6887924 PP_ISKAS needs to be defined in terms of VN_ISKAS for vnodes

6871169 kphysm_add_memory_dynamic panics testing vmstress + DR
6877160 guest domain panic at platsvc:mdeg_notify_client+10c
6874763 memseg_alloc_meta() incorrectly maps page_t pages
6873569 multiple calls to memlist_read_lock() can cause deadlock hazard
6886354 DR failure with "memory span duplication" error
6886782 panic after pagefault in seg_kpm after LDom add-mem on primary
6887644 domain hang/deadlock during ldom mem DR when trying to grab a write lock

7 files changed, 124 insertions, 207 deletions

diff --git a/usr/src/uts/sun4v/io/dr_mem.c b/usr/src/uts/sun4v/io/dr_mem.c
index 2b346a6d48..a32b0af792 100644
--- a/usr/src/uts/sun4v/io/dr_mem.c
+++ b/usr/src/uts/sun4v/io/dr_mem.c
@@ -203,10 +203,6 @@ static mde_cookie_t dr_mem_find_node_md(dr_mem_blk_t *, md_t *, mde_cookie_t *);
 static int mem_add(pfn_t, pgcnt_t);
 static int mem_del(pfn_t, pgcnt_t);
 
-static size_t rsvaddsz;
-extern void i_dr_mem_init(uint64_t *);
-extern void i_dr_mem_fini();
-extern void i_dr_mem_update();
 extern int kphysm_add_memory_dynamic(pfn_t, pgcnt_t);
 
 int
@@ -261,8 +257,6 @@ dr_mem_init(void)
 		return (rv);
 	}
 
-	i_dr_mem_init(&rsvaddsz);
-
 	return (0);
 }
 
@@ -271,8 +265,6 @@ dr_mem_fini(void)
 {
 	int rv;
 
-	i_dr_mem_fini();
-
 	if ((rv = ds_cap_fini(&dr_mem_cap)) != 0) {
 		cmn_err(CE_NOTE, "dr_mem: ds_cap_fini failed: %d", rv);
 	}
@@ -720,6 +712,8 @@ dr_mem_list_query(dr_mem_hdr_t *req, dr_mem_hdr_t **resp, int *resp_len)
 	dr_mem_hdr_t	*rp;
 	dr_mem_query_t	*stat;
 
+	drctl_block();
+
 	/* the incoming array of req_mblks to configure */
 	req_mblks = DR_MEM_CMD_MBLKS(req);
 
@@ -764,6 +758,8 @@ dr_mem_list_query(dr_mem_hdr_t *req, dr_mem_hdr_t **resp, int *resp_len)
 	*resp = rp;
 	*resp_len = rlen;
 
+	drctl_unblock();
+
 	return (0);
 }
 
@@ -832,7 +828,7 @@ static int
 dr_mem_configure(dr_mem_blk_t *mbp, int *status)
 {
 	int rv;
-	uint64_t addr, size, addsz;
+	uint64_t addr, size;
 
 	rv = 0;
 	addr = mbp->addr;
@@ -854,64 +850,10 @@ dr_mem_configure(dr_mem_blk_t *mbp, int *status)
 			*status = DR_MEM_STAT_UNCONFIGURED;
 			rv = DR_MEM_RES_FAILURE;
 		}
-	} else if (rsvaddsz) {
-		addr += size;
-
-		/*
-		 * Add up to the first <rsvaddsz> portion of mblock
-		 * first since that portion has reserved meta pages.
-		 * This will likely guarantee an additional amount of
-		 * free pages from which we may have to allocate the
-		 * rest of the meta pages.
-		 *
-		 * Break up the request in descending order (if needed)
-		 * in order to ensure that cage grows from the high end
-		 * of the original request.
-		 */
-		for (addsz = MIN(size, rsvaddsz); addsz > 0; addsz = size) {
-			ASSERT(addr >= mbp->addr);
-			DR_DBG_MEM("addsz=0x%lx  size=0x%lx\n", addsz, size);
-			if (rv = mem_add(btop(addr - addsz), btop(addsz))) {
-				DR_DBG_MEM("failed to configure span"
-				    " 0x%lx.0x%lx (%d)\n", addr, addsz, rv);
-				break;
-			} else {
-				size -= addsz;
-				addr -= addsz;
-			}
-		}
-
-		/*
-		 * Mark the mblock configured if any span
-		 * in that mblock was successfully added.
-		 *
-		 * In case of partial success:
-		 *
-		 *	rv != DR_MEM_RES_OK
-		 *	status == DR_MEM_STAT_CONFIGURED
-		 *
-		 * mark span actually configured.
-		 */
-		if (size == mbp->size && rv != KPHYSM_ESPAN) {
-			*status = DR_MEM_STAT_UNCONFIGURED;
-		} else {
-			DR_DBG_MEM("failed (partial) to configure span"
-			    " 0x%lx.0x%lx (%d)\n", addr, addsz, rv);
-			*status = DR_MEM_STAT_CONFIGURED;
-			mbp->addr = addr;
-			mbp->size -= size;
-		}
-
-		rv = cvt_err(rv);
-		i_dr_mem_update();
 	} else {
-		/*
-		 * The reserved feature is disabled, add whole mblock.
-		 */
 		rv = mem_add(btop(addr), btop(size));
 		DR_DBG_MEM("addr=0x%lx size=0x%lx rv=%d\n", addr, size, rv);
 		if (rv) {
-			rv = cvt_err(rv);
 			*status = DR_MEM_STAT_UNCONFIGURED;
 		} else {
 			*status = DR_MEM_STAT_CONFIGURED;
@@ -934,7 +876,6 @@ dr_mem_unconfigure(dr_mem_blk_t *mbp, int *status)
 			*status = DR_MEM_STAT_CONFIGURED;
 			rv = DR_MEM_RES_EINVAL;
 	} else if (rv = mem_del(btop(mbp->addr), btop(mbp->size))) {
-		rv = cvt_err(rv);
 		*status = DR_MEM_STAT_CONFIGURED;
 	} else {
 		*status = DR_MEM_STAT_UNCONFIGURED;
@@ -1117,15 +1058,16 @@ mem_add(pfn_t base, pgcnt_t npgs)
 	DR_DBG_MEM("%s: begin base=0x%lx npgs=0x%lx\n", __func__, base, npgs);
 
 	if (npgs == 0)
-		return (0);
+		return (DR_MEM_RES_OK);
 
 	rv = kphysm_add_memory_dynamic(base, npgs);
 	DR_DBG_MEM("%s: kphysm_add(0x%lx, 0x%lx) = %d", __func__, base, npgs,
 	    rv);
-	if (!rv) {
+	if (rv == KPHYSM_OK) {
 		if (rc = kcage_range_add(base, npgs, KCAGE_DOWN))
 			cmn_err(CE_WARN, "kcage_range_add() = %d", rc);
 	}
+	rv = cvt_err(rv);
 	return (rv);
 }
 
@@ -1145,6 +1087,7 @@ static int
 mem_del(pfn_t base, pgcnt_t npgs)
 {
 	int rv, err, del_range = 0;
+	int convert = 1;
 	mem_sync_t ms;
 	memquery_t mq;
 	memhandle_t mh;
@@ -1154,10 +1097,11 @@ mem_del(pfn_t base, pgcnt_t npgs)
 	DR_DBG_MEM("%s: begin base=0x%lx npgs=0x%lx\n", __func__, base, npgs);
 
 	if (npgs == 0)
-		return (0);
+		return (DR_MEM_RES_OK);
 
 	if ((rv = kphysm_del_gethandle(&mh)) != KPHYSM_OK) {
 		cmn_err(CE_WARN, "%s: del_gethandle() = %d", __func__, rv);
+		rv = cvt_err(rv);
 		return (rv);
 	}
 	if ((rv = kphysm_del_span_query(base, npgs, &mq))
@@ -1168,10 +1112,19 @@ mem_del(pfn_t base, pgcnt_t npgs)
 	if (mq.nonrelocatable) {
 		DR_DBG_MEM("%s: non-reloc pages = %ld",
 		    __func__, mq.nonrelocatable);
-		rv = KPHYSM_ENONRELOC;
+		rv  = KPHYSM_ENONRELOC;
 		goto done;
 	}
 	if (rv = kcage_range_delete(base, npgs)) {
+		switch (rv) {
+		case EBUSY:
+			rv = DR_MEM_RES_ENOTVIABLE;
+			break;
+		default:
+			rv = DR_MEM_RES_FAILURE;
+			break;
+		}
+		convert = 0; /* conversion done */
 		cmn_err(CE_WARN, "%s: del_range() = %d", __func__, rv);
 		goto done;
 	} else {
@@ -1183,6 +1136,18 @@ mem_del(pfn_t base, pgcnt_t npgs)
 	}
 	if ((rv = memlist_add_span(ptob(base), ptob(npgs), &d_ml))
 	    != MEML_SPANOP_OK) {
+		switch (rv) {
+		case MEML_SPANOP_ESPAN:
+			rv = DR_MEM_RES_ESPAN;
+			break;
+		case MEML_SPANOP_EALLOC:
+			rv = DR_MEM_RES_ERESOURCE;
+			break;
+		default:
+			rv = DR_MEM_RES_FAILURE;
+			break;
+		}
+		convert = 0; /* conversion done */
 		cmn_err(CE_WARN, "%s: add_span() = %d", __func__, rv);
 		goto done;
 	}
@@ -1245,6 +1210,8 @@ done:
 
 	if ((err = kphysm_del_release(mh)) != KPHYSM_OK)
 		cmn_err(CE_WARN, "%s: del_release() = %d", __func__, err);
+	if (convert)
+		rv = cvt_err(rv);
 
 	DR_DBG_MEM("%s: rv=%d", __func__, rv);
 
diff --git a/usr/src/uts/sun4v/io/drctl.c b/usr/src/uts/sun4v/io/drctl.c
index 5061b1228c..4ca542716f 100644
--- a/usr/src/uts/sun4v/io/drctl.c
+++ b/usr/src/uts/sun4v/io/drctl.c
@@ -156,6 +156,7 @@ _init(void)
 
 	drctlp->drc_inst = -1;
 	mutex_init(&drctlp->drc_lock, NULL, MUTEX_DRIVER, NULL);
+	cv_init(&drctlp->drc_busy_cv, NULL, CV_DRIVER, NULL);
 
 	if ((rv = mod_install(&modlinkage)) != 0)
 		mutex_destroy(&drctlp->drc_lock);
@@ -171,7 +172,7 @@ _fini(void)
 
 	if ((rv = mod_remove(&modlinkage)) != 0)
 		return (rv);
-
+	cv_destroy(&drctlp->drc_busy_cv);
 	mutex_destroy(&drctlp->drc_lock);
 	return (0);
 }
@@ -376,7 +377,6 @@ verify_response(int cmd,
 	return (0);
 }
 
-
 static int
 drctl_config_common(int cmd, int flags, drctl_rsrc_t *res,
     int count, drctl_resp_t **rbuf, size_t *rsize, size_t *rq_size)
@@ -460,7 +460,6 @@ drctl_config_init(int cmd, int flags, drctl_rsrc_t *res,
 	}
 
 	mutex_enter(&drctlp->drc_lock);
-
 	if (drctlp->drc_busy != NULL) {
 		mutex_exit(&drctlp->drc_lock);
 		*rbuf = drctl_generate_err_resp(busy_msg, rsize);
@@ -488,6 +487,7 @@ drctl_config_init(int cmd, int flags, drctl_rsrc_t *res,
 			kmem_free(*rbuf, *rsize);
 			*rbuf = drctl_generate_err_resp(rsp_msg, rsize);
 			drctlp->drc_busy = NULL;
+			cv_broadcast(&drctlp->drc_busy_cv);
 		} else { /* message format is valid */
 			drctlp->drc_busy = ck;
 			drctlp->drc_cmd = cmd;
@@ -511,8 +511,8 @@ drctl_config_init(int cmd, int flags, drctl_rsrc_t *res,
 		drctlp->drc_cmd = -1;
 		drctlp->drc_flags = 0;
 		drctlp->drc_busy = NULL;
+		cv_broadcast(&drctlp->drc_busy_cv);
 	}
-
 	return (rv);
 }
 
@@ -528,12 +528,10 @@ drctl_config_fini(drctl_cookie_t ck, drctl_rsrc_t *res, int count)
 	size_t rq_size;
 
 	mutex_enter(&drctlp->drc_lock);
-
 	if (drctlp->drc_busy != ck) {
 		mutex_exit(&drctlp->drc_lock);
 		return (EBUSY);
 	}
-
 	mutex_exit(&drctlp->drc_lock);
 
 	flags = drctlp->drc_flags;
@@ -579,11 +577,11 @@ drctl_config_fini(drctl_cookie_t ck, drctl_rsrc_t *res, int count)
 	    flags, res, count, NULL, 0, &rq_size);
 
 done:
-	drctlp->drc_cmd = -1;
-	drctlp->drc_flags = 0;
-	drctlp->drc_busy = NULL;
-
-	return (rv);
+		drctlp->drc_cmd = -1;
+		drctlp->drc_flags = 0;
+		drctlp->drc_busy = NULL;
+		cv_broadcast(&drctlp->drc_busy_cv);
+		return (rv);
 }
 
 static int
@@ -697,3 +695,35 @@ pack_message(int cmd,
 
 	return (msgp);
 }
+
+/*
+ * Block DR operations
+ */
+void
+drctl_block(void)
+{
+	/* Wait for any in progress DR operation to complete */
+	mutex_enter(&drctlp->drc_lock);
+	while (drctlp->drc_busy != NULL)
+		(void) cv_wait_sig(&drctlp->drc_busy_cv, &drctlp->drc_lock);
+	/* Mark the link busy */
+	drctlp->drc_busy = (drctl_cookie_t)-1;
+	drctlp->drc_cmd = DRCTL_DRC_BLOCK;
+	drctlp->drc_flags = 0;
+	mutex_exit(&drctlp->drc_lock);
+}
+
+/*
+ * Unblock DR operations
+ */
+void
+drctl_unblock(void)
+{
+	/* Mark the link free */
+	mutex_enter(&drctlp->drc_lock);
+	drctlp->drc_cmd = -1;
+	drctlp->drc_flags = 0;
+	drctlp->drc_busy = NULL;
+	cv_broadcast(&drctlp->drc_busy_cv);
+	mutex_exit(&drctlp->drc_lock);
+}
diff --git a/usr/src/uts/sun4v/io/vlds.c b/usr/src/uts/sun4v/io/vlds.c
index 4eedeee56b..a33978680e 100644
--- a/usr/src/uts/sun4v/io/vlds.c
+++ b/usr/src/uts/sun4v/io/vlds.c
@@ -278,7 +278,6 @@ _init(void)
 	}
 
 	vlds_mdeg_init();
-	(void) vlds_mdeg_register();
 
 	return (s);
 }
@@ -293,8 +292,6 @@ _fini(void)
 
 	ddi_soft_state_fini(&vlds_statep);
 
-	(void) vlds_mdeg_unregister();
-
 	return (s);
 }
 
@@ -338,6 +335,8 @@ vlds_attach(dev_info_t *devi, ddi_attach_cmd_t cmd)
 
 	vlds_minor_init();
 
+	(void) vlds_mdeg_register();
+
 	return (DDI_SUCCESS);
 }
 
@@ -352,6 +351,7 @@ vlds_detach(dev_info_t *devi, ddi_detach_cmd_t cmd)
 
 	vlds_minor_free(vlds_minor_bitmap);
 	ddi_remove_minor_node(devi, NULL);
+	(void) vlds_mdeg_unregister();
 	return (DDI_SUCCESS);
 }
 
diff --git a/usr/src/uts/sun4v/os/memseg.c b/usr/src/uts/sun4v/os/memseg.c
index d562905eec..359588150a 100644
--- a/usr/src/uts/sun4v/os/memseg.c
+++ b/usr/src/uts/sun4v/os/memseg.c
@@ -44,9 +44,6 @@ extern struct memseg *memseg_alloc();
 extern page_t *ppvm_base;
 extern pgcnt_t ppvm_size;
 
-static vnode_t pp_vn, rsv_vn;
-static pgcnt_t rsv_metapgs;
-static int meta_rsv_enable;
 static int sun4v_memseg_debug;
 
 extern struct memseg *memseg_reuse(pgcnt_t);
@@ -77,12 +74,11 @@ extern void remap_to_dummy(caddr_t, pgcnt_t);
 int
 memseg_alloc_meta(pfn_t base, pgcnt_t npgs, void **ptp, pgcnt_t *metap)
 {
-	page_t		*pp, *opp, *epp, *pgpp;
+	page_t		*pp, *opp, *epp;
 	pgcnt_t		metapgs;
-	int		i, rsv;
+	int		i;
 	struct seg	kseg;
 	caddr_t		vaddr;
-	u_offset_t	off;
 
 	/*
 	 * Verify incoming memory is within supported DR range.
@@ -95,7 +91,7 @@ memseg_alloc_meta(pfn_t base, pgcnt_t npgs, void **ptp, pgcnt_t *metap)
 	metapgs = btopr(npgs * sizeof (page_t));
 
 	if (!IS_P2ALIGNED((uint64_t)pp, PAGESIZE) &&
-	    page_find(&pp_vn, (u_offset_t)pp)) {
+	    page_find(&mpvp, (u_offset_t)pp)) {
 		/*
 		 * Another memseg has page_t's in the same
 		 * page which 'pp' resides.  This would happen
@@ -120,7 +116,7 @@ memseg_alloc_meta(pfn_t base, pgcnt_t npgs, void **ptp, pgcnt_t *metap)
 	}
 
 	if (!IS_P2ALIGNED((uint64_t)epp, PAGESIZE) &&
-	    page_find(&pp_vn, (u_offset_t)epp)) {
+	    page_find(&mpvp, (u_offset_t)epp)) {
 		/*
 		 * Another memseg has page_t's in the same
 		 * page which 'epp' resides.  This would happen
@@ -144,59 +140,20 @@ memseg_alloc_meta(pfn_t base, pgcnt_t npgs, void **ptp, pgcnt_t *metap)
 	vaddr = (caddr_t)pp;
 
 	for (i = 0; i < metapgs; i++)
-		if (page_find(&pp_vn, (u_offset_t)(vaddr + i * PAGESIZE)))
+		if (page_find(&mpvp, (u_offset_t)(vaddr + i * PAGESIZE)))
 			panic("page_find(0x%p, %p)\n",
-			    (void *)&pp_vn, (void *)(vaddr + i * PAGESIZE));
+			    (void *)&mpvp, (void *)(vaddr + i * PAGESIZE));
 
 	/*
 	 * Allocate the metadata pages; these are the pages that will
 	 * contain the page_t's for the incoming memory.
-	 *
-	 * If a normal allocation fails, use the reserved metapgs for
-	 * a small allocation; otherwise retry with PG_WAIT.
 	 */
-	rsv = off = 0;
-	if (metapgs <= rsv_metapgs) {
-		MEMSEG_DEBUG("memseg_get: use rsv 0x%lx metapgs", metapgs);
-		ASSERT(meta_rsv_enable);
-		rsv = 1;
-	} else if ((pgpp = page_create_va(&pp_vn, (u_offset_t)pp, ptob(metapgs),
+	if ((page_create_va(&mpvp, (u_offset_t)pp, ptob(metapgs),
 	    PG_NORELOC | PG_EXCL, &kseg, vaddr)) == NULL) {
-		cmn_err(CE_WARN, "memseg_get: can't get 0x%ld metapgs",
+		MEMSEG_DEBUG("memseg_alloc_meta: can't get 0x%ld metapgs",
 		    metapgs);
 		return (KPHYSM_ERESOURCE);
 	}
-	if (rsv) {
-		/*
-		 * The reseve pages must be hashed out of the reserve vnode
-		 * and rehashed by <pp_vn,vaddr>.  The resreved pages also
-		 * must be replenished immedidately at the end of the add
-		 * processing.
-		 */
-		for (i = 0; i < metapgs; i++) {
-			pgpp = page_find(&rsv_vn, off);
-			ASSERT(pgpp);
-			page_hashout(pgpp, 0);
-			hat_devload(kas.a_hat, vaddr, PAGESIZE,
-			    page_pptonum(pgpp), PROT_READ | PROT_WRITE,
-			    HAT_LOAD | HAT_LOAD_REMAP | HAT_LOAD_NOCONSIST);
-			ASSERT(!page_find(&pp_vn, (u_offset_t)vaddr));
-			if (!page_hashin(pgpp, &pp_vn, (u_offset_t)vaddr, 0))
-				panic("memseg_get: page_hashin(0x%p, 0x%p)",
-				    (void *)pgpp, (void *)vaddr);
-			off += PAGESIZE;
-			vaddr += PAGESIZE;
-			rsv_metapgs--;
-		}
-	} else {
-		for (i = 0; i < metapgs; i++) {
-			hat_devload(kas.a_hat, vaddr, PAGESIZE,
-			    page_pptonum(pgpp), PROT_READ | PROT_WRITE,
-			    HAT_LOAD | HAT_LOAD_REMAP | HAT_LOAD_NOCONSIST);
-			pgpp = pgpp->p_next;
-			vaddr += PAGESIZE;
-		}
-	}
 
 	ASSERT(ptp);
 	ASSERT(metap);
@@ -228,7 +185,7 @@ memseg_free_meta(void *ptp, pgcnt_t metapgs)
 	 * Free pages allocated during add.
 	 */
 	for (i = 0; i < metapgs; i++) {
-		pp = page_find(&pp_vn, off);
+		pp = page_find(&mpvp, off);
 		ASSERT(pp);
 		ASSERT(pp->p_szc == 0);
 		page_io_unlock(pp);
@@ -248,7 +205,7 @@ memseg_get_metapfn(void *ptp, pgcnt_t metapg)
 	ASSERT(off);
 	ASSERT(IS_P2ALIGNED((uint64_t)off, PAGESIZE));
 
-	pp = page_find(&pp_vn, off);
+	pp = page_find(&mpvp, off);
 	ASSERT(pp);
 	ASSERT(pp->p_szc == 0);
 	ASSERT(pp->p_pagenum != PFN_INVALID);
@@ -285,7 +242,7 @@ memseg_remap_meta(struct memseg *seg)
 	 */
 
 	if (!IS_P2ALIGNED((uint64_t)pp, PAGESIZE) &&
-	    page_find(&pp_vn, (u_offset_t)(pp - 1)) && !page_deleted(pp - 1)) {
+	    page_find(&mpvp, (u_offset_t)(pp - 1)) && !page_deleted(pp - 1)) {
 		/*
 		 * Another memseg has page_t's in the same
 		 * page which 'pp' resides.  This would happen
@@ -312,7 +269,7 @@ memseg_remap_meta(struct memseg *seg)
 	}
 
 	if (!IS_P2ALIGNED((uint64_t)epp, PAGESIZE) &&
-	    page_find(&pp_vn, (u_offset_t)epp) && !page_deleted(epp)) {
+	    page_find(&mpvp, (u_offset_t)epp) && !page_deleted(epp)) {
 		/*
 		 * Another memseg has page_t's in the same
 		 * page which 'epp' resides.  This would happen
@@ -333,13 +290,13 @@ memseg_remap_meta(struct memseg *seg)
 
 	off = (u_offset_t)pp;
 
-	MEMSEG_DEBUG("memseg_remap: off=0x%lx metapgs=0x%lx\n", (uint64_t)off,
-	    metapgs);
+	MEMSEG_DEBUG("memseg_remap_meta: off=0x%lx metapgs=0x%lx\n",
+	    (uint64_t)off, metapgs);
 	/*
 	 * Free pages allocated during add.
 	 */
 	for (i = 0; i < metapgs; i++) {
-		pp = page_find(&pp_vn, off);
+		pp = page_find(&mpvp, off);
 		ASSERT(pp);
 		ASSERT(pp->p_szc == 0);
 		page_io_unlock(pp);
@@ -347,64 +304,3 @@ memseg_remap_meta(struct memseg *seg)
 		off += PAGESIZE;
 	}
 }
-
-static void
-rsv_alloc()
-{
-	int i;
-	page_t *pp;
-	pgcnt_t metapgs;
-	u_offset_t off;
-	struct seg kseg;
-
-	kseg.s_as = &kas;
-
-	/*
-	 * Reserve enough page_t pages for an add request of
-	 * RSV_SIZE bytes.
-	 */
-	metapgs = btopr(btop(RSV_SIZE) * sizeof (page_t)) - rsv_metapgs;
-
-	for (i = off = 0; i < metapgs; i++, off += PAGESIZE) {
-		(void) page_create_va(&rsv_vn, off, PAGESIZE,
-		    PG_NORELOC | PG_WAIT, &kseg, 0);
-		pp = page_find(&rsv_vn, off);
-		ASSERT(pp);
-		ASSERT(PAGE_EXCL(pp));
-		page_iolock_init(pp);
-		rsv_metapgs++;
-	}
-}
-
-void
-i_dr_mem_init(size_t *hint)
-{
-	if (meta_rsv_enable) {
-		rsv_alloc();
-		if (hint)
-			*hint = RSV_SIZE;
-	}
-}
-
-void
-i_dr_mem_fini()
-{
-	int i;
-	page_t *pp;
-	u_offset_t off;
-
-	for (i = off = 0; i < rsv_metapgs; i++, off += PAGESIZE) {
-		if (pp = page_find(&rsv_vn, off)) {
-			ASSERT(PAGE_EXCL(pp));
-			page_destroy(pp, 0);
-		}
-		ASSERT(!page_find(&rsv_vn, off));
-	}
-	rsv_metapgs = 0;
-}
-
-void
-i_dr_mem_update()
-{
-	rsv_alloc();
-}
diff --git a/usr/src/uts/sun4v/promif/promif_emul.c b/usr/src/uts/sun4v/promif/promif_emul.c
index 143caa1f9d..1adf719ba6 100644
--- a/usr/src/uts/sun4v/promif/promif_emul.c
+++ b/usr/src/uts/sun4v/promif/promif_emul.c
@@ -36,6 +36,7 @@
 #include <sys/mdesc.h>
 #include <sys/mach_descrip.h>
 #include <sys/cpu_module.h>
+#include <vm/seg_kmem.h>
 
 #ifndef _KMDB
 #include <sys/pte.h>
@@ -50,7 +51,7 @@ int cif_cpu_mp_ready;
 int (*prom_cif_handler)(void *) = NULL;
 
 extern struct memlist *phys_avail;
-extern struct vnode prom_ppages;
+extern struct vnode promvp;
 extern void kdi_tlb_page_unlock(caddr_t, int);
 
 #define	COMBINE(hi, lo) (((uint64_t)(uint32_t)(hi) << 32) | (uint32_t)(lo))
@@ -309,7 +310,7 @@ unmap_prom_mappings(struct translation *transroot, size_t ntransroot)
 				ASSERT(PAGE_EXCL(pp));
 				ASSERT(PP_ISNORELOC(pp));
 				ASSERT(!PP_ISFREE(pp));
-				ASSERT(page_find(&prom_ppages, pfn));
+				ASSERT(page_find(&promvp, pfn));
 				ASSERT(page_get_pagecnt(pp->p_szc) == 1);
 
 				if (pp->p_mapping) {
diff --git a/usr/src/uts/sun4v/sys/drctl.h b/usr/src/uts/sun4v/sys/drctl.h
index 75299f2c93..26b4970541 100644
--- a/usr/src/uts/sun4v/sys/drctl.h
+++ b/usr/src/uts/sun4v/sys/drctl.h
@@ -20,7 +20,7 @@
  */
 
 /*
- * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
+ * Copyright 2009 Sun Microsystems, Inc.  All rights reserved.
  * Use is subject to license terms.
  */
 
@@ -46,7 +46,8 @@ typedef enum {
 	DRCTL_IO_CONFIG_REQUEST,
 	DRCTL_IO_CONFIG_NOTIFY,
 	DRCTL_IO_UNCONFIG_REQUEST,
-	DRCTL_IO_UNCONFIG_NOTIFY
+	DRCTL_IO_UNCONFIG_NOTIFY,
+	DRCTL_DRC_BLOCK
 } drctl_cmds_t;
 
 /*
@@ -135,6 +136,8 @@ typedef void *drctl_cookie_t;
 extern int drctl_config_init(int, int,
     drctl_rsrc_t *, int, drctl_resp_t **, size_t *, drctl_cookie_t);
 extern int drctl_config_fini(drctl_cookie_t, drctl_rsrc_t *, int);
+extern void drctl_block(void);
+extern void drctl_unblock(void);
 
 /*
  * Values for the 2nd arg (flags) of drctl_config_init
diff --git a/usr/src/uts/sun4v/vm/mach_kpm.c b/usr/src/uts/sun4v/vm/mach_kpm.c
index f47e74e001..de3a53f7a4 100644
--- a/usr/src/uts/sun4v/vm/mach_kpm.c
+++ b/usr/src/uts/sun4v/vm/mach_kpm.c
@@ -35,11 +35,15 @@
 #include <sys/machsystm.h>
 #include <vm/seg_kpm.h>
 #include <vm/mach_kpm.h>
+#include <vm/faultcode.h>
+
+extern pfn_t memseg_get_start(struct memseg *);
 
 /*
  * Kernel Physical Mapping (kpm) facility
  */
 
+
 void
 mach_kpm_init()
 {
@@ -226,7 +230,15 @@ hat_kpm_addmem_mseg_update(struct memseg *msp, pgcnt_t nkpmpgs,
 	 * if nkpmpgs needs to be used at some point.
 	 */
 
-	base = msp->pages_base;
+	/*
+	 * The meta (page_t) pages for dynamically added memory are allocated
+	 * either from the incoming memory itself or from existing memory.
+	 * In the former case the base of the incoming pages will be different
+	 * than the base of the dynamic segment so call memseg_get_start() to
+	 * get the actual base of the incoming memory for each case.
+	 */
+
+	base = memseg_get_start(msp);
 	end = msp->pages_end;
 
 	hat_devload(kas.a_hat, kpm_vbase + mmu_ptob(base),
@@ -259,7 +271,15 @@ hat_kpm_delmem_mseg_update(struct memseg *msp, struct memseg **mspp)
 {
 	pfn_t base, end;
 
-	base = msp->pages_base;
+	/*
+	 * The meta (page_t) pages for dynamically added memory are allocated
+	 * either from the incoming memory itself or from existing memory.
+	 * In the former case the base of the incoming pages will be different
+	 * than the base of the dynamic segment so call memseg_get_start() to
+	 * get the actual base of the incoming memory for each case.
+	 */
+
+	base = memseg_get_start(msp);
 	end = msp->pages_end;
 
 	hat_unload(kas.a_hat, kpm_vbase +  mmu_ptob(base), mmu_ptob(end - base),