On 02/09/2015 06:27 AM, Catalin Marinas wrote:
On Fri, Feb 06, 2015 at 03:15:20PM -0500, Murali Karicheri wrote:diff --git a/drivers/of/device.c b/drivers/of/device.c index 314c8a9..44209fa 100644 --- a/drivers/of/device.c +++ b/drivers/of/device.c @@ -83,17 +83,18 @@ int of_device_add(struct platform_device *ofdev) */ void of_dma_configure(struct device *dev, struct device_node *np) { - u64 dma_addr, paddr, size; + u64 dma_addr = 0, paddr, size; int ret; bool coherent; - unsigned long offset; + unsigned long offset = 0; struct iommu_ops *iommu; /* * Set default dma-mask to 32 bit. Drivers are expected to setup * the correct supported dma_mask. */ - dev->coherent_dma_mask = DMA_BIT_MASK(32); + if (!dev->coherent_dma_mask) + dev->coherent_dma_mask = DMA_BIT_MASK(32); /* * Set it to coherent_dma_mask by default if the architecture @@ -102,11 +103,14 @@ void of_dma_configure(struct device *dev, struct device_node *np) if (!dev->dma_mask) dev->dma_mask =&dev->coherent_dma_mask; + /* + * Set default size to cover the 32-bit. Drivers are expected to setup + * the correct size and dma_mask. + */Nitpick: drivers don't set up the size, just the mask. We read the size from DT.
Not sure how this came, I will update this.
I think dma_addr and paddr is both modified on error. So this change can't be done. I will revert the logic back to original.+ size = 1ULL<< 32; + ret = of_dma_get_range(np,&dma_addr,&paddr,&size); - if (ret< 0) { - dma_addr = offset = 0; - size = dev->coherent_dma_mask + 1;Do you assume that, on error, of_dma_get_range() does not touch any of the dma_addr/paddr/size? It looks safer to me to leave it as the original.
- } else { + if (!ret) { offset = PFN_DOWN(paddr - dma_addr); /* @@ -128,6 +132,15 @@ void of_dma_configure(struct device *dev, struct device_node *np) dev->dma_pfn_offset = offset; + /* + * Limit coherent and dma mask based on size and default mask + * set by the driver. + */ + dev->coherent_dma_mask = min(dev->coherent_dma_mask, + DMA_BIT_MASK(ilog2(dma_addr + size)));Do we need to cover the case where size is incorrectly set to 0 in the DT? At least some warning and leaving the mask unchanged.
The previous series already warn and return if the size is incorrect (Patch v6 3/7) and leave the dma mask unchanged. So we don't have to do anything in this patch. I will fix the above and re-send.
+ *dev->dma_mask = min((*dev->dma_mask), + DMA_BIT_MASK(ilog2(dma_addr + size))); + coherent = of_dma_is_coherent(np); dev_dbg(dev, "device is%sdma coherent\n", coherent ? " " : " not ");
-- Murali Karicheri Linux Kernel, Texas Instruments -- To unsubscribe from this list: send the line "unsubscribe devicetree" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
