Re: [BISECTED REGRESSION] b43legacy broken on G4 PowerBook

From: Larry Finger
Date: Tue Jun 11 2019 - 18:25:01 EST

On 6/11/19 1:05 AM, Christoph Hellwig wrote:
On Mon, Jun 10, 2019 at 11:09:47AM -0500, Larry Finger wrote:

What might be confusing in your output is that dev->dma_mask is a pointer,
and we are setting it in dma_set_mask. That is before we only check
if the pointer is set, and later we override it. Of course this doesn't
actually explain the failure. But what is even more strange to me
is that you get a return value from dma_supported() that isn't 0 or 1,
as that function is supposed to return a boolean, and I really can't see
how mask >= __phys_to_dma(dev, min_mask), would return anything but 0
or 1. Does the output change if you use the correct printk specifiers?

i.e. with a debug patch like this:

diff --git a/kernel/dma/direct.c b/kernel/dma/direct.c
index 2c2772e9702a..9e5b30b12b10 100644
--- a/kernel/dma/direct.c
+++ b/kernel/dma/direct.c
@@ -378,6 +378,7 @@ EXPORT_SYMBOL(dma_direct_map_resource);
int dma_direct_supported(struct device *dev, u64 mask)
u64 min_mask;
+ bool ret;
@@ -391,7 +392,12 @@ int dma_direct_supported(struct device *dev, u64 mask)
* use __phys_to_dma() here so that the SME encryption mask isn't
* part of the check.
- return mask >= __phys_to_dma(dev, min_mask);
+ ret = (mask >= __phys_to_dma(dev, min_mask));
+ if (!ret)
+ dev_info(dev,
+ "%s: failed (mask = 0x%llx, min_mask = 0x%llx/0x%llx, dma bits = %d\n",
+ __func__, mask, min_mask, __phys_to_dma(dev, min_mask), ARCH_ZONE_DMA_BITS);
+ return ret;
size_t dma_direct_max_mapping_size(struct device *dev)
diff --git a/kernel/dma/mapping.c b/kernel/dma/mapping.c
index f7afdadb6770..6c57ccdee2ae 100644
--- a/kernel/dma/mapping.c
+++ b/kernel/dma/mapping.c
@@ -317,8 +317,14 @@ void arch_dma_set_mask(struct device *dev, u64 mask);
int dma_set_mask(struct device *dev, u64 mask)
- if (!dev->dma_mask || !dma_supported(dev, mask))
+ if (!dev->dma_mask) {
+ dev_info(dev, "no DMA mask set!\n");
return -EIO;
+ }
+ if (!dma_supported(dev, mask)) {
+ printk("DMA not supported\n");
+ return -EIO;
+ }
arch_dma_set_mask(dev, mask);
dma_check_mask(dev, mask);

After I got the correct formatting, the output with this patch only gives the following in dmesg:

b43-pci-bridge 0001:11:00.0: dma_direct_supported: failed (mask = 0x3fffffff, min_mask = 0x5ffff000/0x5ffff000, dma bits = 0x1f
DMA not supported
b43legacy-phy0 ERROR: The machine/kernel does not support the required 30-bit DMA mask

Your first patch did not work as the configuration does not have CONFIG_ZONE_DMA. As a result, the initial value of min_mask always starts at 32 bits and is taken down to 31 with the maximum pfn minimization. When I forced the initial value of min_mask to 30 bits, the device worked.

It is obvious that the case of a mask smaller than min_mask should be handled by the IOMMU. In my system, CONFIG_IOMMU_SUPPORT is selected. All other CONFIG variables containing IOMMU are not selected. When dma_direct_supported() fails, should the system not try for an IOMMU solution? Is the driver asking for the wrong type of memory? It is doing a dma_and_set_mask_coherent() call.