Re: [PATCH v3 10/10] powerpc/uaccess: Implement masked user access

[Gabriel Paubert <paubert@iram.es>]

Hi Christophe,

On Fri, Oct 17, 2025 at 12:21:06PM +0200, Christophe Leroy wrote:
> Masked user access avoids the address/size verification by access_ok().
> Allthough its main purpose is to skip the speculation in the
> verification of user address and size hence avoid the need of spec
> mitigation, it also has the advantage of reducing the amount of
> instructions required so it even benefits to platforms that don't
> need speculation mitigation, especially when the size of the copy is
> not know at build time.
> 
> So implement masked user access on powerpc. The only requirement is
> to have memory gap that faults between the top user space and the
> real start of kernel area.
> 
> On 64 bits platforms the address space is divided that way:
> 
> 	0xffffffffffffffff	+------------------+
> 				|                  |
> 				|   kernel space   |
>  		 		|                  |
> 	0xc000000000000000	+------------------+  <== PAGE_OFFSET
> 				|//////////////////|
> 				|//////////////////|
> 	0x8000000000000000	|//////////////////|
> 				|//////////////////|
> 				|//////////////////|
> 	0x0010000000000000	+------------------+  <== TASK_SIZE_MAX
> 				|                  |
> 				|    user space    |
> 				|                  |
> 	0x0000000000000000	+------------------+
> 
> Kernel is always above 0x8000000000000000 and user always
> below, with a gap in-between. It leads to a 3 instructions sequence:
> 
>   20:	7c 69 fe 76 	sradi   r9,r3,63
>   24:	7c 69 48 78 	andc    r9,r3,r9
>   28:	79 23 00 4c 	rldimi  r3,r9,0,1
> 

Actually there is an even simpler (more obvious) sequence:

sradi r9,r3,63
srdi r9,r9,1  
andc r3,r3,r9

(the second instruction could also be clrldi r9,r9,1)

which translates back to C as:

[snipped]
> +static inline void __user *mask_user_address_simple(const void __user *ptr)
> +{
> +	unsigned long addr = (unsigned long)ptr;
> +	unsigned long sh = BITS_PER_LONG - 1;
> +	unsigned long mask = (unsigned long)((long)addr >> sh);
> +
> +	addr = ((addr & ~mask) & ((1UL << sh) - 1)) | ((mask & 1UL) << sh);
> +
> +	return (void __user *)addr;
> +}
> +

either (srdi):
	unsigned long mask = ((unsigned long)((long)addr >> sh)) >> 1;
or (clrldi):
	unsigned long mask = (unsigned long)(((long)addr >> sh) & LONG_MAX);

followed by:
	return (void __user *)(addr & ~ mask);

the result is the same but I find it easier to read, and it may be
easier for the compiler than to recognize an rl?imi insruction.

Cheers,
Gabriel