Zhu Tom-a2289c | 1 Aug 03:37 2005

RE: NPTL pthread_create() cannot work.

> Yes, I already fixed that recently - that patch is a development one and you're the first tester I heard of.
:-)

> I've succeeded in fixing even this...
Can I get your fix for the further test?

> Is that compiled with NPTL / TLS support? Has 2.3.2 any TLS support at all?
It is said to support NPTL 0.29.

Thanks,
Tom

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
Phill Wombat | 1 Aug 04:53 2005

sysctl -w va.legacy_va_layout=1

Hi Paolo,
 
Looking into another issue (my son's gaming machine - counter-strike source on linux) I came across this interesting parameter whereby winex/Cedega won't run steam unless legacy support for legacy VA layout is enabled on a 2.6.9 kernel.
 
Well, I thought, I'd better try that out......
 
With 2.6.12+bs9 on x86_64 the error now changes
 
On the host set:
sysctl -w va.legacy_va_layout=1
sysctl -w kernel.randomize_va_space=0
 
Starting the uml with init=/bin/sh starts up as usual. Performing the ls command now result in a different thing happening:
Bad op in do_proc_op
 
Where as before it produced:
Child 22915 exited with signal 11

Furthermore, Ctrl-C shuts everything down cleanly. Don't the significance if any of all this, but if feels better.
 
Conducting the same test on 32bit 2.6.12+bs9 results in total disaster (feels very bad):
 
<0>Kernel panic - not syncing: fix_range fixing wrong address space, current=0xabebf6c0
folloed by lost of register dumps and traceback stuff (2 lots actually - which I take to mean it really really crashed).
 
I can only get init=/bin/sh to go when kernel.randomize_va_space=1 and vm.legacy_va_layout=0
Running the ls command then results in a message similar to:
Child 22915 exited with signal 11
 
Perhaps the only thing of value in all this is va.legacy_va_layout which may be useful to UML. It seems to do something, I just don't know what.
 
Cheers
Phill.
 
>Hi Paolo,
>
>>Try also following this (from another email) - it appeared in vanilla
>>2.6.12
>>and may be in your FC4 kernel. Not sure but it may be playing a role
>>here:
>>
>>"It looks like VA space randomization is the culprit.  The problem only
>>shows up when the host is 2.6.12, and can be solved by setting the
>>kernel.randomize_va_space sysctl to zero.  Thanks for the pointer
>>there."
>
>Here are the results with sysctl -w kernel.randomize_va_space=0.
>Essentially no effect observed.

Srivatsa Vaddagiri | 3 Aug 11:48 2005
Picon

Host kernel requirements for UML

Hi,
	I am trying to get a quick answer for a question I had on UML.
As I understand, UML kernel makes use of host kernel's 
system call interface for its functionality. The questions is, does the
host kernel need to support *all* system calls or only a subset?
Does UML rely on /proc or any other such filesystems on host?
This is assuming that we want all possible features inside UML.

--

-- 

Thanks and Regards,
Srivatsa Vaddagiri,
Linux Technology Center,
IBM Software Labs,
Bangalore, INDIA - 560017

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
Jeff Dike | 3 Aug 16:39 2005

Re: Host kernel requirements for UML

On Wed, Aug 03, 2005 at 03:18:05PM +0530, Srivatsa Vaddagiri wrote:
> Hi,
> 	I am trying to get a quick answer for a question I had on UML.
> As I understand, UML kernel makes use of host kernel's 
> system call interface for its functionality. The questions is, does the
> host kernel need to support *all* system calls or only a subset?

Only a small subset.

> Does UML rely on /proc or any other such filesystems on host?

It looks at /proc/cpuinfo for some information about cmov and xmm support
on the processor.

				Jeff

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
antoine | 3 Aug 19:40 2005
Picon

Re: 2.6.12-bs7 vs 2.6.12-bb8

> > I've just stared testing your latest patches on a amd64 box running
> > 2.6.13-rc3-git4, bb8 seems to run (so far)
> Using TT or SKAS0 mode? And is this specific to that host release wrt 2.6.12?
> > but bs7 comes up with: 
(...)
bs works fine on 2.6.13-rc4 now, must have been a host bug?
But all guest 64-bit kernels are still loosing memory fast, gentoo's
"emerge --sync" is enough to eat 300+ MB of ram! (in about 5 minutes)

Antoine

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
Blaisorblade | 3 Aug 20:03 2005
Picon

Re: NPTL pthread_create() cannot work.

On Tuesday 02 August 2005 12:05, Zhu Tom-a2289c wrote:
> Hi Blaisor,
>
> I see your new patch on
> http://www.user-mode-linux.org/~blaisorblade/patches/devel-guest/uml-2.6.13
>-rc3-git-bb-20050731-19:08/ Does this fix the problem?

> I used vinilla 2.6.13-rc3 kernel, no matter whether adding git patch,
> I cannot apply all the patches.
Applying *all* the patches won't work, but until *TLS* it should
> Could you indicate which kernel version I should base?
Well, that should be something along -rc4-git3/4, I gave some indications on 
the "NEWS page".
> Thanks,
> Tom
>
>
>
> -----Original Message-----
> From: Zhu Tom-a2289c
> Sent: 2005年8月1日 9:37
> To: 'Blaisorblade'
> Cc: user-mode-linux-devel <at> lists.sourceforge.net
> Subject: RE: [uml-devel] NPTL pthread_create() cannot work.
>
> > Yes, I already fixed that recently - that patch is a development one and
> > you're the first tester I heard of.
> >
> :-)
> :
> > I've succeeded in fixing even this...
>
> Can I get your fix for the further test?
>
> > Is that compiled with NPTL / TLS support? Has 2.3.2 any TLS support at
> > all?
>
> It is said to support NPTL 0.29.
>
> Thanks,
> Tom

--

-- 
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

-------------------------------------------------------
SF.Net email is sponsored by: Discover Easy Linux Migration Strategies
from IBM. Find simple to follow Roadmaps, straightforward articles,
informative Webcasts and more! Get everything you need to get up to
speed, fast. http://ads.osdn.com/?ad_id=7477&alloc_id=16492&op=click
Srivatsa Vaddagiri | 3 Aug 19:15 2005
Picon

Re: Host kernel requirements for UML

On Wed, Aug 03, 2005 at 10:39:57AM -0400, Jeff Dike wrote:
> On Wed, Aug 03, 2005 at 03:18:05PM +0530, Srivatsa Vaddagiri wrote:
> > Hi,
> > 	I am trying to get a quick answer for a question I had on UML.
> > As I understand, UML kernel makes use of host kernel's 
> > system call interface for its functionality. The questions is, does the
> > host kernel need to support *all* system calls or only a subset?
> 
> Only a small subset.

Thanks for the info. Is there any place where this subset is documented?
I was hoping to avoid looking at the unresoled symbols in UML binary to 
get this info!

--

-- 

Thanks and Regards,
Srivatsa Vaddagiri,
Linux Technology Center,
IBM Software Labs,
Bangalore, INDIA - 560017

-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
antoine | 4 Aug 20:56 2005
Picon

Re: 2.6.12-bs7 vs 2.6.12-bb8

On Wed, 2005-08-03 at 18:40 +0100, antoine wrote:
> > > I've just stared testing your latest patches on a amd64 box running
> > > 2.6.13-rc3-git4, bb8 seems to run (so far)
> > Using TT or SKAS0 mode? And is this specific to that host release wrt 2.6.12?
> > > but bs7 comes up with: 
> (...)
> bs works fine on 2.6.13-rc4 now, must have been a host bug?
> But all guest 64-bit kernels are still loosing memory fast, gentoo's
> "emerge --sync" is enough to eat 300+ MB of ram! (in about 5 minutes)
I can get a similar behavior from 2.6.12.3-bs9 guest (compiled with
SUBARCH=i386) on a 2.6.13-rc5 host!
The strange thing is that the same config file, same guest source but
compiled without SUBARCH (ie: regular 64-bit guest) works ok (well still
loosing memory fast - but at least it runs)

strace ./vmlinux-2.6.12.3-bs9-x86 gives:
(...)
open("/home/antoine/.uml/pX0UwE/pid", O_RDWR|O_CREAT|O_EXCL|O_LARGEFILE,
0644) = 6
getpid()                                = 18680
write(6, "18680\n", 6)                  = 6
close(6)                                = 0
mprotect(0xa0256000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0
write(1, "OK\n", 3)                     = 3
rt_sigaction(SIGPIPE, {0x1000000000000001, [], 0}, {SIG_DFL}, 8) = 0
socketcall(0x8, 0xffffcb70)             = 0
fcntl64(6, F_SETFD, FD_CLOEXEC)         = 0
fcntl64(7, F_SETFD, FD_CLOEXEC)         = 0
rt_sigaction(SIGWINCH, {0x10000000a0019a60, [], 0}, {SIG_DFL}, 8) = 0
getpid()                                = 18680
clone(child_stack=0xa0257fd4, flags=CLONE_FILES|SIGCHLD) = 18681
waitpid(18681, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGSTOP}], WSTOPPED) =
18681
--- SIGCHLD (Child exited)  <at>  0 (0) ---
ptrace(0x15 /* PTRACE_??? */, 18681, 0, 0x1) = 0
ptrace(PTRACE_CONT, 18681, 0, SIG_0)    = 0
rt_sigaction(SIGSEGV, {0x10000000a0019be0, [], 0}, {SIG_DFL}, 8) = 0
rt_sigaction(SIGUSR1, {0x10000000a00184f0, [], 0}, {SIG_DFL}, 8) = 0
waitpid(4294967295, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGSEGV}],
WSTOPPED) = 18681
--- SIGCHLD (Child exited)  <at>  0 (0) ---
ptrace(PTRACE_CONT, 18681, 0, SIGSEGV)  = 0
waitpid(4294967295, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGUSR1}],
WSTOPPED) = 18681
--- SIGCHLD (Child exited)  <at>  0 (0) ---
munmap(0, 2692743168)                   = 0
--- SIGSEGV (Segmentation fault)  <at>  0 (0) ---
--- SIGSEGV (Segmentation fault)  <at>  0 (0) ---
+++ killed by SIGSEGV +++

Antoine

-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
Blaisorblade | 4 Aug 20:22 2005
Picon

Re: Host kernel requirements for UML

On Wednesday 03 August 2005 19:15, Srivatsa Vaddagiri wrote:
> On Wed, Aug 03, 2005 at 10:39:57AM -0400, Jeff Dike wrote:
> > On Wed, Aug 03, 2005 at 03:18:05PM +0530, Srivatsa Vaddagiri wrote:
> > > Hi,
> > > 	I am trying to get a quick answer for a question I had on UML.
> > > As I understand, UML kernel makes use of host kernel's
> > > system call interface for its functionality. The questions is, does the
> > > host kernel need to support *all* system calls or only a subset?

> > Only a small subset.

> Thanks for the info. Is there any place where this subset is documented?
Surely not. However, that *does not* yet include any fancy 2.6 features. Why 
are you looking for it? What are you going to port/run UML to?

> I was hoping to avoid looking at the unresoled symbols in UML binary to
> get this info!

That wouldn't work... how would you discover that recent UMLs require the host 
to support PTRACE_O_SYSGOOD (a ptrace option set through PTRACE_SETOPTIONS)?

That's only in recent ones and that support has been added in 2.4, so it's 
safe.

Hey, if you look at the man page, you won't find out what they're about, 
either!

For the core, it just requires (*heavily*) mmap and friends, standard 
filesystem support (possibly including largefile-aware versions) and signals. 
But it extends a lot for terminal support, for instance it will make the 
stdin/stdout terminal "raw" with cfmakeraw(3), probably through some IOCTLs, 
bundled inside libraries...
--

-- 
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

-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf
Blaisorblade | 4 Aug 20:48 2005
Picon

Re: 2.6.12-bs7 vs 2.6.12-bb8

On Thursday 04 August 2005 20:56, antoine wrote:
> On Wed, 2005-08-03 at 18:40 +0100, antoine wrote:
> > > > I've just stared testing your latest patches on a amd64 box running
> > > > 2.6.13-rc3-git4, bb8 seems to run (so far)
> > >
> > > Using TT or SKAS0 mode? And is this specific to that host release wrt
> > > 2.6.12?
> > >
> > > > but bs7 comes up with:
> >
> > (...)
> > bs works fine on 2.6.13-rc4 now, must have been a host bug?
> > But all guest 64-bit kernels are still loosing memory fast, gentoo's
> > "emerge --sync" is enough to eat 300+ MB of ram! (in about 5 minutes)
>
> I can get a similar behavior from 2.6.12.3-bs9 guest (compiled with
> SUBARCH=i386) on a 2.6.13-rc5 host!
Ehr - can you bug Jeff for these? After trying on a 2.6.12.3 host - I wouldn't 
call -rc5 "stable", by just looking at the changelog *after* the release.

I'm going to be *offline* so unable to handle this stuff.
> The strange thing is that the same config file, same guest source but
> compiled without SUBARCH (ie: regular 64-bit guest) works ok (well still
> loosing memory fast - but at least it runs)
>
> strace ./vmlinux-2.6.12.3-bs9-x86 gives:
> (...)
Which test is it performing when it locks up? It is printing "ok" below, so...

> open("/home/antoine/.uml/pX0UwE/pid", O_RDWR|O_CREAT|O_EXCL|O_LARGEFILE,
> 0644) = 6
> getpid()                                = 18680
> write(6, "18680\n", 6)                  = 6
> close(6)                                = 0
> mprotect(0xa0256000, 8192, PROT_READ|PROT_WRITE|PROT_EXEC) = 0

//Here
> write(1, "OK\n", 3)                     = 3

> rt_sigaction(SIGPIPE, {0x1000000000000001, [], 0}, {SIG_DFL}, 8) = 0
> socketcall(0x8, 0xffffcb70)             = 0
> fcntl64(6, F_SETFD, FD_CLOEXEC)         = 0
> fcntl64(7, F_SETFD, FD_CLOEXEC)         = 0
> rt_sigaction(SIGWINCH, {0x10000000a0019a60, [], 0}, {SIG_DFL}, 8) = 0
> getpid()                                = 18680
> clone(child_stack=0xa0257fd4, flags=CLONE_FILES|SIGCHLD) = 18681
> waitpid(18681, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGSTOP}], WSTOPPED) =
> 18681
> --- SIGCHLD (Child exited)  <at>  0 (0) ---
> ptrace(0x15 /* PTRACE_??? */, 18681, 0, 0x1) = 0
> ptrace(PTRACE_CONT, 18681, 0, SIG_0)    = 0
> rt_sigaction(SIGSEGV, {0x10000000a0019be0, [], 0}, {SIG_DFL}, 8) = 0
> rt_sigaction(SIGUSR1, {0x10000000a00184f0, [], 0}, {SIG_DFL}, 8) = 0
Ok, the below is actually a 32-bit "-1" value (not verified), so it shouldn't 
be bogus.
> waitpid(4294967295, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGSEGV}],
> WSTOPPED) = 18681
> --- SIGCHLD (Child exited)  <at>  0 (0) ---
> ptrace(PTRACE_CONT, 18681, 0, SIGSEGV)  = 0
Same here
> waitpid(4294967295, [{WIFSTOPPED(s) && WSTOPSIG(s) == SIGUSR1}],
> WSTOPPED) = 18681
> --- SIGCHLD (Child exited)  <at>  0 (0) ---
The below *is* IMHO bogus (munmap from 0 is useless).
> munmap(0, 2692743168)                   = 0

> --- SIGSEGV (Segmentation fault)  <at>  0 (0) ---
> --- SIGSEGV (Segmentation fault)  <at>  0 (0) ---
> +++ killed by SIGSEGV +++

> Antoine

--

-- 
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

-------------------------------------------------------
SF.Net email is Sponsored by the Better Software Conference & EXPO
September 19-22, 2005 * San Francisco, CA * Development Lifecycle Practices
Agile & Plan-Driven Development * Managing Projects & Teams * Testing & QA
Security * Process Improvement & Measurement * http://www.sqe.com/bsce5sf

Gmane