On 8/13/26 20:29, Taimuraz Kaitmazov wrote: > On 8/13/26 09:44, Christian König wrote: >> An alternative would be to use DMA_BUF_IOCTL_SYNC from userspace > > Tried it against amdgpu, imported into amdxdna: stale on all 20 runs, no > better than no sync at all. SYNC_BO on the same buffer is clean on all > 20.
That is expected, as I said this is for the case when a device which can't snoop the CPU cache exports a buffer. > Which leaves me no legal way to import a buffer the CPU also reads. Is > there one I'm missing, or should a device like this just not import? Yes, exactly that. For the case where a device which can't snoop imports a buffer we simply don't have an interface at all. In other words amdxdna can export a buffer which others drivers import, but if the xdna HW is not capable of snooping the CPU cache it is not possible for the driver to provide an import functionality. Regards, Christian.
