PAGE_ALIGN(): correctly handle ... breaks AVR32 build
Subsystems (3)
As of
commit 27ac792ca0b0a1e7e65f20342260650516c95864
Author: Andrea Righi <[email protected]> Date: Wed Jul 23 21:28:13 2008 -0700
PAGE_ALIGN(): correctly handle 64-bit values on 32-bit architectures
We no longer have PAGE_ALIGN in asm-avr32/page.h so the AVR32 build dies quite quickly:
CC init/main.o
In file included from include/linux/utsname.h:35,
from init/main.c:20:
include/linux/sched.h: In function 'arch_pick_mmap_layout':include/linux/sched.h:2149: error: implicit declaration of function 'PAGE_ALIGN' make[1]: *** [init/main.o] Error 1
make: *** [init] Error 2
This looks a relatively simple fix, just include linux/mm.h from asm-avr32/processor.h from where sched.h gets the dodgy macro. Unfortunately this, and every other placement of an #include <linux/mm.h> I've tried, ends up in Massive Include Armageddon as per [1].
Ideas?
Thanks,
--Ben.
PATCH 1/1 Uninline arch_pick_mmap_layout
As reported by Ben Nizette:
Quoted 11 lines
We no longer have
PAGE_ALIGNinasm-avr32/page.hso the AVR32 build dies quite quickly:CC
init/main.o
In file included frominclude/linux/utsname.h:35,from init/main.c:20:
include/linux/sched.h: In function 'arch_pick_mmap_layout':include/linux/sched.h:2149: error: implicit declaration of function 'PAGE_ALIGN' make[1]: *** [init/main.o] Error 1
make: *** [init] Error 2
Quoted 4 lines
This looks a relatively simple fix, just include
linux/mm.hfromasm-avr32/processor.hfrom wheresched.hgets the dodgy macro. Unfortunately this, and every other placement of an #include <linux/mm.h> I've tried, ends up in Massive Include Armageddon
So instead of shuffling includes around, just move the definition of arch_pick_mmap_layout() into fs/exec.c. Some architectures, e.g. x86, already define it out-of-line.
--- fs/exec.c | 9 +++++++++ include/linux/sched.h | 9 --------- 2 files changed, 9 insertions(+), 9 deletions(-)
fs/exec.c +9 -0
@@ -74,6 +74,15 @@ int suid_dumpable = 0;static LIST_HEAD (formats);static DEFINE_RWLOCK (binfmt_lock); +#ifndef HAVE_ARCH_PICK_MMAP_LAYOUT +void arch_pick_mmap_layout (struct mm_struct *mm) +{ + mm->mmap_base =TASK_UNMAPPED_BASE ; + mm->get_unmapped_area = arch_get_unmapped_area; + mm->unmap_area = arch_unmap_area; +} +#endif +int register_binfmt (struct linux_binfmt * fmt) {if (!fmt)
include/linux/sched.h +0 -9
@@ -2141,16 +2141,7 @@ static inline void set_task_cpu(struct task_struct *p, unsigned int cpu)#endif /* CONFIG_SMP */ -#ifdef HAVE_ARCH_PICK_MMAP_LAYOUT extern void arch_pick_mmap_layout (struct mm_struct *mm); -#else -static inline void arch_pick_mmap_layout (struct mm_struct *mm) -{ - mm->mmap_base =TASK_UNMAPPED_BASE ; - mm->get_unmapped_area = arch_get_unmapped_area; - mm->unmap_area = arch_unmap_area; -} -#endif #ifdef CONFIG_TRACING extern void
Signature · 1 line
1.5.6.2
Re: [PATCH] Uninline arch_pick_mmap_layout
Quoted 5 lines · On Fri, 25 Jul 2008 12:28:02 +0200 Haavard Skinnemoen…
So instead of shuffling includes around, just move the definition of
arch_pick_mmap_layout()intofs/exec.c. Some architectures, e.g. x86, already define it out-of-line.Signed-off-by Haavard Skinnemoen <[email protected]>
Heh, ok. So we have four patches fixing this regression now. I didn't read all my mail before posting; sorry about that.
The patches which have already been added to -mm look essentially identical to mine, so feel free to ignore my patch.
Haavard