On Mon, Apr 22, 2019 at 11:25 PM Oleg Nesterov oleg@redhat.com wrote:
On 04/22, Zhenliang Wei wrote:
--- a/kernel/signal.c +++ b/kernel/signal.c @@ -2441,6 +2441,7 @@ bool get_signal(struct ksignal *ksig) if (signal_group_exit(signal)) { ksig->info.si_signo = signr = SIGKILL; sigdelset(¤t->pending.signal, SIGKILL);
recalc_sigpending(); goto fatal; }trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, SIG_DFL);
Reviewed-by: Oleg Nesterov oleg@redhat.com
Last time Oleg suggested using SIG_DFL as the third parameter, but its type is 'void (*)(int)', but not expected 'struct k_sigaction *'.
How about trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, &sighand->action[signr - 1]); ?
Although we know that action[SIGKILL] must be SIG_DFL, in order to adapt to the parameter type, I suggest to use the original variable value.
Zhenliang Wei.
On Tue, Apr 23, 2019 at 01:10:52PM +0000, weizhenliang wrote:
On Mon, Apr 22, 2019 at 11:25 PM Oleg Nesterov oleg@redhat.com wrote:
On 04/22, Zhenliang Wei wrote:
--- a/kernel/signal.c +++ b/kernel/signal.c @@ -2441,6 +2441,7 @@ bool get_signal(struct ksignal *ksig) if (signal_group_exit(signal)) { ksig->info.si_signo = signr = SIGKILL; sigdelset(¤t->pending.signal, SIGKILL);
recalc_sigpending(); goto fatal; }trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, SIG_DFL);
Reviewed-by: Oleg Nesterov oleg@redhat.com
Last time Oleg suggested using SIG_DFL as the third parameter, but its type is 'void (*)(int)', but not expected 'struct k_sigaction *'.
Sigh, I should've caught that in the first commit. Although it suggests you didn't even compile your patch...
How about trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, &sighand->action[signr - 1]); ?
That should work, yes.
Christian
On 04/23, weizhenliang wrote:
On Mon, Apr 22, 2019 at 11:25 PM Oleg Nesterov oleg@redhat.com wrote:
On 04/22, Zhenliang Wei wrote:
--- a/kernel/signal.c +++ b/kernel/signal.c @@ -2441,6 +2441,7 @@ bool get_signal(struct ksignal *ksig) if (signal_group_exit(signal)) { ksig->info.si_signo = signr = SIGKILL; sigdelset(¤t->pending.signal, SIGKILL);
recalc_sigpending(); goto fatal; }trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, SIG_DFL);
Reviewed-by: Oleg Nesterov oleg@redhat.com
Last time Oleg suggested using SIG_DFL as the third parameter, but its type is 'void (*)(int)', but not expected 'struct k_sigaction *'.
Yes I misread the signature of TRACE_EVENT(signal_deliver), and I thought you at least compiled the kernel with your patch applied ;)
How about trace_signal_deliver(SIGKILL, SEND_SIG_NOINFO, &sighand->action[signr - 1]); ?
sure, this should fix the problem.
Oleg.
linux-stable-mirror@lists.linaro.org