Re: [PATCH 1/1] kdb: fix stack overflow for large NR_CPUS count

Mike Travis (on Thu, 30 Oct 2008 18:51:58 -0700) wrote:
>The fix I implemented is to move the template to static memory and
>protect it with a spinlock.  The spinlock most likely is not required
>as it's highly unlikely that two bp commands can be entered into kdb
>at the exact same time, but this definitely prevents that.

It is actually impossible.  KDB stops all the cpus and only accepts
console input from one cpu at a time.  The entire KDB command system is
single threaded.  Ditch the spinlock, which removes the need for extra
code on the error paths.

Could you give the static template a kdb_ prefix please.  Even though
template is static, it still shows up in the global kernel namespace.

Apart from that, the patch looks good.

_______________________________________________
kdb mailing list
[email protected]
http://oss.sgi.com/mailman/listinfo/kdb