커널 로그는 다음과 같습니다. UI freeze 상태라 강제 커널 크래시를 유발시켜 램덤프를 추출했습니다.
[ 7164.614352] tegra sound_card.27: Ply shutdown, devnum 7
[ 7165.077580] ------------[ cut here ]------------
[ 7165.077807] WARNING: at /root/rpi/kernel/workqueue.c:2904 __cancel
_work_timer+0x12c/0x178()
[ 7165.087114] Modules linked in: texfat(PO)
[ 7165.092596] CPU: 3 PID: 1373 Comm: mmcqd/0 Tainted: P W O 3.10.44+ #1
[ 7165.101579] [<c0017344>] (unwind_backtrace+0x0/0x144) from [<c0013874>] (show_stack+0x20/0x24)
[ 7165.111750] [<c0013874>] (show_stack+0x20/0x24) from [<c0936350>] (dump_stack+0x20/0x28)
[ 7165.121489] [<c0936350>] (dump_stack+0x20/0x28) from [<c0028848>] (warn_slowpath_common+0x5c/0x7c)
[ 7165.132133] [<c0028848>] (warn_slowpath_common+0x5c/0x7c) from [<c0028894>] (warn_slowpath_null+0x2c/0x34)
[ 7165.143418] [<c0028894>] (warn_slowpath_null+0x2c/0x34) from [<c0048600>] (__cancel_work_timer+0x12c/0x178)
[ 7165.154654] [<c0048600>] (__cancel_work_timer+0x12c/0x178) from [<c0048668>] (cancel_delayed_work_sync+0x1c/0x20)
[ 7165.168097] [<c0048668>] (cancel_delayed_work_sync+0x1c/0x20) from [<c007b588>] (pm_qos_update_request+0x34/0xac)
[ 7165.178711] [<c007b588>] (pm_qos_update_request+0x34/0xac) from [<c064348c>] (sdhci_disable+0x2c/0x48)
[ 7165.189268] [<c064348c>] (sdhci_disable+0x2c/0x48) from [<c062e68c>] (mmc_release_host+0xa8/0xc0)
[ 7165.199855] [<c062e68c>] (mmc_release_host+0xa8/0xc0) from [<c0640794>] (mmc_blk_issue_rq+0x104/0x664)
[ 7165.210600] [<c0640794>] (mmc_blk_issue_rq+0x104/0x664) from [<c0641514>] (mmc_queue_thread+0xb4/0x14c)
[ 7165.221673] [<c0641514>] (mmc_queue_thread+0xb4/0x14c) from [<c004ef1c>] (kthread+0xc0/0xd0)
[ 7165.231685] [<c004ef1c>] (kthread+0xc0/0xd0) from [<c000f788>] (ret_from_fork+0x14/0x20)
[ 7165.241314] ---[ end trace b64c1dfefe8a5e9e ]---
각 cpu별 current task들의 backtrace
0 번 CPU만 key에 의한 강제 crash를 발생 시키기 위한 work thread가 돌고 있고 나머지 cpu들은 idle task였습니다.
PID: 21555 TASK: c3e76400 CPU: 0 COMMAND: "kworker/0:1" ? 강제 crash를 발생한 work
#0 [<c03df0c4>] (crash_kexec_odin) from [<c0931038>]
#1 [<c0931038>] (panic) from [<c0013b4c>]
#2 [<c0013b4c>] (die) from [<c0013ba8>]
#3 [<c0013ba8>] (arm_notify_die) from [<c0008370>]
#4 [<c0008370>] (do_undefinstr) from [<c000f2f0>]
PID: 0 TASK: d909e400 CPU: 1 COMMAND: "swapper/1"
#0 [<c00166b0>] (machine_crash_nonpanic_core) from [<c0098774>]
#1 [<c0098774>] (generic_smp_call_function_single_interrupt) from [<c00158ec>]
#2 [<c00158ec>] (handle_IPI) from [<c00086f0>]
#3 [<c00086f0>] (gic_handle_irq) from [<c000f244>]
#4 [<c000f244>] (__irq_svc) from [<c008dce4>]
PID: 0 TASK: d909ee00 CPU: 2 COMMAND: "swapper/2"
#0 [<c00166b0>] (machine_crash_nonpanic_core) from [<c0098774>]
#1 [<c0098774>] (generic_smp_call_function_single_interrupt) from [<c00158ec>]
#2 [<c00158ec>] (handle_IPI) from [<c00086f0>]
#3 [<c00086f0>] (gic_handle_irq) from [<c000f244>]
#4 [<c000f244>] (__irq_svc) from [<c008dce4>]
#5 [<c007e63c>] (cpu_startup_entry) from [<c092dc10>]
PID: 0 TASK: d92a0000 CPU: 3 COMMAND: "swapper/3"
#0 [<c00166b0>] (machine_crash_nonpanic_core) from [<c0098774>]
#1 [<c0098774>] (generic_smp_call_function_single_interrupt) from [<c00158ec>]
#2 [<c00158ec>] (handle_IPI) from [<c00086f0>]
#3 [<c00086f0>] (gic_handle_irq) from [<c000f244>]
#4 [<c000f244>] (__irq_svc) from [<c008dce4>]
task state 확인
system_server가 UNINTERRUPTIBLE 상태였습니다. 따라서, UI freeze 발생한 것으로 보입니다.
2611 2 3 d79d4600 UN 0.0 0 0 [jbd2/mmcblk0p29]
2627 2 2 d82dc600 UN 0.0 0 0 [jbd2/mmcblk0p10]
2630 2 1 d82dee00 UN 0.0 0 0 [jbd2/mmcblk0p22]
2977 1 3 d82c0a00 UN 0.0 1092 524 rctd
2993 1 2 d6763c00 UN 0.0 3576 776 sdcard
3053 2981 0 d84a6e00 UN 0.0 904 432 logcat
3066 2980 2 d7c91e00 UN 0.0 1312 876 kernel_logger
3070 2984 0 d7f2da00 UN 0.0 900 428 logcat
3071 2985 3 d7c91400 UN 0.0 916 440 logcat
3072 2982 0 d75a8a00 UN 0.0 900 428 logcat
3077 2986 3 d776b200 UN 0.1 3560 1144 tcpdump
3127 2758 1 d66ca800 UN 5.5 1075264 112420 system_server
3304 2 2 d6f69e00 UN 0.0 0 0 [ts_ldisc_tx_loo]
3305 1 3 d7cbe400 UN 0.1 5568 1248 nvm_server
3332 2758 2 d675e400 UN 5.5 1075264 112420 InputDispatcher
3546 2758 3 d7cba800 UN 5.5 1075264 112420 Binder_5
3556 2758 1 d6f6a800 UN 5.5 1075264 112420 watchdog
3598 2758 0 d79f8000 UN 3.5 986664 72460 com.lge.ime
4674 1 2 c6562800 UN 0.0 11324 576 thermalmgr
5103 1 0 c55f0000 UN 0.0 1040 576 dhcpcd
system_server Call Stack
system_server가 file open을 하는 도중, mutex_lock을 얻기 위해, UNINTERRUPTIBLE 상태로 들어간 상태에서 schedule된 상태입니다.
crash> bt 3127
PID: 3127 TASK: d66ca800 CPU: 1 COMMAND: "system_server"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c093e6e0>]
#2 [<c093e6e0>] (schedule_preempt_disabled) from [<c093cab4>]
#3 [<c093cab4>] (mutex_lock_nested) from [<c0145ed8>]
#4 [<c0145ed8>] (do_last) from [<c0146950>]
#5 [<c0146950>] (path_openat) from [<c014707c>]
#6 [<c014707c>] (do_filp_open) from [<c01372d0>]
#7 [<c01372d0>] (do_sys_open) from [<c0137398>]
#8 [<c0137398>] (sys_open) from [<c000f6c0>]
pc : [<400a0918>] lr : [<400ae009>] psr: 200f0010
sp : be807318 ip : 6d773a44 fp : be80736c
r10: 417ed7a0 r9 : 6d773a38 r8 : be807354
r7 : 00000005 r6 : 00000042 r5 : 00000180 r4 : 417ed178
r3 : be807328 r2 : 00000180 r1 : 00020042 r0 : 88d01960
Flags: nzCv IRQs on FIQs on Mode USER_32 ISA ARM
해당 mutex lock 조사
crash> struct mutex 0xD8F0B900
struct mutex {
count = {
counter = -1 // lock 걸려있음
},
<생략>
wait_list = {
next = 0xd626fdb0,
prev = 0xd626fdb0
},
owner = 0xc4815000, // lock의 owner임
name = 0x0,
magic = 0xd8f0b900,
dep_map = {
key = 0xc0fa51e8 <ext4_fs_type+92>,
class_cache = {0xc124114c <lock_classes+231200>, 0xc1241bec <lock_classes+233920>},
name = 0xc0ce3190 "&type->i_mutex_dir_key"
}
}
owner task의 backtrace
PID: 21441 TASK: c4815000 CPU: 4 COMMAND: "BatteryStats-Wr"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c01eb6a0>]
#2 [<c01eb6a0>] (start_this_handle) from [<c01ebb44>]
#3 [<c01ebb44>] (jbd2__journal_start) from [<c01d3810>]
#4 [<c01d3810>] (__ext4_journal_start_sb) from [<c01a66f8>]
#5 [<c01a66f8>] (__ext4_new_inode) from [<c01b5bd4>]
#6 [<c01b5bd4>] (ext4_create) from [<c0143dcc>]
#7 [<c0143dcc>] (vfs_create) from [<c01464a0>]
#8 [<c01464a0>] (do_last) from [<c0146950>]
#9 [<c0146950>] (path_openat) from [<c014707c>]
#10 [<c014707c>] (do_filp_open) from [<c01372d0>]
#11 [<c01372d0>] (do_sys_open) from [<c0137398>]
#12 [<c0137398>] (sys_open) from [<c000f6c0>]
pc : [<400a0918>] lr : [<400ae009>] psr: 20070010
sp : 7f639ad0 ip : 796fbef4 fp : 7f639b24
r10: 87b27138 r9 : 796fbee8 r8 : 7f639b0c
r7 : 00000005 r6 : 00000241 r5 : 00000180 r4 : 86d5cd08
r3 : 7f639ae0 r2 : 00000180 r1 : 00020241 r0 : 80b05e48
Flags: nzCv IRQs on FIQs on Mode USER_32 ISA ARM
0xc4815000 task(PID ? 21441, BatteryStats-Wr ) 도 UNINTERRUPTIBLE 상태임
21441 2758 4 c4815000 UN 5.5 1075264 112420 BatteryStats-Wr
BatteryStats-Wr task가 mutex lock을 accquire한 상태에서, TASK_UNINTERRUPTIBLE로 들어갔고,
system_server는 같은 mutex lock을 accquire하기 위해 대기중 인 상태라 UI Freeze가 발생한 것으로 보입니다.
static int start_this_handle(journal_t *journal, handle_t *handle) {
<생략>
spin_lock(&transaction->t_handle_lock);
needed = transaction->t_outstanding_credits + nblocks;
if (needed > journal->j_max_transaction_buffers) {
DEFINE_WAIT(wait);
jbd_debug(2, "Handle %p starting new commit...\n", handle);
spin_unlock(&transaction->t_handle_lock);
prepare_to_wait(&journal->j_wait_transaction_locked, &wait, TASK_UNINTERRUPTIBLE); //--> 여기서 UN으로 들어감
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
schedule();
finish_wait(&journal->j_wait_transaction_locked, &wait);
goto repeat;
}
}
TASK_UNINTERRUPTIBLE 상태인 다른 몇몇 task의 backtrace를 조사해 봤습니다.
PID: 2611 TASK: d79d4600 CPU: 3 COMMAND: "jbd2/mmcblk0p29"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c01ee80c>]
#2 [<c01ee80c>] (jbd2_journal_commit_transaction) from [<c01f4794>]
#3 [<c01f4794>] (kjournald2) from [<c004ef1c>]
#4 [<c004ef1c>] (kthread) from [<c000f788>]
아래의 코드에서 UNINTERRUPTIBLE로 상태로 변경 후 schedule됩니다.
void jbd2_journal_commit_transaction(journal_t *journal){
// snip
while (atomic_read(&commit_transaction->t_updates)) {
DEFINE_WAIT(wait);
prepare_to_wait(&journal->j_wait_updates, &wait, TASK_UNINTERRUPTIBLE);
if (atomic_read(&commit_transaction->t_updates)) {
spin_unlock(&commit_transaction->t_handle_lock);
write_unlock(&journal->j_state_lock);
schedule();
write_lock(&journal->j_state_lock);
spin_lock(&commit_transaction->t_handle_lock);
}
finish_wait(&journal->j_wait_updates, &wait);
} // snip
아래와 같이 commit_transaction->t_updates가 2입니다.
crash> struct transaction_s.t_updates 0xD6858D80
t_updates = {
counter = 2
}
BatteryStats-Wr task가 mutex lock을 accquire한 상태에서, TASK_UNINTERRUPTIBLE로 들어갔고, system_server는 같은 mutex lock을 accquire하기 위해 대기중 인 상태라 UI Freeze가 발생한 것으로 보입니다.
static int start_this_handle(journal_t *journal, handle_t *handle) {
<생략>
spin_lock(&transaction->t_handle_lock);
needed = transaction->t_outstanding_credits + nblocks;
if (needed > journal->j_max_transaction_buffers) {
DEFINE_WAIT(wait);
jbd_debug(2, "Handle %p starting new commit...\n", handle);
spin_unlock(&transaction->t_handle_lock);
prepare_to_wait(&journal->j_wait_transaction_locked, &wait, TASK_UNINTERRUPTIBLE); //<<-- 여기서 UN으로 들어감
__log_start_commit(journal, transaction->t_tid);
spin_unlock(&journal->j_state_lock);
schedule();
finish_wait(&journal->j_wait_transaction_locked, &wait);
goto repeat;
}
}
.
jbd2/mmcblk0p10 task를 보면, 아래의 코드에서 UNINTERRUPTIBLE로 상태로 변경 후 schedule됐는데, io buffer가 lock상태여서, lock이 풀릴때까지 대기 상태에 있는 것입니다.
void jbd2_journal_commit_transaction(journal_t *journal) {
<생략>
jh = commit_transaction->t_iobuf_list->b_tprev;
bh = jh2bh(jh);
if (buffer_locked(bh)) {
wait_on_buffer(bh);
goto wait_for_iobuf;
}
<생략>
}
PID: 2627 TASK: d82dc600 CPU: 2 COMMAND: "jbd2/mmcblk0p10"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c093e3d4>]
#2 [<c093e3d4>] (io_schedule) from [<c0168084>]
#3 [<c0168084>] (sleep_on_buffer) from [<c093b784>]
#4 [<c093b784>] (__wait_on_bit) from [<c093b834>]
#5 [<c093b834>] (out_of_line_wait_on_bit) from [<c0168064>]
#6 [<c0168064>] (__wait_on_buffer) from [<c01ef218>]
#7 [<c01ef218>] (jbd2_journal_commit_transaction) from [<c01f4794>]
#8 [<c01f4794>] (kjournald2) from [<c004ef1c>]
#9 [<c004ef1c>] (kthread) from [<c000f788>]
TASK_UNINTERRUPTIBLE상태의 task중 아래와 같이 대부분의 task들이 file관련 동작중 TASK_UNINTERRUPTIBLE상태로 진입한 상태입니다.
PID: 2977 TASK: d82c0a00 CPU: 3 COMMAND: "rctd"
PID: 2993 TASK: d6763c00 CPU: 2 COMMAND: "sdcard"
PID: 3053 TASK: d84a6e00 CPU: 0 COMMAND: "logcat"
PID: 3070 TASK: d7f2da00 CPU: 0 COMMAND: "logcat"
PID: 3071 TASK: d7c91400 CPU: 3 COMMAND: "logcat"
PID: 3072 TASK: d75a8a00 CPU: 0 COMMAND: "logcat"
PID: 3077 TASK: d776b200 CPU: 3 COMMAND: "tcpdump"
PID: 3127 TASK: d66ca800 CPU: 1 COMMAND: "system_server"
PID: 3305 TASK: d7cbe400 CPU: 3 COMMAND: "nvm_server"
PID: 3332 TASK: d675e400 CPU: 2 COMMAND: "InputDispatcher"
PID: 3546 TASK: d7cba800 CPU: 3 COMMAND: "Binder_5"
PID: 3556 TASK: d6f6a800 CPU: 1 COMMAND: "watchdog"
PID: 5103 TASK: c55f0000 CPU: 0 COMMAND: "dhcpcd"
PID: 5165 TASK: c5008a00 CPU: 1 COMMAND: "GC"
PID: 5738 TASK: c59e3200 CPU: 2 COMMAND: "Thread-315"
PID: 8773 TASK: c5344600 CPU: 3 COMMAND: "kworker/u16:6"
PID: 21359 TASK: c4826e00 CPU: 2 COMMAND: "OMXCallbackDisp"
PID: 21413 TASK: d6e48000 CPU: 0 COMMAND: "Thread-220"
PID: 21416 TASK: c9ee0a00 CPU: 1 COMMAND: "referencesProto"
PID: 21424 TASK: d84a4600 CPU: 0 COMMAND: "Compiler"
PID: 21431 TASK: c62c1e00 CPU: 4 COMMAND: "UsageStatsServi"
PID: 21441 TASK: c4815000 CPU: 4 COMMAND: "BatteryStats-Wr"
다시 최초의 log를 리뷰하니 PID 1373, mmcqd/0 task에서 WARNING 메시지 출력합니다.
[ 7165.077807] WARNING: at /root/pi/kernel/workqueue.c:2904 __cancel_work_timer+0x12c/0x178()
[ 7165.087114] Modules linked in: texfat(PO)
[ 7165.092596] CPU: 3 PID: 1373 Comm: mmcqd/0 Tainted: P W O 3.10.44+ #1
[ 7165.101579] [<c0017344>] (unwind_backtrace+0x0/0x144) from [<c0013874>] (show_stack+0x20/0x24)
[ 7165.111750] [<c0013874>] (show_stack+0x20/0x24) from [<c0936350>] (dump_stack+0x20/0x28)
[ 7165.121489] [<c0936350>] (dump_stack+0x20/0x28) from [<c0028848>] (warn_slowpath_common+0x5c/0x7c)
[ 7165.132133] [<c0028848>] (warn_slowpath_common+0x5c/0x7c) from [<c0028894>] (warn_slowpath_null+0x2c/0x34)
[ 7165.143418] [<c0028894>] (warn_slowpath_null+0x2c/0x34) from [<c0048600>] (__cancel_work_timer+0x12c/0x178)
[ 7165.154654] [<c0048600>] (__cancel_work_timer+0x12c/0x178) from [<c0048668>] (cancel_delayed_work_sync+0x1c/0x20)
[ 7165.168097] [<c0048668>] (cancel_delayed_work_sync+0x1c/0x20) from [<c007b588>] (pm_qos_update_request+0x34/0xac)
[ 7165.178711] [<c007b588>] (pm_qos_update_request+0x34/0xac) from [<c064348c>] (sdhci_disable+0x2c/0x48)
[ 7165.189268] [<c064348c>] (sdhci_disable+0x2c/0x48) from [<c062e68c>] (mmc_release_host+0xa8/0xc0)
[ 7165.199855] [<c062e68c>] (mmc_release_host+0xa8/0xc0) from [<c0640794>] (mmc_blk_issue_rq+0x104/0x664)
[ 7165.210600] [<c0640794>] (mmc_blk_issue_rq+0x104/0x664) from [<c0641514>] (mmc_queue_thread+0xb4/0x14c)
[ 7165.221673] [<c0641514>] (mmc_queue_thread+0xb4/0x14c) from [<c004ef1c>] (kthread+0xc0/0xd0)
[ 7165.231685] [<c004ef1c>] (kthread+0xc0/0xd0) from [<c000f788>] (ret_from_fork+0x14/0x20)
mmcqd관련 task의 backtrace 조사한 결과를 정리하면 다음과 같습니다.
01 | mmcqd/0 task의 backtrace를 보면, WARNING을 출력했던 backtrace와 동일한 backtrace를 가지고 있음을 알 수 있음.
02 | mmcqd/0는 mmc 동작중 pm_qos를 통해 boosting을 했고, 완료 이후, pm_qos설정을 이전으로 돌리기 위한 작업을 진행하는 과정
03 | sdhci_disable () -> pm_qos_update_request () -> cancel_delayed_work_sync () -> __cancel_work_timer () -> schedule ()
PID: 1373 TASK: d7d29400 CPU: 7 COMMAND: "mmcqd/0"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c0048610>]
#2 [<c0048610>] (__cancel_work_timer) from [<c0048668>]
#3 [<c0048668>] (cancel_delayed_work_sync) from [<c007b588>]
#4 [<c007b588>] (pm_qos_update_request) from [<c064348c>]
#5 [<c064348c>] (sdhci_disable) from [<c062e68c>]
#6 [<c062e68c>] (mmc_release_host) from [<c0640794>]
#7 [<c0640794>] (mmc_blk_issue_rq) from [<c0641514>]
#8 [<c0641514>] (mmc_queue_thread) from [<c004ef1c>]
#9 [<c004ef1c>] (kthread) from [<c000f788>]
PID: 1374 TASK: d7d29e00 CPU: 0 COMMAND: "mmcqd/0boot0"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c0641580>]
#2 [<c0641580>] (mmc_queue_thread) from [<c004ef1c>]
#3 [<c004ef1c>] (kthread) from [<c000f788>]
PID: 1375 TASK: d79d6e00 CPU: 0 COMMAND: "mmcqd/0boot1"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c0641580>]
#2 [<c0641580>] (mmc_queue_thread) from [<c004ef1c>]
#3 [<c004ef1c>] (kthread) from [<c000f788>]
PID: 1376 TASK: d79d6400 CPU: 2 COMMAND: "mmcqd/0rpmb"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c0641580>]
#2 [<c0641580>] (mmc_queue_thread) from [<c004ef1c>]
#3 [<c004ef1c>] (kthread) from [<c000f788>]
PID: 1472 TASK: d8222800 CPU: 2 COMMAND: "mmcqd/1"
#0 [<c093de28>] (__schedule) from [<c093e324>]
#1 [<c093e324>] (schedule) from [<c0641580>]
#2 [<c0641580>] (mmc_queue_thread) from [<c004ef1c>]
#3 [<c004ef1c>] (kthread) from [<c000f788>]
__cancel_work_timer () 함수에서 mmcqd 프로세스가 TASK_INERRUPTIBLE로 진입 후 schedule() 됩니다.
static bool __cancel_work_timer(struct work_struct *work, bool is_dwork) {
<생략>
do {
ret = try_to_grab_pending(work, is_dwork, &flags);
if (unlikely(ret == -ENOENT)) {
if (!flush_work(work)) {
if (work_is_canceling(work)) {
set_current_state(TASK_INTERRUPTIBLE);
schedule(); //<<--INTERRUPTIBLE로 진입후 스케쥴
__set_current_state(TASK_RUNNING);
}
}
}
} while (unlikely(ret < 0));
<생략> }
최초 로그의 WARNING이 출력된 시간과 비슷(7165s) 합니다.
crash> struct task_struct.se 0xd7d29400
se = {
load = {
weight = 1024,
inv_weight = 4194304
},
run_node = {
__rb_parent_color = 1,
rb_right = 0x0,
rb_left = 0x0
},
group_node = {
next = 0xd7d2944c,
prev = 0xd7d2944c
},
on_rq = 0,
exec_start = 7165247678131, //<<-- 최초 로그의 WARNING이 출력된 시간과 비슷함(7165s)
task_struct.se.exec_start는 아래의 코드에서 update가 되도록 되어있습니다.
static void update_curr(struct cfs_rq *cfs_rq){ <생략>
curr->exec_start = now;
<생략>
}
emmc 관련 동작을 하는 mmcqd/0 가 7165s wakeup되어 running중, TASK_INTERRUPTIBLE 상태로 schedule이 되고, 다시 wakeup되지 않습니다.
1. file open/read/write 및 jonrnaling 등의 동작을 시도한 task들이 전부 TASK_UNINTERRUPTIBLE상태로 들어가게 됨.
2. 따라서, UI Freeze로 현상이 나타남.
다음 패치를 적용한 후 UI Freeze 현상은 사라졌습니다.
diff --git a/kernel/workqueue.c b/kernel/workqueue.c
index e1cae9d..cb3a27b 100644
--- a/kernel/workqueue.c
+++ b/kernel/workqueue.c
@@ -2759,6 +2759,7 @@ static bool __cancel_work_timer(struct work_struct *work, bool is_dwork)
* If someone else is canceling, wait for the same event it
* would be waiting for before retrying.
*/
- if (unlikely(ret == -ENOENT))
- flush_work(work);
+ if (unlikely(ret == -ENOENT)) {
+ if (!flush_work(work) && current->policy == SCHED_FIFO && work_is_canceling(work)) {
+ struct sched_param old, new;
+ new.sched_priority = 0;
+ sched_setscheduler(current, SCHED_NORMAL, &new);
+ schedule();
+ sched_setscheduler(current, SCHED_FIFO, &old);
+ }
+ }
"이 포스팅이 유익하다고 생각되시면 댓글로 응원해주시면 감사하겠습니다.
혹시 글을 읽고 궁금점이 있으면 댓글로 질문 남겨주세요. 상세한 답글 올려 드리겠습니다!"
