Hi Jay, Thanks for the quick reply. I read through the last year or so of mailing list archives (stopped when I got to the notification of the 2.6.19 kernel release). Was helpful to catch up on the happenings and see what else had gone on. Didn't see any mentions of anyone having trouble like this (or much trouble at all) with serial ports. Let me see if I can describe the setup and perhaps someone might have a pointer- even if it's just something to try to eliminate a variable.
I am running Ubuntu 7.04 (kernel 2.6.20 + ubuntu patches + latest 2.6.20 KDB patches) on i386 (dual core CPU). I added the line console=ttyS0,9600 to my kernel boot parameters and have the file /etc/event.d/ttyS0 set to spawn a getty process on ttyS0. I get the weird freeze behavior even when I invoke KDB using the ESC+KDB sequence, so it doesn't seem to be just a load-from-panic behavior (I have also loaded KDB from a panic and it exhibits this behavior). Unfortunately my system doesn't have a PS/2 keyboard port, so I'm left with serial and USB (it's an uhci/ehci controller- so sounds like this will not work). I'm really at a loss as to what could be happening here- when KDB is running, that's the only code executing in the CPU, right? So it's not like it's getting swapped out while another process runs, right? The only think I could think is that perhaps there is some weird buffering or polling going on in the serial driver/protocol/terminal setup. My connection on the other end is a PuTTY terminal running on a Windows XP machine with a known-to-be-working USB serial port. PuTTY is in its default settings, though I have also tried kicking the speed on both sides of the connection up to 115200 to see if things improved- it was faster, but I still encountered the intermittent freezes. Tomorrow I'll try a different terminal emulator like TeraTerm, and a different serial port (the native COM port instead of USB) to see if anything changes. If anyone has suggestions, please let me know. Thanks! If I can get this to work properly, it might be nice to see if I can backport the kdump patch into the 2.6.20 version of kdb, since I use kdump on this system, and right now I have to disable kdump before crashes will drop me into kdb. Am I correct in my understand that post-patch, a crash will cause the system to drop into kdb, then runing the "kdump" command will cause the kdump/kexec kernel to be execed? If so that sounds like what I'd want. -Kevin On Thu, Sep 25, 2008 at 8:36 PM, Jay Lan <[EMAIL PROTECTED]> wrote: > Kevin Worth wrote: >> Hi All, > > Hi Kevin, > >> >> I set up a system with KDB and everything runs great, except that when >> I am in KDB on the serial console, it seems to hang intermittently >> (for several minutes) and then continue working again for a few >> seconds or so. I am grateful to have some immediate post-mortem >> debugging capability now on this system (since sometimes kdump fails >> due to memory corruption), but the process goes very slowly when the >> system hangs every couple times you go to type a command. >> >> I know this kernel is a bit old- is this an issue that has been fixed >> on newer versions but was not backported to the 2.6.20 version? > > Which arch were you running on and how did you set up your serial > console? I have not seen the intermittent stall problem you > described before. > >> Any >> ideas of where I can start? I've now tried two different serial ports >> and both are running at 9600 8-n-1, with flow control off. The problem >> seems to occur most often if I leave the console sitting for a few >> seconds, though I've also had it when I'm quickly interacting and >> typing commands. > > I let people more better on serial port setting to respond here. > >> >> Additionally, is the sgi ftp at >> ftp://oss.sgi.com/projects/kdb/download/v4.4/ the only place to find >> files for kdb? > > Yes. > >> It would be nice to have some of the patches that were >> added to the KDB patches for later kernels (i.e. ehci/uhci USB, >> kdump), but it doesn't seem particularly easy to tell which >> modifications are for which feature... > > That is true. You can look at the ChangLog at kdb/ and at > arch/{ia64,x86}/kdb/ directories. Major bug fixes/features are > described in the ChangeLog's. Unzipped patches will have complete > version of ChangeLog's. BTW, uhci USB feature has been backed out > since it caused regression in ohci. > >> I guess perhaps I could search >> the mailing list archives for the original patch submissions... any >> easier way? Does KDB exist in any online source control systems? The >> SGI CVS server no longer appears to contain it... > > Unfortunately, no. > >> are new versions >> just posted straight to the FTP site? > > Yes, it has been that way. We use internal RCS tree to maintain > the patches and newer versions are then placed at the FTP site. > > Regards, > jay > >> >> -Kevin >> --------------------------- >> Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe. > > --------------------------- Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe.
