Keith, I have to go back to "bp" and "bpa". You said "bp" and "bpa" are same. Meaning the current "bp" is a global breakpoint for all cpus? I did some tests, set a break point on cpu 4 by using "bp", then "go". later, when the break point were hit, it was on cpu 3. Is that correct way for "bp"? Thanks! Hua On Sun, 23 Mar 2003, Keith Owens wrote: > On Sat, 22 Mar 2003 09:25:43 -0500 (EST), > Hua Qin <qinhua at networks.ecse.rpi.edu> wrote: > >So "bpa" will have the same problem? > > bp and bpa are the same for software breakpoints. > > >How about "bph" and "bpha", since they are > >using Pentium debug registers other that int3 mechanism, they don't have > >these problem. > > bpha installs the hardware breakpoint on all cpus, bph only installs on > one cpu but is not much use when you do not know which cpu you will be > executing on. > > The current kdb code for handling two cpus dropping into kdb at the > same time has races. Sonic Zhang mailed some test patches recently, > see ftp://oss.sgi.com//projects/kdb/download/v4.0/kdb-smphdr*. I have > not had time to try them yet. > > >