kdb
[Top] [All Lists]

Re: kdb problems with SMP systems

To: jim.houston@xxxxxxxxx
Subject: Re: kdb problems with SMP systems
From: Ethan Solomita <ethan@xxxxxxxxxxxxxxx>
Date: Tue, 30 Apr 2002 11:44:02 -0700
Cc: kdb@xxxxxxxxxxx
References: <3CCEC226.54575A34@xxxxxxxxx>
Sender: owner-kdb@xxxxxxxxxxx

Jim Houston wrote:
> 
> I have been seeing the problems Ethan described on
> SMP i386.  I was impressed by his writing and clear
> analysis of the problems.
> 
        OK, fine: you're flattery has convinced me to just finish the fixes and
make them public. They're about 80-85% done, at least on my sparc64
system.

> I'm very interested in seeing this fixed and would
> like to work with you.  I have some time to work on
> this, and I'm up to speed on the code.  I have  done
> similar work on debuggers for a couple of proprietary
> UNIX systems.  If there is a fix in the works, I would
> like see what you're doing.
> 
        Once I put out the patch, you can figure out why it doesn't work on
i386. 8-)

> I also noticed that single stepping saves and
> restores the interrupt enable flag.  If you single
> step over an instruction which changes this flag, you
> lose its effect.
> 
        On sparc64 this isn't a big issue (%pil vs. pstate.IE), so I hadn't
looked into it.
        -- Ethan

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