[uml-devel] uml-patch-2.4.22-7

[uml-devel] uml-patch-2.4.22-7

[Jeff Dike]

This is another sizable set of changes.  The major items in it are
	Sapan's real-time clock patch - this allows UML to sleep accurately,
assuming enough idle cycles on the host
	highmem support is back - it seems to work, but I haven't banged on
it to much.  One caveat is that using highmem in skas mode is very slow because
of the need to do kmapping in the copy_user routines.
	the start of SMP support in skas mode - it's not totally there, but
it now compiles and runs with one processor
	it will now use /dev/anon if it's present on the host, which it won't
be until I release that patch, which will be fairly soon.  In conjunction with
ubd-mmap, I'm seeing ~25% host memory use reduction.

There are also a number of smaller fixes, code cleanup, improved error
messages, etc.

				Jeff



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Matt Ayres]

Does reboot work inside UML's properly yet? :P (in the sense it will
restart the UML) 

On Tue, 2003-12-16 at 11:53, Jeff Dike wrote:
> This is another sizable set of changes.  The major items in it are
> 	Sapan's real-time clock patch - this allows UML to sleep accurately,
> assuming enough idle cycles on the host
> 	highmem support is back - it seems to work, but I haven't banged on
> it to much.  One caveat is that using highmem in skas mode is very slow because
> of the need to do kmapping in the copy_user routines.
> 	the start of SMP support in skas mode - it's not totally there, but
> it now compiles and runs with one processor
> 	it will now use /dev/anon if it's present on the host, which it won't
> be until I release that patch, which will be fairly soon.  In conjunction with
> ubd-mmap, I'm seeing ~25% host memory use reduction.
> 
> There are also a number of smaller fixes, code cleanup, improved error
> messages, etc.
> 
> 				Jeff
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: IBM Linux Tutorials.
> Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
> Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
> Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
> _______________________________________________
> User-mode-linux-devel mailing list
> User-mode-linux-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
-- 
Matt Ayres <matta@tektonic.net>
TekTonic



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[marco ancillotti]

> 	highmem support is back - it seems to work, but I haven't banged on
> it to much.  

no problem jeff , I will test i hardly for you ;-)

> One caveat is that using highmem in skas mode is very slow because
> of the need to do kmapping in the copy_user routines.

so there's better performance in tt mode when using highmem ?

bye.


-- 
----------------------------------------------------

Marco Ancillotti 
uniX / Linux Tecnical specialist

- IBM Aix Certified Advanced Tecnical Specialist
- IBM Enterprise storage specialist
- Red Hat Certified Engineer

    Lpar manager first developer and mantainer.
( http://savannah.nongnu.org/projects/lparmanager )

S.in.te.si. srl
via cecchi 28
50053 Empoli (FI)
tel: 0571/72400
fax: 0571/72064

m.ancillotti@retesintesi.it
cell: 347/3460086

----------------------------------------------------


-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[William Stearns]

Good day, Matt,

On Tue, 16 Dec 2003, Jeff Dike wrote:

> matta@tektonic.net said:
> > Does reboot work inside UML's properly yet? :P (in the sense it will
> > restart the UML)  
> 
> It works now, and as far as I know, it has always worked.

	I can confirm it works in at least my setup on the UML Coop.  
Reboot shuts down the current VM, starting a new one in its place.  Can 
you describe what doesn't work?
	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 <dirk@loth.demon.co.uk>)
--------------------------------------------------------------------------
William Stearns (wstearns@pobox.com).  Mason, Buildkernel, freedups, p0f,
rsync-backup, ssh-keyinstall, dns-check, more at:   http://www.stearns.org
Linux articles at:                         http://www.opensourcedigest.com
--------------------------------------------------------------------------



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Jeff Dike]

matta@tektonic.net said:
> Does reboot work inside UML's properly yet? :P (in the sense it will
> restart the UML)  

It works now, and as far as I know, it has always worked.

				Jeff



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Matt Ayres]

Well, when reboot/shutdown -r is run from within a UML it shuts down and
then states "INIT: no more processes in this runlevel" running
2.4.22-um6.  I searched Google and saw that some people had this problem
when UM_FASTCALL wasn't set, however I did a grep on my .config file and
didn't not find that option.

Is this most likely why my reboots are not working?  How do I enable it?
Right now I am using the djc 2.4.23 kernel so I would assume he would
turn this on.

A point in the right direction would be very appreciated as this is a
bit of a pain at the moment.  I'll also be testing the /dev/anon
functionality very much as soon as that is released.

On Tue, 2003-12-16 at 12:53, William Stearns wrote:
> Good day, Matt,
> 
> On Tue, 16 Dec 2003, Jeff Dike wrote:
> 
> > matta@tektonic.net said:
> > > Does reboot work inside UML's properly yet? :P (in the sense it will
> > > restart the UML)  
> > 
> > It works now, and as far as I know, it has always worked.
> 
> 	I can confirm it works in at least my setup on the UML Coop.  
> Reboot shuts down the current VM, starting a new one in its place.  Can 
> you describe what doesn't work?
> 	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 <dirk@loth.demon.co.uk>)
> --------------------------------------------------------------------------
> William Stearns (wstearns@pobox.com).  Mason, Buildkernel, freedups, p0f,
> rsync-backup, ssh-keyinstall, dns-check, more at:   http://www.stearns.org
> Linux articles at:                         http://www.opensourcedigest.com
> --------------------------------------------------------------------------
-- 
Matt Ayres <matta@tektonic.net>
TekTonic



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Matt Ayres]

Couple questions regarding these features (i'm fairly new to using UML
so I haven't  been tracking the list for long)

On Tue, 2003-12-16 at 11:53, Jeff Dike wrote:
> This is another sizable set of changes.  The major items in it are
> 	Sapan's real-time clock patch - this allows UML to sleep accurately,
> assuming enough idle cycles on the host

Is there a thread that I can be pointed to or a brief explanation of how
this is a benefit?

> 	it will now use /dev/anon if it's present on the host, which it won't
> be until I release that patch, which will be fairly soon.  In conjunction with
> ubd-mmap, I'm seeing ~25% host memory use reduction.
> 

ubd-mmap is new behavior that doesn't need to be enabled? (meaning, just
upgrade to -7 and we'll seen some better performance using the block
device) ?  Do you know how much performance is gained from ubd-mmap and
/dev/anon individually?





-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Jeff Dike]

mancillotti@interfree.it said:
> so there's better performance in tt mode when using highmem ? 

Ironically, tt mode is faster than skas mode when you have a lot of highmem
in use.  skas mode is slowed down that much.

I'm thinking of ways of improving that.  The current plan is to make kmap/kunmap
do reads and writes instead of mmap/munmap in skas mode.  That will exchange
I/O for the tlb flushing.  We'll see how much difference that makes.

				Jeff



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Jeff Dike]

matta@tektonic.net said:
> Well, when reboot/shutdown -r is run from within a UML it shuts down
> and then states "INIT: no more processes in this runlevel" running
> 2.4.22-um6.

Does it halt?

Where did you get your filesystem image?

I saw this a while back, and it turned out to be the filesystem.  The script
that finally shuts the system down wasn't running, and the reason was that
the running rc script called it through a link, and the link wasn't there.

> Is there a thread that I can be pointed to or a brief explanation of
> how this is a benefit? 

The benefit is that when something does a 'sleep 1' inside UML, it will
sleep for 1 second, and not 2 or 3 seconds.  There have been various complaints
about this on the lists.

> ubd-mmap is new behavior that doesn't need to be enabled? (meaning,
> just upgrade to -7 and we'll seen some better performance using the
> block device) ?  

No, you have to turn it on (by adding ubd=mmap to the command line).

> Do you know how much performance is gained from
> ubd-mmap and /dev/anon individually? 

It's not performance of an individual UML per se (the testing I've done shows
that the performance is about the same), it's lower memory consumption on
the host.  ubd-mmap eliminates the duplicate copies of data on the host (in
the host page cache and the UML page cache) by mapping the host page cache
directly into UML physical memory.  /dev/anon completes the picture by
releasing the overmapped physical memory pages back to the host.

So, it'll increase your hosting capacity, and if the host is swapping, improve
performance by keeping more UMLs in memory.

				Jeff



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Matt Ayres]

On Tue, 2003-12-16 at 17:55, Jeff Dike wrote:
> matta@tektonic.net said:
> > Well, when reboot/shutdown -r is run from within a UML it shuts down
> > and then states "INIT: no more processes in this runlevel" running
> > 2.4.22-um6.
> 
> Does it halt?
> 
> Where did you get your filesystem image?
> 
> I saw this a while back, and it turned out to be the filesystem.  The script
> that finally shuts the system down wasn't running, and the reason was that
> the running rc script called it through a link, and the link wasn't there.
> 

Yes, this was it.  The halt symlink was in runlevel 0 but not in 6.  I
guess I need to find a type to run "cd /etc/rc6.d && ln -sf
../init.d/halt S01reboot" inside each UML.  Thanks.

> > Is there a thread that I can be pointed to or a brief explanation of
> > how this is a benefit? 
> 
> The benefit is that when something does a 'sleep 1' inside UML, it will
> sleep for 1 second, and not 2 or 3 seconds.  There have been various complaints
> about this on the lists.
> 
> > ubd-mmap is new behavior that doesn't need to be enabled? (meaning,
> > just upgrade to -7 and we'll seen some better performance using the
> > block device) ?  
> 
> No, you have to turn it on (by adding ubd=mmap to the command line).
> 
> > Do you know how much performance is gained from
> > ubd-mmap and /dev/anon individually? 
> 
> It's not performance of an individual UML per se (the testing I've done shows
> that the performance is about the same), it's lower memory consumption on
> the host.  ubd-mmap eliminates the duplicate copies of data on the host (in
> the host page cache and the UML page cache) by mapping the host page cache
> directly into UML physical memory.  /dev/anon completes the picture by
> releasing the overmapped physical memory pages back to the host.
> 
> So, it'll increase your hosting capacity, and if the host is swapping, improve
> performance by keeping more UMLs in memory.
> 

Understood.
> 				Jeff
> 
> 
> 
> -------------------------------------------------------
> This SF.net email is sponsored by: IBM Linux Tutorials.
> Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
> Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
> Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
> _______________________________________________
> User-mode-linux-devel mailing list
> User-mode-linux-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel
-- 
Matt Ayres <matta@tektonic.net>
TekTonic



-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
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] uml-patch-2.4.22-7

[Matt Zimmerman]

On Tue, Dec 16, 2003 at 06:32:37PM -0500, Matt Ayres wrote:

> Yes, this was it.  The halt symlink was in runlevel 0 but not in 6.  I
> guess I need to find a type to run "cd /etc/rc6.d && ln -sf
> ../init.d/halt S01reboot" inside each UML.  Thanks.

Where did this broken filesystem come from?

-- 
 - mdz


-------------------------------------------------------
This SF.net email is sponsored by: IBM Linux Tutorials.
Become an expert in LINUX or just sharpen your skills.  Sign up for IBM's
Free Linux Tutorials.  Learn everything from the bash shell to sys admin.
Click now! http://ads.osdn.com/?ad_id=1278&alloc_id=3371&op=click
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel