| lists.openwall.net | lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <CAEf4BzbLJtMGaZoFAaAgnNXe8GCStsw+kZ_3hWoGfySWZ6B5mg@mail.gmail.com>
Date: Fri, 26 Sep 2025 11:52:53 -0700
From: Andrii Nakryiko <andrii.nakryiko@...il.com>
To: Tao Chen <chen.dylane@...ux.dev>
Cc: song@...nel.org, jolsa@...nel.org, ast@...nel.org, daniel@...earbox.net,
andrii@...nel.org, martin.lau@...ux.dev, eddyz87@...il.com,
yonghong.song@...ux.dev, john.fastabend@...il.com, kpsingh@...nel.org,
sdf@...ichev.me, haoluo@...gle.com, bpf@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH bpf-next v2] bpf: Add preempt_disable to protect get_perf_callchain
On Fri, Sep 26, 2025 at 8:40 AM Tao Chen <chen.dylane@...ux.dev> wrote:
>
> As Alexei noted, get_perf_callchain() return values may be reused
> if a task is preempted after the BPF program enters migrate disable
> mode. We therefore use bpf_perf_callchain_entries percpu entries
> similarly to bpf_try_get_buffers to preserve the current task's
> callchain and prevent overwriting by preempting tasks. And we also
> add preempt_disable to protect get_perf_callchain.
>
> Reported-by: Alexei Starovoitov <ast@...nel.org>
> Closes: https://lore.kernel.org/bpf/CAADnVQ+s8B7-fvR1TNO-bniSyKv57cH_ihRszmZV7pQDyV=VDQ@mail.gmail.com
> Signed-off-by: Tao Chen <chen.dylane@...ux.dev>
> ---
> kernel/bpf/stackmap.c | 76 ++++++++++++++++++++++++++++++++++---------
> 1 file changed, 61 insertions(+), 15 deletions(-)
>
> Change list:
> v1 -> v2:
> From Alexei
> - create percpu entris to preserve current task's callchain
> similarly to bpf_try_get_buffers.
> v1: https://lore.kernel.org/bpf/20250922075333.1452803-1-chen.dylane@linux.dev
>
> diff --git a/kernel/bpf/stackmap.c b/kernel/bpf/stackmap.c
> index 2e182a3ac4c..8788c219926 100644
> --- a/kernel/bpf/stackmap.c
> +++ b/kernel/bpf/stackmap.c
> @@ -31,6 +31,55 @@ struct bpf_stack_map {
> struct stack_map_bucket *buckets[] __counted_by(n_buckets);
> };
>
> +struct bpf_perf_callchain_entry {
> + u64 nr;
> + u64 ip[PERF_MAX_STACK_DEPTH];
> +};
> +
> +#define MAX_PERF_CALLCHAIN_PREEMPT 3
> +static DEFINE_PER_CPU(struct bpf_perf_callchain_entry[MAX_PERF_CALLCHAIN_PREEMPT],
> + bpf_perf_callchain_entries);
> +static DEFINE_PER_CPU(int, bpf_perf_callchain_preempt_cnt);
> +
> +static int bpf_get_perf_callchain(struct bpf_perf_callchain_entry **entry,
> + struct pt_regs *regs, u32 init_nr, bool kernel,
> + bool user, u32 max_stack, bool crosstack,
> + bool add_mark)
> +{
> + struct bpf_perf_callchain_entry *bpf_entry;
> + struct perf_callchain_entry *perf_entry;
> + int preempt_cnt;
> +
> + preempt_cnt = this_cpu_inc_return(bpf_perf_callchain_preempt_cnt);
> + if (WARN_ON_ONCE(preempt_cnt > MAX_PERF_CALLCHAIN_PREEMPT)) {
> + this_cpu_dec(bpf_perf_callchain_preempt_cnt);
> + return -EBUSY;
> + }
> +
> + bpf_entry = this_cpu_ptr(&bpf_perf_callchain_entries[preempt_cnt - 1]);
> +
> + preempt_disable();
> + perf_entry = get_perf_callchain(regs, init_nr, kernel, user, max_stack,
> + crosstack, add_mark);
> + if (unlikely(!perf_entry)) {
> + preempt_enable();
> + this_cpu_dec(bpf_perf_callchain_preempt_cnt);
> + return -EFAULT;
> + }
> + memcpy(bpf_entry, perf_entry, sizeof(u64) * (perf_entry->nr + 1));
N copies of a stack trace is not good enough, let's have N + 1 now :)
If we are going with our own buffers, we need to teach
get_perf_callchain to let us pass that buffer directly to avoid that
unnecessary copy.
Also, I know it's about 1KB, but it would be so simple and efficient
to just have this bpf_perf_callchain_entry on the stack. Kernel has a
16KB stack, right? It feels like for something like this using 1KB of
the stack to simplify and speed up stack trace capture is a good
enough reason.
> + *entry = bpf_entry;
> + preempt_enable();
> +
> + return 0;
> +}
> +
> +static void bpf_put_perf_callchain(void)
> +{
> + if (WARN_ON_ONCE(this_cpu_read(bpf_perf_callchain_preempt_cnt) == 0))
> + return;
> + this_cpu_dec(bpf_perf_callchain_preempt_cnt);
> +}
> +
> static inline bool stack_map_use_build_id(struct bpf_map *map)
> {
> return (map->map_flags & BPF_F_STACK_BUILD_ID);
> @@ -303,8 +352,9 @@ BPF_CALL_3(bpf_get_stackid, struct pt_regs *, regs, struct bpf_map *, map,
> u32 max_depth = map->value_size / stack_map_data_size(map);
> u32 skip = flags & BPF_F_SKIP_FIELD_MASK;
> bool user = flags & BPF_F_USER_STACK;
> - struct perf_callchain_entry *trace;
> + struct bpf_perf_callchain_entry *trace;
> bool kernel = !user;
> + int err;
>
> if (unlikely(flags & ~(BPF_F_SKIP_FIELD_MASK | BPF_F_USER_STACK |
> BPF_F_FAST_STACK_CMP | BPF_F_REUSE_STACKID)))
> @@ -314,14 +364,15 @@ BPF_CALL_3(bpf_get_stackid, struct pt_regs *, regs, struct bpf_map *, map,
> if (max_depth > sysctl_perf_event_max_stack)
> max_depth = sysctl_perf_event_max_stack;
>
> - trace = get_perf_callchain(regs, 0, kernel, user, max_depth,
> - false, false);
> + err = bpf_get_perf_callchain(&trace, regs, 0, kernel, user, max_depth,
> + false, false);
> + if (err)
> + return err;
>
> - if (unlikely(!trace))
> - /* couldn't fetch the stack trace */
> - return -EFAULT;
> + err = __bpf_get_stackid(map, (struct perf_callchain_entry *)trace, flags);
> + bpf_put_perf_callchain();
>
> - return __bpf_get_stackid(map, trace, flags);
> + return err;
> }
>
> const struct bpf_func_proto bpf_get_stackid_proto = {
> @@ -443,8 +494,7 @@ static long __bpf_get_stack(struct pt_regs *regs, struct task_struct *task,
> if (sysctl_perf_event_max_stack < max_depth)
> max_depth = sysctl_perf_event_max_stack;
>
> - if (may_fault)
> - rcu_read_lock(); /* need RCU for perf's callchain below */
> + preempt_disable();
>
> if (trace_in)
> trace = trace_in;
> @@ -455,8 +505,7 @@ static long __bpf_get_stack(struct pt_regs *regs, struct task_struct *task,
> crosstask, false);
>
> if (unlikely(!trace) || trace->nr < skip) {
> - if (may_fault)
> - rcu_read_unlock();
> + preempt_enable();
> goto err_fault;
> }
>
> @@ -474,10 +523,7 @@ static long __bpf_get_stack(struct pt_regs *regs, struct task_struct *task,
> } else {
> memcpy(buf, ips, copy_len);
> }
> -
> - /* trace/ips should not be dereferenced after this point */
> - if (may_fault)
> - rcu_read_unlock();
> + preempt_enable();
>
> if (user_build_id)
> stack_map_get_build_id_offset(buf, trace_nr, user, may_fault);
really it's just build_id resolution that can take a while, which is
why we are trying to avoid preemption around it. But for non-build_id
case, can we avoid extra copying?
> --
> 2.48.1
>
Powered by blists - more mailing lists