Re: [PATCH 05/12] unicore32 additional architecture files: low-level lib: uaccess

[Arnd Bergmann <arnd@arndb.de>]

On Sunday 27 February 2011, Guan Xuetao wrote:
> > -----Original Message-----
> > From: Arnd Bergmann [mailto:arnd@arndb.de]
> > Sent: Friday, February 18, 2011 1:45 AM
> > To: Guan Xuetao
> > Cc: linux-kernel@vger.kernel.org; linux-arch@vger.kernel.org; 'Greg KH'
> > Subject: Re: [PATCH 05/12] unicore32 additional architecture files: low-level lib: uaccess
> > 
> > On Wednesday 16 February 2011, Guan Xuetao wrote:
> > > +
> > > +#define __kernel_ok                (segment_eq(get_fs(), KERNEL_DS))
> > > +#define __user_ok(addr, size)      (((size) <= TASK_SIZE)                  \
> > > +                           && ((addr) <= TASK_SIZE - (size)))
> > > +#define __access_ok(addr, size)    (__kernel_ok || __user_ok((addr), (size)))
> > 
> > A nice trick to simplify this is to introduce a per-process variable for comparing
> > the pointer, set_fs() then sets this variable to either TASK_SIZE or
> > UINT_MAX.
> 
> Perhaps like this:
> 1. add a member named kaddr in thread_info
> 2. add following assignment codes in set_fs:
>         current_thread_info()->kaddr =  segment_eq(fs, KERNEL_DS) ? UINT_MAX : TASK_SIZE;
> 3. then __access_ok will be:
>         #define addr 
>         #define __access_ok(addr, size)   (((size) <= TASK_SIZE)                        \
>                                 && ((addr) <= current_thread_info()->kaddr)  - (size)))

That would also work. What I meant was to have

#define set_fs(fs) \
	do { current_thread_info()->kaddr = (fs); } while (0)

#define KERNEL_DS	UINT_MAX
#define USER_DS		TASK_SIZE

Either way is fine with me, just choose whichever works best for you in terms of code size.

	Arnd