linux-ext4 - [PATCH 1/2] libext2fs: use ZERO_RANGE before PUNCH_HOLE in unix

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <152399274434.4654.17509483239699776997.stgit@magnolia>
Date:   Tue, 17 Apr 2018 12:19:04 -0700
From:   "Darrick J. Wong" <darrick.wong@...cle.com>
To:     tytso@....edu, darrick.wong@...cle.com
Cc:     adilger@...ger.ca, linux-ext4@...r.kernel.org
Subject: [PATCH 1/2] libext2fs: use ZERO_RANGE before PUNCH_HOLE in
 unix_zeroout

From: Darrick J. Wong <darrick.wong@...cle.com>

In unix_zeroout() for files, we should try a ZERO_RANGE before we try
PUNCH_HOLE because the former will not cause us to lose preallocated
blocks.  Since block devices have supported fallocate for a few years
now, refactor the fallocate calls into a helper and call it from either
case.

Reported-by: Andreas Dilger <adilger@...ger.ca>
Signed-off-by: Darrick J. Wong <darrick.wong@...cle.com>
---
 lib/ext2fs/unix_io.c |   53 ++++++++++++++++++++++++++++----------------------
 1 file changed, 30 insertions(+), 23 deletions(-)


diff --git a/lib/ext2fs/unix_io.c b/lib/ext2fs/unix_io.c
index c6b058d..1fb6279 100644
--- a/lib/ext2fs/unix_io.c
+++ b/lib/ext2fs/unix_io.c
@@ -1118,6 +1118,31 @@ static errcode_t unix_discard(io_channel channel, unsigned long long block,
 	return EXT2_ET_UNIMPLEMENTED;
 }
 
+/*
+ * If we know about ZERO_RANGE, try that before we try PUNCH_HOLE because
+ * ZERO_RANGE doesn't unmap preallocated blocks.  We prefer fallocate because
+ * it always invalidates page cache, and libext2fs requires that reads after
+ * ZERO_RANGE return zeroes.
+ */
+static int __unix_zeroout(int fd, off_t offset, off_t len)
+{
+	int ret = -1;
+
+#if defined(HAVE_FALLOCATE) && defined(FALLOC_FL_ZERO_RANGE)
+	ret = fallocate(fd, FALLOC_FL_ZERO_RANGE, offset, len);
+	if (ret == 0)
+		return 0;
+#endif
+#if defined(HAVE_FALLOCATE) && defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE)
+	ret = fallocate(fd, FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
+			offset,  len);
+	if (ret == 0)
+		return 0;
+#endif
+	errno = EOPNOTSUPP;
+	return ret;
+}
+
 /* parameters might not be used if OS doesn't support zeroout */
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wunused-parameter"
@@ -1134,10 +1159,7 @@ static errcode_t unix_zeroout(io_channel channel, unsigned long long block,
 	if (safe_getenv("UNIX_IO_NOZEROOUT"))
 		goto unimplemented;
 
-	if (channel->flags & CHANNEL_FLAGS_BLOCK_DEVICE) {
-		/* Not implemented until the BLKZEROOUT mess is fixed */
-		goto unimplemented;
-	} else {
+	if (!(channel->flags & CHANNEL_FLAGS_BLOCK_DEVICE)) {
 		/* Regular file, try to use truncate/punch/zero. */
 		struct stat statbuf;
 
@@ -1157,26 +1179,11 @@ static errcode_t unix_zeroout(io_channel channel, unsigned long long block,
 			if (ret)
 				goto err;
 		}
-#if defined(HAVE_FALLOCATE) && (defined(FALLOC_FL_ZERO_RANGE) || \
-	(defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE)))
-#if defined(FALLOC_FL_PUNCH_HOLE) && defined(FALLOC_FL_KEEP_SIZE)
-		ret = fallocate(data->dev,
-				FALLOC_FL_PUNCH_HOLE | FALLOC_FL_KEEP_SIZE,
-				(off_t)(block) * channel->block_size + data->offset,
-				(off_t)(count) * channel->block_size);
-		if (ret == 0)
-			goto err;
-#endif
-#ifdef FALLOC_FL_ZERO_RANGE
-		ret = fallocate(data->dev,
-				FALLOC_FL_ZERO_RANGE,
-				(off_t)(block) * channel->block_size + data->offset,
-				(off_t)(count) * channel->block_size);
-#endif
-#else
-		goto unimplemented;
-#endif /* HAVE_FALLOCATE && (ZERO_RANGE || (PUNCH_HOLE && KEEP_SIZE)) */
 	}
+
+	ret = __unix_zeroout(data->dev,
+			(off_t)(block) * channel->block_size + data->offset,
+			(off_t)(count) * channel->block_size);
 err:
 	if (ret < 0) {
 		if (errno == EOPNOTSUPP)