This patchset is some fixes to MemoryRegionOps that were
accidentally setting odd .impl values, plus an extra assert
that we can now add that a MemoryRegionOps doesn't set
.impl.unaligned true but .valid.unaligned false. They're
originally from the CJ Chen / Tomoyuki Hirose patchset that
aims to support proper synthesis of unaligned accesses in
the memory system core. I still want to get that patchset
upstream, but for the moment this is a set of simple
cleanups:

 * riscv_iommu_trap_ops set .impl.unaligned but not
   .valid.unaligned
 * npcm7xx_fiu_flash_ops failed to set any .impl at all
 * xtensa_mx_pic_ops failed to set any .impl, and was
   less strict than it could be about .valid

Patches 1 and 4 have been reviewed.

thanks
-- PMM

CJ Chen (2):
  hw/riscv: iommu-trap: remove .impl.unaligned = true
  system/memory: assert on invalid MemoryRegionOps .unaligned combo

Peter Maydell (2):
  hw/npcm7xx_fiu: Specify .impl for npcm7xx_fiu_flash_ops
  hw/xtensa/mx_pic: Specify xtensa_mx_pic_ops .impl settings

 hw/riscv/riscv-iommu.c |  1 -
 hw/ssi/npcm7xx_fiu.c   |  5 +++++
 hw/xtensa/mx_pic.c     | 27 +++++++++++++++++++++++++++
 system/memory.c        |  1 +
 4 files changed, 33 insertions(+), 1 deletion(-)

-- 
2.43.0

On Tue, Apr 28, 2026 at 10:33:35AM +0100, Peter Maydell wrote:
> This patchset is some fixes to MemoryRegionOps that were
> accidentally setting odd .impl values, plus an extra assert
> that we can now add that a MemoryRegionOps doesn't set
> .impl.unaligned true but .valid.unaligned false. They're
> originally from the CJ Chen / Tomoyuki Hirose patchset that
> aims to support proper synthesis of unaligned accesses in
> the memory system core. I still want to get that patchset
> upstream, but for the moment this is a set of simple
> cleanups:
> 
>  * riscv_iommu_trap_ops set .impl.unaligned but not
>    .valid.unaligned
>  * npcm7xx_fiu_flash_ops failed to set any .impl at all
>  * xtensa_mx_pic_ops failed to set any .impl, and was
>    less strict than it could be about .valid
> 
> Patches 1 and 4 have been reviewed.

I queued it..

-- 
Peter Xu

On Tue, Apr 28, 2026 at 10:33:35AM +0100, Peter Maydell wrote:
> This patchset is some fixes to MemoryRegionOps that were
> accidentally setting odd .impl values, plus an extra assert
> that we can now add that a MemoryRegionOps doesn't set
> .impl.unaligned true but .valid.unaligned false. They're
> originally from the CJ Chen / Tomoyuki Hirose patchset that
> aims to support proper synthesis of unaligned accesses in
> the memory system core. I still want to get that patchset
> upstream, but for the moment this is a set of simple
> cleanups:
> 
>  * riscv_iommu_trap_ops set .impl.unaligned but not
>    .valid.unaligned
>  * npcm7xx_fiu_flash_ops failed to set any .impl at all
>  * xtensa_mx_pic_ops failed to set any .impl, and was
>    less strict than it could be about .valid
> 
> Patches 1 and 4 have been reviewed.
> 
> thanks
> -- PMM
> 
> CJ Chen (2):
>   hw/riscv: iommu-trap: remove .impl.unaligned = true
>   system/memory: assert on invalid MemoryRegionOps .unaligned combo
> 
> Peter Maydell (2):
>   hw/npcm7xx_fiu: Specify .impl for npcm7xx_fiu_flash_ops
>   hw/xtensa/mx_pic: Specify xtensa_mx_pic_ops .impl settings

Reviewed-by: Peter Xu <peterx@redhat.com>

-- 
Peter Xu