Hello
I have a program that generates a lot of signals (perf_event_open
overflows). The signals are set up to be RT signals with sigaction().
The program sometimes gets enough signals that the RT signal queues fill
up. In that case the kernel responds by ignoring the signal settings
and unconditionally sends a SIGIO.
The code that does this is (I think) in fs/fcntl.h, line 472
if (!do_send_sig_info(signum, &si, p, group))
break;
/* fall-through: fall back on the old plain SIGIO signal */
case 0:
do_send_sig_info(SIGIO, SEND_SIG_PRIV, p, group);
The probelm is this SIGIO is sent ignoring the
sa_flags = SA_SIGINFO;
setting. So the si_fd value isn't set, making it hard to tell which
fd has overflowed (which would be useful so I could shut down the
offending fd to stop the signal flood).
Is this the expected behavior? It dates back to pre-git.
A related issue:
If I try to read the si->si_fd value anyway, I get garbage.
It seems that the code that implementes SEND_SIG_PRIV
only sets the uid and pid fields, and possibly is leaking kernel
state through the rest of the siginfo_t structure.
as best I can tell this happens in kernel/signal.c line 1106:
case (unsigned long) SEND_SIG_PRIV:
q->info.si_signo = sig;
q->info.si_errno = 0;
q->info.si_code = SI_KERNEL;
q->info.si_pid = 0;
q->info.si_uid = 0;
break;
values past the pid and uid fields aren't zerod when read from the
siginfo_t field passed to the signal handler.
static void sigio_handler(int signum, siginfo_t *info, void *uc);
Thanks,
Vince
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
