2024-02-26 17:09:54

by Oleg Nesterov

[permalink] [raw]
Subject: [PATCH 0/3] get_signal: minor cleanups and fix

Wen, as you can see I have changed my mind again ;)

Lets remove this clear_siginfo() right now. It is incomplete (and thus
looks confusing) and unnecessary.

See also 2/3, PF_USER_WORKER's already don't get a fully initialized ksig
anyway.

Do you see anything wrong?

Oleg.



2024-02-26 17:11:28

by Oleg Nesterov

[permalink] [raw]
Subject: [PATCH 3/3] get_signal: don't initialize ksig->info if SIGNAL_GROUP_EXIT/group_exec_task

This initialization is incomplete and unnecessary, neither do_group_exit()
nor PF_USER_WORKER need ksig->info.

Signed-off-by: Oleg Nesterov <[email protected]>
---
kernel/signal.c | 9 ++++++---
1 file changed, 6 insertions(+), 3 deletions(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index a69d3069067a..9c6a5ccac328 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2727,12 +2727,15 @@ bool get_signal(struct ksignal *ksig)
/* Has this task already been marked for death? */
if ((signal->flags & SIGNAL_GROUP_EXIT) ||
signal->group_exec_task) {
- clear_siginfo(&ksig->info);
- ksig->info.si_signo = signr = SIGKILL;
+ signr = SIGKILL;
sigdelset(&current->pending.signal, SIGKILL);
trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO,
- &sighand->action[SIGKILL - 1]);
+ &sighand->action[SIGKILL-1]);
recalc_sigpending();
+ /*
+ * implies do_group_exit() or return to PF_USER_WORKER,
+ * no need to initialize ksig->info/etc.
+ */
goto fatal;
}

--
2.25.1.362.g51ebf55


2024-02-26 17:13:56

by Oleg Nesterov

[permalink] [raw]
Subject: [PATCH 2/3] get_signal: hide_si_addr_tag_bits: fix the usage of uninitialized ksig

ksig->ka and ksig->info are not initialized if get_signal() returns 0 or
if the caller is PF_USER_WORKER.

Check signr != 0 before SA_EXPOSE_TAGBITS and move the "out" label down.

The latter means that ksig->sig won't be initialized if a PF_USER_WORKER
thread gets a fatal signal but this is fine, PF_USER_WORKER's don't use
ksig. And there is nothing new, in this case ksig->ka and ksig-info are
not initialized anyway. Add a comment.

Signed-off-by: Oleg Nesterov <[email protected]>
---
kernel/signal.c | 11 ++++++-----
1 file changed, 6 insertions(+), 5 deletions(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index 09a6dd07cf6b..a69d3069067a 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2881,8 +2881,9 @@ bool get_signal(struct ksignal *ksig)

/*
* PF_USER_WORKER threads will catch and exit on fatal signals
- * themselves. They have cleanup that must be performed, so
- * we cannot call do_exit() on their behalf.
+ * themselves. They have cleanup that must be performed, so we
+ * cannot call do_exit() on their behalf. Note that ksig won't
+ * be properly initialized, PF_USER_WORKER's shouldn't use it.
*/
if (current->flags & PF_USER_WORKER)
goto out;
@@ -2894,12 +2895,12 @@ bool get_signal(struct ksignal *ksig)
/* NOTREACHED */
}
spin_unlock_irq(&sighand->siglock);
-out:
+
ksig->sig = signr;

- if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
+ if (signr && !(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
hide_si_addr_tag_bits(ksig);
-
+out:
return signr > 0;
}

--
2.25.1.362.g51ebf55


2024-02-26 17:39:05

by Oleg Nesterov

[permalink] [raw]
Subject: [PATCH 1/3] get_signal: don't abuse ksig->info.si_signo and ksig->sig

Cleanup and preparation for the next changes.

get_signal() uses signr or ksig->info.si_signo or ksig->sig in a chaotic
way, this looks confusing. Change it to always use signr.

Signed-off-by: Oleg Nesterov <[email protected]>
---
kernel/signal.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)

diff --git a/kernel/signal.c b/kernel/signal.c
index c9c57d053ce4..09a6dd07cf6b 100644
--- a/kernel/signal.c
+++ b/kernel/signal.c
@@ -2842,7 +2842,7 @@ bool get_signal(struct ksignal *ksig)
spin_lock_irq(&sighand->siglock);
}

- if (likely(do_signal_stop(ksig->info.si_signo))) {
+ if (likely(do_signal_stop(signr))) {
/* It released the siglock. */
goto relock;
}
@@ -2866,7 +2866,7 @@ bool get_signal(struct ksignal *ksig)

if (sig_kernel_coredump(signr)) {
if (print_fatal_signals)
- print_fatal_signal(ksig->info.si_signo);
+ print_fatal_signal(signr);
proc_coredump_connector(current);
/*
* If it was able to dump core, this kills all
@@ -2890,7 +2890,7 @@ bool get_signal(struct ksignal *ksig)
/*
* Death signals, no core dump.
*/
- do_group_exit(ksig->info.si_signo);
+ do_group_exit(signr);
/* NOTREACHED */
}
spin_unlock_irq(&sighand->siglock);
@@ -2900,7 +2900,7 @@ bool get_signal(struct ksignal *ksig)
if (!(ksig->ka.sa.sa_flags & SA_EXPOSE_TAGBITS))
hide_si_addr_tag_bits(ksig);

- return ksig->sig > 0;
+ return signr > 0;
}

/**
--
2.25.1.362.g51ebf55