user-mode-linux-devel

boot hang at mounting proc filesystem, was Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[William S. <wst...@po...>]

Good afternoon, Jeff,

On Wed, 8 Nov 2000, Jeff Dike wrote:

> If you really can't get the console to respond, gdb the tracing thread and
> 'printf "%s", log_buf'.  This will dump the contents of the kernel log.  Maybe
> there's something strange there.
>
> Also put 'debug' on the kernel command line, and when it hangs, ^C the gdb and
> 'bt' to see what it thinks is up.

	I compiled my own test10 and left in the symbols.  Added "debug"
to the command line and left off "</dev/null" for the moment.  It's now
hanging at mounting proc filesystem.  When that happens, I hit ctrl-c and
do a backtrace:

600000e
att 1
b start_kernel
c
GNU gdb 5.0
[snip]
This GDB was configured as "i386-redhat-linux"...
(gdb) att 1
Attaching to program: /home/wstearns/uml/rh6.2/linux, Pid 1
0x100b4701 in __kill ()
(gdb) b start_kernel
Breakpoint 1 at 0x100fcc5f: file init/main.c, line 514.
(gdb) c
Continuing.

Breakpoint 1, start_kernel () at init/main.c:514
514             printk(linux_banner);
(gdb) c
Continuing.

Program received signal SIGINT, Interrupt.
0x100ae7bd in set_user_thread (t=0x0, on=0, restore_state=0)
    at process_kern.c:72
72      {
(gdb) bt
#0  0x100ae7bd in set_user_thread (t=0x0, on=0, restore_state=0)
    at process_kern.c:72
#1  0x100ad779 in segv_handler (sig=11) at trap_user.c:263
#2  <signal handler called>
#3  0x10025722 in copy_mount_options (data=0x8058378, where=0x52d3fc80)
    at
/usr/src/uml-linux/linux-2.4.0-test10.uml/include/asm/arch/string.h:202
#4  0x10025a70 in sys_mount (dev_name=0x8058358 "/proc",
    dir_name=0x8058368 "/proc", type=0x8058378 "proc", flags=3236757504,
    data=0x0) at super.c:1430
#5  0x100ab7e2 in execute_syscall (syscall=21, args=0x52d3fce4)
    at syscall_kern.c:340
#6  0x100abc19 in syscall_handler (unused=0) at syscall_user.c:113
#7  <signal handler called>
#8  0x100b4701 in __kill ()
(gdb) printf "%s", log_buf
<4>tracing thread pid = 2617
<4>Linux version 2.4.0-test10-1um (ro...@sp...) (gcc version
2.96
20000731 (Red Hat Linux 7.0)) #1 Sat Nov 11 13:39:17 EST 2000
<4>On node 0 totalpages: 12288
<4>zone(0): 0 pages.
<4>zone(1): 12288 pages.
<4>zone(2): 0 pages.
<4>Kernel command line: mem=48M debug root=/dev/ubd0
<4>Calibrating delay loop... 710.41 BogoMIPS
<4>Memory: 48324k available
<4>Dentry-cache hash table entries: 8192 (order: 4, 65536 bytes)
<4>Buffer-cache hash table entries: 1024 (order: 0, 4096 bytes)
<4>Page-cache hash table entries: 16384 (order: 4, 65536 bytes)
<4>Inode-cache hash table entries: 4096 (order: 3, 32768 bytes)
<5>VFS: Diskquotas version dquot_6.4.0 initialized
<4>POSIX conformance testing by UNIFIX
<6>Linux NET4.0 for Linux 2.4
<6>Based upon Swansea University Computer Society NET3.039
<4>Starting kswapd v1.8
<4>pty: 256 Unix98 ptys configured
<4>RAMDISK driver initialized: 16 RAM disks of 4096K size 1024 blocksize
<6>loop: enabling 8 loop devices
<4>User-mode Linux network interface 0.005 (eth0)
---Type <return> to continue, or q <return> to quit---
<4>User-mode Linux network interface 0.005 (eth1)
<4>User-mode Linux network interface 0.005 (eth2)
<4>User-mode Linux network interface 0.005 (eth3)
<6>NET4: Linux TCP/IP 1.0 for NET4.0
<6>IP Protocols: ICMP, UDP, TCP
<4>IP: routing cache hash table of 512 buckets, 4Kbytes
<4>TCP: Hash tables configured (established 4096 bind 4096)
<4>ip_conntrack (384 buckets, 3072 max)
<4>ip_tables: (c)2000 Netfilter core team
<6>NET4: Unix domain sockets 1.0/SMP for Linux NET4.0.
<4>Initializing stdio console driver
<6>Initializing software serial port version 0
<4>serial line 0 assigned pty /dev/ptyp3
<4>ssl receive thread is pid 2627
<4>devfs: v0.102 (20000622) Richard Gooch (rg...@at...)
<4>devfs: devfs_debug: 0x0
<4>devfs: boot_options: 0x0
<4>VFS: Mounted root (ext2 filesystem) readonly.
<4>Mounted devfs on /dev
(gdb)

	For reference, the console shows the following lines after the
above devfs:

INIT: version 2.78 booting
Started device management daemon for /dev
                        Welcome to Red Hat Linux
                Press 'I' to enter interactive startup.
Mounting proc filesystem

	This appears to be a different problem than the "Hey, what
happened to fd0?" as I'm not using "</dev/null".
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"Unix _is_ user friendly.  It's just very selective about who its
friends are.  And sometimes even best friends have fights."
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts,
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[William S. <wst...@po...>]

Good afternoon, Jeff,

On Wed, 8 Nov 2000, Jeff Dike wrote:

> wst...@po... said:
> > I moved up the serial/0 line after rc.sysinit but before the runlevel
> > specific lines and was able to connect to ttyp9 (I assume it's the
> > ttypX equivalent of the ptypX that the uml kernel opens on FD7) with
> > minicom:
> 
> Why didn't it come up in an xterm?  Did you put 'no-xterm' on the command line?

	I hadn't moved the ttys/[012] lines in inittab up before the
runlevel lines before; I just did.
	The xterms come up and I can even log in - if I do it
quickly.  Just like the serial terminal, I lose the ability to type on
the xterms partway through the boot process.  I can start up something
like top on a term; as soon as fd0 closes, I can no longer type on it and
top freezes.

> In any case '-o' on the minicom command line will stop it from trying to init 
> the modem.

	Cool...

> If you really can't get the console to respond, gdb the tracing thread and 
> 'printf "%s", log_buf'.  This will dump the contents of the kernel log.  Maybe 
> there's something strange there.

	Your kernel is stripped - I'll rebuild and try again.

> Also put 'debug' on the kernel command line, and when it hangs, ^C the gdb and 
> 'bt' to see what it thinks is up.

	Hmmm - when I add debug, it starts the idle and tracing thread and
doesn't get any further.
	Since you and Adam can both run it correctly, I'll play a little
more on this end and see if I can come up with more details.  I'm almost
wondering if there's something in rh / rh7 that's causing problems.
	Cheers,
	- Bill

---------------------------------------------------------------------------
    "I can picture in my mind a world without war, a world without hate.
And I can picture us attacking that world, because they'd never expect
it."
    --Bad Mojo
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[uml-devel] Re: UML Documentation/Configure.help typo?

[Jeff D. <jd...@ka...>]

da...@th... said:
> Actually, I was going to suggest this patch, but I wasn't sure if
> you'd want it

Yeah, I do.  I was planning on making people do "make linux ARCH=um".

> SUBARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/) 
> ARCH := $(SUBARCH) 

I like that one.  It's in my pool.  Except I added a comment to explain what's 
happening.

> In any case, would you care to comment on the topic of integrating UML
> into the standard Linux kernel tree? Are you for it? 

It's going to happen sooner rather than later.  I'm going to send it to Alan 
sometime after 2.4.0 and before Linus forks off the 2.5 pool.  That way I 
don't have to send it in twice.

> What does Linus
> think (if he's commented on the subject in the past)?

The only comment I've heard was from Rusty, who mentioned it to Linus about a 
year ago.  Linus said he thought it should go in for sheer hack value...

				Jeff

[uml-devel] Re: UML Documentation/Configure.help typo?

[Dave Z. <da...@th...>]

On Fri, 10 Nov 2000, Dave Zarzycki wrote:

> -ARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/)
> -
> -SUBARCH := $(ARCH)
> -ARCH = um
> +SUBARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/)
> +ifneq ($(ARCH),um)
> +ARCH := $(SUBARCH)
> +endif

Hmmm... Now that I think about it, this would work just as well:

SUBARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/)
ARCH := $(SUBARCH)

davez

-- 
Dave Zarzycki
http://thor.sbay.org/~dave/

[uml-devel] Re: UML Documentation/Configure.help typo?

[Dave Z. <da...@th...>]

On Sat, 11 Nov 2000, Jeff Dike wrote:

> Maybe everyone is complaining about my writing because they can't find 
> anything wrong with my code :-)

Actually, I was going to suggest this patch, but I wasn't sure if you'd
want it (this patch might be handy if you're interested in eventually
submitting UML for inclusion into Linus's kernel):

--- linux-orig/Makefile	Sun Nov  5 10:19:12 2000
+++ linux/Makefile	Fri Nov 10 21:05:23 2000
@@ -5,10 +5,10 @@
 
 KERNELRELEASE=$(VERSION).$(PATCHLEVEL).$(SUBLEVEL)$(EXTRAVERSION)
 
-ARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/)
-
-SUBARCH := $(ARCH)
-ARCH = um
+SUBARCH := $(shell uname -m | sed -e s/i.86/i386/ -e s/sun4u/sparc64/ -e s/arm.*/arm/ -e s/sa110/arm/)
+ifneq ($(ARCH),um)
+ARCH := $(SUBARCH)
+endif
 
 CONFIG_SHELL := $(shell if [ -x "$$BASH" ]; then echo $$BASH; \
 	  else if [ -x /bin/bash ]; then echo /bin/bash; \


So if you want to build a kernel for your host system, you wouldn't need
to do anything different. If you want to build a user-mode Linux kernel
for your system, all you need to do is one of two options:

make ARCH=um config dep clean linux

or:

"export ARCH=um" or "setenv ARCH um" in your shell.


In any case, would you care to comment on the topic of integrating UML
into the standard Linux kernel tree? Are you for it? What does Linus think
(if he's commented on the subject in the past)?

davez

-- 
Dave Zarzycki
http://thor.sbay.org/~dave/

Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[Jeff D. <jd...@ka...>]

wst...@po... said:
> I moved up the serial/0 line after rc.sysinit but before the runlevel
> specific lines and was able to connect to ttyp9 (I assume it's the
> ttypX equivalent of the ptypX that the uml kernel opens on FD7) with
> minicom:

Why didn't it come up in an xterm?  Did you put 'no-xterm' on the command line?

In any case '-o' on the minicom command line will stop it from trying to init 
the modem.

If you really can't get the console to respond, gdb the tracing thread and 
'printf "%s", log_buf'.  This will dump the contents of the kernel log.  Maybe 
there's something strange there.

Also put 'debug' on the kernel command line, and when it hangs, ^C the gdb and 
'bt' to see what it thinks is up.

				Jeff

Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[Jeff D. <jd...@ka...>]

wst...@po... said:
> I'd love to hear from others on this list; can you successfully start
> up a uml kernel with:
> linux-2.blah.blah </dev/null

I did it.  Otherwise, I wouldn't have claimed that the bug was fixed...

				Jeff

Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[William S. <wst...@po...>]

Good afternoon, Jeff, all,
	A real quick request - I'd love to hear from others on this
list; can you successfully start up a uml kernel with:

linux-2.blah.blah </dev/null

	?  You won't get a console, but if the boot works successfully
you'll get the xterms and you'll get networking.
	Please respond, and include your host distribution, host
kernel, uml distribution, and uml kernel.  I don't mind being the only
problematic child in the classroom, but I'd like to know whether this is
working for others or not.
	I'd especially like to hear whether linux-2.4.0-test10 as a uml
kernel works or not, as Jeff made some fixes to the "</dev/null" code
there.

On Wed, 8 Nov 2000, Jeff Dike wrote:

> wst...@po... said:
> > the boot proceeds pretty well until it gets to the named boot script.
> > If I don't use the </dev/null, umnl boots to xterms and a login
> > prompt.  Once I add </dev/null, it stops somewhere in the named script
> > (a "ping the hotst" script run before named executes, one run after
> > named does not).
> 
> Have you tried enabling a getty on a virtual console at run-level 2?  That 
> would let you log in before named hoses the boot process and you could poke 
> around to see what's happening.

	Good idea.
	I moved up the serial/0 line after rc.sysinit but before the
runlevel specific lines and was able to connect to ttyp9 (I assume it's
the ttypX equivalent of the ptypX that the uml kernel opens on FD7) with
minicom:

Red Hat Linux release 6.2 (Zoot)                                                               
Kernel 2.4.0-test10-1um on a um                                              
                                                                             
redhat62 login: AT S7=45 S0=0 L1 V1 X4 &c1 E1 Q0
                                        
root                                    

	, but I can't actually get it to respond to keystrokes.  Even if
it didn't like minicom's AT..., I should still be able to hit return once
or twice and get to a usable login.
	OK, try it again.  Before we get to the named death, here are the
file descriptors for the parent tracing thread (pts/10 is the Eterm in
which I'm running the uml):

/proc/15198/fd:
total 4
dr-x------    2 wstearns wstearns        0 Nov  8 16:11 .
dr-xr-xr-x    3 wstearns wstearns        0 Nov  8 16:11 ..
lr-x------    1 wstearns wstearns       64 Nov  8 16:11 0 -> /dev/null
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 1 -> /dev/pts/10
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 10 -> /home/wstearns/uml/rh6.2/root_fs
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 2 -> /dev/pts/10
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 3 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 4 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 5 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 6 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 7 -> /dev/ptyp9
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 8 -> /dev/ptmx
lrwx------    1 wstearns wstearns       64 Nov  8 16:11 9 -> /dev/pts/29

	Once we get to named (again, I'm not claiming named is the
culprit, it just seems to happen around that time):

/proc/15198/fd:
total 8
dr-x------    2 wstearns wstearns        0 Nov  8 16:12 .
dr-xr-xr-x    3 wstearns wstearns        0 Nov  8 16:12 ..
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 1 -> /dev/pts/10
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 10 -> /home/wstearns/uml/rh6.2/root_fs
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 11 -> socket:[4900667]
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 2 -> /dev/pts/10
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 3 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 4 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 5 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 6 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 7 -> /dev/ptyp9
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 8 -> /dev/ptmx
lrwx------    1 wstearns wstearns       64 Nov  8 16:12 9 -> /dev/pts/29

	, and no more stdin.
	The host lists the following threads:

[root@sparrow /root]# ps axf | grep linu[x]-2.4 
15198 pts/10   S      0:03      \_ ./linux-2.4.0-test10 [(tracing thread)]
15199 pts/10   S      0:01          \_ ./linux-2.4.0-test10 [(idle thread)]
15204 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [init]
15209 pts/10   S      0:00          |   \_ ./linux-2.4.0-test10 [init]
15210 pts/10   S      0:00          |   \_ ./linux-2.4.0-test10 [init]
15205 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [(kernel thread)]
15206 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [(kernel thread)]
15207 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [(kernel thread)]
15208 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [(kernel thread)]
15215 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [/sbin/devfsd]
15376 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [/sbin/mingetty]
15377 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [/bin/bash]
15557 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [portmap]
15610 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [syslogd]
15621 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [klogd]
15637 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [identd]
15638 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [identd]
15639 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [identd]
15640 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [identd]
15641 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [identd]
15657 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [crond]
15673 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [inetd]
15677 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [/bin/sh]
15689 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [named]
15690 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [initlog]
15772 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [crond]
15778 pts/10   T      0:00          \_ ./linux-2.4.0-test10 [CROND]

	A similar run shows the same type of behavior on tap1 I've seen
before (uml-rh62 is the uml, uml-router is the masqing host kernel):

User level filter, protocol ALL, datagram packet socket
tcpdump: listening on tap1

(start up uml </dev/null)

16:25:05.150023 B arp who-has uml-router tell uml-rh62
16:25:05.150049 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)

(uml finds host via arp)

16:25:05.153059 < uml-rh62.1024 > j.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:05.502358 > j.root-servers.net.domain > uml-rh62.1024: 47718*- 13/0/13 NS A.ROOT-SERVERS.NET., NS H.ROOT-SERVERS.NET., NS C.ROOT-SERVERS.NET., NS G.ROOT-SERVERS.NET., NS F.ROOT-SERVERS.NET., (436) (DF)

(root name service request sent and reply returned)

16:25:09.013729 < uml-rh62.1024 > i.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:09.392363 > i.root-servers.net.domain > uml-rh62.1024: 47718*- 13/0/13 NS L.ROOT-SERVERS.NET., NS M.ROOT-SERVERS.NET., NS I.ROOT-SERVERS.NET., NS E.ROOT-SERVERS.NET., NS D.ROOT-SERVERS.NET., (436)

(another root name service request, we'll assume that the uml named was
just impatient)

16:25:10.502303 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:11.502278 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:12.502277 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)

(Now the host isn't responding to arps anymore and the uml doesn't seem to
be getting the earlier responses or it would have moved on.)

16:25:13.013696 < uml-rh62.1024 > f.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:13.502277 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:13.812356 > f.root-servers.net.domain > uml-rh62.1024: 47718*- 13/0/13 NS E.ROOT-SERVERS.NET., NS D.ROOT-SERVERS.NET., NS A.ROOT-SERVERS.NET., NS H.ROOT-SERVERS.NET., NS C.ROOT-SERVERS.NET., (436) (DF)
16:25:14.502281 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:15.502277 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:17.013653 < uml-rh62.1024 > G.ROOT-SERVERS.NET.domain: 47718 NS? . (17) (DF)
16:25:21.013519 < uml-rh62.1024 > b.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:21.552362 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:22.552280 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:23.552297 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:25.017482 < uml-rh62.1024 > m.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:25.472362 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:26.472283 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:27.472285 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:29.013485 < uml-rh62.1024 > l.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:33.013447 < uml-rh62.1024 > k.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:33.372368 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:34.372294 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:35.372296 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:37.013483 < uml-rh62.1024 > d.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:37.482357 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:38.482280 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:39.482287 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:41.013688 < uml-rh62.1024 > a.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:41.272365 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:42.272284 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:43.272286 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:45.013502 < uml-rh62.1024 > c.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:45.902357 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:46.902282 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:47.902282 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:49.013467 < uml-rh62.1024 > h.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:53.014102 < uml-rh62.1024 > E.ROOT-SERVERS.NET.domain: 47718 NS? . (17) (DF)
16:25:53.402357 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:54.402279 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:55.402279 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:57.013430 < uml-rh62.1024 > j.root-servers.net.domain: 47718 NS? . (17) (DF)
16:25:57.352368 > arp who-has uml-rh62 tell uml-router (fe:fd:0:0:0:0)
16:25:58.012690 < arp who-has uml-router tell uml-rh62
16:25:58.012708 > j.root-servers.net.domain > uml-rh62.1024: 47718*- 13/0/13 NS A.ROOT-SERVERS.NET., NS H.ROOT-SERVERS.NET., NS C.ROOT-SERVERS.NET., NS G.ROOT-SERVERS.NET., NS F.ROOT-SERVERS.NET., (436) (DF)
16:25:58.012714 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:25:59.012641 < arp who-has uml-router tell uml-rh62
16:25:59.012658 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)

(We're back to arp deafness.)

16:26:00.012637 < arp who-has uml-router tell uml-rh62
16:26:00.012652 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:01.012694 B arp who-has uml-router tell uml-rh62
16:26:01.012710 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:02.012676 B arp who-has uml-router tell uml-rh62
16:26:02.012691 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:03.012670 B arp who-has uml-router tell uml-rh62
16:26:03.012686 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:05.013449 B arp who-has uml-router tell uml-rh62
16:26:05.013466 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:06.012704 B arp who-has uml-router tell uml-rh62
16:26:06.012719 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:07.012738 B arp who-has uml-router tell uml-rh62
16:26:07.012756 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:13.013505 B arp who-has uml-router tell uml-rh62
16:26:13.013521 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:14.012645 B arp who-has uml-router tell uml-rh62
16:26:14.012660 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
16:26:15.012651 B arp who-has uml-router tell uml-rh62
16:26:15.012667 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)

72 packets received by filter

	Sorry for the long trace, but essentially it looks like the
communication starts, but then the host and the uml get deaf to each
other.  Without being able to do a dump on the uml side I can't be sure,
but it looks like requests are making it out but the responses just never
make it back.

	Put down the shotgun, Jeff, that's not what they mean by
troubleshooting.  Jeff, put it down.  /me runs...
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"I don't have any solution but I certainly admire the problem."
	- Ashleigh Brilliant
(Courtesy of Steve Dodd <di...@lo...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

Re: [uml-devel] null stdin/stdout strangeness - but getting there!

[Jeff D. <jd...@ka...>]

wst...@po... said:
> the boot proceeds pretty well until it gets to the named boot script.
> If I don't use the </dev/null, umnl boots to xterms and a login
> prompt.  Once I add </dev/null, it stops somewhere in the named script
> (a "ping the hotst" script run before named executes, one run after
> named does not).

Have you tried enabling a getty on a virtual console at run-level 2?  That 
would let you log in before named hoses the boot process and you could poke 
around to see what's happening.

				Jeff

[uml-devel] null stdin/stdout strangeness - but getting there!

[William S. <wst...@po...>]

Good day, Jeff and all,
(Host 2.4.0-test10, rh7, uml root is rh62)
	Thanks for the infinite printk fix, Jeff.  The printk meltdown
found in previous versions ("Hmmm, can't printk that message?  Must alert
the user with a printk!") has disappeared.
	I'm trying to run test10 with stdin and stdout redirected from/to
/dev/null - I'm hoping to use it as a server only platform.  When I use

e2fsck -v -y root_fs ; ./linux-2.4.0-test10 mem=48M </dev/null

	the boot proceeds pretty well until it gets to the named boot
script.  If I don't use the </dev/null, umnl boots to xterms and a login
prompt.  Once I add </dev/null, it stops somewhere in the named script (a
"ping the hotst" script run before named executes, one run after named
does not).
	Partway through that script, fd0 is completely closed on all uml
threads.  I can watch it with:

while /bin/true ; do clear ; for X in $(ps axf | grep [l]inux-2.4.0-test10
| awk '{print $1}') ; do ls -al /proc/$X/fd/0 ; done ; sleep 1 ; done

	Up until named, I get:

lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25751/fd/0 -> /dev/null
lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25752/fd/0 -> /dev/null
lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25763/fd/0 -> /dev/null
...
lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25865/fd/0 -> /dev/null
lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25884/fd/0 -> /dev/null
lr-x------    1 wstearns wstearns       64 Nov  7 14:31 /proc/25920/fd/0 -> /dev/null

	with the odd "No such file or directory" from the delay between
looking up the pids and actually taking a dir.  Once the named script gets
underway, all threads no longer have stdin open:

ls: /proc/26443/fd/0: No such file or directory
ls: /proc/25751/fd/0: No such file or directory
ls: /proc/25752/fd/0: No such file or directory
ls: /proc/25763/fd/0: No such file or directory
ls: /proc/25768/fd/0: No such file or directory
ls: /proc/25769/fd/0: No such file or directory
ls: /proc/25764/fd/0: No such file or directory
ls: /proc/25765/fd/0: No such file or directory
ls: /proc/25766/fd/0: No such file or directory
ls: /proc/25767/fd/0: No such file or directory
ls: /proc/25795/fd/0: No such file or directory
ls: /proc/25828/fd/0: No such file or directory
ls: /proc/25884/fd/0: No such file or directory
ls: /proc/25920/fd/0: No such file or directory
ls: /proc/26074/fd/0: No such file or directory
ls: /proc/26266/fd/0: No such file or directory
ls: /proc/26319/fd/0: No such file or directory
ls: /proc/26330/fd/0: No such file or directory
ls: /proc/26345/fd/0: No such file or directory
ls: /proc/26346/fd/0: No such file or directory
ls: /proc/26347/fd/0: No such file or directory
ls: /proc/26348/fd/0: No such file or directory
ls: /proc/26349/fd/0: No such file or directory
ls: /proc/26365/fd/0: No such file or directory
ls: /proc/26381/fd/0: No such file or directory
ls: /proc/26385/fd/0: No such file or directory
ls: /proc/26397/fd/0: No such file or directory
ls: /proc/26398/fd/0: No such file or directory

	For the brief second between when the network is brought up and
when fd0 is closed, I can ping uml from the host.  Once fd0 is closed, I
can't anymore.  I'll guess that named isn't itself the problem; it just
happens to be executing and trying to do outside lookups when the fd is
closed:

tcpdump: listening on tap1
14:47:51.592086 B arp who-has uml-router tell 192.168.210.141
14:47:51.592101 > arp reply uml-router (fe:fd:0:0:0:0) is-at fe:fd:0:0:0:0 (fe:fd:c0:a8:d2:8d)
14:47:51.597167 < 192.168.210.141.1024 > a.root-servers.net.domain: 31496 NS? . (17) (DF)
14:47:51.842545 > a.root-servers.net.domain > 192.168.210.141.1024: 31496*- 13/0/13 NS A.ROOT-SERVERS.NET., NS H.ROOT-SERVERS.NET., NS C.ROOT-SERVERS.NET., NS G.ROOT-SERVERS.NET., NS F.ROOT-SERVERS.NET., (436)
14:47:55.006502 < 192.168.210.141.1024 > d.root-servers.net.domain: 31496 NS? . (17) (DF)
14:47:55.532558 > d.root-servers.net.domain > 192.168.210.141.1024: 31496*- 13/0/13 NS C.ROOT-SERVERS.NET., NS G.ROOT-SERVERS.NET., NS F.ROOT-SERVERS.NET., NS B.ROOT-SERVERS.NET., NS J.ROOT-SERVERS.NET., (436)
14:47:59.007525 < 192.168.210.141.1024 > k.root-servers.net.domain: 31496 NS? . (17) (DF)
14:47:59.302574 > k.root-servers.net.domain > 192.168.210.141.1024: 31496*- 13/0/13 NS K.ROOT-SERVERS.NET., NS L.ROOT-SERVERS.NET., NS M.ROOT-SERVERS.NET., NS I.ROOT-SERVERS.NET., NS E.ROOT-SERVERS.NET., (436)

	It looks like the responses are being passed back to the uml
correctly.  (Note that the host is doing Masq/NAT for the uml).

	I'm not sure if these are all just symptoms of some other problem
or even if a closed fd0 matters.  I'm bringing it up on the off chance
that close is being called accidentally in one of the uml threads, or
being called correctly but with an incorrect parameter of 0.

	Just for reference, here are the fd's for the parent tracing
thread once it's stopped at named:

[root@sparrow rh6.2]# ls -al /proc/29904/fd
total 8
dr-x------    2 wstearns wstearns        0 Nov  7 14:45 .
dr-xr-xr-x    3 wstearns wstearns        0 Nov  7 14:45 ..
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 1 -> /dev/pts/8
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 10 -> /home/wstearns/uml/rh6.2/root_fs
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 11 -> socket:[12210072]
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 2 -> /dev/pts/8
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 3 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 4 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 5 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 6 -> /home/wstearns/uml/rh6.2/vm_file (deleted)
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 7 -> /dev/ptyp0
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 8 -> /dev/ptmx
lrwx------    1 wstearns wstearns       64 Nov  7 14:45 9 -> /dev/pts/20

	Cheers,
	- Bill 'eating red herrings again for dinner?' Stearns

---------------------------------------------------------------------------
	"Assuming that piece of ear weighed about 1/2 an ounce, Bill
[Gates] could afford to eat 443.39 pounds of Evander Holyfield if he were
so inclined." 
	-=- http://web.quuxuum.org/~evan/bgnw.html
(Courtesy of James Mastros <ro...@je...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

[uml-devel] Re: Makefile clean up gcc -> $(CC)

[Jeff D. <jd...@ka...>]

da...@th... said:
> > The one thing to look for is the ability to intercept and 
> > modify a process system calls.
> I knew about that. In fact that is the least of my worries as it turns
> out.

Actually, it's the most of your worries.  If you can't do that, then UML just 
won't work.

> Well, other gotchas include the fact that memory access exceptions are
> delivered via mach ports, signals are delivered on the main/first
> thread of a process only, and the kernel's reliance of being compile
> into an ELF  executable (Mac OS X is Mach-O.)

All these things can probably be worked around.  But if you can't intercept 
system calls, there's no point in even thinking about them.

				Jeff

[uml-devel] Re: Makefile clean up gcc -> $(CC)

[Dave Z. <da...@th...>]

On Mon, 6 Nov 2000, Jeff Dike wrote:

> You really need to look at the OS before trying to port UML to it.  It's not a 
> simple recompile.  The one thing to look for is the ability to intercept and 
> modify a process system calls.

I knew about that. In fact that is the least of my worries as it turns
out.

> If that's there, then probably everything else can be made to work.

Well, other gotchas include the fact that memory access exceptions are
delivered via mach ports, signals are delivered on the main/first thread
of a process only, and the kernel's reliance of being compile into an ELF 
executable (Mac OS X is Mach-O.)

There are probably more gotchas, but I don't remember them off hand.

davez

-- 
Dave Zarzycki
http://thor.sbay.org/~dave/

Re: [uml-devel] Makefile clean up gcc -> $(CC)

[Jeff D. <jd...@ka...>]

da...@th... said:
> I got bored and tried building UML on Mac OS X
> The more I got into it, the more I learned that this we be a less than
>  trivial port.

You really need to look at the OS before trying to port UML to it.  It's not a 
simple recompile.  The one thing to look for is the ability to intercept and 
modify a process system calls.  If that's there, then probably everything else 
can be made to work.

> Please consider applying the attached patch.

Applied.  Thanks.

				Jeff

[uml-devel] Makefile clean up gcc -> $(CC)

[Dave Z. <da...@th...>]

I got bored and tried building UML on Mac OS X, the first thing I needed
to fix was changing every instance of gcc to $(CC), since the the Mac OS X
version of gcc is called cc. Please consider applying the attached patch.
It's also useful for those people that have Linux distributions where gcc
is not the preferred compiler (for example, kgcc on Red Hat Linux 7.0.)

For the curious individuals out there, I got a few files to build before I
gave up (for now.) Maybe I'll try again later. The more I got into it, the
more I learned that this we be a less than trivial port.

davez

-- 
Dave Zarzycki
http://thor.sbay.org/~dave/

[User-mode-linux-devel] test10 checkins

[Jeff D. <jd...@ka...>]

This batch updates the pool to 2.4.0-test10.

There is the start of a hostfs filesystem.  It is a virtual filesystem that 
provides access to the host root filesystem.  You can mount it and cd into it, 
but when you run ls, the kernel will panic.

printk was fixed so that it no longer goes into an infinite recursion when the 
main console is not a terminal (or non-existent).

I broke the ability to mount host devices inside UML a while ago.  That is 
fixed again.

I fixed a race which occasionally caused strange kernel memory faults.

				Jeff

Re: [User-mode-linux-devel] *Unofficial* patch from 10/30 cvs and test10-pre7

[Jeff D. <jd...@ka...>]

wst...@po... said:
> tlb.c:73: structure has no member named `pte'
> tlb.c:73: structure has no member named `pte' 

Change that line to look like:
	if(opte->pte_low != npte->pte_low){

				Jeff

Re: [User-mode-linux-devel] *Unofficial* patch from 10/30 cvs and test10-pre7

[William S. <wst...@po...>]

Good morning, Jeff, all,

On Tue, 31 Oct 2000, Jeff Dike wrote:

> > memory.c: In function `copy_page_range':
> > memory.c:230: warning: implicit declaration of function
> > `ptep_clear_wrprotect'
> 
> Put 
> 
> #include <asm-generic/pgtable.h>
> 
> at the bottom of asm-um/pgtable.h.  That should clear up the missing 
> declarations.  And it might fix the others as well.

	Bingo!

> This is just from staring at the patch and the code being patched, YMMV...
> 
> > blk.h from cvs had a "#define DEVICE_INTR do_ida" under compaq_smart2
> > that didn't seem to have relevance for uml and wasn't even mentioned
> > in 2.4.0-test1 through 2.4.0-test10-pre7.
> 
> Oops.  I guess I wasn't paying enough attention :-)

	No biggie.  The only changed files in this patch from the previous
one are:
	blk.h: useless do_ida removed
	pgtable.h: included asm-generic/pgtable.h
	param.h: CLOCKS_PER_SEC defined to be 20 - correct?

> > 	Here's hoping this saves Jeff a little work on the official merge. 
> 
> Let me know if it boots :-)

	It's getting a lot closer:

gcc -D__KERNEL__ -I/usr/src/uml-linux/linux-2.4.0-test10-pre7.uml/include
-Wall -Wstrict-prototypes -O2  -fno-strict-aliasing -g  -U__i386__ -Ui386
-D__arch_um__ -DSUBARCH="i386" -DNESTING=0  -I../include  -c -o tlb.o tlb.c
tlb.c: In function `fix_range':
tlb.c:73: structure has no member named `pte'
tlb.c:73: structure has no member named `pte'
tlb.c:75: warning: passing arg 2 of `map' makes integer from pointer without a cast
tlb.c:77: warning: passing arg 1 of `munmap' makes pointer from integer without a cast
tlb.c:78: warning: passing arg 2 of `map' makes integer from pointer without a cast
tlb.c:86: warning: passing arg 1 of `munmap' makes pointer from integer without a cast
tlb.c:92: warning: passing arg 1 of `munmap' makes pointer from integer without a cast
tlb.c: In function `init_flush_vm':
tlb.c:128: warning: passing arg 2 of `map' makes integer from pointer without a cast
tlb.c: In function `add_process_vmas':
tlb.c:260: warning: implicit declaration of function `malloc'
tlb.c:260: warning: assignment makes pointer from integer without a cast
make[1]: *** [tlb.o] Error 1
make[1]: Leaving directory `/usr/src/uml-linux/linux-2.4.0-test10-pre7.uml/arch/um/kernel'
make: ***
[_dir_/usr/src/uml-linux/linux-2.4.0-test10-pre7.uml/arch/um/kernel] Error 2

	Cheers,
	- Bill

---------------------------------------------------------------------------
	Q: Is it possible to set up masquerading timeouts that TCP 
connection never expires even if there are no any packets traveling?
	A: Sure.
	# ipchains -M -S 13564800 0 0
	That'll last you up until January 1, 2000, and after the rioting
will start and you won't have to worry about masquerading any more. 8-)
	-- Paul Rusty Russell <Pau...@ru...>
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

Re: [User-mode-linux-devel] *Unofficial* patch from 10/30 cvs and test10-pre7

[Jeff D. <jd...@ka...>]

> memory.c: In function `copy_page_range':
> memory.c:230: warning: implicit declaration of function
> `ptep_clear_wrprotect'

[snip]

> make[1]: *** [first_rule] Error 2
> make[1]: Leaving directory
> `/usr/src/uml-linux/linux-2.4.0-test10-pre7.uml.orig/mm'
> make: *** [_dir_mm] Error 2\

Put 

#include <asm-generic/pgtable.h>

at the bottom of asm-um/pgtable.h.  That should clear up the missing 
declarations.  And it might fix the others as well.

This is just from staring at the patch and the code being patched, YMMV...

> blk.h from cvs had a "#define DEVICE_INTR do_ida" under compaq_smart2
> that didn't seem to have relevance for uml and wasn't even mentioned
> in 2.4.0-test1 through 2.4.0-test10-pre7.

Oops.  I guess I wasn't paying enough attention :-)

> 	Here's hoping this saves Jeff a little work on the official merge. 

Let me know if it boots :-)

				Jeff

[User-mode-linux-devel] *Unofficial* patch from 10/30 cvs and test10-pre7

[William S. <wst...@po...>]

Good evening, all,
	This is an unofficial attempt to merge the cvs tree as of 2:30 EST
today with the test10-pre7 patch.
	Only one section didn't quite make sense; blk.h from cvs had a
"#define DEVICE_INTR do_ida" under compaq_smart2 that didn't seem to have
relevance for uml and wasn't even mentioned in 2.4.0-test1 through
2.4.0-test10-pre7.  Considering UML doesn't even do hardware, this
definitely falls in the who-gives-a-darn category.  :-)

[root@sparrow linux]# pwd
/usr/src/uml-linux/linux-2.4.0-test10-pre7.uml/include/linux
[root@sparrow linux]# diff -ud blk.h.uml blk.h | less
--- blk.h.uml   Tue Oct  3 20:09:20 2000
+++ blk.h       Mon Oct  2 14:02:34 2000
@@ -314,19 +314,9 @@
 #elif (MAJOR_NR == COMPAQ_SMART2_MAJOR)
 
 #define DEVICE_NAME "ida"
-#define DEVICE_INTR do_ida
 #define TIMEOUT_VALUE (25*HZ)
 #define DEVICE_REQUEST do_ida_request0
 #define DEVICE_NR(device) (MINOR(device) >> 4)
-
-#elif (MAJOR_NR == UBD_MAJOR)
-
-#define DEVICE_NAME "User-mode block device"
-#define DEVICE_INTR do_ubd
-#define DEVICE_REQUEST do_ubd_request
-#define DEVICE_NR(device) (MINOR(device))
-#define DEVICE_ON(device)
-#define DEVICE_OFF(device)
 
 #endif /* MAJOR_NR == whatever */
 
	Here's hoping this saves Jeff a little work on the official merge.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"Love is a snowmobile racing across the tundra and then suddenly it
flips over, pinning you underneath.  At night, the ice weasels come."
	-- Matt Groening
(Courtesy of Steve Dodd <di...@lo...>)
--------------------------------------------------------------------------
William Stearns (wst...@po...).  Mason, Buildkernel, named2hosts, 
and ipfwadm2ipchains are at:                http://www.pobox.com/~wstearns
LinuxMonth; articles for Linux Enthusiasts! http://www.linuxmonth.com
--------------------------------------------------------------------------

Re: [User-mode-linux-devel] Re: LINE 0.1 available

[Jeff D. <jd...@ka...>]

epa...@cs... said:
> So ideally, a UML port to NT would be with VC  (which I realize would
> be a tremendous pain in the ass)

The kernel parts of UML, at least, would have to be compiled with gcc since 
it's very gcc-specific in places.

				Jeff

Re: [User-mode-linux-devel] Re: LINE 0.1 available

[Michael V. <mi...@me...>]

On Thu, 19 Oct 2000, Erik Paulson wrote:

> Cygwin requires that all the cygwin dll's be installed on the executing 
> machine though, right? So ideally, a UML port to NT would be with VC 
> (which I realize would be a tremendous pain in the ass)

I think the only DLL you really need is cygwin.dll, which just needs to be
located in the same directory as the executable.  

However you only need the cygwin dll's if you are using the Cygwin
libraries.  You can use Cygwin to build normal Win32 applications that do
not depend on cygwin.dll at all.

         Mike

Re: [User-mode-linux-devel] Re: LINE 0.1 available

[Erik P. <epa...@cs...>]

On Thu, Oct 19, 2000 at 05:57:09PM -0400, Michael Vines wrote:
> On Thu, 19 Oct 2000, Jeff Dike wrote:
> 
> > Very cool.  I'll fiddle with my site some to add pointers to your stuff.
> 
> Thanks
> 
> > I'm curious as to why, once you had decided to do this, you didn't just link 
> > the kernel in to get the real system calls.  That would have resulted in a 
> > completely independent implementation of UML...  On Windows, no less...
> 
> That's sortof what I was working towards when I got swamped with real work
> and had to stop.  You can see that with the source for 'version2', how
> it's more laid out like the real kernel source.  But the fact that the
> kernel is quite gcc specific was a real pain considering I was using VC++. 
> I know cygwin would have been the logical choice, but I prefer the VC++
> debugger over gdb :)

Cygwin requires that all the cygwin dll's be installed on the executing 
machine though, right? So ideally, a UML port to NT would be with VC 
(which I realize would be a tremendous pain in the ass)

-Erik

Re: [User-mode-linux-devel] Re: LINE 0.1 available

[Michael V. <mi...@me...>]

On Thu, 19 Oct 2000, Jeff Dike wrote:

> Very cool.  I'll fiddle with my site some to add pointers to your stuff.

Thanks

> I'm curious as to why, once you had decided to do this, you didn't just link 
> the kernel in to get the real system calls.  That would have resulted in a 
> completely independent implementation of UML...  On Windows, no less...

That's sortof what I was working towards when I got swamped with real work
and had to stop.  You can see that with the source for 'version2', how
it's more laid out like the real kernel source.  But the fact that the
kernel is quite gcc specific was a real pain considering I was using VC++. 
I know cygwin would have been the logical choice, but I prefer the VC++
debugger over gdb :)

> In retrospect, I'm glad I didn't get slasdotted until last week. 

I was actually really surprised by how long it took.  You should have been
slashdotted half a year ago IMO

         Mike

[User-mode-linux-devel] Re: LINE 0.1 available

[Jeff D. <jd...@ka...>]

mi...@me... said:
> I've made the Linux application runner utility for Windows that I was
> working on over the summer available on the web.

Very cool.  I'll fiddle with my site some to add pointers to your stuff.

> The basic difference between them is that 'Version1' executes the
> syscalls in the Linux application's process space and 'Version2'
> executes the syscalls in the monitor's process space.

With UML, it has to be in the app.  You had only one thread, so you could do 
syscalls wherever you wanted.  The UML tracing thread has to serve all threads 
in the kernel, so when a syscall sleeps, it has to be in the app.

> Only a very small subset of the Linux syscalls have actually been
> implemented and the ones that have been implemented are buggy!

That's OK.  I've already got the system calls covered :-)

I'm curious as to why, once you had decided to do this, you didn't just link 
the kernel in to get the real system calls.  That would have resulted in a 
completely independent implementation of UML...  On Windows, no less...

> It's
> somewhat embarrassing releasing code in such a pre-alpha condition :P

You should have seen some early versions of UML...

In retrospect, I'm glad I didn't get slasdotted until last week.  Much 
earlier, and a lot of people could have been turned off by bugs, things that 
didn't work, and things that weren't there.  Now, the bugs aren't that 
noticable, there aren't that many missing things, and I can claim that those 
are on the way.  The one downside of having it happen last week was that was 
the middle of the stack overflow thing, but most slashdotters are probably 
running 2.2, and didn't notice that.

				Jeff

[User-mode-linux-devel] LINE 0.1 available

[Michael V. <mi...@me...>]

Hi Jeff,

I've made the Linux application runner utility for Windows that I was
working on over the summer available on the web.  The cleanup took much
less time than I was expecting...mainly because I spent all Wednesday
working it :)

It has been released under the GPL.

I actually found two completely different versions of the utility on my
filesystem so I included them both because I though it might be
interesting to contrast the two.  The basic difference between them is
that 'Version1' executes the syscalls in the Linux application's process
space and 'Version2' executes the syscalls in the monitor's process space.

Only a very small subset of the Linux syscalls have actually been 
implemented and the ones that have been implemented are buggy!  It's
somewhat embarrassing releasing code in such a pre-alpha condition :P

The URL is http://neomueller.org/~isamu/line/

         Mike