| Message ID | ffdc5455d698d2525f2d126d0adabf9418fd76b1.1418990609.git.robin.murphy@arm.com (mailing list archive) |
|---|---|
| State | Not Applicable |
| Headers | show |
On Fri, Dec 19, 2014 at 05:39:09PM +0000, Robin Murphy wrote: > Many DMA controllers and other devices set max_segment_size to > indicate their scatter-gather capability, but have no interest in > segment_boundary_mask. However, the existence of a dma_parms structure > precludes the use of any default value, leaving them as zeros (assuming > a properly kzalloc'ed structure). If a well-behaved IOMMU (or SWIOTLB) > then tries to respect this by ensuring a mapped segment does not cross > a zero-byte boundary, hilarity ensues. > > Since zero is a nonsensical value for either parameter, treat it as an > indicator for "default", as might be expected. In the process, clean up > a bit by replacing the bare constants with slightly more meaningful > macros and removing the superfluous "else" statements. > > Signed-off-by: Robin Murphy <robin.murphy@arm.com> > --- > > Hi Vinod, dmaengine folks, > > This isn't strictly a dmaengine patch, but get_maintainer.pl pointed at you, > and there do happen to be more dmaengine drivers potentially affected by this > than anything else - the current PL330 driver blows up arm64's SWIOTLB when > running dmatest on the Juno platform, which is what brought the underlying > problem to light. That's due to include/linux/dma*, I will fix that up. I think arm folks should be able to help review this... Last few commit point to Will Decon carrying these
diff --git a/include/linux/dma-mapping.h b/include/linux/dma-mapping.h index c3007cb..99ba736 100644 --- a/include/linux/dma-mapping.h +++ b/include/linux/dma-mapping.h @@ -141,7 +141,9 @@ static inline void arch_teardown_dma_ops(struct device *dev) { } static inline unsigned int dma_get_max_seg_size(struct device *dev) { - return dev->dma_parms ? dev->dma_parms->max_segment_size : 65536; + if (dev->dma_parms && dev->dma_parms->max_segment_size) + return dev->dma_parms->max_segment_size; + return SZ_64K; } static inline unsigned int dma_set_max_seg_size(struct device *dev, @@ -150,14 +152,15 @@ static inline unsigned int dma_set_max_seg_size(struct device *dev, if (dev->dma_parms) { dev->dma_parms->max_segment_size = size; return 0; - } else - return -EIO; + } + return -EIO; } static inline unsigned long dma_get_seg_boundary(struct device *dev) { - return dev->dma_parms ? - dev->dma_parms->segment_boundary_mask : 0xffffffff; + if (dev->dma_parms && dev->dma_parms->segment_boundary_mask) + return dev->dma_parms->segment_boundary_mask; + return DMA_BIT_MASK(32); } static inline int dma_set_seg_boundary(struct device *dev, unsigned long mask) @@ -165,8 +168,8 @@ static inline int dma_set_seg_boundary(struct device *dev, unsigned long mask) if (dev->dma_parms) { dev->dma_parms->segment_boundary_mask = mask; return 0; - } else - return -EIO; + } + return -EIO; } #ifndef dma_max_pfn
Many DMA controllers and other devices set max_segment_size to indicate their scatter-gather capability, but have no interest in segment_boundary_mask. However, the existence of a dma_parms structure precludes the use of any default value, leaving them as zeros (assuming a properly kzalloc'ed structure). If a well-behaved IOMMU (or SWIOTLB) then tries to respect this by ensuring a mapped segment does not cross a zero-byte boundary, hilarity ensues. Since zero is a nonsensical value for either parameter, treat it as an indicator for "default", as might be expected. In the process, clean up a bit by replacing the bare constants with slightly more meaningful macros and removing the superfluous "else" statements. Signed-off-by: Robin Murphy <robin.murphy@arm.com> --- Hi Vinod, dmaengine folks, This isn't strictly a dmaengine patch, but get_maintainer.pl pointed at you, and there do happen to be more dmaengine drivers potentially affected by this than anything else - the current PL330 driver blows up arm64's SWIOTLB when running dmatest on the Juno platform, which is what brought the underlying problem to light. include/linux/dma-mapping.h | 17 ++++++++++------- 1 file changed, 10 insertions(+), 7 deletions(-)