[RFC][PATCH] Demultiplexing SIGTRAP signal

Ingo Molnar mingo at elte.hu
Mon Sep 22 20:42:02 EST 2008


* Srinivasa Ds <srinivasa at in.ibm.com> wrote:

> Currently a SIGTRAP signal can denote any one of below reasons.
> 	- Breakpoint hit
> 	- H/W debug register hit
> 	- Single step
> 	- SIGTRAP signal sent through kill() or rasie()
> 
> Architectures like powerpc/parisc provides infrastructure to 
> demultiplex SIGTRAP signal by passing down the information for 
> receiving SIGTRAP through si_code of siginfot_t structure. Here is an 
> attempt is generalise this infrastructure by extending it to x86 and 
> x86_64 archs.

no fundamental objections - assuming existing x86 apps have not grown an 
ABI dependency on the existing send_sigtrap() semantics. (Debuggers and 
JITs would be a candidate for such dependencies.)

a small implementational detail, this bit:

> @@ -935,8 +936,22 @@ void __kprobes do_debug(struct pt_regs *
>  			goto clear_TF_reenable;
>  	}
>  
> -	/* Ok, finally something we can handle */
> -	send_sigtrap(tsk, regs, error_code);
> +	tsk->thread.trap_no = 1;
> +	tsk->thread.error_code = error_code;
> +
> +	memset(&info, 0, sizeof(info));
> +	info.si_signo = SIGTRAP;
> +	if (condition & DR_STEP)
> +		info.si_code = TRAP_TRACE;
> +	else if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3))
> +		info.si_code = TRAP_HWBKPT;
> +	else
> +		info.si_code = TRAP_BRKPT;
> +	/* User-mode ip? */
> +	info.si_addr = user_mode_vm(regs) ? (void __user *) regs->ip : NULL;
> +
> +	/* Send us the fake SIGTRAP */
> +	force_sig_info(SIGTRAP, &info, tsk);

should be pushed into [a sufficiently extended] send_sigtrap() instead.

and this bit:

> -	info.si_code = TRAP_BRKPT;
> +	if (condition & DR_STEP)
> +		info.si_code = TRAP_TRACE;
> +	else if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3))
> +		info.si_code = TRAP_HWBKPT;
> +	else
> +		info.si_code = TRAP_BRKPT;

should be separated into a helper function as well i guess.

Roland, any objections to the core idea (or to the implementation)?

	Ingo



More information about the Linuxppc-dev mailing list