Re: [PATCH 5.4 40/50] dma-direct: exclude dma_direct_map_resource from the min_low_pfn check

From: Greg Kroah-Hartman
Date: Fri May 08 2020 - 09:27:43 EST


On Fri, May 08, 2020 at 06:50:02PM +0530, Naresh Kamboju wrote:
> On Fri, 8 May 2020 at 18:23, Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> >
> > From: Christoph Hellwig <hch@xxxxxx>
> >
> > commit 68a33b1794665ba8a1d1ef1d3bfcc7c587d380a6 upstream.
> >
> > The valid memory address check in dma_capable only makes sense when mapping
> > normal memory, not when using dma_map_resource to map a device resource.
> > Add a new boolean argument to dma_capable to exclude that check for the
> > dma_map_resource case.
> >
> > Fixes: b12d66278dd6 ("dma-direct: check for overflows on 32 bit DMA addresses")
> > Reported-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > Signed-off-by: Christoph Hellwig <hch@xxxxxx>
> > Acked-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > Tested-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
> > Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxxxxxxxxxxxxxx>
> <trim>
> >
> > --- a/kernel/dma/direct.c
> > +++ b/kernel/dma/direct.c
> > @@ -327,7 +327,7 @@ static inline bool dma_direct_possible(s
> > size_t size)
> > {
> > return swiotlb_force != SWIOTLB_FORCE &&
> > - dma_capable(dev, dma_addr, size);
> > + dma_capable(dev, dma_addr, size, true);
>
> While building kernel Image for arm architecture the following error noticed
> on stale-rc 5.4 kernel branch.
>
> # make -sk KBUILD_BUILD_USER=TuxBuild -C/linux -j16 ARCH=arm
> CROSS_COMPILE=arm-linux-gnueabihf- HOSTCC=gcc CC="sccache
> arm-linux-gnueabihf-gcc" O=build zImage
> #
> ../kernel/dma/direct.c: In function âdma_direct_possibleâ:
> ../kernel/dma/direct.c:330:3: error: too many arguments to function
> âdma_capableâ
> 330 | dma_capable(dev, dma_addr, size, true);
> | ^~~~~~~~~~~
> In file included from ../include/linux/dma-direct.h:12,
> from ../kernel/dma/direct.c:10:
> ../arch/arm/include/asm/dma-direct.h:17:20: note: declared here
> 17 | static inline bool dma_capable(struct device *dev, dma_addr_t
> addr, size_t size)
> | ^~~~~~~~~~~
> In file included from ../include/linux/init.h:5,
> from ../include/linux/memblock.h:12,
> from ../kernel/dma/direct.c:7:
> ../kernel/dma/direct.c: In function âdma_direct_map_resourceâ:
> ../kernel/dma/direct.c:379:16: error: too many arguments to function
> âdma_capableâ
> 379 | if (unlikely(!dma_capable(dev, dma_addr, size, false))) {
> | ^~~~~~~~~~~
> ../include/linux/compiler.h:78:42: note: in definition of macro âunlikelyâ
> 78 | # define unlikely(x) __builtin_expect(!!(x), 0)
> | ^
> In file included from ../include/linux/dma-direct.h:12,
> from ../kernel/dma/direct.c:10:
> ../arch/arm/include/asm/dma-direct.h:17:20: note: declared here
> 17 | static inline bool dma_capable(struct device *dev, dma_addr_t
> addr, size_t size)
> | ^~~~~~~~~~~
> make[3]: *** [../scripts/Makefile.build:266: kernel/dma/direct.o] Error 1
> In file included from ../include/linux/string.h:6,
> from ../include/linux/dma-mapping.h:6,
> from ../include/linux/dma-direct.h:5,
> from ../kernel/dma/swiotlb.c:24:
> ../kernel/dma/swiotlb.c: In function âswiotlb_mapâ:
> ../kernel/dma/swiotlb.c:681:16: error: too many arguments to function
> âdma_capableâ
> 681 | if (unlikely(!dma_capable(dev, *dma_addr, size, true))) {
> | ^~~~~~~~~~~
> ../include/linux/compiler.h:78:42: note: in definition of macro âunlikelyâ
> 78 | # define unlikely(x) __builtin_expect(!!(x), 0)
> | ^
> In file included from ../include/linux/dma-direct.h:12,
> from ../kernel/dma/swiotlb.c:24:
> ../arch/arm/include/asm/dma-direct.h:17:20: note: declared here
> 17 | static inline bool dma_capable(struct device *dev, dma_addr_t
> addr, size_t size)
> | ^~~~~~~~~~

Ah crap, I think I remember trying to apply this in the past and running
into these issues. I'll go drop it now and push out a -rc2, thanks!

greg k-h