NFSv4.1: lseg refcounting

Prepare put_lseg and get_lseg to be called from the pNFS I/O code.
Pull common code from pnfs_lseg_locked to call from pnfs_lseg.
Inline pnfs_lseg_locked into it's only caller.

Signed-off-by: Fred Isaman <iisaman@netapp.com>
Signed-off-by: Benny Halevy <bhalevy@panasas.com>
Signed-off-by: Trond Myklebust <Trond.Myklebust@netapp.com>

2 files changed, 59 insertions, 23 deletions

diff --git a/fs/nfs/pnfs.c b/fs/nfs/pnfs.c
index 7d031cd7d920..3afa82e45438 100644
--- a/fs/nfs/pnfs.c
+++ b/fs/nfs/pnfs.c
@@ -230,32 +230,41 @@ static void free_lseg(struct pnfs_layout_segment *lseg)
         put_layout_hdr(NFS_I(ino)->layout);
 }
 
-/* The use of tmp_list is necessary because pnfs_curr_ld->free_lseg
- * could sleep, so must be called outside of the lock.
- * Returns 1 if object was removed, otherwise return 0.
- */
-static int
-put_lseg_locked(struct pnfs_layout_segment *lseg,
-                struct list_head *tmp_list)
+static void
+put_lseg_common(struct pnfs_layout_segment *lseg)
+{
+        struct inode *inode = lseg->pls_layout->plh_inode;
+
+        BUG_ON(test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
+        list_del_init(&lseg->pls_list);
+        if (list_empty(&lseg->pls_layout->plh_segs)) {
+                set_bit(NFS_LAYOUT_DESTROYED, &lseg->pls_layout->plh_flags);
+                /* Matched by initial refcount set in alloc_init_layout_hdr */
+                put_layout_hdr_locked(lseg->pls_layout);
+        }
+        rpc_wake_up(&NFS_SERVER(inode)->roc_rpcwaitq);
+}
+
+static void
+put_lseg(struct pnfs_layout_segment *lseg)
 {
+        struct inode *inode;
+
+        if (!lseg)
+                return;
+
         dprintk("%s: lseg %p ref %d valid %d\n", __func__, lseg,
                 atomic_read(&lseg->pls_refcount),
                 test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
-        if (atomic_dec_and_test(&lseg->pls_refcount)) {
-                struct inode *ino = lseg->pls_layout->plh_inode;
+        inode = lseg->pls_layout->plh_inode;
+        if (atomic_dec_and_lock(&lseg->pls_refcount, &inode->i_lock)) {
+                LIST_HEAD(free_me);
 
-                BUG_ON(test_bit(NFS_LSEG_VALID, &lseg->pls_flags));
-                list_del(&lseg->pls_list);
-                if (list_empty(&lseg->pls_layout->plh_segs)) {
-                        set_bit(NFS_LAYOUT_DESTROYED, &lseg->pls_layout->plh_flags);
-                        /* Matched by initial refcount set in alloc_init_layout_hdr */
-                        put_layout_hdr_locked(lseg->pls_layout);
-                }
-                rpc_wake_up(&NFS_SERVER(ino)->roc_rpcwaitq);
-                list_add(&lseg->pls_list, tmp_list);
-                return 1;
+                put_lseg_common(lseg);
+                list_add(&lseg->pls_list, &free_me);
+                spin_unlock(&inode->i_lock);
+                pnfs_free_lseg_list(&free_me);
         }
-        return 0;
 }
 
 static bool
@@ -276,7 +285,13 @@ static int mark_lseg_invalid(struct pnfs_layout_segment *lseg,
                  * list.  It will now be removed when all
                  * outstanding io is finished.
                  */
-                rv = put_lseg_locked(lseg, tmp_list);
+                dprintk("%s: lseg %p ref %d\n", __func__, lseg,
+                        atomic_read(&lseg->pls_refcount));
+                if (atomic_dec_and_test(&lseg->pls_refcount)) {
+                        put_lseg_common(lseg);
+                        list_add(&lseg->pls_list, tmp_list);
+                        rv = 1;
+                }
         }
         return rv;
 }
@@ -689,7 +704,7 @@ pnfs_find_lseg(struct pnfs_layout_hdr *lo, u32 iomode)
         list_for_each_entry(lseg, &lo->plh_segs, pls_list) {
                 if (test_bit(NFS_LSEG_VALID, &lseg->pls_flags) &&
                     is_matching_lseg(lseg, iomode)) {
-                        ret = lseg;
+                        ret = get_lseg(lseg);
                         break;
                 }
                 if (cmp_layout(iomode, lseg->pls_range.iomode) > 0)
@@ -769,6 +784,7 @@ pnfs_update_layout(struct inode *ino,
 out:
         dprintk("%s end, state 0x%lx lseg %p\n", __func__,
                 nfsi->layout ? nfsi->layout->plh_flags : -1, lseg);
+        put_lseg(lseg); /* STUB - callers currently ignore return value */
         return lseg;
 out_unlock:
         spin_unlock(&ino->i_lock);
@@ -821,7 +837,7 @@ pnfs_layout_process(struct nfs4_layoutget *lgp)
         }
         init_lseg(lo, lseg);
         lseg->pls_range = res->range;
-        *lgp->lsegpp = lseg;
+        *lgp->lsegpp = get_lseg(lseg);
         pnfs_insert_layout(lo, lseg);
 
         if (res->return_on_close) {
diff --git a/fs/nfs/pnfs.h b/fs/nfs/pnfs.h
index e2612ea0cbed..9a994bc9899f 100644
--- a/fs/nfs/pnfs.h
+++ b/fs/nfs/pnfs.h
@@ -177,6 +177,16 @@ static inline int lo_fail_bit(u32 iomode)
                          NFS_LAYOUT_RW_FAILED : NFS_LAYOUT_RO_FAILED;
 }
 
+static inline struct pnfs_layout_segment *
+get_lseg(struct pnfs_layout_segment *lseg)
+{
+        if (lseg) {
+                atomic_inc(&lseg->pls_refcount);
+                smp_mb__after_atomic_inc();
+        }
+        return lseg;
+}
+
 /* Return true if a layout driver is being used for this mountpoint */
 static inline int pnfs_enabled_sb(struct nfs_server *nfss)
 {
@@ -194,6 +204,16 @@ static inline void pnfs_destroy_layout(struct nfs_inode *nfsi)
 }
 
 static inline struct pnfs_layout_segment *
+get_lseg(struct pnfs_layout_segment *lseg)
+{
+        return NULL;
+}
+
+static inline void put_lseg(struct pnfs_layout_segment *lseg)
+{
+}
+
+static inline struct pnfs_layout_segment *
 pnfs_update_layout(struct inode *ino, struct nfs_open_context *ctx,
                    enum pnfs_iomode access_type)
 {