Re: [PATCH 08/16] mm/unicore32: use vm_flags_t for vma flags

From: Guan Xuetao
Date: Tue Mar 27 2012 - 03:49:52 EST

On 03/27/2012 01:58 PM, Konstantin Khlebnikov wrote:
Guan Xuetao wrote:
On 03/21/2012 02:56 PM, Konstantin Khlebnikov wrote:
The same magic like in arm: assembler code wants to test VM_EXEC,
but for big-endian we should get upper word for this.

Signed-off-by: Konstantin Khlebnikov<khlebnikov@xxxxxxxxxx>
Cc: Guan Xuetao<gxt@xxxxxxxxxxxxxxx>
arch/unicore32/kernel/asm-offsets.c | 6 +++++-
arch/unicore32/mm/fault.c | 2 +-
2 files changed, 6 insertions(+), 2 deletions(-)

diff --git a/arch/unicore32/kernel/asm-offsets.c b/arch/unicore32/kernel/asm-offsets.c
index ffcbe75..e3199b5 100644
--- a/arch/unicore32/kernel/asm-offsets.c
+++ b/arch/unicore32/kernel/asm-offsets.c
@@ -87,9 +87,13 @@ int main(void)
DEFINE(S_FRAME_SIZE, sizeof(struct pt_regs));
DEFINE(VMA_VM_MM, offsetof(struct vm_area_struct, vm_mm));
+#if defined(CONFIG_CPU_BIG_ENDIAN)&& (NR_VMA_FLAGS> 32)
+ DEFINE(VMA_VM_FLAGS, offsetof(struct vm_area_struct, vm_flags) + 4);
CONFIG_CPU_BIG/LITTLE_ENDIAN is defined only in some archs, and not
supported by unicore32.

Ok, I'll drop this in v2

DEFINE(VMA_VM_FLAGS, offsetof(struct vm_area_struct, vm_flags));
+ DEFINE(VM_EXEC, (__force unsigned int)VM_EXEC);

Is this check useful for asm-offsets.h?

this forced-typecast to make sparse happy, because we use here only (int) part of vma->vm_flags.
Perhaps, vm_flags_t, not unsigned int is more proper.

diff --git a/arch/unicore32/mm/fault.c b/arch/unicore32/mm/fault.c
index 283aa4b..9137996 100644
--- a/arch/unicore32/mm/fault.c
+++ b/arch/unicore32/mm/fault.c
@@ -158,7 +158,7 @@ void do_bad_area(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
static inline bool access_error(unsigned int fsr, struct vm_area_struct *vma)
- unsigned int mask = VM_READ | VM_WRITE | VM_EXEC;
+ vm_flags_t mask = VM_READ | VM_WRITE | VM_EXEC;

I am confused for the type of vm_flags in vm_area_struct being
'unsigned long', not vm_flags_t.

Second patch in this patchset changes it.
vm_flags_t will be unsigned long or or unsigned long long depending on vma flags count.
But more likely it will be unsigned int, because I have another patchset in work,
which currently drop four bits in vm_flags, so we can postpone its expansion.
I see. Thanks for your explanation.

Guan Xuetao

To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at
Please read the FAQ at