* [PATCH] hungtask: add filter kthread/check comm
@ 2021-05-19 12:13 chenguanyou
2021-05-19 13:06 ` Michal Hocko
2021-05-19 17:53 ` Randy Dunlap
0 siblings, 2 replies; 11+ messages in thread
From: chenguanyou @ 2021-05-19 12:13 UTC (permalink / raw)
To: linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
Some kernel threads are always in D state, when we enable hung_task,
it will misjudge, we should skip these to narrow the scope.
exp mtk mobilephone:
root 420 420 2 0 0 kwdt_thread 0 D wdtk-0
root 421 421 2 0 0 kwdt_thread 0 D wdtk-1
root 422 422 2 0 0 kwdt_thread 0 D wdtk-2
root 423 423 2 0 0 kwdt_thread 0 D wdtk-3
root 424 424 2 0 0 kwdt_thread 0 D wdtk-4
root 425 425 2 0 0 kwdt_thread 0 D wdtk-5
root 426 426 2 0 0 kwdt_thread 0 D wdtk-6
root 427 427 2 0 0 kwdt_thread 0 D wdtk-7
root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
---
include/linux/sched/sysctl.h | 4 ++++
kernel/hung_task.c | 17 +++++++++++++++++
kernel/sysctl.c | 15 +++++++++++++++
lib/Kconfig.debug | 23 +++++++++++++++++++++++
4 files changed, 59 insertions(+)
diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index 3c31ba88aca5..e8a9a28215bf 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
#define sysctl_hung_task_all_cpu_backtrace 0
#endif /* CONFIG_SMP */
+#define TASK_COMM_LEN 16
+
extern int sysctl_hung_task_check_count;
extern unsigned int sysctl_hung_task_panic;
extern unsigned long sysctl_hung_task_timeout_secs;
extern unsigned long sysctl_hung_task_check_interval_secs;
extern int sysctl_hung_task_warnings;
+extern unsigned int sysctl_hung_task_filter_kthread;
+extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos);
#else
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index 396ebaebea3f..e018563d4882 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -23,6 +23,7 @@
#include <linux/sched/sysctl.h>
#include <trace/events/sched.h>
+#include <linux/string.h>
/*
* The number of tasks checked:
@@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
*/
unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
+/*
+ * Non zero means no checking kthread
+ */
+unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
+
+/*
+ * Only one
+ */
+char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
+
int __read_mostly sysctl_hung_task_warnings = 10;
static int __read_mostly did_panic;
@@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
{
unsigned long switch_count = t->nvcsw + t->nivcsw;
+ if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
+ return;
+
+ if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
+ return;
+
/*
* Ensure the task is not frozen.
* Also, skip vfork and any other user process that freezer should skip.
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 62fbd09b5dc1..157c47a8430a 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2519,6 +2519,21 @@ static struct ctl_table kern_table[] = {
.proc_handler = proc_dointvec_minmax,
.extra1 = &neg_one,
},
+ {
+ .procname = "hung_task_filter_kthread",
+ .data = &sysctl_hung_task_filter_kthread,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ },
+ {
+ .procname = "hung_task_check_comm",
+ .data = &sysctl_hung_task_check_comm,
+ .maxlen = TASK_COMM_LEN,
+ .mode = 0644,
+ .proc_handler = proc_dostring,
+ },
#endif
#ifdef CONFIG_RT_MUTEXES
{
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2c7f46b366f1..59cfa9e230ee 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1101,6 +1101,29 @@ config DEFAULT_HUNG_TASK_TIMEOUT
A timeout of 0 disables the check. The default is two minutes.
Keeping the default should be fine in most cases.
+config DEFAULT_HUNG_TASK_FILTER_KTHREAD
+ bool "Default filter kthread for hung task"
+ depends on DETECT_HUNG_TASK
+ default 0
+ help
+ This option controls filter kthread used to determine when
+ a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
+
+ It can be adjusted at runtime via the kernel.hung_task_filter_kthread
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_filter_kthread.
+
+ A filter of 1 disables the check
+
+config DEFAULT_HUNG_TASK_CHECK_COMM
+ string "Default check only one comm"
+ depends on DETECT_HUNG_TASK
+ default ""
+ help
+ It can be adjusted at runtime via the kernel.hung_task_check_comm
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_check_comm.
+
config BOOTPARAM_HUNG_TASK_PANIC
bool "Panic (Reboot) On Hung Tasks"
depends on DETECT_HUNG_TASK
--
2.17.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
2021-05-19 12:13 [PATCH] hungtask: add filter kthread/check comm chenguanyou
@ 2021-05-19 13:06 ` Michal Hocko
2021-05-19 17:53 ` Randy Dunlap
1 sibling, 0 replies; 11+ messages in thread
From: Michal Hocko @ 2021-05-19 13:06 UTC (permalink / raw)
To: chenguanyou
Cc: linux-kernel, akpm, keescook, lukas.bulwahn, vbabka, gpiccoli,
chenguanyou
On Wed 19-05-21 20:13:25, chenguanyou wrote:
> Some kernel threads are always in D state, when we enable hung_task,
> it will misjudge, we should skip these to narrow the scope.
Why they are in the D state in the first place?
Also you are proposing a new user interface which would need to be
maintained for ever. This means that any such a proposal has to be very
carefully justified. Always make sure to cc linux-api mailing list in
these cases.
From the (not much) information you have provided this interface seems
more like a workaround for buggy code so far.
--
Michal Hocko
SUSE Labs
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
2021-05-19 12:13 [PATCH] hungtask: add filter kthread/check comm chenguanyou
2021-05-19 13:06 ` Michal Hocko
@ 2021-05-19 17:53 ` Randy Dunlap
[not found] ` <CAHS3RMVYbHTzD6JnOmE331qSbcnvuYnBe0jraNuuLc0Z2NnStg@mail.gmail.com>
1 sibling, 1 reply; 11+ messages in thread
From: Randy Dunlap @ 2021-05-19 17:53 UTC (permalink / raw)
To: chenguanyou, linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
Hi,
Along with what Micahl has said, please see below.
On 5/19/21 5:13 AM, chenguanyou wrote:
> Some kernel threads are always in D state, when we enable hung_task,
> it will misjudge, we should skip these to narrow the scope.
>
> exp mtk mobilephone:
> root 420 420 2 0 0 kwdt_thread 0 D wdtk-0
> root 421 421 2 0 0 kwdt_thread 0 D wdtk-1
> root 422 422 2 0 0 kwdt_thread 0 D wdtk-2
> root 423 423 2 0 0 kwdt_thread 0 D wdtk-3
> root 424 424 2 0 0 kwdt_thread 0 D wdtk-4
> root 425 425 2 0 0 kwdt_thread 0 D wdtk-5
> root 426 426 2 0 0 kwdt_thread 0 D wdtk-6
> root 427 427 2 0 0 kwdt_thread 0 D wdtk-7
> root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
> root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
> root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
> root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
> root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
> root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
> root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
> root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
>
> Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
> ---
> include/linux/sched/sysctl.h | 4 ++++
> kernel/hung_task.c | 17 +++++++++++++++++
> kernel/sysctl.c | 15 +++++++++++++++
> lib/Kconfig.debug | 23 +++++++++++++++++++++++
> 4 files changed, 59 insertions(+)
>
> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
> index 396ebaebea3f..e018563d4882 100644
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -23,6 +23,7 @@
> #include <linux/sched/sysctl.h>
>
> #include <trace/events/sched.h>
> +#include <linux/string.h>
>
> /*
> * The number of tasks checked:
> @@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
> */
> unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
>
> +/*
> + * Non zero means no checking kthread
Non-zero
> + */
> +unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
> +
> +/*
> + * Only one
> + */
> +char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
> +
> int __read_mostly sysctl_hung_task_warnings = 10;
>
> static int __read_mostly did_panic;
> @@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
> {
> unsigned long switch_count = t->nvcsw + t->nivcsw;
>
> + if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
> + return;
> +
> + if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
> + return;
> +
> /*
> * Ensure the task is not frozen.
> * Also, skip vfork and any other user process that freezer should skip.
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 62fbd09b5dc1..157c47a8430a 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -2519,6 +2519,21 @@ static struct ctl_table kern_table[] = {
> .proc_handler = proc_dointvec_minmax,
> .extra1 = &neg_one,
> },
> + {
> + .procname = "hung_task_filter_kthread",
> + .data = &sysctl_hung_task_filter_kthread,
> + .maxlen = sizeof(int),
> + .mode = 0644,
> + .extra1 = SYSCTL_ZERO,
> + .extra2 = SYSCTL_ONE,
> + },
> + {
> + .procname = "hung_task_check_comm",
> + .data = &sysctl_hung_task_check_comm,
> + .maxlen = TASK_COMM_LEN,
> + .mode = 0644,
> + .proc_handler = proc_dostring,
> + },
> #endif
These new sysctls should be documented in Documentation/admin-guide/sysctl/kernel.rst.
> #ifdef CONFIG_RT_MUTEXES
> {
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 2c7f46b366f1..59cfa9e230ee 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1101,6 +1101,29 @@ config DEFAULT_HUNG_TASK_TIMEOUT
> A timeout of 0 disables the check. The default is two minutes.
> Keeping the default should be fine in most cases.
>
> +config DEFAULT_HUNG_TASK_FILTER_KTHREAD
> + bool "Default filter kthread for hung task"
> + depends on DETECT_HUNG_TASK
> + default 0
> + help
> + This option controls filter kthread used to determine when
uses
> + a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
skipped.
> +
> + It can be adjusted at runtime via the kernel.hung_task_filter_kthread
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_filter_kthread.
> +
> + A filter of 1 disables the check
check.
> +
> +config DEFAULT_HUNG_TASK_CHECK_COMM
> + string "Default check only one comm"
> + depends on DETECT_HUNG_TASK
> + default ""
> + help
> + It can be adjusted at runtime via the kernel.hung_task_check_comm
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_check_comm.
That help text doesn't tell how the Kconfig symbol is used.
> +
> config BOOTPARAM_HUNG_TASK_PANIC
> bool "Panic (Reboot) On Hung Tasks"
> depends on DETECT_HUNG_TASK
>
thanks.
--
~Randy
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
[not found] ` <CAHS3RMVYbHTzD6JnOmE331qSbcnvuYnBe0jraNuuLc0Z2NnStg@mail.gmail.com>
@ 2021-05-21 8:31 ` Michal Hocko
0 siblings, 0 replies; 11+ messages in thread
From: Michal Hocko @ 2021-05-21 8:31 UTC (permalink / raw)
To: 陈冠有
Cc: Randy Dunlap, linux-kernel, akpm, Kees Cook, lukas.bulwahn,
vbabka, gpiccoli, chenguanyou
On Fri 21-05-21 14:38:06, 陈冠有 wrote:
> when we no skip kthread, exp log:
> 3,25164,832483138,-; (1)[67:khungtaskd]INFO: task amms_task:51 blocked for
> more than 120 seconds.
> 3,25165,832483169,-; (1)[67:khungtaskd] Tainted: P S W O
> 4.14.186-g9d5f2ff-dirty #4
> 3,25166,832483186,-; (1)[67:khungtaskd]"echo 0 >
> /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> 6,25167,832483205,-; (1)[67:khungtaskd]amms_task D 0 51 2
> 0x00000000
> 4,25168,832483232,-; (1)[67:khungtaskd]Call trace:
> 4,25169,832483268,-; (1)[67:khungtaskd] __switch_to+0x134/0x150
> 4,25170,832483297,-; (1)[67:khungtaskd] __schedule+0xd5c/0x1100
> 4,25171,832483318,-; (1)[67:khungtaskd] schedule+0x70/0x90
> 4,25172,832483343,-; (1)[67:khungtaskd] kthread+0xfc/0x18c
> 4,25173,832483365,-; (1)[67:khungtaskd] ret_from_fork+0x10/0x18
> 3,25174,832483482,-; (1)[67:khungtaskd]INFO: task mdrt_thread:123 blocked
> for more than 120 seconds.
> 3,25175,832483501,-; (1)[67:khungtaskd] Tainted: P S W O
> 4.14.186-g9d5f2ff-dirty #4
> 3,25176,832483516,-; (1)[67:khungtaskd]"echo 0 >
> /proc/sys/kernel/hung_task_timeout_secs" disables this message.
> 6,25177,832483533,-; (1)[67:khungtaskd]mdrt_thread D 0 123 2
> 0x00000000
> 4,25178,832483555,-; (1)[67:khungtaskd]Call trace:
> 4,25179,832483574,-; (1)[67:khungtaskd] __switch_to+0x134/0x150
> 4,25180,832483595,-; (1)[67:khungtaskd] __schedule+0xd5c/0x1100
> 4,25181,832483615,-; (1)[67:khungtaskd] schedule+0x70/0x90
> 4,25182,832483635,-; (1)[67:khungtaskd] kthread+0xfc/0x18c
> 4,25183,832483655,-; (1)[67:khungtaskd] ret_from_fork+0x10/0x18
> 3,25184,832483737,-; (1)[67:khungtaskd]INFO: task scp_power_reset:227
> blocked for more than 120 seconds.
What are all these kernel threads doing and why it is ok to inhibit them
in the report.
--
Michal Hocko
SUSE Labs
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
[not found] ` <CAHS3RMWjuB98TzvcYyQ0qtNYOxDeM7W1YmyqDYs=H-cn-VsVdw@mail.gmail.com>
@ 2021-05-25 7:26 ` Michal Hocko
0 siblings, 0 replies; 11+ messages in thread
From: Michal Hocko @ 2021-05-25 7:26 UTC (permalink / raw)
To: 陈冠有
Cc: Andrew Morton, linux-kernel, Kees Cook, lukas.bulwahn, vbabka,
gpiccoli, chenguanyou
On Tue 25-05-21 11:27:16, 陈冠有 wrote:
[...]
> These are MTK's LPM designs.
No idea what those are but it seems like an out of tree code.
> If we have linux-api, hungtask can choose to skip kthread, helpful for us
> to debug user space threads in "state=D" when enable hungtask panic.
No, this approach is wrong. You are trying to workaround an incorrect
out of tree code by creating a user visible API. I have already brought
that up in earlier version of this patch.
Nacked-by: Michal Hocko <mhocko@suse.com>
Andrew has already given you a lead on how to fix the said code. Use
TASK_IDLE when waiting for an event. This will both hide that waiting
task from load average accounting and also from the hung task detector.
--
Michal Hocko
SUSE Labs
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
2021-05-21 13:25 chenguanyou
@ 2021-05-22 21:51 ` Andrew Morton
[not found] ` <CAHS3RMWjuB98TzvcYyQ0qtNYOxDeM7W1YmyqDYs=H-cn-VsVdw@mail.gmail.com>
0 siblings, 1 reply; 11+ messages in thread
From: Andrew Morton @ 2021-05-22 21:51 UTC (permalink / raw)
To: chenguanyou
Cc: linux-kernel, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli,
chenguanyou
On Fri, 21 May 2021 21:25:44 +0800 chenguanyou <chenguanyou9338@gmail.com> wrote:
> Some kernel threads are always in D state, when we enable hung_task,
> it will misjudge, we should skip these to narrow the scope.
>
> exp mtk mobilephone:
> root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
> root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
> root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
> root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
> root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
> root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
> root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
> root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
Maybe convert these threads to use TASK_IDLE?
^ permalink raw reply [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
2021-05-21 13:25 chenguanyou
@ 2021-05-21 16:37 ` Randy Dunlap
0 siblings, 0 replies; 11+ messages in thread
From: Randy Dunlap @ 2021-05-21 16:37 UTC (permalink / raw)
To: chenguanyou, linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
On 5/21/21 6:25 AM, chenguanyou wrote:
> Some kernel threads are always in D state, when we enable hung_task,
> it will misjudge, we should skip these to narrow the scope.
>
> exp mtk mobilephone:
> root 420 420 2 0 0 kwdt_thread 0 D wdtk-0
> root 421 421 2 0 0 kwdt_thread 0 D wdtk-1
> root 422 422 2 0 0 kwdt_thread 0 D wdtk-2
> root 423 423 2 0 0 kwdt_thread 0 D wdtk-3
> root 424 424 2 0 0 kwdt_thread 0 D wdtk-4
> root 425 425 2 0 0 kwdt_thread 0 D wdtk-5
> root 426 426 2 0 0 kwdt_thread 0 D wdtk-6
> root 427 427 2 0 0 kwdt_thread 0 D wdtk-7
> root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
> root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
> root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
> root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
> root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
> root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
> root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
> root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
>
> Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
> ---
> include/linux/sched/sysctl.h | 4 ++++
> kernel/hung_task.c | 17 +++++++++++++++++
> kernel/sysctl.c | 16 ++++++++++++++++
> lib/Kconfig.debug | 24 ++++++++++++++++++++++++
> 4 files changed, 61 insertions(+)
>
> diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
> index 3c31ba88aca5..e8a9a28215bf 100644
> --- a/include/linux/sched/sysctl.h
> +++ b/include/linux/sched/sysctl.h
> @@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
> #define sysctl_hung_task_all_cpu_backtrace 0
> #endif /* CONFIG_SMP */
>
> +#define TASK_COMM_LEN 16
> +
> extern int sysctl_hung_task_check_count;
> extern unsigned int sysctl_hung_task_panic;
> extern unsigned long sysctl_hung_task_timeout_secs;
> extern unsigned long sysctl_hung_task_check_interval_secs;
> extern int sysctl_hung_task_warnings;
> +extern unsigned int sysctl_hung_task_filter_kthread;
> +extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
> int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
> void *buffer, size_t *lenp, loff_t *ppos);
> #else
> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
> index 396ebaebea3f..e018563d4882 100644
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -23,6 +23,7 @@
> #include <linux/sched/sysctl.h>
>
> #include <trace/events/sched.h>
> +#include <linux/string.h>
>
> /*
> * The number of tasks checked:
> @@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
> */
> unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
>
> +/*
> + * Non zero means no checking kthread
Non-zero
(also mentioned in v1 review, whereas this is v2, I guess, although
it doesn't say that it is v2, but it should say that.)
> + */
> +unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
> +
> +/*
> + * Only one
> + */
> +char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
> +
> int __read_mostly sysctl_hung_task_warnings = 10;
>
> static int __read_mostly did_panic;
> @@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
> {
> unsigned long switch_count = t->nvcsw + t->nivcsw;
>
> + if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
> + return;
> +
> + if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
> + return;
> +
> /*
> * Ensure the task is not frozen.
> * Also, skip vfork and any other user process that freezer should skip.
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 62fbd09b5dc1..1daede87c88d 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -2519,6 +2519,22 @@ static struct ctl_table kern_table[] = {
> .proc_handler = proc_dointvec_minmax,
> .extra1 = &neg_one,
> },
> + {
> + .procname = "hung_task_filter_kthread",
> + .data = &sysctl_hung_task_filter_kthread,
> + .maxlen = sizeof(int),
> + .mode = 0644,
> + .proc_handler = proc_dointvec_minmax,
> + .extra1 = SYSCTL_ZERO,
> + .extra2 = SYSCTL_ONE,
> + },
> + {
> + .procname = "hung_task_check_comm",
> + .data = &sysctl_hung_task_check_comm,
> + .maxlen = TASK_COMM_LEN,
> + .mode = 0644,
> + .proc_handler = proc_dostring,
> + },
(copy-paste from v1 review:)
These new sysctls should be documented in Documentation/admin-guide/sysctl/kernel.rst.
> #endif
> #ifdef CONFIG_RT_MUTEXES
> {
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 2c7f46b366f1..6eab8cf0c37f 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1101,6 +1101,30 @@ config DEFAULT_HUNG_TASK_TIMEOUT
> A timeout of 0 disables the check. The default is two minutes.
> Keeping the default should be fine in most cases.
>
> +config DEFAULT_HUNG_TASK_FILTER_KTHREAD
> + int "Default filter kthread for hung task"
> + depends on DETECT_HUNG_TASK
> + range 0 1
> + default 0
> + help
> + This option controls filter kthread used to determine when
(again:)
uses
> + a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
(again:)
skipped.
> +
> + It can be adjusted at runtime via the kernel.hung_task_filter_kthread
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_filter_kthread.
> +
> + A filter of 1 disables the check
(again:)
check.
> +
> +config DEFAULT_HUNG_TASK_CHECK_COMM
> + string "Default check only one comm"
> + depends on DETECT_HUNG_TASK
> + default ""
> + help
> + It can be adjusted at runtime via the kernel.hung_task_check_comm
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_check_comm.
> +
(again:)
That help text doesn't tell how the Kconfig symbol is used.
> config BOOTPARAM_HUNG_TASK_PANIC
> bool "Panic (Reboot) On Hung Tasks"
> depends on DETECT_HUNG_TASK
>
--
~Randy
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH] hungtask: add filter kthread/check comm
@ 2021-05-21 13:25 chenguanyou
2021-05-22 21:51 ` Andrew Morton
0 siblings, 1 reply; 11+ messages in thread
From: chenguanyou @ 2021-05-21 13:25 UTC (permalink / raw)
To: linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
Some kernel threads are always in D state, when we enable hung_task,
it will misjudge, we should skip these to narrow the scope.
exp mtk mobilephone:
root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
---
Documentation/admin-guide/sysctl/kernel.rst | 16 ++++++++++++
include/linux/sched/sysctl.h | 4 +++
kernel/hung_task.c | 17 +++++++++++++
kernel/sysctl.c | 16 ++++++++++++
lib/Kconfig.debug | 27 +++++++++++++++++++++
5 files changed, 80 insertions(+)
diff --git a/Documentation/admin-guide/sysctl/kernel.rst b/Documentation/admin-guide/sysctl/kernel.rst
index 1d56a6b73a4e..081ca22db4d5 100644
--- a/Documentation/admin-guide/sysctl/kernel.rst
+++ b/Documentation/admin-guide/sysctl/kernel.rst
@@ -405,6 +405,22 @@ This file shows up if ``CONFIG_DETECT_HUNG_TASK`` is enabled.
-1: report an infinite number of warnings.
+hung_task_filter_kthread
+========================
+
+We should skip kthread when a hung task is detected.
+This file shows up if ``CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD`` is enabled.
+
+= =========================================================
+0 Not skip detect kthread.
+1 Skip detect kthread.
+= =========================================================
+
+hung_task_check_comm
+====================
+
+We should skip non ``hung_task_check_comm`` when a hung task is detected.
+This file shows up if ``CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM`` is enabled.
hyperv_record_panic_msg
=======================
diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index 3c31ba88aca5..e8a9a28215bf 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
#define sysctl_hung_task_all_cpu_backtrace 0
#endif /* CONFIG_SMP */
+#define TASK_COMM_LEN 16
+
extern int sysctl_hung_task_check_count;
extern unsigned int sysctl_hung_task_panic;
extern unsigned long sysctl_hung_task_timeout_secs;
extern unsigned long sysctl_hung_task_check_interval_secs;
extern int sysctl_hung_task_warnings;
+extern unsigned int sysctl_hung_task_filter_kthread;
+extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos);
#else
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index 396ebaebea3f..baee8466b902 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -23,6 +23,7 @@
#include <linux/sched/sysctl.h>
#include <trace/events/sched.h>
+#include <linux/string.h>
/*
* The number of tasks checked:
@@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
*/
unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
+/*
+ * Non-zero means no checking kthread
+ */
+unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
+
+/*
+ * Only one
+ */
+char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
+
int __read_mostly sysctl_hung_task_warnings = 10;
static int __read_mostly did_panic;
@@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
{
unsigned long switch_count = t->nvcsw + t->nivcsw;
+ if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
+ return;
+
+ if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
+ return;
+
/*
* Ensure the task is not frozen.
* Also, skip vfork and any other user process that freezer should skip.
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 62fbd09b5dc1..1daede87c88d 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2519,6 +2519,22 @@ static struct ctl_table kern_table[] = {
.proc_handler = proc_dointvec_minmax,
.extra1 = &neg_one,
},
+ {
+ .procname = "hung_task_filter_kthread",
+ .data = &sysctl_hung_task_filter_kthread,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = proc_dointvec_minmax,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ },
+ {
+ .procname = "hung_task_check_comm",
+ .data = &sysctl_hung_task_check_comm,
+ .maxlen = TASK_COMM_LEN,
+ .mode = 0644,
+ .proc_handler = proc_dostring,
+ },
#endif
#ifdef CONFIG_RT_MUTEXES
{
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2c7f46b366f1..63570b1fec35 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1101,6 +1101,33 @@ config DEFAULT_HUNG_TASK_TIMEOUT
A timeout of 0 disables the check. The default is two minutes.
Keeping the default should be fine in most cases.
+config DEFAULT_HUNG_TASK_FILTER_KTHREAD
+ int "Default filter kthread for hung task"
+ depends on DETECT_HUNG_TASK
+ range 0 1
+ default 0
+ help
+ This option controls filter kthread uses to determine when
+ a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skipped.
+
+ It can be adjusted at runtime via the kernel.hung_task_filter_kthread
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_filter_kthread.
+
+ A filter of 1 disables the check.
+
+config DEFAULT_HUNG_TASK_CHECK_COMM
+ string "Default check only one comm"
+ depends on DETECT_HUNG_TASK
+ default ""
+ help
+ This option controls only detect "task.comm = kernel.hung_task_check_comm" when
+ it become "state=TASK_UNINTERRUPTIBLE", skip other threads.
+
+ It can be adjusted at runtime via the kernel.hung_task_check_comm
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_check_comm.
+
config BOOTPARAM_HUNG_TASK_PANIC
bool "Panic (Reboot) On Hung Tasks"
depends on DETECT_HUNG_TASK
--
2.17.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* [PATCH] hungtask: add filter kthread/check comm
@ 2021-05-21 13:25 chenguanyou
2021-05-21 16:37 ` Randy Dunlap
0 siblings, 1 reply; 11+ messages in thread
From: chenguanyou @ 2021-05-21 13:25 UTC (permalink / raw)
To: linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
Some kernel threads are always in D state, when we enable hung_task,
it will misjudge, we should skip these to narrow the scope.
exp mtk mobilephone:
root 420 420 2 0 0 kwdt_thread 0 D wdtk-0
root 421 421 2 0 0 kwdt_thread 0 D wdtk-1
root 422 422 2 0 0 kwdt_thread 0 D wdtk-2
root 423 423 2 0 0 kwdt_thread 0 D wdtk-3
root 424 424 2 0 0 kwdt_thread 0 D wdtk-4
root 425 425 2 0 0 kwdt_thread 0 D wdtk-5
root 426 426 2 0 0 kwdt_thread 0 D wdtk-6
root 427 427 2 0 0 kwdt_thread 0 D wdtk-7
root 435 435 2 0 0 mtk_lpm_monitor_thread 0 D LPM-0
root 436 436 2 0 0 mtk_lpm_monitor_thread 0 D LPM-1
root 437 437 2 0 0 mtk_lpm_monitor_thread 0 D LPM-2
root 438 438 2 0 0 mtk_lpm_monitor_thread 0 D LPM-3
root 439 439 2 0 0 mtk_lpm_monitor_thread 0 D LPM-4
root 440 440 2 0 0 mtk_lpm_monitor_thread 0 D LPM-5
root 441 441 2 0 0 mtk_lpm_monitor_thread 0 D LPM-6
root 442 442 2 0 0 mtk_lpm_monitor_thread 0 D LPM-7
Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
---
include/linux/sched/sysctl.h | 4 ++++
kernel/hung_task.c | 17 +++++++++++++++++
kernel/sysctl.c | 16 ++++++++++++++++
lib/Kconfig.debug | 24 ++++++++++++++++++++++++
4 files changed, 61 insertions(+)
diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index 3c31ba88aca5..e8a9a28215bf 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
#define sysctl_hung_task_all_cpu_backtrace 0
#endif /* CONFIG_SMP */
+#define TASK_COMM_LEN 16
+
extern int sysctl_hung_task_check_count;
extern unsigned int sysctl_hung_task_panic;
extern unsigned long sysctl_hung_task_timeout_secs;
extern unsigned long sysctl_hung_task_check_interval_secs;
extern int sysctl_hung_task_warnings;
+extern unsigned int sysctl_hung_task_filter_kthread;
+extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos);
#else
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index 396ebaebea3f..e018563d4882 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -23,6 +23,7 @@
#include <linux/sched/sysctl.h>
#include <trace/events/sched.h>
+#include <linux/string.h>
/*
* The number of tasks checked:
@@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
*/
unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
+/*
+ * Non zero means no checking kthread
+ */
+unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
+
+/*
+ * Only one
+ */
+char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
+
int __read_mostly sysctl_hung_task_warnings = 10;
static int __read_mostly did_panic;
@@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
{
unsigned long switch_count = t->nvcsw + t->nivcsw;
+ if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
+ return;
+
+ if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
+ return;
+
/*
* Ensure the task is not frozen.
* Also, skip vfork and any other user process that freezer should skip.
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 62fbd09b5dc1..1daede87c88d 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2519,6 +2519,22 @@ static struct ctl_table kern_table[] = {
.proc_handler = proc_dointvec_minmax,
.extra1 = &neg_one,
},
+ {
+ .procname = "hung_task_filter_kthread",
+ .data = &sysctl_hung_task_filter_kthread,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .proc_handler = proc_dointvec_minmax,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ },
+ {
+ .procname = "hung_task_check_comm",
+ .data = &sysctl_hung_task_check_comm,
+ .maxlen = TASK_COMM_LEN,
+ .mode = 0644,
+ .proc_handler = proc_dostring,
+ },
#endif
#ifdef CONFIG_RT_MUTEXES
{
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2c7f46b366f1..6eab8cf0c37f 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1101,6 +1101,30 @@ config DEFAULT_HUNG_TASK_TIMEOUT
A timeout of 0 disables the check. The default is two minutes.
Keeping the default should be fine in most cases.
+config DEFAULT_HUNG_TASK_FILTER_KTHREAD
+ int "Default filter kthread for hung task"
+ depends on DETECT_HUNG_TASK
+ range 0 1
+ default 0
+ help
+ This option controls filter kthread used to determine when
+ a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
+
+ It can be adjusted at runtime via the kernel.hung_task_filter_kthread
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_filter_kthread.
+
+ A filter of 1 disables the check
+
+config DEFAULT_HUNG_TASK_CHECK_COMM
+ string "Default check only one comm"
+ depends on DETECT_HUNG_TASK
+ default ""
+ help
+ It can be adjusted at runtime via the kernel.hung_task_check_comm
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_check_comm.
+
config BOOTPARAM_HUNG_TASK_PANIC
bool "Panic (Reboot) On Hung Tasks"
depends on DETECT_HUNG_TASK
--
2.17.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
* Re: [PATCH] hungtask: add filter kthread/check comm
2021-05-18 11:16 chenguanyou
@ 2021-05-19 11:28 ` Vlastimil Babka
0 siblings, 0 replies; 11+ messages in thread
From: Vlastimil Babka @ 2021-05-19 11:28 UTC (permalink / raw)
To: chenguanyou, linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, gpiccoli, chenguanyou
On 5/18/21 1:16 PM, chenguanyou wrote:
The description is missing.
> Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
> ---
> include/linux/sched/sysctl.h | 4 ++++
> kernel/hung_task.c | 17 +++++++++++++++++
> kernel/sysctl.c | 15 +++++++++++++++
> lib/Kconfig.debug | 23 +++++++++++++++++++++++
> 4 files changed, 59 insertions(+)
>
> diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
> index 3c31ba88aca5..e8a9a28215bf 100644
> --- a/include/linux/sched/sysctl.h
> +++ b/include/linux/sched/sysctl.h
> @@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
> #define sysctl_hung_task_all_cpu_backtrace 0
> #endif /* CONFIG_SMP */
>
> +#define TASK_COMM_LEN 16
> +
> extern int sysctl_hung_task_check_count;
> extern unsigned int sysctl_hung_task_panic;
> extern unsigned long sysctl_hung_task_timeout_secs;
> extern unsigned long sysctl_hung_task_check_interval_secs;
> extern int sysctl_hung_task_warnings;
> +extern unsigned int sysctl_hung_task_filter_kthread;
> +extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
> int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
> void *buffer, size_t *lenp, loff_t *ppos);
> #else
> diff --git a/kernel/hung_task.c b/kernel/hung_task.c
> index 396ebaebea3f..e018563d4882 100644
> --- a/kernel/hung_task.c
> +++ b/kernel/hung_task.c
> @@ -23,6 +23,7 @@
> #include <linux/sched/sysctl.h>
>
> #include <trace/events/sched.h>
> +#include <linux/string.h>
>
> /*
> * The number of tasks checked:
> @@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
> */
> unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
>
> +/*
> + * Non zero means no checking kthread
> + */
> +unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
> +
> +/*
> + * Only one
> + */
> +char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
> +
> int __read_mostly sysctl_hung_task_warnings = 10;
>
> static int __read_mostly did_panic;
> @@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
> {
> unsigned long switch_count = t->nvcsw + t->nivcsw;
>
> + if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
> + return;
> +
> + if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
> + return;
> +
> /*
> * Ensure the task is not frozen.
> * Also, skip vfork and any other user process that freezer should skip.
> diff --git a/kernel/sysctl.c b/kernel/sysctl.c
> index 62fbd09b5dc1..157c47a8430a 100644
> --- a/kernel/sysctl.c
> +++ b/kernel/sysctl.c
> @@ -2519,6 +2519,21 @@ static struct ctl_table kern_table[] = {
> .proc_handler = proc_dointvec_minmax,
> .extra1 = &neg_one,
> },
> + {
> + .procname = "hung_task_filter_kthread",
> + .data = &sysctl_hung_task_filter_kthread,
> + .maxlen = sizeof(int),
> + .mode = 0644,
> + .extra1 = SYSCTL_ZERO,
> + .extra2 = SYSCTL_ONE,
> + },
> + {
> + .procname = "hung_task_check_comm",
> + .data = &sysctl_hung_task_check_comm,
> + .maxlen = TASK_COMM_LEN,
> + .mode = 0644,
> + .proc_handler = proc_dostring,
> + },
> #endif
> #ifdef CONFIG_RT_MUTEXES
> {
> diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
> index 2c7f46b366f1..59cfa9e230ee 100644
> --- a/lib/Kconfig.debug
> +++ b/lib/Kconfig.debug
> @@ -1101,6 +1101,29 @@ config DEFAULT_HUNG_TASK_TIMEOUT
> A timeout of 0 disables the check. The default is two minutes.
> Keeping the default should be fine in most cases.
>
> +config DEFAULT_HUNG_TASK_FILTER_KTHREAD
> + bool "Default filter kthread for hung task"
> + depends on DETECT_HUNG_TASK
> + default 0
> + help
> + This option controls filter kthread used to determine when
> + a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
> +
> + It can be adjusted at runtime via the kernel.hung_task_filter_kthread
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_filter_kthread.
> +
> + A filter of 1 disables the check
> +
> +config DEFAULT_HUNG_TASK_CHECK_COMM
> + string "Default check only one comm"
> + depends on DETECT_HUNG_TASK
> + default ""
> + help
> + It can be adjusted at runtime via the kernel.hung_task_check_comm
> + sysctl or by writing a value to
> + /proc/sys/kernel/hung_task_check_comm.
> +
> config BOOTPARAM_HUNG_TASK_PANIC
> bool "Panic (Reboot) On Hung Tasks"
> depends on DETECT_HUNG_TASK
>
^ permalink raw reply [flat|nested] 11+ messages in thread
* [PATCH] hungtask: add filter kthread/check comm
@ 2021-05-18 11:16 chenguanyou
2021-05-19 11:28 ` Vlastimil Babka
0 siblings, 1 reply; 11+ messages in thread
From: chenguanyou @ 2021-05-18 11:16 UTC (permalink / raw)
To: linux-kernel
Cc: akpm, keescook, mhocko, lukas.bulwahn, vbabka, gpiccoli, chenguanyou
Signed-off-by: chenguanyou <chenguanyou@xiaomi.com>
---
include/linux/sched/sysctl.h | 4 ++++
kernel/hung_task.c | 17 +++++++++++++++++
kernel/sysctl.c | 15 +++++++++++++++
lib/Kconfig.debug | 23 +++++++++++++++++++++++
4 files changed, 59 insertions(+)
diff --git a/include/linux/sched/sysctl.h b/include/linux/sched/sysctl.h
index 3c31ba88aca5..e8a9a28215bf 100644
--- a/include/linux/sched/sysctl.h
+++ b/include/linux/sched/sysctl.h
@@ -14,11 +14,15 @@ extern unsigned int sysctl_hung_task_all_cpu_backtrace;
#define sysctl_hung_task_all_cpu_backtrace 0
#endif /* CONFIG_SMP */
+#define TASK_COMM_LEN 16
+
extern int sysctl_hung_task_check_count;
extern unsigned int sysctl_hung_task_panic;
extern unsigned long sysctl_hung_task_timeout_secs;
extern unsigned long sysctl_hung_task_check_interval_secs;
extern int sysctl_hung_task_warnings;
+extern unsigned int sysctl_hung_task_filter_kthread;
+extern char sysctl_hung_task_check_comm[TASK_COMM_LEN];
int proc_dohung_task_timeout_secs(struct ctl_table *table, int write,
void *buffer, size_t *lenp, loff_t *ppos);
#else
diff --git a/kernel/hung_task.c b/kernel/hung_task.c
index 396ebaebea3f..e018563d4882 100644
--- a/kernel/hung_task.c
+++ b/kernel/hung_task.c
@@ -23,6 +23,7 @@
#include <linux/sched/sysctl.h>
#include <trace/events/sched.h>
+#include <linux/string.h>
/*
* The number of tasks checked:
@@ -48,6 +49,16 @@ unsigned long __read_mostly sysctl_hung_task_timeout_secs = CONFIG_DEFAULT_HUNG_
*/
unsigned long __read_mostly sysctl_hung_task_check_interval_secs;
+/*
+ * Non zero means no checking kthread
+ */
+unsigned int __read_mostly sysctl_hung_task_filter_kthread = CONFIG_DEFAULT_HUNG_TASK_FILTER_KTHREAD;
+
+/*
+ * Only one
+ */
+char __read_mostly sysctl_hung_task_check_comm[TASK_COMM_LEN] = CONFIG_DEFAULT_HUNG_TASK_CHECK_COMM;
+
int __read_mostly sysctl_hung_task_warnings = 10;
static int __read_mostly did_panic;
@@ -88,6 +99,12 @@ static void check_hung_task(struct task_struct *t, unsigned long timeout)
{
unsigned long switch_count = t->nvcsw + t->nivcsw;
+ if (unlikely(strlen(sysctl_hung_task_check_comm) && !strstr(t->comm, sysctl_hung_task_check_comm)))
+ return;
+
+ if (unlikely(sysctl_hung_task_filter_kthread && t->flags & PF_KTHREAD))
+ return;
+
/*
* Ensure the task is not frozen.
* Also, skip vfork and any other user process that freezer should skip.
diff --git a/kernel/sysctl.c b/kernel/sysctl.c
index 62fbd09b5dc1..157c47a8430a 100644
--- a/kernel/sysctl.c
+++ b/kernel/sysctl.c
@@ -2519,6 +2519,21 @@ static struct ctl_table kern_table[] = {
.proc_handler = proc_dointvec_minmax,
.extra1 = &neg_one,
},
+ {
+ .procname = "hung_task_filter_kthread",
+ .data = &sysctl_hung_task_filter_kthread,
+ .maxlen = sizeof(int),
+ .mode = 0644,
+ .extra1 = SYSCTL_ZERO,
+ .extra2 = SYSCTL_ONE,
+ },
+ {
+ .procname = "hung_task_check_comm",
+ .data = &sysctl_hung_task_check_comm,
+ .maxlen = TASK_COMM_LEN,
+ .mode = 0644,
+ .proc_handler = proc_dostring,
+ },
#endif
#ifdef CONFIG_RT_MUTEXES
{
diff --git a/lib/Kconfig.debug b/lib/Kconfig.debug
index 2c7f46b366f1..59cfa9e230ee 100644
--- a/lib/Kconfig.debug
+++ b/lib/Kconfig.debug
@@ -1101,6 +1101,29 @@ config DEFAULT_HUNG_TASK_TIMEOUT
A timeout of 0 disables the check. The default is two minutes.
Keeping the default should be fine in most cases.
+config DEFAULT_HUNG_TASK_FILTER_KTHREAD
+ bool "Default filter kthread for hung task"
+ depends on DETECT_HUNG_TASK
+ default 0
+ help
+ This option controls filter kthread used to determine when
+ a kernel task has become "state=TASK_UNINTERRUPTIBLE" and should be skip.
+
+ It can be adjusted at runtime via the kernel.hung_task_filter_kthread
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_filter_kthread.
+
+ A filter of 1 disables the check
+
+config DEFAULT_HUNG_TASK_CHECK_COMM
+ string "Default check only one comm"
+ depends on DETECT_HUNG_TASK
+ default ""
+ help
+ It can be adjusted at runtime via the kernel.hung_task_check_comm
+ sysctl or by writing a value to
+ /proc/sys/kernel/hung_task_check_comm.
+
config BOOTPARAM_HUNG_TASK_PANIC
bool "Panic (Reboot) On Hung Tasks"
depends on DETECT_HUNG_TASK
--
2.17.1
^ permalink raw reply related [flat|nested] 11+ messages in thread
end of thread, other threads:[~2021-05-25 7:26 UTC | newest]
Thread overview: 11+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2021-05-19 12:13 [PATCH] hungtask: add filter kthread/check comm chenguanyou
2021-05-19 13:06 ` Michal Hocko
2021-05-19 17:53 ` Randy Dunlap
[not found] ` <CAHS3RMVYbHTzD6JnOmE331qSbcnvuYnBe0jraNuuLc0Z2NnStg@mail.gmail.com>
2021-05-21 8:31 ` Michal Hocko
-- strict thread matches above, loose matches on Subject: below --
2021-05-21 13:25 chenguanyou
2021-05-22 21:51 ` Andrew Morton
[not found] ` <CAHS3RMWjuB98TzvcYyQ0qtNYOxDeM7W1YmyqDYs=H-cn-VsVdw@mail.gmail.com>
2021-05-25 7:26 ` Michal Hocko
2021-05-21 13:25 chenguanyou
2021-05-21 16:37 ` Randy Dunlap
2021-05-18 11:16 chenguanyou
2021-05-19 11:28 ` Vlastimil Babka
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).