page_wait_on_bit(pg_locked_map:0) [ 182.476744] [E] dept_page_clear_bit(pg_locked_map:0) [ 182.476884] [ 182.476943] [S]: start of the event context [ 182.477055] [W]: the wait blocked [ 182.477158] [E]: the event not reachable [ 182.477268] --------------------------------------------------- [ 182.477415] context A's detail [ 182.477504] --------------------------------------------------- [ 182.477652] context A [ 182.477711] [S] (unknown)(pg_locked_map:0) [ 182.477852] [W] dept_page_wait_on_bit(pg_locked_map:0) [ 182.477992] [E] dept_page_clear_bit(pg_locked_map:0) [ 182.478132] [ 182.478190] [S] (unknown)(pg_locked_map:0): [ 182.478300] (N/A) [ 182.478359] [ 182.478418] [W] dept_page_wait_on_bit(pg_locked_map:0): [ 182.478558] [<ffffffff8cf9a5ae>] kthread+0xfe/0x200 [ 182.478698] stacktrace: [ 182.478757] kthread+0xfe/0x200 [ 182.478867] ret_from_fork+0x29d/0x2e0 [ 182.478977] ret_from_fork_asm+0x1a/0x30 [ 182.479118] [ 182.479177] [E] dept_page_clear_bit(pg_locked_map:0): [ 182.479316] [<ffffffff8d989fe3>] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.479485] stacktrace: [ 182.479545] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.479685] kthread+0xfe/0x200 [ 182.479796] ret_from_fork+0x29d/0x2e0 [ 182.479906] ret_from_fork_asm+0x1a/0x30 [ 182.480045] --------------------------------------------------- [ 182.480192] information that might be helpful [ 182.480310] --------------------------------------------------- [ 182.480458] CPU: 2 UID: 0 PID: 915 Comm: lkdtm_folio_a Not tainted 6.19.0-virtme #20 PREEMPT(voluntary) [ 182.480464] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 182.480466] Call Trace: [ 182.480469] <TASK> [ 182.480473] dump_stack_lvl+0x69/0xa0 [ 182.480481] cb_check_dl+0x6be/0x760 [ 182.480492] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480497] ? lock_acquire+0x26b/0x2b0 [ 182.480507] bfs+0x138/0x1c0 [ 182.480511] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480522] add_dep+0xd6/0x1c0 [ 182.480527] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480531] ? __pfx_bfs_init_check_dl+0x10/0x10 [ 182.480535] ? __pfx_bfs_extend_dep+0x10/0x10 [ 182.480539] ? __pfx_bfs_dequeue_dep+0x10/0x10 [ 182.480543] ? __pfx_cb_check_dl+0x10/0x10 [ 182.480551] __dept_event+0x489/0x520 [ 182.480558] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480564] ? lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.480570] dept_event+0x99/0xc0 [ 182.480581] folio_unlock+0x3c/0x60 [ 182.480587] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.480594] ? __pfx_lkdtm_folio_ab_kthread+0x10/0x10 [ 182.480598] kthread+0xfe/0x200 [ 182.480605] ? __pfx_kthread+0x10/0x10 [ 182.480613] ret_from_fork+0x29d/0x2e0 [ 182.480616] ? __pfx_kthread+0x10/0x10 [ 182.480622] ret_from_fork_asm+0x1a/0x30 [ 182.480644] </TASK>
# echo FOLIO_MUTEX_LOCK_ABBA > /sys/kernel/debug/provoke-crash/DIRECT [ 25.744189] lkdtm: Performing direct entry FOLIO_MUTEX_LOCK_ABBA [ 25.750265] =================================================== [ 25.750826] DEPT: Circular dependency has been detected. [ 25.750911] 6.19.0-virtme #20 Not tainted [ 25.750995] --------------------------------------------------- [ 25.751098] summary [ 25.751149] --------------------------------------------------- [ 25.751252] *** DEADLOCK *** [ 25.751252] [ 25.751338] context A [ 25.751389] [S] lock(mutex_b:0) [ 25.751458] [W] dept_page_wait_on_bit(pg_lcked_map:0) [[ 25.751652] ? [E] unlock(mutex_b:0) 2[ 25.751781] 0[ 25.751861] context B 0[ 25.751937] [S] (unknown)(pg_locked_map:0)4 h[ 25.752088] [W] lock(mutex_b:0) [ 25.752175] [E] dept_page_clear_bit(pg_locked_map:0) [ 25.752265] [ 25.752315] [S]: start of the event context [ 25.752382] [W]: the wait blocked [ 25.752452] [E]: the event not reachable [ 25.752520] --------------------------------------------------- [ 25.752621] context A's detairl [ 25.752726] -o--------------------------------o------------------ t[ 25.752894] context A @[ 25.752960] [S] lock(mutexv_b:0) i[ 25.753074] [W] dept_page_rwait_on_bit(pg_locked_map:0) t[ 25.753210] [E] unlock(mutmex_b:0) e[ 25.753324] -[ 25.753394] [S] lock(mutex_b:n0): g[ 25.753513] [:<ffffffffb1989d90>] lkdtm_mutex_/folio_kthread+0x40/0xe0 [ 25.753705] stacktrace: [ 25.753757] h lkdtm_mutex_folio_kthread+0x40/0xe0 o[ 25.753890] m kthread+0xfee/0x200 /[ 25.754049] ret_from_fokrk+0x29d/0x2e0 i[ 25.754175] m ret_from_fo/rk_asm+0x1a/0x30s a[ 25.754390] [ 25.754496] [gW] dept_page_waiit_on_bit(pg_lockned_map:0): g[ 25.754678] [<ffffffffb0f9a5ae>] kthread+0xfe/0x200 [ 25.754796] s#tacktrace: [ 25.754874] kthread+0xfe/0x200 [ 25.754981] ret_from_fork+0x29d/0x2e0 [ 25.755039] ret_from_fork_asm+0x1a/0x30 [ 25.755130] [ 25.755185] [E] unlock(mutex_b:0): [ 25.755254] (N/A) [ 25.755304] --------------------------------------------------- [ 25.755406] context B's detail [ 25.755473] --------------------------------------------------- [ 25.755576] context B [ 25.755625] [S] (unknown)(pg_locked_map:0) [ 25.755724] [W] lock(mutex_b:0) [ 25.755791] [E] dept_page_clear_bit(pg_locked_map:0) [ 25.755893] [ 25.755955] [S] (unknown)(pg_locked_map:0): [ 25.756039] (N/A) [ 25.756102] [ 25.756164] [W] lock(mutex_b:0): [ 25.756248] [<ffffffffb1989e68>] lkdtm_folio_mutex_kthread+0x28/0xe0 [ 25.756376] stacktrace: [ 25.756438] lkdtm_folio_mutex_kthread+0x28/0xe0 [ 25.756545] kthread+0xfe/0x200 [ 25.756629] ret_from_fork+0x29d/0x2e0 [ 25.756726] ret_from_fork_asm+0x1a/0x30 [ 25.756853] [ 25.756918] [E] dept_page_clear_bit(pg_locked_map:0): [ 25.757081] [<ffffffffb1989e80>] lkdtm_folio_mutex_kthread+0x40/0xe0 [ 25.757241] stacktrace: [ 25.757315] lkdtm_folio_mutex_kthread+0x40/0xe0 [ 25.757449] kthread+0xfe/0x200 [ 25.757555] ret_from_fork+0x29d/0x2e0 [ 25.757672] ret_from_fork_asm+0x1a/0x30 [ 25.757816] --------------------------------------------------- [ 25.758014] information that might be helpful [ 25.758180] --------------------------------------------------- [ 25.758370] CPU: 15 UID: 0 PID: 922 Comm: lkdtm_mutex_fol Not tainted 6.19.0-virtme #20 PREEMPT(voluntary) [ 25.758381] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 25.758386] Call Trace: [ 25.758391] <TASK> [ 25.758403] dump_stack_lvl+0x69/0xa0 [ 25.758424] cb_check_dl+0x6be/0x760 [ 25.758454] bfs+0x17d/0x1c0 [ 25.758459] ? srso_alias_return_thunk+0x5/0xfbef5 [ 25.758471] add_dep+0xd6/0x1c0 [ 25.758477] ? lkdtm_mutex_folio_kthread+0x40/0xe0 [ 25.758482] ? __pfx_bfs_init_check_dl+0x10/0x10 [ 25.758488] ? __pfx_bfs_extend_dep+0x10/0x10 [ 25.758492] ? __pfx_bfs_dequeue_dep+0x10/0x10 [ 25.758496] ? __pfx_cb_check_dl+0x10/0x10 [ 25.758505] __dept_wait+0x274/0x6a0 [ 25.758514] ? kthread+0xfe/0x200 [ 25.758520] ? __mutex_lock+0xae3/0x1230 [ 25.758532] ? srso_alias_return_thunk+0x5/0xfbef5 [ 25.758536] ? dept_enter+0x68/0xa0 [ 25.758545] ? kthread+0xfe/0x200 [ 25.758551] dept_wait+0xa7/0xc0 [ 25.758562] ? __pfx_lkdtm_mutex_folio_kthread+0x10/0x10 [ 25.758567] lkdtm_mutex_folio_kthread+0x9d/0xe0 [ 25.758573] kthread+0xfe/0x200 [ 25.758579] ? __pfx_kthread+0x10/0x10 [ 25.758587] ret_from_fork+0x29d/0x2e0 [ 25.758591] ? __pfx_kthread+0x10/0x10 [ 25.758597] ret_from_fork_asm+0x1a/0x30 [ 25.758619] </TASK>
# echo FOLIO_DEFERRED_EVENT_ABBA > /sys/kernel/debug/provoke-crash/DIRECT
[ 100.907960] lkdtm: Performing direct entry FOLIO_DEFERRED_EVENT_ABBA
[ 100.908476] lkdtm: [Waiter Thread] Securing the folio_lock...
[ 100.908662] lkdtm: [Waiter Thread] Lock secured. Sleeping on wait queue.
[ 100.908691] lkdtm: [Trigger] Calling wake_up() to initiate deferred deadlock.
[ 100.908928] lkdtm: [Waker Context] Atomic callback triggered. Deferring work...
[ 100.909265] lkdtm: [Worker Context] Attempting to acquire folio_lock...
[ 151.210039] INFO: task kworker/14:1:124 blocked for more than 30 seconds.
[ 151.210422] Not tainted 6.19.0-virtme #23
[ 151.210559] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 151.210695] task:kworker/14:1 state:D stack:14344 pid:124 tgid:124 ppid:2 task_flags:0x4208060 flags:0x00080000
[ 151.210880] Workqueue: events deferred_deadlocking
[ 151.210979] Call Trace:
[ 151.211029] <TASK>
[ 151.211088] __schedule+0x5e7/0x1170
[ 151.211182] schedule+0x3a/0x130
[ 151.211256] io_schedule+0x46/0x70
[ 151.211324] folio_wait_bit_common+0x125/0x2d0
[ 151.211423] ? __pfx_wake_page_function+0x10/0x10
[ 151.211563] deferred_deadlocking+0x68/0x70
[ 151.211664] process_one_work+0x205/0x690
[ 151.211782] ? lock_is_held_type+0x9e/0x110
[ 151.211900] worker_thread+0x188/0x330
[ 151.211994] ? __pfx_worker_thread+0x10/0x10
[ 151.212111] kthread+0xfe/0x200
[ 151.212199] ? __pfx_kthread+0x10/0x10
[ 151.212295] ret_from_fork+0x2b2/0x2e0
[ 151.212385] ? __pfx_kthread+0x10/0x10
[ 151.212473] ret_from_fork_asm+0x1a/0x30
[ 151.212599] </TASK>
[ 151.212676] INFO: task lkdtm_waiter:916 blocked for more than 30 seconds.
[ 151.212844] Not tainted 6.19.0-virtme #23
[ 151.212955] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 151.213105] task:lkdtm_waiter state:D stack:14936 pid:916 tgid:916 ppid:2 task_flags:0x208040 flags:0x00080000
[ 151.213309] Call Trace:
[ 151.213372] <TASK>
[ 151.213444] __schedule+0x5e7/0x1170
[ 151.213550] ? __pfx_lkdtm_waiter_thread+0x10/0x10
[ 151.213666] schedule+0x3a/0x130
[ 151.213776] lkdtm_waiter_thread+0xab/0x120
[ 151.213869] ? __pfx_wake_for_deferred_work+0x10/0x10
[ 151.213986] kthread+0xfe/0x200
[ 151.214077] ? __pfx_kthread+0x10/0x10
[ 151.214173] ret_from_fork+0x2b2/0x2e0
[ 151.214260] ? __pfx_kthread+0x10/0x10
[ 151.214347] ret_from_fork_asm+0x1a/0x30
[ 151.214472] </TASK>
[ 151.214588]
[ 151.214588] Showing all locks held in the system:
[ 151.214766] 1 lock held by khungtaskd/116:
[ 151.214878] #0: ffffffffb676d760 (rcu_read_lock){....}-{1:3}, at: debug_show_all_locks+0x36/0x1c0
[ 151.215078] 2 locks held by kworker/14:1/124:
[ 151.215191] #0: ffff8ee181098948 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x55c/0x690
[ 151.215386] #1: ffffd401404a3e40 ((work_completion)(&deadlock_work)){+.+.}-{0:0}, at: process_one_work+0x1c4/0x690
[ 151.215619]
[ 151.215677] =============================================--- drivers/misc/lkdtm/Makefile | 1 + drivers/misc/lkdtm/core.c | 1 + drivers/misc/lkdtm/deadlock.c | 304 ++++++++++++++++++++++++ drivers/misc/lkdtm/lkdtm.h | 1 + tools/testing/selftests/lkdtm/tests.txt | 4 + 5 files changed, 311 insertions(+) create mode 100644 drivers/misc/lkdtm/deadlock.c
drivers/misc/lkdtm/Makefile +1 -0
@@ -3,6 +3,7 @@ obj-$(CONFIG_LKDTM) += lkdtm.o lkdtm-$(CONFIG_LKDTM) += core.o lkdtm-$(CONFIG_LKDTM) += bugs.o +lkdtm-$(CONFIG_LKDTM) += deadlock.o lkdtm-$(CONFIG_LKDTM) += heap.o lkdtm-$(CONFIG_LKDTM) += perms.o lkdtm-$(CONFIG_LKDTM) += refcount.o
drivers/misc/lkdtm/core.c +1 -0
@@ -89,6 +89,7 @@ static struct crashpoint crashpoints[] = {/* List of possible types for crashes that can be triggered. */ static const struct crashtype_category *crashtype_categories[] = { &bugs_crashtypes, + &deadlock_crashtypes, &heap_crashtypes, &perms_crashtypes, &refcount_crashtypes,
drivers/misc/lkdtm/deadlock.c new file +304 -0
304 lines not on this page · raw message
drivers/misc/lkdtm/lkdtm.h +1 -0
@@ -77,6 +77,7 @@ struct crashtype_category {/* Each category's crashtypes list. */ extern struct crashtype_category bugs_crashtypes; +extern struct crashtype_category deadlock_crashtypes;extern struct crashtype_category heap_crashtypes;extern struct crashtype_category perms_crashtypes;extern struct crashtype_category refcount_crashtypes;
tools/testing/selftests/lkdtm/tests.txt +4 -0
@@ -87,3 +87,7 @@ FORTIFY_STR_MEMBER detected buffer overflow FORTIFY_MEM_OBJECT detected buffer overflow FORTIFY_MEM_MEMBER detected field-spanning write PPC_SLB_MULTIHIT Recovered +#FOLIO_LOCK_AA Hangs the system +#FOLIO_LOCK_ABBA Hangs the system +#FOLIO_MUTEX_LOCK_ABBA Hangs the system +#FOLIO_DEFERRED_EVENT_ABBA Hangs the system
Signature · 1 line
2.39.5
Quoted 24 lines · Yunseong Kim
Introduces four new crash types to LKDTM to reproduce deadlock patterns involving
folio_lock(), which operates on a wait-on-bit mechanism:1.
FOLIO_LOCK_AA:
Triggers a self-deadlock (AA) by attempting to acquire the same folio lock twice in the same execution context.2.
FOLIO_LOCK_ABBA:
Triggers a classic ABBA deadlock between two threads trying to folio lock two different folios in reverse order.3.
FOLIO_MUTEX_LOCK_ABBA:
Reproduces an ABBA deadlock involving afolio_lock()and a mutex.4.
FOLIO_DEFERRED_EVENT_ABBA:
Creates a deferred deadlock where a thread holding afolio_lock()waits on a wait queue. By deferring its lock acquisition to a workqueue, the waker forms a circular dependency that blocks both the waiter and the kworker.These tests allow developers to validate the kernel's behavior (e.g., hung task detection, DEPT[1][2][3] report) under wait/event-based deadlock conditions.
Quoted 7 lines
[1] https://lwn.net/Articles/1036222/
[2] https://youtu.be/pfWxBuMzxks?si=mW699Yz6dp38diiM
[3] https://lore.kernel.org/lkml/[email protected]/Below are the call traces from hungtaskd (for 1.
FOLIO_LOCK_AAand
4.FOLIO_DEFERRED_EVENT_ABBA) and DEPT (for 2.FOLIO_LOCK_ABBAand
3.FOLIO_MUTEX_LOCK_ABBA):
Why don't you attach 'the call traces from hungtaskd for all, 1 to 4', and '3 from DEPT' that can be reported by DEPT? Worth noting that 2 is not covered by DEPT yet.
Quoted 46 lines
# echo FOLIO_LOCK_AA > /sys/kernel/debug/provoke-crash/DIRECT
[ 26.841460] lkdtm: Performing direct entry FOLIO_LOCK_AA
[ 61.151664] INFO: task bash:915 blocked for more than 30 seconds.
[ 61.152107] Not tainted 6.19.0-virtme #20
[ 61.152482] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
[ 61.152843] task:bash state:D stack:12536 pid:915 tgid:915 ppid:909 task_flags:0x400100 flags:0x00080000
[ 61.153440] Call Trace:
[ 61.153585] <TASK>
[ 61.153732] ? __schedule+0x5e9/0x11b0
[ 61.153941] __schedule+0x61c/0x11b0
[ 61.154157] schedule+0x3a/0x130
[ 61.154447] io_schedule+0x46/0x70
[ 61.154649] folio_wait_bit_common+0x1ab/0x440
[ 61.154918] ? __pfx_wake_page_function+0x10/0x10
[ 61.155167] lkdtm_FOLIO_LOCK_AA+0x10c/0x1b0
[ 61.155562] lkdtm_do_action+0x18/0x30
[ 61.155754] direct_entry+0x8d/0xe0
[ 61.155955] full_proxy_write+0x69/0xa0
[ 61.156157] vfs_write+0xea/0x600
[ 61.156448] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.156706] ? find_held_lock+0x2b/0x80
[ 61.156902] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157141] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157469] ? from_pool+0x7d/0x190
[ 61.157664] ? srso_alias_return_thunk+0x5/0xfbef5
[ 61.157900] ? dept_enter+0x68/0xa0
[ 61.158098] ksys_write+0x76/0xf0
[ 61.158396] do_syscall_64+0xc2/0xf80
[ 61.158599] entry_SYSCALL_64_after_hwframe+0x77/0x7f
[ 61.158851] RIP: 0033:0x7fe2e7d58340
[ 61.159063] RSP: 002b:00007ffc370df3a8 EFLAGS: 00000202 ORIG_RAX: 0000000000000001
[ 61.159526] RAX: ffffffffffffffda RBX: 000000000000000e RCX: 00007fe2e7d58340
[ 61.159863] RDX: 000000000000000e RSI: 0000559c6fb13ed0 RDI: 0000000000000001
[ 61.160353] RBP: 0000559c6fb13ed0 R08: 0000000000000007 R09: 0000000000000073
[ 61.160695] R10: 0000000000001000 R11: 0000000000000202 R12: 000000000000000e
[ 61.161032] R13: 00007fe2e7e34760 R14: 000000000000000e R15: 00007fe2e7e2f9e0
[ 61.161490] </TASK>
[ 61.161639]
[ 61.161639] Showing all locks held in the system:
[ 61.161942] 1 lock held by khungtaskd/116:
[ 61.162142] #0: ffffffffa2380620 (rcu_read_lock){....}-{1:3}, at: debug_show_all_locks+0x36/0x1c0
[ 61.162884] 1 lock held by bash/915:
[ 61.163422] #0: ffff89af81beb570 (sb_writers#8){.+.+}-{0:0}, at: ksys_write+0x76/0xf0
[ 61.164077]
[ 61.164367] =============================================Quoted 9 lines
# echo FOLIO_LOCK_ABBA > /sys/kernel/debug/provoke-crash/DIRECT [ 182.473798] lkdtm: Performing direct entry FOLIO_LOCK_ABBA [ 182.475397] =================================================== [ 182.475637] DEPT: Circular dependency has been detected. [ 182.475774] 6.19.0-virtme #20 Not tainted [ 182.475885] --------------------------------------------------- [ 182.476054] summary [ 182.476112] --------------------------------------------------- [ 182.476259] *** AA DEADLOCK ***
At the moment, DEPT considers all the folio locks as the same class. I'm working on it to split it to multiple classes according to its usage. Once it's done, folio ABBA deadlock detection works well.
Quoted 67 lines
[ 182.476259] [ 182.476379] context A [ 182.476463] [S] (unknown)(pg_locked_map:0) [ 182.476604] [W] dept_page_wait_on_bit(pg_locked_map:0) [ 182.476744] [E] dept_page_clear_bit(pg_locked_map:0) [ 182.476884] [ 182.476943] [S]: start of the event context [ 182.477055] [W]: the wait blocked [ 182.477158] [E]: the event not reachable [ 182.477268] --------------------------------------------------- [ 182.477415] context A's detail [ 182.477504] --------------------------------------------------- [ 182.477652] context A [ 182.477711] [S] (unknown)(pg_locked_map:0) [ 182.477852] [W] dept_page_wait_on_bit(pg_locked_map:0) [ 182.477992] [E] dept_page_clear_bit(pg_locked_map:0) [ 182.478132] [ 182.478190] [S] (unknown)(pg_locked_map:0): [ 182.478300] (N/A) [ 182.478359] [ 182.478418] [W] dept_page_wait_on_bit(pg_locked_map:0): [ 182.478558] [<ffffffff8cf9a5ae>] kthread+0xfe/0x200 [ 182.478698] stacktrace: [ 182.478757] kthread+0xfe/0x200 [ 182.478867] ret_from_fork+0x29d/0x2e0 [ 182.478977] ret_from_fork_asm+0x1a/0x30 [ 182.479118] [ 182.479177] [E] dept_page_clear_bit(pg_locked_map:0): [ 182.479316] [<ffffffff8d989fe3>] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.479485] stacktrace: [ 182.479545] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.479685] kthread+0xfe/0x200 [ 182.479796] ret_from_fork+0x29d/0x2e0 [ 182.479906] ret_from_fork_asm+0x1a/0x30 [ 182.480045] --------------------------------------------------- [ 182.480192] information that might be helpful [ 182.480310] --------------------------------------------------- [ 182.480458] CPU: 2 UID: 0 PID: 915 Comm: lkdtm_folio_a Not tainted 6.19.0-virtme #20 PREEMPT(voluntary) [ 182.480464] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 182.480466] Call Trace: [ 182.480469] <TASK> [ 182.480473] dump_stack_lvl+0x69/0xa0 [ 182.480481] cb_check_dl+0x6be/0x760 [ 182.480492] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480497] ? lock_acquire+0x26b/0x2b0 [ 182.480507] bfs+0x138/0x1c0 [ 182.480511] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480522] add_dep+0xd6/0x1c0 [ 182.480527] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480531] ? __pfx_bfs_init_check_dl+0x10/0x10 [ 182.480535] ? __pfx_bfs_extend_dep+0x10/0x10 [ 182.480539] ? __pfx_bfs_dequeue_dep+0x10/0x10 [ 182.480543] ? __pfx_cb_check_dl+0x10/0x10 [ 182.480551] __dept_event+0x489/0x520 [ 182.480558] ? srso_alias_return_thunk+0x5/0xfbef5 [ 182.480564] ? lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.480570] dept_event+0x99/0xc0 [ 182.480581] folio_unlock+0x3c/0x60 [ 182.480587] lkdtm_folio_ab_kthread+0xb3/0x1a0 [ 182.480594] ? __pfx_lkdtm_folio_ab_kthread+0x10/0x10 [ 182.480598] kthread+0xfe/0x200 [ 182.480605] ? __pfx_kthread+0x10/0x10 [ 182.480613] ret_from_fork+0x29d/0x2e0 [ 182.480616] ? __pfx_kthread+0x10/0x10 [ 182.480622] ret_from_fork_asm+0x1a/0x30 [ 182.480644] </TASK>
Quoted 1 line
# echo FOLIO_MUTEX_LOCK_ABBA > /sys/kernel/debug/provoke-crash/DIRECT
Good to see DEPT works for this case.
Quoted 6 lines
[ 25.744189] lkdtm: Performing direct entry FOLIO_MUTEX_LOCK_ABBA [ 25.750265] =================================================== [ 25.750826] DEPT: Circular dependency has been detected. [ 25.750911] 6.19.0-virtme #20 Not tainted [ 25.750995] --------------------------------------------------- [ 25.751098] summary
Quoted 8 lines
[ 25.751149] --------------------------------------------------- [ 25.751252] *** DEADLOCK *** [ 25.751252] [ 25.751338] context A [ 25.751389] [S] lock(mutex_b:0) [ 25.751458] [W] dept_page_wait_on_bit(pg_lcked_map:0) [[ 25.751652] ? [E] unlock(mutex_b:0) 2[ 25.751781]
Remove the broken letters.
Quoted 196 lines
0[ 25.751861] context B 0[ 25.751937] [S] (unknown)(pg_locked_map:0)4 h[ 25.752088] [W] lock(mutex_b:0) [ 25.752175] [E] dept_page_clear_bit(pg_locked_map:0) [ 25.752265] [ 25.752315] [S]: start of the event context [ 25.752382] [W]: the wait blocked [ 25.752452] [E]: the event not reachable [ 25.752520] --------------------------------------------------- [ 25.752621] context A's detairl [ 25.752726] -o--------------------------------o------------------ t[ 25.752894] context A @[ 25.752960] [S] lock(mutexv_b:0) i[ 25.753074] [W] dept_page_rwait_on_bit(pg_locked_map:0) t[ 25.753210] [E] unlock(mutmex_b:0) e[ 25.753324] -[ 25.753394] [S] lock(mutex_b:n0): g[ 25.753513] [:<ffffffffb1989d90>] lkdtm_mutex_/folio_kthread+0x40/0xe0 [ 25.753705] stacktrace: [ 25.753757] h lkdtm_mutex_folio_kthread+0x40/0xe0 o[ 25.753890] m kthread+0xfee/0x200 /[ 25.754049] ret_from_fokrk+0x29d/0x2e0 i[ 25.754175] m ret_from_fo/rk_asm+0x1a/0x30s a[ 25.754390] [ 25.754496] [gW] dept_page_waiit_on_bit(pg_lockned_map:0): g[ 25.754678] [<ffffffffb0f9a5ae>] kthread+0xfe/0x200 [ 25.754796] s#tacktrace: [ 25.754874] kthread+0xfe/0x200 [ 25.754981] ret_from_fork+0x29d/0x2e0 [ 25.755039] ret_from_fork_asm+0x1a/0x30 [ 25.755130] [ 25.755185] [E] unlock(mutex_b:0): [ 25.755254] (N/A) [ 25.755304] --------------------------------------------------- [ 25.755406] context B's detail [ 25.755473] --------------------------------------------------- [ 25.755576] context B [ 25.755625] [S] (unknown)(pg_locked_map:0) [ 25.755724] [W] lock(mutex_b:0) [ 25.755791] [E] dept_page_clear_bit(pg_locked_map:0) [ 25.755893] [ 25.755955] [S] (unknown)(pg_locked_map:0): [ 25.756039] (N/A) [ 25.756102] [ 25.756164] [W] lock(mutex_b:0): [ 25.756248] [<ffffffffb1989e68>] lkdtm_folio_mutex_kthread+0x28/0xe0 [ 25.756376] stacktrace: [ 25.756438] lkdtm_folio_mutex_kthread+0x28/0xe0 [ 25.756545] kthread+0xfe/0x200 [ 25.756629] ret_from_fork+0x29d/0x2e0 [ 25.756726] ret_from_fork_asm+0x1a/0x30 [ 25.756853] [ 25.756918] [E] dept_page_clear_bit(pg_locked_map:0): [ 25.757081] [<ffffffffb1989e80>] lkdtm_folio_mutex_kthread+0x40/0xe0 [ 25.757241] stacktrace: [ 25.757315] lkdtm_folio_mutex_kthread+0x40/0xe0 [ 25.757449] kthread+0xfe/0x200 [ 25.757555] ret_from_fork+0x29d/0x2e0 [ 25.757672] ret_from_fork_asm+0x1a/0x30 [ 25.757816] --------------------------------------------------- [ 25.758014] information that might be helpful [ 25.758180] --------------------------------------------------- [ 25.758370] CPU: 15 UID: 0 PID: 922 Comm: lkdtm_mutex_fol Not tainted 6.19.0-virtme #20 PREEMPT(voluntary) [ 25.758381] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.2-debian-1.16.2-1 04/01/2014 [ 25.758386] Call Trace: [ 25.758391] <TASK> [ 25.758403] dump_stack_lvl+0x69/0xa0 [ 25.758424] cb_check_dl+0x6be/0x760 [ 25.758454] bfs+0x17d/0x1c0 [ 25.758459] ? srso_alias_return_thunk+0x5/0xfbef5 [ 25.758471] add_dep+0xd6/0x1c0 [ 25.758477] ? lkdtm_mutex_folio_kthread+0x40/0xe0 [ 25.758482] ? __pfx_bfs_init_check_dl+0x10/0x10 [ 25.758488] ? __pfx_bfs_extend_dep+0x10/0x10 [ 25.758492] ? __pfx_bfs_dequeue_dep+0x10/0x10 [ 25.758496] ? __pfx_cb_check_dl+0x10/0x10 [ 25.758505] __dept_wait+0x274/0x6a0 [ 25.758514] ? kthread+0xfe/0x200 [ 25.758520] ? __mutex_lock+0xae3/0x1230 [ 25.758532] ? srso_alias_return_thunk+0x5/0xfbef5 [ 25.758536] ? dept_enter+0x68/0xa0 [ 25.758545] ? kthread+0xfe/0x200 [ 25.758551] dept_wait+0xa7/0xc0 [ 25.758562] ? __pfx_lkdtm_mutex_folio_kthread+0x10/0x10 [ 25.758567] lkdtm_mutex_folio_kthread+0x9d/0xe0 [ 25.758573] kthread+0xfe/0x200 [ 25.758579] ? __pfx_kthread+0x10/0x10 [ 25.758587] ret_from_fork+0x29d/0x2e0 [ 25.758591] ? __pfx_kthread+0x10/0x10 [ 25.758597] ret_from_fork_asm+0x1a/0x30 [ 25.758619] </TASK># echo FOLIO_DEFERRED_EVENT_ABBA > /sys/kernel/debug/provoke-crash/DIRECT [ 100.907960] lkdtm: Performing direct entry FOLIO_DEFERRED_EVENT_ABBA [ 100.908476] lkdtm: [Waiter Thread] Securing the folio_lock... [ 100.908662] lkdtm: [Waiter Thread] Lock secured. Sleeping on wait queue. [ 100.908691] lkdtm: [Trigger] Calling wake_up() to initiate deferred deadlock. [ 100.908928] lkdtm: [Waker Context] Atomic callback triggered. Deferring work... [ 100.909265] lkdtm: [Worker Context] Attempting to acquire folio_lock... [ 151.210039] INFO: task kworker/14:1:124 blocked for more than 30 seconds. [ 151.210422] Not tainted 6.19.0-virtme #23 [ 151.210559] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 151.210695] task:kworker/14:1 state:D stack:14344 pid:124 tgid:124 ppid:2 task_flags:0x4208060 flags:0x00080000 [ 151.210880] Workqueue: events deferred_deadlocking [ 151.210979] Call Trace: [ 151.211029] <TASK> [ 151.211088] __schedule+0x5e7/0x1170 [ 151.211182] schedule+0x3a/0x130 [ 151.211256] io_schedule+0x46/0x70 [ 151.211324] folio_wait_bit_common+0x125/0x2d0 [ 151.211423] ? __pfx_wake_page_function+0x10/0x10 [ 151.211563] deferred_deadlocking+0x68/0x70 [ 151.211664] process_one_work+0x205/0x690 [ 151.211782] ? lock_is_held_type+0x9e/0x110 [ 151.211900] worker_thread+0x188/0x330 [ 151.211994] ? __pfx_worker_thread+0x10/0x10 [ 151.212111] kthread+0xfe/0x200 [ 151.212199] ? __pfx_kthread+0x10/0x10 [ 151.212295] ret_from_fork+0x2b2/0x2e0 [ 151.212385] ? __pfx_kthread+0x10/0x10 [ 151.212473] ret_from_fork_asm+0x1a/0x30 [ 151.212599] </TASK> [ 151.212676] INFO: task lkdtm_waiter:916 blocked for more than 30 seconds. [ 151.212844] Not tainted 6.19.0-virtme #23 [ 151.212955] "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message. [ 151.213105] task:lkdtm_waiter state:D stack:14936 pid:916 tgid:916 ppid:2 task_flags:0x208040 flags:0x00080000 [ 151.213309] Call Trace: [ 151.213372] <TASK> [ 151.213444] __schedule+0x5e7/0x1170 [ 151.213550] ? __pfx_lkdtm_waiter_thread+0x10/0x10 [ 151.213666] schedule+0x3a/0x130 [ 151.213776] lkdtm_waiter_thread+0xab/0x120 [ 151.213869] ? __pfx_wake_for_deferred_work+0x10/0x10 [ 151.213986] kthread+0xfe/0x200 [ 151.214077] ? __pfx_kthread+0x10/0x10 [ 151.214173] ret_from_fork+0x2b2/0x2e0 [ 151.214260] ? __pfx_kthread+0x10/0x10 [ 151.214347] ret_from_fork_asm+0x1a/0x30 [ 151.214472] </TASK> [ 151.214588] [ 151.214588] Showing all locks held in the system: [ 151.214766] 1 lock held by khungtaskd/116: [ 151.214878] #0: ffffffffb676d760 (rcu_read_lock){....}-{1:3}, at: debug_show_all_locks+0x36/0x1c0 [ 151.215078] 2 locks held by kworker/14:1/124: [ 151.215191] #0: ffff8ee181098948 ((wq_completion)events){+.+.}-{0:0}, at: process_one_work+0x55c/0x690 [ 151.215386] #1: ffffd401404a3e40 ((work_completion)(&deadlock_work)){+.+.}-{0:0}, at: process_one_work+0x1c4/0x690 [ 151.215619] [ 151.215677] =============================================Assisted-by Gemini:gemini-3.1-proCc Byungchul Park <[email protected]>Cc Yeoreum Yun <[email protected]>Signed-off-by Yunseong Kim <[email protected]>--- drivers/misc/lkdtm/Makefile | 1 + drivers/misc/lkdtm/core.c | 1 + drivers/misc/lkdtm/deadlock.c | 304 ++++++++++++++++++++++++ drivers/misc/lkdtm/lkdtm.h | 1 + tools/testing/selftests/lkdtm/tests.txt | 4 + 5 files changed, 311 insertions(+) create mode 100644 drivers/misc/lkdtm/deadlock.cdrivers/misc/lkdtm/Makefile +1
-0@@ -3,6 +3,7 @@ obj-$(CONFIG_LKDTM) += lkdtm.o lkdtm-$(CONFIG_LKDTM) += core.o lkdtm-$(CONFIG_LKDTM) += bugs.o +lkdtm-$(CONFIG_LKDTM) += deadlock.o lkdtm-$(CONFIG_LKDTM) += heap.o lkdtm-$(CONFIG_LKDTM) += perms.o lkdtm-$(CONFIG_LKDTM) += refcount.odrivers/misc/lkdtm/core.c +1
-0@@ -89,6 +89,7 @@ static struct crashpoint crashpoints[] = {/* List of possible types for crashes that can be triggered. */ static const struct crashtype_category *crashtype_categories[] = { &bugs_crashtypes, + &deadlock_crashtypes, &heap_crashtypes, &perms_crashtypes, &refcount_crashtypes,drivers/misc/lkdtm/deadlock.c new file +4
-0@@ -0,0 +1,304 @@ +// SPDX-License-Identifier: GPL-2.0 +/* +* This is for all the tests related to deadlock. +*/
drivers/misc/lkdtm/deadlock.c line 14
Quoted 10 lines
drivers/misc/lkdtm/deadlock.c new file +10
-0+#include "lkdtm.h" +#include <linux/mm.h> +#include <linux/kthread.h> +#include <linux/mutex.h> +#include <linux/workqueue.h> +#include <linux/delay.h> +#include <linux/wait.h> +#include <linux/pagemap.h> + +static struct folio *folio_common;
Either declare this way and alloc them at the very beginning of lkdtm:
static struct folio *folio_common;
static struct folio *folio_a;
static struct folio *folio_b;
Or use struct'ed parameter to pass the stack variables to kthread_run() e.g:
struct folio *folio_a = alloc();
struct folio *folio_b = alloc();
struct two_locks folios = { .a = folio_a; .b = folio b; };Quoted 117 lines
drivers/misc/lkdtm/deadlock.c new file +117
-0+ +/* +* Triggering a simple AA deadlock on a folio, Attempting to acquire the same +* folio twice in the same execution context, resulting in a self-deadlock. +*/ +static void lkdtm_FOLIO_LOCK_AA (void ) +{ + folio_common =folio_alloc (GFP_KERNEL | __GFP_ZERO,0 ); + +if (!folio_common) { +pr_err ("folio_alloc() failed.\n" ); +return ; + } + +folio_lock (folio_common); +folio_lock (folio_common); + +/* Unreachable */ +folio_unlock (folio_common); + +folio_put (folio_common); +} + +/* +* Attempting the 'AB' order for ABBA deadlock +*/ +static int lkdtm_folio_ab_kthread (void *folio_b) +{ +while (true) { +folio_lock (folio_common); +folio_lock ((struct folio *)folio_b); +folio_unlock ((struct folio *)folio_b); +folio_unlock (folio_common); + } + +return 0 ; +} + +/* +* Attempting the 'BA' order for ABBA deadlock +*/ +static int lkdtm_folio_ba_kthread (void *folio_b) +{ +while (true) { +folio_lock ((struct folio *)folio_b); +folio_lock (folio_common); +folio_unlock (folio_common); +folio_unlock ((struct folio *)folio_b); + } + +return 0 ; +} + +/* +* Spawning kthreads that attempt to acquire Waiter A and Waiter B in reverse +* order. Leading to a state where Thread A holds Waiter A and waits for +* Waiter B, while Thread B holds Waiter B and waits for Waiter A. +*/ +static void lkdtm_FOLIO_LOCK_ABBA (void ) +{ +struct folio *folio_b; +struct task_struct *t0, *t1; + + folio_common =folio_alloc (GFP_KERNEL | __GFP_ZERO,0 ); + folio_b =folio_alloc (GFP_KERNEL | __GFP_ZERO,0 ); + +if (!folio_common || !folio_b) { +pr_err ("folio_alloc() failed.\n" ); +return ; + } + + t0 =kthread_run (lkdtm_folio_ab_kthread, folio_b,"lkdtm_folio_a" ); + t1 =kthread_run (lkdtm_folio_ba_kthread, folio_b,"lkdtm_folio_b" ); + +if (IS_ERR (t0) ||IS_ERR (t1)) +pr_err ("failed to start kthread.\n" ); + +folio_put (folio_common); +folio_put (folio_b); +} + +DEFINE_MUTEX (mutex_b); + +/* Attempting 'folio_lock() A then Mutex B' order */ +static int lkdtm_folio_mutex_kthread (void *) +{ +while (true) { +folio_lock (folio_common); +mutex_lock (&mutex_b); +mutex_unlock (&mutex_b); +folio_unlock (folio_common); + } + +return 0 ; +} + +/* Attempting 'Mutex B then folio_lock() A' order */ +static int lkdtm_mutex_folio_kthread (void *) +{ +while (true) { +mutex_lock (&mutex_b); +folio_lock (folio_common); +folio_unlock (folio_common); +mutex_unlock (&mutex_b); + } + +return 0 ; +} + +/* Triggering ABBA deadlock between folio_lock() and mutex. */ +static void lkdtm_FOLIO_MUTEX_LOCK_ABBA (void ) +{ +struct task_struct *t0, *t1; + + folio_common =folio_alloc (GFP_KERNEL | __GFP_ZERO,0 ); + + t0 =kthread_run (lkdtm_folio_mutex_kthread,NULL ,"lkdtm_folio_mutex" );
drivers/misc/lkdtm/deadlock.c line 141
Quoted 10 lines
drivers/misc/lkdtm/deadlock.c new file +10
-0+ t1 =kthread_run (lkdtm_mutex_folio_kthread,NULL ,"lkdtm_mutex_folio" ); + +if (IS_ERR (t0) ||IS_ERR (t1)) +pr_err ("failed to start kthreads\n" ); + +folio_put (folio_common); +} + +/* +* Deferred AB-BA Deadlock Scenario
Can you explain what this test is for? Hard to understand to me. If you meant to demo ABBA deadlock where A is folio wait and B is general wait, I think this example looks too complicated. Can you make this straightforward?
My questions are:
1. What is exactly B?
2. Why do you use deferred workqueue?
Byungchul
Quoted 189 lines
drivers/misc/lkdtm/deadlock.c new file +163
-0+* +* Deferring a lock acquisition from an atomic wake-up callback to a +* sleepable workqueue context. +* +* ---------------------------------------------------------------------- +* 'lkdtm_waiter' kthread Waker kworker thread +* [Sleepable Context] (LKDTM Trigger) [Sleepable Context] +* | | | +* 1. folio_lock(folio_common) | | +* [Holds Folio] | | +* | | | +* [Waits for Wakeup] <--- 2. wake_up(&wq_deadlock) | +* | | | +* | 3. wake_for_deferred_work() | +* | [Inside wq_deadlock->lock] | +* | | | +* | 4. schedule_work() ---> | +* | | 5. deferred_deadlocking() +* | | | +* | | 6. folio_lock(folio_common) +* | | [Waits for Folio] +* | | | +* ---------------------------------------------------------------------- +*/ +static DECLARE_WAIT_QUEUE_HEAD (wq_deadlock); +static DECLARE_COMPLETION (waiter_ready); +static struct folio *folio_common; +static struct work_struct deadlock_work; + +/** +* deferred_deadlocking - The deferred task executed by a kworker thread. +* @work: The work structure associated with this task. +* +* Since this runs in a kworker thread, it is a safe sleepable context. +* Attempting to acquire the folio_lock here will not cause an atomic +* scheduling violation, but it will create a logical deadlock and a +* circular dependency. +*/ +static void deferred_deadlocking (struct work_struct *work) +{ +pr_info ("[Worker Context] Attempting to acquire folio_lock...\n" ); + +/* +* DEADLOCK POINT: +* The kworker blocks here indefinitely because the lkdtm_waiter +* thread holds the PG_locked bit of folio_common. +*/ +folio_lock (folio_common); + +/* Unreachable */ +folio_unlock (folio_common); + +folio_put (folio_common); + +wake_up_all (&wq_deadlock); +} + +/** +* wake_for_deferred_work - Invoking the deferred_waker_work(). +* @wq_entry: The wait queue entry being woken up. +* @mode: Wakeup mode (e.g., TASK_NORMAL). +* @sync: Indicates if the wakeup is synchronous. +* @key: Event-specific key passed to wake_up(). +* +* Return: Always 0, meaning the waiter is not woken up and +* remains in the wait queue. +*/ +static int wake_for_deferred_work (struct wait_queue_entry *wq_entry, +unsigned int mode,int sync,void *key) +{ +pr_emerg ( +"[Waker Context] Atomic callback triggered. Deferring work...\n" ); + +schedule_work (&deadlock_work); + +return 0 ; +} + +/** +* lkdtm_waiter_thread - The background thread holding the lock. +* +* It acquires the folio lock, signals readiness to the trigger process, +* and then goes to sleep on the custom wait queue. +* +* Return: 0 on exit (unreachable in successful deadlock). +*/ +static int lkdtm_waiter_thread (void *) +{ +struct wait_queue_entry custom_wait; + +init_waitqueue_func_entry (&custom_wait, wake_for_deferred_work); + +pr_info ("[Waiter Thread] Securing the folio_lock...\n" ); +folio_lock (folio_common); + +complete (&waiter_ready); + +pr_info ("[Waiter Thread] Lock secured. Sleeping on wait queue.\n" ); + +add_wait_queue (&wq_deadlock, &custom_wait); + +/* +* Manual sleep logic. We sleep without a condition because we +* expect the deferred work to eventually wake us up. +*/ +set_current_state (TASK_UNINTERRUPTIBLE ); +schedule (); + +/* Unreachable */ +__set_current_state (TASK_RUNNING ); +remove_wait_queue (&wq_deadlock, &custom_wait); +folio_unlock (folio_common); + +return 0 ; +} + +/* +* Spawns the waiter thread, and triggers the wait queue wakeup mechanism to +* initiate the deferred deadlock. +*/ +static void lkdtm_FOLIO_DEFERRED_EVENT_ABBA (void ) +{ +struct task_struct *waiter_task; + + folio_common =folio_alloc (GFP_KERNEL | __GFP_ZERO,0 ); +if (!folio_common) { +pr_err ("Failed to allocate folio.\n" ); +return ; + } + + +INIT_WORK (&deadlock_work, deferred_deadlocking); +reinit_completion (&waiter_ready); + + waiter_task =kthread_run (lkdtm_waiter_thread,NULL ,"lkdtm_waiter" ); +if (IS_ERR (waiter_task)) { +pr_err ("Failed to create waiter thread.\n" ); +folio_put (folio_common); +return ; + } + +wait_for_completion (&waiter_ready); + +pr_info ("[Trigger] Calling wake_up() to initiate deferred deadlock.\n" ); + +/* +* Triggers wake_for_deferred_work() in the current atomic context, +* which in turn schedules deferred_deadlocking(). +*/ +wake_up (&wq_deadlock); +} + +static struct crashtype crashtypes[] = { +CRASHTYPE (FOLIO_LOCK_AA ), +CRASHTYPE (FOLIO_LOCK_ABBA ), +CRASHTYPE (FOLIO_MUTEX_LOCK_ABBA ), +CRASHTYPE (FOLIO_DEFERRED_EVENT_ABBA ), +}; + +struct crashtype_category deadlock_crashtypes = { + .crashtypes = crashtypes, + .len =ARRAY_SIZE (crashtypes), +};drivers/misc/lkdtm/lkdtm.h +1
-0@@ -77,6 +77,7 @@ struct crashtype_category {/* Each category's crashtypes list. */ extern struct crashtype_category bugs_crashtypes; +extern struct crashtype_category deadlock_crashtypes;extern struct crashtype_category heap_crashtypes;extern struct crashtype_category perms_crashtypes;extern struct crashtype_category refcount_crashtypes;tools/testing/selftests/lkdtm/tests.txt +4
-1@@ -87,3 +87,7 @@ FORTIFY_STR_MEMBER detected buffer overflow FORTIFY_MEM_OBJECT detected buffer overflow FORTIFY_MEM_MEMBER detected field-spanning write PPC_SLB_MULTIHIT Recovered +#FOLIO_LOCK_AA Hangs the system +#FOLIO_LOCK_ABBA Hangs the system +#FOLIO_MUTEX_LOCK_ABBA Hangs the system +#FOLIO_DEFERRED_EVENT_ABBA Hangs the system --2.39.5
kernel test robot
build botkernel test robot noticed the following build warnings:
Hi Yunseong,
kernel test robot noticed the following build warnings:
[auto build test WARNING on char-misc/char-misc-testing]
[also build test WARNING on char-misc/char-misc-next char-misc/char-misc-linus kees/for-next/pstore kees/for-next/kspp shuah-kselftest/next shuah-kselftest/fixes soc/for-next linus/master v7.0-rc6 next-20260403] [If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Yunseong-Kim/lkdtm-Add-folio_lock-deadlock-scenarios/20260405-145013 base: char-misc/char-misc-testing patch link: https://lore.kernel.org/r/20260402143947.162844-1-ysk%40kzalloc.com patch subject: [PATCH] lkdtm: Add folio_lock deadlock scenarios config: s390-randconfig-002-20260405 (https://download.01.org/0day-ci/archive/20260405/[email protected]/config) compiler: s390-linux-gcc (GCC) 14.3.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260405/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All warnings (new ones prefixed by >>):
>> Warning: drivers/misc/lkdtm/deadlock.c:228 function parameter '' not described in 'lkdtm_waiter_thread' >> Warning: drivers/misc/lkdtm/deadlock.c:228 function parameter '' not described in 'lkdtm_waiter_thread'
Kconfig warnings: (for reference only)
WARNING: unmet direct dependencies detected for OF_GPIO
Depends on [n]: GPIOLIB [=y] && OF [=y] && HAS_IOMEM [=n]
Selected by [y]:
- REGULATOR_RT5133 [=y] && REGULATOR [=y] && I2C [=y] && GPIOLIB [=y] && OF [=y]
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
kernel test robot
build botkernel test robot noticed the following build errors:
Hi Yunseong,
kernel test robot noticed the following build errors:
[auto build test ERROR on char-misc/char-misc-testing]
[also build test ERROR on char-misc/char-misc-next char-misc/char-misc-linus kees/for-next/pstore kees/for-next/kspp shuah-kselftest/next shuah-kselftest/fixes soc/for-next linus/master v7.0-rc6 next-20260403] [If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Yunseong-Kim/lkdtm-Add-folio_lock-deadlock-scenarios/20260405-145013 base: char-misc/char-misc-testing patch link: https://lore.kernel.org/r/20260402143947.162844-1-ysk%40kzalloc.com patch subject: [PATCH] lkdtm: Add folio_lock deadlock scenarios config: riscv-randconfig-002-20260405 (https://download.01.org/0day-ci/archive/20260405/[email protected]/config) compiler: riscv64-linux-gcc (GCC) 9.5.0 reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260405/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All errors (new ones prefixed by >>):
drivers/misc/lkdtm/deadlock.c: In function 'lkdtm_folio_mutex_kthread':
>> drivers/misc/lkdtm/deadlock.c:99:38: error: parameter name omitted
99 | static int lkdtm_folio_mutex_kthread(void *)
| ^~~~~~
drivers/misc/lkdtm/deadlock.c: In function 'lkdtm_mutex_folio_kthread':
drivers/misc/lkdtm/deadlock.c:112:38: error: parameter name omitted
112 | static int lkdtm_mutex_folio_kthread(void *)
| ^~~~~~
drivers/misc/lkdtm/deadlock.c: In function 'lkdtm_waiter_thread':
drivers/misc/lkdtm/deadlock.c:228:32: error: parameter name omitted
228 | static int lkdtm_waiter_thread(void *)
| ^~~~~~
vim +99 drivers/misc/lkdtm/deadlock.c
97
98 /* Attempting 'folio_lock() A then Mutex B' order */
> 99 static int lkdtm_folio_mutex_kthread(void *)
100 {
101 while (true) {
102 folio_lock(folio_common);
103 mutex_lock(&mutex_b);
104 mutex_unlock(&mutex_b);
105 folio_unlock(folio_common);
106 }
107
108 return 0;
109 }
110
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki
kernel test robot
build botkernel test robot noticed the following build warnings:
Hi Yunseong,
kernel test robot noticed the following build warnings:
[auto build test WARNING on char-misc/char-misc-testing]
[also build test WARNING on char-misc/char-misc-next char-misc/char-misc-linus kees/for-next/pstore kees/for-next/kspp shuah-kselftest/next shuah-kselftest/fixes soc/for-next linus/master v7.0-rc6 next-20260403] [If your patch is applied to the wrong git tree, kindly drop us a note.
And when submitting patch, we suggest to use '--base' as documented in
https://git-scm.com/docs/git-format-patch#_base_tree_information]
url: https://github.com/intel-lab-lkp/linux/commits/Yunseong-Kim/lkdtm-Add-folio_lock-deadlock-scenarios/20260405-145013 base: char-misc/char-misc-testing patch link: https://lore.kernel.org/r/20260402143947.162844-1-ysk%40kzalloc.com patch subject: [PATCH] lkdtm: Add folio_lock deadlock scenarios config: s390-randconfig-001-20260405 (https://download.01.org/0day-ci/archive/20260405/[email protected]/config) compiler: clang version 23.0.0git (https://github.com/llvm/llvm-project c80443cd37b2e2788cba67ffa180a6331e5f0791) reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20260405/[email protected]/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <[email protected]>
| Closes: https://lore.kernel.org/oe-kbuild-all/[email protected]/
All warnings (new ones prefixed by >>):
>> drivers/misc/lkdtm/deadlock.c:99:44: warning: omitting the parameter name in a function definition is a C23 extension [-Wc23-extensions]
99 | static int lkdtm_folio_mutex_kthread(void *)
| ^
drivers/misc/lkdtm/deadlock.c:112:44: warning: omitting the parameter name in a function definition is a C23 extension [-Wc23-extensions]
112 | static int lkdtm_mutex_folio_kthread(void *)
| ^
drivers/misc/lkdtm/deadlock.c:228:38: warning: omitting the parameter name in a function definition is a C23 extension [-Wc23-extensions]
228 | static int lkdtm_waiter_thread(void *)
| ^
3 warnings generated.
vim +99 drivers/misc/lkdtm/deadlock.c
97
98 /* Attempting 'folio_lock() A then Mutex B' order */
> 99 static int lkdtm_folio_mutex_kthread(void *)
100 {
101 while (true) {
102 folio_lock(folio_common);
103 mutex_lock(&mutex_b);
104 mutex_unlock(&mutex_b);
105 folio_unlock(folio_common);
106 }
107
108 return 0;
109 }
110
Signature · 2 lines
0-DAY CI Kernel Test Service https://github.com/intel/lkp-tests/wiki