> From: Todd Blevins <tblevins@cisco.com> > Date: Fri, 12 Apr 2002 12:53:02 -0400 (EDT) > Well, here's a trace if it means anything to you...doing a load psw...if i > understand the opcode, we load the doubleword from the address of the 2nd > operand in the operation...it contains all 0s: > > I 00C > Tracing active at IPL OK, this is what I would do. First get that kernel.img into a working Linux, then do this: $ strings kernel.img | grep 'Linux version' Linux version 2.4.9-27-t1 (zaitcev@test1.stuttgart.redhat.com) (gcc version 2.95.3 20010315 (release)) #1 SMP Mon Feb 11 22:09:45 CET 2002 Send the version string to Florian and ask him to find you the corresponding System.map. I hope it lingers somewhere in the build area... In System.map find the log_buf value, say 2b1cc0. IPL from the reader and when the thing crashes, say d rux2b1cc0.200 512 bytes if often not enough, add more until you reach the end of the trace. This assumes that your VM is fresh enough to support EBCDIC to ASCII translation in the display. By the way, since the trace always starts with "Linux", you may try to skip getting System.map by asking the trace facility to search core for the pattern "Linux" in ASCII. Here's the hex values: [zaitcev@s390 zaitcev]$ echo 'Linux version'| od -x 0000000 4c69 6e75 7820 7665 7273 696f 6e0a 0000016 [zaitcev@s390 zaitcev]$ -- Pete
