Re: [PATCH] Fix NR_KEYS off-by-one error

[OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>]

Andries Brouwer <aebr@win.tue.nl> writes:

> On Sat, Jul 17, 2004 at 01:35:59AM +0900, OGAWA Hirofumi wrote:
> 
> :: KDGKBENT ioctl can use 256 entries (0-255), but it was defined as
> :: key_map[NR_KEYS] (NR_KEYS == 255). The code seems also thinking it's 256.
> :: 
> :: 	key_map[0] = U(K_ALLOCATED);
> :: 	for (j = 1; j < NR_KEYS; j++)
> :: 		key_map[j] = U(K_HOLE);
> 
> I think the code has no opinion. It was 128 in 2.4.
> I am not aware of assumptions on NR_KEYS.
> So, do not think this is an off-by-one error.

My point is that key_map is 0-254 array. But KDGKBENT uses 255

	case KDGKBENT:
		key_map = key_maps[s];
		if (key_map) {
		    val = U(key_map[i]);
		    if (kbd->kbdmode != VC_UNICODE && KTYP(val) >= NR_TYPES)
			val = K_HOLE;
		} else
		    val = (i ? K_HOLE : K_NOSUCHMAP);
		return put_user(val, &user_kbe->kb_value);

Thanks.
-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>