Add code to help debug the rmmod hang.

diff --git a/lustre/obdfs/flushd.c b/lustre/obdfs/flushd.c
index 2458720..88245b2 100644 (file)
--- a/lustre/obdfs/flushd.c
+++ b/lustre/obdfs/flushd.c
@@ -127,10 +127,13 @@ void obdfs_dequeue_pages(struct inode *inode)
         struct list_head *tmp;
 
         ENTRY;
+       printk ("down0\n");
         obd_down(&obdfs_i2sbi(inode)->osi_list_mutex);
+       printk ("got down0\n");
         tmp = obdfs_islist(inode);
         if ( list_empty(tmp) ) {
                 CDEBUG(D_INFO, "no dirty pages for inode %ld\n", inode->i_ino);
+               printk ("up 0a");
                 obd_up(&obdfs_i2sbi(inode)->osi_list_mutex);
                 EXIT;
                 return;
@@ -153,6 +156,7 @@ void obdfs_dequeue_pages(struct inode *inode)
                 put_page(page);
         }
 
+       printk ("up 0b");
         obd_up(&obdfs_i2sbi(inode)->osi_list_mutex);
 
         /* decrement inode reference for page cache */
@@ -187,9 +191,12 @@ int obdfs_flush_reqs(struct list_head *inode_list, unsigned long check_time)
 
         sbi = list_entry(inode_list, struct obdfs_sb_info, osi_inodes);
 
+       printk ("down1\n");
         obd_down(&sbi->osi_list_mutex);
+       printk ("got down1\n");
         if ( list_empty(inode_list) ) {
                 CDEBUG(D_INFO, "list empty\n");
+               printk ("up 1a\n");
                 obd_up(&sbi->osi_list_mutex);
                 EXIT;
                 return 0;
@@ -254,6 +261,7 @@ int obdfs_flush_reqs(struct list_head *inode_list, unsigned long check_time)
                         num_obdos++;
 
                         if ( num_io == MAX_IOVEC ) {
+                               printk ("up 1b\n");
                                 obd_up(&sbi->osi_list_mutex);
                                 err = obdfs_do_vec_wr(inodes, num_io, num_obdos,
                                                       obdos, bufs_per_obdo,
@@ -266,13 +274,16 @@ int obdfs_flush_reqs(struct list_head *inode_list, unsigned long check_time)
                                         EXIT;
                                         goto ERR;
                                 }
+                               printk ("down2\n");
                                 obd_down(&sbi->osi_list_mutex);
+                               printk ("got down2\n");
                                 goto again;
                         }
                 }
         }
 
 BREAK:
+       printk ("up 2\n");
         obd_up(&sbi->osi_list_mutex);
 
         /* flush any remaining I/Os */
@@ -290,7 +301,9 @@ BREAK:
          * Make sure we don't point at the current inode with tmp
          * when we re-init the list on the inode, or we will loop.
          */
+       printk ("down3\n");
         obd_down(&sbi->osi_list_mutex);
+       printk ("got down3\n");
         tmp = inode_list;
         while ( (tmp = tmp->prev) != inode_list ) {
                 struct obdfs_inode_info *ii;
@@ -309,6 +322,7 @@ BREAK:
                         INIT_LIST_HEAD(obdfs_islist(inode));
                 }
         }
+       printk ("up 3\n");
         obd_up(&sbi->osi_list_mutex);
 
         CDEBUG(D_INFO, "flushed %ld pages in total\n", total_io);

diff --git a/lustre/obdfs/super.c b/lustre/obdfs/super.c
index 9af4896..5f97dcc 100644 (file)
--- a/lustre/obdfs/super.c
+++ b/lustre/obdfs/super.c
@@ -192,7 +192,7 @@ static struct super_block * obdfs_read_super(struct super_block *sb,
         CDEBUG(D_INFO, "\n"); 
         /* list of dirty inodes, and a mutex to hold while modifying it */
         INIT_LIST_HEAD(&sbi->osi_inodes);
-        sema_init(&sbi->osi_list_mutex, 1);
+        init_MUTEX (&sbi->osi_list_mutex);
 
         CDEBUG(D_INFO, "\n"); 
         sbi->osi_super = sb;
@@ -286,7 +286,6 @@ static void obdfs_put_super(struct super_block *sb)
 
         OPS(sb,disconnect)(ID(sb));
         list_del(&sbi->osi_list);
-        memset(sbi, 0, sizeof(*sbi));
         
         printk(KERN_INFO "OBDFS: Bye bye.\n");