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  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]
Date:   Sun, 10 May 2020 12:38:23 +0530
From:   Ritesh Harjani <riteshh@...ux.ibm.com>
To:     linux-ext4@...r.kernel.org
Cc:     linux-fsdevel@...r.kernel.org, Jan Kara <jack@...e.com>,
        tytso@....edu, "Aneesh Kumar K . V" <aneesh.kumar@...ux.ibm.com>,
        linux-kernel@...r.kernel.org,
        Ritesh Harjani <riteshh@...ux.ibm.com>
Subject: [RFCv4 3/6] ext4: mballoc: Optimize ext4_mb_good_group_nolock further if grp needs init

Currently in case if EXT4_MB_GRP_NEED_INIT(grp) is true, then we first
check for few things like grp->bb_free etc with spinlock (ext4_lock_group)
held. Then we drop the lock only to initialize the group's buddy cache
and then again take the lock and check for ext4_mb_good_group().

Once this step is done we return to ext4_mb_regular_allocator(), load
the buddy and then again take the lock only to check
ext4_mb_good_group(), which was anyways done in previous step.

I believe we can optimize one step here where if the group needs init
we can check for only few things and return early. Then recheck for
ext4_mb_good_group() only once after loading the buddy cache.

Signed-off-by: Ritesh Harjani <riteshh@...ux.ibm.com>
---
 fs/ext4/mballoc.c | 44 +++++++++++++++++++++-----------------------
 1 file changed, 21 insertions(+), 23 deletions(-)

diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index dcd05ff7c012..7d766dc34cdd 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -2134,33 +2134,34 @@ static bool ext4_mb_good_group(struct ext4_allocation_context *ac,
  * during ext4_mb_init_group(). This should not be called with
  * ext4_lock_group() held.
  */
-static int ext4_mb_good_group_nolock(struct ext4_allocation_context *ac,
+static bool ext4_mb_good_group_nolock(struct ext4_allocation_context *ac,
 				     ext4_group_t group, int cr)
 {
 	struct ext4_group_info *grp = ext4_get_group_info(ac->ac_sb, group);
 	struct super_block *sb = ac->ac_sb;
 	ext4_grpblk_t free;
-	int ret = 0;
+	bool ret = false, need_init = EXT4_MB_GRP_NEED_INIT(grp);
 
 	ext4_lock_group(sb, group);
-	free = grp->bb_free;
-	if (free == 0)
-		goto out;
-	if (cr <= 2 && free < ac->ac_g_ex.fe_len)
-		goto out;
-	if (unlikely(EXT4_MB_GRP_BBITMAP_CORRUPT(grp)))
-		goto out;
-	ext4_unlock_group(sb, group);
-
-	/* We only do this if the grp has never been initialized */
-	if (unlikely(EXT4_MB_GRP_NEED_INIT(grp))) {
-		ret = ext4_mb_init_group(ac->ac_sb, group, GFP_NOFS);
-		if (ret)
-			return ret;
+	/*
+	 * If the group needs init then no need to call ext4_mb_init_group()
+	 * after dropping the lock. It's better we check bb_free/other things
+	 * here and if it meets the criteria than return true. Later we
+	 * will anyway check for good group after loading the buddy cache
+	 * which, if required will call ext4_mb_init_group() from within.
+	 */
+	if (unlikely(need_init)) {
+		free = grp->bb_free;
+		if (free == 0)
+			goto out;
+		if (cr <= 2 && free < ac->ac_g_ex.fe_len)
+			goto out;
+		if (unlikely(EXT4_MB_GRP_BBITMAP_CORRUPT(grp)))
+			goto out;
+		ret = true;
+	} else {
+		ret = ext4_mb_good_group(ac, group, cr);
 	}
-
-	ext4_lock_group(sb, group);
-	ret = ext4_mb_good_group(ac, group, cr);
 out:
 	ext4_unlock_group(sb, group);
 	return ret;
@@ -2252,11 +2253,8 @@ ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
 
 			/* This now checks without needing the buddy page */
 			ret = ext4_mb_good_group_nolock(ac, group, cr);
-			if (ret <= 0) {
-				if (!first_err)
-					first_err = ret;
+			if (ret == 0)
 				continue;
-			}
 
 			err = ext4_mb_load_buddy(sb, group, &e4b);
 			if (err)
-- 
2.21.0

Powered by blists - more mailing lists