[PATCH 10/15] xkb: Fix possible NULL pointer dereference

[Peter Hutterer]

On Tue, Jul 27, 2010 at 03:09:51PM +0300, Pauli Nieminen wrote:
> If search for device failed sli is NULL. In that case we have to protect
> dereference to prevent server crash.
> 
> Signed-off-by: Pauli Nieminen <ext-pauli.nieminen at nokia.com>
> ---
>  xkb/ddxLoad.c |    6 +++++-
>  xkb/xkbLEDs.c |   10 ++++++----
>  2 files changed, 11 insertions(+), 5 deletions(-)
> 
> diff --git a/xkb/ddxLoad.c b/xkb/ddxLoad.c
> index 5e6ab87..2fdf71e 100644
> --- a/xkb/ddxLoad.c
> +++ b/xkb/ddxLoad.c
> @@ -342,7 +342,11 @@ char		fileName[PATH_MAX];
>  unsigned	missing;
>  
>      *xkbRtrn = NULL;
> -    if ((keybd==NULL)||(keybd->key==NULL)||(keybd->key->xkbInfo==NULL))
> +    if (!keybd) {
> +        LogMessage(X_ERROR, "XKB: %s no device given as parameter.\n", __func__);
> +        return 0;
> +    }
> +    if ((keybd->key==NULL)||(keybd->key->xkbInfo==NULL))
>  	 xkb= NULL;
>      else xkb= keybd->key->xkbInfo->desc;
>      if ((names->keycodes==NULL)&&(names->types==NULL)&&

what was the error message that prompted this hunk? Cursory inspection of
the code seems like a keybd of NULL is valid.

> diff --git a/xkb/xkbLEDs.c b/xkb/xkbLEDs.c
> index f617537..1682671 100644
> --- a/xkb/xkbLEDs.c
> +++ b/xkb/xkbLEDs.c
> @@ -714,10 +714,12 @@ XkbSrvLedInfoPtr	sli;
>  	    }	
>  	}
>      }
> -    if ((sli->names==NULL)&&(needed_parts&XkbXI_IndicatorNamesMask))
> -	sli->names= calloc(XkbNumIndicators, sizeof(Atom));
> -    if ((sli->maps==NULL)&&(needed_parts&XkbXI_IndicatorMapsMask))
> -	sli->maps= calloc(XkbNumIndicators, sizeof(XkbIndicatorMapRec));
> +    if (sli) {
> +	if ((sli->names==NULL)&&(needed_parts&XkbXI_IndicatorNamesMask))
> +	    sli->names= calloc(XkbNumIndicators, sizeof(Atom));
> +	if ((sli->maps==NULL)&&(needed_parts&XkbXI_IndicatorMapsMask))
> +	    sli->maps= calloc(XkbNumIndicators, sizeof(XkbIndicatorMapRec));
> +    }
>      return sli;
>  }
>  
> -- 
> 1.6.3.3

these seem to be two unrelated hunks, should be two separate patches.
Reviewed-by: Peter Hutterer <peter.hutterer at who-t.net> for the second hunk
though with the commit message above.
 
Cheers,
  Peter