Bug#353461: [2.6.16-rc3,powerpc] asm/highmem.h not found (ARCH=ppc vs. ARCH=powerpc)
On Sat, Feb 18, 2006 at 06:52:41PM +0100, Rene Engelhard wrote:
> Package: linux-2.6
> Severity: important
> Version: 2.6.15+2.6.16-rc3-0experimental.0snapshot.5916
> [ Sven tells me to file a bug here; although this package is unofficial
> and not-in-Debian I'll do so. complain at him if you think like me... ]
> while trying to build bcm43xx against 2.6.16-rc3:
> mkdir -p /home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx/.tmp_versions
> make -f scripts/Makefile.build obj=/home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx
> gcc -Wp,-MD,/home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx/.bcm43xx_main.o.d -nostdinc -isystem /usr/lib/gcc/powerpc-linux-gnu/4.0.3/include -D__KERNEL__ -Iinclude -include include/linux/autoconf.h -Iarch/powerpc -Iarch/powerpc/include -DCONFIG_BCM43XX=1 -DCONFIG_BCM43XX_DMA=1 -DCONFIG_BCM43XX_PIO=1 -DCONFIG_BCM43XX_DEBUG=1 -DMODULE -D"KBUILD_STR(s)=#s" -D"KBUILD_BASENAME=KBUILD_STR(bcm43xx_main)" -D"KBUILD_MODNAME=KBUILD_STR(bcm43xx)" -c -o /home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx/.tmp_bcm43xx_main.o /home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx/bcm43xx_main.c
> In file included from include/linux/skbuff.h:27,
> from include/linux/if_ether.h:110,
> from include/linux/netdevice.h:29,
> from include/linux/if_arp.h:26,
> from /home/rene/Debian/Pakete/bcm43xx/bcm43xx-20060215/drivers/net/wireless/bcm43xx/bcm43xx_main.c:34:
> include/linux/highmem.h:12:25: error: asm/highmem.h: No such file or directory
08:51 < miltonm> svenl: ok well it is still in the include/ppc diretory ... if
(s)he is trying to build a seperate modules against ARCH=ppc then there were
some issues with that directory ... otherwise i wouuld look
at which headers include that file and then look at those in
08:56 < svenl> miltonm: he is trying to build with ARCH=powerpc.
08:56 < svenl> miltonm: as supposedly 2.6.16-rc* now defaults to ARCH=powerpc.
08:59 < miltonm> svenl yes it does .... i don't have a tree handy, but i would
guess asm/highmem.h should be indirectly included by some other mm header, and
that was absorbed. hence grep highmem.h include/ssm-p*pc/*.h
09:00 < svenl> miltonm: so you think it should be working ?
09:00 < miltonm> svenl: if its included under powerpc somwhere then it is the
kbuild for ext modules (seperate objdir?) problems
09:00 < svenl> ok.
09:01 < miltonm> svenl: well, it could be no one has built with highmem
defined since the move (since ppc64 has !config_highmem and ben and paul might
09:02 < svenl> miltonm: that is also my guess.
09:02 < svenl> since most of you guys build with ppc64 those days :)
09:02 < miltonm> svenl: well they also build for their laptops but like i
said, maybe 2g/2g
09:03 < miltonm> svenl: or 32 bit grabs the old headers via
arch/powerpc/include symlink ??
09:04 < miltonm> like i said i don't have a src tree here , laptop is
09:04 < svenl> miltonm: oh, there is still such a trick, i thought it would
die for arch/powerpc.
09:04 < svenl> hehe.
09:04 < miltonm> well, yea, that link should die, and i forget which way it
09:05 < miltonm> svenl maybe linux/highmem.h vs asm/highmem.h ? or like you
said, no one tested highmem
09:05 < svenl> miltonm: there was a link from arch/ppc/include to the powerpc
stuff, i didn't know of a link the other way around.
09:05 * miltonm should stop speculating
09:05 < svenl> miltonm: i will be offline next week, will ask _rene_ to
investigate and eventually come here.
09:05 < svenl> ok ?
09:06 < miltonm> yea probably right ... is there anohter file that defines
those functions under include/asm-powerpc?
Rene, hope this helps, can you investigate and comment on this bug report.
Bastian i will be offline or mostly so until fosdem, can you fix this if it
ends up being a linux-2.6 header packaging bug ?