Jan, I have produced this version of kallsyms which will run on x86, and cross compile to arm/xscale (BE) (tested), maybe arm/xscale (LE) (untested), maybe ppc (BE) (untested), and x86_64 (tested). Its not full blown cross (ie: you can't compile on BE 32 or LE 64 to any other architecture, just LE 32 to BE 32, maybe LE 32, LE 64.
Let me know if it works for you. Keep in mind this is just a hack, but best of luck. Keith, also would you consider adding this to the website? Regards -steve On Tue, 2004-11-23 at 09:25, Jan Rovins wrote: > Hello KDB List, > > Has anyone done up a cross compile version > of kallsyms x68 LE host to Xscale BE target. > > I saw some talk of it in the List archives, and was wondering > if anybody has succeeded with something like this yet? > > We are using kernel version 2.4.20 > > Jan > ----- Original Message ----- > From: "marks" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]>; "Paul Wilcox-Baker" <[EMAIL PROTECTED]> > Cc: "Matt Gierlach" <[EMAIL PROTECTED]> > Sent: Monday, November 22, 2004 6:16 PM > Subject: Re: Linux adventures > > > > Hi, > > > > If we could get our hands on a kallsyms program (which is a symlink to > > insmod by the way) which could run in the x86 environment but operate on > > and produce arm compatible files we could perform a build without the need > > to utilize the ARM system. Currently, when kallsyms runs on x86 it chokes > > on the arm object files. We then boot the ARM system and run "make > > kallsyms" which uses the arm version of kallsyms on the arm objects (our > > kernel source is in the NFS'd file system so is accessible from both the > > x86 and arm environments at the same time). We then switch back to the > > x86 environment, edit the Makefile to comment out the section that calls > > kallsyms, and finish the kernel build. Important: the edit to the > > Makefile must preserve its date/time stamp otherwise dependency checking > > forces the entire kernel build to start from scratch which is not what we > > want at this point. > > > > Note: the kallsyms processing in this particular kernel build was added to > > the kernel when I applied the kdb patch to the kernel source. This is not > > KGDB (which requires a 2nd computer to use) but instead is kdb built > > directly into the kernel image. The kallsyms step collects up all the > > kernel symbols for use by the embedded kdb. > > > > The last time I tried to compile this kernel witout the CONFIG_DEVFS_FS > > kernel configuration setting it failed. I assumed it was requried. > > Sounds like this is not the case for the kernel source Paul is using. > > > > On Mon, 22 Nov 2004, Paul Wilcox-Baker wrote: > > > > > Date: Mon, 22 Nov 2004 14:46 PST > > > From: Paul Wilcox-Baker <[EMAIL PROTECTED]> > > > To: janr > > > Cc: mjg, marks > > > Subject: Re: Linux adventures > > > > > > Dear Jan, > > > > > > > I will Try a build with the Montavista kernel that you sent me. > > > > and see how far I can get, the last time I tried it, it wouldn't build > > > > on my machine, and I need to find out why. > > > > I have the older Montavista 3.0 tools on that machine, but the kernel > > > > still should build. > > > > > > We do have one problem here with creating > > > the bootable images. There is a phase > > > where some sort symbol table mangling goes > > > on. This is performed by "make kallsyms" > > > For some reason, the x86 tools try and use a > > > version of kallsyms that expects x86 objects, > > > not ARM objects. The solution here is > > > to run one of the NFS kernels on the ADI ARM > > > board. This performs the "kallsyms" operation > > > successfully. > > > > > > Presumably, somewhere there is a "kallsyms" > > > the runs on the x86 and takes ARM objects, > > > bit neither Mark nor I could find it. > > > > > > After doing the make "kallsyms" on the > > > ARM, the "kallsyms" part of the > > > makefile has to be commented out of the > > > makefile, because, once again, the dependencies > > > aren't right and it tries to run the x86 "kallsyms" > > > again. After the edit, "make bootpImage ...." can > > > be run again to produce the complete thing. > > > Thanks Paul. > > > > > > > -- > > Thanks, > > > > Mark Sincerbox > > Adax, Inc. > > 1-510-548-7047 x129 > --------------------------- > Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe. -- Binary/unsupported file stripped by Ecartis -- -- Type: text/x-c -- File: kallsyms.c --------------------------- Use http://oss.sgi.com/ecartis to modify your settings or to unsubscribe.
