I've just pushed patch #1 to drm-misc-fixes. I'm still wondering if #2 is actually optimal. From what I've heard only 2MiB should be used and especially 64k doesn't make sense any more on modern HW.
Regards, Christian. On 10/7/26 14:07, Davidlohr Bueso wrote: > ping? if nothing else patch 1 should be picked up. > > Thanks, > Davidlohr > > On Mon, 28 Sep 2026, Davidlohr Bueso wrote: > >> Hello, >> >> Changes from v1 >> (https://lore.kernel.org/all/[email protected]/) >> >> - The bug fix is now patch 1, unchanged from v1, picked up tags. >> - Removed patch that adjusts per base page size (arm). >> - The 2MB order is x86 only, so Arm keeps what it has. Left other platforms >> as is as well to reduce the change impact. (Christian) >> - The orders and the fragmentation trade-off are documented, both in the >> code and in the userspace heap documentation. (Christian) >> >> Thanks! >> >> Davidlohr Bueso (2): >> dma-buf: validate the attachment before asserting its reservation lock >> dma-buf: system_heap: allocate 2MB chunks on x86 >> >> Documentation/userspace-api/dma-buf-heaps.rst | 4 ++++ >> drivers/dma-buf/dma-buf-mapping.c | 4 ++-- >> drivers/dma-buf/heaps/system_heap.c | 24 ++++++++++++++----- >> 3 files changed, 24 insertions(+), 8 deletions(-) >> >> -- >> 2.39.5 >>
