kdb
[Top] [All Lists]

Re: [PATCH] kdb on SMP (breakpoint hits multiple cpus simultaneously)

To: "Zhang, Sonic" <sonic.zhang@xxxxxxxxx>
Subject: Re: [PATCH] kdb on SMP (breakpoint hits multiple cpus simultaneously)
From: "Vamsi Krishna S ." <vamsi@xxxxxxxxxx>
Date: Thu, 3 Apr 2003 14:16:07 +0530
Cc: Keith Owens <kaos@xxxxxxx>, Hua Qin <qinhua@xxxxxxxxxxxxxxxxxxxxx>, kdb@xxxxxxxxxxx
In-reply-to: <37FBBA5F3A361C41AB7CE44558C3448E488CEB@pdsmsx403.ccr.corp.intel.com>; from sonic.zhang@intel.com on Thu, Apr 03, 2003 at 03:20:01PM +0800
References: <37FBBA5F3A361C41AB7CE44558C3448E488CEB@pdsmsx403.ccr.corp.intel.com>
Reply-to: vamsi@xxxxxxxxxx
Sender: kdb-bounce@xxxxxxxxxxx
User-agent: Mutt/1.2.5i
Yes, that is what I thought. My earlier patch fixed only the int3
points. I will do similar fix for debugreg breakpoints too
and update the man pages and code comments accordingly.

Thanks,
Vamsi.


On Thu, Apr 03, 2003 at 03:20:01PM +0800, Zhang, Sonic wrote:
> Hi,
> 
>       Yes, the case you mentioned is possible in current KDB. I remember some 
> related instructions in the KDB man pages. It is said that don't remove any 
> break point when you enter KDB through DEBUG trap or Int3 trap, disable it 
> instead and remove it safely later when you enter KDB via keyboard trap or 
> serial port trap. 
> 
>       Maybe your patch can clean up these comments in the man pages?
> 
>       Regards.
> 
>       Sonic Zhang
> 
> -----Original Message-----
> From: Vamsi Krishna S . [mailto:vamsi@xxxxxxxxxx]
> Sent: 2003?4?3? 14:59
> To: Zhang, Sonic
> Cc: Keith Owens; Hua Qin; kdb@xxxxxxxxxxx
> Subject: Re: [PATCH] kdb on SMP (breakpoint hits multiple cpus simultaneously)
> 
> 
> Hi,
> 
> Thanks for the clarification. I thought from my quick
> glance through that currently all hard breakpoints
> are global, whereas your patches enabled per-cpu
> breakpoints. On second look, you are right, the current
> code allows only per-cpu debugreg-breakpoints, global
> ones are much more generally useful, which your patch
> enables.
> 
> I also agree with the name changes and the fact that
> per-cpu breakpoints may be useful for debugging certain
> paths (processes) bound to a specific cpu.
> 
> smphdr* patches are very much needed.
> 
> Thanks,
> Vamsi.
> 
> BTW: Can you think about what happens if two debugreg
> breakpoints are triggered at about the same time on
> different cpus? One of them wins the race to enter
> the kdb, and then the user clears all breakpoints..
> will kdb be able still recognise this as one of its 
> breakpoints? I think not, but can you give it a thought?
> 
> -- 
> Vamsi Krishna S.
> Linux Technology Center,
> IBM Software Lab, Bangalore.
> Ph: +91 80 5044959
> Internet: vamsi@xxxxxxxxxx

-- 
Vamsi Krishna S.
Linux Technology Center,
IBM Software Lab, Bangalore.
Ph: +91 80 5044959
Internet: vamsi@xxxxxxxxxx

<Prev in Thread] Current Thread [Next in Thread>