ET and HOMAM3 broken with Linux 2.6 emulation [WAS: Moving to a more recent linux base, when?]

[Michal Varga]

On Fri, 2007-08-24 at 01:21, Alexander Leidinger wrote:
> > And, one more bad news, Heroes of Might And Magic 3 (Loki's version)
> > too doesn't start (as with Enemy Territory, I noticed that not long
> > after 2.6 switch, but didn't have much time for games back then and
> > soon forgot about it. I just assumed that this will be fixed later
> > with more 2.6 progress and didn't care much then. Well, it still
> > doesn't work).
> 
> Without reports there's probably no fix.

Yes but my point there was, imagine that right after the initial import
of the new experimental linux 2.6 support into the tree, people start to
flood you with reports like "omg my linuz gamez stopped working!11 plz
fix it 4 me asap!!!tnxz".

It's not like they are really mission critical and I'm running my
desktop on -CURRENT for, let's see.. last 6 years? I've seen so many
things so hideously broken, that it's almost like running Windows for a
week. So I won't report a broken game just days after thing like 2.6
support was commited.


> Run the programs within "ktrace -i". Then download linux_kdump from  
> http://www.leidinger.net/FreeBSD/ (i386 version for an old version of
> -current...) or compile it yourself. Use it to decode the ktrace dump.
> The resulting text may help to find the cause.
> 
> Bye,
> Alexander.

Thanks for pointing out the linux_kdump, now it's obvious (to me) why I
wasn't able to get anything meaningful from the native ktrace/kdump
combo. But then, as far as I can trace out a library or disk/io problem
here and there, I'm not very proficient in low level and kernel
internals. In this case it seems it's not a stuff I can handle:

Heroes of Might And Magic 3 breaks with:
  6578 heroes3  CALL  linux_getppid                                                                         
  6578 heroes3  RET   linux_getppid 6577/0x19b1                                                             
  6578 heroes3  CALL  read(0x6,0x84784a0,0x94)                                                              
  6578 heroes3  GIO   fd 6 read 148 bytes                                                                   
       " D@\b\0\0\0\0�ֿ�$�%\b\0006G\b\0\0\0\M^@\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\  
        \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\ 
        \0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0 2G\b\0\0\0\0H�G\b�\M^]@\b\^X\0\0\0�\0\0\0С@\b"                 
  6578 heroes3  RET   read 148/0x94                                                                         
  6578 heroes3  CALL  linux_mmap(0x84783ec)                                                                 
  6578 heroes3  RET   linux_mmap -1090519040/0xbf000000                                                     
  6578 heroes3  PSIG  SIGSEGV SIG_DFL  

Enemy Territory breaks with:
  5538 et.x86   CALL  gettimeofday(0xbfbfdcb0,0)
  5538 et.x86   RET   gettimeofday 0
  5538 et.x86   CALL  linux_open(0x2ed330fc,0x2,0x400)
  5538 et.x86   NAMI  "/compat/linux/dev/zero"
  5538 et.x86   NAMI  "/dev/zero"
  5538 et.x86   RET   linux_open 7
  5538 et.x86   CALL  linux_mmap2(0,0x2000,0x7,0x2,0x7,0)
  5538 et.x86   RET   linux_mmap2 759926784/0x2d4b9000
  5538 et.x86   CALL  close(0x7)
  5538 et.x86   RET   close 0
  5538 et.x86   CALL  linux_mmap2(0,0x5b000,0x3,0x22,0xffffffff,0)
  5538 et.x86   RET   linux_mmap2 795836416/0x2f6f8000
  5538 et.x86   CALL  linux_getpid
  5538 et.x86   RET   linux_getpid 5538/0x15a2
  5538 et.x86   CALL  linux_getpid
  5538 et.x86   RET   linux_getpid 5538/0x15a2
  5538 et.x86   CALL  linux_modify_ldt(0x11,0xbfbfdc64,0x10)
  5538 et.x86   RET   linux_modify_ldt 666/0x29a
  5538 et.x86   PSIG  SIGSEGV caught handler=0x808c720 mask=0x0 code=0x0


I don't want to bother anyone just for a few games broken with
experimental linux support, I said that in the beginning. I have a
plenty of linux apps that work flawlessly and I'm fine with that. But if
you see something interesting in the snippets above, that can possibly
help you in fixing a few bugs in 2.6 emulation, I'll be more than happy
to upload full dumps somewhere so someone can investigate.

m.

-- 
Michal Varga <varga at stonehenge.sk>
Stonehenge