-
Notifications
You must be signed in to change notification settings - Fork 169
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Add drgn helpers and script for locking primitives #387
Open
imran-kn
wants to merge
15
commits into
osandov:main
Choose a base branch
from
imran-kn:lock_helpers_work
base: main
Could not load branches
Branch not found: {{ refName }}
Loading
Could not load tags
Nothing to show
Loading
Are you sure you want to change the base?
Some commits from the old base branch may be removed from the timeline,
and old review comments may become outdated.
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sometimes it is useful to get cpu that owns a given per-cpu variable. For example osq locks and qspinlocks use per-cpu entities when a CPU needs to spin, while waiting to acquire the lock. This happens when there are multiple spinners for a given lock. In such cases, this helper can let us know the spinners for a given qspinlock or osq lock. Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
These don't have tets cases so can't go under helpers. I have put these under contrib so that they are available for use. python3 -m drgn -s vmlinux -c vmcore contrib/locks.py --help usage: locks.py [-h] lock_type info_type [locks ...] positional arguments: lock_type type of lock i.e mutex. semaphore, rwsemaphore etc. info_type "owner" or "waiter" or "all" locks list of lock addresses options: -h, --help show this help message and exit For example following command will give us call stack for owner waiters of specified mutex(es): python3 -m drgn -s vmlinux -c vmcore contrib/locks.py mutex all ffffffffc0143400 .................. Dumping call stack for waiter of mutex: ffffffffc0143400 call stack for pid: 215 #0 context_switch (kernel/sched/core.c:5238:2) osandov#1 __schedule (kernel/sched/core.c:6551:8) osandov#2 schedule (kernel/sched/core.c:6627:3) osandov#3 schedule_preempt_disabled (kernel/sched/core.c:6686:2) osandov#4 __mutex_lock_common (kernel/locking/mutex.c:679:3) osandov#5 __mutex_lock (kernel/locking/mutex.c:747:9) osandov#6 0xffffffffc01411f6 .................. call stack for pid: 216 #0 context_switch (kernel/sched/core.c:5238:2) osandov#1 __schedule (kernel/sched/core.c:6551:8) osandov#2 schedule (kernel/sched/core.c:6627:3) osandov#3 schedule_preempt_disabled (kernel/sched/core.c:6686:2) osandov#4 __mutex_lock_common (kernel/locking/mutex.c:679:3) osandov#5 __mutex_lock (kernel/locking/mutex.c:747:9) osandov#6 0xffffffffc014112b .................. Dumping call stack for owner of mutex: ffffffffc0143400 call stack for pid: 214 #0 delay_tsc (arch/x86/lib/delay.c:79:3) osandov#1 0xffffffffc0141308 .................. Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
python3 -m drgn -s vmlinux -c vmcore contrib/locks.py --help usage: locks.py [-h] lock_type info_type [locks ...] positional arguments: lock_type type of lock i.e mutex. semaphore, rwsemaphore etc. info_type "owner" or "waiter" or "all" locks list of lock addresses options: -h, --help show this help message and exit For example following command will give us call stack for owner waiters of specified mutex(es): python3 -m drgn -s vmlinux -c vmcore contrib/locks.py semaphore waiter ffffffffc0097340 Dumping call stack for waiter(s) of semaphore: ffffffffc0097340 call stack for pid: 1178 #0 context_switch (kernel/sched/core.c:2811:2) osandov#1 __schedule (kernel/sched/core.c:3387:8) osandov#2 schedule (kernel/sched/core.c:3431:3) osandov#3 schedule_timeout (kernel/time/timer.c:1724:3) osandov#4 __down_common (kernel/locking/semaphore.c:221:13) osandov#5 __down (kernel/locking/semaphore.c:238:2) osandov#6 down (kernel/locking/semaphore.c:62:3) osandov#7 0xffffffffc0095045 .................. call stack for pid: 1180 #0 context_switch (kernel/sched/core.c:2811:2) osandov#1 __schedule (kernel/sched/core.c:3387:8) osandov#2 schedule (kernel/sched/core.c:3431:3) osandov#3 schedule_timeout (kernel/time/timer.c:1724:3) osandov#4 __down_common (kernel/locking/semaphore.c:221:13) osandov#5 __down (kernel/locking/semaphore.c:238:2) osandov#6 down (kernel/locking/semaphore.c:62:3) osandov#7 0xffffffffc0095045 ............. Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
Signed-off-by: Imran Khan <[email protected]>
Also include its test. Signed-off-by: Imran Khan <[email protected]>
Add some ready-made helpers for rwsem and make the script modular so that options can be added/removed for one lock type, independently from other lock types. Some example of using the script have been given below: python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py -h usage: drgn script to dump lock information [-h] {mutex,semaphore,rwsem} ... options: -h, --help show this help message and exit subcommands: {mutex,semaphore,rwsem} mutex get mutex info. semaphore get semaphore info. rwsem get read-write semaphore info. python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py semaphore -h usage: drgn script to dump lock information semaphore [-h] [--info | --waiter-list | --waiter-callstack] [locks ...] positional arguments: locks list of lock addresses options: -h, --help show this help message and exit --info dump given semaphore's info like waiter(s) etc. --waiter-list provide a list, of waiters of given semaphore(s) --waiter-callstack provide callstack of all waiters of given semaphore(s) python3 -m drgn -s vmlinux -c vmcore contrib/locks.py semaphore --waiter-list ffffffffc0097340 The waiters of semaphore: ffffffffc0097340 are as follows: (struct task_struct *)0xffff992afc618000 pid: 1178 state: D (struct task_struct *)0xffff992afc7ae200 pid: 1181 state: D (struct task_struct *)0xffff992afc61ee40 pid: 1179 state: D (struct task_struct *)0xffff992afc619880 pid: 1180 state: D python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py semaphore --waiter-list ffffffffc0262340 The waiters of semaphore: ffffffffc0262340 are as follows: (struct task_struct *)0xffff96c3fc6ec980 pid: 1178 state: S (struct task_struct *)0xffff96c3fc6eee40 pid: 1177 state: S (struct task_struct *)0xffff96c3fc5a55c0 pid: 1175 state: S (struct task_struct *)0xffff96c3fc6ee200 pid: 1176 state: S python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py semaphore --waiter-callstack ffffffffc0262340 Dumping call stack for waiter(s) of semaphore: ffffffffc0262340 call stack for pid: 1178 #0 context_switch (kernel/sched/core.c:2811:2) osandov#1 __schedule (kernel/sched/core.c:3387:8) osandov#2 schedule (kernel/sched/core.c:3431:3) osandov#3 schedule_timeout (kernel/time/timer.c:1724:3) osandov#4 __down_common (kernel/locking/semaphore.c:221:13) osandov#5 __down_interruptible (kernel/locking/semaphore.c:243:9) osandov#6 down_interruptible (kernel/locking/semaphore.c:85:12) osandov#7 0xffffffffc0260045 ..................... call stack for pid: 1176 #0 context_switch (kernel/sched/core.c:2811:2) osandov#1 __schedule (kernel/sched/core.c:3387:8) osandov#2 schedule (kernel/sched/core.c:3431:3) osandov#3 schedule_timeout (kernel/time/timer.c:1724:3) osandov#4 __down_common (kernel/locking/semaphore.c:221:13) osandov#5 __down_interruptible (kernel/locking/semaphore.c:243:9) osandov#6 down_interruptible (kernel/locking/semaphore.c:85:12) osandov#7 0xffffffffc0260045 call stack for pid: 1176 ..................... python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py mutex -h usage: drgn script to dump lock information mutex [-h] [--info | --waiter-list | --waiter-callstack | --owner-callstack] [locks ...] positional arguments: locks list of lock addresses options: -h, --help show this help message and exit --info dump given mutex's info like owner, waiter(s) etc. --waiter-list provide a list, of waiters of given mutex(es) --waiter-callstack provide callstack of all waiters of given mutex(es) --owner-callstack provide callstack of owner of given mutex(es) python3 -m drgn -s vmlinux -c vmcore contrib/locks.py mutex --info ffffffffc0143400 mutex: ffffffffc0143400 is owned by (struct task_struct *)0xffffa1fe42393900 pid: 214 state: R The waiters of mutex: ffffffffc0143400 are as follows: (struct task_struct *)0xffffa1fe42395580 pid: 215 state: D (struct task_struct *)0xffffa1fe42a58000 pid: 216 state: D (struct task_struct *)0xffffa1fe42389c80 pid: 217 state: D python3 -m drgn -s vmlinux -c vmcore contrib/locks.py mutex --waiter-list ffffffffc0143400 The waiters of mutex: ffffffffc0143400 are as follows: (struct task_struct *)0xffffa1fe42395580 pid: 215 state: D (struct task_struct *)0xffffa1fe42a58000 pid: 216 state: D (struct task_struct *)0xffffa1fe42389c80 pid: 217 state: D python3 -m drgn -s vmlinux -c vmcore contrib/locks.py mutex --waiter-callstack ffffffffc0143400 Dumping call stack for waiter of mutex: ffffffffc0143400 call stack for pid: 215 #0 context_switch (kernel/sched/core.c:5238:2) osandov#1 __schedule (kernel/sched/core.c:6551:8) osandov#2 schedule (kernel/sched/core.c:6627:3) osandov#3 schedule_preempt_disabled (kernel/sched/core.c:6686:2) osandov#4 __mutex_lock_common (kernel/locking/mutex.c:679:3) osandov#5 __mutex_lock (kernel/locking/mutex.c:747:9) osandov#6 0xffffffffc01411f6 ........................ python3 -m drgn -s vmlinux -c vmcore-multiple-readers contrib/locks.py rwsem -h usage: drgn script to dump lock information rwsem [-h] [--info | --waiter-list | --waiter-callstack | --owner-callstack] [locks ...] positional arguments: locks list of lock addresses options: -h, --help show this help message and exit --info dump given rwsem's info like owner, waiter(s) etc. --waiter-list provide a list, of waiters of given rwsem(s) --waiter-callstack provide callstack of all waiters of given rwsem(s) --owner-callstack provide callstack of owner of given rwsem(s) python3 -m drgn -s vmlinux -c vmcore-reader-writer-reader-reader contrib/locks.py rwsem --info ffffffffc036d3c0 rwsem: ffffffffc036d3c0 is owned by one or more readers. The waiters of rwsem are as follows: (struct task_struct *)0xffff9e593d4e3100: (pid)1175: type: down_write state: D (struct task_struct *)0xffff9e593d4e1880: (pid)1176: type: down_read state: D (struct task_struct *)0xffff9e593d4e6200: (pid)1177: type: down_read state: D python3 -m drgn -s vmlinux -c vmcore-reader-writer-reader-reader contrib/locks.py rwsem --waiter-callstack ffffffffc036d3c0 Dumping call stack for waiter of rwsem: ffffffffc036d3c0 call stack for pid: 1175 #0 context_switch (kernel/sched/core.c:2814:2) osandov#1 __schedule (kernel/sched/core.c:3389:8) osandov#2 schedule (kernel/sched/core.c:3433:3) osandov#3 __rwsem_down_write_failed_common (kernel/locking/rwsem-xadd.c:588:4) osandov#4 call_rwsem_down_write_failed+0x13/0x1f (arch/x86/lib/rwsem.S:105) osandov#5 __down_write (./arch/x86/include/asm/rwsem.h:126:2) osandov#6 down_write (kernel/locking/rwsem.c:56:2) osandov#7 0xffffffffc036b2b6 ................................ python3 -m drgn -s vmlinux -c vmcore-writer-reader-reader-reader contrib/locks.py rwsem --info ffffffffc02033c0 rwsem: ffffffffc02033c0 owned by writer (struct task_struct *)0xffff9162fdb58c40 (pid)1173 (state)R The waiters of rwsem are as follows: (struct task_struct *)0xffff9162fdb5ee40: (pid)1174: type: down_read state: D (struct task_struct *)0xffff9162fd8355c0: (pid)1175: type: down_read state: D (struct task_struct *)0xffff9162fd836200: (pid)1176: type: down_read state: D For newer kernels number of read owners can be obtained: python3 -m drgn -s vmlinux -c vmcore-multiple-readers contrib/locks.py rwsem --info ffffffffc00c9340 rwsem: ffffffffc00c9340 is owned by 2 reader(s). There are no waiters for rwsem: ffffffffc00c9340. Signed-off-by: Imran Khan <[email protected]>
Since osq_lock/unlock functions are not available to modules, I could not provide test cases for these helpers and hence I have put these under contrib. These helpers are useful in high sys usage cases where high sys usage is due to optimistic spinning on sleeping locks like rwsems and/or mutexes. In such cases we can use the script as follows to get the CPUs which are optimistically spinning to acquire the rwsem or mutex: python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py osq --dump There are spinners on one or more osq_lock CPU(s): [21, 17, 59, 67, 6, 0, 28, 53, 4, 41, 55, 38, 63, 23, 27, 19, 36] are spinning on same osq_lock CPU(s): [42, 56, 35, 18, 22, 3, 1] are spinning on same osq_lock CPU(s): [70, 44, 9] are spinning on same osq_lock CPU(s): [16, 26] are spinning on same osq_lock Since an osq_lock belongs to a sleeping lock, the CPUs spinning on same osq_lock are spinning to acquire the same sleeping lock. Above vmcore was obtained with a fork-bomb test. Signed-off-by: Imran Khan <[email protected]>
For example: python3 -m drgn -s vmlinux -c vmcore-writer-reader-reader-reader contrib/locks.py rwsem --spinner-callstack ffffffffc03083c0 rwsem: ffffffffc03083c0 has 1 spinners and their call-stack is as follows: call stack for pid: 239 #0 __read_once_size (./include/linux/compiler.h:268:2) osandov#1 arch_atomic64_read (./arch/x86/include/asm/atomic64_64.h:22:9) osandov#2 atomic64_read (./include/asm-generic/atomic-instrumented.h:837:9) osandov#3 atomic_long_read (./include/asm-generic/atomic-long.h:28:9) osandov#4 rwsem_owner_flags (kernel/locking/rwsem.c:298:24) osandov#5 rwsem_spin_on_owner (kernel/locking/rwsem.c:737:9) osandov#6 rwsem_optimistic_spin (kernel/locking/rwsem.c:812:17) osandov#7 rwsem_down_read_slowpath (kernel/locking/rwsem.c:1018:6) osandov#8 __down_read_killable (kernel/locking/rwsem.c:1366:14) osandov#9 down_read_killable (kernel/locking/rwsem.c:1532:6) osandov#10 0xffffffffc030622c ................. Signed-off-by: Imran Khan <[email protected]>
For example: python3 -m drgn -s vmlinux -c vmcore-1 contrib/locks.py mutex --spinner-callstack ffffffffc02af340 mutex: ffffffffc02af340 has 4 spinners and their call-stack is as follows: call stack for pid: 250 #0 __read_once_size (./include/linux/compiler.h:268:2) osandov#1 osq_lock (kernel/locking/osq_lock.c:137:10) osandov#2 mutex_optimistic_spin (kernel/locking/mutex.c:667:8) osandov#3 __mutex_lock_common (kernel/locking/mutex.c:971:6) osandov#4 __mutex_lock (kernel/locking/mutex.c:1109:9) osandov#5 0xffffffffc02ad045 ...................... call stack for pid: 251 #0 __read_once_size (./include/linux/compiler.h:268:2) osandov#1 osq_lock (kernel/locking/osq_lock.c:137:10) osandov#2 mutex_optimistic_spin (kernel/locking/mutex.c:667:8) osandov#3 __mutex_lock_common (kernel/locking/mutex.c:971:6) osandov#4 __mutex_lock (kernel/locking/mutex.c:1109:9) osandov#5 0xffffffffc02ad045 ..................... call stack for pid: 248 #0 __read_once_size (./include/linux/compiler.h:268:2) osandov#1 osq_lock (kernel/locking/osq_lock.c:137:10) osandov#2 mutex_optimistic_spin (kernel/locking/mutex.c:667:8) osandov#3 __mutex_lock_common (kernel/locking/mutex.c:971:6) osandov#4 __mutex_lock (kernel/locking/mutex.c:1109:9) osandov#5 0xffffffffc02ad045 ..................... call stack for pid: 249 #0 __read_once_size (./include/linux/compiler.h:268:2) osandov#1 arch_atomic64_read (./arch/x86/include/asm/atomic64_64.h:22:9) osandov#2 atomic64_read (./include/asm-generic/atomic-instrumented.h:837:9) osandov#3 atomic_long_read (./include/asm-generic/atomic-long.h:28:9) osandov#4 __mutex_owner (kernel/locking/mutex.c:75:32) osandov#5 mutex_spin_on_owner (kernel/locking/mutex.c:566:9) osandov#6 mutex_optimistic_spin (kernel/locking/mutex.c:683:8) osandov#7 __mutex_lock_common (kernel/locking/mutex.c:971:6) osandov#8 __mutex_lock (kernel/locking/mutex.c:1109:9) osandov#9 0xffffffffc02ad045 .................... Signed-off-by: Imran Khan <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This change set includes helpers for following locking primitives:
Since osq_lock/unlock functions are not available to modules,. I could not add test cases for these and hence I have put all osq helpers in a script under contrib. Besides osq helpers the contrib script also contains helper functions for above mentioned other 3 locking primitives. These helper functions take a lock address as argument and dump various info pertaining to these locks. I have included some examples in commit for reference.
P.S: I also have per-cpu rwsem helpers almost ready but will add them after some more testing.