bcachefs: Fix __bch2_truncate_page()
authorKent Overstreet <kent.overstreet@gmail.com>
Fri, 9 Oct 2020 04:09:20 +0000 (00:09 -0400)
committerKent Overstreet <kent.overstreet@linux.dev>
Sun, 22 Oct 2023 21:08:44 +0000 (17:08 -0400)
__bch2_truncate_page() will mark some of the blocks in a page as
unallocated. But, if the page is mmapped (and writable), every block in
the page needs to be marked dirty, else those blocks won't be written by
__bch2_writepage().

The solution is to change those userspace mappings to RO, so that we
force bch2_page_mkwrite() to be called again.

Signed-off-by: Kent Overstreet <kent.overstreet@gmail.com>
Signed-off-by: Kent Overstreet <kent.overstreet@linux.dev>
fs/bcachefs/fs-io.c

index c099572..0290f74 100644 (file)
@@ -27,6 +27,7 @@
 #include <linux/migrate.h>
 #include <linux/mmu_context.h>
 #include <linux/pagevec.h>
+#include <linux/rmap.h>
 #include <linux/sched/signal.h>
 #include <linux/task_io_accounting_ops.h>
 #include <linux/uio.h>
@@ -2160,6 +2161,12 @@ static int __bch2_truncate_page(struct bch_inode_info *inode,
        ret = bch2_get_page_disk_reservation(c, inode, page, false);
        BUG_ON(ret);
 
+       /*
+        * This removes any writeable userspace mappings; we need to force
+        * .page_mkwrite to be called again before any mmapped writes, to
+        * redirty the full page:
+        */
+       page_mkclean(page);
        filemap_dirty_folio(mapping, page_folio(page));
 unlock:
        unlock_page(page);