[PATCH] inode dirty blocks Re: test12-pre4

From: Alexander Viro (viro@math.psu.edu)
Date: Mon Dec 04 2000 - 01:01:36 EST


On Sun, 3 Dec 2000, Linus Torvalds wrote:

>
> Synching up with Alan and various other stuff. The most important one
> being the fix to the inode dirty block list.

It doesn't solve the problem. If you unlink a file with dirty metadata
you have a nice chance to hit the BUG() in inode.c:83. I hope that patch
below closes all remaining holes. See analysis in previous posting
(basically, bforget() is not enough when we free the block; bh should
be removed from the inode's list regardless of the ->b_count).
                                                        Cheers,
                                                                Al

diff -urN rc12-pre4/fs/buffer.c rc12-pre4-dirty_blocks/fs/buffer.c
--- rc12-pre4/fs/buffer.c Mon Dec 4 01:01:43 2000
+++ rc12-pre4-dirty_blocks/fs/buffer.c Mon Dec 4 01:11:42 2000
@@ -1164,6 +1164,31 @@
 }
 
 /*
+ * Call it when you are going to free the block. The difference between
+ * that and bforget() is that we remove the thing from inode queue
+ * unconditionally.
+ */
+void bforget_inode(struct buffer_head * buf)
+{
+ /* grab the lru lock here to block bdflush. */
+ spin_lock(&lru_list_lock);
+ write_lock(&hash_table_lock);
+ remove_inode_queue(buf);
+ if (!atomic_dec_and_test(&buf->b_count) || buffer_locked(buf))
+ goto in_use;
+ __hash_unlink(buf);
+ write_unlock(&hash_table_lock);
+ __remove_from_lru_list(buf, buf->b_list);
+ spin_unlock(&lru_list_lock);
+ put_last_free(buf);
+ return;
+
+ in_use:
+ write_unlock(&hash_table_lock);
+ spin_unlock(&lru_list_lock);
+}
+
+/*
  * bread() reads a specified block and returns the buffer that contains
  * it. It returns NULL if the block was unreadable.
  */
@@ -1460,6 +1485,9 @@
                 clear_bit(BH_Mapped, &bh->b_state);
                 clear_bit(BH_Req, &bh->b_state);
                 clear_bit(BH_New, &bh->b_state);
+ spin_lock(&lru_list_lock);
+ remove_inode_queue(bh);
+ spin_unlock(&lru_list_lock);
         }
 }
 
diff -urN rc12-pre4/fs/ext2/inode.c rc12-pre4-dirty_blocks/fs/ext2/inode.c
--- rc12-pre4/fs/ext2/inode.c Mon Dec 4 01:01:43 2000
+++ rc12-pre4-dirty_blocks/fs/ext2/inode.c Mon Dec 4 01:13:10 2000
@@ -416,7 +416,7 @@
 
         /* Allocation failed, free what we already allocated */
         for (i = 1; i < n; i++)
- bforget(branch[i].bh);
+ bforget_inode(branch[i].bh);
         for (i = 0; i < n; i++)
                 ext2_free_blocks(inode, le32_to_cpu(branch[i].key), 1);
         return err;
@@ -484,7 +484,7 @@
 
 changed:
         for (i = 1; i < num; i++)
- bforget(where[i].bh);
+ bforget_inode(where[i].bh);
         for (i = 0; i < num; i++)
                 ext2_free_blocks(inode, le32_to_cpu(where[i].key), 1);
         return -EAGAIN;
@@ -854,7 +854,7 @@
                                            (u32*)bh->b_data,
                                            (u32*)bh->b_data + addr_per_block,
                                            depth);
- bforget(bh);
+ bforget_inode(bh);
                         /* Writer: ->i_blocks */
                         inode->i_blocks -= inode->i_sb->s_blocksize / 512;
                         /* Writer: end */
diff -urN rc12-pre4/include/linux/fs.h rc12-pre4-dirty_blocks/include/linux/fs.h
--- rc12-pre4/include/linux/fs.h Mon Dec 4 01:01:47 2000
+++ rc12-pre4-dirty_blocks/include/linux/fs.h Mon Dec 4 01:12:03 2000
@@ -1201,6 +1201,7 @@
                 __brelse(buf);
 }
 extern void __bforget(struct buffer_head *);
+extern void bforget_inode(struct buffer_head *);
 static inline void bforget(struct buffer_head *buf)
 {
         if (buf)
diff -urN rc12-pre4/kernel/ksyms.c rc12-pre4-dirty_blocks/kernel/ksyms.c
--- rc12-pre4/kernel/ksyms.c Mon Dec 4 01:01:49 2000
+++ rc12-pre4-dirty_blocks/kernel/ksyms.c Mon Dec 4 01:12:19 2000
@@ -188,6 +188,7 @@
 EXPORT_SYMBOL(breada);
 EXPORT_SYMBOL(__brelse);
 EXPORT_SYMBOL(__bforget);
+EXPORT_SYMBOL(bforget_inode);
 EXPORT_SYMBOL(ll_rw_block);
 EXPORT_SYMBOL(__wait_on_buffer);
 EXPORT_SYMBOL(___wait_on_page);

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
Please read the FAQ at http://www.tux.org/lkml/



This archive was generated by hypermail 2b29 : Thu Dec 07 2000 - 21:00:10 EST