On Fri, May 28, 2021 at 7:07 AM Lin, Ming <minggr@xxxxxxxxx> wrote: > > Does something like following draft patch on the right track? No, I don't think this can work: > + _dst_pte = pte_mkspecial(pfn_pte(my_zero_pfn(dst_addr), > + vma->vm_page_prot)); You can't just blindly insert the zero pfn - for a shared write mapping, that would actually allow writes to the zeropage. That would be horrible. So it would have to do all the same things that it does for a page that is inside the inode size. I do also dislike how it's a per-inode flag - so it would affect other mappings of the same shared memory segment too. But considering that the page would have to be part of the page cache for that shmem inode, that may be inevitable. But it sure does smell a bit. Oh, and if we make this kind of magic shmem extension, Hugh Dickins should be part of the conversation too. Hugh, you probably saw the original on linux-mm, but I'm adding you explicitly to the participants here. .. and if you didn't see the background, here it is https://lore.kernel.org/linux-mm/vs1Us2sm4qmfvLOqNat0-r16GyfmWzqUzQ4KHbXJwEcjhzeoQ4sBTxx7QXDG9B6zk5AeT7FsNb3CSr94LaKy6Novh1fbbw8D_BBxYsbPLms=@emersion.fr/ for your edification.. Linus
