[rh7,v2] pfcache: do not drop pfcache checksum during mmap(SHARED) of ro-files

Submitted by Konstantin Khorenko on Feb. 28, 2020, 9:48 a.m.


Message ID 20200228094824.29474-1-khorenko@virtuozzo.com
State New
Series "pfcache: do not drop pfcache checksum during mmap(SHARED) of ro-files"
Headers show

Commit Message

Konstantin Khorenko Feb. 28, 2020, 9:48 a.m.
From: Vasily Averin <vvs@virtuozzo.com>

Currently any mmap(SHARED) removes pfcache mark, even for
read_only-opened files. It is required to guarantee that shared mappings
will see changes if mapped file will be write-opened and modified.

On the other hand we do not want to drop pfcache mark without the
necessity. At the moment ldconfig calls mmap(SHARED) before adding
any new files into library cache. It removes pfcache mark and
does not allow to add libraries into pfcache.

This patch helps to avoid the problem: during mmap(SHARED) for read-only
files kernel does not drop pfcache mark from inode but closes file in
pfcache area and use its local copy.


Signed-off-by: Vasily Averin <vvs@virtuozzo.com>
Acked-by: Konstantin Khorenko <khorenko@virtuozzo.com>

* if we do mmap(SHARED) in ro mode, we still can use pfcache file
  mapping safely because any open in rw mode will close peer (pfcache)
  file anyway, so don't close pfcache on ro mmap(SHARED) move the check

* for open mode a step above, so don't need to call prep
  ext4_file_mmap(vma==NULL) if open in ro mode.
 mm/util.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

Patch hide | download patch | download mbox

diff --git a/mm/util.c b/mm/util.c
index 69dd74900ae6d..6b1cf7620e846 100644
--- a/mm/util.c
+++ b/mm/util.c
@@ -331,7 +331,8 @@  unsigned long vm_mmap_pgoff(struct file *file, unsigned long addr,
 		/* Ugly fix for PSBM-23133 vdavydov@ */
 		if (file && file->f_op && (flag & MAP_TYPE) == MAP_SHARED &&
 		    S_ISREG(file_inode(file)->i_mode) &&
-		    (file_inode(file)->i_sb->s_type->fs_flags & FS_HAS_MMAP_PREP))
+		    (file_inode(file)->i_sb->s_type->fs_flags & FS_HAS_MMAP_PREP) &&
+		    (file->f_mode & FMODE_WRITE))
 			file->f_op->mmap(file, NULL);
 		ret = do_mmap_pgoff(file, addr, len, prot, flag, pgoff,