Re: [Bug-cpio] Bug or problem in CPIO?

[Ondrej Vasik]

On Sun, 2014-03-23 at 17:59 +0100, Roy van der Veen wrote:
> Dear all,
> 
> 
> We face a very serious problem using CPIO.
> 
> 
> At a customer site we made back-up's using cpio, and gzip:
> 
> 
> find erp s3 -print|cpio -ocvB > /u/download/bu/erp.cpio 2>&1
> gzip erp.cpio
> 
> 
> 
> So far, so good.
> 
> 
> The server is based on Red Hat ES5.4. Version of cpio: cpio (GNU cpio)
> 2.6

When using enterprise Linux distributions, better way is to contact
directly product support of the vendor ( access.redhat.com )

> Last week the server died, and we build exactly the same Linux-server,
> in a virtual environment.

Do you mean "building from sources" or just new virtual machine using
the old kickstart?
Is the filesystem and partition layout used on the system the same? Are
the versions really the same? I know, it will be hard to find out, as
the old server died - but maybe some package downgrade (in the failing
stack, so glibc/cpio) could help .

> The restore is aborted (far too early) with the message:
> *** glibc detected *** cpio: free(): invalid next size (fast):
> 0x0933a298 ***
> ======= Backtrace: =========
> /lib/libc.so.6[0xa42585]
> /lib/libc.so.6(cfree+0x59)[0xa42959]
> cpio[0x804b55a]
> cpio[0x804ba2c]
> cpio[0x804bff3]
> cpio[0x8050cb1]
> /lib/libc.so.6(__libc_start_main+0xdc)[0x9eee9c]
> cpio[0x8049ac1]
> ======= Memory map: ========
> We tried several things to solve this:
> * Trying several levels of gunzip
> * trying several options of cpio (block-size, format, with/without -c,
> etc.)


> Can you give us a hint how to solve this? We are desperate, and our
> customer in panic.

First of all, I would suggest to install debuginfo packages for the cpio
and glibc. It would make the backtrace much more useful.

Greetings,
          Ondrej Vasik