On Fri 25-02-22 16:14:14, Chaitanya Kulkarni wrote: > On 2/25/22 04:05, Jan Kara wrote: > > On Fri 25-02-22 00:50:18, John Hubbard wrote: > >> Hi, > >> > >> Summary: > >> > >> This puts some prerequisites in place, including a CONFIG parameter, > >> making it possible to start converting and testing the Direct IO part of > >> each filesystem, from get_user_pages_fast(), to pin_user_pages_fast(). > >> > >> It will take "a few" kernel releases to get the whole thing done. > >> > >> Details: > >> > >> As part of fixing the "get_user_pages() + file-backed memory" problem > >> [1], and to support various COW-related fixes as well [2], we need to > >> convert the Direct IO code from get_user_pages_fast(), to > >> pin_user_pages_fast(). Because pin_user_pages*() calls require a > >> corresponding call to unpin_user_page(), the conversion is more > >> elaborate than just substitution. > >> > >> Further complicating the conversion, the block/bio layers get their > >> Direct IO pages via iov_iter_get_pages() and iov_iter_get_pages_alloc(), > >> each of which has a large number of callers. All of those callers need > >> to be audited and changed so that they call unpin_user_page(), rather > >> than put_page(). > >> > >> After quite some time exploring and consulting with people as well, it > >> is clear that this cannot be done in just one patchset. That's because, > >> not only is this large and time-consuming (for example, Chaitanya > >> Kulkarni's first reaction, after looking into the details, was, "convert > >> the remaining filesystems to use iomap, *then* convert to FOLL_PIN..."), > >> but it is also spread across many filesystems. > > > > With having modified fs/direct-io.c and fs/iomap/direct-io.c which > > filesystems do you know are missing conversion? Or is it that you just want > > to make sure with audit everything is fine? The only fs I could find > > unconverted by your changes is ceph. Am I missing something? > > if I understand your comment correctly file systems which are listed in > the list see [1] (all the credit goes to John to have a complete list) > that are not using iomap but use XXX_XXX_direct_IO() should be fine, > since in the callchain going from :- > > XXX_XXX_direct_io() > __blkdev_direct_io() > do_direct_io() > > ... > > submit_page_selection() > get/put_page() <--- > > will take care of itself ? Yes, John's changes to fs/direct-io.c should take care of these filesystems using __blkdev_direct_io(). Honza > [1] > > jfs_direct_IO() > nilfs_direct_IO() > ntfs_dirct_IO() > reiserfs_direct_IO() > udf_direct_IO() > ocfs2_dirct_IO() > affs_direct_IO() > exfat_direct_IO() > ext2_direct_IO() > fat_direct_IO() > hfs_direct_IO() > hfs_plus_direct_IO() > > -ck > > -- Jan Kara <jack@xxxxxxxx> SUSE Labs, CR
