[uml-devel] Re: Problems applying patches to 2.6.12-rc5-mm1

[uml-devel] Re: Problems applying patches to 2.6.12-rc5-mm1

[antoine]

Note: this is an attempt at running it on x86_64 host, so the errors
may or may not be the same...

I applied the patches on the website to 2.6.12-rc5 (mm1 and mm2: most
patches apply cleanly to both, some have been merged in mm2 obviously),
I got an error here:

gcc   -Wall -Wstrict-prototypes -Wno-trigraphs -fno-strict-aliasing
-fno-common -ffreestanding -O2 -fno-omit-frame-pointer
-fno-optimize-sibling-calls -g -U__x86_64__ -fno-builtin -D__arch_um__
-DSUBARCH=\"x86_64\" -Iarch/um/include
-I/usr/src/linux-2.6.12-rc5-mm2-uml/arch/um/kernel/tt/include
-D__x86_64__ -D_GNU_SOURCE -D_LARGEFILE64_SOURCE -c -o
arch/um/sys-x86_64/stub_segv.o arch/um/sys-x86_64/stub_segv.c
In file included from /usr/include/asm/ucontext.h:2,
                 from arch/um/sys-x86_64/stub_segv.c:8:
/usr/include/asm/../asm-x86_64/ucontext.h:4: error: redefinition of
`struct ucontext'

I removed the #include "ucontext", and added #include "uml-config.h" to
stub_segv.c. Then it got stuck on:
fixdep: arch/um/sys-x86_64/.stub_segv.o.d: No such file or directory
Re-running the build as suggested by Jeff did not help, but
creating an empty file to make it proceed seems to work...
`touch arch/um/sys-x86_64/.stub_segv.o.d`
(I haven't tried removing the av-skas patch yet)

Then it builds and almost runs:
(...)
[42949373.260000] kjournald starting.  Commit interval 5 seconds
[42949373.260000] EXT3-fs: mounted filesystem with ordered data mode.
[42949373.260000] VFS: Mounted root (ext3 filesystem) readonly.
[42949373.260000] Kernel panic - not syncing: Kernel mode fault at addr
0x0, ip 0x0
[42949373.260000]
[42949373.260000] Pid: 1, comm: bash Not tainted 2.6.12-rc5-mm2
[42949373.260000] RIP: 0000:[<0000000000000000>]
[42949373.260000] RSP: 0000000060453c30  EFLAGS: 00010202
[42949373.260000] RAX: 0000000000000001 RBX: 0000000000000000 RCX:
0000000060194ec1
[42949373.260000] RDX: 000000006002840d RSI: 0000000000000001 RDI:
0000000060453ad0
[42949373.260000] RBP: 0000000000000000 R08: 0000000000000000 R09:
0000000000000000
[42949373.260000] R10: 0000000000000008 R11: 0000000000000246 R12:
0000000000000000
[42949373.260000] R13: 0000000000000001 R14: 000000000000000a R15:
0000000000000000
[42949373.260000] Call Trace:
[42949373.260000] 604536d8:  [<6001575f>] panic_exit+0x2f/0x50
[42949373.260000] 604536f8:  [<60040c2b>] notifier_call_chain+0x2b/0x50
[42949373.260000] 60453728:  [<6002fe24>] panic+0xe4/0x180
[42949373.260000] 60453768:  [<60194ec1>] __sigprocmask+0x11/0x40
[42949373.260000] 60453788:  [<6010e0b3>] __up_read+0xb3/0xf0
[42949373.260000] 604537a8:  [<60014908>] handle_page_fault+0xc8/0x280
[42949373.260000] 604537d8:  [<6002a1ef>] __do_user_copy+0x4f/0xd0
[42949373.260000] 60453818:  [<60014caa>] segv+0x1ea/0x2b0
[42949373.260000] 60453828:  [<60194f85>] sigemptyset+0x15/0x40
[42949373.260000] 60453838:  [<60028d71>] change_sig+0x61/0x80
[42949373.260000] 604538a8:  [<60194f85>] sigemptyset+0x15/0x40
[42949373.260000] 604538b8:  [<60028de1>] change_signals+0x51/0x80
[42949373.260000] 60453928:  [<60015056>] segv_handler+0x106/0x120
[42949373.260000] 60453958:  [<60014f50>] segv_handler+0x0/0x120
[42949373.260000] 60453968:  [<60017fc9>] sig_handler_common_tt
+0x119/0x1b0
[42949373.260000] 604539d8:  [<60028a80>] sig_handler+0x10/0x20
[42949373.260000] 604539e8:  [<60194c40>] __restore_rt+0x0/0x10
[42949373.260000] 60453a78:  [<6002840d>] run_kernel_thread+0x2d/0x50
[42949373.260000] 60453a88:  [<60194ec1>] __sigprocmask+0x11/0x40
[42949373.260000] 60453ae0:  [<6000f1b0>] init+0x0/0x100
[42949373.260000] 60453b08:  [<6002840d>] run_kernel_thread+0x2d/0x50
[42949373.260000] 60453b88:  [<6000f1b0>] init+0x0/0x100
[42949373.260000] 60453b98:  [<60028e2e>] unblock_signals+0xe/0x10
[42949373.260000] 60453ba8:  [<60016456>] new_thread_handler+0x166/0x1a0
[42949373.260000] 60453cd8:  [<60194ec1>] __sigprocmask+0x11/0x40
[42949373.260000]
[42949373.260000]

Now, I am really stuck.
Any ideas?

Cheers
Antoine


On Sat, 2005-05-28 at 20:07 +0100, antoine wrote:
> Hi Jeff,
> 
> Now that the ptrace host bug has been fixed in 2.6.11.11, I can run
> tests on shared systems without taking too many risks, but I am a little
> bit lost as to which patches I would need to apply, between your
> patches, Blaisorblade's, al's,.. Which all seem to contain x86_64
> patches. Let me know what I can do to help testing. I am sure I can find
> some new bugs, in fact I am working on a Java VM bug atm.
> 
> Cheers
> Antoine
> 
> 
> On Mon, 2005-05-09 at 01:57 +0100, Antoine Martin wrote:
> > Crashing it is quite easy in fact, reiserfs as root_fs just won't work.
> > I don't think this trace is going to help, but here it is anyway:
> > 
> > [42949381.990000] REISERFS: panic (device Null superblock):
> > reiserfs[175]: assertion !( comp_keys( &MAX_KEY, p_s_key ) && !
> > key_in_buffer(p_s_search_path, p_s_key, p_s_sb) ) failed at
> > fs/reiserfs/stree.c:685:search_by_key: PAP-5130: key is not in the
> > buffer
> > [42949381.990000] Kernel panic - not syncing: BUG!
> > [42949381.990000]  [42949381.990000]
> > 
> > 
> > On Sun, 2005-05-08 at 23:20 +0100, Antoine Martin wrote:
> > > All good, builds fine and hasn't crashed yet!
> > > I'll get back to you with (hopefully) more useful debugging information
> > > as soon as Andi Kleen comes up with a patch for the host bug.
> > > 
> > > Antoine
> > > 
> > > On Sun, 2005-05-08 at 14:35 -0400, Jeff Dike wrote:
> > > > I uploaded a new set of patches which patches, builds, and runs on both
> > > > i386 and x86_64.
> > > > 
> > > > On Sun, May 08, 2005 at 06:37:19PM +0100, Antoine Martin wrote:
> > > > > I noticed that you updated the website with patches against -rc4 so I
> > > > > built the beast, As before, I had to do this to fix it:
> > > > > cd include/asm
> > > > > ln -sf elf-x86_64.h elf.h
> > > > 
> > > > I get an empty elf.h which obviously causes build problems.  Removing it
> > > > and rebuilding fixes it.  I haven't looked at this yet.
> > > > 
> > > > > I also had a reject on:
> > > > > arch/um/sys-x86_64/Makefile
> > > > 
> > > > This should be OK now.
> > > > 
> > > > 				Jeff



-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[Jeff Dike]

On Thu, Jun 02, 2005 at 06:51:22AM +0100, antoine wrote:
> I applied the patches on the website to 2.6.12-rc5 (mm1 and mm2: most
> patches apply cleanly to both, some have been merged in mm2 obviously),
> I got an error here:

I pushed out a fixed set of patches (against -mm2) that work for me.

There were some include changes in stub_segv.c.

The .d file thing is still there.  That seems to have been caused by my 
changing stub_segv.c to a userspace file.

				Jeff


-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel

[uml-devel] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[antoine]

Thanks, almost all applied cleanly - I skipped the s390 ones for now.
Compiling still requires a `touch arch/um/sys-x86_64/.stub_segv.o.d` to
make it proceed. Works for me! (tm)
But I can't get it to boot into init properly, I can only boot with
init=/bin/bash. (then remount root rw, mount proc)

Every time I run a command, I get something like:
bash: child setpgid (15484 to 15484): No such process
(I remember seeing this one before - but I can't remember the
solution :-(

I can bring the network up manually (ifconfig) and ping (in and out),
but running any of the /etc/init.d/ scripts locks up, same goes for
starting sshd or even "Hello World" in hello.sh

Sometimes when it goes into a spin, it loos on this (to infinity):
"[42949450.500000] fix_range_common: failed, killing current process"

Not sure where do go from here... Maybe patching with just the minimum?
It needs roughly at least up to fix-tt-USR1-handlers to be able to boot
(otherwise you get the error I posted yesterday).

Antoine




On Thu, 2005-06-02 at 17:06 -0400, Jeff Dike wrote:
> On Thu, Jun 02, 2005 at 06:51:22AM +0100, antoine wrote:
> > I applied the patches on the website to 2.6.12-rc5 (mm1 and mm2: most
> > patches apply cleanly to both, some have been merged in mm2 obviously),
> > I got an error here:
> 
> I pushed out a fixed set of patches (against -mm2) that work for me.
> 
> There were some include changes in stub_segv.c.
> 
> The .d file thing is still there.  That seems to have been caused by my 
> changing stub_segv.c to a userspace file.
> 
> 				Jeff



-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
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] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[Jeff Dike]

On Fri, Jun 03, 2005 at 03:33:36AM +0100, antoine wrote:
> Thanks, almost all applied cleanly - I skipped the s390 ones for now.

I wouldn't cherry-pick patches.  You may be right, but it also means you're
running different code than me.

> I can bring the network up manually (ifconfig) and ping (in and out),
> but running any of the /etc/init.d/ scripts locks up, same goes for
> starting sshd or even "Hello World" in hello.sh

Dunno, it works here, and I've beat on it a bunch.  If you can get stack
traces or anything from the hangs, that might help.

				Jeff


-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
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] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[antoine]

On Thu, 2005-06-02 at 22:38 -0400, Jeff Dike wrote:
> On Fri, Jun 03, 2005 at 03:33:36AM +0100, antoine wrote:
> > Thanks, almost all applied cleanly - I skipped the s390 ones for now.
> 
> I wouldn't cherry-pick patches.  You may be right, but it also means you're
> running different code than me.
I wouldn't normally do that, but there is a patch that does not apply
cleanly: fp-state (can't find file to patch at input line 28 ...)
So I started thinking that there were things I did not necessarily need
or want (s390, x11fb, audio, etc) - I guess I was wrong...

I started again, applying everything else, some files complained about
missing skas.h or proc_mm.h, another case of empty file, easy fix:
rm ./arch/um/kernel/skas/include/skas.h

I got tired of doing the same steps over and over again (to be sure I
wasn't reporting my own mistakes) so I made this very simple script to
untar kernel source + apply MM + apply UML patches:
http://213.228.237.37/uml/2.6.12-rc5-mm2/make-uml64.sh
I've also dumped the compile errors, kernel binary, .config in the same
directory.

Everything kinda works now! And it is *very* fast (compared to TT x86
guest) And this time it is a bit more stable.
I think there is still a memory leak lurking, OOM seems to kick in (see
bottom of this email)

Very minor thing: I noticed that the TT threads generally do not show
which process they are running inside the UML (as it used to), any way
to restore this?

Thanks
Antoine


> 
> > I can bring the network up manually (ifconfig) and ping (in and out),
> > but running any of the /etc/init.d/ scripts locks up, same goes for
> > starting sshd or even "Hello World" in hello.sh
> 
> Dunno, it works here, and I've beat on it a bunch.  If you can get stack
> traces or anything from the hangs, that might help.
> 
> 				Jeff
*********
* OOM
* 128RAM + 128swap should be enough to emerge things...
*********

[42951416.180000] oom-killer: gfp_mask=0xd2
[42951416.180000] DMA per-cpu:
[42951416.180000] cpu 0 hot: low 30, high 90, batch 15
[42951416.180000] cpu 0 cold: low 0, high 30, batch 15
[42951416.180000] Normal per-cpu: empty
[42951416.180000] HighMem per-cpu: empty
[42951416.180000]
[42951416.180000] Free pages:        1432kB (0kB HighMem)
[42951416.180000] Active:166 inactive:333 dirty:0 writeback:0 unstable:0
free:358 slab:999 mapped:82 pagetables:9547
[42951416.180000] DMA free:1432kB min:1448kB low:1808kB high:2172kB
active:664kB inactive:1332kB present:131072kB pages_scanned:1301
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] Normal free:0kB min:0kB low:0kB high:0kB active:0kB
inactive:0kB present:0kB pages_scanned:0
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] HighMem free:0kB min:128kB low:160kB high:192kB
active:0kB inactive:0kB present:0kB pages_scanned:0
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] DMA: 0*4kB 5*8kB 1*16kB 1*32kB 1*64kB 0*128kB 1*256kB
0*512kB 1*1024kB 0*2048kB 0*4096kB = 1432kB
[42951416.180000] Normal: empty
[42951416.180000] HighMem: empty
[42951416.180000] Swap cache: add 314384, delete 314009, find
48066/104534, race 0+0
[42951416.180000] Free swap  = 112048kB
[42951416.180000] Total swap = 131064kB
[42951416.180000] Out of Memory: Killed process 978 (emerge).
[42951416.180000] oom-killer: gfp_mask=0xd2
[42951416.180000] DMA per-cpu:
[42951416.180000] cpu 0 hot: low 30, high 90, batch 15
[42951416.180000] cpu 0 cold: low 0, high 30, batch 15
[42951416.180000] Normal per-cpu: empty
[42951416.180000] HighMem per-cpu: empty
[42951416.180000]
[42951416.180000] Free pages:        1432kB (0kB HighMem)
[42951416.180000] Active:166 inactive:333 dirty:0 writeback:0 unstable:0
free:358 slab:1000 mapped:82 pagetables:9519
[42951416.180000] DMA free:1432kB min:1448kB low:1808kB high:2172kB
active:664kB inactive:1332kB present:131072kB pages_scanned:1301
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] Normal free:0kB min:0kB low:0kB high:0kB active:0kB
inactive:0kB present:0kB pages_scanned:0
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] HighMem free:0kB min:128kB low:160kB high:192kB
active:0kB inactive:0kB present:0kB pages_scanned:0
[42951416.180000] lowmem_reserve[]: 0 0 0
[42951416.180000] DMA: 0*4kB 5*8kB 1*16kB 1*32kB 1*64kB 0*128kB 1*256kB
0*512kB 1*1024kB 0*2048kB 0*4096kB = 1432kB
[42951416.180000] Normal: empty
[42951416.180000] HighMem: empty
[42951416.180000] Swap cache: add 314384, delete 314009, find
48066/104535, race 0+0
[42951416.180000] Free swap  = 123236kB
[42951416.180000] Total swap = 131064kB
[42951416.180000] Out of Memory: Killed process 5713 (portageq).




-------------------------------------------------------
This SF.Net email is sponsored by Yahoo.
Introducing Yahoo! Search Developer Network - Create apps using Yahoo!
Search APIs Find out how you can build Yahoo! directly into your own
Applications - visit http://developer.yahoo.net/?fr=offad-ysdn-ostg-q22005
_______________________________________________
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] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[Jeff Dike]

On Fri, Jun 03, 2005 at 08:28:10PM +0100, antoine wrote:
> I think there is still a memory leak lurking, OOM seems to kick in (see
> bottom of this email)

Yeah, it's a page leak, I think from page tables, which I haven't tracked
down yet.

> Very minor thing: I noticed that the TT threads generally do not show
> which process they are running inside the UML (as it used to), any way
> to restore this?

Those aren't tt threads, they are skas0 threads, and we could probably restore 
the command names, but I haven't bothered doing so.

				Jeff


-------------------------------------------------------
This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
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] Re: 2.6.11.8 + UML/x86_64 (2.6.12-rc3+) = oops

[Bodo Stroesser]

Jeff Dike wrote:
> On Fri, Jun 03, 2005 at 08:28:10PM +0100, antoine wrote:
> 
>>I think there is still a memory leak lurking, OOM seems to kick in (see
>>bottom of this email)
> 
> 
> Yeah, it's a page leak, I think from page tables, which I haven't tracked
> down yet.
> 
> 
>>Very minor thing: I noticed that the TT threads generally do not show
>>which process they are running inside the UML (as it used to), any way
>>to restore this?
> 
> 
> Those aren't tt threads, they are skas0 threads, and we could probably restore 
> the command names, but I haven't bothered doing so.
I thought about doing so, as the output of ps really looks bad.
Unfortunately, I found only very nasty ways.
On older host's, we could try to expand or strip down the environment in a similar
way as is used it tt. Then, names could be placed into stub-(stack?)-page.
Nasty, but it could work.
An even worse case is a host using stack-randomization. We need a pointer in the
host poining to stub-pages. If the host has left a "hole" above the stack, would
we have to "exec" until accidentally we get a stack starting at upper address-range
limit?
		Bodo
> 
> 				Jeff
> 
> 
> -------------------------------------------------------
> This SF.Net email is sponsored by: NEC IT Guy Games.  How far can you shotput
> a projector? How fast can you ride your desk chair down the office luge track?
> If you want to score the big prize, get to know the little guy.  
> Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
> _______________________________________________
> 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: NEC IT Guy Games.  How far can you shotput
a projector? How fast can you ride your desk chair down the office luge track?
If you want to score the big prize, get to know the little guy.  
Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel