On Mon, Jan 06, 2014 at 08:07:58PM +1100, Stephen Rothwell wrote: > Hi Andrew, > > After merging the akpm-current tree, today's linux-next build (arm > multi_v7_defconfig) failed like this: > > arch/arm/mm/init.c:199:13: error: conflicting types for 'arm_dma_zone_size' > include/linux/bootmem.h:259:11: note: previous declaration of 'arm_dma_zone_size' was here > > Caused by commit a3ae9362fafe ("mm/memblock: add memblock memory > allocation apis"). > > For this build, BOOTMEM_LOW_LIMIT is defined as __pa(MAX_DMA_ADDRESS) and > MAX_DMA_ADDRESS is defined as: > > #define MAX_DMA_ADDRESS ({ \ > extern unsigned long arm_dma_zone_size; \ > arm_dma_zone_size ? \ > (PAGE_OFFSET + arm_dma_zone_size) : 0xffffffffUL; }) > > and in arch/arm/mm/init.c, arm_dma_zone_size is declared as: > > phys_addr_t arm_dma_zone_size __read_mostly; > > Urk! :-( > > OK, so commit 364230b995214 ("ARM: use phys_addr_t for DMA zone sizes") > changed the definition of arm_dma_zone_size except it missed the one in > arch/arm/include/asm/dma.h. > > I have applied this merge fix patch for today: There's more here than just this change required here. We also need to limit it if PAGE_OFFSET + arm_dma_zone_size is greater than 4GB, as MAX_DMA_ADDRESS is a virtual address - we really don't want the 32-bit value of this wrapping. Exactly how we go about this, I'm not sure at the moment, but maybe something like this: #define MAX_DMA_ADDRESS ({ \ extern phys_addr_t arm_dma_zone_size; \ arm_dma_zone_size && arm_dma_zone_size < (0x10000000 - PAGE_OFFSET) ? \ (PAGE_OFFSET + arm_dma_zone_size) : 0xffffffffUL; }) -- Russell King ARM architecture Linux Kernel maintainer -- To unsubscribe from this list: send the line "unsubscribe linux-next" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html