Re: [PATCH v2 2/2] staging: ion: chunk_heap: use %pad for printing dma_addr_t's

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Mon, May 04, 2015 at 10:59:39PM -0700, Mitchel Humpherys wrote:
> On Mon, May 04 2015 at 01:05:50 PM, Colin Cross <ccross@android.com> wrote:
> > On Mon, May 4, 2015 at 1:22 AM, Dan Carpenter <dan.carpenter@oracle.com> wrote:
> >> On Thu, Apr 09, 2015 at 06:10:04PM -0700, Mitchel Humpherys wrote:
> >>> We're currently using %lu and %ld to print some variables of type
> >>> dma_addr_t, which results in the following warning when dma_addr_t is
> >>> 64-bits wide:
> >>>
> >>>     drivers/staging/android/ion/ion_chunk_heap.c: In function 'ion_chunk_heap_create':
> >>>     drivers/staging/android/ion/ion_chunk_heap.c:176:2: warning: format '%lu' expects argument of type 'long unsigned int', but argument 3 has type 'dma_addr_t' [-Wformat=]
> >>>       pr_info("%s: base %lu size %zu align %ld\n", __func__, chunk_heap->base,
> >>>       ^
> >>>     drivers/staging/android/ion/ion_chunk_heap.c:176:2: warning: format '%ld' expects argument of type 'long int', but argument 5 has type 'dma_addr_t' [-Wformat=]
> >>>
> >>> Fix this by using %pad as instructed in printk-formats.txt.
> >>>
> >>> Signed-off-by: Mitchel Humpherys <mitchelh@codeaurora.org>
> >>
> >> This one was just merged and I was about to email you that it introduces
> >> some new Smatch warnings, but actually looking at it, it's just wrong.
> >>
> >> We want to print "chunk_heap->base" and not "&chunk_heap->base".
> >
> > This would be correct if base was a dma_addr_t...
> >
> >> And anyway "&chunk_heap->base" is a regular pointer, not a dma_addr_t.
> >
> > But it is actually an ion_phys_addr_t, which is currently typedef'd to
> > unsigned long.  Are you using a local patch that replaces
> > ion_phys_addr_t with dma_addr_t?
> >
> >> So please send a new patch that removes the &.
> >
> > Removing the & is not correct, lib/vsprintf.c will dereference the arg
> > for %pad or %pap.  I think this patch should just be dropped, the old
> > %lu was correct for what is in Linus' tree.
> 
> Ah, you're absolutely correct.  We have a local patch that makes
> ion_phys_addr_t a dma_addr_t (needed for LPAE), which is why I needed to
> convert the printk format...
> 
> Greg, can you please drop this patch from your tree?  We'll only need
> this if/when mainline Ion gets LPAE support...

Now dropped, thanks.

greg k-h