Hi,
Here are 2 patches I am thinking of sending with next coresched series. Let me
know what you think. These are for not slowing down certain AMD devices if they
don't have vulnerabilities that require coresched.
1/2 - keep coresched disabled if device doesn't have vulnerability.
2/2 - add debug information.
In the future, if needed we could add more options to make it possible to
force-enable coresched. But right now I don't see a need for that, till a
usecase arises.
Joel Fernandes (Google) (2):
x86/bugs: Disable coresched on hardware that does not need it
sched/debug: Add debug information about whether coresched is enabled
arch/x86/kernel/cpu/bugs.c | 8 ++++++++
kernel/sched/core.c | 7 +++++++
kernel/sched/debug.c | 4 ++++
kernel/sched/sched.h | 5 +++++
4 files changed, 24 insertions(+)
--
2.29.2.222.g5d2a92d10f8-goog
It is useful to see whether coresched is enabled or not, especially in
devices that don't need it. Add information about the same to
/proc/sched_debug.
Signed-off-by: Joel Fernandes (Google) <[email protected]>
---
kernel/sched/debug.c | 4 ++++
1 file changed, 4 insertions(+)
diff --git a/kernel/sched/debug.c b/kernel/sched/debug.c
index 88bf45267672..935b68be18cd 100644
--- a/kernel/sched/debug.c
+++ b/kernel/sched/debug.c
@@ -782,6 +782,10 @@ static void sched_debug_header(struct seq_file *m)
"sysctl_sched_tunable_scaling",
sysctl_sched_tunable_scaling,
sched_tunable_scaling_names[sysctl_sched_tunable_scaling]);
+#ifdef CONFIG_SCHED_CORE
+ SEQ_printf(m, " .%-40s: %d\n", "core_sched_enabled",
+ !!static_branch_likely(&__sched_core_enabled));
+#endif
SEQ_printf(m, "\n");
}
--
2.29.2.222.g5d2a92d10f8-goog