bpf: Add support for sleepable tracepoint programs by kernel-patches-daemon-bpf[bot] · Pull Request #11398 · kernel-patches/bpf

kernel-patches-daemon-bpf · 2026-03-13T21:15:47Z

Pull request for series with
subject: bpf: Add support for sleepable tracepoint programs
version: 4
url: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584

kernel-patches-daemon-bpf · 2026-03-13T21:15:48Z

Upstream branch: ca0f39a
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

kernel-patches-review-bot · 2026-03-13T21:31:57Z

commit 2b9c4f9a741437688351ae80490daa5b8ce7556e
Author: Mykyta Yatsenko <yatsenko@meta.com>

selftests/bpf: Add tests for sleepable tracepoint programs

Adds functional tests for sleepable tracepoint programs (tp_btf.s,
raw_tp.s, tp.s) attaching to nanosleep, plus a negative test for
non-faultable tracepoints, and updated verifier/sleepable.c tests.

> Add a negative test (test_sleepable_raw_tp_fail) that verifies
> sleepable programs are rejected on non-faultable tracepoints
> (sched_switch).

The commit message references test_sleepable_raw_tp_fail, but this
name doesn't appear anywhere in the committed code. It looks like a
leftover from v2 of the series, where the negative test lived in a
separate file (progs/test_sleepable_raw_tp_fail.c).

In this version the negative test is handle_sched_switch in
progs/test_sleepable_tracepoints.c, tested via the RUN_TESTS macro.
Should the commit message be updated to reflect the current naming?

AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

In-Reply-To-Subject: selftests/bpf: Add tests for sleepable tracepoint programs
CI run summary: https://github.com/kernel-patches/bpf/actions/runs/23070834076

kernel-patches-review-bot · 2026-03-13T21:33:10Z

commit b7b6029199ace23f1ddaf7529035791cd2799bf4
Author: Mykyta Yatsenko <yatsenko@meta.com>

libbpf: Add section handlers for sleepable tracepoints

Add SEC_DEF entries for sleepable tracepoint variants and update
attach_tp() and attach_raw_tp() to parse the new section name
prefixes.

> diff --git a/tools/lib/bpf/libbpf.c b/tools/lib/bpf/libbpf.c
> index 0662d72bad20..ef79a980c0bb 100644
> --- a/tools/lib/bpf/libbpf.c
> +++ b/tools/lib/bpf/libbpf.c

[ ... ]

> @@ -12997,11 +13009,19 @@ static int attach_tp(const struct bpf_program *prog, long cookie, struct bpf_lin
>  	if (!sec_name)
>  		return -ENOMEM;
>
> -	/* extract "tp/<category>/<name>" or "tracepoint/<category>/<name>" */
> -	if (str_has_pfx(prog->sec_name, "tp/"))
> -		tp_cat = sec_name + sizeof("tp/") - 1;
> -	else
> -		tp_cat = sec_name + sizeof("tracepoint/") - 1;
> +	/* extract "<prefix><category>/<name>" */
> +	tp_cat = NULL;
> +	for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
> +		if (str_has_pfx(prog->sec_name, prefixes[i])) {
> +			tp_cat = sec_name + strlen(prefixes[i]);
> +			break;
> +		}
> +	}
> +	if (!tp_cat) {
> +		free(sec_name);
> +		return -EINVAL;
> +	}

Should attach_tp() handle bare SEC("tp.s") and SEC("tracepoint.s")
the same way it handles bare SEC("tp") and SEC("tracepoint")?

sec_def_matches() allows bare section names for "+" suffixed
SEC_DEF entries -- when sec_name[len] == '\0' it returns true.
So SEC("tp.s") matches SEC_DEF("tp.s+") and reaches attach_tp().

The existing bare-name guard only checks for "tp" and "tracepoint":

    if (strcmp(prog->sec_name, "tp") == 0 ||
        strcmp(prog->sec_name, "tracepoint") == 0)
        return 0;

Since none of the prefixes ("tp.s/", "tp/", "tracepoint.s/",
"tracepoint/") match the bare string "tp.s", tp_cat stays NULL
and the function returns -EINVAL. In bpf_object__attach_skeleton()
this becomes a fatal error that stops the entire skeleton attach.

attach_raw_tp() in the same patch handles the equivalent bare case
correctly -- when sec_name[pfx_len] == '\0' it returns 0:

    for (i = 0; i < ARRAY_SIZE(prefixes); i++) {
        ...
        if (prog->sec_name[pfx_len] == '\0')
            return 0;
        ...
    }

The convention across libbpf attach handlers (attach_kprobe,
attach_raw_tp) is to return 0 with *link = NULL for bare section
names, signaling "no auto-attach" without error.

AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

In-Reply-To-Subject: libbpf: Add section handlers for sleepable tracepoints
CI run summary: https://github.com/kernel-patches/bpf/actions/runs/23070834076

kernel-patches-review-bot · 2026-03-13T21:38:44Z

commit ac1bb22ffd7320682f2a63c12ab84193029f707b
Author: Mykyta Yatsenko <yatsenko@meta.com>

bpf: Add sleepable support for classic tracepoint programs

Adds trace_call_bpf_faultable() for faultable tracepoints supporting
sleepable BPF programs, restructures perf_syscall_enter/exit to run
BPF before preempt_disable, and gates sleepable tracepoint programs
to syscall tracepoints at attach time.

> diff --git a/kernel/trace/trace_syscalls.c b/kernel/trace/trace_syscalls.c
> index 37317b81fcda..e9ddf9db2b56 100644
> --- a/kernel/trace/trace_syscalls.c
> +++ b/kernel/trace/trace_syscalls.c
> @@ -1372,26 +1372,27 @@ static DECLARE_BITMAP(enabled_perf_exit_syscalls, NR_syscalls);
>  static int sys_perf_refcount_enter;
>  static int sys_perf_refcount_exit;
>
> -static int perf_call_bpf_enter(struct trace_event_call *call, struct pt_regs *regs,
> +static int perf_call_bpf_enter(struct trace_event_call *call,
>  			       struct syscall_metadata *sys_data,
> -			       struct syscall_trace_enter *rec)
> +			       int syscall_nr, unsigned long *args)
>  {
>  	struct syscall_tp_t {
>  		struct trace_entry ent;
>  		int syscall_nr;
>  		unsigned long args[SYSCALL_DEFINE_MAXARGS];
>  	} __aligned(8) param;
> +	struct pt_regs regs;
>  	int i;
>
>  	BUILD_BUG_ON(sizeof(param.ent) < sizeof(void *));
>
> -	/* bpf prog requires 'regs' to be the first member in the ctx (a.k.a. &param) */
> -	perf_fetch_caller_regs(regs);
> +	/* bpf prog requires 'regs' to be the first member in the ctx */
> +	perf_fetch_caller_regs(&regs);

Should regs be zero-initialized here? The perf_fetch_caller_regs()
comment in include/linux/perf_event.h states:

    NOTE: assumes @regs is otherwise already 0 filled; this is
    important for things like PERF_SAMPLE_REGS_INTR.

The function only sets ip, sp, cs, and flags, leaving all other
register fields at whatever happened to be on the stack.

Previously, the regs pointer came from perf_trace_buf_alloc() which
returned this_cpu_ptr(&__perf_regs[rctx]), a per-cpu variable that
is zero-initialized at boot. The new stack-local struct does not
get that zeroing.

If a BPF program calls bpf_perf_event_output(), bpf_perf_event_output_tp()
extracts this pointer:

    struct pt_regs *regs = *(struct pt_regs **)tp_buff;

and passes it through to perf_event_output(), where uninitialized
register fields could end up in perf samples.

Would initializing to zero fix this?

    struct pt_regs regs = {};

> +	*(struct pt_regs **)&param = &regs;

[ ... ]

> @@ -1515,21 +1519,22 @@ static void perf_sysenter_disable(struct trace_event_call *call)
>  		syscall_fault_buffer_disable();
>  }
>
> -static int perf_call_bpf_exit(struct trace_event_call *call, struct pt_regs *regs,
> -			      struct syscall_trace_exit *rec)
> +static int perf_call_bpf_exit(struct trace_event_call *call,
> +			      int syscall_nr, long ret_val)
>  {
>  	struct syscall_tp_t {
>  		struct trace_entry ent;
>  		int syscall_nr;
>  		unsigned long ret;
>  	} __aligned(8) param;
> -
> -	/* bpf prog requires 'regs' to be the first member in the ctx (a.k.a. &param) */
> -	perf_fetch_caller_regs(regs);
> +	struct pt_regs regs;
> +
> +	/* bpf prog requires 'regs' to be the first member in the ctx */
> +	perf_fetch_caller_regs(&regs);

Same uninitialized regs concern as perf_call_bpf_enter() above.

> +	*(struct pt_regs **)&param = &regs;

[ ... ]

AI reviewed your patch. Please fix the bug or email reply why it's not a bug.
See: https://github.com/kernel-patches/vmtest/blob/master/ci/claude/README.md

In-Reply-To-Subject: bpf: Add sleepable support for classic tracepoint programs
CI run summary: https://github.com/kernel-patches/bpf/actions/runs/23070834076

kernel-patches-daemon-bpf · 2026-03-13T21:41:25Z

Forwarding comment 4058102918 via email
In-Reply-To: 20260313-sleepable_tracepoints-v4-5-debc688a66b3@meta.com
Patch: https://patchwork.kernel.org/project/netdevbpf/patch/20260313-sleepable_tracepoints-v4-5-debc688a66b3@meta.com/

kernel-patches-daemon-bpf · 2026-03-13T21:41:27Z

Forwarding comment 4058107515 via email
In-Reply-To: 20260313-sleepable_tracepoints-v4-4-debc688a66b3@meta.com
Patch: https://patchwork.kernel.org/project/netdevbpf/patch/20260313-sleepable_tracepoints-v4-4-debc688a66b3@meta.com/

kernel-patches-daemon-bpf · 2026-03-13T21:41:28Z

Forwarding comment 4058130661 via email
In-Reply-To: 20260313-sleepable_tracepoints-v4-2-debc688a66b3@meta.com
Patch: https://patchwork.kernel.org/project/netdevbpf/patch/20260313-sleepable_tracepoints-v4-2-debc688a66b3@meta.com/

kernel-patches-daemon-bpf · 2026-03-14T02:15:33Z

Upstream branch: bb41fce
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

kernel-patches-daemon-bpf · 2026-03-16T16:15:56Z

Upstream branch: bb41fce
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

kernel-patches-daemon-bpf · 2026-03-16T16:35:17Z

Upstream branch: 202e42e
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

kernel-patches-daemon-bpf · 2026-03-16T18:37:59Z

Upstream branch: 6c8e1a9
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

Rework __bpf_trace_run() to support sleepable BPF programs by using explicit RCU flavor selection, following the uprobe_prog_run() pattern. For sleepable programs, use rcu_read_lock_trace() for lifetime protection and add a might_fault() annotation. For non-sleepable programs, use the regular rcu_read_lock(). Replace the combined rcu_read_lock_dont_migrate() with separate rcu_read_lock()/ migrate_disable() calls, since sleepable programs need rcu_read_lock_trace() instead of rcu_read_lock(). Remove the preempt_disable_notrace/preempt_enable_notrace pair from the faultable tracepoint BPF probe wrapper in bpf_probe.h, since preemption management is now handled inside __bpf_trace_run(). This enables both BTF-based raw tracepoints (tp_btf.s) and classic raw tracepoints (raw_tp.s) to run sleepable BPF programs when attached to faultable tracepoints (e.g. syscall tracepoints). Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>

Add trace_call_bpf_faultable(), a variant of trace_call_bpf() for faultable tracepoints that supports sleepable BPF programs. It uses rcu_read_lock_trace() for lifetime protection and bpf_prog_run_array_uprobe() for per-program RCU flavor selection, following the uprobe_prog_run() pattern. Uses preempt-safe this_cpu_inc_return/this_cpu_dec for the bpf_prog_active recursion counter since preemption is enabled in this context. Restructure perf_syscall_enter() and perf_syscall_exit() to run BPF filter before perf event processing. Previously, BPF ran after the per-cpu perf trace buffer was allocated under preempt_disable, requiring cleanup via perf_swevent_put_recursion_context() on filter. Now BPF runs in faultable context before preempt_disable, reading syscall arguments from local variables instead of the per-cpu trace record, removing the dependency on buffer allocation. This allows sleepable BPF programs to execute and avoids unnecessary buffer allocation when BPF filters the event. The perf event submission path (buffer allocation, fill, submit) remains under preempt_disable as before. Add an attach-time check in __perf_event_set_bpf_prog() to reject sleepable BPF_PROG_TYPE_TRACEPOINT programs on non-syscall tracepoints, since only syscall tracepoints run in faultable context. Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>

Allow BPF_PROG_TYPE_RAW_TRACEPOINT, BPF_PROG_TYPE_TRACEPOINT, and BPF_TRACE_RAW_TP (tp_btf) programs to be sleepable by adding them to can_be_sleepable(). For BTF-based raw tracepoints (tp_btf), add a load-time check in bpf_check_attach_target() that rejects sleepable programs attaching to non-faultable tracepoints with a descriptive error message. For classic raw tracepoints (raw_tp), add an attach-time check in bpf_raw_tp_link_attach() that rejects sleepable programs on non-faultable tracepoints. The attach-time check is needed because the tracepoint name is not known at load time for classic raw_tp. Replace the verbose error message that enumerates allowed program types with a generic "program of this type cannot be sleepable" message, since the list of sleepable-capable types keeps growing. Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>

Add SEC_DEF entries for sleepable tracepoint variants: - "tp_btf.s+" for sleepable BTF-based raw tracepoints - "raw_tp.s+" for sleepable classic raw tracepoints - "raw_tracepoint.s+" (alias) - "tp.s+" for sleepable classic tracepoints - "tracepoint.s+" (alias) Update attach_raw_tp() to recognize "raw_tp.s" and "raw_tracepoint.s" prefixes when extracting the tracepoint name. Rewrite attach_tp() to use a prefix array including "tp.s/" and "tracepoint.s/" variants for proper section name parsing. Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>

Add functional tests for sleepable tracepoint programs that attach to the nanosleep syscall and use bpf_copy_from_user() to read user memory: - tp_btf: BTF-based raw tracepoint using SEC("tp_btf.s/sys_enter") with PT_REGS_PARM1_SYSCALL (non-CO-RE macro for BTF programs). - classic: Classic raw tracepoint using SEC("raw_tp.s/sys_enter") with PT_REGS_PARM1_CORE_SYSCALL (CO-RE macro needed for classic). - tracepoint: Classic tracepoint using SEC("tp.s/syscalls/sys_enter_nanosleep") receiving struct syscall_trace_enter with direct access to args[]. Add a negative test (test_sleepable_raw_tp_fail) that verifies sleepable programs are rejected on non-faultable tracepoints (sched_switch). Update verifier/sleepable.c tests: - Add "sleepable raw tracepoint accept" test for sys_enter. - Rename reject test and update error message to match the new descriptive "Sleepable program cannot attach to non-faultable tracepoint" message. Signed-off-by: Mykyta Yatsenko <yatsenko@meta.com>

kernel-patches-daemon-bpf · 2026-03-16T21:21:25Z

Upstream branch: 2364959
series: https://patchwork.kernel.org/project/netdevbpf/list/?series=1066584
version: 4

kernel-patches-daemon-bpf Bot added new bpf-next V4 labels Mar 13, 2026

kernel-patches-review-bot Bot added the ai-review label Mar 13, 2026

kernel-patches-daemon-bpf Bot added V4-ci-fail and removed ai-review labels Mar 13, 2026

kernel-patches-daemon-bpf Bot force-pushed the bpf-next_base branch from 15b24d7 to 43a1b48 Compare March 14, 2026 02:11

kernel-patches-daemon-bpf Bot force-pushed the series/1066584=>bpf-next branch from 2b9c4f9 to 3c91638 Compare March 14, 2026 02:15

kernel-patches-daemon-bpf Bot force-pushed the bpf-next_base branch from 43a1b48 to 912a12b Compare March 16, 2026 16:11

kernel-patches-daemon-bpf Bot force-pushed the series/1066584=>bpf-next branch from 3c91638 to d0ab5cf Compare March 16, 2026 16:16

kernel-patches-daemon-bpf Bot force-pushed the bpf-next_base branch from 912a12b to 71c9f15 Compare March 16, 2026 16:27

kernel-patches-daemon-bpf Bot force-pushed the series/1066584=>bpf-next branch from d0ab5cf to 24ac3d1 Compare March 16, 2026 16:35

kernel-patches-daemon-bpf Bot force-pushed the bpf-next_base branch from 71c9f15 to 45bc2d7 Compare March 16, 2026 18:33

kernel-patches-daemon-bpf Bot force-pushed the series/1066584=>bpf-next branch from 24ac3d1 to 55fe0a0 Compare March 16, 2026 18:38

kernel-patches-daemon-bpf Bot added V4-ci-pass and removed V4-ci-fail labels Mar 16, 2026

kernel-patches-daemon-bpf Bot force-pushed the bpf-next_base branch from 45bc2d7 to 1e267ee Compare March 16, 2026 21:16

mykyta5 added 2 commits March 16, 2026 14:21

mykyta5 added 3 commits March 16, 2026 14:21

kernel-patches-daemon-bpf Bot force-pushed the series/1066584=>bpf-next branch from 55fe0a0 to 31f8330 Compare March 16, 2026 21:21

kernel-patches-daemon-bpf Bot closed this Mar 16, 2026

kernel-patches-daemon-bpf Bot deleted the series/1066584=>bpf-next branch March 18, 2026 22:11

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

bpf: Add support for sleepable tracepoint programs#11398

bpf: Add support for sleepable tracepoint programs#11398
kernel-patches-daemon-bpf[bot] wants to merge 5 commits intobpf-next_basefrom
series/1066584=>bpf-next

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 14, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-review-bot Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 13, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 14, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

kernel-patches-daemon-bpf Bot commented Mar 16, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant