lkml.org 
[lkml]   [2018]   [May]   [18]   [last100]   RSS Feed
Views: [wrap][no wrap]   [headers]  [forward] 
 
Messages in this thread
/
Subjectdma_sync_*_for_cpu and direction=TO_DEVICE (was Re: [PATCH 02/20] dma-mapping: provide a generic dma-noncoherent implementation)
From
Date
On 05/18/2018 06:11 AM, Alexey Brodkin wrote:
> void arch_sync_dma_for_device(struct device *dev, phys_addr_t paddr,
> size_t size, enum dma_data_direction dir)
> {
> + if (dir != DMA_TO_DEVICE){
> + dump_stack();
> + printk(" *** %s@%d: DMA direction is %s instead of %s\n",
> + __func__, __LINE__, dir_to_str(dir), dir_to_str(DMA_TO_DEVICE));
> + }
> +
> return _dma_cache_sync(dev, paddr, size, dir);
> }
>
> void arch_sync_dma_for_cpu(struct device *dev, phys_addr_t paddr,
> size_t size, enum dma_data_direction dir)
> {
> + if (dir != DMA_FROM_DEVICE) {
> + dump_stack();
> + printk(" *** %s@%d: DMA direction is %s instead of %s\n",
> + __func__, __LINE__, dir_to_str(dir), dir_to_str(DMA_FROM_DEVICE));
> + }
> +
> return _dma_cache_sync(dev, paddr, size, dir);
> }

...
> In case of MMC/DW_MCI (AKA DesignWare MobileStorage controller) that's an execution flow:
> 1) __dw_mci_start_request()
> 2) dw_mci_pre_dma_transfer()
> 3) dma_map_sg(..., mmc_get_dma_dir(data))
>
> Note mmc_get_dma_dir() is just "data->flags & MMC_DATA_WRITE ? DMA_TO_DEVICE : DMA_FROM_DEVICE".
> I.e. if we're preparing for sending data dma_noncoherent_map_sg() will have DMA_TO_DEVICE which
> is quite OK for passing to dma_noncoherent_sync_sg_for_device() but in case of reading we'll have
> DMA_FROM_DEVICE which we'll pass to dma_noncoherent_sync_sg_for_device() in dma_noncoherent_map_sg().
>
> I'd say this is not entirely correct because IMHO arch_sync_dma_for_cpu() is supposed to only be used
> in case of DMA_FROM_DEVICE and arch_sync_dma_for_device() only in case of DMA_TO_DEVICE.

So roughly 10 years ago, some kernel rookie name Vineet Gupta, asked the exact
same question :-)

http://kernelnewbies.kernelnewbies.narkive.com/aGW1QcDv/query-about-dma-sync-for-cpu-and-direction-to-device

I never understood the need for this direction. And if memory serves me right, at
that time I was seeing twice the amount of cache flushing !

> But the real fix of my problem is:
> ---------------------------------------->8------------------------------------
> --- a/lib/dma-noncoherent.c
> +++ b/lib/dma-noncoherent.c
> @@ -35,7 +35,7 @@ static dma_addr_t dma_noncoherent_map_page(struct device *dev, struct page *page
>
> addr = dma_direct_map_page(dev, page, offset, size, dir, attrs);
> if (!dma_mapping_error(dev, addr) && !(attrs & DMA_ATTR_SKIP_CPU_SYNC))
> - arch_sync_dma_for_device(dev, page_to_phys(page), size, dir);
> + arch_sync_dma_for_device(dev, page_to_phys(page) + offset, size, dir);
> return addr;
> }
> ---------------------------------------->8------------------------------------
>
> You seem to lost an offset in the page so if we happen to have a buffer not aligned to
> a page boundary then we were obviously corrupting data outside our data :)

Neat !

\
 
 \ /
  Last update: 2018-05-18 19:21    [W:0.078 / U:28.204 seconds]
©2003-2018 Jasper Spaans|hosted at Digital Ocean and TransIP|Read the blog|Advertise on this site