From: John P. <jo...@sn...> - 2006-04-04 18:51:57
|
Hi, I'm hoping someone can help guide me to a fully working UML setup. On the host, I'm running the AMD64 port of Debian Sarge with a custom kernel built from the standard kernel.org 2.6.16.1 source with the SKAS3-v8.2 patch applied. uname -a gives: Linux bedlam 2.6.16.1-jp5-skas3-v8.2 #1 SMP PREEMPT Fri Mar 31 12:34:36 BST 2006 x86_64 GNU/Linux The guest kernel is again built from the 2.6.16.1 source, in a seperate location (not /usr/src/linux) with the bb1 patch applied (but not the skas3 one). uname -a within the guest gives: Linux umlpcloc1 2.6.16.1-bb1 #1 Fri Mar 31 19:49:41 BST 2006 x86_64 GNU/Linux While the basic things work (i.e. I can boot into the UML, get networking going, use SSH etc) I am plagued with segfaults. I can't install firefox for example, it segfaults during the "firefox-bin -register" phase. I also get random segfaults when using "apt-get install". Xnest also works but frequently dies due to segfaults. When I boot the UML I notice it says I'm running in SKAS0 mode: ... Checking for the skas3 patch in the host: - /proc/mm...not found - PTRACE_FAULTINFO...not found - PTRACE_LDT...not found UML running in SKAS0 mode Linux version 2.6.16.1-bb1 (john@bedlam) (gcc version 3.4.4 20050314 (prerelease) (Debian 3.4.3-13)) #1 Fri Mar 31 19:49:41 BST 2006 ... Does skas3 work with AMD64? Am I missing something fundamental here? I have a 32-bit chroot on the system, and I've tried using that to build a 32-bit (686) UML but again I had problems with segfaults - not necessarily in the same places. I'm ripping out my hair here and would welcome any advice you have to offer. If you need any more information just let me know, Cheers, John |
From: Jeff D. <jd...@ad...> - 2006-04-04 21:30:30
|
On Tue, Apr 04, 2006 at 07:51:33PM +0100, John Parsons wrote: > Does skas3 work with AMD64? On x86_64, add noprocmm to the command line if it tries to run in skas3 mode. In your case, it's not seeing any skas3, so that's not an issue. As to why it's not seeing skas3, I don't know. /proc/mm should exist, for one thing. If that's not there, something went wrong with your build. > Am I missing something fundamental here? I have a > 32-bit chroot on the system, and I've tried using that to build a > 32-bit (686) UML but again I had problems with segfaults - not > necessarily in the same places. I'm ripping out my hair here and > would welcome any advice you have to offer. On i386 TLS is working well, but on x86_64, it's still dodgy. I don't know what this means for an i386 UML running on an x86_64 host, though. So, if you can avoid NTPL until this is fixed, that will help with the segfaults. Jeff |
From: David L. <da...@la...> - 2006-04-05 01:47:41
|
On Tue, 4 Apr 2006, Jeff Dike wrote: > On Tue, Apr 04, 2006 at 07:51:33PM +0100, John Parsons wrote: >> Does skas3 work with AMD64? > > On x86_64, add noprocmm to the command line if it tries to run in > skas3 mode. In your case, it's not seeing any skas3, so that's not an > issue. As to why it's not seeing skas3, I don't know. /proc/mm > should exist, for one thing. If that's not there, something went > wrong with your build. is this in the 2.6.16-rc1 patch? or does it still require an external patch. David Lang |
From: Blaisorblade <bla...@ya...> - 2006-04-05 14:02:09
|
On Wednesday 05 April 2006 03:47, David Lang wrote: > On Tue, 4 Apr 2006, Jeff Dike wrote: > > On Tue, Apr 04, 2006 at 07:51:33PM +0100, John Parsons wrote: > >> Does skas3 work with AMD64? > > On x86_64, add noprocmm to the command line if it tries to run in > > skas3 mode. In your case, it's not seeing any skas3, so that's not an > > issue. As to why it's not seeing skas3, I don't know. /proc/mm > > should exist, for one thing. If that's not there, something went > > wrong with your build. > is this in the 2.6.16-rc1 patch? You mean 2.6.17-rc1 I assume, right? Anyway the answer is the same. > or does it still require an external > patch. There's no chance whatsoever of SKAS3 being merged as-is in mainline. As I explained you, David, we'll have to rewrite the API to get SKAS4, and that'll be long; to be clearer, when it'll be merged we'll have to change UML to use it. -- Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!". Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894) http://www.user-mode-linux.org/~blaisorblade ___________________________________ Yahoo! Mail: gratis 1GB per i messaggi e allegati da 10MB http://mail.yahoo.it |
From: Blaisorblade <bla...@ya...> - 2006-04-05 14:17:53
|
On Tuesday 04 April 2006 22:31, Jeff Dike wrote: > On Tue, Apr 04, 2006 at 07:51:33PM +0100, John Parsons wrote: > > Does skas3 work with AMD64? > > On x86_64, add noprocmm to the command line if it tries to run in > skas3 mode. In your case, it's not seeing any skas3, so that's not an > issue. As to why it's not seeing skas3, I don't know. /proc/mm > should exist, for one thing. If that's not there, something went > wrong with your build. SKAS3-V8 has no SKAS3 support at all for amd64. You need -V9-pre for that. And even then, it only works for 32-bit UMLs. > > Am I missing something fundamental here? I have a > > 32-bit chroot on the system, and I've tried using that to build a > > 32-bit (686) UML but again I had problems with segfaults - not > > necessarily in the same places. I'm ripping out my hair here and > > would welcome any advice you have to offer. > On i386 TLS is working well, but on x86_64, it's still dodgy. I don't > know what this means for an i386 UML running on an x86_64 host, > though. I can confirm that with 2.6.16-bb1 and 2.6.17-rc1 for the UML kernel, an i386 UML on x86_64 host works well, even for TLS. > So, if you can avoid NTPL until this is fixed, that will help with the > segfaults. Indeed, I've not even started coding TLS for x86_64. -- Inform me of my mistakes, so I can keep imitating Homer Simpson's "Doh!". Paolo Giarrusso, aka Blaisorblade (Skype ID "PaoloGiarrusso", ICQ 215621894) http://www.user-mode-linux.org/~blaisorblade ___________________________________ Yahoo! Messenger with Voice: chiama da PC a telefono a tariffe esclusive http://it.messenger.yahoo.com |