On Tuesday 21 June 2011, Nicolas Pitre wrote:

On Mon, 20 Jun 2011, Arnd Bergmann wrote:

This example is flawed. The DMA API documentation already forbids DMA to 
the stack because of cache line sharing issues.  If you declare your 
buffer outside of the function body, the compiler can't optimize away 
the buffer store anymore, and this example works as expected without any 
memory clobber.

Ok, another example, even simpler:

int f(int *dma_buf, volatile int *mmio_reg)
{
        (void) *mmio_reg; /* wait for DMA to complete */
        return *dma_buf;
}


gcc-4.4, 4.5 and 4.6 all turn this into:

        ldr     r0, [r0, #0]
        ldr     r3, [r1, #0]
        bx      lr

which means that the dma_buf variable is dereferenced before the
volatile mmio_reg variable, which opens up a race: An interrupt may have
signalled us that a DMA is in progress, so we read a MMIO register from
the device (this is guaranteed to flush the DMA on PCI and similar buses).
If we read the dma_buf before we read the mmio register, the data we get
back may be stale.

Adding a barrier() between the two turns the assembly into the expected

        ldr     r3, [r1, #0]
        ldr     r0, [r0, #0]
        bx      lr


	Arnd