User-mode Linux kernel port

I rediscovered a bug that someone noticed with UML + reiserfs a while ago that 
I never fixed.

It is now fixed, submitted to CVS, and the patch has been updated.

				Jeff

I'm not having much luck checking it into CVS at the moment, but it is 
available as a patch from sourceforge and from ftp.nl.linux.org:/pub/uml/uml-pa
tch-2.4.1.bz2

There is now page dirty and access bit support.

The serial line and console drivers now support pts devices and file 
descriptors as I/O sources.  Input and output channels can also be speciifed 
separately.  This makes it possible to do 'no-xterm' again, with something 
like 'con0=fd:0,fd:1 con=pty'.

Exported kernel_thread and __const_udelay.

Fixed various bugs.

				Jeff

Hello,
I have a small & simple OS running on SH3.
I want to create a Linux environment on that.

I have a few questions in this regard.
Is there any precondition to run Linux on such an OS? Or should the port 
address this need.
If I have some devices which have to be controlled by the native OS , and a 
linux app tries to write to the same device,
would I have to emulate it.
Is it possible to have the foll scenario.
device used only by my OS
device used used only by Linux
device used by both the OS and Linux -here virtualization would be required 
right.
But in the UML port I could not find any code where virtualization is being 
done.
Can someone point me to this section in the code.

Thanks
Santosh

Hi,
=A0 =A0 =A0 =A0 I was investigating at SMP support provided by user-mod=
e-linux. As
of today, there is none. If anyone is working on adding SMP support, ca=
n
they give me a alpha (or beta) version of it? I am looking at using SMP=

enabled user-mode-linux to study scalability limitations of Linux.

Thanks,
vatsa
=

jay.robertson@... said:
> Mounted devfs on /dev 
> Kernel panic: Unexpectedly got signal 4 in signals 

That's a SIGILL that something is getting unexpectedly.  A few other people 
have seen this, but I never have.

> Unfortunately, I get the same error on the  kernel that I made--but I
> could at least get a backtrace from my kernel.   A backtrace yielded
> the following:

That was the wrong backtrace.  What I'd like to see is a backtrace of the 
process that got the illegal instruction (see http://user-mode-linux.sourceforg
e.net/trouble.html for the details).  I'd also like to know what process that 
was (presumably init, but look at current_task.comm anyway, and show me 
current_task.thread while you're at it).

And I'd like to know about the instruction that faulted.  What is it, and how 
did it get there?

				Jeff

I've been seeing this consistently and am trying to track down the
cause. The symptoms are all the xterms and the gdb session are locked
up.
I've been able to use gdb to trace the symptom down to the point where
the lockd kernel daemon terminates. The process is dead (ps -ef on the
real host shows it's gone). But it remains the current uml task.

Once I'm in the locked stated I've managed to attach gdb yet again to
the tracing deamon via 'gdb linux <pid>'. If I then ^C the first
('locked up') gdb session the signal does make the tracing thread return
from waitpid and process it. It goes into proxy.c:debugger_syscall and
tries to pass the signal on to the nonexisting lockd process.

In general, it looks like no more rescheduling occurs once lockd
terminates. I think the umount process should be ready to run, but this
does no good if nobody ever invokes the scheduler. But this never seems
to happen. If I can trust gdb, the idle loop is no longer running. And
shouldn't the idle_timer at least be firing? Seems like we must be stuck
with SIGVTALRM and SIGALRM blocked out, which suggests we're constantly
handling some other signal like SIGSEGV or SIGTRAP. I think someone
mentioned seeing a situation where UML was getting a constant stream of
SIGSEGV's. I tried using gdb's 'handle SIGSEGV print ...' command but
nothing took the bait.

Any ideas?

Thanks,
--gmcnutt

Hello,

I'm getting the following error when trying to run the 2.4.0 User-mode 
Linux kernel.  Upon running "./linux", I get the following output:

Initializing stdio console driver
Initializing software serial port version 1
devfs: v0.102 (20000622) Richard Gooch (rgooch@...)
devfs: devfs_debug: 0x0
devfs: boot_options: 0x0
VFS: Mounted root (ext2 filesystem) readonly.
Mounted devfs on /dev
Kernel panic: Unexpectedly got signal 4 in signals

The kernel off of the download site did not appear to have the ability to 
do a "linux debug" command.  Therefore, I tried to get around the problem 
by compiling my own kernel.  Unfortunately, I get the same error on the 
kernel that I made--but I could at least get a backtrace from my kernel.  
A backtrace yielded the following:

#0  panic (fmt=0x10119b60 "Unexpectedly got signal %d in signals")
    at panic.c:100
#1  0x100a5de1 in signals (init_proc=0x100a6058 <start_kernel_proc>, 
    sp=0x1012bffc) at trap_user.c:215
#2  0x100a6573 in linux_main (argc=3, argv=0xbffffad4) at um_arch.c:196
#3  0x100012a7 in main (argc=3, argv=0xbffffad4, envp=0xbffffae4)
    at /fs1/vm/support/kernel-vm/linux/arch/um/main.c:69

After the panic, the kernel seems to take up all the CPU.  The following 
processes continue to run, which have to be killed manually:

22474 pts/1    R      0:03 ./linux [(tracing thread)]
22475 pts/1    T      0:01 ./linux [(idle thread)]
22476 pts/1    T      0:00 ./linux [init]
22477 pts/1    T      0:00 ./linux [(kernel thread)]
22478 pts/1    T      0:00 ./linux [(kernel thread)]
22479 pts/1    T      0:00 ./linux [(kernel thread)]
22480 pts/1    T      0:00 ./linux [(kernel thread)]
22481 pts/1    T      0:00 ./linux [(kernel thread)]
22482 pts/1    S      0:00 ./linux [init]

I am running this on a RedHat 6.2 system with a patched 2.2.14 kernel 
(which is the stock kernel for that distribution).  The user-mode kernel I 
am using is, of course, the latest 2.4.0 release.  I have tried running 
the kernel against two different root filesystems, but the results are all 
the same.

As far as compiling the kernel, I followed the exact procedure that you 
outlined, with the exception that /usr/src/uml/linux is a symbolic link to 
the real location of the kernel.

Suggestions??

- Jay

Today I did some more playing around with HZ in order to get rid of
warnings ps and w display in UML. I raised HZ to 100 from 20, and
commands like sleep paused for longer periods of time. In fact, w was
inaccurate, instead of showing 5 minutes up for example it would show
only one minute.

Therefore this change seems to be more determinal just to get rid of
warnings. I will look for another way to get rid of these warnings.

--
Mooneer Salem
Sysadmin, Ultraspeed UK (http://www.ultraspeed.co.uk/)
GPLTrans (http://www.translator.cx/)
IRC: irc.slashnet.org/mind21_98
Personal Home Page (http://msalem.translator.cx/)

At Sunday, 07 January 2001, you wrote:

>

>Yeah, I don't boot a lot of those filesystems all that often, so 
they can fall 
>out of date, but it's generally not fatal.

I did have a lot of failure with your version of toms root/boot disk.


Cheers
J




===================================================================
EASY and FREE access to your email anywhere: http://Mailreader.com/
===================================================================

mooneer@... said:
> 1. I removed the pauses after "execlp of xterm failed"...that way I
> can start it automatically via system startup scripts. 

What pauses?  I don't see any pauses in there.  And how is that preventing it 
from being started from rc?

> (Note to Jeff: I think I figured out why it was stalling on startup
> after the HZ change, so everything seems to work fine with HZ=100 now).

What was the problem?

> 3. I was able to get UML to run when I compiled it with pgcc and -O9.
> This also brought an additional performance improvement. 

You're living on the edge.  The main kernel doesn't go over -O2 because (maybe 
among other reasons) it does over-agressive inlining which actually hurts 
performance because the duplicated code shrinks the effective size of the 
i-cache.

				Jeff

I messed around with UML a bit right now and I made the following changes
in my source tree:

1. I removed the pauses after "execlp of xterm failed"...that way I can
start
   it automatically via system startup scripts.
2. I changed HZ to 100 instead of 20, perliminary testing shows it brought a
   performance improvement (this was discussed a bit on IRC in
#kernelnewbies)
   (Note to Jeff: I think I figured out why it was stalling on startup after
the HZ
    change, so everything seems to work fine with HZ=100 now). This change
also
   removed the annoying warning some Linux utilities generated.
3. I was able to get UML to run when I compiled it with pgcc and -O9. This
also
   brought an additional performance improvement.

If anyone's interested I'll clean up my changes a bit and submit a patch.

Thanks,

--
Mooneer Salem
Sysadmin, Ultraspeed UK (http://www.ultraspeed.co.uk/)
GPLTrans (http://www.translator.cx/)
IRC: irc.slashnet.org/mind21_98
Personal Home Page (http://msalem.translator.cx/)

> You've got two problems here, and one of them is mine:
> 
> > In uml I continue the debian installation off of cdrom and as I say ok
> > to the final screen I get a "Kernel panic: Kernel mode fault at addr
> > 0xbefffe90, ip 0x1009f315" from user-mode linux which is running as
> > me, not as root.
> 
> Can you get me a stack trace from the panic?

Here is the info from my latest crash, linux-2.4.0-prerelease + uml-linux-2.4.0-
prerelease.  If you need more backtrace, or I have gathered info on the
wrong address, please let me know.  I'm leaving gdb running for now so
I can gather any more info you might need.

FWIW, when I'm not finding creative ways to corrupt my filesystem, uml
is quite impressive.  Also, one note, I believe that the error occurs when
the debian installer moves inittab out of the way and replaces it with
inittab.real and I assume sends "telinit q" to reread the inttab file during
the end of the second stage of the installation.

Kernel panic: Kernel mode fault at addr 0xbeffffcc, ip 0x1009ba16
(gdb) i sym 0x1009ba16
signal_deliverer + 6 in section .text
(gdb) i line *0x1009ba16
Line 42 of "syscall_user.c" starts at address 0x1009ba10 <signal_deliverer>
   and ends at 0x1009ba19 <signal_deliverer+9>.
(gdb)bt
#0  panic (fmt=0x1010c080 "Kernel mode fault at addr 0x%lx, ip 0x%lx")
    at panic.c:100
#1  0x1009cf2f in segv (address=3204448204, ip=269072918, is_write=2, 
    is_user=0) at trap_kern.c:66
#2  0x1009d824 in segv_handler (sig=11) at trap_user.c:290
#3  0x100a4108 in __restore ()
    at ../sysdeps/unix/sysv/linux/i386/sigaction.c:127
#4  0x1009baf6 in process_stack_handler (sig=12) at syscall_user.c:68
#5  0x100a4108 in __restore ()
    at ../sysdeps/unix/sysv/linux/i386/sigaction.c:127
#6  0x10098af2 in signal_handler (task=0x51268000, h=134550864, sig=1)
    at process.c:61
#7  0x1009f6ec in deliver_signals (t=0x51268000, altstack=0)
    at signal_kern.c:74
#8  0x1009ba74 in signal_deliverer (altstack=0) at syscall_user.c:57
#9  0x1009baf6 in process_stack_handler (sig=12) at syscall_user.c:68
#10 0x100a4108 in __restore ()
    at ../sysdeps/unix/sysv/linux/i386/sigaction.c:127
#11 0x10098af2 in signal_handler (task=0x51268000, h=134550864, sig=1)
    at process.c:61
#12 0x1009f6ec in deliver_signals (t=0x51268000, altstack=0)
    at signal_kern.c:74
#13 0x1009ba74 in signal_deliverer (altstack=0) at syscall_user.c:57
#14 0x1009baf6 in process_stack_handler (sig=12) at syscall_user.c:68
#15 0x100a4108 in __restore ()
    at ../sysdeps/unix/sysv/linux/i386/sigaction.c:127
#16 0x10098af2 in signal_handler (task=0x51268000, h=134550864, sig=1)
    at process.c:61
#17 0x1009f6ec in deliver_signals (t=0x51268000, altstack=0)
    at signal_kern.c:74
#18 0x1009ba74 in signal_deliverer (altstack=0) at syscall_user.c:57
#19 0x1009baf6 in process_stack_handler (sig=12) at syscall_user.c:68
#20 0x100a4108 in __restore ()
    at ../sysdeps/unix/sysv/linux/i386/sigaction.c:127

You've got two problems here, and one of them is mine:

> In uml I continue the debian installation off of cdrom and as I say ok
> to the final screen I get a "Kernel panic: Kernel mode fault at addr
> 0xbefffe90, ip 0x1009f315" from user-mode linux which is running as
> me, not as root.

Can you get me a stack trace from the panic?  See http://user-mode-linux.source
forge.net/trouble.html if you need information on doing that.

				Jeff

daw@... said:
> My telnetd is able to be run from inetd, which attaches fds 0,1,2 to
> the network socket before execing telnetd.  Have you tried running
> telnetd with fds 0,1,2 attached to the device?  It just might work.

My telnetd doesn't appear to be able to do that.  My cheezy
	telnetd 2>&1 > /dev/tty/2 < /dev/tty2

resulted in telnetd bombing out because it couldn't run getpeername on its 
input fd because it's not a socket.

So I think I'm back to either nc, which is great except for the stty thing, or 
my scheme of attaching the uml device to a pty, attaching telnetd to the 
socket, and have the -L helper copy data between them.

				Jeff