ufs: remove the BKL

This introduces a new per-superblock mutex in UFS to replace
the big kernel lock. I have been careful to avoid nested
calls to lock_ufs and to get the lock order right with
respect to other mutexes, in particular lock_super.

I did not make any attempt to prove that the big kernel
lock is not needed in a particular place in the code,
which is very possible.

The mutex has a significant performance impact, so it is only
used on SMP or PREEMPT configurations.

As Nick Piggin noticed, any allocation inside of the lock
may end up deadlocking when we get to ufs_getfrag_block
in the reclaim task, so we now use GFP_NOFS.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Tested-by: Nick Bowler <nbowler@elliptictech.com>
Cc: Evgeniy Dushistov <dushistov@mail.ru>
Cc: Nick Piggin <npiggin@gmail.com>

1 files changed, 21 insertions, 57 deletions

diff --git a/fs/ufs/inode.c b/fs/ufs/inode.c
index 2b251f2093af..03c255f12df5 100644
--- a/fs/ufs/inode.c
+++ b/fs/ufs/inode.c
@@ -34,7 +34,6 @@
 #include <linux/stat.h>
 #include <linux/string.h>
 #include <linux/mm.h>
-#include <linux/smp_lock.h>
 #include <linux/buffer_head.h>
 #include <linux/writeback.h>
 
@@ -43,7 +42,7 @@
 #include "swab.h"
 #include "util.h"
 
-static u64 ufs_frag_map(struct inode *inode, sector_t frag);
+static u64 ufs_frag_map(struct inode *inode, sector_t frag, bool needs_lock);
 
 static int ufs_block_to_path(struct inode *inode, sector_t i_block, sector_t offsets[4])
 {
@@ -82,7 +81,7 @@ static int ufs_block_to_path(struct inode *inode, sector_t i_block, sector_t off
  * the begining of the filesystem.
  */
 
-static u64 ufs_frag_map(struct inode *inode, sector_t frag)
+static u64 ufs_frag_map(struct inode *inode, sector_t frag, bool needs_lock)
 {
 	struct ufs_inode_info *ufsi = UFS_I(inode);
 	struct super_block *sb = inode->i_sb;
@@ -107,7 +106,8 @@ static u64 ufs_frag_map(struct inode *inode, sector_t frag)
 
 	p = offsets;
 
-	lock_kernel();
+	if (needs_lock)
+		lock_ufs(sb);
 	if ((flags & UFS_TYPE_MASK) == UFS_TYPE_UFS2)
 		goto ufs2;
 
@@ -152,7 +152,8 @@ ufs2:
 	ret = temp + (u64) (frag & uspi->s_fpbmask);
 
 out:
-	unlock_kernel();
+	if (needs_lock)
+		unlock_ufs(sb);
 	return ret;
 }
 
@@ -415,14 +416,16 @@ out:
 int ufs_getfrag_block(struct inode *inode, sector_t fragment, struct buffer_head *bh_result, int create)
 {
 	struct super_block * sb = inode->i_sb;
-	struct ufs_sb_private_info * uspi = UFS_SB(sb)->s_uspi;
+	struct ufs_sb_info * sbi = UFS_SB(sb);
+	struct ufs_sb_private_info * uspi = sbi->s_uspi;
 	struct buffer_head * bh;
 	int ret, err, new;
 	unsigned long ptr,phys;
 	u64 phys64 = 0;
+	bool needs_lock = (sbi->mutex_owner != current);
 	
 	if (!create) {
-		phys64 = ufs_frag_map(inode, fragment);
+		phys64 = ufs_frag_map(inode, fragment, needs_lock);
 		UFSD("phys64 = %llu\n", (unsigned long long)phys64);
 		if (phys64)
 			map_bh(bh_result, sb, phys64);
@@ -436,7 +439,8 @@ int ufs_getfrag_block(struct inode *inode, sector_t fragment, struct buffer_head
 	ret = 0;
 	bh = NULL;
 
-	lock_kernel();
+	if (needs_lock)
+		lock_ufs(sb);
 
 	UFSD("ENTER, ino %lu, fragment %llu\n", inode->i_ino, (unsigned long long)fragment);
 	if (fragment >
@@ -498,7 +502,9 @@ out:
 		set_buffer_new(bh_result);
 	map_bh(bh_result, sb, phys);
 abort:
-	unlock_kernel();
+	if (needs_lock)
+		unlock_ufs(sb);
+
 	return err;
 
 abort_too_big:
@@ -506,48 +512,6 @@ abort_too_big:
 	goto abort;
 }
 
-static struct buffer_head *ufs_getfrag(struct inode *inode,
-				       unsigned int fragment,
-				       int create, int *err)
-{
-	struct buffer_head dummy;
-	int error;
-
-	dummy.b_state = 0;
-	dummy.b_blocknr = -1000;
-	error = ufs_getfrag_block(inode, fragment, &dummy, create);
-	*err = error;
-	if (!error && buffer_mapped(&dummy)) {
-		struct buffer_head *bh;
-		bh = sb_getblk(inode->i_sb, dummy.b_blocknr);
-		if (buffer_new(&dummy)) {
-			memset(bh->b_data, 0, inode->i_sb->s_blocksize);
-			set_buffer_uptodate(bh);
-			mark_buffer_dirty(bh);
-		}
-		return bh;
-	}
-	return NULL;
-}
-
-struct buffer_head * ufs_bread (struct inode * inode, unsigned fragment,
-	int create, int * err)
-{
-	struct buffer_head * bh;
-
-	UFSD("ENTER, ino %lu, fragment %u\n", inode->i_ino, fragment);
-	bh = ufs_getfrag (inode, fragment, create, err);
-	if (!bh || buffer_uptodate(bh)) 		
-		return bh;
-	ll_rw_block (READ, 1, &bh);
-	wait_on_buffer (bh);
-	if (buffer_uptodate(bh))
-		return bh;
-	brelse (bh);
-	*err = -EIO;
-	return NULL;
-}
-
 static int ufs_writepage(struct page *page, struct writeback_control *wbc)
 {
 	return block_write_full_page(page,ufs_getfrag_block,wbc);
@@ -900,9 +864,9 @@ static int ufs_update_inode(struct inode * inode, int do_sync)
 int ufs_write_inode(struct inode *inode, struct writeback_control *wbc)
 {
 	int ret;
-	lock_kernel();
+	lock_ufs(inode->i_sb);
 	ret = ufs_update_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
-	unlock_kernel();
+	unlock_ufs(inode->i_sb);
 	return ret;
 }
 
@@ -922,22 +886,22 @@ void ufs_evict_inode(struct inode * inode)
 	if (want_delete) {
 		loff_t old_i_size;
 		/*UFS_I(inode)->i_dtime = CURRENT_TIME;*/
-		lock_kernel();
+		lock_ufs(inode->i_sb);
 		mark_inode_dirty(inode);
 		ufs_update_inode(inode, IS_SYNC(inode));
 		old_i_size = inode->i_size;
 		inode->i_size = 0;
 		if (inode->i_blocks && ufs_truncate(inode, old_i_size))
 			ufs_warning(inode->i_sb, __func__, "ufs_truncate failed\n");
-		unlock_kernel();
+		unlock_ufs(inode->i_sb);
 	}
 
 	invalidate_inode_buffers(inode);
 	end_writeback(inode);
 
 	if (want_delete) {
-		lock_kernel();
+		lock_ufs(inode->i_sb);
 		ufs_free_inode (inode);
-		unlock_kernel();
+		unlock_ufs(inode->i_sb);
 	}
 }