[uClibc]e2fsck problem: (Bad address) while doing inode scan

Erik Andersen andersen at codepoet.org
Mon Mar 31 19:23:32 UTC 2003


On Mon Mar 31, 2003 at 09:19:16PM +0200, Giulio Orsero wrote:
> On Mon, 31 Mar 2003 11:37:51 -0700, Erik Andersen <andersen at codepoet.org>
> wrote:
> 
> >> I might have messed up with uclibc. Maybe e2fsprogs misdetects uclibc
> >> headers and self-configure with wrong assumptions at compile time?
> >
> >This sounds fairly likely...  Tried doing an strace and comparing
> >calls to lseek?  When you _compile_ e2fsck are you also compiling 
> >it using 2.0 or 2.2 kernel header files?  Perhaps e2fsprogs is
> I compiled e2fsprogs with and without any headers installed, because I
> thinks it should always use automatically the headers inside uclibc dirs,
> correct?
> I compiled it on a 2.2 system too (which has never seen 2.4.x), just to be
> sure.
> 
> So far, I always tried wih uclibc compiled on a 2.4 system (Red Hat 7.3)
> with kernel source pointing to a 2.2.19 tree; I verified version.h inside
> uclibc installed/runtime/devel tree is 2.2.x.
> 
> I tried compiling uclibc too on a 2.2 system (Red Hat 6.x) but it fails due
> to ld not supporting -init.

Ok.  Well if uClibc is using 2.2 kernel headers, then that is
probably ok.

Next step...  Try running an strace if the libc5 binary and
compare the set of lseek calls being made to the strace you
already made of the uClibc binary.  I suspect that will tell
us a great deal, since the libc5 binary works,

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--



More information about the uClibc mailing list