Re: libdvdread: Can't seek to block 256

Mon Dec 08 2003, Felix Kurth wrote:
> Hi 
> 
> using kernel 2.6.0-test11, i encounter, the following by invoking:
> tcprobe -i /dev/dvd
> (of course xine is broken too)
> 
(I'm not sure what tcprobe is but anyway...)

> libdvdread: Using libdvdcss version 1.2.8 for DVD access
> libdvdread: Can't seek to block 256
> libdvdread: Can't seek to block 256
> libdvdread: Can't open file VIDEO_TS.IFO.
> (iodump.c) unable to open directory "/dev/dvd"
> 
> im not using ide-scsi, since its deprecaded on that kernel, also cd/dvd
> -burning ist broken with it. (Tried it with ide-scsi, dvd-reading does
> work, but not cd-burning)
> 
dvd-reading with tcprobe or just mount ls/cp and such?


> at last, here is the output of a strace:
> 
[-snip-]
> felix@~ 116 files 81Mb -> strace tcprobe -i /dev/dvd
> open("/dev/dvd", O_RDONLY|O_LARGEFILE)  = 3
> fstat64(3, {st_mode=S_IFBLK|0600, st_rdev=makedev(22, 0), ...}) = 0
> ioctl(3, 0x5390, 0xbfffbf30)            = 0
> read(3, "", 2048)                       = 0
> _llseek(3, 0, [0], SEEK_SET)            = 0
> _llseek(3, 524288, 0xbfffc5a0, SEEK_SET) = -1 EINVAL (Invalid argument)
> write(2, "libdvdread: Can\'t seek to block "..., 36libdvdread: Can't seek to
> block 256
> ) = 36

_llseek(3, 524288, 0xbfffc5a0, SEEK_SET) = -1 EINVAL

Well reading the man page

       int _llseek(unsigned int fd, unsigned long offset_high,  unsigned  long
       offset_low, loff_t *result, unsigned int whence);

       The  _llseek  function repositions the offset of the file descriptor fd
       to (offset_high<<32) | offset_low bytes relative to  the  beginning  of
       the  file,  the  current  position in the file, or the end of the file,
       depending on whether whence is SEEK_SET, SEEK_CUR, or SEEK_END, respec-
       tively.

So the offset seems to be awfully large here... much larger than the DVD
could possibly be... in fact it's close to INT_MAX Mbyte.

I'm no sure what could be causing this, some possiblilities are wrong/buggy
code in libdvdread, in libdvdcss, in the dvdcss.h file, the Makefile of
either of these packages, a compiler bug (i.e. gcc misscompiling it), a
broken gnulibc header, broken largefile system code in gnulibc, a kernel
bug.

So which of these can we rule out?

> so the problem is related to the atapi stuff ?

Well it would seem so if it all the same binaries work when using ide-scsi.

> Drive is a brand new Plextor PX708A DVD+-RW.
> If more information are needed please contact me. I tried to debug
> libdvdread, recompiled transcode and libdvdread with -g, but i had no fun,
> since gdb crashed very often (GNU gdb 5.3). So i gave up. Maybe i should
> try it again with some help. Or is this a kernel bug that i should report ?
> 
> has anybody an idea/fix ?

You could try to write out the block offset value in the
UDFReadBlocksRaw function just before it calls dvdinput_seek.  
and/or inside the css_seek() wrapper function 
and/or inside libdvdcss's dvdcss_seek() (see src/libdvdcss.c)
and/or inside it's various pf_seek() implementations.  For you it
would probably be libc_seek (device.c).  The cast/multiplication from
int (blocks) to off_t is in the libc_seek function...

-- 
Håkan Hjort