[uml-devel] process start time in future

[uml-devel] process start time in future

[Johannes Formann]

Hi,

I've a problem, when I'm running an UML for a few days, the starttime 
of new prozesses are slowly drifting into the future.

I'ven seen that with 2.4.21-6, 2.4.22-5um and 2.4.22-6um.

Can anyone explain/fix it?

An exampel is shown below.

$  ps -f ; date
UID        PID  PPID  C STIME TTY          TIME CMD
johannes  8967  8966  0 23:54 pts/1    00:00:00 -bash
johannes  9117  8967  0 Nov28 pts/1    00:00:00 ps -f
Thu Nov 27 19:27:52 UTC 2003

# ps -f ; date
UID        PID  PPID  C STIME TTY          TIME CMD
root      5256  5254  0 23:55 pts/0    00:00:00 -bash
root      5376  5256  0 Nov28 pts/0    00:00:00 ps -f
Thu Nov 27 19:27:46 UTC 2003


Greetings Johannes


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] process start time in future

[roland]

hi !
also a effect caused by the serveral times reported "clock skew issue"?

maybe it is solved, soon:
http://sourceforge.net/mailarchive/message.php?msg_id=6499133

btw: is jeff the only person doing development or patchmerging/bugfixing most of 
the time? is uml primarily a "one man" thing and uml dies, if jeff looses interest 
one day? i don`t hope so... but i really wonder anyway, that there is so few traffic 
on the mailinglists. is uml still so "exotic" that so few people actually seem to be
interested? maybe everybody has vmware already... ;)


regards
roland


----- Original Message ----- 
From: "Johannes Formann" <johannes@formann.de>
To: <user-mode-linux-devel@lists.sourceforge.net>
Sent: Thursday, November 27, 2003 8:31 PM
Subject: [uml-devel] process start time in future


> Hi,
> 
> I've a problem, when I'm running an UML for a few days, the starttime 
> of new prozesses are slowly drifting into the future.
> 
> I'ven seen that with 2.4.21-6, 2.4.22-5um and 2.4.22-6um.
> 
> Can anyone explain/fix it?
> 
> An exampel is shown below.
> 
> $  ps -f ; date
> UID        PID  PPID  C STIME TTY          TIME CMD
> johannes  8967  8966  0 23:54 pts/1    00:00:00 -bash
> johannes  9117  8967  0 Nov28 pts/1    00:00:00 ps -f
> Thu Nov 27 19:27:52 UTC 2003
> 
> # ps -f ; date
> UID        PID  PPID  C STIME TTY          TIME CMD
> root      5256  5254  0 23:55 pts/0    00:00:00 -bash
> root      5376  5256  0 Nov28 pts/0    00:00:00 ps -f
> Thu Nov 27 19:27:46 UTC 2003
> 
> 
> Greetings Johannes
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: SF.net Giveback Program.
> Does SourceForge.net help you be more productive?  Does it
> help you create better code?  SHARE THE LOVE, and help us help
> YOU!  Click Here: http://sourceforge.net/donate/
> _______________________________________________
> User-mode-linux-devel mailing list
> User-mode-linux-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
> 


-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] process start time in future

[BlaisorBlade]

Alle 23:26, giovedì 27 novembre 2003, roland ha scritto:
> btw: is jeff the only person doing development or patchmerging/bugfixing
> most of the time?
He is not the only one who develops UML: there is a whole bunch of patches 
coming in, even if he is the only one with write access to the official 
CVS/BitKeeper(I don't know which one) repository, as long as I know. Just as 
the main kernel.
> is uml primarily a "one man" thing and uml dies, if jeff
> looses interest one day? i don`t hope so... but i really wonder anyway,
> that there is so few traffic on the mailinglists.

That's not true, provided you don't compare to the LKML...
> is uml still so "exotic"
> that so few people actually seem to be interested?
A few people? It's inside Debian(Matt Zimmerman is the maintainer), it's 
bundled inside SuSE 9.0(we have often mail from SuSE labs)

1) it is very interesting: some things simply can't be done without it.

For instance, if you have a switched LAN, instead than one with a hub, to 
install NIDS you normally need hardware support(i.e. a port on the switch, 
connected to the NIDS, replicating what goes out from another port); and then 
a devoted machine, only to run Snort...

2) it's widely used and will be more, since it has been included inside 
official 2.6(even if you still need some patches :-( ); there are various 
hosting providers providing virtual servers using it; in Italy, for instance, 
one of these advertises on a lot of widely distributed informatic reviews...; 
I've also read a lot of articles speaking about it.
> maybe everybody has
> vmware already... ;)
VMware is not free software and since it goes emulating hardware, it should 
have more overhead; said that, I have no benchmark to show this. But if UML 
were slower, it would only need tuning. Note that normally, UML memory is put 
on a disk file, so it IS slower. But if you put /tmp on tmpfs(i.e. in RAM), 
this goes away.
-- 
cat <<EOSIGN
Paolo Giarrusso, aka Blaisorblade
Linux Kernel 2.4.21/2.6.0-test on an i686; Linux registered user n. 292729
EOSIGN




-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

Re: [uml-devel] process start time in future

[Henrik Nordstrom]

On Fri, 28 Nov 2003, BlaisorBlade wrote:

> VMware is not free software and since it goes emulating hardware, it should 
> have more overhead; said that, I have no benchmark to show this. But if UML 
> were slower, it would only need tuning. Note that normally, UML memory is put 
> on a disk file, so it IS slower. But if you put /tmp on tmpfs(i.e. in RAM), 
> this goes away.

This is not one of the major bottlenecks of UML.

How it manages the virtual memory may be based on what I am seeing when 
trying to debug my memory management related issue with iptables-restore 
failing..

The disk I/O drivers in UML is a sad chapter and is a major bottleneck
(both ubd and hostfs), but at the same time is not beyond repair. The
reason to this is manly the lack of reasonable async io methods on the 
host..

The number of context switches required on the host for servicing a system
call within the UML is also a major bottleneck, but is something which
Jeff is agressively trying to address with SKAS.

Regards
Henrik



-------------------------------------------------------
This SF.net email is sponsored by: SF.net Giveback Program.
Does SourceForge.net help you be more productive?  Does it
help you create better code?  SHARE THE LOVE, and help us help
YOU!  Click Here: http://sourceforge.net/donate/
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel