sched/tracing: fix __trace_sched_switch_state (5.18-rc7+)
PATCH 1/1 sched/tracing: fix __trace_sched_switch_state (5.18-rc7+)
From: Bruce Ashfield <[email protected]>
The commit [fix: sched/tracing: Don't re-read p->state when emitting sched_switch event (v5.18)] was correct, but the kernel changed their mind with the following commit:
commit 9c2136be0878c88c53dea26943ce40bb03ad8d8d
Author: Delyan Kratunov <[email protected]>
Date: Wed May 11 18:28:36 2022 +0000
sched/tracing: Append prev_state to tp args instead
Commit fa2c3254d7cf (sched/tracing: Don't re-read p->state when emitting sched_switch event, 2022-01-20) added a new prev_state argument to the sched_switch tracepoint, before the prev task_struct pointer.
This reordering of arguments broke BPF programs that use the raw tracepoint (e.g. tp_btf programs). The type of the second argument has changed and existing programs that assume a task_struct* argument (e.g. for bpf_task_storage access) will now fail to verify.
If we instead append the new argument to the end, all existing programs would continue to work and can conditionally extract the prev_state argument on supported kernel versions.
By reordering the parameters (again) we can get back up and building.
---
Hi all,
This is more than likely NOT a correct fix, but while working on the yocto -dev reference kernel, I ran into this build failure against 5.18-rc7.
I didn't see any sign of another fix on the mailing list, so I wanted to send this in case anyone else runs into the failure and to check to see if there's a better fix in progress.
Cheers,
Bruce
include/instrumentation/events/sched.h | 15 ++++++++------- 1 file changed, 8 insertions(+), 7 deletions(-)
include/instrumentation/events/sched.h +8 -7
@@ -23,8 +23,9 @@#if (LTTNG_LINUX_VERSION_CODE >=LTTNG_KERNEL_VERSION (5 ,18 ,0 ))static inline long __trace_sched_switch_state (bool preempt, -unsigned int prev_state, -struct task_struct *p) +struct task_struct *p, +struct task_struct *n, +unsigned int prev_state ) {unsigned int state;
@@ -356,20 +357,20 @@ LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new, LTTNG_TRACEPOINT_EVENT(sched_switch, TP_PROTO(bool preempt, - unsigned int prev_state, struct task_struct *prev, - struct task_struct *next), + struct task_struct *next, + unsigned int prev_state), - TP_ARGS(preempt, prev_state, prev, next), + TP_ARGS(preempt, prev, next, prev_state), TP_FIELDS( ctf_array_text(char, prev_comm, prev->comm, TASK_COMM_LEN) ctf_integer(pid_t, prev_tid, prev->pid) ctf_integer(int, prev_prio, prev->prio - MAX_RT_PRIO) #ifdef CONFIG_LTTNG_EXPERIMENTAL_BITWISE_ENUM - ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev_state, prev)) + ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev, next, prev_state)) #else - ctf_integer(long, prev_state, __trace_sched_switch_state(preempt, prev_state, prev)) + ctf_integer(long, prev_state, __trace_sched_switch_state(preempt, prev, next, prev_state)) #endif ctf_array_text(char, next_comm, next->comm, TASK_COMM_LEN) ctf_integer(pid_t, next_tid, next->pid)
Signature · 6 lines
2.19.1 _______________________________________________ lttng-dev mailing list [email protected] https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
Quoted 40 lines · ----- On May 19, 2022, at 11:02 AM, Bruce Ashfield via lttng-dev…
From: Bruce Ashfield <[email protected]>
The commit [fix: sched/tracing: Don't re-read
p->statewhen emittingsched_switchevent (v5.18)] was correct, but the kernel changed their mind with the following commit:commit
9c2136be0878c88c53dea26943ce40bb03ad8d8d
Author: Delyan Kratunov <[email protected]>
Date: Wed May 11 18:28:36 2022 +0000sched/tracing: Append
prev_stateto tp args insteadCommit
fa2c3254d7cf(sched/tracing: Don't re-readp->statewhen emittingsched_switchevent, 2022-01-20) added a newprev_stateargument to thesched_switchtracepoint, before the prevtask_structpointer.This reordering of arguments broke BPF programs that use the raw tracepoint (e.g.
tp_btfprograms). The type of the second argument has changed and existing programs that assume a task_struct* argument (e.g. forbpf_task_storageaccess) will now fail to verify.If we instead append the new argument to the end, all existing programs would continue to work and can conditionally extract the
prev_stateargument on supported kernel versions.Fixes fa2c3254d7cf (sched/tracing: Don't re-read p->state when emitting
sched_switchevent, 2022-01-20)Signed-off-by Delyan Kratunov <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>Link:
https://lkml.kernel.org/r/[email protected]By reordering the parameters (again) we can get back up and building.
Signed-off-by Bruce Ashfield <[email protected]>---
Hi all,
Quoted 7 lines
This is more than likely NOT a correct fix, but while working on the yocto -dev reference kernel, I ran into this build failure against 5.18-rc7.
I didn't see any sign of another fix on the mailing list, so I wanted to send this in case anyone else runs into the failure and to check to see if there's a better fix in progress.
Hi Bruce,
Please see: https://review.lttng.org/c/lttng-modules/+/8045
I was planning to merge it today, unless you have objections.
I notice that in your own patch, you modify the arguments passed to __trace_sched_switch_state(), althrough I don't see them changing in the upstream kernel. Am I missing something ?
Thanks,
Mathieu
Quoted 64 lines
Cheers,
Bruce
include/instrumentation/events/sched.h| 15 ++++++++-------1 file changed, 8 insertions(+), 7 deletions(-)include/instrumentation/events/sched.h +7
-7@@ -23,8 +23,9 @@#if (LTTNG_LINUX_VERSION_CODE >=LTTNG_KERNEL_VERSION (5 ,18 ,0 ))static inline long __trace_sched_switch_state (bool preempt, -unsigned int prev_state, -struct task_struct *p) +struct task_struct *p, +struct task_struct *n, +unsigned int prev_state ) {unsigned int state;@@ -356,20 +357,20 @@ LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,LTTNG_TRACEPOINT_EVENT (sched_switch, TP_PROTO(bool preempt, - unsigned int prev_state, struct task_struct *prev, - struct task_struct *next), + struct task_struct *next, + unsigned int prev_state), -TP_ARGS (preempt, prev_state, prev, next), +TP_ARGS (preempt, prev, next, prev_state),TP_FIELDS ( ctf_array_text(char, prev_comm, prev->comm, TASK_COMM_LEN) ctf_integer(pid_t, prev_tid, prev->pid) ctf_integer(int, prev_prio, prev->prio - MAX_RT_PRIO) #ifdef CONFIG_LTTNG_EXPERIMENTAL_BITWISE_ENUM - ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev_state, prev)) +ctf_enum (task_state, long, prev_state,__trace_sched_switch_state (preempt, prev, next, prev_state)) #else - ctf_integer(long, prev_state, __trace_sched_switch_state(preempt, prev_state,prev))
include/instrumentation/events/sched.h +1
-0+ ctf_integer(long, prev_state, __trace_sched_switch_state(preempt, prev, next,
prev_state))#endif ctf_array_text(char, next_comm, next->comm, TASK_COMM_LEN) ctf_integer(pid_t, next_tid, next->pid)--
2.19.1_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
Signature · 7 lines
Mathieu Desnoyers EfficiOS Inc. http://www.efficios.com _______________________________________________ lttng-dev mailing list [email protected] https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev
Quoted 54 lines · Mathieu Desnoyers
Quoted 40 lines · ----- On May 19, 2022, at 11:02 AM, Bruce Ashfield via lttng-dev…
From: Bruce Ashfield <[email protected]>
The commit [fix: sched/tracing: Don't re-read
p->statewhen emittingsched_switchevent (v5.18)] was correct, but the kernel changed their mind with the following commit:commit
9c2136be0878c88c53dea26943ce40bb03ad8d8d
Author: Delyan Kratunov <[email protected]>
Date: Wed May 11 18:28:36 2022 +0000sched/tracing: Append
prev_stateto tp args insteadCommit
fa2c3254d7cf(sched/tracing: Don't re-readp->statewhen emittingsched_switchevent, 2022-01-20) added a newprev_stateargument to thesched_switchtracepoint, before the prevtask_structpointer.This reordering of arguments broke BPF programs that use the raw tracepoint (e.g.
tp_btfprograms). The type of the second argument has changed and existing programs that assume a task_struct* argument (e.g. forbpf_task_storageaccess) will now fail to verify.If we instead append the new argument to the end, all existing programs would continue to work and can conditionally extract the
prev_stateargument on supported kernel versions.Fixes fa2c3254d7cf (sched/tracing: Don't re-read p->state when emitting
sched_switchevent, 2022-01-20)Signed-off-by Delyan Kratunov <[email protected]>Signed-off-by Peter Zijlstra (Intel) <[email protected]>Link:
https://lkml.kernel.org/r/[email protected]By reordering the parameters (again) we can get back up and building.
Signed-off-by Bruce Ashfield <[email protected]>---
Hi all,
Quoted 7 lines
This is more than likely NOT a correct fix, but while working on the yocto -dev reference kernel, I ran into this build failure against 5.18-rc7.
I didn't see any sign of another fix on the mailing list, so I wanted to send this in case anyone else runs into the failure and to check to see if there's a better fix in progress.
Hi Bruce,
Quoted 7 lines
Please see: https://review.lttng.org/c/lttng-modules/+/8045
I was planning to merge it today, unless you have objections.
I notice that in your own patch, you modify the arguments passed to
__trace_sched_switch_state(), althrough I don't see them changing in the upstream kernel. Am I missing something ?
No objects here. That is obviously better than mine :) I swear I saw the order of the arguments changing when I looked at the kernel, and had a secondary compile error in those calls .. but I very easily could have been mistaken.
Bruce
Quoted 74 lines
Thanks,
Mathieu
Quoted 57 lines
Cheers,
Bruce
include/instrumentation/events/sched.h| 15 ++++++++-------1 file changed, 8 insertions(+), 7 deletions(-)include/instrumentation/events/sched.h +7
-7@@ -23,8 +23,9 @@#if (LTTNG_LINUX_VERSION_CODE >=LTTNG_KERNEL_VERSION (5 ,18 ,0 ))static inline long __trace_sched_switch_state (bool preempt, -unsigned int prev_state, -struct task_struct *p) +struct task_struct *p, +struct task_struct *n, +unsigned int prev_state ) {unsigned int state;@@ -356,20 +357,20 @@ LTTNG_TRACEPOINT_EVENT_INSTANCE(sched_wakeup_template, sched_wakeup_new,LTTNG_TRACEPOINT_EVENT (sched_switch, TP_PROTO(bool preempt, - unsigned int prev_state, struct task_struct *prev, - struct task_struct *next), + struct task_struct *next, + unsigned int prev_state), -TP_ARGS (preempt, prev_state, prev, next), +TP_ARGS (preempt, prev, next, prev_state),TP_FIELDS ( ctf_array_text(char, prev_comm, prev->comm, TASK_COMM_LEN) ctf_integer(pid_t, prev_tid, prev->pid) ctf_integer(int, prev_prio, prev->prio - MAX_RT_PRIO) #ifdef CONFIG_LTTNG_EXPERIMENTAL_BITWISE_ENUM - ctf_enum(task_state, long, prev_state, __trace_sched_switch_state(preempt, prev_state, prev)) +ctf_enum (task_state, long, prev_state,__trace_sched_switch_state (preempt, prev, next, prev_state)) #else - ctf_integer(long, prev_state, __trace_sched_switch_state(preempt, prev_state,prev))
+ ctf_integer(long,prev_state, __trace_sched_switch_state(preempt, prev, next,prev_state))#endif ctf_array_text(char, next_comm, next->comm, TASK_COMM_LEN) ctf_integer(pid_t, next_tid, next->pid)Quoted 7 lines
--
2.19.1_______________________________________________
lttng-dev mailing list
[email protected]
https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev--
Mathieu Desnoyers
EfficiOS Inc.
http://www.efficios.com
Signature · 7 lines
- Thou shalt not follow the NULL pointer, for chaos and madness await thee at its end - "Use the force Harry" - Gandalf, Star Trek II _______________________________________________ lttng-dev mailing list [email protected] https://lists.lttng.org/cgi-bin/mailman/listinfo/lttng-dev