On Wed, Jun 14, 2006 at 04:39:21PM -0700, Piet Delaney wrote: > On Wed, 2006-06-14 at 16:04 -0700, Tom Rini wrote: > > On Tue, Jun 13, 2006 at 03:41:52PM -0700, Piet Delaney wrote: > > [snip] > > > I don't see any good reason why we shouldn't provide debug support in > > > the patch to facilitate installing the kgdb stub. It's been part of kgdb > > > stubs as far back as the old SunOS 4.1.4 patch I used back in 1995. > > > > The goal of this project is something that's mergable with kernel.org > > and something like that is far to ugly to live. > > Then all of our stuff should be #ifdef KGDB; while merging from 2.6.12 > to 2.6.13 I recall it noticing a few of the #ifdef KGDB's had been > removed.
Yes, there really shouldn't be any #ifdef KGDB's in the kernel tree and we're down to very few[1] and I think it's really just one in the end we'll need. > I don't think a few keep printk's, like George had would be > unacceptable; Andrew had no problems with it in the mm series. I don't know if that's true as Andrew dropped George's version and had no desire to try and change any code in there for reasons like that. > Something like the SOCK_DEBUG() macro might be reasonable; KGDB_DEBUG(); > where the function name is prefixed to all printk output. Macro is > about 6 lines of code and it's cosmetic impact in very small. What use is it? > The kgdb stub is special; you can't use kgdb on it to debug it. No, you have to debug is the old fashioned way. But thankfully that's fairly easy to do. And the problems we have are more fine grained, or fat-grained, where this isn't helpful either. > BTW, there isn't any incompatibility between the KGDB and KDUMP > patches is there? Don't know. In the end, there shouldn't be as both would use the same hooks. No comment about trying to use KGDB and KDUMP, or any other combination of multiple debugging mechanisms. > > The other problem is I don't see the use for it, and I've had a lot of > > problems I've had to fix. Since you can see what kgdb sent in GDB, you > > don't need to try and send it along again. If the two can communicate, > > this is redundant, if the two cannot communicate, it's not helpful > > either. > > I problems like Steve's, where the stub isn't responding it's useful. No it's not. You know KGDB is sending packets, you know GDB isn't receiving them. Look in the middle :) > I used it a few times; my bet is that it is useful if used judicially > like George's patch. I've done similar things earlier on too. But this type of problem is more likely a network setup issue or a "please configure/connect to kgdb once the system is up" oddity. [1]: OK, I'm ignoring the serial hell some platforms have. -- Tom Rini _______________________________________________ Kgdb-bugreport mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/kgdb-bugreport
