Re: oom-killer problem

Re: oom-killer problem

[Daniel Ritz]

hi

i got the same here. reason is a recent change that made modules always
shows as module.mod. it breaks modprobe and probably many scripts..besides
lsmod looking horrible.
stuff like this in modprobe.conf:
	install pcmcia_core /sbin/modprobe --ignore-install pcmcia_core; /sbin/modprobe pcmcia
makes modprobe fork/exec endlessly calling itself...until oom interrupts it...
reverting the attached patch fixes the problem...

rgds
-daniel


From: Sam Ravnborg <sam@mars.ravnborg.org>
Date: Sat, 24 Jun 2006 20:50:18 +0000 (+0200)
Subject: kbuild: fix make -rR breakage
X-Git-Url: http://www.kernel.org/git/gitweb.cgi?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=e5c44fd88c146755da6941d047de4d97651404a9

kbuild: fix make -rR breakage

make failed to supply the filename when using make -rR and using $(*F)
to get target filename without extension.
This bug was not reproduceable in small scale but using:
$(basename $(notdir $@)) fixes it with same functionality.

Signed-off-by: Sam Ravnborg <sam@ravnborg.org>
---

--- a/scripts/Kbuild.include
+++ b/scripts/Kbuild.include
@@ -13,6 +13,11 @@ space   := $(empty) $(empty)
 depfile = $(subst $(comma),_,$(@D)/.$(@F).d)
 
 ###
+# basetarget equals the filename of the target with no extension.
+# So 'foo/bar.o' becomes 'bar'
+basetarget = $(basename $(notdir $@))
+
+###
 # Escape single quote for use in echo statements
 escsq = $(subst $(squote),'\$(squote)',$1)
 
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -117,7 +117,7 @@ $(real-objs-m:.o=.lst): quiet_modtag := 
 $(obj-m)              : quiet_modtag := [M]
 
 # Default for not multi-part modules
-modname = $(*F)
+modname = $(basetarget)
 
 $(multi-objs-m)         : modname = $(modname-multi)
 $(multi-objs-m:.o=.i)   : modname = $(modname-multi)
--- a/scripts/Makefile.host
+++ b/scripts/Makefile.host
@@ -80,8 +80,10 @@ obj-dirs += $(host-objdirs)
 #####
 # Handle options to gcc. Support building with separate output directory
 
-_hostc_flags   = $(HOSTCFLAGS)   $(HOST_EXTRACFLAGS)   $(HOSTCFLAGS_$(*F).o)
-_hostcxx_flags = $(HOSTCXXFLAGS) $(HOST_EXTRACXXFLAGS) $(HOSTCXXFLAGS_$(*F).o)
+_hostc_flags   = $(HOSTCFLAGS)   $(HOST_EXTRACFLAGS)   \
+                 $(HOSTCFLAGS_$(basetarget).o)
+_hostcxx_flags = $(HOSTCXXFLAGS) $(HOST_EXTRACXXFLAGS) \
+                 $(HOSTCXXFLAGS_$(basetarget).o)
 
 ifeq ($(KBUILD_SRC),)
 __hostc_flags	= $(_hostc_flags)
--- a/scripts/Makefile.lib
+++ b/scripts/Makefile.lib
@@ -82,12 +82,12 @@ obj-dirs	:= $(addprefix $(obj)/,$(obj-di
 #       than one module. In that case KBUILD_MODNAME will be set to foo_bar,
 #       where foo and bar are the name of the modules.
 name-fix = $(subst $(comma),_,$(subst -,_,$1))
-basename_flags = -D"KBUILD_BASENAME=KBUILD_STR($(call name-fix,$(*F)))"
+basename_flags = -D"KBUILD_BASENAME=KBUILD_STR($(call name-fix,$(basetarget)))"
 modname_flags  = $(if $(filter 1,$(words $(modname))),\
                  -D"KBUILD_MODNAME=KBUILD_STR($(call name-fix,$(modname)))")
 
-_c_flags       = $(CFLAGS) $(EXTRA_CFLAGS) $(CFLAGS_$(*F).o)
-_a_flags       = $(AFLAGS) $(EXTRA_AFLAGS) $(AFLAGS_$(*F).o)
+_c_flags       = $(CFLAGS) $(EXTRA_CFLAGS) $(CFLAGS_$(basetarget).o)
+_a_flags       = $(AFLAGS) $(EXTRA_AFLAGS) $(AFLAGS_$(basetarget).o)
 _cpp_flags     = $(CPPFLAGS) $(EXTRA_CPPFLAGS) $(CPPFLAGS_$(@F))
 
 # If building the kernel in a separate objtree expand all occurrences
--- a/scripts/Makefile.modpost
+++ b/scripts/Makefile.modpost
@@ -72,7 +72,7 @@ $(modules:.ko=.mod.c): __modpost ;
 # Step 5), compile all *.mod.c files
 
 # modname is set to make c_flags define KBUILD_MODNAME
-modname = $(*F)
+modname = $(basetarget)
 
 quiet_cmd_cc_o_c = CC      $@
       cmd_cc_o_c = $(CC) $(c_flags) $(CFLAGS_MODULE)	\

Re: oom-killer problem

[Linus Torvalds]

On Tue, 27 Jun 2006, Daniel Ritz wrote:
>
> reverting the attached patch fixes the problem...

Michal, can you also confirm that just doing a simple revert of that one 
commit makes things work for you?

Sam, if I don't hear otherwise from you, and Michael confirms, I'll just 
revert it for now, and you can figure out how to fix it without breakage?

			Linus

Re: oom-killer problem

[Michal Piotrowski]

On 27/06/06, Linus Torvalds <torvalds@osdl.org> wrote:
>
>
> On Tue, 27 Jun 2006, Daniel Ritz wrote:
> >
> > reverting the attached patch fixes the problem...
>
> Michal, can you also confirm that just doing a simple revert of that one
> commit makes things work for you?

Yes I can confirm that.
(http://www.ussg.iu.edu/hypermail/linux/kernel/0606.3/0827.html)

>
> Sam, if I don't hear otherwise from you, and Michael confirms, I'll just
> revert it for now, and you can figure out how to fix it without breakage?
>
>                         Linus
>

Regards,
Michal

-- 
Michal K. K. Piotrowski
LTG - Linux Testers Group
(http://www.stardust.webpages.pl/ltg/wiki/)

Re: oom-killer problem

[Sam Ravnborg]

On Mon, Jun 26, 2006 at 04:05:40PM -0700, Linus Torvalds wrote:
> 
> 
> On Tue, 27 Jun 2006, Daniel Ritz wrote:
> >
> > reverting the attached patch fixes the problem...
> 
> Michal, can you also confirm that just doing a simple revert of that one 
> commit makes things work for you?
> 
> Sam, if I don't hear otherwise from you, and Michael confirms, I'll just 
> revert it for now, and you can figure out how to fix it without breakage?
I will try to find time during the weekend to track down the cause of
this.
But by reverting said patch you also have to revert:
566f81ca598f80de03e80a9a743e94b65b4e017e

This is the patch where make -rR is enabled and that one initally caused
problems for ia64.

	Sam

Re: oom-killer problem

[Daniel Ritz]

On Tuesday 27 June 2006 07.39, Sam Ravnborg wrote:
> On Mon, Jun 26, 2006 at 04:05:40PM -0700, Linus Torvalds wrote:
> > On Tue, 27 Jun 2006, Daniel Ritz wrote:
> > >
> > > reverting the attached patch fixes the problem...
> > 
> > Michal, can you also confirm that just doing a simple revert of that one 
> > commit makes things work for you?
> > 
> > Sam, if I don't hear otherwise from you, and Michael confirms, I'll just 
> > revert it for now, and you can figure out how to fix it without breakage?
> I will try to find time during the weekend to track down the cause of
> this.
> But by reverting said patch you also have to revert:
> 566f81ca598f80de03e80a9a743e94b65b4e017e
> 
> This is the patch where make -rR is enabled and that one initally caused
> problems for ia64.
> 
> 	Sam
> 

this little something on top (instead of reverting the patch) also fixes the
problem for me. because the module name depends on the name of module.mod.o
it's necessary to call basename twice on that, otherwise the name ends up
being module.mod

maybe we could also define basetarget in Kbuild.include as:
	basetarget = $(basename $(basename $(notdir $@)))
but that might be too much...dunno


diff --git a/scripts/Makefile.build b/scripts/Makefile.build
index 3cb445c..e66f6dc 100644
--- a/scripts/Makefile.build
+++ b/scripts/Makefile.build
@@ -117,7 +117,7 @@ quiet_modtag := $(empty)   $(empty)
 $(obj-m)              : quiet_modtag := [M]
 
 # Default for not multi-part modules
-modname = $(basetarget)
+modname = $(basename $(basetarget))
 
 $(multi-objs-m)         : modname = $(modname-multi)
 $(multi-objs-m:.o=.i)   : modname = $(modname-multi)
diff --git a/scripts/Makefile.modpost b/scripts/Makefile.modpost
index e83613e..17f3d31 100644
--- a/scripts/Makefile.modpost
+++ b/scripts/Makefile.modpost
@@ -72,7 +72,7 @@ # Declare generated files as targets for
 # Step 5), compile all *.mod.c files
 
 # modname is set to make c_flags define KBUILD_MODNAME
-modname = $(basetarget)
+modname = $(basename $(basetarget))
 
 quiet_cmd_cc_o_c = CC      $@
       cmd_cc_o_c = $(CC) $(c_flags) $(CFLAGS_MODULE)	\

Re: oom-killer problem

[Linus Torvalds]

On Tue, 27 Jun 2006, Daniel Ritz wrote:
>  
>  # Default for not multi-part modules
> -modname = $(basetarget)
> +modname = $(basename $(basetarget))

Is there some way to make it clear _what_ the suffix we expect to remove 
actually is in GNU make? Ie the "shell" kind of "basename" logic.

Ie, I'd personally be happier with a 

	modname = $(basename $(basetarget) .mod)

kind of thing (yeah, this obviously does _not_ work)

Gah. I've happily been trying to avoid having to know all the GNU Makefile 
magic. Sam, does Dans patch make sense and can you explain what went 
wrong? I'd happily revert the revert and instead use Dans patch, I just 
want to understand it..

		Linus

Re: oom-killer problem

[Keith Owens]

Linus Torvalds (on Tue, 27 Jun 2006 18:15:46 -0700 (PDT)) wrote:
>
>
>On Tue, 27 Jun 2006, Daniel Ritz wrote:
>>  
>>  # Default for not multi-part modules
>> -modname = $(basetarget)
>> +modname = $(basename $(basetarget))
>
>Is there some way to make it clear _what_ the suffix we expect to remove 
>actually is in GNU make? Ie the "shell" kind of "basename" logic.
>
>Ie, I'd personally be happier with a 
>
>	modname = $(basename $(basetarget) .mod)
>
>kind of thing (yeah, this obviously does _not_ work)

modname = $(patsubst %.mod,%,$(basetarget))

should do it (untested).

Re: oom-killer problem

[Daniel Ritz]

On Wednesday 28 June 2006 03.42, Keith Owens wrote:
> Linus Torvalds (on Tue, 27 Jun 2006 18:15:46 -0700 (PDT)) wrote:
> >
> >
> >On Tue, 27 Jun 2006, Daniel Ritz wrote:
> >>  
> >>  # Default for not multi-part modules
> >> -modname = $(basetarget)
> >> +modname = $(basename $(basetarget))
> >
> >Is there some way to make it clear _what_ the suffix we expect to remove 
> >actually is in GNU make? Ie the "shell" kind of "basename" logic.
> >
> >Ie, I'd personally be happier with a 
> >
> >	modname = $(basename $(basetarget) .mod)
> >
> >kind of thing (yeah, this obviously does _not_ work)
> 
> modname = $(patsubst %.mod,%,$(basetarget))
> 
> should do it (untested).
> 

yep, works fine.

Re: oom-killer problem

[Kyle Moffett]

On Jun 27, 2006, at 21:42:13, Keith Owens wrote:
> Linus Torvalds (on Tue, 27 Jun 2006 18:15:46 -0700 (PDT)) wrote:
>> Ie, I'd personally be happier with a
>>
>> 	modname = $(basename $(basetarget) .mod)
>>
>> kind of thing (yeah, this obviously does _not_ work)
>
> modname = $(patsubst %.mod,%,$(basetarget))

Sorry to come in so late on this; this email's been sitting in my  
outbox for a week.  Here's a mildly simpler alternative:

modname = $(basetarget:.mod=)

A slightly more flexible alternative:

modname = $(basetarget:%.mod=%)

Cheers,
Kyle Moffett

oom-killer problem

[Michal Piotrowski]

Hi Linus,

I have noticed a small problem with
2.6.17-5fd571cbc13db113bda26c20673e1ec54bfd26b4 - in fact, it doesn't
work.

http://www.stardust.webpages.pl/files/linux/bug1.jpg
http://www.stardust.webpages.pl/files/linux/bug2.jpg

Here is a config file
http://www.stardust.webpages.pl/files/linux/config

Any ideas?

Regards,
Michal

-- 
Michal K. K. Piotrowski
LTG - Linux Testers Group
(http://www.stardust.webpages.pl/ltg/wiki/)

Re: oom-killer problem

[Linus Torvalds]

On Mon, 26 Jun 2006, Michal Piotrowski wrote:
> 
> I have noticed a small problem with
> 2.6.17-5fd571cbc13db113bda26c20673e1ec54bfd26b4 - in fact, it doesn't
> work.

Well, it looks to me like you have IDE problems (and shaky hands ;)

Can you pinpoint when these things started happening? "git bisect" is your 
friend..

		Linus

Re: oom-killer problem

[Michal Piotrowski]

On 26/06/06, Linus Torvalds <torvalds@osdl.org> wrote:
>
>
> On Mon, 26 Jun 2006, Michal Piotrowski wrote:
> >
> > I have noticed a small problem with
> > 2.6.17-5fd571cbc13db113bda26c20673e1ec54bfd26b4 - in fact, it doesn't
> > work.
>
> Well, it looks to me like you have IDE problems (and shaky hands ;)
>
> Can you pinpoint when these things started happening? "git bisect" is your
> friend..

[michal@ltg01-fedora linux-git]$ git-bisect bad
e5c44fd88c146755da6941d047de4d97651404a9
e5c44fd88c146755da6941d047de4d97651404a9 is first bad commit
commit e5c44fd88c146755da6941d047de4d97651404a9
Author: Sam Ravnborg <sam@mars.ravnborg.org>
Date:   Sat Jun 24 22:50:18 2006 +0200

    kbuild: fix make -rR breakage

    make failed to supply the filename when using make -rR and using $(*F)
    to get target filename without extension.
    This bug was not reproduceable in small scale but using:
    $(basename $(notdir $@)) fixes it with same functionality.

    Signed-off-by: Sam Ravnborg <sam@ravnborg.org>

:040000 040000 88f5ba77585f29510879579cf2737470e4f5eaef
a7e1133f110192016e5a34b123eb8929c2f5d40e M      scripts

Please revert this commit.


BTW "git show e5c44fd88c146755da6941d047de4d97651404a9" doesn't show
any IDE specific changes.

>
>                 Linus
>

Regards,
Michal

-- 
Michal K. K. Piotrowski
LTG - Linux Testers Group
(http://www.stardust.webpages.pl/ltg/wiki/)