Re: [dm-crypt] Data Corruption using cryptsetup luks on ARM arch

[Arno Wagner <arno@wagner.name>]

On Tue, Sep 15, 2009 at 07:48:55AM -0300, Salatiel Filho wrote:
> On Mon, Sep 14, 2009 at 18:10, Arno Wagner <arno@wagner.name> wrote:
> > On Mon, Sep 14, 2009 at 06:25:10AM -0300, Salatiel Filho wrote:
> >> On Mon, Sep 14, 2009 at 00:35, Arno Wagner <arno@wagner.name> wrote:
> >> > On Sun, Sep 13, 2009 at 07:19:00PM -0300, Salatiel Filho wrote:
> >> >> Where should i report that i get lots of corruption when using
> >> >> cryptsetup on arm architecture ?
> >> >>
> >> >> - Can not use crypted swap , cause page allocator will fail and my
> >> >> programs will start to seg fault.
> >> >> - Some files just get corrupted on copying...
> >> >>
> >> >> Anyone can help me ?debug it , so i can report ?developers and they
> >> >> can fix this.
> >> >
> >> > If it generally works, but sometimes corrupts, then this
> >> > very likely has nothing to to with dm-crypt. In fact it
> >> > sounds like you have defective hardware, but more detail
> >> > is needed for even a rough guess.
> >> >
> >> I do not think ?it is a hardware problem since the disk is ?fine and
> >> it works fine in other architectures with no corruption at all. And
> >> the page allocator failure always happens when using crypted swap but
> >> there is no problem when using uncrypted swap.
> >
> > Determine the exact nature of the corruption. It is really
> > a waste of time to speculate before you have that. Best
> > option is a binary compare of a corrupted file with the
> > uncorrupted version. Use something like
> >
> > cmp -b -l <file1> <file2>
> I tried two corrupted files ...
> Attached the cmp -b -l 'corrupted' 'ok' :
> 
> -rw-r--r-- 1 root root 50000000 2009-09-15 07:04 file1.ori
> -rw-r--r-- 1 root root 50000000 2009-09-15 06:21 file1.corrupted
> 
> -rw-r--r-- 1 root root 50000000 2009-09-15 07:04 file2.ori
> -rw-r--r-- 1 root root 50000000 2009-09-15 06:22 file2.corrupted

Ok, at a guess (assuming quite a bit of zeros in the non-corrupted
files), this looks like full sectors are written with zeros instead 
of the data they should hold.  This leaves out simple memory or
other hardware corruption. I would be inclined to assume this
is a software problem now.

Arno
-- 
Arno Wagner, Dr. sc. techn., Dipl. Inform., CISSP -- Email: arno@wagner.name 
GnuPG:  ID: 1E25338F  FP: 0C30 5782 9D93 F785 E79C  0296 797F 6B50 1E25 338F
----
Cuddly UI's are the manifestation of wishful thinking. -- Dylan Evans

If it's in the news, don't worry about it.  The very definition of 
"news" is "something that hardly ever happens." -- Bruce Schneier