Until now, only 32-bit DMA addressing was allowed, following a report on
some old Intel machine that dropped 64-bit PCIe packets, even though
pci_set_dma_mask() was successful with DMA_BIT_MASK(64).
But then came TI's Keystone II chip (ARM Cortex A15 + DSPs), which refuses
32-bit DMA addressing (for good reasons). So 64-bit DMA is allowed as a
fallback option.
Signed-off-by: Eli Billauer <[email protected]>
---
drivers/char/xillybus/xillybus_pcie.c | 10 ++++++----
1 files changed, 6 insertions(+), 4 deletions(-)
diff --git a/drivers/char/xillybus/xillybus_pcie.c b/drivers/char/xillybus/xillybus_pcie.c
index d8266bc..9418300 100644
--- a/drivers/char/xillybus/xillybus_pcie.c
+++ b/drivers/char/xillybus/xillybus_pcie.c
@@ -193,14 +193,16 @@ static int xilly_probe(struct pci_dev *pdev,
}
/*
- * In theory, an attempt to set the DMA mask to 64 and dma_using_dac=1
- * is the right thing. But some unclever PCIe drivers report it's OK
- * when the hardware drops those 64-bit PCIe packets. So trust
- * nobody and use 32 bits DMA addressing in any case.
+ * Some (old and buggy?) hardware drops 64-bit addressed PCIe packets,
+ * even when the PCIe driver claims that a 64-bit mask is OK. On the
+ * other hand, on some architectures, 64-bit addressing is mandatory.
+ * So go for the 64-bit mask only when failing is the other option.
*/
if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(32))) {
endpoint->dma_using_dac = 0;
+ } else if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
+ endpoint->dma_using_dac = 1;
} else {
dev_err(endpoint->dev, "Failed to set DMA mask. Aborting.\n");
return -ENODEV;
--
1.7.2.3