Re: [PATCH v11 08/12] signal, x86: add SIGSYS info and make it synchronous.
From: Will Drewry
Date: Tue Feb 28 2012 - 12:07:01 EST
On Tue, Feb 28, 2012 at 10:02 AM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
> On 02/27, Will Drewry wrote:
>>
>> On Mon, Feb 27, 2012 at 11:22 AM, Oleg Nesterov <oleg@xxxxxxxxxx> wrote:
>> > On 02/24, Will Drewry wrote:
>> >>
>> >> To ensure that SIGSYS delivery occurs on return from the triggering
>> >> system call, SIGSYS is added to the SYNCHRONOUS_MASK macro.
>> >
>> > Hmm. Can't understand... please help.
>> >
>> >> #define SYNCHRONOUS_MASK \
>> >> (sigmask(SIGSEGV) | sigmask(SIGBUS) | sigmask(SIGILL) | \
>> >> - sigmask(SIGTRAP) | sigmask(SIGFPE))
>> >> + sigmask(SIGTRAP) | sigmask(SIGFPE) | sigmask(SIGSYS))
>> >
>> > Why?
>> >
>> > SYNCHRONOUS_MASK just tells dequeue_signal() "pick them first".
>> > This is needed to make sure that the handler for, say SIGSEGV,
>> > can use ucontext->ip as a faulting addr.
>>
>> I think that Roland covered this. (Since the syscall_rollback was
>> called it's nice to let our handler get first go.)
>
> OK, except I do not really understand the "our handler get first go".
Err I meant "gets to go first".
> Suppose SIGSYS "races" with the pending SIGHUP. With this change
> the caller for SIGHUP will be called first. But yes, setup_frame()
> will be called for SIGSYS first. Hopefully this is what you want.
I believe it is. I just want ucontext_t to be properly populate since
the registers were just rolled back for it.
>> > But seccomp adds info->si_call_addr, this looks unneeded.
>>
>> True enough. I can drop it.
>
> Hmm. I meant, the change in SYNCHRONOUS_MASK looks unneeded. Please
> keep ->si_call_addr, it is much more convenient than ucontext_t in
> userspace.
Sorry for the confusion
>> It'd only be useful if the SIGSYS wasn't
>> being forced and the signal was being handled without ucontext_t
>> access.
>
> No, force_ doesn't make any difference in this sense...
I guess I was thinking about users of signalfd wanting the call site
but force_ avoids it being blockable so that seems largely irrelevant
at present.
> In short, the patch looks fine to me, but if you resend it may be
> you can update the changelog.
Thanks! I will try to clarify the changelog.
will
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/