LU-14644 vvp: wait for nrpages to be updated

truncate_inode_pages() says there still may be a page in a process
of deletion upon return. wait for another thread which is doing
__delete_from_page_cache() to get nrpages updated.

Lustre-change: https://review.whamcloud.com/43464
Lustre-commit: 7d5d004506650c3739898e70d72c9a86b8aeeb88

Signed-off-by: Vitaly Fertman <c17818@cray.com>
Change-Id: I165b3d0866efaf2eb7e977520ebba4ee831874ab
HPE-bug-id: LUS-8842

diff --git a/lustre/include/lustre_compat.h b/lustre/include/lustre_compat.h
index c36f2bf..e2734c2 100644 (file)
--- a/lustre/include/lustre_compat.h
+++ b/lustre/include/lustre_compat.h
@@ -811,6 +811,8 @@ static inline bool bdev_integrity_enabled(struct block_device *bdev, int rw)
 #define page_tree i_pages
 #else
 #define i_pages tree_lock
+#define xa_lock_irqsave(lockp, flags) spin_lock_irqsave(lockp, flags)
+#define xa_unlock_irqrestore(lockp, flags) spin_unlock_irqrestore(lockp, flags)
 #define xa_lock_irq(lockp) spin_lock_irq(lockp)
 #define xa_unlock_irq(lockp) spin_unlock_irq(lockp)
 #endif

diff --git a/lustre/llite/llite_internal.h b/lustre/llite/llite_internal.h
index 4bfa305..4f156ee 100644 (file)
--- a/lustre/llite/llite_internal.h
+++ b/lustre/llite/llite_internal.h
@@ -957,6 +957,7 @@ int ll_statfs_internal(struct ll_sb_info *sbi, struct obd_statfs *osfs,
 int ll_update_inode(struct inode *inode, struct lustre_md *md);
 void ll_update_inode_flags(struct inode *inode, int ext_flags);
 int ll_read_inode2(struct inode *inode, void *opaque);
+void ll_truncate_inode_pages_final(struct inode *inode);
 void ll_delete_inode(struct inode *inode);
 int ll_iocontrol(struct inode *inode, struct file *file,
                  unsigned int cmd, unsigned long arg);

diff --git a/lustre/llite/llite_lib.c b/lustre/llite/llite_lib.c
index 80b3293..3b9809e 100644 (file)
--- a/lustre/llite/llite_lib.c
+++ b/lustre/llite/llite_lib.c
@@ -1427,7 +1427,7 @@ static int ll_update_lsm_md(struct inode *inode, struct lustre_md *md)
                lsm_md_dump(D_ERROR, lsm);
                GOTO(unlock, rc = -EINVAL);
        }
-  
+
        up_read(&lli->lli_lsm_sem);
        down_write(&lli->lli_lsm_sem);
        /* clear existing lsm */
@@ -1440,7 +1440,7 @@ static int ll_update_lsm_md(struct inode *inode, struct lustre_md *md)
        up_write(&lli->lli_lsm_sem);
        if (rc)
                RETURN(rc);
- 
+
        /* set md->lmv to NULL, so the following free lustre_md will not free
         * this lsm.
         */
@@ -1453,7 +1453,7 @@ static int ll_update_lsm_md(struct inode *inode, struct lustre_md *md)
        OBD_ALLOC_PTR(attr);
        if (!attr)
                GOTO(unlock, rc = -ENOMEM);
- 
+
        /* validate the lsm */
        rc = md_merge_attr(ll_i2mdexp(inode), lli->lli_lsm_md, attr,
                           ll_md_blocking_ast);
@@ -2053,6 +2053,34 @@ int ll_update_inode(struct inode *inode, struct lustre_md *md)
        return 0;
 }
 
+void ll_truncate_inode_pages_final(struct inode *inode)
+{
+       struct address_space *mapping = &inode->i_data;
+       unsigned long nrpages;
+       unsigned long flags;
+
+       truncate_inode_pages_final(mapping);
+
+       /* Workaround for LU-118: Note nrpages may not be totally updated when
+        * truncate_inode_pages() returns, as there can be a page in the process
+        * of deletion (inside __delete_from_page_cache()) in the specified
+        * range. Thus mapping->nrpages can be non-zero when this function
+        * returns even after truncation of the whole mapping.  Only do this if
+        * npages isn't already zero.
+        */
+       nrpages = mapping->nrpages;
+       if (nrpages) {
+               xa_lock_irqsave(&mapping->i_pages, flags);
+               nrpages = mapping->nrpages;
+               xa_unlock_irqrestore(&mapping->i_pages, flags);
+       } /* Workaround end */
+
+       LASSERTF(nrpages == 0, "%s: inode="DFID"(%p) nrpages=%lu, "
+                "see https://jira.whamcloud.com/browse/LU-118\n",
+                ll_get_fsname(inode->i_sb, NULL, 0),
+                PFID(ll_inode2fid(inode)), inode, nrpages);
+}
+
 int ll_read_inode2(struct inode *inode, void *opaque)
 {
         struct lustre_md *md = opaque;
@@ -2110,8 +2138,6 @@ int ll_read_inode2(struct inode *inode, void *opaque)
 void ll_delete_inode(struct inode *inode)
 {
        struct ll_inode_info *lli = ll_i2info(inode);
-       struct address_space *mapping = &inode->i_data;
-       unsigned long nrpages;
        ENTRY;
 
        if (S_ISREG(inode->i_mode) && lli->lli_clob != NULL) {
@@ -2125,26 +2151,8 @@ void ll_delete_inode(struct inode *inode)
                cl_sync_file_range(inode, 0, OBD_OBJECT_EOF, inode->i_nlink ?
                                   CL_FSYNC_LOCAL : CL_FSYNC_DISCARD, 1);
        }
-       truncate_inode_pages_final(mapping);
 
-       /* Workaround for LU-118: Note nrpages may not be totally updated when
-        * truncate_inode_pages() returns, as there can be a page in the process
-        * of deletion (inside __delete_from_page_cache()) in the specified
-        * range. Thus mapping->nrpages can be non-zero when this function
-        * returns even after truncation of the whole mapping.  Only do this if
-        * npages isn't already zero.
-        */
-       nrpages = mapping->nrpages;
-       if (nrpages) {
-               xa_lock_irq(&mapping->i_pages);
-               nrpages = mapping->nrpages;
-               xa_unlock_irq(&mapping->i_pages);
-       } /* Workaround end */
-
-       LASSERTF(nrpages == 0, "%s: inode="DFID"(%p) nrpages=%lu, "
-                "see https://jira.whamcloud.com/browse/LU-118\n",
-                ll_get_fsname(inode->i_sb, NULL, 0),
-                PFID(ll_inode2fid(inode)), inode, nrpages);
+       ll_truncate_inode_pages_final(inode);
 
 #ifdef HAVE_SBOPS_EVICT_INODE
        ll_clear_inode(inode);

diff --git a/lustre/llite/vvp_object.c b/lustre/llite/vvp_object.c
index c3bf715..de4ecd8 100644 (file)
--- a/lustre/llite/vvp_object.c
+++ b/lustre/llite/vvp_object.c
@@ -168,13 +168,8 @@ static int vvp_prune(const struct lu_env *env, struct cl_object *obj)
                RETURN(rc);
        }
 
-       truncate_inode_pages(inode->i_mapping, 0);
-       if (inode->i_mapping->nrpages) {
-               CDEBUG(D_VFSTRACE, DFID ": still has %lu pages remaining\n",
-                      PFID(lu_object_fid(&obj->co_lu)),
-                      inode->i_mapping->nrpages);
-               RETURN(-EIO);
-       }
+       ll_truncate_inode_pages_final(inode);
+       clear_bit(AS_EXITING, &inode->i_mapping->flags);
 
        RETURN(0);
 }