Skip to content

Use scx_bpf_cpu_curr() instead of deprecated scx_bpf_cpu_rq() #2734

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.

Sign up for GitHub

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

Jump to bottom
Open
wants to merge 7 commits into
base: main
Choose a base branch
from
Open

Use scx_bpf_cpu_curr() instead of deprecated scx_bpf_cpu_rq() #2734

wants to merge 7 commits into from
+29 −21

Conversation

cloehle
Copy link
Contributor

@cloehle cloehle commented Sep 4, 2025 •
edited
Loading

The new scx_bpf_cpu_curr() is queued for v6.18.
Use them instead of the soon deprecated scx_bpf_cpu_rq().

Error handling (on NULL) is kept as is for the schedulers

This needs the queued:
https://lore.kernel.org/lkml/[email protected]/

@cloehle cloehle mentioned this pull request Sep 4, 2025
Closed
htejun
htejun reviewed Sep 4, 2025
View reviewed changes
Copy link
Contributor

@htejun htejun left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh, wait, you need to port the compat macro too.

arighi
arighi reviewed Sep 4, 2025
View reviewed changes
scheds/rust/scx_cosmos/src/bpf/main.bpf.c
@@ -325,13 +325,13 @@ static inline const struct cpumask *get_idle_smtmask(s32 cpu)
*/
static inline bool is_cpu_idle(s32 cpu)
{
struct rq *rq = scx_bpf_cpu_rq(cpu);
struct task_struct *p = __COMPAT_scx_bpf_cpu_curr(cpu);
Copy link
Contributor

@arighi arighi Sep 4, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

We need to add __COMPAT_scx_bpf_cpu_curr() https://git.kernel.org/pub/scm/linux/kernel/git/tj/sched_ext.git/tree/tools/sched_ext/include/scx/compat.bpf.h?h=for-6.18#n238 to scheds/include/scx/compat.bpf.h or this won't build.

Copy link
Contributor Author

@cloehle cloehle Sep 4, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Ah I thought these were getting pulled in from the kernel source (at some point)?
Do we update both compat.bpf.h and common.h separately for kernel and scx repo?

Edit: Nevermind, I misunderstood the process, I've added the two commits (definitions and Andrea's compat helper) here now, i.e. this PR compiles and runs self-sufficient if the kernel includes the mentioned series.

scheds/rust/scx_cosmos/src/bpf/main.bpf.c
return false;
}
return rq->curr->flags & PF_IDLE;
return !!(p->flags & PF_IDLE);
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

nit: we probably don't need !! here, since the result would be cast to bool anyway.

@cloehle
scx: Add scx_bpf_locked_rq() and scx_bpf_cpu_curr()
d2d830d 
The new functions provide a safer way to access a rq or a remote
rq->curr respectively. They are both part of v6.18.

See kernel commits:
("sched_ext: Introduce scx_bpf_locked_rq()")
("sched_ext: Introduce scx_bpf_cpu_curr()")

Signed-off-by: Christian Loehle <[email protected]>
@cloehle cloehle force-pushed the cloehle/scx_bpf_cpu_rq-compat branch from 8d41e53 to 11bb34c Compare September 4, 2025 22:50
@multics69
Copy link
Contributor

Overall LGTM.

@cloehle cloehle force-pushed the cloehle/scx_bpf_cpu_rq-compat branch from 11bb34c to 9deff4b Compare September 4, 2025 23:32
@arighi arighi added this pull request to the merge queue Sep 6, 2025
@github-merge-queue github-merge-queue bot removed this pull request from the merge queue due to failed status checks Sep 6, 2025
arighi
arighi requested changes Sep 6, 2025
View reviewed changes
Copy link
Contributor

@arighi arighi left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hm... I'm a bit confused.

We have:

BTF_ID_FLAGS(func, scx_bpf_cpu_curr, KF_RET_NULL | KF_RCU)

I thought the verifier would complain if we try to use scx_bpf_cpu_curr() outside of a bpf_rcu_read_lock/unlock() section, which is exactly what we're doing here. But the verifier is still happy and the kernel is (correctly) complaining:

[    5.938735] =============================
[    5.939878] WARNING: suspicious RCU usage
[    5.941002] sched_ext: BPF scheduler "cosmos_1.0.2_gd0e71ca_x86_64_unknown_linux_gnu_debug" enabled
[    5.943254] 6.17.0-rc1 #1-NixOS Not tainted
[    5.944477] -----------------------------
[    5.945636] kernel/sched/ext.c:6415 suspicious rcu_dereference_check() usage!

So we can fix this PR adding the proper bpf_rcu_read_lock/unlock() around __COMPAT_scx_bpf_cpu_curr(), but I'd like to understand why the verifier is happy.

@etsal do you have any idea? Thanks.

multics69
multics69 approved these changes Sep 7, 2025
View reviewed changes
Copy link
Contributor

@multics69 multics69 left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Thanks! It looks good to me.

@cloehle cloehle force-pushed the cloehle/scx_bpf_cpu_rq-compat branch from 9deff4b to eb700fc Compare September 7, 2025 21:02
@cloehle
compat: Add compat helper for scx_bpf_cpu_curr()
1acbf73 
Introduce a compatibility helper that allows BPF schedulers to use
scx_bpf_cpu_curr() on older kernels.

See kernel commit
("sched_ext: Add compat helper for scx_bpf_cpu_curr()")

Signed-off-by: Andrea Righi <[email protected]>
Signed-off-by: Christian Loehle <[email protected]>
@cloehle
scx_cosmos: Use scx_bpf_cpu_curr()
b2bae17 
Use the new scx_bpf_cpu_curr() introduced in v6.18 as a safer way to
access rq->curr instead of the deprecated scx_bpf_cpu_rq().

Signed-off-by: Christian Loehle <[email protected]>
@cloehle
scx_flash: Use scx_bpf_cpu_curr()
360b971 
Use the new scx_bpf_cpu_curr() introduced in v6.18 as a safer way to
access rq->curr instead of the deprecated scx_bpf_cpu_rq().

Signed-off-by: Christian Loehle <[email protected]>
@cloehle
scx_lavd: Use scx_bpf_cpu_curr()
cd98947 
Use the new scx_bpf_cpu_curr() introduced in v6.18 as a safer way to
access rq->curr instead of the deprecated scx_bpf_cpu_rq().

Signed-off-by: Christian Loehle <[email protected]>
@cloehle
scx_layered: Use scx_bpf_cpu_curr()
fa74176 
Use the new scx_bpf_cpu_curr() introduced in v6.18 as a safer way to
access rq->curr instead of the deprecated scx_bpf_cpu_rq().

Signed-off-by: Christian Loehle <[email protected]>
@cloehle
scx_tickless: Use scx_bpf_cpu_curr()
5cd7c5f 
Use the new scx_bpf_cpu_curr() introduced in v6.18 as a safer way to
access rq->curr instead of the deprecated scx_bpf_cpu_rq().

Signed-off-by: Christian Loehle <[email protected]>
@cloehle cloehle force-pushed the cloehle/scx_bpf_cpu_rq-compat branch from eb700fc to 5cd7c5f Compare September 7, 2025 21:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@htejun htejun htejun approved these changes

@arighi arighi arighi requested changes

@multics69 multics69 multics69 approved these changes

Requested changes must be addressed to merge this pull request.

Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@cloehle @multics69 @arighi @htejun