Re: linux kernel warnings/errors

[Josh Triplett <josh@joshtriplett.org>]

On Tue, Feb 23, 2010 at 11:49:06AM -0800, Randy Dunlap wrote:
> Hi,
> 
> You probably know that sparse produces a ton of errors & warnings when
> run on the Linux kernel tree (a little over 1 MB in my latest 'make C=1'
> on x86_64 arch.).
> 
> I'm wondering if all of these are valid.
> 
> Examples:
> 
> 
> 1.  Use of the BUILD_BUG_ON() macro causes this error from sparse:
> 
> arch/x86/kernel/paravirt.c:101:9: error: invalid bitfield width, -1.
> 
> include/linux/kernel.h:
> 
> /* Force a compilation error if condition is true */
> #define BUILD_BUG_ON(condition) ((void)BUILD_BUG_ON_ZERO(condition))
> 
> /* Force a compilation error if condition is true, but also produce a
>    result (of value 0 and type size_t), so the expression can be used
>    e.g. in a structure initializer (or where-ever else comma expressions
>    aren't permitted). */
> #define BUILD_BUG_ON_ZERO(e) (sizeof(struct { int:-!!(e); }))

In theory that should only happen if the BUILD_BUG_ON_ZERO actually
triggers, and thus sparse fails to build just as GCC would.  If sparse
has this problem and GCC doesn't, could you please provide a minimal
test case?

> 2.  drivers/block/drbd/drbd_int.h uses __protected_by(var);
> this seems to be unknown to sparse.
> 
> #ifdef __CHECKER__
> # define __protected_by(x)       __attribute__((require_context(x,1,999,"rdwr")))
> 
> sparse says many times:
> 
> drivers/block/drbd/drbd_int.h:887:39: error: attribute 'require_context': unknown attribute

That looks broken.  A few patches went around for new Sparse
context-tracking features, one of which got reverted before the 0.4.2
release, but I don't think any of them would have allowed *that*.

- Josh Triplett