[PATCH] Move kasprintf.o to obj-y

[PATCH] Move kasprintf.o to obj-y

[Alexey Dobriyan]

Modular puppyvisor started giving linking errors

	  MODPOST 1 modules
	ERROR: "kasprintf" [drivers/lguest/lg.ko] undefined!

Signed-off-by: Alexey Dobriyan <adobriyan@sw.ru>
---

 lib/Makefile |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

--- a/lib/Makefile
+++ b/lib/Makefile
@@ -2,7 +2,7 @@
 # Makefile for some libs needed in the kernel.
 #
 
-lib-y := ctype.o string.o vsprintf.o kasprintf.o cmdline.o \
+lib-y := ctype.o string.o vsprintf.o cmdline.o \
 	 rbtree.o radix-tree.o dump_stack.o \
 	 idr.o int_sqrt.o bitmap.o extable.o prio_tree.o \
 	 sha1.o irq_regs.o reciprocal_div.o argv_split.o
@@ -13,7 +13,7 @@ lib-$(CONFIG_SMP) += cpumask.o
 lib-y	+= kobject.o kref.o kobject_uevent.o klist.o
 
 obj-y += div64.o sort.o parser.o halfmd4.o debug_locks.o random32.o \
-	 bust_spinlocks.o hexdump.o
+	 bust_spinlocks.o hexdump.o kasprintf.o
 
 ifeq ($(CONFIG_DEBUG_KOBJECT),y)
 CFLAGS_kobject.o += -DDEBUG

Re: [PATCH] Move kasprintf.o to obj-y

[Christoph Hellwig]

On Mon, Sep 24, 2007 at 11:18:26AM +0400, Alexey Dobriyan wrote:
> Modular puppyvisor started giving linking errors
> 
> 	  MODPOST 1 modules
> 	ERROR: "kasprintf" [drivers/lguest/lg.ko] undefined!

Please kill lib-y while you're at it.  It's useless and a constant
source of pain like this.

Re: [PATCH] Move kasprintf.o to obj-y

[Sam Ravnborg]

On Mon, Sep 24, 2007 at 08:41:10AM +0100, Christoph Hellwig wrote:
> On Mon, Sep 24, 2007 at 11:18:26AM +0400, Alexey Dobriyan wrote:
> > Modular puppyvisor started giving linking errors
> > 
> > 	  MODPOST 1 modules
> > 	ERROR: "kasprintf" [drivers/lguest/lg.ko] undefined!
> 
> Please kill lib-y while you're at it.  It's useless and a constant
> source of pain like this.
Kernel-bloat is another "constant source of pain".
But the troubles are that increased blot does not result in compiler erros.

And your proposal to kil lib-y is a counter-act here.

	Sam

Re: [PATCH] Move kasprintf.o to obj-y

[Christoph Hellwig]

On Mon, Sep 24, 2007 at 09:59:49AM +0200, Sam Ravnborg wrote:
> > Please kill lib-y while you're at it.  It's useless and a constant
> > source of pain like this.
> Kernel-bloat is another "constant source of pain".
> But the troubles are that increased blot does not result in compiler erros.
> 
> And your proposal to kil lib-y is a counter-act here.

Killing lib-y doesn't create kernel bloat if done right.  Just introduce
proper Kconfig dependencies for the truely optional parts.  But if you
look at the list of objects in lib-y you'll see that they fall in basically
three categories:

 (1) always used by core code - should be obj-y
 (2) generic implementation for arch-specific functionality, should be
     guarded by Kconfig for clarity reasons anyway
 (3) library code often but not always used by code that's not always
     built in.  Here we run into the move to obj-y to avoid compiler
     warning when used as module issue all the time.  Adding a CONFIG
     for the bigger onces actually avoids bloat over obj-y here, and
     when it's small enough obj-y is a lot safer in the presence of
     modular users

Really, trying to use the old static library use on demand concept simply
doesn't fit the way the kernel is built with it's modules and dependencies.

Re: [PATCH] Move kasprintf.o to obj-y

[Andrew Morton]

On Mon, 24 Sep 2007 09:09:48 +0100 Christoph Hellwig <hch@infradead.org> wrote:

> On Mon, Sep 24, 2007 at 09:59:49AM +0200, Sam Ravnborg wrote:
> > > Please kill lib-y while you're at it.  It's useless and a constant
> > > source of pain like this.
> > Kernel-bloat is another "constant source of pain".
> > But the troubles are that increased blot does not result in compiler erros.
> > 
> > And your proposal to kil lib-y is a counter-act here.
> 
> Killing lib-y doesn't create kernel bloat if done right.  Just introduce
> proper Kconfig dependencies for the truely optional parts.  But if you
> look at the list of objects in lib-y you'll see that they fall in basically
> three categories:
> 
>  (1) always used by core code - should be obj-y
>  (2) generic implementation for arch-specific functionality, should be
>      guarded by Kconfig for clarity reasons anyway
>  (3) library code often but not always used by code that's not always
>      built in.  Here we run into the move to obj-y to avoid compiler
>      warning when used as module issue all the time.  Adding a CONFIG
>      for the bigger onces actually avoids bloat over obj-y here, and
>      when it's small enough obj-y is a lot safer in the presence of
>      modular users
> 
> Really, trying to use the old static library use on demand concept simply
> doesn't fit the way the kernel is built with it's modules and dependencies.

yup, it duplicates what the config system does and is a common source of
problems.

Re: [PATCH] Move kasprintf.o to obj-y

[Sam Ravnborg]

On Mon, Sep 24, 2007 at 09:09:48AM +0100, Christoph Hellwig wrote:
> On Mon, Sep 24, 2007 at 09:59:49AM +0200, Sam Ravnborg wrote:
> > > Please kill lib-y while you're at it.  It's useless and a constant
> > > source of pain like this.
> > Kernel-bloat is another "constant source of pain".
> > But the troubles are that increased blot does not result in compiler erros.
> > 
> > And your proposal to kil lib-y is a counter-act here.
> 
> Killing lib-y doesn't create kernel bloat if done right.  Just introduce
> proper Kconfig dependencies for the truely optional parts.  But if you
> look at the list of objects in lib-y you'll see that they fall in basically
> three categories:
> 
>  (1) always used by core code - should be obj-y
>  (2) generic implementation for arch-specific functionality, should be
>      guarded by Kconfig for clarity reasons anyway
>  (3) library code often but not always used by code that's not always
>      built in.  Here we run into the move to obj-y to avoid compiler
>      warning when used as module issue all the time.  Adding a CONFIG
>      for the bigger onces actually avoids bloat over obj-y here, and
>      when it's small enough obj-y is a lot safer in the presence of
>      modular users
> 
> Really, trying to use the old static library use on demand concept simply
> doesn't fit the way the kernel is built with it's modules and dependencies.

I objected to just killing lib-y (the way I read your original proposal).
If we introduce proper kconfig dependencies as in (3) then I am all for it.

	Sam