On 2021-11-11 06:50, Christoph Hellwig wrote:
Hi all,
Linus complained about the complex flow in dma_direct_alloc, so this
tries to simplify it a bit, and while I was at it I also made sure that
unencrypted pages never leak back into the page allocator.
Before I forget, I've had a quick skim of the remaining patches and
nothing more stands out. Let me know if you'd like me to find time to
check everything over in detail again for a proper review, but otherwise
I reckon we may as well get this baking in -next sooner rather than later.
Cheers,
Robin.
Changes since v2:
- don't call dma_set_decrypted on remapped memory
- move the leak printk into dma_set_encrypted
- add another local variable to clean up dma_direct_alloc
- return NULL when the is no way to make the memory coherent
Changes since v1:
- fix a missing return
- add a new patch to fix a pre-existing missing unmap
- various additional cleanups
Diffstat:
direct.c | 234
+++++++++++++++++++++++++++++++++++++--------------------------
1 file changed, 138 insertions(+), 96 deletions(-)
_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu
