1 files changed, 148 insertions, 11 deletions

diff --git a/arch/powerpc/platforms/pseries/mobility.c b/arch/powerpc/platforms/pseries/mobility.c
index e83e0891272d..95fe802ccdfd 100644
--- a/arch/powerpc/platforms/pseries/mobility.c
+++ b/arch/powerpc/platforms/pseries/mobility.c
@@ -24,8 +24,11 @@
 #include <linux/stringify.h>
 
 #include <asm/machdep.h>
+#include <asm/nmi.h>
 #include <asm/rtas.h>
 #include "pseries.h"
+#include "vas.h"	/* vas_migration_handler() */
+#include "papr-hvpipe.h"	/* hvpipe_migration_handler() */
 #include "../../kernel/cacheinfo.h"
 
 static struct kobject *mobility_kobj;
@@ -47,6 +50,30 @@ struct update_props_workarea {
 #define MIGRATION_SCOPE	(1)
 #define PRRN_SCOPE -2
 
+#ifdef CONFIG_PPC_WATCHDOG
+static unsigned int nmi_wd_lpm_factor = 200;
+
+#ifdef CONFIG_SYSCTL
+static const struct ctl_table nmi_wd_lpm_factor_ctl_table[] = {
+	{
+		.procname	= "nmi_wd_lpm_factor",
+		.data		= &nmi_wd_lpm_factor,
+		.maxlen		= sizeof(int),
+		.mode		= 0644,
+		.proc_handler	= proc_douintvec_minmax,
+	},
+};
+
+static int __init register_nmi_wd_lpm_factor_sysctl(void)
+{
+	register_sysctl("kernel", nmi_wd_lpm_factor_ctl_table);
+
+	return 0;
+}
+device_initcall(register_nmi_wd_lpm_factor_sysctl);
+#endif /* CONFIG_SYSCTL */
+#endif /* CONFIG_PPC_WATCHDOG */
+
 static int mobility_rtas_call(int token, char *buf, s32 scope)
 {
 	int rc;
@@ -63,6 +90,27 @@ static int mobility_rtas_call(int token, char *buf, s32 scope)
 
 static int delete_dt_node(struct device_node *dn)
 {
+	struct device_node *pdn;
+	bool is_platfac;
+
+	pdn = of_get_parent(dn);
+	is_platfac = of_node_is_type(dn, "ibm,platform-facilities") ||
+		     of_node_is_type(pdn, "ibm,platform-facilities");
+	of_node_put(pdn);
+
+	/*
+	 * The drivers that bind to nodes in the platform-facilities
+	 * hierarchy don't support node removal, and the removal directive
+	 * from firmware is always followed by an add of an equivalent
+	 * node. The capability (e.g. RNG, encryption, compression)
+	 * represented by the node is never interrupted by the migration.
+	 * So ignore changes to this part of the tree.
+	 */
+	if (is_platfac) {
+		pr_notice("ignoring remove operation for %pOFfp\n", dn);
+		return 0;
+	}
+
 	pr_debug("removing node %pOFfp\n", dn);
 	dlpar_detach_node(dn);
 	return 0;
@@ -140,7 +188,7 @@ static int update_dt_node(struct device_node *dn, s32 scope)
 	u32 nprops;
 	u32 vd;
 
-	update_properties_token = rtas_token("ibm,update-properties");
+	update_properties_token = rtas_function_token(RTAS_FN_IBM_UPDATE_PROPERTIES);
 	if (update_properties_token == RTAS_UNKNOWN_SERVICE)
 		return -EINVAL;
 
@@ -161,7 +209,7 @@ static int update_dt_node(struct device_node *dn, s32 scope)
 		nprops = be32_to_cpu(upwa->nprops);
 
 		/* On the first call to ibm,update-properties for a node the
-		 * the first property value descriptor contains an empty
+		 * first property value descriptor contains an empty
 		 * property name, the property value length encoded as u32,
 		 * and the property value is the node path being updated.
 		 */
@@ -222,6 +270,19 @@ static int add_dt_node(struct device_node *parent_dn, __be32 drc_index)
 	if (!dn)
 		return -ENOENT;
 
+	/*
+	 * Since delete_dt_node() ignores this node type, this is the
+	 * necessary counterpart. We also know that a platform-facilities
+	 * node returned from dlpar_configure_connector() has children
+	 * attached, and dlpar_attach_node() only adds the parent, leaking
+	 * the children. So ignore these on the add side for now.
+	 */
+	if (of_node_is_type(dn, "ibm,platform-facilities")) {
+		pr_notice("ignoring add operation for %pOF\n", dn);
+		dlpar_free_cc_nodes(dn);
+		return 0;
+	}
+
 	rc = dlpar_attach_node(dn, parent_dn);
 	if (rc)
 		dlpar_free_cc_nodes(dn);
@@ -231,14 +292,14 @@ static int add_dt_node(struct device_node *parent_dn, __be32 drc_index)
 	return rc;
 }
 
-int pseries_devicetree_update(s32 scope)
+static int pseries_devicetree_update(s32 scope)
 {
 	char *rtas_buf;
 	__be32 *data;
 	int update_nodes_token;
 	int rc;
 
-	update_nodes_token = rtas_token("ibm,update-nodes");
+	update_nodes_token = rtas_function_token(RTAS_FN_IBM_UPDATE_NODES);
 	if (update_nodes_token == RTAS_UNKNOWN_SERVICE)
 		return 0;
 
@@ -392,6 +453,43 @@ static int wait_for_vasi_session_suspending(u64 handle)
 	return ret;
 }
 
+static void wait_for_vasi_session_completed(u64 handle)
+{
+	unsigned long state = 0;
+	int ret;
+
+	pr_info("waiting for memory transfer to complete...\n");
+
+	/*
+	 * Wait for transition from H_VASI_RESUMED to H_VASI_COMPLETED.
+	 */
+	while (true) {
+		ret = poll_vasi_state(handle, &state);
+
+		/*
+		 * If the memory transfer is already complete and the migration
+		 * has been cleaned up by the hypervisor, H_PARAMETER is return,
+		 * which is translate in EINVAL by poll_vasi_state().
+		 */
+		if (ret == -EINVAL || (!ret && state == H_VASI_COMPLETED)) {
+			pr_info("memory transfer completed.\n");
+			break;
+		}
+
+		if (ret) {
+			pr_err("H_VASI_STATE return error (%d)\n", ret);
+			break;
+		}
+
+		if (state != H_VASI_RESUMED) {
+			pr_err("unexpected H_VASI_STATE result %lu\n", state);
+			break;
+		}
+
+		msleep(500);
+	}
+}
+
 static void prod_single(unsigned int target_cpu)
 {
 	long hvrc;
@@ -417,11 +515,15 @@ static void prod_others(void)
 
 static u16 clamp_slb_size(void)
 {
+#ifdef CONFIG_PPC_64S_HASH_MMU
 	u16 prev = mmu_slb_size;
 
 	slb_set_size(SLB_MIN_SIZE);
 
 	return prev;
+#else
+	return 0;
+#endif
 }
 
 static int do_suspend(void)
@@ -526,10 +628,13 @@ retry:
 		prod_others();
 	}
 	/*
-	 * Execution may have been suspended for several seconds, so
-	 * reset the watchdog.
+	 * Execution may have been suspended for several seconds, so reset
+	 * the watchdogs. touch_nmi_watchdog() also touches the soft lockup
+	 * watchdog.
 	 */
+	rcu_cpu_stall_reset();
 	touch_nmi_watchdog();
+
 	return ret;
 }
 
@@ -626,17 +731,49 @@ static int pseries_suspend(u64 handle)
 static int pseries_migrate_partition(u64 handle)
 {
 	int ret;
+	unsigned int factor = 0;
+
+#ifdef CONFIG_PPC_WATCHDOG
+	factor = nmi_wd_lpm_factor;
+#endif
+	/*
+	 * When the migration is initiated, the hypervisor changes VAS
+	 * mappings to prepare before OS gets the notification and
+	 * closes all VAS windows. NX generates continuous faults during
+	 * this time and the user space can not differentiate these
+	 * faults from the migration event. So reduce this time window
+	 * by closing VAS windows at the beginning of this function.
+	 */
+	vas_migration_handler(VAS_SUSPEND);
+	hvpipe_migration_handler(HVPIPE_SUSPEND);
 
 	ret = wait_for_vasi_session_suspending(handle);
 	if (ret)
-		return ret;
+		goto out;
+
+	if (factor)
+		watchdog_hardlockup_set_timeout_pct(factor);
 
 	ret = pseries_suspend(handle);
-	if (ret == 0)
+	if (ret == 0) {
 		post_mobility_fixup();
-	else
+		/*
+		 * Wait until the memory transfer is complete, so that the user
+		 * space process returns from the syscall after the transfer is
+		 * complete. This allows the user hooks to be executed at the
+		 * right time.
+		 */
+		wait_for_vasi_session_completed(handle);
+	} else
 		pseries_cancel_migration(handle, ret);
 
+	if (factor)
+		watchdog_hardlockup_set_timeout_pct(0);
+
+out:
+	vas_migration_handler(VAS_RESUME);
+	hvpipe_migration_handler(HVPIPE_RESUME);
+
 	return ret;
 }
 
@@ -645,8 +782,8 @@ int rtas_syscall_dispatch_ibm_suspend_me(u64 handle)
 	return pseries_migrate_partition(handle);
 }
 
-static ssize_t migration_store(struct class *class,
-			       struct class_attribute *attr, const char *buf,
+static ssize_t migration_store(const struct class *class,
+			       const struct class_attribute *attr, const char *buf,
 			       size_t count)
 {
 	u64 streamid;