On 8/5/20 11:27 PM, John Paul Adrian Glaubitz wrote:
> with the same result. That's on 4.19.137. Can try a newer kernel tomorrow.
Looking at the change [1], I noticed that the ia64-specific "ia64_swiotlb_alloc_coherent()"
contained an additional check whether the DMA_BIT_MASK is 64 bits, so I added that back into
swiotlb_alloc():
diff --git a/kernel/dma/swiotlb.c b/kernel/dma/swiotlb.c
index 2a8c41f12d45..e51654180189 100644
--- a/kernel/dma/swiotlb.c
+++ b/kernel/dma/swiotlb.c
@@ -1018,6 +1018,9 @@ void *swiotlb_alloc(struct device *dev, size_t size, dma_addr_t *dma_handle,
if (gfp & __GFP_NOWARN)
attrs |= DMA_ATTR_NO_WARN;
+ if (dev->coherent_dma_mask != DMA_BIT_MASK(64))
+ gfp |= GFP_DMA32;
+
/*
* Don't print a warning when the first allocation attempt fails.
* swiotlb_alloc_coherent() will print a warning when the DMA memory
No success, unfortunately even though this additional check was the only ia64-specific
part in the old code. But I assume the problem is also that swiotlb_alloc_coherent()
got replaced by swiotlb_alloc()?
Adrian
> [1] https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux.git/commit/?id=543cea9accd9804307541cb93d3ed7ec94b07237
--
.''`. John Paul Adrian Glaubitz
: :' : Debian Developer - glaubitz@xxxxxxxxxx
`. `' Freie Universitaet Berlin - glaubitz@xxxxxxxxxxxxxxxxxxx
`- GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
|