shmfs: don't allocate pages on read

Commit Message

Mikulas Patocka July 21, 2020, 9:12 a.m. UTC

Hi

Do we still need the patch a0ee5ec520ede?

We are using the loop device for testing and when we read something from 
the loop device, it allocates the pages on the underlying shmfs 
filesystem. See this example:

# mkdir -p /mnt/test
# mount -t tmpfs /dev/null /mnt/test
# cd /mnt/test
# truncate -s 1GiB file
# du -hs file
0       file
# losetup /dev/loop0 file
# du -hs file
1,1M    file
# dd if=/dev/loop0 of=/dev/null
2097152+0 záznamů přečteno
2097152+0 zĂĄznamĹŻ zapsĂĄno
1073741824 bajtĹŻ (1,1 GB, 1,0 GiB) zkopĂ­rovĂĄno, 4,06865 s, 264 MB/s
# du -hs file
1,0G    file

This patch turns off the allocation on read.

Signed-off-by: Mikulas Patocka <[e-mail]>

---
 mm/shmem.c |    8 --------
 1 file changed, 8 deletions(-)

Patch

Index: linux-2.6/mm/shmem.c
===================================================================
--- linux-2.6.orig/mm/shmem.c	2020-06-29 14:50:06.000000000 +0200
+++ linux-2.6/mm/shmem.c	2020-07-16 19:22:58.000000000 +0200
@@ -2507,14 +2507,6 @@  static ssize_t shmem_file_read_iter(stru
 	ssize_t retval = 0;
 	loff_t *ppos = &iocb->ki_pos;
 
-	/*
-	 * Might this read be for a stacking filesystem?  Then when reading
-	 * holes of a sparse file, we actually need to allocate those pages,
-	 * and even mark them dirty, so it cannot exceed the max_blocks limit.
-	 */
-	if (!iter_is_iovec(to))
-		sgp = SGP_CACHE;
-
 	index = *ppos >> PAGE_SHIFT;
 	offset = *ppos & ~PAGE_MASK;