# HG changeset patch
# User maybee
# Date 1185248587 25200
# Node ID dc10a713d1a07f67a7994bc04adef332c1604ea6
# Parent  83b15139bb8620277f9504bada6c447b42a38c15
6453407 rm a file when the root file system is at its quota limit reports ENOSPC
6483887 without direct management, arc ghost lists can run amok
6557767 assertion failed: error == 17 || lr->lr_length <= zp->z_blksz
6577156 zfs_putapage discards pages too easily
6581978 assertion failed: koff <= filesz, file: ../../common/fs/zfs/zfs_vnops.c, line: 2834
6582642 cannot set property while quota exceed or really out of space
6583710 assertion failed: new_state->arcs_size + to_delta >= *size (0x5a4cc00 >= 0x5a50a00)

diff -r 83b15139bb86 -r dc10a713d1a0 usr/src/uts/common/fs/zfs/arc.c
--- a/usr/src/uts/common/fs/zfs/arc.c	Mon Jul 23 18:07:13 2007 -0700
+++ b/usr/src/uts/common/fs/zfs/arc.c	Mon Jul 23 20:43:07 2007 -0700
@@ -379,6 +379,7 @@
 static void arc_get_data_buf(arc_buf_t *buf);
 static void arc_access(arc_buf_hdr_t *buf, kmutex_t *hash_lock);
 static int arc_evict_needed(arc_buf_contents_t type);
+static void arc_evict_ghost(arc_state_t *state, int64_t bytes);
 
 #define	GHOST_STATE(state)	\
 	((state) == arc_mru_ghost || (state) == arc_mfu_ghost)
@@ -842,7 +843,6 @@
 				to_delta = ab->b_size;
 			}
 			atomic_add_64(size, to_delta);
-			ASSERT3U(new_state->arcs_size + to_delta, >=, *size);
 
 			if (use_mutex)
 				mutex_exit(&new_state->arcs_mtx);
@@ -1258,6 +1258,27 @@
 	if (missed)
 		ARCSTAT_INCR(arcstat_mutex_miss, missed);
 
+	/*
+	 * We have just evicted some date into the ghost state, make
+	 * sure we also adjust the ghost state size if necessary.
+	 */
+	if (arc_no_grow &&
+	    arc_mru_ghost->arcs_size + arc_mfu_ghost->arcs_size > arc_c) {
+		int64_t mru_over = arc_anon->arcs_size + arc_mru->arcs_size +
+		    arc_mru_ghost->arcs_size - arc_c;
+
+		if (mru_over > 0 && arc_mru_ghost->arcs_lsize[type] > 0) {
+			int64_t todelete =
+			    MIN(arc_mru_ghost->arcs_lsize[type], mru_over);
+			arc_evict_ghost(arc_mru_ghost, todelete);
+		} else if (arc_mfu_ghost->arcs_lsize[type] > 0) {
+			int64_t todelete = MIN(arc_mfu_ghost->arcs_lsize[type],
+			    arc_mru_ghost->arcs_size +
+			    arc_mfu_ghost->arcs_size - arc_c);
+			arc_evict_ghost(arc_mfu_ghost, todelete);
+		}
+	}
+
 	return (stolen);
 }
 
diff -r 83b15139bb86 -r dc10a713d1a0 usr/src/uts/common/fs/zfs/dsl_dir.c
--- a/usr/src/uts/common/fs/zfs/dsl_dir.c	Mon Jul 23 18:07:13 2007 -0700
+++ b/usr/src/uts/common/fs/zfs/dsl_dir.c	Mon Jul 23 20:43:07 2007 -0700
@@ -698,6 +698,7 @@
 	uint64_t est_used, quota, parent_rsrv;
 	int edquot = EDQUOT;
 	int txgidx = txg & TXG_MASK;
+	boolean_t ismos;
 	int i;
 	struct tempreserve *tr;
 
@@ -713,30 +714,31 @@
 	for (i = 0; i < TXG_SIZE; i++)
 		est_used += dd->dd_tempreserved[i];
 
-	quota = UINT64_MAX;
-
-	if (dd->dd_phys->dd_quota)
+	/*
+	 * If this transaction will result in a net free of space, we want
+	 * to let it through.
+	 */
+	if (netfree || dd->dd_phys->dd_quota == 0)
+		quota = UINT64_MAX;
+	else
 		quota = dd->dd_phys->dd_quota;
 
 	/*
-	 * If this transaction will result in a net free of space, we want
-	 * to let it through, but we have to be careful: the space that it
-	 * frees won't become available until *after* this txg syncs.
-	 * Therefore, to ensure that it's possible to remove files from
-	 * a full pool without inducing transient overcommits, we throttle
+	 * Adjust the quota against the actual pool size at the root.
+	 * To ensure that it's possible to remove files from a full
+	 * pool without inducing transient overcommits, we throttle
 	 * netfree transactions against a quota that is slightly larger,
 	 * but still within the pool's allocation slop.  In cases where
 	 * we're very close to full, this will allow a steady trickle of
 	 * removes to get through.
 	 */
-	if (dd->dd_parent == NULL) {
+	ismos = (dd->dd_phys->dd_head_dataset_obj == 0);
+	if (dd->dd_parent == NULL || ismos) {
 		uint64_t poolsize = dsl_pool_adjustedsize(dd->dd_pool, netfree);
 		if (poolsize < quota) {
 			quota = poolsize;
 			edquot = ENOSPC;
 		}
-	} else if (netfree) {
-		quota = UINT64_MAX;
 	}
 
 	/*
@@ -771,7 +773,7 @@
 	list_insert_tail(tr_list, tr);
 
 	/* see if it's OK with our parent */
-	if (dd->dd_parent && parent_rsrv) {
+	if (dd->dd_parent && parent_rsrv && !ismos) {
 		return (dsl_dir_tempreserve_impl(dd->dd_parent,
 		    parent_rsrv, netfree, tr_list, tx));
 	} else {
diff -r 83b15139bb86 -r dc10a713d1a0 usr/src/uts/common/fs/zfs/zfs_vnops.c
--- a/usr/src/uts/common/fs/zfs/zfs_vnops.c	Mon Jul 23 18:07:13 2007 -0700
+++ b/usr/src/uts/common/fs/zfs/zfs_vnops.c	Mon Jul 23 20:43:07 2007 -0700
@@ -875,7 +875,8 @@
 		lr->lr_blkoff = off - boff;
 		error = dmu_sync(zio, db, &lr->lr_blkptr,
 		    lr->lr_common.lrc_txg, zfs_get_done, zgd);
-		ASSERT(error == EEXIST || lr->lr_length <= zp->z_blksz);
+		ASSERT((error && error != EINPROGRESS) ||
+		    lr->lr_length <= zp->z_blksz);
 		if (error == 0) {
 			zil_add_vdev(zfsvfs->z_log,
 			    DVA_GET_VDEV(BP_IDENTITY(&lr->lr_blkptr)));
@@ -2810,8 +2811,10 @@
 	rl_t		*rl;
 	u_offset_t	off, koff;
 	size_t		len, klen;
+	uint64_t	filesz;
 	int		err;
 
+	filesz = zp->z_phys->zp_size;
 	off = pp->p_offset;
 	len = PAGESIZE;
 	/*
@@ -2819,9 +2822,7 @@
 	 * muiltiple pages so that we write a full block (thus avoiding
 	 * a read-modify-write).
 	 */
-	if (zp->z_blksz > PAGESIZE) {
-		uint64_t filesz = zp->z_phys->zp_size;
-
+	if (off < filesz && zp->z_blksz > PAGESIZE) {
 		if (!ISP2(zp->z_blksz)) {
 			/* Only one block in the file. */
 			klen = P2ROUNDUP((ulong_t)zp->z_blksz, PAGESIZE);
@@ -2841,20 +2842,20 @@
 	/*
 	 * Can't push pages past end-of-file.
 	 */
-	if (off >= zp->z_phys->zp_size) {
-		/* discard all pages */
-		flags |= B_INVAL;
+	filesz = zp->z_phys->zp_size;
+	if (off >= filesz) {
+		/* ignore all pages */
 		err = 0;
 		goto out;
-	} else if (off + len > zp->z_phys->zp_size) {
-		int npages = btopr(zp->z_phys->zp_size - off);
+	} else if (off + len > filesz) {
+		int npages = btopr(filesz - off);
 		page_t *trunc;
 
 		page_list_break(&pp, &trunc, npages);
-		/* discard pages past end of file */
+		/* ignore pages past end of file */
 		if (trunc)
-			pvn_write_done(trunc, B_INVAL | flags);
-		len = zp->z_phys->zp_size - off;
+			pvn_write_done(trunc, flags);
+		len = filesz - off;
 	}
 
 	tx = dmu_tx_create(zfsvfs->z_os);
@@ -2890,7 +2891,7 @@
 
 out:
 	zfs_range_unlock(rl);
-	pvn_write_done(pp, (err ? B_ERROR : 0) | B_WRITE | flags);
+	pvn_write_done(pp, (err ? B_ERROR : 0) | flags);
 	if (offp)
 		*offp = off;
 	if (lenp)