Re: [PATCH net v1 1/1] bnx2: Fix unused data compilation warning

[Jakub Kicinski <kuba@kernel.org>]

On Tue, 4 Mar 2025 15:41:19 +0200 Andy Shevchenko wrote:
> > > diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
> > > index b79925b1c433..927884f10b0f 100644
> > > --- a/include/linux/dma-mapping.h
> > > +++ b/include/linux/dma-mapping.h
> > > @@ -629,7 +629,7 @@ static inline int dma_mmap_wc(struct device *dev,
> > >  #else
> > >  #define DEFINE_DMA_UNMAP_ADDR(ADDR_NAME)
> > >  #define DEFINE_DMA_UNMAP_LEN(LEN_NAME)
> > > -#define dma_unmap_addr(PTR, ADDR_NAME)           (0)
> > > +#define dma_unmap_addr(PTR, ADDR_NAME)           (((PTR)->ADDR_NAME), 0)
> > >  #define dma_unmap_addr_set(PTR, ADDR_NAME, VAL)  do { } while (0)
> > >  #define dma_unmap_len(PTR, LEN_NAME)             (0)
> > >  #define dma_unmap_len_set(PTR, LEN_NAME, VAL)    do { } while (0)
> > > ---
> > > 
> > > Would that work?  
> 
> Actually it won't work because the variable is under the same ifdeffery.
> What will work is to spreading the ifdeffery to the users, but it doesn't any
> better than __maybe_unsused, which is compact hack (yes, I admit that it is not
> the nicest solution, but it's spread enough in the kernel).

I meant something more like (untested):

diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h
index b79925b1c433..a7ebcede43f6 100644
--- a/include/linux/dma-mapping.h
+++ b/include/linux/dma-mapping.h
@@ -629,10 +629,10 @@ static inline int dma_mmap_wc(struct device *dev,
 #else
 #define DEFINE_DMA_UNMAP_ADDR(ADDR_NAME)
 #define DEFINE_DMA_UNMAP_LEN(LEN_NAME)
-#define dma_unmap_addr(PTR, ADDR_NAME)           (0)
-#define dma_unmap_addr_set(PTR, ADDR_NAME, VAL)  do { } while (0)
-#define dma_unmap_len(PTR, LEN_NAME)             (0)
-#define dma_unmap_len_set(PTR, LEN_NAME, VAL)    do { } while (0)
+#define dma_unmap_addr(PTR, ADDR_NAME)           ({ typeof(PTR) __p __maybe_unused = PTR; 0; )}
+#define dma_unmap_addr_set(PTR, ADDR_NAME, VAL)  do { typeof(PTR) __p __maybe_unused = PTR; } while (0)
+#define dma_unmap_len(PTR, LEN_NAME)             ({ typeof(PTR) __p __maybe_unused = PTR; 0; )}
+#define dma_unmap_len_set(PTR, LEN_NAME, VAL)    do { typeof(PTR) __p __maybe_unused = PTR; } while (0)
 #endif
 
 #endif /* _LINUX_DMA_MAPPING_H */


I just don't know how much code out there depends on PTR not
existing if !CONFIG_NEED_DMA_MAP_STATE