On Thu, Jan 31, 2019 at 09:32:27PM +0100, Takashi Iwai wrote:
On Thu, 31 Jan 2019 20:54:33 +0100, Zach Riggle 🖖 wrote:
(1) Can the buffer be mremap()ed with a different offset into the buffer? This was a concern in the past and the reason for the anon_inode stuff at all. I believe that as long as the *size* of the mapping doesn't change, Linux mm will gladly permit mremap() without informing the driver.
Could you elaborate which perspective of mremap() can be a big problem? The driver interface does nothing but a standard mmap for now.
I believe the issue was that if someone could remap the buffer to gain access to memory outside the memory allocated for the PCM buffer that would be a problem. To be honest I'm surprised that this might be a general issue with mmap().