user-mode-linux-user

Re: [User-mode-linux-user] FeedBack on latest kernel

[James M. <mac...@ED...>]

On Thu, 5 Oct 2000, Jeff Dike wrote:

> mac...@ed... said:
> > if it even acknowledges the enter ;). It makes logging in a very
> > challenging experience :). 
> So, it happens so often you can't log in...

Yes, the password is hard to tell if you get a character or not :(.

> Does it happen only on the main console?  Have you checked it on a virtual 
> console, the serial line, or in a telnet session?

Happens on ssh logins too. I believe it happend on other virtuals, but I
have not tested that way lately.

> Can you tell me exactly how you're running UML?

In a directory I have a small root filesystem that was basically a boot
disk with vi and ping added and some extra space to see if I could get
apache on it :).

I do not have experience with devfs so I mknod the ptys that ssh wants
after the boot, but I know the problem was there before that :).

I do the ./um_.... eth0 100 and then ./linux and up she comes. I have only
one virtual screen, as I prefer to play via ssh so I can rsync stuff
around.

> As for that networking thing, have you tried both interfaces to see if they 
> both have the problem?

You mean the umn method? Not from home because I did not have slip in the
kernel. I will see about getting something there to test that.

> 				Jeff
> 

thanks again,
JES
--
James B. MacLean        mac...@ed...
Department of Education http://www.ednet.ns.ca/~macleajb
Nova Scotia, Canada
B3M 4B2
     

Re: [User-mode-linux-user] FeedBack on latest kernel

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

mac...@ed... said:
> if it even acknowledges the enter ;). It makes logging in a very
> challenging experience :).

So, it happens so often you can't log in...

Does it happen only on the main console?  Have you checked it on a virtual 
console, the serial line, or in a telnet session?

Can you tell me exactly how you're running UML?

As for that networking thing, have you tried both interfaces to see if they 
both have the problem?

				Jeff

Re: [User-mode-linux-user] FeedBack on latest kernel

[James M. <mac...@ED...>]

On Wed, 4 Oct 2000, Jeff Dike wrote:

> mac...@ED... said:
> > > Does the shell act like it sees the missing characters?
> > I do not understand.
> If you type "abcdef" and see "abdf" on the screen, does bash say "abcdef: 
> command not found" or "abdf: command not found"?

It would be command not found... if it even acknowledges the enter ;). It
makes logging in a very challenging experience :).
 
> > I even tried the tap0 approach. tcpdump on the physical sees the
> > traffic but the physical kernel/TCP/IP does not apply, act on it.
> I'd suspect something's going on with UML, but if traffic is disappearing into 
> the host networking and disappearing, that doesn't give me much to go on.

Agreed. And the fact that it works differently on a different machine
is odd. 2 of the 3 machines at work were Mandrake installs, but one was
RedHat like here at home, so I can't even suggest that :(.

How can tcpdump see it, but the kernel not? I can understand it on
interfaces where you inject it after the TCP/IP stack where only the
wire/tapX gets to deal with it, but since it was working at work I don't
see how it works differently here.

I remember something in the tap devices where if you ran tcpdump on the
tap, the data did not go any further than to see it on tcpdump, then if
you stopped tcpdump, it would then go properly. I believe somebody added
something about copying some struct in the kernel for each thing using the
tap so that they all got to react and not just the first one in line.
Likely not related, but mentioned anyhow :).

cheers,
JES
--
James B. MacLean        mac...@ed...
Department of Education http://www.ednet.ns.ca/~macleajb
Nova Scotia, Canada
B3M 4B2
     

Re: [User-mode-linux-user] FeedBack on latest kernel

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

ri...@co... said:
> Btw, how "efficient" is networking between different UMLs within the
> same box? ;)

Probably depends on the interface you use.  The old umn device uses the host 
kernel as the router.  The new eth device uses a usermode daemon, but those 
packets also go through the kernel, so I'd guess there's slightly more 
overhead.

Aside from the virtuality overhead, which is there on UML <-> anything 
networking, it's basically fancy IPC, which Linux ought to be pretty good at.

				Jeff

Re: [User-mode-linux-user] FeedBack on latest kernel

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

mac...@ED... said:
> > Does the shell act like it sees the missing characters?
> I do not understand.

If you type "abcdef" and see "abdf" on the screen, does bash say "abcdef: 
command not found" or "abdf: command not found"?

> I even tried the tap0 approach. tcpdump on the physical sees the
> traffic but the physical kernel/TCP/IP does not apply, act on it.

I'd suspect something's going on with UML, but if traffic is disappearing into 
the host networking and disappearing, that doesn't give me much to go on.

				Jeff

Re: [User-mode-linux-user] Kernel panic with 2.4.0-test9

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

tec...@li... said:
> So, unless I can find test8 somewhere (it's not on the sourceforge
> archive, although I've not looked at CVS yet).

You can check it out of cvs with the tag 'v_2_4_0_test8'.

I just tried resurrecting it on the project download page (which used to be 
possible), but it doesn't look like that works any more.

				Jef

Re: [User-mode-linux-user] Kernel panic with 2.4.0-test9

[David C. <tec...@li...>]

On Wed, 4 Oct 2000, Jeff Dike wrote:
> Other people are complaining about the same thing.  I'd just prefer just to
> consider that they're nuts and ignore them, but they now outnumber me and they
> might beat me up if I did that.

Heh, well, 2.4.0-test2 and 4 work fine, but I'd like to be able to use the
virtual ethernet thingy. So, unless I can find test8 somewhere (it's not
on the sourceforge archive, although I've not looked at CVS yet).

> So, I'm looking either to reproduce it or for someone to send me a stack trace.

Hrm, okay, I'll do that tomorrow.

-- 
David Coulson
tec...@so...

Re: [User-mode-linux-user] FeedBack on latest kernel

[James M. <mac...@ED...>]

On Wed, 4 Oct 2000, Rik van Riel wrote:

> On Wed, 4 Oct 2000, Jeff Dike wrote:
> > mac...@ed... said:
> > > My machine at work (P600) seems to act still like it is missing
> > > signals from the ttys/0 that I boot on. Appears as missed keystrokes.
> > So, for example, you could type "abcdef" and on the screen would appear "abdf"?
> > Does the shell act like it sees the missing characters?
> I've seen this happen when I tested UML yesterday ...

So far on all boxes at work I've seen it and on all the kernels I tried,
yet at home, it has not appeared, or else I just miss it :).
 
> 	[UML networking]
> Hmmm, I will need to play with this. Btw, how "efficient"
> is networking between different UMLs within the same box? ;)

Well, bursty would be a word. Sometimes real good, other times it seems to
stall. It's like signals are not getting through (like the keyboard I
guess :).
 
> regards,
> Rik

take care,
JES
--
James B. MacLean        mac...@ed...
Department of Education http://www.ednet.ns.ca/~macleajb
Nova Scotia, Canada
B3M 4B2
     

Re: [User-mode-linux-user] FeedBack on latest kernel

[James M. <mac...@ED...>]

On Wed, 4 Oct 2000, Jeff Dike wrote:

> mac...@ed... said:
> > My machine at work (P600) seems to act still like it is missing
> > signals from the ttys/0 that I boot on. Appears as missed keystrokes. 
> So, for example, you could type "abcdef" and on the screen would appear "abdf"?

Yes. In its worse state you can type for a while and nothing, then
suddenly it begins again, but misses all the in  between strokes.

> Does the shell act like it sees the missing characters?

I do not understand. The modem lights flicker (I'm at home connecting to
work), but I see nothing in the screen to suggest it is getting anything.

> > Tcpdump on home box shows arp replies from UML, but for some reason
> > they are not picked up by the physical kernel.
> Can you see anything that the UML is doing wrong?

Not by my looking. I can basically get at all boxes other than the
physical box that are on the local network (no need for forwarding). Yet
forwarding is on in both physical and UML, and I am already using this box
as a gateway for other traffic.

I tried multiple ip's with no luck. I even tried the tap0 approach.
tcpdump on the physical sees the traffic but the physical kernel/TCP/IP
does not apply, act on it. Even when I manually add the arp into the
table, it just acts as if there is no traffic for it to see.

> > An attempted rsync to my work box of a few programs resulted in: Stack
> > overflowed onto current_task page In interrupt handler - not syncing
> > which resulted in me killing the linux processes.
> I need a stack trace of this.  See the latest traffic on the devel list for 
> how to get one.

> 				Jeff

Saw about 6 messages. I will look into getting that done. Thanks,
JES
--
James B. MacLean        mac...@ed...
Department of Education http://www.ednet.ns.ca/~macleajb
Nova Scotia, Canada
B3M 4B2
     

Re: [User-mode-linux-user] FeedBack on latest kernel

[Rik v. R. <ri...@co...>]

On Wed, 4 Oct 2000, Jeff Dike wrote:
> mac...@ed... said:
> > My machine at work (P600) seems to act still like it is missing
> > signals from the ttys/0 that I boot on. Appears as missed keystrokes.
> 
> So, for example, you could type "abcdef" and on the screen would appear "abdf"?
> 
> Does the shell act like it sees the missing characters?

I've seen this happen when I tested UML yesterday ...

	[UML networking]

Hmmm, I will need to play with this. Btw, how "efficient"
is networking between different UMLs within the same box? ;)

regards,

Rik
--
"What you're running that piece of shit Gnome?!?!"
       -- Miguel de Icaza, UKUUG 2000

http://www.conectiva.com/		http://www.surriel.com/

Re: [User-mode-linux-user] Kernel panic with 2.4.0-test9

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

tec...@li... said:
> Kernel panic: Stack overflowed onto current_task page In interrupt
> handler - not syncing 

Other people are complaining about the same thing.  I'd just prefer just to 
consider that they're nuts and ignore them, but they now outnumber me and they 
might beat me up if I did that.

So, I'm looking either to reproduce it or for someone to send me a stack trace.

				Jeff

Re: [User-mode-linux-user] FeedBack on latest kernel

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

mac...@ed... said:
> My machine at work (P600) seems to act still like it is missing
> signals from the ttys/0 that I boot on. Appears as missed keystrokes.

So, for example, you could type "abcdef" and on the screen would appear "abdf"?

Does the shell act like it sees the missing characters?

> Tcpdump on home box shows arp replies from UML, but for some reason
> they are not picked up by the physical kernel.

Can you see anything that the UML is doing wrong?

> An attempted rsync to my work box of a few programs resulted in: Stack
> overflowed onto current_task page In interrupt handler - not syncing
> which resulted in me killing the linux processes.

I need a stack trace of this.  See the latest traffic on the devel list for 
how to get one.

				Jeff

[User-mode-linux-user] FeedBack on latest kernel

[James M. <mac...@ED...>]

Hi Folks,

Newbie with UML, but this is the third kernel that I have downloaded and
tried from SourceForge and wanted to give some feedback.

. My machine at work (P600) seems to act still like it is missing signals
from the ttys/0 that I boot on. Appears as missed keystrokes. My home box
(P166) does not show this behaviour. When I started the P600 (work) UML in
an ssh session, it even seemed to be sending back characters almost like
when my modem is at the wrong speed :).

. I can ping my UML from my physical machine at work (P600), but on my
home box (P166) arp does not get filled and ping fails using the um
networking. I can connect from my physical to the UML at work, and not
from my physical to my UML at home. Tcpdump on home box shows arp replies
from UML, but for some reason they are not picked up by the physical
kernel.

. Both machines will allow connections from other machines (I am using the
virtual eth0 method)

. An attempted rsync to my work box of a few programs resulted in:
Stack overflowed onto current_task page In interrupt handler - not syncing
which resulted in me killing the linux processes.

My physical boxes are running 2.2.17pre19/20.

I enjoy seeing this project bearing fruit. My boot is very fast for UML
even on the P166.

Hope this is of use,
JES
--
James B. MacLean        mac...@ed...
Department of Education http://www.ednet.ns.ca/~macleajb
Nova Scotia, Canada
B3M 4B2
     

[User-mode-linux-user] Kernel panic with 2.4.0-test9

[David C. <tec...@li...>]

When trying to start 2.4.0-test9 with the root_fs_tomrtbt_1.7.205 image I
get the following;

VFS: Mounted root (ext2 filesystem) readonly.
Mounted devfs on /dev
INIT: version 2.60 booting
Kernel panic: Stack overflowed onto current_task page
In interrupt handler - not syncing
Kernel panic: Stack overflowed onto current_task page
In interrupt handler - not syncing

Host is running 2.4.0-test9 with SMP support. Occasionally it gets
further and starts some of the init.d stuff, then it does the same thing.

Any ideas?

-- 
David Coulson
tec...@so...

[User-mode-linux-user] user-mode port 0.32-2.4.0-test9

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

The user-mode port of 2.4.0-test9 is available.

The bug that caused bash to occasionally segfault on address 0 was fixed.

I also went on a breakpoint-fixing binge.  The problems caused by the kernel 
debugger hitting breakpoints are fixed.  Also, gdb no longer panics the kernel 
when it hits a breakpoint.  Both the kernel debugger and gdb now seem to work 
well.

Bill Stearns produced a number of new bootable filesystems, including Red Hat 
7.0, Mandrake 7.1, and Immunix 6.2.  These are available from the project's 
download page.

The project's home page is http://user-mode-linux.sourceforge.net

The project's download page is http://sourceforge.net/project/filelist.php?grou
p_id=429

				Jeff

[User-mode-linux-user] UML will work on test9 and later hosts

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

I'm going through the patch, and I noticed that the ptrace patch that went 
into test8 has been backed out.

It would have been nice to know exactly what that fix broke, though.

				Jeff

Re: [User-mode-linux-user] Ptrace broken since 2.4.0-test8pre4?...

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

On Tue, 19 Sep 2000, Jeff Dike wrote:

> yu...@as... said:
> > I tested vanilla test7 with  ptrace() patch. It breaks uml exactly
> > like I see with any kernel > test7. 
> 
> > exec_user.c:29 ptrace(PTRACE_SYSCALL, 4901, 0, 0) = 0
> > And voila, we got SIGSEGV instead of happy running child:
> > Child 4901 exited with signal 11 
> 
> Yuri, I apologize for doubting you.  That's about the best evidence possible 
> that the test8 ptrace fix is what's breaking UML and not the thread group 
> stuff.
> 
> If anyone has any idea why that fix breaks PTRACE_SYSCALL, I'd love to know.  
> I'm kind of short on time at the moment, so I'm not in a great position to 
> delve into this at the moment.
> 
> Until this is fixed, UML doesn't work on any host kernels beyond test7.

	As a further verification that this change is causing problems for
UML, test9-pre5 allows uml to run if the ptrace.c change is
reverted.  I do understand that this is not the correct fix for the
problem, but for those who would like to independently verify this, the
(incorrect) patch is attached.

	Cheers,
	- Bill

---------------------------------------------------------------------------
	Things you Do Not Want To See On IRC: your husband commenting on the
S390 port and in the next comment, announcing that he expects a new toy.
He tells me the two are unrelated. I do hope so. 
	- Telsa Gwynn, Alan Cox' wife
--------------------------------------------------------------------------
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-user] Ptrace broken since 2.4.0-test8pre4?...

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

yu...@as... said:
> I tested vanilla test7 with  ptrace() patch. It breaks uml exactly
> like I see with any kernel > test7. 

> exec_user.c:29 ptrace(PTRACE_SYSCALL, 4901, 0, 0) = 0
> And voila, we got SIGSEGV instead of happy running child:
> Child 4901 exited with signal 11 

Yuri, I apologize for doubting you.  That's about the best evidence possible 
that the test8 ptrace fix is what's breaking UML and not the thread group 
stuff.

If anyone has any idea why that fix breaks PTRACE_SYSCALL, I'd love to know.  
I'm kind of short on time at the moment, so I'm not in a great position to 
delve into this at the moment.

Until this is fixed, UML doesn't work on any host kernels beyond test7.

				Jeff

Re: [User-mode-linux-user] Ptrace broken since 2.4.0-test8pre4?...

[Yuri P. <yu...@as...>]

Hello Jeff,

I tested vanilla test7 with  ptrace() patch.
It breaks uml exactly like I see with any kernel > test7.

Seems like the ORIG_EAX != -1 is needed  to correctly restart syscall
after PTRACE_SYSCALL, but I did not check this codepath thoroughly.

Following what is going with uml, just for the very first exec:

[lots of trap_user.c:228, trap_user.c:149 skipped]

start_syscall():
process.c:181 ptrace(PTRACE_GETREGS, 4901, 0, 501242c0) = 0

finish_exec():
exec_user.c:26 ptrace(PTRACE_POKEUSER, 4901, UESP, bf7fff20) = 0
exec_user.c:27 ptrace(PTRACE_POKEUSER, 4901, EIP, 400019a0) = 0
exec_user.c:28 ptrace(PTRACE_POKEUSER, 4901, EDX, 0) = 0
exec_user.c:29 ptrace(PTRACE_SYSCALL, 4901, 0, 0) = 0

And voila, we got SIGSEGV instead of happy running child:
Child 4901 exited with signal 11

> Have you tried putting that ptrace patch into test7 and seeing if UML stops
> working with that as the host?
>
> test8 is a known bad platform for UML, so I'm going to consider that the
> ptrace patch is OK unless it's possible to turn a good host into a bad one by
> adding that ptrace fix.


--- v2.4.0-test7/linux/arch/i386/kernel/ptrace.c Fri Jun 23 21:55:07 2000
+++ linux/arch/i386/kernel/ptrace.c Sat Sep 2 12:00:02 2000
@@ -99,6 +99,11 @@
                case EFL:
                        value &= FLAG_MASK;
                        value |= get_stack_long(child, EFL_OFFSET) & ~FLAG_MASK;
+                       break;
+               case EIP:
+                       /* Mark us as not being in a system call, so that no restart issues happen
*/
+                       put_stack_long(child, 4*ORIG_EAX - sizeof(struct pt_regs), -1);
+                       break;
        }
        if (regno > GS*4)
                regno -= 2*4;

Re: [User-mode-linux-user] Ptrace broken since 2.4.0-test8pre4?...

[Yuri P. <yu...@as...>]

Jeff Dike wrote:

> yu...@as... said:
> > Beeing an active user mode linux user :-) I can say that since
> > 2.4.0-test8 (host kernel) I cannot run uml-linux successfully.
>
> > In contrast with popular feeling that "threaded programes screwed
> > signal handling on test8.", it is actually a  small change to arch/
> > i386/ptrace.c introduced since test8pre4.
>
> I don't see a problem with the ptrace fix.  Before changing EIP, UML saves all
> of the process registers into the thread struct.  So that change to ORIG_EAX
> shouldn't matter as far as UML is concerned.
>
>                                 Jeff

But it does matter, one of user-process threads gets SEGV very early
during uml boot under host kernel >= test8.

I can try to localize this place in case you are not able to reproduce
the crash, but I suppose you can see it yourself.

Did you actually try to run uml on that kernels?


Regards,
Yuri Pudgorodsky

Re: [User-mode-linux-user] Ptrace broken since 2.4.0-test8pre4?...

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

yu...@as... said:
> Beeing an active user mode linux user :-) I can say that since
> 2.4.0-test8 (host kernel) I cannot run uml-linux successfully.

> In contrast with popular feeling that "threaded programes screwed
> signal handling on test8.", it is actually a  small change to arch/
> i386/ptrace.c introduced since test8pre4. 

I don't see a problem with the ptrace fix.  Before changing EIP, UML saves all 
of the process registers into the thread struct.  So that change to ORIG_EAX 
shouldn't matter as far as UML is concerned.

				Jeff

[User-mode-linux-user] Ptrace broken since 2.4.0-test8pre4?...

[Yuri P. <yu...@as...>]

Hi!

Beeing an active user mode linux user :-) I can say that since
2.4.0-test8 (host kernel) I cannot run uml-linux successfully.

In contrast with popular feeling that "threaded programes screwed
signal handling on test8.", it is actually a  small change to
arch/i386/ptrace.c introduced since test8pre4.

Also, I remember complains from  Andi Kleen noticed that new kernels
break ups (an alternative debugger).

See the following postings -

  http://kernelnotes.org/lnxlists/linux-kernel/lk_0009_01/msg00265.html
  http://kernelnotes.org/lnxlists/linux-kernel/lk_0009_01/msg00283.html

resulted in this change -

--- v2.4.0-test7/linux/arch/i386/kernel/ptrace.c Fri Jun 23 21:55:07 2000
+++ linux/arch/i386/kernel/ptrace.c Sat Sep 2 12:00:02 2000
@@ -99,6 +99,11 @@
                case EFL:
                        value &= FLAG_MASK;
                        value |= get_stack_long(child, EFL_OFFSET) & ~FLAG_MASK;
+                       break;
+               case EIP:
+                       /* Mark us as not being in a system call, so that no restart issues happen */
+                       put_stack_long(child, 4*ORIG_EAX - sizeof(struct pt_regs), -1);
+                       break;
        }
        if (regno > GS*4)
                regno -= 2*4;

While I cannot comment on the above change from technical point of view,
it seems the patch breaks more then it cures. Time to consider reversing?


Regards,
Yuri Pudgorodsky

[User-mode-linux-user] Very poor signal handling Test8

[Frank J. <fx...@in...>]

I concur that Test8 has very POOR signal handling of threaded programs.

I also returned to Test7 and things have greatly stabilized.

:)

Re: [User-mode-linux-user] user-mode port 0.31-2.4.0-test8

[David W. <dw...@in...>]

dw...@in... said:
>  Doesn't work for me. Fails after mounting the root filesystem.

Ignore that. Running on a test8 host doesn't work because threaded 
programs have screwed signal handling on test8. Backing the host back down
to test7 works fine.

--
dwmw2

[User-mode-linux-user] user-mode port 0.31-2.4.0-test8

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

The user-mode port of 2.4.0-test8 is available.

It now boots on LFS kernels.
The block driver now handles out-of-range accesses better.
Various other bugs were fixed.

There is also a Debian 2.2 filesystem, a test harness, and the beginning of a 
test suite available.

The project's home page is http://user-mode-linux.sourceforge.net

The project's download page is http://sourceforge.net/project/filelist.php?grou
p_id=429

				Jeff