Re: infinite loop on startup

Re: infinite loop on startup

[Stas Sergeev]

Hello.

James B. Hiller wrote:
> Starting Windows 95...
> (my dos is from Win95)
> and it hangs.
[]
> I rolled back to 1.2.0.0 since I didn't seem to be having the problem
> with that version, but now it's doing the same thing.
So it is not a dosemu problem, right?
What kernel are you using? I think I
can reproduce your problem on 2.6
kernels now, except that in my case
the presense of dosdebug changes
nothing.

> Any ideas?  And how should I tailor the debug to provide a useful
> (and reasonably sized) file for someone to look at?
Please do a -D9+r log. If that indeed
is the same problem that I have, you
will get a fix.

Re: infinite loop on startup

[James B. Hiller]

> > Starting Windows 95...
> > (my dos is from Win95)
> > and it hangs.
> []
> > I rolled back to 1.2.0.0 since I didn't seem to be having the problem
> > with that version, but now it's doing the same thing.
> So it is not a dosemu problem, right?

I suspect not.

> What kernel are you using? I think I

Am using 2.6.6-rc3-mm2, but as you imply, it certainly has been behaving
in this way for most if not all of the 2.6 kernels.

> can reproduce your problem on 2.6
> kernels now, except that in my case
> the presense of dosdebug changes
> nothing.

And that may actually be a poor observation on my part.  I've noticed
since reporting this that it happens sometimes and sometimes not, with
no pattern yet discernible to me.  I did start it with dosdebug a number
of times and it never exhibited the behavior, but I may not have done
so enough.

> 
> > Any ideas?  And how should I tailor the debug to provide a useful
> > (and reasonably sized) file for someone to look at?
> Please do a -D9+r log. If that indeed
> is the same problem that I have, you
> will get a fix.

Ok.  Thx for guiding me on the debug options.  There's just so many options
and so much output, when I don't see a clear connection, it's rather hard
to guess.

Will be within a few hours, I hope.

thx,
jbh

Re: infinite loop on startup

[Stas Sergeev]

Hello.

James B. Hiller wrote:
> Running now with preemption enabled.  linux boots fine, no oopses (yet);
> dosemu starts fine all the time (20 starts successful) and apps seem
> to run fine.
Ok, thanks for testing.
I am CC'ing that part to linux-msdos to
make people aware about the problem and
the solution.

Re: infinite loop on startup

[Stas Sergeev]

[-- Attachment #1: Type: text/plain, Size: 820 bytes --]


Hello.

James B. Hiller wrote:
> Please let me know if this looks like your issue, and if you need
> anything else.
Yes, from the log it looks exactly the
same problem I have observed.
And here is the fix:
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0405.0/1242.html

It is SMP-unsafe though, so in case you
have an SMP box with preemption enabled,
you'd better wait till the next kernel
-pre release:
http://www.uwsg.indiana.edu/hypermail/linux/kernel/0405.0/1265.html

Or you can use the attached patch instead,
which does exactly what Linus said. But I
haven't tested that one due to a lack of an
SMP box handy.

The problem is that noone yet have audithed
the 2.6 kernel for dosemu-compatibility the
way it was done with 2.4. So using 2.6 with
dosemu is not recommended yet, but I already
started that work.

[-- Attachment #2: tss_iobtm1.diff --]
[-- Type: text/plain, Size: 508 bytes --]


--- linux/arch/i386/kernel/process.c	2004-04-14 09:41:14.000000000 +0400
+++ linux/arch/i386/kernel/process.c	2004-05-07 14:54:13.000000000 +0400
@@ -292,8 +292,12 @@
 
 	/* The process may have allocated an io port bitmap... nuke it. */
 	if (unlikely(NULL != tsk->thread.io_bitmap_ptr)) {
+		int cpu = get_cpu();
+		struct tss_struct *tss = init_tss + cpu;
 		kfree(tsk->thread.io_bitmap_ptr);
 		tsk->thread.io_bitmap_ptr = NULL;
+		tss->io_bitmap_base = INVALID_IO_BITMAP_OFFSET;
+		put_cpu();
 	}
 }
 

[-- Attachment #3: Type: text/plain, Size: 82 bytes --]


Scanned by evaluation version of Dr.Web antivirus Daemon 
http://drweb.ru/unix/

Infinite loop on startup

[James B. Hiller]

Hi.

I just went from dosemu 1.2.0 to 1.2.1, and I think something broke
along the way.

For installation, I simply compiled from source (no issues) and did
the usual make install.  No other changes.

Now, when I start xdosemu, I get:

Linux DOS emulator 1.2.1.0 $Date: 2004/03/06$
Last configured at Sun Mar 14 11:49:18 EST 2004 on linux
This is work in progress.
Please test against a recent version before reporting bugs and problems.
Submit Bug Reports, Patches & New Code to linux-msdos@vger.kernel.org or via
the SourceForge tracking system at http://www.sourceforge.net/projects/dosemu
 
DPMI-Server Version 0.9 installed
 
Starting Windows 95...

(my dos is from Win95)

and it hangs.

When I do the same thing and turn on all debugging, I get a couple million
lines that seems stuck with:

..
Do INT0x16: Using caller_function()
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Do INT0x16: Using caller_function()
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
Return for FORCE_PIC
..

For booting, I am using an hdimage file in /var/lib/dosemu/hdimage that
contains the following autoexec.bat and config.sys:

autoexec.bat:

@echo off
echo "Welcome to dosemu 1.2.0!"
SET PATH=C:\;C:\DOS;C:\DOS\COMMAND;C:\BUDGET;C:\BATCH;C:\MENU;C:\MSCMOUSE;
lredir e: linux\fs
lredir d: linux\fs\dosd
lredir c: linux\fs\dosc
rem gmouse

config.sys:

rem this sets up the redirector to look at the Linux root directory.
rem this is probably not a good idea once you have dosemu installed
rem device=c:\emufs.sys /

FILES=40
device=c:\ems.sys
DOS=UMB,HIGH
shell=c:\command.com /e:4096 /p

rem DOS=UMB,HIGH
rem FILES=80
rem rem install=c:\command.com /E:4096 /P
rem LASTDRIVE=I
rem devicehigh=c:\ems.sys

and I have identical such files on my hard drive dos partition.

I scanned through the inital portion of the debug output and nothing
SEEMS to be out of line, other than that area that seems to be doing
the reads of all the directories on /dosc doesn't seem to be getting
ALL of /dosc.  It's pretty big, and has a lot of directories.  This
was a topic of discussion a couple months ago with Bart, in that he
said there may be a problem reading in a very big directory, but it
wasn't causing an I-never-finish-initializing scenario.

I rolled back to 1.2.0.0 since I didn't seem to be having the problem
with that version, but now it's doing the same thing.

Any ideas?  And how should I tailor the debug to provide a useful
(and reasonably sized) file for someone to look at?

thx,
jbh