Linuxulator on RELENG_7 causes games to hang

Hi,

today someone reported a breakage in the linux emulation
(see PR
ports/118213). As games//HeroesOfMightAndMagic worked fine
on RELENG_6,
some change between RELENG_6 and RELENG_7 must have caused a
bug with
such programs.

games/HeroesOfMightAndMagic starts as supposed and goes into
a wait
state including a <defunct> zombie process.
In contrast to other reported issues, this does not seem to
be related
to the sound directly (although it might be caused by it) as
the process
hangs no matter if sound is enabled or disabled.

games/rt2-demo suffers from the same issue, although it can
be started
without issues if you disable the sound with the -s
parameter. 

System and other information:
RELENG_7 (freshly built two days ago)

# cat /dev/sndstat
FreeBSD Audio Driver (newpcm: 32bit 2007061600/i386)
Installed devices:
pcm0: <Intel ICH5 (82801EB)> at io 0xfebffa00,
0xfebff900 irq 17 bufsz 16384 kld snd_ich [MPSAFE]
(1p:1v/1r:1v channels duplex default)

# dmesg |fgrep pcm
pcm0: <Intel ICH5 (82801EB)> port
0xee00-0xeeff,0xedc0-0xedff mem
0xfebffa00-0xfebffbff,0xfebff900-0xfebff9ff irq 17 at device
31.5 on pci0
pcm0: [ITHREAD]
pcm0: primary codec not ready!
pcm0: <Analog Devices AD1980 AC97 Codec>

# sysctl compat
compat.linux.oss_version: 198144
compat.linux.osrelease: 2.4.2
compat.linux.osname: Linux

Last lines of truss on games/rt2-demo:

linux_open("/dev/dsp",0x1,00)			 = 10 (0xa)
linux_ioctl(0xa,0xc004500a,0xbfbfc434,0x400,0x10,0x6) = 0
(0x0)
linux_ioctl(0xa,0xc0045005,0xbfbfc430,0xbfbfc430,0x10,0x6) =
0 (0x0)
linux_ioctl(0xa,0xc0045003,0xbfbfc444,0xbfbfc430,0x0,0x6) =
0 (0x0)
linux_ioctl(0xa,0xc0045002,0xbfbfc430,0xbfbfc430,0x0,0x6) =
0 (0x0)
linux_getpid(0x82c9580,0x0,0x0,0xbfbfc46c,0x10,0x6) = 50040
(0xc378)
linux_rt_sigprocmask(0x2,0x0,0xbfbfc7dc,0x8,0x2,0x6) = 0
(0x0)
write(7,"`v,bM^DM-HM-?M-?`!"...,148)
= 148 (0x94)
linux_rt_sigprocmask(0x2,0x0,0xbfbfc748,0x8,0x2,0x6)
-- Waiting -- 

Last lines of truss on games/HeroesOfMightAndMagic:

linux_open("/usr/local/games/Heroes3/data/video/3dologo
.mpg",0x0,0666) = 11 (0xb)
linux_newfstat(11,0xbfbfc604)			 = 0 (0x0)
linux_mmap(0xbfbfc63c,0xbfbfc604,0x0,0xbfbfc67c,0x84ec7b0,0x
6) = 680394752 (0x288e0000)
linux_brk(0x85a6000,0x2980,0x859d000,0x9000,0x859a680,0x6) =
140140544 (0x85a6000)
linux_brk(0x85ba000,0x2d68,0x85a6000,0x14000,0x85a3298,0x6)
= 140222464 (0x85ba000)
linux_brk(0x85cd000,0x34e0,0x85ba000,0x13000,0x85b6b20,0x6)
= 140300288 (0x85cd000)
linux_brk(0x85e0000,0x38d8,0x85cd000,0x13000,0x85c9728,0x6)
= 140378112 (0x85e0000)
linux_brk(0x85e5000,0x3cd0,0x85e0000,0x5000,0x85dc330,0x6) =
140398592 (0x85e5000)
linux_brk(0x85ea000,0x41c8,0x85e5000,0x5000,0x85e0e38,0x6) =
140419072 (0x85ea000)
linux_brk(0x85fd000,0x46c0,0x85ea000,0x13000,0x85e5940,0x6)
= 140496896 (0x85fd000)
linux_brk(0x8602000,0x4ab8,0x85fd000,0x5000,0x85f8548,0x6) =
140517376 (0x8602000)
linux_rt_sigprocmask(0x2,0x0,0xbfbfcb64,0x8,0x2,0x6) = 0
(0x0)
write(6,"r9bfM-LM-?M-?^Pd$"...,148) = 148
(0x94)
linux_rt_sigprocmask(0x2,0x0,0xbfbfcad0,0x8,0x2,0x6) = 0
(0x0)
-- Waiting -- 

Changing the sysctl hw.snd.compat_linux_mmap to 0 or 1 does
not show any
difference for both programs.

Any help to sort that issue out is appreciated and please CC
me as I am
not subscribed to the list.

Regards
Marcus

On Sat, Nov 24, 2007 at 10:13:32AM +0100, Marcus von Appen
wrote:
> Hi,
> 
> today someone reported a breakage in the linux
emulation (see PR
> ports/118213). As games//HeroesOfMightAndMagic worked
fine on RELENG_6,
> some change between RELENG_6 and RELENG_7 must have
caused a bug with
> such programs.
> 
> games/HeroesOfMightAndMagic starts as supposed and goes
into a wait
> state including a <defunct> zombie process.
> In contrast to other reported issues, this does not
seem to be related
> to the sound directly (although it might be caused by
it) as the process
> hangs no matter if sound is enabled or disabled.
> 
> games/rt2-demo suffers from the same issue, although it
can be started
> without issues if you disable the sound with the -s
parameter. 

please try to bisect code changes to identify the exact
change that broke it.
I'll investigate that from that point further..

thnx

roman
_______________________________________________
freebsd-emulationfreebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-emu
lation
To unsubscribe, send any mail to
"freebsd-emulation-unsubscribefreebsd.org"

On, Sat Nov 24, 2007, Roman Divacky wrote:

> On Sat, Nov 24, 2007 at 10:13:32AM +0100, Marcus von
Appen wrote:
> > Hi,
> > 
> > today someone reported a breakage in the linux
emulation (see PR
> > ports/118213). As games//HeroesOfMightAndMagic
worked fine on RELENG_6,
> > some change between RELENG_6 and RELENG_7 must
have caused a bug with
> > such programs.
> > 
> > games/HeroesOfMightAndMagic starts as supposed and
goes into a wait
> > state including a <defunct> zombie process.
> > In contrast to other reported issues, this does
not seem to be related
> > to the sound directly (although it might be caused
by it) as the process
> > hangs no matter if sound is enabled or disabled.
> > 
> > games/rt2-demo suffers from the same issue,
although it can be started
> > without issues if you disable the sound with the
-s parameter. 
> 
> please try to bisect code changes to identify the exact
change that broke it.
> I'll investigate that from that point further..
> 

If you give me a detailled introduction about how the linux
emulation
system is organized in both, application layer, interfaces
and source
code layout, I'll gladly check through it.

Regards
Marcus

On Sat, Nov 24, 2007 at 08:17:07PM +0100, Marcus von Appen
wrote:
> On, Sat Nov 24, 2007, Roman Divacky wrote:
> 
> > On Sat, Nov 24, 2007 at 10:13:32AM +0100, Marcus
von Appen wrote:
> > > Hi,
> > > 
> > > today someone reported a breakage in the
linux emulation (see PR
> > > ports/118213). As
games//HeroesOfMightAndMagic worked fine on RELENG_6,
> > > some change between RELENG_6 and RELENG_7
must have caused a bug with
> > > such programs.
> > > 
> > > games/HeroesOfMightAndMagic starts as
supposed and goes into a wait
> > > state including a <defunct> zombie
process.
> > > In contrast to other reported issues, this
does not seem to be related
> > > to the sound directly (although it might be
caused by it) as the process
> > > hangs no matter if sound is enabled or
disabled.
> > > 
> > > games/rt2-demo suffers from the same issue,
although it can be started
> > > without issues if you disable the sound with
the -s parameter. 
> > 
> > please try to bisect code changes to identify the
exact change that broke it.
> > I'll investigate that from that point further..
> > 
> 
> If you give me a detailled introduction about how the
linux emulation
> system is organized in both, application layer,
interfaces and source
> code layout, I'll gladly check through it.

just go by date... try oldest_date and see if it works, try
newest_date
and see if it works. then just half the interval. bisection.
you should
be able to identify day on which it broke. assigning an
actual commit
that broke it should be easy.

or I dont understand your question...
_______________________________________________
freebsd-emulationfreebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-emu
lation
To unsubscribe, send any mail to
"freebsd-emulation-unsubscribefreebsd.org"

On, Sat Nov 24, 2007, Roman Divacky wrote:

[...]
> > If you give me a detailled introduction about how
the linux emulation
> > system is organized in both, application layer,
interfaces and source
> > code layout, I'll gladly check through it.
> 
> just go by date... try oldest_date and see if it works,
try newest_date
> and see if it works. then just half the interval.
bisection. you should
> be able to identify day on which it broke. assigning an
actual commit
> that broke it should be easy.
> 
> or I dont understand your question...

I just misinterpreted the 'bisect' and though of bisecting
the whole
code changes, not the CVS dates .

It'll take some days, I guess, as it has to be done in a
qemu image
(currently there's no second test system at hand).

Regards
Marcus

On, Sat Nov 24, 2007, Roman Divacky wrote:

> On Sat, Nov 24, 2007 at 10:13:32AM +0100, Marcus von
Appen wrote:
> > Hi,
> > 
> > today someone reported a breakage in the linux
emulation (see PR
> > ports/118213). As games//HeroesOfMightAndMagic
worked fine on RELENG_6,
> > some change between RELENG_6 and RELENG_7 must
have caused a bug with
> > such programs.
> > 
> > games/HeroesOfMightAndMagic starts as supposed and
goes into a wait
> > state including a <defunct> zombie process.
> > In contrast to other reported issues, this does
not seem to be related
> > to the sound directly (although it might be caused
by it) as the process
> > hangs no matter if sound is enabled or disabled.
> > 
> > games/rt2-demo suffers from the same issue,
although it can be started
> > without issues if you disable the sound with the
-s parameter. 
> 
> please try to bisect code changes to identify the exact
change that broke it.
> I'll investigate that from that point further..
> 

At least I could track the issue down the RELENG_6 as well.
It happens
on the most recent RELENG_6 as well as on the 200710
snapshot from the
servers. Unfortunately I had no chance to track it down on
RELENG_7 so
far - getting back in time turned out to be a real pity and
installing a
clean system for every dated cvs snapshot is more than a
pain.

I'll try to step forward from RELENG_6_2 within the next
weeks but
that'll take a lot of time, so it'd be quite helpful, if you
can isolate
the possibly affecting changes between RELENG_6_2 and
RELENG_6 as well.

Regards
Marcus

> servers. Unfortunately I had no chance to track it down
on RELENG_7 so
> far - getting back in time turned out to be a real pity
and installing a
> clean system for every dated cvs snapshot is more than
a pain.

you have to update only the kernel.. that should not be such
a big problem
_______________________________________________
freebsd-emulationfreebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-emu
lation
To unsubscribe, send any mail to
"freebsd-emulation-unsubscribefreebsd.org"

On, Sun Dec 02, 2007, Roman Divacky wrote:

> > servers. Unfortunately I had no chance to track it
down on RELENG_7 so
> > far - getting back in time turned out to be a real
pity and installing a
> > clean system for every dated cvs snapshot is more
than a pain.
> 
> you have to update only the kernel.. that should not be
such a big problem

In fact it was. The kernel did not like to be built without
cleaning
anything else first. Various compile time errors occured,
which made it
impossible to update it separately from the rest - no idea
why that
happened, I guess, that some missing environment setting
messed up the
tree.

Anyhow, I could track the RELENG_6 issue down to the changes
made on
2007-07-08, which were a big MFC made by netchild. A kernel
build
starting from the 8th does not work, the 7th is still okay.

I'm starting once more with RELENG_7 now I have an idea
about the
time frame of RELENG_6.

Regards
Marcus

Quoting Marcus von Appen <mvasysfault.org> (from Tue,
4 Dec 2007  
11:28:58 +0100):

> On, Sun Dec 02, 2007, Roman Divacky wrote:
>
>> > servers. Unfortunately I had no chance to
track it down on RELENG_7 so
>> > far - getting back in time turned out to be a
real pity and installing a
>> > clean system for every dated cvs snapshot is
more than a pain.
>>
>> you have to update only the kernel.. that should
not be such a big problem
>
> In fact it was. The kernel did not like to be built
without cleaning
> anything else first. Various compile time errors
occured, which made it
> impossible to update it separately from the rest - no
idea why that
> happened, I guess, that some missing environment
setting messed up the
> tree.
>
> Anyhow, I could track the RELENG_6 issue down to the
changes made on
> 2007-07-08, which were a big MFC made by netchild. A
kernel build
> starting from the 8th does not work, the 7th is still
okay.
>
> I'm starting once more with RELENG_7 now I have an idea
about the
> time frame of RELENG_6.

You have to look up the timeframe of the corresponding
changes in 7-current.
Just go to cvsweb.freebsd.org/src/sys/ and have a look at
the commit  
history for compat/linux/* and {i386,amd64}/linux{,32}/* to
find  
suitable points in time to look at.

Bye,
Alexander.

-- 
Cthulhu Cthucks!

http://www.Leidinger.net
   Alexander  Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org 
     netchild  FreeBSD.org  : PGP ID = 72077137
_______________________________________________
freebsd-emulationfreebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-emu
lation
To unsubscribe, send any mail to
"freebsd-emulation-unsubscribefreebsd.org"

>
linux_mmap(0xbfbfc63c,0xbfbfc604,0x0,0xbfbfc67c,0x84ec7b0,0x
6) = 680394752 (0x288e0000)
                                       ^^^^^^^^^^
				       this is flags? looks crazy
_______________________________________________
freebsd-emulationfreebsd.org mailing list
http://lists.freebsd.org/mailman/listinfo/freebsd-emu
lation
To unsubscribe, send any mail to
"freebsd-emulation-unsubscribefreebsd.org"