On Wed, Nov 8, 2017 at 6:54 AM, Greentime Hu [off-list ref] wrote:

+/*
+ * The DMA mask corresponding to the maximum bus address allocatable
+ * using GFP_DMA.  The default here places no restriction on DMA
+ * allocations.  This must be the smallest DMA mask in the system,
+ * so a successful GFP_DMA allocation will always satisfy this.
+ */
+#ifndef ISA_DMA_THRESHOLD
+#define ISA_DMA_THRESHOLD      (0xffffffffULL)
+#endif

I see this one only in powerpc, I think it can be removed.

+
+/*
+ * Optional device DMA address remapping. Do _not_ use directly!
+ * We should really eliminate virt_to_bus() here - it's deprecated.
+ */
+#define page_to_dma(dev, page)         ((dma_addr_t)__virt_to_phys((unsigned long)page_address(page)))
+#define dma_to_virt(dev, addr)         ((void *)__phys_to_virt(addr))
+#define virt_to_dma(dev, addr)         ((dma_addr_t)__virt_to_phys((unsigned long)(addr)))

This looks like an older comment, I think we have eliminated them from the
mainline kernel in all drivers, so you should remove them here too.

+
+/*
+ * This should be the size of the virtually indexed cache/ways,
+ * whichever is greater since the cache aliases every size/ways
+ * bytes.
+ */
+#define        SHMLBA  (4 * PAGE_SIZE) /* attach addr a multiple of this */
+#define        REALSHMLBA      SHMLBA

I don't see REALSHMLBA anywhere in the kernel, do you need it?

For SHMLBA, I think it should be defined as an absolute number, using
the maximum
that you might need for any possible value of PAGE_SIZE. Otherwise user space
has a much harder time figuring out what it should use.

        Arnd