Date: Sun, 8 Sep 2002 20:15:33 +1000 (EST) From: Bruce Evans <bde@zeta.org.au> To: Garrett Wollman <wollman@lcs.mit.edu> Cc: jmallett@FreeBSD.ORG, <arch@FreeBSD.ORG> Subject: Re: CFR: signalinfo(3) Message-ID: <20020908195306.N2533-100000@gamplex.bde.org> In-Reply-To: <200209071711.g87HBLxO048550@khavrinen.lcs.mit.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 7 Sep 2002, Garrett Wollman wrote:
> In article <20020907022003.A66983@FreeBSD.org> you write:
>
> >See, signal(3) installs handlers for the sa_handler variety of signal
> >handlers, and yet this is not the only type of handler one might want
> >to have, one might also want the "traditional" sa_sigaction handler
> >style, which includes siginfo_t, and struct sigcontext, both of which
> >are very useful, depending on the signal.
>
> Actually, those are new-style (POSIX Real-Time Signals) handlers. The
> traditional (Standard C) style takes a single argument. The
> really-old-fashioned (4.2BSD, pre-C89) handlers provided three
> arguments but they weren't the same as the three arguments used in
> RTS handlers and in any case we don't support that now.
We still support 2 of the args of 4.2BSD handlers in a binary compatible
way, at least on i386's. All signal handlers get passed extra args no
matter how they are declared.
4.2BSD args: int signum, int sigcode, struct sigcontext *scp
FreeBSD-old args: same
FreeBSD-not-so-old args: int signum, int sigcode, struct sigcontext *scp,
vm_offset_t faultaddr
FreeBSD args: int signum, siginfo_t *sip, struct sigcontext *scp,
vm_offset_t faultaddr
In -current, SA_SIGINFO just causes sip to be non-NULL. In my test, sip
pointed to not very interesting and apparently broken info:
% Program received signal SIGUSR1, User defined signal 1.
% 0x280a5963 in kill () from /usr/lib/libc.so.5
% (gdb) c
% Continuing.
%
% Breakpoint 1, sig (code=30, sip=0xbfbff84c, scp=0xbfbff708, addr=0) at z.c:7
% 7 }
% (gdb) where
% #0 sig (code=30, sip=0xbfbff84c, scp=0xbfbff708, addr=0) at z.c:7
% #1 <signal handler called>
% #2 0x280a5963 in kill () from /usr/lib/libc.so.5
% #3 0x080484e1 in _start ()
% (gdb) p *sip
% $1 = {si_signo = 30, si_errno = 0, si_code = 0, si_pid = 0, si_uid = 0,
% si_status = 0, si_addr = 0x2, si_value = {sigval_int = 0, sigval_ptr = 0x0},
% si_band = 0, __spare__ = {0, 0, 0, 0, 0, 0, 0}}
I expected at least si_pid and si_uid to be nonzero here.
> >So I wrote signalinfo(3)
> >which installs *those* kinds of signal handlers, no fuss.
>
> I'm confused. Why would you want to use this? There is already a
> perfectly good interface for installing Real-Time Signal handlers,
> and it's specified in the standard: sigaction(3). I don't think
> that this convenience function belongs in the C library.
I agree. There are approximately
2^(<number of SA_ flags> + <number of signal bits>) convenience functions
that can be built from sigaction(). Which ones are actually convenient
is application-specific.
Bruce
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20020908195306.N2533-100000>