Received-SPF: pass (google.com: domain of linux-kernel+bounces-36196-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240123153313.GA21832@redhat.com> <20240123153357.GA21857@redhat.com>
In-Reply-To: <20240123153357.GA21857@redhat.com>
From: Dylan Hatch <dylanbhatch@google.com>
Date: Tue, 23 Jan 2024 15:50:27 -0800
Message-ID: <CADBMgpw4m-PZJrqW4YBbM_ssFB210wbDvxAS69pEWTDCGdKGnA@mail.gmail.com>
Subject: Re: [PATCH 2/3] fs/proc: do_task_stat: use sig->stats_lock to gather
 the threads/children stats
To: Oleg Nesterov <oleg@redhat.com>
Cc: Andrew Morton <akpm@linux-foundation.org>, "Eric W. Biederman" <ebiederm@xmission.com>, 
	linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Tue, Jan 23, 2024 at 7:35=E2=80=AFAM Oleg Nesterov <oleg@redhat.com> wro=
te:
>
> lock_task_sighand() can trigger a hard lockup. If NR_CPUS threads call
> do_task_stat() at the same time and the process has NR_THREADS, it will
> spin with irqs disabled O(NR_CPUS * NR_THREADS) time.
>
> Change do_task_stat() to use sig->stats_lock to gather the statistics
> outside of ->siglock protected section, in the likely case this code
> will run lockless.
>
> Signed-off-by: Oleg Nesterov <oleg@redhat.com>
> ---
>  fs/proc/array.c | 58 +++++++++++++++++++++++++++----------------------
>  1 file changed, 32 insertions(+), 26 deletions(-)
>
> diff --git a/fs/proc/array.c b/fs/proc/array.c
> index 45ba91863808..34a47fb0c57f 100644
> --- a/fs/proc/array.c
> +++ b/fs/proc/array.c
> @@ -477,13 +477,13 @@ static int do_task_stat(struct seq_file *m, struct =
pid_namespace *ns,
>         int permitted;
>         struct mm_struct *mm;
>         unsigned long long start_time;
> -       unsigned long cmin_flt =3D 0, cmaj_flt =3D 0;
> -       unsigned long  min_flt =3D 0,  maj_flt =3D 0;
> -       u64 cutime, cstime, utime, stime;
> -       u64 cgtime, gtime;
> +       unsigned long cmin_flt, cmaj_flt, min_flt, maj_flt;
> +       u64 cutime, cstime, cgtime, utime, stime, gtime;
>         unsigned long rsslim =3D 0;
>         unsigned long flags;
>         int exit_code =3D task->exit_code;
> +       struct signal_struct *sig =3D task->signal;
> +       unsigned int seq =3D 1;
>
>         state =3D *get_task_state(task);
>         vsize =3D eip =3D esp =3D 0;
> @@ -511,12 +511,8 @@ static int do_task_stat(struct seq_file *m, struct p=
id_namespace *ns,
>
>         sigemptyset(&sigign);
>         sigemptyset(&sigcatch);
> -       cutime =3D cstime =3D 0;
> -       cgtime =3D gtime =3D 0;
>
>         if (lock_task_sighand(task, &flags)) {
> -               struct signal_struct *sig =3D task->signal;
> -
>                 if (sig->tty) {
>                         struct pid *pgrp =3D tty_get_pgrp(sig->tty);
>                         tty_pgrp =3D pid_nr_ns(pgrp, ns);
> @@ -527,27 +523,9 @@ static int do_task_stat(struct seq_file *m, struct p=
id_namespace *ns,
>                 num_threads =3D get_nr_threads(task);
>                 collect_sigign_sigcatch(task, &sigign, &sigcatch);
>
> -               cmin_flt =3D sig->cmin_flt;
> -               cmaj_flt =3D sig->cmaj_flt;
> -               cutime =3D sig->cutime;
> -               cstime =3D sig->cstime;
> -               cgtime =3D sig->cgtime;
>                 rsslim =3D READ_ONCE(sig->rlim[RLIMIT_RSS].rlim_cur);
>
> -               /* add up live thread stats at the group level */
>                 if (whole) {
> -                       struct task_struct *t;
> -
> -                       __for_each_thread(sig, t) {
> -                               min_flt +=3D t->min_flt;
> -                               maj_flt +=3D t->maj_flt;
> -                               gtime +=3D task_gtime(t);
> -                       }
> -
> -                       min_flt +=3D sig->min_flt;
> -                       maj_flt +=3D sig->maj_flt;
> -                       gtime +=3D sig->gtime;
> -
>                         if (sig->flags & (SIGNAL_GROUP_EXIT | SIGNAL_STOP=
_STOPPED))
>                                 exit_code =3D sig->group_exit_code;
>                 }
> @@ -562,6 +540,34 @@ static int do_task_stat(struct seq_file *m, struct p=
id_namespace *ns,
>         if (permitted && (!whole || num_threads < 2))
>                 wchan =3D !task_is_running(task);
>
> +       do {
> +               seq++; /* 2 on the 1st/lockless path, otherwise odd */
> +               flags =3D read_seqbegin_or_lock_irqsave(&sig->stats_lock,=
 &seq);
> +
> +               cmin_flt =3D sig->cmin_flt;
> +               cmaj_flt =3D sig->cmaj_flt;
> +               cutime =3D sig->cutime;
> +               cstime =3D sig->cstime;
> +               cgtime =3D sig->cgtime;
> +
> +               if (whole) {
> +                       struct task_struct *t;
> +
> +                       min_flt =3D sig->min_flt;
> +                       maj_flt =3D sig->maj_flt;
> +                       gtime =3D sig->gtime;
> +
> +                       rcu_read_lock();
> +                       __for_each_thread(sig, t) {
> +                               min_flt +=3D t->min_flt;
> +                               maj_flt +=3D t->maj_flt;
> +                               gtime +=3D task_gtime(t);
> +                       }
> +                       rcu_read_unlock();
> +               }
> +       } while (need_seqretry(&sig->stats_lock, seq));
> +       done_seqretry_irqrestore(&sig->stats_lock, seq, flags);
> +
>         if (whole) {
>                 thread_group_cputime_adjusted(task, &utime, &stime);
>         } else {
> --
> 2.25.1.362.g51ebf55
>

Signed-off-by: Dylan Hatch <dylanbhatch@google.com>