On Wed, Jul 24, 2024 at 11:34:57AM +1000, Dave Chinner wrote: > This is to prevent the deadlocks on upper->lower->upper and > lower->upper->lower filesystem recursion via GFP_KERNEL memory > allocation and reclaim recursing between the two filesystems. This > is especially relevant for filesystems with ->writepage methods that > can be called from direct reclaim. Hence allocations in this path > need to be at least NOFS to prevent recursion back into the upper > filesystem from writeback into the lower filesystem. FYI, we're making good progress on removing ->writepage. $ git grep '\.writepage\>.*=' fs/ceph/addr.c: .writepage = ceph_writepage, fs/ecryptfs/mmap.c: .writepage = ecryptfs_writepage, fs/f2fs/checkpoint.c: .writepage = f2fs_write_meta_page, fs/f2fs/data.c: .writepage = f2fs_write_data_page, fs/f2fs/node.c: .writepage = f2fs_write_node_page, fs/gfs2/aops.c: .writepage = gfs2_jdata_writepage, fs/gfs2/meta_io.c: .writepage = gfs2_aspace_writepage, fs/gfs2/meta_io.c: .writepage = gfs2_aspace_writepage, fs/hostfs/hostfs_kern.c: .writepage = hostfs_writepage, fs/nilfs2/inode.c: .writepage = nilfs_writepage, fs/nilfs2/mdt.c: .writepage = nilfs_mdt_write_page, fs/orangefs/inode.c: .writepage = orangefs_writepage, fs/vboxsf/file.c: .writepage = vboxsf_writepage, mm/shmem.c: .writepage = shmem_writepage, mm/swap_state.c: .writepage = swap_writepage, so mostly just the usual stragglers. I sent a series to fix up gfs2 earlier this week: https://lore.kernel.org/linux-fsdevel/20240719175105.788253-1-willy@xxxxxxxxxxxxx/
