user-mode-linux-devel

Re: [uml-devel] RE: [uml-user] [PATCH net-tools] debug improvements

[Rodrigo B. (a. morcego) <rod...@co...>]

On Wed, Dec 06, 2000 at 12:26:43AM -0500, Jeff Dike wrote:
> mo...@ea... said:
> >      TCP    Unix   Difference
> > -----------------------------=20
> > Min| 0.8    0.3    0.5=20
> > Max| 29.1   0.3    28.8=20
> > Avg| 110.0  1.8    108.2=20
> > Pak| 6%     0%     0%=20
> > -----------------------------=20
>=20
> Nice.  You sold me.  Is your patch ready for public viewing?  I'd like to=
=20
> reproduce these results here.
>=20

Same for me ...
I would like to integrate it with my patch. With both performance improveme=
nts,
it will be great.

[]s

--=20
 Rodrigo Barbosa (morcego)  - rod...@co...
 Conectiva R&D Team         - http://distro.conectiva.com.br
 "Quis custodiet custodes?" - http://www.conectiva.com

Re: [uml-devel] RE: [uml-user] [PATCH net-tools] debug improvements

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

mo...@ea... said:
>      TCP    Unix   Difference
> ----------------------------- 
> Min| 0.8    0.3    0.5 
> Max| 29.1   0.3    28.8 
> Avg| 110.0  1.8    108.2 
> Pak| 6%     0%     0% 
> ----------------------------- 

Nice.  You sold me.  Is your patch ready for public viewing?  I'd like to 
reproduce these results here.

				Jeff

[uml-devel] RE: [uml-user] [PATCH net-tools] debug improvements

[Mooneer S. <mo...@ea...>]

Kernel parameters: mem=64m no-xterm fake_arch
Host kernel: 2.2.17
UML kernel: 2.4.0-test11

Flood pinged the UML kernel with 1,000 ICMP packets and kept track of
packet loss/ping time. This was performed three times, with the best
result for each used.

"Traditional Style" (TCP to localhost):

root@apoc [~]# ping -c 1000 -f 10.0.0.1
PING 10.0.0.1 (10.0.0.1) from 10.0.0.2 : 56(84) bytes of data.
..........................................................................
--- 10.0.0.1 ping statistics ---
1073 packets transmitted, 1000 packets received, 6% packet loss
round-trip min/avg/max = 0.8/29.1/110.0 ms
root@apoc [~]#

Unix socket (/tmp/tap0):

root@apoc [~]# ping -c 1000 -f 10.0.0.1
PING 10.0.0.1 (10.0.0.1) from 10.0.0.2 : 56(84) bytes of data.
.
--- 10.0.0.1 ping statistics ---
1000 packets transmitted, 1000 packets received, 0% packet loss
round-trip min/avg/max = 0.3/0.3/1.8 ms
root@apoc [~]#

Table of Results (times in ms):

     TCP    Unix   Difference
-----------------------------
Min| 0.8    0.3    0.5
Max| 29.1   0.3    28.8
Avg| 110.0  1.8    108.2
Pak| 6%     0%     0%
-----------------------------
Pak = Packet loss (percentage)
Min = Minimum ping time
Max = Maximum ping time
Avg = Average ping time

TCP = UML communicates with um_eth_net_util through a TCP/IP connection
Unix = UML communicates with um_eth_net_util through a named pipe/Unix
socket

I believe the reason for the speed increase is because the kernel
does not have to go through the overhead that a TCP/IP connection
entails, even when using the loopback device.

-----Original Message-----
From: Jeff Dike [mailto:jd...@ka...]
Sent: Tuesday, December 05, 2000 5:27 PM
To: Mooneer Salem
Cc: Rodrigo Barbosa (aka morcego);
use...@li...
Subject: Re: [uml-user] [PATCH net-tools] debug improvements


rod...@co... said:
> Have you benckmarked it ?

Rodrigo is right.  Measure the latency across both types of sockets.  A
connection through localhost goes through the loopback interface, which I
think is about as low-overhead as you could want.

				Jeff

[uml-devel] Re: [uml-user] [PATCH net-tools] debug improvements

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

rod...@co... said:
> Have you benckmarked it ? 

Rodrigo is right.  Measure the latency across both types of sockets.  A 
connection through localhost goes through the loopback interface, which I 
think is about as low-overhead as you could want.

				Jeff

[uml-devel] Re: [uml-user] [PATCH net-tools] debug improvements

[Rodrigo B. (a. morcego) <rod...@co...>]

On Tue, Dec 05, 2000 at 02:56:18PM -0800, Mooneer Salem wrote:
> [sent to user-mode-linux-devel as well]
>=20
> I'm messing with the network code at the moment. I noticed that
> the virtual ethernet driver makes a TCP connection to localhost.
> How much less overhead would be incurred if this was changed to
> use Unix sockets instead (e.g. like /tmp/mysql.sock for MySQL)?
>=20
> Right now I made a few hacks to both the user-mode utilities and
> the test11 kernel to use Unix sockets instead of TCP/IP.
> Basically the kernel opens /tmp/tapx for each virtual ethernet device
> it detects, instead of opening a connection to localhost. The user-mode
> utilities make sockets on /tmp/tapx and use those instead of TCP/IP. This
> should cut down on overhead. Once I see how stable or unstable it
> is I'll make the diffs and send them to Jeff and/or the list.

Well, I may be wrong, but I think UNIX Socket handling is actualy
slower then network socket handling (expecialy when conserning the
lo interface).
Have you benckmarked it ?

[]s

--=20
 Rodrigo Barbosa (morcego)  - rod...@co...
 Conectiva R&D Team         - http://distro.conectiva.com.br
 "Quis custodiet custodes?" - http://www.conectiva.com

[uml-devel] RE: [uml-user] [PATCH net-tools] debug improvements

[Mooneer S. <mo...@ea...>]

[sent to user-mode-linux-devel as well]

I'm messing with the network code at the moment. I noticed that
the virtual ethernet driver makes a TCP connection to localhost.
How much less overhead would be incurred if this was changed to
use Unix sockets instead (e.g. like /tmp/mysql.sock for MySQL)?

Right now I made a few hacks to both the user-mode utilities and
the test11 kernel to use Unix sockets instead of TCP/IP.
Basically the kernel opens /tmp/tapx for each virtual ethernet device
it detects, instead of opening a connection to localhost. The user-mode
utilities make sockets on /tmp/tapx and use those instead of TCP/IP. This
should cut down on overhead. Once I see how stable or unstable it
is I'll make the diffs and send them to Jeff and/or the list.

-----Original Message-----
From: use...@li...
[mailto:use...@li...]On Behalf Of
Rodrigo Barbosa (aka morcego)
Sent: Tuesday, December 05, 2000 12:32 PM
To: Jeff Dike
Cc: use...@li...
Subject: Re: [uml-user] [PATCH net-tools] debug improvements


On Tue, Dec 05, 2000 at 01:19:20AM -0500, Jeff Dike wrote:
> It's up to Jim Leu, but I would be inclined not to take this patch.  The
> reason is that you're chasing a big problem, and big problems have big
fixes
> (in the sense that you fix it and it makes a big difference).  So, if I
were
> you, I'd ignore all this little stuff until you find the real bug.  Also,
by
> fixing little things, you run the risk of getting enough small
improvements to
> add up to something that's barely acceptable.  Then you stop chasing the
real
> problem because things are now sort of OK.

I don't necessarily agree with you. I have seen big problems that only
needed small fixes.
But in this particular case, we only think that it's just one big problem.
But we can have a series of small problems that looks just like a big one.

> I think this particular patch is also a bad idea because it's good to be
able
> to turn on debugging in the standard executable with a command line
option,
> and not have to recompile it.  And if debugging is the default, as you
have
> it, then hardly anyone is going to turn it off, so the ifdefs aren't doing
> anything except cluttering the code.

Well, my proposal was to minimize changes. The code didn't have suport
for a debug flag, the debug option was hardcoded and turned on by default.
What I'm planning to the future (don't scream or cluebat me, please), is to
use multithreading to have debugging with very small performance hits.
I keep the main thread for routing, and another one for debugging, so I can
have async debug, causing minimal performance hit.

> I also have a hard time believing that those tests make any kind of
> performance difference.  And if it is measurable, then we're back to my
> argument above.

They have an effect, yes. Very small. But I'm not yet sure that we are
dealing
with just one big problem here.

> Cleaning up excessive output might be a good idea, but all of the
fprintf's
> that you've ifdefed look like fairly rare occurrences.

Not so rare. And fprintf is a samewhat costy instruction.

> In the interest of not being totally negative, I have to say that moving
that
> fork does look like a good idea.

Thanks.
And no, I don't think you are being totaly negative. Totaly negative is like
some people that simply refuses a patch, without telling why they did.
Once you explained your reason, I can work to improve it, remove problematic
points, and maybe come up with something better.

Re: [uml-devel] bug in uml?

[Michael V. <mj...@un...>]

On Tue, 5 Dec 2000, Erik Paulson wrote:

> One cool trick that you can do is checkpoint under UML, and (after hacking
> on LINE for a little bit) restart on NT :)

Or even better...include a NT box in a Mosix cluster :)


Mike

Re: [uml-devel] bug in uml?

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

On Mon, Dec 04, 2000 at 11:01:19PM -0500, Jeff Dike wrote:
> 
> BTW, have you tried checkpointing UML itself?  That would be way cool.

I'd thought about that, and it's something to look into later, but for now
I just want to do a single process. 

One cool trick that you can do is checkpoint under UML, and (after hacking
on LINE for a little bit) restart on NT :)

-Erik

> 
> 				Jeff
> 
> 

[uml-devel] cvs checkins

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

I've checked in the fixes to the block driver.  "dd if=/dev/ubd/0 
of=/dev/null" no longer causes hangs, and dbench no longer causes file 
corruption.

At the request of Daniel Phillips, I'm also maintaining a patch against the 
latest cvs as uml-patch-latest.2.4.0-test11.  It's available from the project 
download page.

				Jeff

Re: [uml-devel] bug in uml?

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

epa...@cs... said:
> The problem is that it segfaults restoring the last page of memory.
> Maybe it's something wonky with the way UML handles it's stack, or
> maybe there's a bug in the mmap() part of UML.

OK, I'll have a look at it.  There is some funkiness going on, but it should 
work.

BTW, have you tried checkpointing UML itself?  That would be way cool.

				Jeff

[uml-devel] bug in uml?

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

Hi,

	I've been working on getting checkpointing working under user-mode-linux,
and last night I found an interesting utility that can checkpoint any 
currently-running process all in userspace. /proc gives you most of what you
need, and the little bit it doesn't (current registers, position info for
the currently open fd's) it swipes out of the process with ptrace. Now, I 
could easily believe that the checkpointing wouldn't work under UML, so I'm
planning on adding the bits to /proc it needs so it doesn't have to mess with
ptrace(). However, I don't see any reason the restart code included here
wouldn't work - basically it creates a new stack to work out of, 
mmap()'s in the old process, opens up the old files and seeks in them, and
restores the registers.

The problem is that it segfaults restoring the last page of memory. Maybe
it's something wonky with the way UML handles it's stack, or maybe there's a
bug in the mmap() part of UML.

The code I was trying was here: 
ftp://ftp.gin.cz/pub/local/feela/src/freezer.tgz

It builds two apps, freeze and warm. I would 'freeze' a process that was
just doing a 
for(i=0;i<1000;i++){printf(i);sleep(1);}

under the host kernel and then 'warm' it under UML.

Thanks!

-Erik

[uml-devel] Re: stdin</dev/null - darn close!

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

wst...@po... said:
> 	I get _reams_ - hundreds - of:
> Failed to forward 0 to pid 25385, errno = 9 
> 	With each keystroke on one of the xterms after boot, I get:
> stdio_rcv_proc read failed, errno = 5 
> The console doesn't seem to be reset on halt

Well, I don't see any of that.  It looks like I may have to pay another visit 
to Dartmouth and break fingers until you see that there is nothing wrong with 
UML...

> I take it you split up the file descriptors?  It seems to work as
> advertised.

Yup, the consoles now have separate in and out descriptors, although in the 
normal case that they're attached to a tty, they're the same.

> error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp2: No such file or directory
> error: parent pid should not be init 
> error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp4: No such file or directory 

The first and third are from xterm.  I have no idea what's going on in its 
little mind, but it seems harmless.

I guess the second is, too, but I've never seen it.

				Jeff

[uml-devel] stdin</dev/null - darn close!

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

Good day, Jeff, all,
	(Host rh7, 2.4.0-test11-ac1, uml rh6.2, Jeff's 2.4.0-test11-1um #1
Mon Nov 20 01:07:52 EST 2000 um unknown).

-------- STDIN=/dev/null --------
	./linux-2.4.0-test11 mem=48M </dev/null

	I get _reams_ - hundreds - of:
Failed to forward 0 to pid 25385, errno = 9
Failed to forward 0 to pid 25396, errno = 9
Failed to forward 0 to pid 25385, errno = 9
Failed to forward 0 to pid 25574, errno = 9
Failed to forward 0 to pid 25385, errno = 9
Failed to forward 0 to pid 25396, errno = 9
Failed to forward 0 to pid 25385, errno = 9
Failed to forward 0 to pid 25574, errno = 9
Failed to forward 0 to pid 25385, errno = 9
Failed to forward 0 to pid 25396, errno = 9
Failed to forward 0 to pid 25623, errno = 9

	These end as the loging prompts show up.
	With each keystroke on one of the xterms after boot, I get:

stdio_rcv_proc read failed, errno = 5

	That said, the redirect does seem to work; typing on the console
has no effect, as it should, but the console messages correctly show up
(they didn't in earlier kernels, as we discussed).  I take it you split up
the file descriptors?  It seems to work as advertised.
	The console doesn't seem to be reset on halt (nothing I type
after halt shows up and <Enter> just prints a new copy of the prompt on
the same line), so I have to type "clear ; reset" blind, but it comes back
after that.

-------- STDOUT=/dev/null --------
	e2fsck -v root_fs ; ./linux-2.4.0-test11 mem=48M >/dev/null

	Way cool.  All console output suppressed, but I can still login
blind on the console and type commands.  "ls -al >/dev/ttys/1" spits out a
directory listing to the first xterm.
	The original shell even works correctly after a halt.  Darn,
nothing to complain about.  ;-)

-------- STDIN=STDOUT=/dev/null --------
	e2fsck -v root_fs ; ./linux-2.4.0-test11 mem=48M </dev/null
>/dev/null

	Pretty close.  I no longer get the "Failed to forward..." or
"Stdio_rvc_proc..." messages, although those might still exist but be
heading off to /dev/null so I don't see them.
	I do get the following on the console.  The first and third are
well-known; they show up at every xterm respawn.  I don't remember the
second.

error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp2: No such file or directory
error: parent pid should not be init
error stat'ing /dev/ttydirs/xxx/xxxxxxxxxxxxp4: No such file or directory

	Logging in blind correctly fails to work.  Good.


	In all three cases the xterms start up without problem.  Also,
adding "no-xterm" to the command line works correctly; the xterms don't
start, but servers are available.
	Waaaaaaaaay cool!  Nice work, Jeff.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	Last week TechWeb reported that Eric S. Raymond was invited to
speak to a Microsoft Research Group. Given Raymond's status in the Open
Source Community, this is somewhat akin to the Vatican asking Martin
Luther to pay a visit in 1517 and chat about the note that he left.
	-- Joe "Zonker" Brockmeier
--------------------------------------------------------------------------
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] cvs checkins

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

This batch contains the "debug=go" and "ubd=sync" options.

				Jeff

Re: [uml-devel] Swap space in 2.4.0-test11

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

wst...@po... said:
> If you can't reproduce it, I'll compile my own non-stripped one.

You guessed it.  It Works For Me (TM):

darkstar:~# swapon /dev/ubd/1
darkstar:~#

Run it debug and get a stack trace from the panic.

				Jeff

[uml-devel] Swap space in 2.4.0-test11

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

Good day, Jeff, all,
	I've got a quirk in 2.4.0-test11.  The root filesystem in question
is rh62., host kernel is 2.4.0-test11-ac1, host root is rh7, uml kernel is
the stock one from the web site, swap is a 126MB swapfile.

-rwxr-xr-x    1 wstearns wstearns  1306884 Nov 20 01:07 linux-2.4.0-test11
lrwxrwxrwx    1 wstearns wstearns       18 Nov 20 01:22 linux -> linux-2.4.0-test11
[wstearns@sparrow rh6.2]$ ./linux mem=20M ubd1=swap
tracing thread pid = 3353
Linux version 2.4.0-test11-1um (jd...@cc...) (gcc version
egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)) #1 Mon Nov 20 01:07:52 EST 2000
[snip]
Setting clock : Sat Nov 25 15:46:14 EST 2000 [  OK  ]
Activating swap partitions Kernel panic: Double fault on 0x51800000 - panicing because it wasn't fixed the first time

	OK, let's try it with a non-sparse swapfile on the host:

[wstearns@sparrow rh6.2]$ dd if=/dev/zero of=swap-notsparse bs=$((1024*1024)) count=126
126+0 records in
126+0 records out
[wstearns@sparrow rh6.2]$ sudo mkswap swap-notsparse
Password:
Setting up swapspace version 1, size = 132116480 bytes

	And:

[wstearns@sparrow rh6.2]$ ./linux mem=20M ubd1=swap-notsparse
tracing thread pid = 3625
Linux version 2.4.0-test11-1um (jd...@cc...) (gcc version egcs-2.91.66 19990314/Linux (egcs-1.1.2 release)) #1 Mon Nov 20 01:07:52 EST 2000
[snip]
Setting clock : Sat Nov 25 15:57:01 EST 2000 [  OK  ]
Activating swap partitions Kernel panic: Double fault on 0x51800000 - panicing because it wasn't fixed the first time

	Identical.
	OK, one more test.  Tell the host kernel to stop using the
physical swap partition on the physical drive, give wstearnbs
ownership of the partition, then:

./linux mem=20M ubd1=/dev/hda5
	and
Activating swap partitions Kernel panic: Double fault on 0x51800000 - panicing because it wasn't fixed the first time
	Same error.  Your kernel doesn't have symbol information, so gdb
is of limited use.  If you can't reproduce it, I'll compile my own
non-stripped one.
	Cheers,
	- Bill

---------------------------------------------------------------------------
	"Put down those Windows disks, Dave..."
	-- HAL
--------------------------------------------------------------------------
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] cvs checkins

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

I checked in a bunch of changes which cause clone failures to be handled much 
more gracefully.  Now, you'll get a failed fork rather than a panic.

				Jeff

[uml-devel] test11 checkins

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

The 2.4.0-test11 updates are checked in.

hostfs now works better.  You can mount it, cd around inside it, and cat 
things.  Writing things is right out.

It turned out that my stack guard page wasn't really guarding anything.  This 
has been fixed.

				Jeff

Re: [uml-devel] First pass at some UML benchmarks

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

epa...@cs... said:
> I've run lmbench through a UML session, and again natively on a PIII
> machine here to see how things looked.

I run lmbench as a stress test on every kernel before I release it.  I don't 
pay any attention to the numbers because they are all over the map.  My 
activities on the host apparently affect the lmbench numbers of the UML.

How many runs did you make?  And how consistent were the numbers?

If your numbers aren't consistent, I would try something extreme like bring up 
the host only as far as needed to run UML (like single-user with the necessary 
filesystems mounted) and do the lmbench.  That would put the host in as 
predictable a state as possible.  And then do the same with the UML to get it 
in as consistent a state as possible.  Then do a bunch of lmbench runs and see 
if you get good numbers.

Also, if you're going to compare UML vs host numbers, they'd better be running 
the same kernel version.

If you stay away from the kernel, UML performance is basically the same as the 
host, which isn't surprising.  Once you start going into the kernel, 
performance starts hurting.

Every UML system call involves four context switches on the host.  Adding a 
new system call tracing mechanism to the host is on my list of things to do 
for 2.5.

Also, I/O bound tests are hurt by the block driver, which can only have one 
outstanding request at a time.  Fixing this would probably greatly help the 
performance of things like kernel builds.

				Jeff

[uml-devel] First pass at some UML benchmarks

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

Hello,
	I've run lmbench through a UML session, and again natively on a PIII
machine here to see how things looked. It's not totally fair, I should have
probably mounted the lmbench directory over the loopback device on the native
kernel, and I think I've got some memory stuff misconfigured (UML blew up
when I tried to tell it lmbench that it had more than 10 megs of RAM, so I 
think I'm missing a command-line flag or something. It also failed in 
alloc_pages during a run, but I haven't had a chance to take a look at that)

Has anyone else taken a look at some benchmarks?

The machine I'm using is a 650mhz PIII with 512 megs of RAM, and a 10gig IDE
drive (I'll have to ask around to see who made it). I'm using UML 2.4.0-test10
as downloaded off the website tonight (18 Nov), on a 2.2.16smp host. 

This work is for CS736 (Operating Systems)  here at the University of Wisconsin.My project involves using UML as a testbed for secure sandboxing and process
migration in a distributed system, and we're trying to get a feeling for some
performance numbers. I'm going to be running SPEC through soon as well, since
many of our target apps are CPU-bound and shouldn't hit into the OS too much
(except for VM-issues)

Thanks, and I'm looking forward to your comments.

-Erik

(lmbench can be found at:
http://www.bitmover.com/lmbench/ )

                 L M B E N C H  2 . 0   S U M M A R Y
                 ------------------------------------
                 (Alpha software, do not distribute)

Basic system parameters
----------------------------------------------------
Host                 OS Description              Mhz
                                                    
--------- ------------- ----------------------- ----
perdita.c Linux 2.2.16-       i686-pc-linux-gnu  652
uml       Linux 2.4.0-t       i686-pc-linux-gnu  650

Processor, Processes - times in microseconds - smaller is better
----------------------------------------------------------------
Host                 OS  Mhz null null      open selct sig  sig  fork exec sh  
                             call  I/O stat clos TCP   inst hndl proc proc proc
--------- ------------- ---- ---- ---- ---- ---- ----- ---- ---- ---- ---- ----
perdita.c Linux 2.2.16-  652  0.5  0.8  4.9  6.2    28  1.4  2.1 0.2K 1.2K 27.K
uml       Linux 2.4.0-t  650 25.8 35.4 64.5 101.       36.2 105. 8.3K 16.K 47.K

Context switching - times in microseconds - smaller is better
-------------------------------------------------------------
Host                 OS 2p/0K 2p/16K 2p/64K 8p/16K 8p/64K 16p/16K 16p/64K
                        ctxsw  ctxsw  ctxsw ctxsw  ctxsw   ctxsw   ctxsw
--------- ------------- ----- ------ ------ ------ ------ ------- -------
perdita.c Linux 2.2.16-     1      5            10             28        
uml       Linux 2.4.0-t   359    368           418            413        

*Local* Communication latencies in microseconds - smaller is better
-------------------------------------------------------------------
Host                 OS 2p/0K  Pipe AF     UDP  RPC/   TCP  RPC/ TCP
                        ctxsw       UNIX         UDP         TCP conn
--------- ------------- ----- ----- ---- ----- ----- ----- ----- ----
perdita.c Linux 2.2.16-     1     7   13    26          41        132
uml       Linux 2.4.0-t   359   693  782                             

File & VM system latencies in microseconds - smaller is better
--------------------------------------------------------------
Host                 OS   0K File      10K File      Mmap    Prot    Page
                        Create Delete Create Delete  Latency Fault   Fault 
--------- ------------- ------ ------ ------ ------  ------- -----   ----- 
perdita.c Linux 2.2.16-      9      1     19      2      636     1    0.5K
uml       Linux 2.4.0-t     24      8    169     12     2772          0.1K

*Local* Communication bandwidths in MB/s - bigger is better
-----------------------------------------------------------
Host                OS  Pipe AF    TCP  File   Mmap  Bcopy  Bcopy  Mem   Mem
                             UNIX      reread reread (libc) (hand) read write
--------- ------------- ---- ---- ---- ------ ------ ------ ------ ---- -----
perdita.c Linux 2.2.16-  524  270   76    373    468    181    142  482   160
uml       Linux 2.4.0-t    6   35   -1    270    460    187    150  460   190

Memory latencies in nanoseconds - smaller is better
    (WARNING - may not be correct, check graphs)
---------------------------------------------------
Host                 OS   Mhz  L1 $   L2 $    Main mem    Guesses
--------- -------------  ---- ----- ------    --------    -------
perdita.c Linux 2.2.16-   652  4.60  10.76       122.4
uml       Linux 2.4.0-t   650  4.64  10.86       194.1
 

[uml-devel] CVS checkins

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

If you want to build from source from now on, you have to put ARCH=um on the 
make command line or in the environment.  I added Dave Zarzycki's changes to 
the top-level Makefile, which make the default behavior to build a native 
kernel.  This is in preparation to getting UML in the main pool.

hostfs still doesn't work, but it's configurable.

UML works better as a daemon.

The umn driver doesn't put "-n" on the ifconfig command line.

Also, the problems with checking out arch/um/fs seem to have been fixed.

				Jeff

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

[Matt C. <ma...@co...>]

Jeff Dike wrote:
> 
> ma...@co... said:
> > When I run UML for the first time (to install Slackware), it hangs
> > during the boot process after:
> >
> > none on /proc type proc (rw)
> >
> > If I press CTRL+C, UML boots up fine.  Once I've installed Slackware,
> > I don't have this problem anymore.  Any idea what this is, or if it's
> > related to the problem Bill was having?
> 
> I don't think so.  I think that hang was a hard one.  If it happens with a
> kernel built by me that you got off my site, then it's not.

It happens every time with the linux-2.4.0-test9 and linux-2.4.0-test10
kernel binaries I downloaded from your site.  I'm running on Slackware
7.0, kernel version 2.2.16.

I don't mind it so much since it goes away after I install Slackware,
but I was curious as to what causes it.

-- 

 - Matt Clay
 - Cowboyz.com
 - (503) 241-1990

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

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

ma...@co... said:
> When I run UML for the first time (to install Slackware), it hangs
> during the boot process after:
>
> none on /proc type proc (rw)
>
> If I press CTRL+C, UML boots up fine.  Once I've installed Slackware,
> I don't have this problem anymore.  Any idea what this is, or if it's
> related to the problem Bill was having?

I don't think so.  I think that hang was a hard one.  If it happens with a 
kernel built by me that you got off my site, then it's not.  If you're running 
on a distro that has a "kgcc", then build with that and see if the problem 
goes away.

				Jeff

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

[Matt C. <ma...@co...>]

Jeff Dike wrote:
> 
> wst...@po... said:
> > 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.
> 
> Just to fill everyone in on what happened with this - I looked at this with
> Bill today and it turns out to be a compiler problem.
> 
> At -O2, the default gcc on whatever distro Bill was running calculates
> addresses of labels incorrectly.  Switching to kgcc fixed the problem.

When I run UML for the first time (to install Slackware), it hangs
during the boot process after:

none on /proc type proc (rw)

If I press CTRL+C, UML boots up fine.  Once I've installed Slackware, I
don't have this problem anymore.  Any idea what this is, or if it's
related to the problem Bill was having?

-- 

 - Matt Clay
 - Cowboyz.com
 - (503) 241-1990

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

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

wst...@po... said:
> 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.

Just to fill everyone in on what happened with this - I looked at this with 
Bill today and it turns out to be a compiler problem.

At -O2, the default gcc on whatever distro Bill was running calculates 
addresses of labels incorrectly.  Switching to kgcc fixed the problem.

				Jeff