Re: [RFC v3 PATCH 2/5] mm: introduce VM_DEAD flag

[Michal Hocko]

On Sat 30-06-18 06:39:42, Yang Shi wrote:
> VM_DEAD flag is used to mark a vma is being unmapped, access to this
> area will trigger SIGSEGV.
> 
> This flag will be used by the optimization for unmapping large address
> space (>= 1GB) in the later patch. It is 64 bit only at the moment,
> since:
>   * we used up vm_flags bit for 32 bit
>   * 32 bit machine typically will not have such large mapping
> 
> All architectures, which support 64 bit, need check this flag in their
> page fault handler. This is implemented in later patches.

Please add a new flag with its users. There is simply no way to tell the
semantic from the above description and the patch as well.
 
> Suggested-by: Michal Hocko <mhocko@xxxxxxxxxx>
> Signed-off-by: Yang Shi <yang.shi@xxxxxxxxxxxxxxxxx>
> ---
>  include/linux/mm.h | 6 ++++++
>  1 file changed, 6 insertions(+)
> 
> diff --git a/include/linux/mm.h b/include/linux/mm.h
> index a0fbb9f..28a3906 100644
> --- a/include/linux/mm.h
> +++ b/include/linux/mm.h
> @@ -242,6 +242,12 @@ extern int overcommit_kbytes_handler(struct ctl_table *, int, void __user *,
>  #endif
>  #endif /* CONFIG_ARCH_HAS_PKEYS */
>  
> +#ifdef CONFIG_64BIT
> +#define VM_DEAD			BIT(37)	/* bit only usable on 64 bit kernel */
> +#else
> +#define VM_DEAD			0
> +#endif
> +
>  #if defined(CONFIG_X86)
>  # define VM_PAT		VM_ARCH_1	/* PAT reserves whole VMA at once (x86) */
>  #elif defined(CONFIG_PPC)
> -- 
> 1.8.3.1

-- 
Michal Hocko
SUSE Labs