Re: [PATCH AUTOSEL 6.14 295/642] nvme: map uring_cmd data even if address is 0

On Mon, May 05, 2025 at 04:15:47PM -0700, Caleb Sander Mateos wrote:

I wouldn't backport this change to any releases. It's a potential behavior change if a userspace application was submitting NVMe passthru commands with a NULL data pointer but nonzero data length and expecting the data buffer to be ignored. And supporting the data field set to 0 is only necessary for ublk zero-copy, which is introduced in 6.15.

I'll drop it, thanks!