Menu
▾
▴
user-mode-linux-user
|
From: Snoopy <sn...@do...> - 2002-11-07 11:42:26
|
Hello dear friends, I am slowly making oprogress and now I am attempting to set up netowrking for my UM Linux. I have folloed the advice of networking.html on sourceforge.net and I plumped for the TUN/TAP method of doing this, apparently Ethertap is now considered old. OK, so my host runs eth0 with an IP Adress of 192.168.10.8. I want my UML to use 192.168.10.51 - yes I want the UML to have the same net. So I did the whole doohicky with the bridge and brctl etc. So now my hosts eth0 has no more P address, rather the bidge now has that, or rather my hosts end of the bridge. So then I log into the UML (which was invoked with eth0=tuntap,tap0) and try to do ifconfig eth0 192.168.10.51 up and I get: bash-2.05# ifconfig lo 127.0.0.1 up bash-2.05# ifconfig eth0 192.168.10.51 up SIOCSIFFLAGS: Device or resource busy SIOCSIFFLAGS: Device or resource busy bash-2.05# ifconfig eth0 eth0 Link encap:Ethernet HWaddr FE:FD:C0:A8:0A:33 inet addr:192.168.10.51 Bcast:192.168.10.255 Mask:255.255.255.0 BROADCAST MULTICAST MTU:1500 Metric:1 RX packets:0 errors:0 dropped:0 overruns:0 frame:0 TX packets:0 errors:0 dropped:0 overruns:0 carrier:0 collisions:0 txqueuelen:100 RX bytes:0 (0.0 b) TX bytes:0 (0.0 b) Interrupt:5 bash-2.05# ifconfig eth0 up SIOCSIFFLAGS: Device or resource busy bash-2.05# So despite the fact, that the 1st ifconfig got a message, it did eat the IP address. But I cannot bring the eth0 interfce up. Hence of course I cannot ping the host nor can I ping the UML from the host. I do not really know what this message implies. I know its the "other end" of the bridge - is that why it will not allow me to set things, because the otther end is already up ? Thank you. Love Snoopy |
|
From: David C. <da...@da...> - 2002-11-07 16:29:40
|
Snoopy wrote: > This turned out to be a problem on my SuSe host: when I redid everything with a new tap1 interface, everything went fine. What exactly was the problem? Might be useful for others to know incase the experience the same thing. David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: Snoopy <sn...@do...> - 2002-11-08 08:49:27
|
On Thu, 07 Nov 2002 16:30:43 +0000 David Coulson <da...@da...> wrote: Dear David, dear list, gladly so, I just posted this message quickly so as not to send everyone off on a wild goose chase for nothing. In truth I really don't yet know why tap0 (the first one I made) is the problem. I found that I get this "Device or resource busy" not only when I try to ifconfig things within the uml but also on the host. So I tried to delete it with tunctl -d tap0 and it came back with the same error, except I was of course sure there was no uml "hanging on" to it or similar. I tried to tunctl -d it even after reboot, in single-user mode but to no avail. Its simply stuck there. Persists across reboots and everything. Very strange. So I will badger away at it more today and then as soon as I find out more I will gladly post it, because I also want to give back some help when I get some. of course. Love Snooopy > What exactly was the problem? Might be useful for others to know incase > the experience the same thing. > > David > > -- > David Coulson http://davidcoulson.net/ |
|
From: Net L. <net...@li...> - 2002-11-08 13:30:47
|
Did you add a call in modules.conf? On Fri, 8 Nov 2002, Snoopy wrote: > On Thu, 07 Nov 2002 16:30:43 +0000 > David Coulson <da...@da...> wrote: > > Dear David, dear list, > > gladly so, I just posted this message quickly so as not to send everyone off on a wild goose chase for nothing. > > In truth I really don't yet know why tap0 (the first one I made) is the problem. I found that I get this "Device or resource busy" not only when I try to ifconfig things within the uml but also on the host. So I tried to delete it with tunctl -d tap0 and it came back with the same error, except I was of course sure there was no uml "hanging on" to it or similar. > > I tried to tunctl -d it even after reboot, in single-user mode but to no avail. Its simply stuck there. Persists across reboots and everything. Very strange. > > So I will badger away at it more today and then as soon as I find out more I will gladly post it, because I also want to give back some help when I get some. of course. > > Love > Snooopy > > > > What exactly was the problem? Might be useful for others to know incase > > the experience the same thing. > > > > David > > > > -- > > David Coulson http://davidcoulson.net/ > > > ------------------------------------------------------- > This sf.net email is sponsored by: See the NEW Palm > Tungsten T handheld. Power & Color in a compact size! > http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en > _______________________________________________ > User-mode-linux-user mailing list > Use...@li... > https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user > -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Lonni J Friedman net...@li... Linux Step-by-step & TyGeMo http://netllama.ipfox.com |
|
From: Steven T. <to...@to...> - 2002-11-08 14:09:44
|
When booting my uml kernel I get near constant disk access on the host system. (Top shows about 80% of it's time in system 10% in user) I moved my root and swap to another drive so as to be sure they where not the cause of the access. Once it finally boots I can log in and if I try to compile a problem the drive light on the host again is nearly always on durring the ./configure portion. Needless to say this makes things extreamly slow. Does anyone have any idea what the cause could be? Thanks in advance. Steven |
|
From: David C. <da...@da...> - 2002-11-08 14:24:50
|
Steven Tower wrote: > Needless to say this makes things extreamly slow. Does anyone have any > idea what the cause could be? Is /tmp a filesystem on a hard disk? If so, mount it tmpfs and things should improve. UML uses /tmp to create 'physical' memory for the system. David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: Steven T. <to...@to...> - 2002-11-08 16:39:43
|
David, Thanks much. Fixed my problems. Steven On Fri, 2002-11-08 at 09:26, David Coulson wrote: > Steven Tower wrote: > > Needless to say this makes things extreamly slow. Does anyone have any > > idea what the cause could be? > > Is /tmp a filesystem on a hard disk? If so, mount it tmpfs and things > should improve. > > UML uses /tmp to create 'physical' memory for the system. > > David > > -- > David Coulson http://davidcoulson.net/ > d...@vi... http://journal.davidcoulson.net/ > > > > ------------------------------------------------------- > This sf.net email is sponsored by: See the NEW Palm > Tungsten T handheld. Power & Color in a compact size! > http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en > _______________________________________________ > User-mode-linux-user mailing list > Use...@li... > https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user |
|
From: Steven T. <to...@to...> - 2002-11-08 16:40:35
|
Can anyone point me to some documentation on skaz? What is it, is it automatically enabled, what do you need to do to enable it? Thanks, Steven On Fri, 2002-11-08 at 09:26, David Coulson wrote: > Steven Tower wrote: > > Needless to say this makes things extreamly slow. Does anyone have any > > idea what the cause could be? > > Is /tmp a filesystem on a hard disk? If so, mount it tmpfs and things > should improve. > > UML uses /tmp to create 'physical' memory for the system. > > David > > -- > David Coulson http://davidcoulson.net/ > d...@vi... http://journal.davidcoulson.net/ > > > > ------------------------------------------------------- > This sf.net email is sponsored by: See the NEW Palm > Tungsten T handheld. Power & Color in a compact size! > http://ads.sourceforge.net/cgi-bin/redirect.pl?palm0001en > _______________________________________________ > User-mode-linux-user mailing list > Use...@li... > https://lists.sourceforge.net/lists/listinfo/user-mode-linux-user |
|
From: David C. <da...@da...> - 2002-11-08 21:43:44
|
Steven Tower wrote: > Can anyone point me to some documentation on skaz? What is it, is it > automatically enabled, what do you need to do to enable it? Check Jeff's diary page on http://user-mode-linux.sf.net/. You need a kernel patch for the host in order to use it. Might be worth subscribing to user-mode-linux-devel if you're wanting to use skas. David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: Snoopy <sn...@do...> - 2002-11-08 15:41:41
|
On Fri, 8 Nov 2002 08:30:27 -0500 (EST) Net Llama! <net...@li...> wrote: > Did you add a call in modules.conf? > Well, I am not sure what you mean by that. I compiled my kernel with the TUN/TAP firml;y in place, i.e. not as a module but with a 'yes' in the right place. Ditto for the bridging and ethertap stuff. I know ethertap is obsolete but then I thought it might be a good idea to have a fall back. I am still uncertain abuot which interfaces need an IP address. In my host I have eth0, this I use to also talk to the outside world., its 192.168.10.8. My UMLs get something like 192.168.10.51, 52 etc. on their eth0, this is of course their end of the tap. However what about br0, the "bridge" ? What about tap0 ? I would have thought those need IPs too ? And basically if my UMLs cannot talk to the outside world much because when I do like described in the paper, then my hosts eth0 is set to IP 0.0.0.0 which is kind of useless. The tap0 end of the bridge br0 gets this IP address, but obviously the packets sent to that bridge do not come out of eth0 because otherwise I should still see the outside. So I thought host eth0 has to remain as is, tap0 should get a different IP and so should the UML end of the tap, this is clear. I am a bear of small brain and long words bother me...(Winnie the Pooh) Love, Snoopy |
|
From: Net L. <net...@li...> - 2002-11-08 15:50:42
|
On Fri, 8 Nov 2002, Snoopy wrote: > On Fri, 8 Nov 2002 08:30:27 -0500 (EST) > Net Llama! <net...@li...> wrote: > > > Did you add a call in modules.conf? > > > > Well, I am not sure what you mean by that. I compiled my kernel with the TUN/TAP firml;y in place, i.e. not as a module but with a 'yes' in the right place. Ditto for the bridging and ethertap stuff. I know ethertap is obsolete but then I thought it might be a good idea to have a fall back. PLease fix your linewrap. Right now you're at about 400 char/line, instead of the standard 72. I think the problem is your monolithic kernel. If tun/tap support is compiled in (or 'firmly in place' as you put it) then its going to be there automatically at bootup. If you built it as a module, then perhaps you'd have greater flexibilitiy. Modules get called in modules.conf. > > I am still uncertain abuot which interfaces need an IP address. Depends on your setup. eth0 inside the UML does. If you're not playing with bridging, then all the interfaces on the HOST need IPs. > > In my host I have eth0, this I use to also talk to the outside world., its 192.168.10.8. > > My UMLs get something like 192.168.10.51, 52 etc. on their eth0, this is of course their end of the tap. > > However what about br0, the "bridge" ? What about tap0 ? See http://sxs.sf.net/uml.html -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Lonni J Friedman net...@li... Linux Step-by-step & TyGeMo http://netllama.ipfox.com |
|
From: Ian C. <mai...@ic...> - 2002-11-08 16:21:14
|
Hello, > See http://sxs.sf.net/uml.html "File Not Found" Bye for Now, Ian |
|
From: Net L. <net...@li...> - 2002-11-08 16:30:26
|
Sorry, try http://sxs.sf.net/sxs/uml.html On Fri, 8 Nov 2002, Ian Chilton wrote: > Hello, > > > See http://sxs.sf.net/uml.html > > "File Not Found" > > > Bye for Now, > > Ian > -- ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Lonni J Friedman net...@li... Linux Step-by-step & TyGeMo http://netllama.ipfox.com |
|
From: Ian C. <ia...@ic...> - 2002-11-08 16:33:57
|
Hello, > Sorry, try http://sxs.sf.net/sxs/uml.html That works :) Thanks Ian |
|
From: Steven T. <to...@to...> - 2002-11-08 19:57:41
|
Are there any tips/tricks for performance optimizations for uml sessions? For example is one file system better then another concerning performance and stability what are the trade offs when used for the root_fs. What are the benefits of compiling it with SMP support and without? Thanks for any help, Steven |
|
From: Steven P. <st...@si...> - 2002-11-08 20:16:25
|
On Fri, Nov 08, 2002 at 02:56:31PM -0500, Steven Tower wrote: > Are there any tips/tricks for performance optimizations for uml > sessions? The first thing you need to do is make sure the TMPDIR environment variable points to a tmpfs directory. Generally speaking, it's a bad idea to use tmpfs for /tmp, since it breaks things that use loop mounts like mkinitrd. Steve -- st...@si... | Southern Illinois Linux Users Group (618)398-7360 | See web site for meeting details. Steven Pritchard | http://www.silug.org/ |
|
From: David C. <da...@da...> - 2002-11-08 21:50:47
|
Steven Tower wrote: > Are there any tips/tricks for performance optimizations for uml > sessions? 1) Mount /tmp as a tmpfs filesystem 2) For lots of UMLs, use COW with a single backing store 3) Remember - UMLs can have their own swap space 4) Use skas (and cross fingers it doesn't break on you) 5) Lots of RAM (to stop host swapping) and fast CPUs 6) More of #5. Even more RAM. David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: Ian C. B. <icb...@nk...> - 2002-11-08 23:09:39
|
On Friday 08 November 2002 16:53, David Coulson wrote: > Steven Tower wrote: > > Are there any tips/tricks for performance optimizations for uml > > sessions? > > 1) Mount /tmp as a tmpfs filesystem > 2) For lots of UMLs, use COW with a single backing store Using COWs with a single backing store only seems to impede performance, = not=20 improve it. If you're looking for performance, strongly reconsider unique= =20 images for each UML instance. --=20 - Ian C. Blenke <icb...@nk...> (This message bound by the following: http://www.nks.net/email_disclaimer.html) |
|
From: Matthew B. <ma...@by...> - 2002-11-08 23:26:13
|
On Friday 08 November 2002 23:09, Ian C. Blenke wrote:
> On Friday 08 November 2002 16:53, David Coulson wrote:
> > Steven Tower wrote:
> > > Are there any tips/tricks for performance optimizations for uml
> > > sessions?
> >
> > 1) Mount /tmp as a tmpfs filesystem
> > 2) For lots of UMLs, use COW with a single backing store
>
> Using COWs with a single backing store only seems to impede performance,
> not improve it. If you're looking for performance, strongly reconsider
> unique images for each UML instance.
Does anyone know why this is the case? Like I said earlier, I assumed that
having the kernel being able to cache a bunch of commonly-used disc blocks in
memory, especially if they were shared between many UMLs, would improve
performance. I can understand that COW discs may be slower intrinsically
because the kernel would have to potentially check two disc blocks rather
than one per read.
--
Matthew Bloch Bytemark Computer Consulting Limited
http://www.bytemark.co.uk/
tel. +44 (0) 8707 455026
|
|
From: David C. <da...@da...> - 2002-11-09 03:08:21
|
Ian C. Blenke wrote: > Using COWs with a single backing store only seems to impede performance, not > improve it. If you're looking for performance, strongly reconsider unique > images for each UML instance. Interesting - Ever tried to boot 25 UMLs at once using individual files and comparing it to a single backing store and COWs? David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: Snoopy <sn...@do...> - 2002-11-11 07:56:25
|
On Sat, 09 Nov 2002 03:10:49 +0000 David Coulson <da...@da...> wrote: > > Interesting - Ever tried to boot 25 UMLs at once using individual files > and comparing it to a single backing store and COWs? > Hmm, I have read the passages in the docs where this is claimed, but then I think that I should optimise my UMLs for run-time performance, not how fast they boot. When my server crshes and my disks need fscking, with 2 x 120gig on them, then it will take a few minutes for everything to be ready. And frankly I don't think it worries me then that my UMLs take a minute or two longer to come up. I wanted to definetely try some performance things with COWs vs. one file system per UML. Basically I think that the "usage pattern" will be different: people will upload some SW into their UMLs and then simply run it. So they will write a lot during the initial configuration phase of their UMl and then more or less leave it alone, perhaps put some Web Pages into their homes, doument roots whatever. But after an initial flurry of changes I would expect root and usr etc. to be fairly stable, hence intuitively I would wager the speed advantages of COW will simply diminish over time. And once you don't go to certain disk blocks they will fall out of the caches anyways and you have to refill the cache even using COW. Once I get some data I will gladly post it. Love, Snoopy |
|
From: David C. <da...@da...> - 2002-11-11 14:40:57
|
Snoopy wrote: > When my server crshes and my disks need fscking, with 2 x 120gig on them, then it will take a few minutes for everything to be ready. And frankly I don't think it worries me then that my UMLs take a minute or two longer to come up. Well, use a journaled filesystem. David -- David Coulson http://davidcoulson.net/ d...@vi... http://journal.davidcoulson.net/ |
|
From: James N. <jn...@nk...> - 2002-11-13 20:52:39
|
On Fri, 2002-11-08 at 22:10, David Coulson wrote: > Ian C. Blenke wrote: > > Using COWs with a single backing store only seems to impede performance, not > > improve it. If you're looking for performance, strongly reconsider unique > > images for each UML instance. > > Interesting - Ever tried to boot 25 UMLs at once using individual files > and comparing it to a single backing store and COWs? ..This seemed worth an experiment. I used sar to watch CPU time, and recorded the time it took to go from 100% idle to 100% idle. All systems were started at the same time. Boot time for 22 UML images, without COW files: 9m6s Boot time for 22 UML images, with COW files and one backing store: 3m7s -James |
|
From: Johannes F. <jfo...@mu...> - 2002-11-13 21:13:55
|
At 15:52 Uhr -0500 13.11.2002, James Neal wrote: >On Fri, 2002-11-08 at 22:10, David Coulson wrote: >> Ian C. Blenke wrote: >> > Using COWs with a single backing store only seems to impede >>performance, not >> > improve it. If you're looking for performance, strongly reconsider unique >> > images for each UML instance. >> >> Interesting - Ever tried to boot 25 UMLs at once using individual files >> and comparing it to a single backing store and COWs? > >..This seemed worth an experiment. > >I used sar to watch CPU time, and recorded the time it took to go from >100% idle to 100% idle. All systems were started at the same time. > >Boot time for 22 UML images, without COW files: 9m6s >Boot time for 22 UML images, with COW files and one backing store: 3m7s Could the good disk cache using COW play a role in this test? The Data is quite identical. So there might be other results when the caching has no such a huge effect. Greetings Johannes |
|
From: Rio M. <ro...@vb...> - 2002-11-13 09:49:04
|
Yes .. i have the same problem too.. eventhough we have bring interface tap0 down, tunctcl -d tap0 sometimes=20 results in error like that.. Regards, Rio Martin. On Wednesday 13 November 2002 15:46, Snoopy wrote: > YEs yes of course, > bringing it down, all the other ends of the bridge it was attached to, > everything I tried. > Its still there... > Aliens, simply just Aliens. Reopen the X-Files :-) > Love > Snoopy > On Wed, 13 Nov 2002 15:41:58 +0700 > Rio Martin <ro...@vb...> wrote: > > On Wednesday 13 November 2002 15:26, Snoopy wrote: > > > Dear List, > > > tunctl -d tap0 > > > - fails and comes back with Device or resource is busy. > > > > have you bring down interface tap0 ? > > try it .. > > > > Regards, > > Rio Martin. > > > > -- > > You will be surprised by a loud noise. --=20 Ask Not for whom the Bell Tolls, and You will Pay only the Station-to-Station rate. |
Thanks for helping keep SourceForge clean.
X