Introduce a bpf hook to handle psi events. The primary intended
purpose of this hook is to declare OOM events based on the reaching
a certain memory pressure level, similar to what systemd-oomd and oomd
are doing in userspace.
Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev>
---
kernel/sched/psi.c | 36 +++++++++++++++++++++++++++++++++++-
1 file changed, 35 insertions(+), 1 deletion(-)
diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
index 1396674fa722..4c4eb4ead8f6 100644
--- a/kernel/sched/psi.c
+++ b/kernel/sched/psi.c
@@ -176,6 +176,32 @@ static void psi_avgs_work(struct work_struct *work);
static void poll_timer_fn(struct timer_list *t);
+#ifdef CONFIG_BPF_SYSCALL
+__bpf_hook_start();
+
+__weak noinline int bpf_handle_psi_event(struct psi_trigger *t)
+{
+ return 0;
+}
+
+__bpf_hook_end();
+
+BTF_KFUNCS_START(bpf_psi_hooks)
+BTF_ID_FLAGS(func, bpf_handle_psi_event, KF_SLEEPABLE)
+BTF_KFUNCS_END(bpf_psi_hooks)
+
+static const struct btf_kfunc_id_set bpf_psi_hook_set = {
+ .owner = THIS_MODULE,
+ .set = &bpf_psi_hooks,
+};
+
+#else
+static inline int bpf_handle_psi_event(struct psi_trigger *t)
+{
+ return 0;
+}
+#endif
+
static void group_init(struct psi_group *group)
{
int cpu;
@@ -489,6 +515,7 @@ static void update_triggers(struct psi_group *group, u64 now,
/* Generate an event */
if (cmpxchg(&t->event, 0, 1) == 0) {
+ bpf_handle_psi_event(t);
if (t->of)
kernfs_notify(t->of->kn);
else
@@ -1655,6 +1682,8 @@ static const struct proc_ops psi_irq_proc_ops = {
static int __init psi_proc_init(void)
{
+ int err = 0;
+
if (psi_enable) {
proc_mkdir("pressure", NULL);
proc_create("pressure/io", 0666, NULL, &psi_io_proc_ops);
@@ -1662,9 +1691,14 @@ static int __init psi_proc_init(void)
proc_create("pressure/cpu", 0666, NULL, &psi_cpu_proc_ops);
#ifdef CONFIG_IRQ_TIME_ACCOUNTING
proc_create("pressure/irq", 0666, NULL, &psi_irq_proc_ops);
+#endif
+#ifdef CONFIG_BPF_SYSCALL
+ err = register_btf_fmodret_id_set(&bpf_psi_hook_set);
+ if (err)
+ pr_err("error while registering bpf psi hooks: %d", err);
#endif
}
- return 0;
+ return err;
}
module_init(psi_proc_init);
--
2.49.0.901.g37484f566f-googOn Sun, Apr 27, 2025 at 8:37 PM Roman Gushchin <roman.gushchin@linux.dev> wrote:
>
> Introduce a bpf hook to handle psi events. The primary intended
> purpose of this hook is to declare OOM events based on the reaching
> a certain memory pressure level, similar to what systemd-oomd and oomd
> are doing in userspace.
It's a bit awkward that this requires additional userspace action to
create PSI triggers. I have almost no experience with BPF, so this
might be a stupid question, but maybe we could provide a bpf kfunc for
the BPF handler to register its PSI trigger(s) upon handler
registration?
>
> Signed-off-by: Roman Gushchin <roman.gushchin@linux.dev>
> ---
> kernel/sched/psi.c | 36 +++++++++++++++++++++++++++++++++++-
> 1 file changed, 35 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/sched/psi.c b/kernel/sched/psi.c
> index 1396674fa722..4c4eb4ead8f6 100644
> --- a/kernel/sched/psi.c
> +++ b/kernel/sched/psi.c
> @@ -176,6 +176,32 @@ static void psi_avgs_work(struct work_struct *work);
>
> static void poll_timer_fn(struct timer_list *t);
>
> +#ifdef CONFIG_BPF_SYSCALL
> +__bpf_hook_start();
> +
> +__weak noinline int bpf_handle_psi_event(struct psi_trigger *t)
> +{
> + return 0;
> +}
> +
> +__bpf_hook_end();
> +
> +BTF_KFUNCS_START(bpf_psi_hooks)
> +BTF_ID_FLAGS(func, bpf_handle_psi_event, KF_SLEEPABLE)
> +BTF_KFUNCS_END(bpf_psi_hooks)
> +
> +static const struct btf_kfunc_id_set bpf_psi_hook_set = {
> + .owner = THIS_MODULE,
> + .set = &bpf_psi_hooks,
> +};
> +
> +#else
> +static inline int bpf_handle_psi_event(struct psi_trigger *t)
> +{
> + return 0;
> +}
> +#endif
> +
> static void group_init(struct psi_group *group)
> {
> int cpu;
> @@ -489,6 +515,7 @@ static void update_triggers(struct psi_group *group, u64 now,
>
> /* Generate an event */
> if (cmpxchg(&t->event, 0, 1) == 0) {
> + bpf_handle_psi_event(t);
> if (t->of)
> kernfs_notify(t->of->kn);
> else
> @@ -1655,6 +1682,8 @@ static const struct proc_ops psi_irq_proc_ops = {
>
> static int __init psi_proc_init(void)
> {
> + int err = 0;
> +
> if (psi_enable) {
> proc_mkdir("pressure", NULL);
> proc_create("pressure/io", 0666, NULL, &psi_io_proc_ops);
> @@ -1662,9 +1691,14 @@ static int __init psi_proc_init(void)
> proc_create("pressure/cpu", 0666, NULL, &psi_cpu_proc_ops);
> #ifdef CONFIG_IRQ_TIME_ACCOUNTING
> proc_create("pressure/irq", 0666, NULL, &psi_irq_proc_ops);
> +#endif
> +#ifdef CONFIG_BPF_SYSCALL
> + err = register_btf_fmodret_id_set(&bpf_psi_hook_set);
> + if (err)
> + pr_err("error while registering bpf psi hooks: %d", err);
> #endif
> }
> - return 0;
> + return err;
> }
> module_init(psi_proc_init);
>
> --
> 2.49.0.901.g37484f566f-goog
>
On Tue, Apr 29, 2025 at 05:28:59PM -0700, Suren Baghdasaryan wrote: > On Sun, Apr 27, 2025 at 8:37 PM Roman Gushchin <roman.gushchin@linux.dev> wrote: > > > > Introduce a bpf hook to handle psi events. The primary intended > > purpose of this hook is to declare OOM events based on the reaching > > a certain memory pressure level, similar to what systemd-oomd and oomd > > are doing in userspace. > > It's a bit awkward that this requires additional userspace action to > create PSI triggers. I have almost no experience with BPF, so this > might be a stupid question, but maybe we could provide a bpf kfunc for > the BPF handler to register its PSI trigger(s) upon handler > registration? It looks like it's doable using struct_ops path: the .init callback can create psi triggers and "attach" them to the loaded bpf program. But I need to figure out the details. Good point, thank you!
Hi Roman,
kernel test robot noticed the following build warnings:
[auto build test WARNING on akpm-mm/mm-everything]
url: https://github.com/intel-lab-lkp/linux/commits/Roman-Gushchin/mm-introduce-a-bpf-hook-for-OOM-handling/20250428-113742
base: https://git.kernel.org/pub/scm/linux/kernel/git/akpm/mm.git mm-everything
patch link: https://lore.kernel.org/r/20250428033617.3797686-10-roman.gushchin%40linux.dev
patch subject: [PATCH rfc 09/12] sched: psi: bpf hook to handle psi events
config: sh-randconfig-001-20250428 (https://download.01.org/0day-ci/archive/20250428/202504281309.smYiDStM-lkp@intel.com/config)
compiler: sh4-linux-gcc (GCC) 10.5.0
reproduce (this is a W=1 build): (https://download.01.org/0day-ci/archive/20250428/202504281309.smYiDStM-lkp@intel.com/reproduce)
If you fix the issue in a separate patch/commit (i.e. not just a new version of
the same patch/commit), kindly add following tags
| Reported-by: kernel test robot <lkp@intel.com>
| Closes: https://lore.kernel.org/oe-kbuild-all/202504281309.smYiDStM-lkp@intel.com/
All warnings (new ones prefixed by >>):
In file included from kernel/sched/build_utility.c:94:
>> kernel/sched/psi.c:193:38: warning: 'bpf_psi_hook_set' defined but not used [-Wunused-const-variable=]
193 | static const struct btf_kfunc_id_set bpf_psi_hook_set = {
| ^~~~~~~~~~~~~~~~
vim +/bpf_psi_hook_set +193 kernel/sched/psi.c
192
> 193 static const struct btf_kfunc_id_set bpf_psi_hook_set = {
194 .owner = THIS_MODULE,
195 .set = &bpf_psi_hooks,
196 };
197
--
0-DAY CI Kernel Test Service
https://github.com/intel/lkp-tests/wiki
© 2016 - 2026 Red Hat, Inc.