On Wed, Dec 15, 2010 at 12:39:20PM +0000, Catalin Marinas wrote:
> On 13 December 2010 16:29, Russell King - ARM Linux
> <linux@xxxxxxxxxxxxxxxx> wrote:
> > On Mon, Dec 13, 2010 at 03:52:20PM +0000, Catalin Marinas wrote:
> >> On 10 December 2010 17:03, Russell King - ARM Linux
> >> <linux@xxxxxxxxxxxxxxxx> wrote:
> >> > On Fri, Dec 10, 2010 at 12:03:07PM +0100, Janusz Krzysztofik wrote:
> >> >> void __init omap1_camera_init(void *info)
> >> >> {
> >> >> struct platform_device *dev = &omap1_camera_device;
> >> >> + dma_addr_t paddr = omap1_camera_phys_mempool_base;
> >> >> + dma_addr_t size = omap1_camera_phys_mempool_size;
> >> >> int ret;
> >> >>
> >> >> dev->dev.platform_data = info;
> >> >>
> >> >> + if (paddr) {
> >> >> + if (dma_declare_coherent_memory(&dev->dev, paddr, paddr, size,
> >> >> + DMA_MEMORY_MAP | DMA_MEMORY_EXCLUSIVE))
> >> >
> >> > Although this works, you're ending up with SDRAM being mapped via
> >> > ioremap, which uses MT_DEVICE - so what is SDRAM ends up being
> >> > mapped as if it were a device.
> >>
> >> BTW, does the generic dma_declare_coherent_memory() does the correct
> >> thing in using ioremap()?
> >
> > I argue it doesn't, as I said above. It means we map SDRAM as device
> > memory, and as I understand the way interconnects work, it's entirely
> > possible that this may not result in the SDRAM being accessible.
> [...]
> > So, not only does this fail the kernel's own rules, but as we already know,
> > it fails the architecture's restrictions with multiple mappings of memory
> > when used with SDRAM, and it also maps main memory as a device. I wonder
> > how many more things this broken API needs to do wrong before it's current
> > implementation is consigned to the circular filing cabinet.
>
> Should we not try to fix the generic code and still allow platforms to
> use dma_declare_coherent_memory() in a safer way? I guess it may need
> some arguing/explanation on linux-arch.
I think so - one of the issues with dma_declare_coherent_memory() is that
it's original intention (as I understand it) was to allow drivers to use
on-device dma coherent memory.
Eg, a network controller with its own local SRAM which it can fetch DMA
descriptors from, which tells it where in the bus address space to fetch
packets from. This SRAM is not part of the hosts memory, but is on the
peripheral's bus, and so ioremap() (or maybe ioremap_wc()) would be
appropriate for it.
However, ioremap() on system RAM (even that which has been taken out on
the memory map) may be problematical.
I think the correct solution would be to revise the interface so it takes
a void * pointer, which can be handed out by dma_alloc_coherent() directly
without the API having to worry about how to map the memory. IOW, push
the mapping of that memory up a level to the caller of
dma_declare_coherent_memory().
We can then sanely do preallocations via dma_coherent_alloc() and caching
them back into dma_declare_coherent_memory() without creating additional
mappings which cause architectural violations.
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
