linux.git/arch/openrisc/kernel/dma.c, branch v6.2

openrisc: Fix pagewalk usage in arch_dma_{clear, set}_uncached

2022-10-08T05:11:29+00:00

Since commit 8782fb61cc848 ("mm: pagewalk: Fix race between unmap and page
walker"), walk_page_range() on kernel ranges won't work anymore,
walk_page_range_novma() must be used instead.

Note: I don't have an openrisc development setup, so this is completely
untested.

Fixes: 8782fb61cc848 ("mm: pagewalk: Fix race between unmap and page walker")
Signed-off-by: Jann Horn 
Signed-off-by: Stafford Horne

openrisc: fix SMP tlb flush NULL pointer dereference

2021-11-03T11:48:29+00:00

Throughout the OpenRISC kernel port VMA is passed as NULL when flushing
kernel tlb entries.  Somehow this was missed when I was testing
c28b27416da9 ("openrisc: Implement proper SMP tlb flushing") and now the
SMP kernel fails to completely boot.

In OpenRISC VMA is used only to determine which cores need to have their
TLB entries flushed.

This patch updates the logic to flush tlbs on all cores when the VMA is
passed as NULL.  Also, we update places VMA is passed as NULL to use
flush_tlb_kernel_range instead.  Now, the only place VMA is passed as
NULL is in the implementation of flush_tlb_kernel_range.

Fixes: c28b27416da9 ("openrisc: Implement proper SMP tlb flushing")
Reported-by: Jan Henrik Weinstock 
Signed-off-by: Stafford Horne

dma-mapping: merge into

2020-10-06T05:07:06+00:00

Move more nitty gritty DMA implementation details into the common
internal header.

Signed-off-by: Christoph Hellwig

openrisc: fix boot oops when DEBUG_VM is enabled

2020-06-26T07:27:36+00:00

Since v5.8-rc1 OpenRISC Linux fails to boot when DEBUG_VM is enabled.
This has been bisected to commit 42fc541404f2 ("mmap locking API: add
mmap_assert_locked() and mmap_assert_write_locked()").

The added locking checks exposed the issue that OpenRISC was not taking
this mmap lock when during page walks for DMA operations.  This patch
locks and unlocks the mmap lock for page walking.

Link: http://lkml.kernel.org/r/20200617090247.1680188-1-shorne@gmail.com
Fixes: 42fc541404f2 ("mmap locking API: add mmap_assert_locked() and mmap_assert_write_locked()"
Signed-off-by: Stafford Horne 
Reviewed-by: Michel Lespinasse 
Cc: Jonas Bonn 
Cc: Stefan Kristiansson 
Cc: Jason Gunthorpe 
Cc: Steven Price 
Cc: Thomas Hellstrom 
Cc: Robin Murphy 
Cc: Vlastimil Babka 
Cc: Daniel Jordan 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

openrisc: use the generic in-place uncached DMA allocator

2020-03-16T09:48:16+00:00

Switch openrisc to use the dma-direct allocator and just provide the
hooks for setting memory uncached or cached.

Signed-off-by: Christoph Hellwig 
Reviewed-by: Stafford Horne 
Reviewed-by: Robin Murphy

dma-mapping: drop the dev argument to arch_sync_dma_for_*

2019-11-20T19:31:38+00:00

These are pure cache maintainance routines, so drop the unused
struct device argument.

Signed-off-by: Christoph Hellwig 
Suggested-by: Daniel Vetter

pagewalk: separate function pointers from iterator data

2019-09-07T07:28:04+00:00

The mm_walk structure currently mixed data and code.  Split out the
operations vectors into a new mm_walk_ops structure, and while we are
changing the API also declare the mm_walk structure inside the
walk_page_range and walk_page_vma functions.

Based on patch from Linus Torvalds.

Link: https://lore.kernel.org/r/20190828141955.22210-3-hch@lst.de
Signed-off-by: Christoph Hellwig 
Reviewed-by: Thomas Hellstrom 
Reviewed-by: Steven Price 
Reviewed-by: Jason Gunthorpe 
Signed-off-by: Jason Gunthorpe

mm: split out a new pagewalk.h header from mm.h

2019-09-07T07:28:04+00:00

Add a new header for the two handful of users of the walk_page_range /
walk_page_vma interface instead of polluting all users of mm.h with it.

Link: https://lore.kernel.org/r/20190828141955.22210-2-hch@lst.de
Signed-off-by: Christoph Hellwig 
Reviewed-by: Thomas Hellstrom 
Reviewed-by: Steven Price 
Reviewed-by: Jason Gunthorpe 
Signed-off-by: Jason Gunthorpe

Merge tag 'dma-mapping-5.3' of git://git.infradead.org/users/hch/dma-mapping

2019-07-12T22:13:55+00:00

Pull dma-mapping updates from Christoph Hellwig:

 - move the USB special case that bounced DMA through a device bar into
   the USB code instead of handling it in the common DMA code (Laurentiu
   Tudor and Fredrik Noring)

 - don't dip into the global CMA pool for single page allocations
   (Nicolin Chen)

 - fix a crash when allocating memory for the atomic pool failed during
   boot (Florian Fainelli)

 - move support for MIPS-style uncached segments to the common code and
   use that for MIPS and nios2 (me)

 - make support for DMA_ATTR_NON_CONSISTENT and
   DMA_ATTR_NO_KERNEL_MAPPING generic (me)

 - convert nds32 to the generic remapping allocator (me)

* tag 'dma-mapping-5.3' of git://git.infradead.org/users/hch/dma-mapping: (29 commits)
  dma-mapping: mark dma_alloc_need_uncached as __always_inline
  MIPS: only select ARCH_HAS_UNCACHED_SEGMENT for non-coherent platforms
  usb: host: Fix excessive alignment restriction for local memory allocations
  lib/genalloc.c: Add algorithm, align and zeroed family of DMA allocators
  nios2: use the generic uncached segment support in dma-direct
  nds32: use the generic remapping allocator for coherent DMA allocations
  arc: use the generic remapping allocator for coherent DMA allocations
  dma-direct: handle DMA_ATTR_NO_KERNEL_MAPPING in common code
  dma-direct: handle DMA_ATTR_NON_CONSISTENT in common code
  dma-mapping: add a dma_alloc_need_uncached helper
  openrisc: remove the partial DMA_ATTR_NON_CONSISTENT support
  arc: remove the partial DMA_ATTR_NON_CONSISTENT support
  arm-nommu: remove the partial DMA_ATTR_NON_CONSISTENT support
  ARM: dma-mapping: allow larger DMA mask than supported
  dma-mapping: truncate dma masks to what dma_addr_t can hold
  iommu/dma: Apply dma_{alloc,free}_contiguous functions
  dma-remap: Avoid de-referencing NULL atomic_pool
  MIPS: use the generic uncached segment support in dma-direct
  dma-direct: provide generic support for uncached kernel segments
  au1100fb: fix DMA API abuse
  ...

openrisc: remove the partial DMA_ATTR_NON_CONSISTENT support

2019-06-25T06:14:39+00:00

The openrisc DMA code supports DMA_ATTR_NON_CONSISTENT allocations, but
does not provide a cache_sync operation.  This means any user of it
will never be able to actually transfer cache ownership and thus cause
coherency bugs.

Signed-off-by: Christoph Hellwig 
Acked-by: Stafford Horne