[uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[BlaisorBlade]

I've started working also on UML/2.4.

Today I've split the changes between 2.4.24-1um and 2.4.24-2um into several 
patches; the HUMFS ones are carefully split out, but I've separated even the 
others. Now I need testers. Lots of people cannot use UML patches later than 
2.4.24-1um, since they contain a lot of Hostfs bugs and some stability 
issues. Now, if you use just some patches, you'll avoid the bugs in the 
other.

You should apply surely the first 3 patches, IMHO; the other ones should not 
hurt; inside txt/ there is a description of them.

mini-changes.patch
  Various little changes which don't hurt.

use_setjmp_wrapper.patch
  Adds setjmp_wrapper and use it, to fix klogd bug.

use-gettimeofday.patch
  Replace rdtsc with gettimeofday.

remapper-fix.patch
  Fixes probably to the ubd-mmap code

SIG-fileHandle-changes.patch
  Cope with signals from various places.

exec-shield.patch
  Patch for exec-shield users (IMHO unsafe)

expand-os-for-HUMFS.patch
  Expands os_ layer, seems to be for HUMFS only.

HUMFS.patch
  Adds HUMFS support, but create problems for Hostfs


The actual patches can be found inside the patches/ subdir in the tarball, and 
the "series" file contains the order.
THEY ARE AGAINST 2.4.24-1uml. If you want to apply it onto 2.4.26, don't 
forget to include the following patch (just for 2.4.26): 
http://vs165141.v.sectoor.de/uml-2.4.26-cmpxchg.patch

But you should move the content of 2.4.24-2-inc-split/ folder into the root of 
the linux tree (i.e. under Uml-linux) and then run "pushpatch 30" after 
installing patch-scripts.

Here there is the tarball:
http://www.user-mode-linux.org/~blaisorblade/patches/2.4.24-2-inc-split.tar.bz2

Here you can download patch-scripts; to install them you must just expand the 
tarball into /usr/local/bin (or any folder you like; remember to update the 
path if needed).

http://www.zip.com.au/~akpm/linux/patches/patch-scripts-0.18/

Bye
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729



-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
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] Fixing the 2.4 bugs - patches against 2.4.24-1um

[Frank Sorenson]

On Thu, 1 Jul 2004, BlaisorBlade wrote:
> I've started working also on UML/2.4.
> 
> Today I've split the changes between 2.4.24-1um and 2.4.24-2um into several 
> patches; the HUMFS ones are carefully split out, but I've separated even the 
> others. Now I need testers. Lots of people cannot use UML patches later than 
> 2.4.24-1um, since they contain a lot of Hostfs bugs and some stability 
> issues. Now, if you use just some patches, you'll avoid the bugs in the 
> other.
> 
> You should apply surely the first 3 patches, IMHO; the other ones should not 
> hurt; inside txt/ there is a description of them.
<snip>

BlaisorBlade,

Thank you for splitting these patches out.  Using them, I was able to
track down and squash a kernel panic issue I've had with all 2.4 patch
versions after 2.4.24-1um!

My panic, for reference - 
http://marc.theaimsgroup.com/?l=user-mode-linux-devel&m=108344473226217&w=2

In this case, the problem was actualy in the first 'mini-changes' patch, 
and the fix (diff to the 2.4.24-2um patch):
--- uml-patch-2.4.24-2  2004-04-07 14:31:13.000000000 -0600
+++ uml-patch-2.4.24-2_frank    2004-07-07 14:26:49.000000000 -0600
@@ -35588,7 +35588,7 @@
 +
 +ARCH_CFLAGS = -U__$(SUBARCH)__ -U$(SUBARCH)
 +
-+ifeq ($(CONFIG_GPROF),n)
++ifneq ($(CONFIG_GPROF),y)
 +ARCH_CFLAGS += -DUM_FASTCALL
 +endif
 +

The configuration program will either set CONFIG_GPROF to 'y', or it will 
put "# CONFIG_GPROF is not set" into the .config file.  CONFIG_GPROF will 
never be 'n', at least in my understanding.

I believe this bug may affect stability for a number of people, and it has
existed since 2.4.24-2um first came out.  It may not solve any underlying
problems, but I'd like to hear if it helps anyone else.

Frank
---------------------------------------------------------------------------
Frank Sorenson - KD7TZK
Systems Manager, Computer Science Department
Brigham Young University
frank@tuxrocks.com


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
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-user] Re: [uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[BlaisorBlade]

Alle 22:51, mercoledì 7 luglio 2004, Frank Sorenson ha scritto:
> On Thu, 1 Jul 2004, BlaisorBlade wrote:
> > I've started working also on UML/2.4.
> >
> > Today I've split the changes between 2.4.24-1um and 2.4.24-2um into
> > several patches; the HUMFS ones are carefully split out, but I've
> > separated even the others. Now I need testers. Lots of people cannot use
> > UML patches later than 2.4.24-1um, since they contain a lot of Hostfs
> > bugs and some stability issues. Now, if you use just some patches, you'll
> > avoid the bugs in the other.
> >
> > You should apply surely the first 3 patches, IMHO; the other ones should
> > not hurt; inside txt/ there is a description of them.

> BlaisorBlade,
>
> Thank you for splitting these patches out.  Using them, I was able to
> track down and squash a kernel panic issue I've had with all 2.4 patch
> versions after 2.4.24-1um!
>
> My panic, for reference -
> http://marc.theaimsgroup.com/?l=user-mode-linux-devel&m=108344473226217&w=2
>
> In this case, the problem was actualy in the first 'mini-changes' patch,
> and the fix (diff to the 2.4.24-2um patch):
> --- uml-patch-2.4.24-2  2004-04-07 14:31:13.000000000 -0600
> +++ uml-patch-2.4.24-2_frank    2004-07-07 14:26:49.000000000 -0600
> @@ -35588,7 +35588,7 @@
>  +
>  +ARCH_CFLAGS = -U__$(SUBARCH)__ -U$(SUBARCH)
>  +
> -+ifeq ($(CONFIG_GPROF),n)
> ++ifneq ($(CONFIG_GPROF),y)
>  +ARCH_CFLAGS += -DUM_FASTCALL
>  +endif
>  +
>
> The configuration program will either set CONFIG_GPROF to 'y', or it will
> put "# CONFIG_GPROF is not set" into the .config file.  CONFIG_GPROF will
> never be 'n', at least in my understanding.

Yes, you are right. Compliments for that - I would have never seen such a 
panic (I understand you perfectly, but I missed it and would have missed it 
again). So thanks a lot for helping with it - I'll save that for when I'm 
back (I'm leaving now, so have no time at all to do anything). Your catch also 
means that UML will break without -DUM_FASTCALL, which is not nice: the buggy 
change you see has actually uncovered some other bug; if it weren't so, the 
bug you have fixed would only have hurted performance.

> I believe this bug may affect stability for a number of people, and it has
> existed since 2.4.24-2um first came out.  It may not solve any underlying
> problems, but I'd like to hear if it helps anyone else.
-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729




-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
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-user] Re: [uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[Frank Sorenson]

On Thu, 8 Jul 2004, BlaisorBlade wrote:
> Yes, you are right. Compliments for that - I would have never seen such a
> panic (I understand you perfectly, but I missed it and would have missed it
> again). So thanks a lot for helping with it - I'll save that for when I'm
> back (I'm leaving now, so have no time at all to do anything). Your catch also
> means that UML will break without -DUM_FASTCALL, which is not nice: the buggy
> change you see has actually uncovered some other bug; if it weren't so, the
> bug you have fixed would only have hurted performance.

Correct.  I've verified that UML does break without -DUM_FASTCALL, even on 
the 'stable' 2.4.24-1um patch, so the problem was probably introduced at 
some point earlier.  Does anyone know the history behind UM_FASTCALL, its' 
effects, and when related modifications occurred?

Frank
---------------------------------------------------------------------------
Frank Sorenson - KD7TZK
Systems Manager, Computer Science Department
Brigham Young University
frank@tuxrocks.com


-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
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-user] Re: [uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[Jeff Dike]

frank@tuxrocks.com said:
> The configuration program will either set CONFIG_GPROF to 'y', or it
> will  put "# CONFIG_GPROF is not set" into the .config file.
> CONFIG_GPROF will  never be 'n', at least in my understanding. 

Jeez, thanks for finding that.

> I believe this bug may affect stability for a number of people, and it
> has existed since 2.4.24-2um first came out.  It may not solve any
> underlying problems, but I'd like to hear if it helps anyone else. 

I don't think it should, unless you're profiling.  The problem is that -pg
and regparam(3) don't (or didn't) interact nicely.  The -pg code trashed
the arguments that regparam was passing in registers, and you'd get strange
panics early in boot.

Have you checked that this actually fixes anything?  It looks like just a 
stupid typo to me.

				Jeff



-------------------------------------------------------
This SF.Net email sponsored by Black Hat Briefings & Training.
Attend Black Hat Briefings & Training, Las Vegas July 24-29 - 
digital self defense, top technical experts, no vendor pitches, 
unmatched networking opportunities. Visit www.blackhat.com
_______________________________________________
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-user] Re: [uml-devel] Fixing the 2.4 bugs - patches against 2.4.24-1um

[Frank Sorenson]

On Tue, 13 Jul 2004, Jeff Dike wrote:
> I don't think it should, unless you're profiling.  The problem is that -pg
> and regparam(3) don't (or didn't) interact nicely.  The -pg code trashed
> the arguments that regparam was passing in registers, and you'd get strange
> panics early in boot.
>
> Have you checked that this actually fixes anything?  It looks like just a
> stupid typo to me.

Yes, it is definitely a stupid typo, and I don't think this alone fixes 
anything significant, but there appears to be another bug underneath. 
Without -DUM_FASTCALL, I've been seeing the kernel panics we previously 
discussed.  There appears to be some other bug that exists at least as 
early as 2.4.24-1um, triggered when UM_FASTCALL is not defined.  I've 
verified that this is the case with the following versions:

2.4.24-1um
2.4.24-2um
2.4.24-3um
2.4.25-1um
2.4.26-1um

The panic I'm getting is very repeatable--every time I run the j2eesdk 
installer, it'll panic, but if UM_FASTCALL is defined, it runs just fine.

Frank
---------------------------------------------------------------------------
Frank Sorenson - KD7TZK
Systems Manager, Computer Science Department
Brigham Young University
frank@tuxrocks.com


-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_id=4721&alloc_id=10040&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] Fixing the 2.4 bugs - patches against 2.4.24-1um - new ones

[BlaisorBlade]

Alle 20:50, giovedì 1 luglio 2004, BlaisorBlade ha scritto:
> I've started working also on UML/2.4.
>
> Today I've split the changes between 2.4.24-1um and 2.4.24-2um into several
> patches; the HUMFS ones are carefully split out, but I've separated even
> the others. Now I need testers. Lots of people cannot use UML patches later
> than 2.4.24-1um, since they contain a lot of Hostfs bugs and some stability
> issues. Now, if you use just some patches, you'll avoid the bugs in the
> other.

I've caught up until 2.4.26-2um, and I'm distributing the new tarball in my 
homepage:

http://www.user-mode-linux.org/~blaisorblade/patches/UML-splitout-2.4.26-2.tar.bz2

The instructions stay the same; I'm not attaching the description because I 
have no time; there should be no bug with a known fix...

> The actual patches can be found inside the patches/ subdir in the tarball,
> and the "series" file contains the order.

THEY ARE AGAINST 2.4.26 vanilla + 2.4.24-1uml patch: the fixlet for 2.4.26 is 
included (so please unapply it before applying this patchset).

> But you should move the content of 2.4.24-2-inc-split/ folder into the root
> of the linux tree (i.e. under Uml-linux) and then run "pushpatch 30" after
> installing patch-scripts.

The folder name changed, and now use pushpatch 100

> Here you can download patch-scripts; to install them you must just expand
> the tarball into /usr/local/bin (or any folder you like; remember to update
> the path if needed).
>
> http://www.zip.com.au/~akpm/linux/patches/patch-scripts-0.18/

-- 
Paolo Giarrusso, aka Blaisorblade
Linux registered user n. 292729



-------------------------------------------------------
This SF.Net email is sponsored by BEA Weblogic Workshop
FREE Java Enterprise J2EE developer tools!
Get your free copy of BEA WebLogic Workshop 8.1 today.
http://ads.osdn.com/?ad_idG21&alloc_id\x10040&opÌk
_______________________________________________
User-mode-linux-devel mailing list
User-mode-linux-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/user-mode-linux-devel