Annotate KM_PUSHPAGE call paths with PF_NOFS

The txg_sync(), zfs_putpage(), zvol_write(), and zvol_discard()
call paths must only use KM_PUSHPAGE to avoid potential deadlocks
during direct reclaim.

This patch annotates these call paths so any accidental use of
KM_SLEEP will be quickly detected.   In the interest of stability
if debugging is disabled the offending allocation will have its
GFP flags automatically corrected.  When debugging is enabled
any misuse will be treated as a fatal error.

This patch is entirely for debugging.  We should be careful to
NOT become dependant on it fixing up the incorrect allocations.

Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>

diff --git a/module/zfs/txg.c b/module/zfs/txg.c
index c234567d7e5295f6437f8b417f7b5d3c4c3ce130..aefda6f69149086601ee1f35bfd492c3219f1c40 100644 (file)
--- a/module/zfs/txg.c
+++ b/module/zfs/txg.c
@@ -382,6 +382,15 @@ txg_sync_thread(dsl_pool_t *dp)
        callb_cpr_t cpr;
        uint64_t start, delta;
 
+#ifdef _KERNEL
+       /*
+        * Annotate this process with a flag that indicates that it is
+        * unsafe to use KM_SLEEP during memory allocations due to the
+        * potential for a deadlock.  KM_PUSHPAGE should be used instead.
+        */
+       current->flags |= PF_NOFS;
+#endif /* _KERNEL */
+
        txg_thread_enter(tx, &cpr);
 
        start = delta = 0;

diff --git a/module/zfs/zpl_file.c b/module/zfs/zpl_file.c
index 2e9f72ad12d99f3452b2dbe47c5983f374dd885f..9c27b7f1d3b2f614a0e996f980da858780ab8548 100644 (file)
--- a/module/zfs/zpl_file.c
+++ b/module/zfs/zpl_file.c
@@ -357,8 +357,16 @@ zpl_putpage(struct page *pp, struct writeback_control *wbc, void *data)
 
        ASSERT(PageLocked(pp));
        ASSERT(!PageWriteback(pp));
-
+       ASSERT(!(current->flags & PF_NOFS));
+
+       /*
+        * Annotate this call path with a flag that indicates that it is
+        * unsafe to use KM_SLEEP during memory allocations due to the
+        * potential for a deadlock.  KM_PUSHPAGE should be used instead.
+        */
+       current->flags |= PF_NOFS;
        (void) zfs_putpage(mapping->host, pp, wbc);
+       current->flags &= ~PF_NOFS;
 
        return (0);
 }

diff --git a/module/zfs/zvol.c b/module/zfs/zvol.c
index 07bda6dbaa4e742f649d7e7ecdcf2104be4ce78f..60ff64ebe7b03dfa41e96de42e611ecf02da7986 100644 (file)
--- a/module/zfs/zvol.c
+++ b/module/zfs/zvol.c
@@ -540,6 +540,14 @@ zvol_write(void *arg)
        dmu_tx_t *tx;
        rl_t *rl;
 
+       /*
+        * Annotate this call path with a flag that indicates that it is
+        * unsafe to use KM_SLEEP during memory allocations due to the
+        * potential for a deadlock.  KM_PUSHPAGE should be used instead.
+        */
+       ASSERT(!(current->flags & PF_NOFS));
+       current->flags |= PF_NOFS;
+
        if (req->cmd_flags & VDEV_REQ_FLUSH)
                zil_commit(zv->zv_zilog, ZVOL_OBJ);
 
@@ -548,7 +556,7 @@ zvol_write(void *arg)
         */
        if (size == 0) {
                blk_end_request(req, 0, size);
-               return;
+               goto out;
        }
 
        rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
@@ -562,7 +570,7 @@ zvol_write(void *arg)
                dmu_tx_abort(tx);
                zfs_range_unlock(rl);
                blk_end_request(req, -error, size);
-               return;
+               goto out;
        }
 
        error = dmu_write_req(zv->zv_objset, ZVOL_OBJ, req, tx);
@@ -578,6 +586,8 @@ zvol_write(void *arg)
                zil_commit(zv->zv_zilog, ZVOL_OBJ);
 
        blk_end_request(req, -error, size);
+out:
+       current->flags &= ~PF_NOFS;
 }
 
 #ifdef HAVE_BLK_QUEUE_DISCARD
@@ -592,14 +602,22 @@ zvol_discard(void *arg)
        int error;
        rl_t *rl;
 
+       /*
+        * Annotate this call path with a flag that indicates that it is
+        * unsafe to use KM_SLEEP during memory allocations due to the
+        * potential for a deadlock.  KM_PUSHPAGE should be used instead.
+        */
+       ASSERT(!(current->flags & PF_NOFS));
+       current->flags |= PF_NOFS;
+
        if (offset + size > zv->zv_volsize) {
                blk_end_request(req, -EIO, size);
-               return;
+               goto out;
        }
 
        if (size == 0) {
                blk_end_request(req, 0, size);
-               return;
+               goto out;
        }
 
        rl = zfs_range_lock(&zv->zv_znode, offset, size, RL_WRITER);
@@ -613,6 +631,8 @@ zvol_discard(void *arg)
        zfs_range_unlock(rl);
 
        blk_end_request(req, -error, size);
+out:
+       current->flags &= ~PF_NOFS;
 }
 #endif /* HAVE_BLK_QUEUE_DISCARD */