175eb3d9bSMahesh Salgaonkar /*
2c28237f1SMarkus Elfring  * OPAL asynchronus Memory error handling support in PowerNV.
375eb3d9bSMahesh Salgaonkar  *
475eb3d9bSMahesh Salgaonkar  * This program is free software; you can redistribute it and/or modify
575eb3d9bSMahesh Salgaonkar  * it under the terms of the GNU General Public License as published by
675eb3d9bSMahesh Salgaonkar  * the Free Software Foundation; either version 2 of the License, or
775eb3d9bSMahesh Salgaonkar  * (at your option) any later version.
875eb3d9bSMahesh Salgaonkar  *
975eb3d9bSMahesh Salgaonkar  * This program is distributed in the hope that it will be useful,
1075eb3d9bSMahesh Salgaonkar  * but WITHOUT ANY WARRANTY; without even the implied warranty of
1175eb3d9bSMahesh Salgaonkar  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
1275eb3d9bSMahesh Salgaonkar  * GNU General Public License for more details.
1375eb3d9bSMahesh Salgaonkar  *
1475eb3d9bSMahesh Salgaonkar  * You should have received a copy of the GNU General Public License
1575eb3d9bSMahesh Salgaonkar  * along with this program; if not, write to the Free Software
1675eb3d9bSMahesh Salgaonkar  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
1775eb3d9bSMahesh Salgaonkar  *
1875eb3d9bSMahesh Salgaonkar  * Copyright 2013 IBM Corporation
1975eb3d9bSMahesh Salgaonkar  * Author: Mahesh Salgaonkar <mahesh@linux.vnet.ibm.com>
2075eb3d9bSMahesh Salgaonkar  */
2175eb3d9bSMahesh Salgaonkar 
2275eb3d9bSMahesh Salgaonkar #undef DEBUG
2375eb3d9bSMahesh Salgaonkar 
2475eb3d9bSMahesh Salgaonkar #include <linux/kernel.h>
2575eb3d9bSMahesh Salgaonkar #include <linux/init.h>
2675eb3d9bSMahesh Salgaonkar #include <linux/of.h>
2775eb3d9bSMahesh Salgaonkar #include <linux/mm.h>
2875eb3d9bSMahesh Salgaonkar #include <linux/slab.h>
2975eb3d9bSMahesh Salgaonkar 
30b14726c5SMichael Ellerman #include <asm/machdep.h>
3175eb3d9bSMahesh Salgaonkar #include <asm/opal.h>
3275eb3d9bSMahesh Salgaonkar #include <asm/cputable.h>
3375eb3d9bSMahesh Salgaonkar 
3475eb3d9bSMahesh Salgaonkar static int opal_mem_err_nb_init;
3575eb3d9bSMahesh Salgaonkar static LIST_HEAD(opal_memory_err_list);
3675eb3d9bSMahesh Salgaonkar static DEFINE_SPINLOCK(opal_mem_err_lock);
3775eb3d9bSMahesh Salgaonkar 
3875eb3d9bSMahesh Salgaonkar struct OpalMsgNode {
3975eb3d9bSMahesh Salgaonkar 	struct list_head list;
4075eb3d9bSMahesh Salgaonkar 	struct opal_msg msg;
4175eb3d9bSMahesh Salgaonkar };
4275eb3d9bSMahesh Salgaonkar 
4375eb3d9bSMahesh Salgaonkar static void handle_memory_error_event(struct OpalMemoryErrorData *merr_evt)
4475eb3d9bSMahesh Salgaonkar {
4575eb3d9bSMahesh Salgaonkar 	uint64_t paddr_start, paddr_end;
4675eb3d9bSMahesh Salgaonkar 
476e8a9279SColin Ian King 	pr_debug("%s: Retrieved memory error event, type: 0x%x\n",
4875eb3d9bSMahesh Salgaonkar 		  __func__, merr_evt->type);
4975eb3d9bSMahesh Salgaonkar 	switch (merr_evt->type) {
5075eb3d9bSMahesh Salgaonkar 	case OPAL_MEM_ERR_TYPE_RESILIENCE:
51223ca9d8SAnton Blanchard 		paddr_start = be64_to_cpu(merr_evt->u.resilience.physical_address_start);
52223ca9d8SAnton Blanchard 		paddr_end = be64_to_cpu(merr_evt->u.resilience.physical_address_end);
5375eb3d9bSMahesh Salgaonkar 		break;
5475eb3d9bSMahesh Salgaonkar 	case OPAL_MEM_ERR_TYPE_DYN_DALLOC:
55223ca9d8SAnton Blanchard 		paddr_start = be64_to_cpu(merr_evt->u.dyn_dealloc.physical_address_start);
56223ca9d8SAnton Blanchard 		paddr_end = be64_to_cpu(merr_evt->u.dyn_dealloc.physical_address_end);
5775eb3d9bSMahesh Salgaonkar 		break;
5875eb3d9bSMahesh Salgaonkar 	default:
5975eb3d9bSMahesh Salgaonkar 		return;
6075eb3d9bSMahesh Salgaonkar 	}
6175eb3d9bSMahesh Salgaonkar 
6275eb3d9bSMahesh Salgaonkar 	for (; paddr_start < paddr_end; paddr_start += PAGE_SIZE) {
6383b57531SEric W. Biederman 		memory_failure(paddr_start >> PAGE_SHIFT, 0);
6475eb3d9bSMahesh Salgaonkar 	}
6575eb3d9bSMahesh Salgaonkar }
6675eb3d9bSMahesh Salgaonkar 
6775eb3d9bSMahesh Salgaonkar static void handle_memory_error(void)
6875eb3d9bSMahesh Salgaonkar {
6975eb3d9bSMahesh Salgaonkar 	unsigned long flags;
7075eb3d9bSMahesh Salgaonkar 	struct OpalMemoryErrorData *merr_evt;
7175eb3d9bSMahesh Salgaonkar 	struct OpalMsgNode *msg_node;
7275eb3d9bSMahesh Salgaonkar 
7375eb3d9bSMahesh Salgaonkar 	spin_lock_irqsave(&opal_mem_err_lock, flags);
7475eb3d9bSMahesh Salgaonkar 	while (!list_empty(&opal_memory_err_list)) {
7575eb3d9bSMahesh Salgaonkar 		 msg_node = list_entry(opal_memory_err_list.next,
7675eb3d9bSMahesh Salgaonkar 					   struct OpalMsgNode, list);
7775eb3d9bSMahesh Salgaonkar 		list_del(&msg_node->list);
7875eb3d9bSMahesh Salgaonkar 		spin_unlock_irqrestore(&opal_mem_err_lock, flags);
7975eb3d9bSMahesh Salgaonkar 
8075eb3d9bSMahesh Salgaonkar 		merr_evt = (struct OpalMemoryErrorData *)
8175eb3d9bSMahesh Salgaonkar 					&msg_node->msg.params[0];
8275eb3d9bSMahesh Salgaonkar 		handle_memory_error_event(merr_evt);
8375eb3d9bSMahesh Salgaonkar 		kfree(msg_node);
8475eb3d9bSMahesh Salgaonkar 		spin_lock_irqsave(&opal_mem_err_lock, flags);
8575eb3d9bSMahesh Salgaonkar 	}
8675eb3d9bSMahesh Salgaonkar 	spin_unlock_irqrestore(&opal_mem_err_lock, flags);
8775eb3d9bSMahesh Salgaonkar }
8875eb3d9bSMahesh Salgaonkar 
8975eb3d9bSMahesh Salgaonkar static void mem_error_handler(struct work_struct *work)
9075eb3d9bSMahesh Salgaonkar {
9175eb3d9bSMahesh Salgaonkar 	handle_memory_error();
9275eb3d9bSMahesh Salgaonkar }
9375eb3d9bSMahesh Salgaonkar 
9475eb3d9bSMahesh Salgaonkar static DECLARE_WORK(mem_error_work, mem_error_handler);
9575eb3d9bSMahesh Salgaonkar 
9675eb3d9bSMahesh Salgaonkar /*
9775eb3d9bSMahesh Salgaonkar  * opal_memory_err_event - notifier handler that queues up the opal message
9875eb3d9bSMahesh Salgaonkar  * to be preocessed later.
9975eb3d9bSMahesh Salgaonkar  */
10075eb3d9bSMahesh Salgaonkar static int opal_memory_err_event(struct notifier_block *nb,
10175eb3d9bSMahesh Salgaonkar 			  unsigned long msg_type, void *msg)
10275eb3d9bSMahesh Salgaonkar {
10375eb3d9bSMahesh Salgaonkar 	unsigned long flags;
10475eb3d9bSMahesh Salgaonkar 	struct OpalMsgNode *msg_node;
10575eb3d9bSMahesh Salgaonkar 
10675eb3d9bSMahesh Salgaonkar 	if (msg_type != OPAL_MSG_MEM_ERR)
10775eb3d9bSMahesh Salgaonkar 		return 0;
10875eb3d9bSMahesh Salgaonkar 
10975eb3d9bSMahesh Salgaonkar 	msg_node = kzalloc(sizeof(*msg_node), GFP_ATOMIC);
11075eb3d9bSMahesh Salgaonkar 	if (!msg_node) {
11175eb3d9bSMahesh Salgaonkar 		pr_err("MEMORY_ERROR: out of memory, Opal message event not"
11275eb3d9bSMahesh Salgaonkar 		       "handled\n");
11375eb3d9bSMahesh Salgaonkar 		return -ENOMEM;
11475eb3d9bSMahesh Salgaonkar 	}
115a0828cf5SMarkus Elfring 	memcpy(&msg_node->msg, msg, sizeof(msg_node->msg));
11675eb3d9bSMahesh Salgaonkar 
11775eb3d9bSMahesh Salgaonkar 	spin_lock_irqsave(&opal_mem_err_lock, flags);
11875eb3d9bSMahesh Salgaonkar 	list_add(&msg_node->list, &opal_memory_err_list);
11975eb3d9bSMahesh Salgaonkar 	spin_unlock_irqrestore(&opal_mem_err_lock, flags);
12075eb3d9bSMahesh Salgaonkar 
12175eb3d9bSMahesh Salgaonkar 	schedule_work(&mem_error_work);
12275eb3d9bSMahesh Salgaonkar 	return 0;
12375eb3d9bSMahesh Salgaonkar }
12475eb3d9bSMahesh Salgaonkar 
12575eb3d9bSMahesh Salgaonkar static struct notifier_block opal_mem_err_nb = {
12675eb3d9bSMahesh Salgaonkar 	.notifier_call	= opal_memory_err_event,
12775eb3d9bSMahesh Salgaonkar 	.next		= NULL,
12875eb3d9bSMahesh Salgaonkar 	.priority	= 0,
12975eb3d9bSMahesh Salgaonkar };
13075eb3d9bSMahesh Salgaonkar 
13175eb3d9bSMahesh Salgaonkar static int __init opal_mem_err_init(void)
13275eb3d9bSMahesh Salgaonkar {
13375eb3d9bSMahesh Salgaonkar 	int ret;
13475eb3d9bSMahesh Salgaonkar 
13575eb3d9bSMahesh Salgaonkar 	if (!opal_mem_err_nb_init) {
13675eb3d9bSMahesh Salgaonkar 		ret = opal_message_notifier_register(
13775eb3d9bSMahesh Salgaonkar 					OPAL_MSG_MEM_ERR, &opal_mem_err_nb);
13875eb3d9bSMahesh Salgaonkar 		if (ret) {
13975eb3d9bSMahesh Salgaonkar 			pr_err("%s: Can't register OPAL event notifier (%d)\n",
14075eb3d9bSMahesh Salgaonkar 			       __func__, ret);
14175eb3d9bSMahesh Salgaonkar 			return ret;
14275eb3d9bSMahesh Salgaonkar 		}
14375eb3d9bSMahesh Salgaonkar 		opal_mem_err_nb_init = 1;
14475eb3d9bSMahesh Salgaonkar 	}
14575eb3d9bSMahesh Salgaonkar 	return 0;
14675eb3d9bSMahesh Salgaonkar }
14796e023e7SAlistair Popple machine_device_initcall(powernv, opal_mem_err_init);
148