On Mon, Oct 3, 2011 at 7:27 AM, Michael Kerrisk <mtk.manpages@xxxxxxxxx> wrote: > Thanks for the detailed responses. Some comments to your remarks > below, and a couple of open questions (marked "????"). If you send me > the answers, then I can get another draft for review. > >>>> - SIGSTOP _can_ be injected. >>> >>> Was this true at one time? If yes, then we should document past and >>> current behavior, and note when the change occurred. >>> >>> In the Linux 2.4 sources, I see the following in >>> arch/i386/kernel/signal.c::do_signal(): >>> >>> /* The debugger continued. Ignore SIGSTOP. */ >>> if (signr == SIGSTOP) >>> continue; >>> >>> Did that code prevent SIGSTOP being injected in the 2.4 series? >> >> Looks like it is indeed the code. > > ???? > Sorry -- I'm not quite clear there. You're confirming that SIGSTOP > could not be injected in 2.4, right? Yes. In 2.4, SIGSTOP can't be injected. >> No need to do PTRACE_GETSIGINFO. >> Remember, requiring PTRACE_GETSIGINFO on every ptrace stop >> is a performance hit. > > Thanks. So I'll change that sentence (and the others): > > A subsequent PTRACE_GETSIGINFO on the stopped tracee will return a > siginfo_t structure with si_code set to SIGTRAP|PTRACE_EVENT_FORK<<8. > > to: > > A waitpid() by the tracer will return SIGTRAP|PTRACE_EVENT_FORK<<8 as > the status of the tracee. Word "status" above is ambiguous. Is it waitpid status? Is it si_code field in PTRACE_GETSIGINFO result? We probably need to be ridiculously verbose here to avoid confusion: "A waitpid() by the tracer will return status value which will have SIGTRAP | (PTRACE_EVENT_FORK << 8) in its most significant 24 bits. IOW: (status >> 8) will be equal to SIGTRAP | (PTRACE_EVENT_FORK << 8)." >> As of kernel 2.6.38, >> after the tracer sees the tracee ptrace-stop and until it >> restarts or kills it, the tracee will not run, >> and will not send notifications (except >> .B SIGKILL >> death) to the tracer, even if the tracer enters into another >> .BR waitpid (2) >> call. >> .LP >> .\" >> .\" FIXME ??? referrent of "it" in the next line is unclear >> .\" What does "it" refer to? >> Currently, it causes a problem with transparent handling of stopping >> signals: if the tracer restarts the tracee after group-stop, >> .B SIGSTOP >> is effectively ignored: the tracee doesn't remain stopped, it runs. >> If the tracer doesn't restart the tracee before entering into the next >> .BR waitpid (2), >> future >> .B SIGCONT >> signals will not be reported to the tracer. >> This would cause >> .B SIGCONT >> to have no effect. >> >> "it" refers to ptrace behavior versus group-stops and SIGCONT, >> as described. Feel free to rephrase. > > ???? > Help! I'm still having problems here. The problem may possibly be > this: when one uses a pronoun like "it" in English, it's normally a > back reference to some text already given. Is this "it" a back > reference (In that case, could you please send me a rewritten version > of the sentence that replaces "it" by some descriptive text), or is it > a reference to the current paragraph (in other words, should this > paragraph rather start with the words "Currently, here is a problem > with...")? I think replacing "it" with "this kernel behavior" will do: "Currently, this kernel behavior causes a problem with transparent handling of stopping signals: if the tracer restarts the tracee after group-stop, the stopping signal is effectively ignored: the tracee doesn't remain stopped, it runs. ..." (^^^^^^ also, replaced SIGSTOP with "the stopping signal" - all stopping signals are equally affected). >> No, it is not ok. Please consult sigaction(2) manpage and >> /usr/include/bits/siginfo.h >> For example, si_code == SI_TIMER (-2) can be sent by timer >> expiration, which is not a system call. There are many other signal >> sources which are not systcalls. > > Okay. So how about the following: > > was delivered as a result of a userspace action, > for example, a direct system call > .RB ( tgkill (2), > .BR kill (2), > .BR sigqueue (3), > etc.), > expiration of a POSIX timer, > change of state on a POSIX message queue, > or completion of an asynchronous I/O request. Yes, this looks ok. -- vda -- To unsubscribe from this list: send the line "unsubscribe linux-man" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
