Don't clear PinPending without a good reason.

This definitely is wrong.  If an inode change is waiting for
previous-phase data to be written, we don't want to lose
the PinPending.

Signed-off-by: NeilBrown <neilb@suse.de>

diff --git a/cluster.c b/cluster.c
index 0bf1db5adcf025cffd3d2ab192816f2bb9356967..13f5f727e241a335889d53163b8695bd669fac27 100644 (file)
--- a/cluster.c
+++ b/cluster.c
@@ -703,7 +703,6 @@ int lafs_cluster_allocate(struct block *b, int cnum)
                                if (!test_and_clear_bit(B_ICredit, &b2->flags))
                                        credits--;
                }
-               clear_bit(B_PinPending, &b2->flags); // FIXME not sure this is right
                LAFS_BUG(credits < 0, b2);
                lafs_space_return(fs, credits);
                /* make sure 'dirty' status is registered */