From: Arnd Bergmann
> Sent: 14 August 2018 12:08
...
> > There are also a whole load of crappy __packed in that header file.
> > There might be one or two 64bit items on 32bit boundaries but
> > that can be solved without using __packed.
>
> Agreed, this likely causes problems on architectures without unaligned
> load/store instructions that end up doing byte accesses to the descriptor
> fields, which in turn can confuse the hardware, and can become very
> slow when they live in dma_alloc_coherent() memory. That looks
> like a completely unrelated issue though.
If you ever define a variable of one of those types (or embed one
in another structure that contains non-word sized items) you'll
get the entire structure misaligned.
I doubt that is what you had in mind.
Maybe you could use __packed __attribute__((aligned(4))).
But it is much better just to use a 64bit type with only 4 byte alignment
(I think there is a standard one in one of the header files).
I'd also add a compile-time assert on the length of the non-trivial
structures. That will detect all sorts of errors.
David
-
Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT,
UK
Registration No: 1397386 (Wales)
