cramfs root filesystem corruption

cramfs root filesystem corruption

[Paul Ruhland]

I've found a few posts similar to this from about a year back in the main
kernel list but I could never follow the thread to any conclusion.

I'm running a cramfs root filesystem on an 850 system, linux-2.4.4 ( from denx
).  Everything has been running fine but the cramfs filesystem has started
behaving badly.

The cramfs image is in flash, mounted as root via mtd blockdevice.

Typically the error occurs during flash reprogramming ( using mtd char device
'/dev/mtd0' ) ...during/after flash programming the cramfs filesystem 'loses'
information ( files disappear ).  The errors from syslog are 'error during
decompression', typically the error numbers ( from 'linux/zlib_fs.h' ) are
Z_BUF_ERROR, for the first error,  and Z_DATA_ERROR for the rest.  I've got
my BDI breaking at the uncompress error and the source length is huge:

(gdb) frame 0
#0  cramfs_uncompress_block (dst=0xc078c000, dstlen=4096, src=0xc0145ac0,
srclen=4969176) at uncompress.c:49

The reprogramming of the flash is successful, all new data has been written to
the correct location(s).  The cramfs image ( root filesystem ) in flash is
NOT corrupt.  A hard reboot and everything is fine.

Has anyone seen such cramfs problem before?  I'm just looking to see if there
is any 'known' problems before I get to far into debugging this thing.

--
Paul Ruhland

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: cramfs root filesystem corruption

[Wolfgang Denk]

In message <200205301450.50112.pruhland@microwavedata.com> you wrote:
>
> The cramfs image is in flash, mounted as root via mtd blockdevice.
>
> Typically the error occurs during flash reprogramming ( using mtd char device
> '/dev/mtd0' ) ...during/after flash programming the cramfs filesystem 'loses'
> information ( files disappear ).  The errors from syslog are 'error during
> decompression', typically the error numbers ( from 'linux/zlib_fs.h' ) are
> Z_BUF_ERROR, for the first error,  and Z_DATA_ERROR for the rest.  I've got

Ummm... can you please explain EXACTLY what you are doing? You have a
cramfs in flash, mounted as root, and you are erasing and overwriting
the SAME cramfs partition WHILE RUNNING FROM IT?

Wolfgang Denk

--
Software Engineering:  Embedded and Realtime Systems,  Embedded Linux
Phone: (+49)-8142-4596-87  Fax: (+49)-8142-4596-88  Email: wd@denx.de
Wisdom is one of the few things that looks bigger the further away it
is.                               - Terry Pratchett, _Witches Abroad_

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: cramfs root filesystem corruption

[Kári Davíðsson]

Let me guess?

Are you updateing the cramfs on a running system, i.e. a system where the cramfs
is mounted somewhere?

K.D.

> -----Original Message-----
> From: Paul Ruhland [mailto:pruhland@microwavedata.com]
> Sent: 30. maí 2002 18:51
> To: linuxppc-embedded@lists.linuxppc.org
> Subject: cramfs root filesystem corruption
>
>
>
> I've found a few posts similar to this from about a year back
> in the main
> kernel list but I could never follow the thread to any conclusion.
>
> I'm running a cramfs root filesystem on an 850 system,
> linux-2.4.4 ( from denx
> ).  Everything has been running fine but the cramfs
> filesystem has started
> behaving badly.
>
> The cramfs image is in flash, mounted as root via mtd blockdevice.
>
> Typically the error occurs during flash reprogramming ( using
> mtd char device
> '/dev/mtd0' ) ...during/after flash programming the cramfs
> filesystem 'loses'
> information ( files disappear ).  The errors from syslog are
> 'error during
> decompression', typically the error numbers ( from
> 'linux/zlib_fs.h' ) are
> Z_BUF_ERROR, for the first error,  and Z_DATA_ERROR for the
> rest.  I've got
> my BDI breaking at the uncompress error and the source length is huge:
>
> (gdb) frame 0
> #0  cramfs_uncompress_block (dst=0xc078c000, dstlen=4096,
> src=0xc0145ac0,
> srclen=4969176) at uncompress.c:49
>
> The reprogramming of the flash is successful, all new data
> has been written to
> the correct location(s).  The cramfs image ( root filesystem
> ) in flash is
> NOT corrupt.  A hard reboot and everything is fine.
>
> Has anyone seen such cramfs problem before?  I'm just looking
> to see if there
> is any 'known' problems before I get to far into debugging this thing.
>
> --
> Paul Ruhland
>


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

RE: cramfs root filesystem corruption

[Kári Davíðsson]

Inactive meaning "unmounted" or just not beeing used?

K.D.

> -----Original Message-----
> From: Paul Ruhland [mailto:pruhland@microwavedata.com]
> Sent: 31. maí 2002 14:04
> To: wd@denx.de
> Cc: linuxppc-embedded@lists.linuxppc.org
> Subject: Re: cramfs root filesystem corruption
>
>
>
> 200205301450.50112.pruhland@microwavedata.com> you wrote:
> >> The cramfs image is in flash, mounted as root via mtd blockdevice.
> >> Typically the error occurs during flash reprogramming (
> using mtd char
> >> device '/dev/mtd0' ) ...during/after flash programming the cramfs
> >> filesystem 'loses' information ( files disappear ).  The
> errors from syslog
> >> are 'error during decompression', typically the error
> numbers ( from
> >> 'linux/zlib_fs.h' ) are Z_BUF_ERROR, for the first error,
> and Z_DATA_ERROR
> >> for the rest.
> >
> > Ummm... can you please explain EXACTLY what you
> > are doing? You have a
> > cramfs in flash, mounted as root, and you are erasing and
> overwriting
> > the SAME cramfs partition WHILE RUNNING FROM IT?
> >
>
> No, I have redundant cramfs images in flash.  I am erasing
> and overwriting the
> inactive image while the other, active image,  is mounted as root.
>
>
>
>

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: cramfs root filesystem corruption

[Paul Ruhland]

If it was mounted it certainly wouldn't qualify as 'inactive', at least by my
standards.  It is not mounted or being used in any way by the running system.


> -----Original Message-----
> From: Kári Davíðsson [mailto:kd@flaga.is]
> Sent: Friday, May 31, 2002 10:07 AM
> To: pruhland@microwavedata.com; wd@denx.de
> Cc: linuxppc-embedded@lists.linuxppc.org
> Subject: RE: cramfs root filesystem corruption
>
>
> Inactive meaning "unmounted" or just not beeing used?
>
> K.D.
>
> > -----Original Message-----
> > From: Paul Ruhland [mailto:pruhland@microwavedata.com]
> > Sent: 31. maí 2002 14:04
> > To: wd@denx.de
> > Cc: linuxppc-embedded@lists.linuxppc.org
> > Subject: Re: cramfs root filesystem corruption
> >
> > 200205301450.50112.pruhland@microwavedata.com> you wrote:
> > >> The cramfs image is in flash, mounted as root via mtd blockdevice.
> > >> Typically the error occurs during flash reprogramming (
> >
> > using mtd char
> >
> > >> device '/dev/mtd0' ) ...during/after flash programming the cramfs
> > >> filesystem 'loses' information ( files disappear ).  The
> >
> > errors from syslog
> >
> > >> are 'error during decompression', typically the error
> >
> > numbers ( from
> >
> > >> 'linux/zlib_fs.h' ) are Z_BUF_ERROR, for the first error,
> >
> > and Z_DATA_ERROR
> >
> > >> for the rest.
> > >
> > > Ummm... can you please explain EXACTLY what you
> > > are doing? You have a
> > > cramfs in flash, mounted as root, and you are erasing and
> >
> > overwriting
> >
> > > the SAME cramfs partition WHILE RUNNING FROM IT?
> >
> > No, I have redundant cramfs images in flash.  I am erasing
> > and overwriting the
> > inactive image while the other, active image,  is mounted as root.
> >
> >


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/