[RFC PATCH v4 0/5] ARM: Fix dma_alloc_coherent() and friends for NOMMU

[Vladimir Murzin]

On 11/01/17 13:17, Benjamin Gaignard wrote:
> 2017-01-10 15:18 GMT+01:00 Vladimir Murzin <vladimir.murzin at arm.com>:
>> Hi,
>>
>> It seem that addition of cache support for M-class cpus uncovered
>> latent bug in DMA usage. NOMMU memory model has been treated as being
>> always consistent; however, for R/M classes of cpu memory can be
>> covered by MPU which in turn might configure RAM as Normal
>> i.e. bufferable and cacheable. It breaks dma_alloc_coherent() and
>> friends, since data can stuck in caches now or be buffered.
>>
>> This patch set is trying to address the issue by providing region of
>> memory suitable for consistent DMA operations. It is supposed that
>> such region is marked by MPU as non-cacheable. Robin suggested to
>> advertise such memory as reserved shared-dma-pool, rather then using
>> homebrew command line option, and extend dma-coherent to provide
>> default DMA area in the similar way as it is done for CMA (PATCH
>> 2/5). It allows us to offload all bookkeeping on generic coherent DMA
>> framework, and it is seems that it might be reused by other
>> architectures like c6x and blackfin.
>>
>> Dedicated DMA region is required for cases other than:
>>  - MMU/MPU is off
>>  - cpu is v7m w/o cache support
>>  - device is coherent
>>
>> In case one of the above conditions is true dma operations are forced
>> to be coherent and wired with dma_noop_ops.
>>
>> To make life easier NOMMU dma operations are kept in separate
>> compilation unit.
>>
>> Since the issue was reported in the same time as Benjamin sent his
>> patch [1] to allow mmap for NOMMU, his case is also addressed in this
>> series (PATCH 1/5 and PATCH 3/5).
>>
>> Thanks!
> 
> I have tested this v4 on my setup (stm32f4, no cache, no MPU) and unfortunately
> it doesn't work with my drm/kms driver.

I guess the same is for fbmem, but would be better to have confirmation since
amba-clcd I use has not been ported to drm/kms (yet), so I can't test.

> I haven't any errors but nothing is displayed unlike what I have when
> using current dma-mapping
> code.
> I guess the issue is coming from dma-noop where __get_free_pages() is
> used instead of alloc_pages()
> in dma-mapping.

Unless I've missed something bellow is a call stack for both

#1
__alloc_simple_buffer
	__dma_alloc_buffer
		alloc_pages
		split_page
		__dma_clear_buffer
			memset
	page_address

#2
__get_free_pages
	alloc_pages
	page_address

So the difference is that nommu case in dma-mapping.c memzeros memory, handles
DMA_ATTR_NO_KERNEL_MAPPING and does optimisation of memory usage.

Is something from above critical for your driver?

> 
> Since my hardware doesn't have cache or MPU (and so use dma-noop) I
> haven't reserved specific memory region.
> Buffer addresses and vma parameters look correct... What could I have
> miss here ?

No ideas, sorry...

Cheers
Vladimir

> 
> Benjamin
> 
>>
>> [1] http://www.armlinux.org.uk/developer/patches/viewpatch.php?id=8633/1
>>
>> Vladimir Murzin (5):
>>   dma: Add simple dma_noop_mmap
>>   drivers: dma-coherent: Introduce default DMA pool
>>   ARM: NOMMU: Introduce dma operations for noMMU
>>   ARM: NOMMU: Set ARM_DMA_MEM_BUFFERABLE for M-class cpus
>>   ARM: dma-mapping: Remove traces of NOMMU code
>>
>>  .../bindings/reserved-memory/reserved-memory.txt   |   3 +
>>  arch/arm/include/asm/dma-mapping.h                 |   3 +-
>>  arch/arm/mm/Kconfig                                |   2 +-
>>  arch/arm/mm/Makefile                               |   5 +-
>>  arch/arm/mm/dma-mapping-nommu.c                    | 252 +++++++++++++++++++++
>>  arch/arm/mm/dma-mapping.c                          |  26 +--
>>  drivers/base/dma-coherent.c                        |  59 ++++-
>>  lib/dma-noop.c                                     |  21 ++
>>  8 files changed, 335 insertions(+), 36 deletions(-)
>>  create mode 100644 arch/arm/mm/dma-mapping-nommu.c
>>
>> --
>> 2.0.0
>>
> 
> 
>