mm: flush dcache before writing into page to avoid alias
authoranfei zhou <anfei.zhou@gmail.com>
Tue, 2 Feb 2010 21:44:02 +0000 (13:44 -0800)
committerLinus Torvalds <torvalds@linux-foundation.org>
Wed, 3 Feb 2010 02:11:21 +0000 (18:11 -0800)
The cache alias problem will happen if the changes of user shared mapping
is not flushed before copying, then user and kernel mapping may be mapped
into two different cache line, it is impossible to guarantee the coherence
after iov_iter_copy_from_user_atomic.  So the right steps should be:

flush_dcache_page(page);
kmap_atomic(page);
write to page;
kunmap_atomic(page);
flush_dcache_page(page);

More precisely, we might create two new APIs flush_dcache_user_page and
flush_dcache_kern_page to replace the two flush_dcache_page accordingly.

Here is a snippet tested on omap2430 with VIPT cache, and I think it is
not ARM-specific:

int val = 0x11111111;
fd = open("abc", O_RDWR);
addr = mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_SHARED, fd, 0);
*(addr+0) = 0x44444444;
tmp = *(addr+0);
*(addr+1) = 0x77777777;
write(fd, &val, sizeof(int));
close(fd);

The results are not always 0x11111111 0x77777777 at the beginning as expected.  Sometimes we see 0x44444444 0x77777777.

Signed-off-by: Anfei <anfei.zhou@gmail.com>
Cc: Russell King <rmk@arm.linux.org.uk>
Cc: Miklos Szeredi <miklos@szeredi.hu>
Cc: Nick Piggin <nickpiggin@yahoo.com.au>
Cc: <linux-arch@vger.kernel.org>
Cc: <stable@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
fs/fuse/file.c
mm/filemap.c

index c18913a777ae3865148c9bbd679d0c09ff092f4f..a9f5e137f1d31547f86bbfdb87df11c07daf963f 100644 (file)
@@ -828,6 +828,9 @@ static ssize_t fuse_fill_write_pages(struct fuse_req *req,
                if (!page)
                        break;
 
+               if (mapping_writably_mapped(mapping))
+                       flush_dcache_page(page);
+
                pagefault_disable();
                tmp = iov_iter_copy_from_user_atomic(page, ii, offset, bytes);
                pagefault_enable();
index e3736923220e86a6274828e52286c69691ee67be..698ea80f21022bf709dacc1101c614327109d828 100644 (file)
@@ -2232,6 +2232,9 @@ again:
                if (unlikely(status))
                        break;
 
+               if (mapping_writably_mapped(mapping))
+                       flush_dcache_page(page);
+
                pagefault_disable();
                copied = iov_iter_copy_from_user_atomic(page, i, offset, bytes);
                pagefault_enable();