[linux-2.6-block.git] / drivers / net / ethernet / mellanox / mlx4 / catas.c

/*
 * Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
 * Copyright (c) 2007, 2008 Mellanox Technologies. All rights reserved.
 *
 * This software is available to you under a choice of one of two
 * licenses.  You may choose to be licensed under the terms of the GNU
 * General Public License (GPL) Version 2, available from the file
 * COPYING in the main directory of this source tree, or the
 * OpenIB.org BSD license below:
 *
 *     Redistribution and use in source and binary forms, with or
 *     without modification, are permitted provided that the following
 *     conditions are met:
 *
 *      - Redistributions of source code must retain the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer.
 *
 *      - Redistributions in binary form must reproduce the above
 *        copyright notice, this list of conditions and the following
 *        disclaimer in the documentation and/or other materials
 *        provided with the distribution.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
 * SOFTWARE.
 */

#include <linux/workqueue.h>
#include <linux/module.h>

#include "mlx4.h"

enum {
	MLX4_CATAS_POLL_INTERVAL	= 5 * HZ,
};


int mlx4_internal_err_reset = 1;
module_param_named(internal_err_reset, mlx4_internal_err_reset,  int, 0644);
MODULE_PARM_DESC(internal_err_reset,
		 "Reset device on internal errors if non-zero"
		 " (default 1, in SRIOV mode default is 0)");

static int read_vendor_id(struct mlx4_dev *dev)
{
	u16 vendor_id = 0;
	int ret;

	ret = pci_read_config_word(dev->persist->pdev, 0, &vendor_id);
	if (ret) {
		mlx4_err(dev, "Failed to read vendor ID, ret=%d\n", ret);
		return ret;
	}

	if (vendor_id == 0xffff) {
		mlx4_err(dev, "PCI can't be accessed to read vendor id\n");
		return -EINVAL;
	}

	return 0;
}

static int mlx4_reset_master(struct mlx4_dev *dev)
{
	int err = 0;

	if (!pci_channel_offline(dev->persist->pdev)) {
		err = read_vendor_id(dev);
		/* If PCI can't be accessed to read vendor ID we assume that its
		 * link was disabled and chip was already reset.
		 */
		if (err)
			return 0;

		err = mlx4_reset(dev);
		if (err)
			mlx4_err(dev, "Fail to reset HCA\n");
	}

	return err;
}

void mlx4_enter_error_state(struct mlx4_dev_persistent *persist)
{
	int err;
	struct mlx4_dev *dev;

	if (!mlx4_internal_err_reset)
		return;

	mutex_lock(&persist->device_state_mutex);
	if (persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR)
		goto out;

	dev = persist->dev;
	mlx4_err(dev, "device is going to be reset\n");
	err = mlx4_reset_master(dev);
	BUG_ON(err != 0);

	dev->persist->state |= MLX4_DEVICE_STATE_INTERNAL_ERROR;
	mlx4_err(dev, "device was reset successfully\n");
	mutex_unlock(&persist->device_state_mutex);

	/* At that step HW was already reset, now notify clients */
	mlx4_dispatch_event(dev, MLX4_DEV_EVENT_CATASTROPHIC_ERROR, 0);
	mlx4_cmd_wake_completions(dev);
	return;

out:
	mutex_unlock(&persist->device_state_mutex);
}

static void mlx4_handle_error_state(struct mlx4_dev_persistent *persist)
{
	int err = 0;

	mlx4_enter_error_state(persist);
	mutex_lock(&persist->interface_state_mutex);
	if (persist->interface_state & MLX4_INTERFACE_STATE_UP &&
	    !(persist->interface_state & MLX4_INTERFACE_STATE_DELETION)) {
		err = mlx4_restart_one(persist->pdev);
		mlx4_info(persist->dev, "mlx4_restart_one was ended, ret=%d\n",
			  err);
	}
	mutex_unlock(&persist->interface_state_mutex);
}

static void dump_err_buf(struct mlx4_dev *dev)
{
	struct mlx4_priv *priv = mlx4_priv(dev);

	int i;

	mlx4_err(dev, "Internal error detected:\n");
	for (i = 0; i < priv->fw.catas_size; ++i)
		mlx4_err(dev, "  buf[%02x]: %08x\n",
			 i, swab32(readl(priv->catas_err.map + i)));
}

static void poll_catas(unsigned long dev_ptr)
{
	struct mlx4_dev *dev = (struct mlx4_dev *) dev_ptr;
	struct mlx4_priv *priv = mlx4_priv(dev);

	if (readl(priv->catas_err.map)) {
		dump_err_buf(dev);
		goto internal_err;
	}

	if (dev->persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR) {
		mlx4_warn(dev, "Internal error mark was detected on device\n");
		goto internal_err;
	}

	mod_timer(&priv->catas_err.timer,
		  round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL));
	return;

internal_err:
	if (mlx4_internal_err_reset)
		queue_work(dev->persist->catas_wq, &dev->persist->catas_work);
}

static void catas_reset(struct work_struct *work)
{
	struct mlx4_dev_persistent *persist =
		container_of(work, struct mlx4_dev_persistent,
			     catas_work);

	mlx4_handle_error_state(persist);
}

void mlx4_start_catas_poll(struct mlx4_dev *dev)
{
	struct mlx4_priv *priv = mlx4_priv(dev);
	phys_addr_t addr;

	/*If we are in SRIOV the default of the module param must be 0*/
	if (mlx4_is_mfunc(dev))
		mlx4_internal_err_reset = 0;

	INIT_LIST_HEAD(&priv->catas_err.list);
	init_timer(&priv->catas_err.timer);
	priv->catas_err.map = NULL;

	addr = pci_resource_start(dev->persist->pdev, priv->fw.catas_bar) +
		priv->fw.catas_offset;

	priv->catas_err.map = ioremap(addr, priv->fw.catas_size * 4);
	if (!priv->catas_err.map) {
		mlx4_warn(dev, "Failed to map internal error buffer at 0x%llx\n",
			  (unsigned long long) addr);
		return;
	}

	priv->catas_err.timer.data     = (unsigned long) dev;
	priv->catas_err.timer.function = poll_catas;
	priv->catas_err.timer.expires  =
		round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL);
	add_timer(&priv->catas_err.timer);
}

void mlx4_stop_catas_poll(struct mlx4_dev *dev)
{
	struct mlx4_priv *priv = mlx4_priv(dev);

	del_timer_sync(&priv->catas_err.timer);

	if (priv->catas_err.map) {
		iounmap(priv->catas_err.map);
		priv->catas_err.map = NULL;
	}

	if (dev->persist->interface_state & MLX4_INTERFACE_STATE_DELETION)
		flush_workqueue(dev->persist->catas_wq);
}

int  mlx4_catas_init(struct mlx4_dev *dev)
{
	INIT_WORK(&dev->persist->catas_work, catas_reset);
	dev->persist->catas_wq = create_singlethread_workqueue("mlx4_health");
	if (!dev->persist->catas_wq)
		return -ENOMEM;

	return 0;
}

void mlx4_catas_end(struct mlx4_dev *dev)
{
	if (dev->persist->catas_wq) {
		destroy_workqueue(dev->persist->catas_wq);
		dev->persist->catas_wq = NULL;
	}
}
Commit	Line	Data
225c7b1f RD	1	/*
225c7b1f RD	2	* Copyright (c) 2007 Cisco Systems, Inc. All rights reserved.
51a379d0	3	* Copyright (c) 2007, 2008 Mellanox Technologies. All rights reserved.
225c7b1f RD	4	*
	5	* This software is available to you under a choice of one of two
	6	* licenses. You may choose to be licensed under the terms of the GNU
	7	* General Public License (GPL) Version 2, available from the file
	8	* COPYING in the main directory of this source tree, or the
	9	* OpenIB.org BSD license below:
	10	*
	11	* Redistribution and use in source and binary forms, with or
	12	* without modification, are permitted provided that the following
	13	* conditions are met:
	14	*
	15	* - Redistributions of source code must retain the above
	16	* copyright notice, this list of conditions and the following
	17	* disclaimer.
	18	*
	19	* - Redistributions in binary form must reproduce the above
	20	* copyright notice, this list of conditions and the following
	21	* disclaimer in the documentation and/or other materials
	22	* provided with the distribution.
	23	*
	24	* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
	25	* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
	26	* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
	27	* NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
	28	* BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
	29	* ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
	30	* CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
	31	* SOFTWARE.
	32	*/
	33
ee49bd93	34	#include <linux/workqueue.h>
9d9779e7	35	#include <linux/module.h>
ee49bd93	36
225c7b1f RD	37	#include "mlx4.h"
225c7b1f RD	38
ee49bd93 JM	39	enum {
	40	MLX4_CATAS_POLL_INTERVAL = 5 * HZ,
	41	};
	42
ee49bd93	43
ee49bd93	44
f5aef5aa YH	45	int mlx4_internal_err_reset = 1;
f5aef5aa YH	46	module_param_named(internal_err_reset, mlx4_internal_err_reset, int, 0644);
ee49bd93	47	MODULE_PARM_DESC(internal_err_reset,
d81c7186 JM	48	"Reset device on internal errors if non-zero"
d81c7186 JM	49	" (default 1, in SRIOV mode default is 0)");
ee49bd93	50
f6bc11e4 YH	51	static int read_vendor_id(struct mlx4_dev *dev)
	52	{
	53	u16 vendor_id = 0;
	54	int ret;
	55
	56	ret = pci_read_config_word(dev->persist->pdev, 0, &vendor_id);
	57	if (ret) {
	58	mlx4_err(dev, "Failed to read vendor ID, ret=%d\n", ret);
	59	return ret;
	60	}
	61
	62	if (vendor_id == 0xffff) {
	63	mlx4_err(dev, "PCI can't be accessed to read vendor id\n");
	64	return -EINVAL;
	65	}
	66
	67	return 0;
	68	}
	69
	70	static int mlx4_reset_master(struct mlx4_dev *dev)
	71	{
	72	int err = 0;
	73
	74	if (!pci_channel_offline(dev->persist->pdev)) {
	75	err = read_vendor_id(dev);
	76	/* If PCI can't be accessed to read vendor ID we assume that its
	77	* link was disabled and chip was already reset.
	78	*/
	79	if (err)
	80	return 0;
	81
	82	err = mlx4_reset(dev);
	83	if (err)
	84	mlx4_err(dev, "Fail to reset HCA\n");
	85	}
	86
	87	return err;
	88	}
	89
	90	void mlx4_enter_error_state(struct mlx4_dev_persistent *persist)
	91	{
	92	int err;
	93	struct mlx4_dev *dev;
	94
f5aef5aa	95	if (!mlx4_internal_err_reset)
f6bc11e4 YH	96	return;
	97
	98	mutex_lock(&persist->device_state_mutex);
	99	if (persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR)
	100	goto out;
	101
	102	dev = persist->dev;
	103	mlx4_err(dev, "device is going to be reset\n");
	104	err = mlx4_reset_master(dev);
	105	BUG_ON(err != 0);
	106
	107	dev->persist->state \|= MLX4_DEVICE_STATE_INTERNAL_ERROR;
	108	mlx4_err(dev, "device was reset successfully\n");
	109	mutex_unlock(&persist->device_state_mutex);
	110
	111	/* At that step HW was already reset, now notify clients */
	112	mlx4_dispatch_event(dev, MLX4_DEV_EVENT_CATASTROPHIC_ERROR, 0);
f5aef5aa	113	mlx4_cmd_wake_completions(dev);
f6bc11e4 YH	114	return;
	115
	116	out:
	117	mutex_unlock(&persist->device_state_mutex);
	118	}
	119
	120	static void mlx4_handle_error_state(struct mlx4_dev_persistent *persist)
	121	{
	122	int err = 0;
	123
	124	mlx4_enter_error_state(persist);
c69453e2 YH	125	mutex_lock(&persist->interface_state_mutex);
	126	if (persist->interface_state & MLX4_INTERFACE_STATE_UP &&
	127	!(persist->interface_state & MLX4_INTERFACE_STATE_DELETION)) {
	128	err = mlx4_restart_one(persist->pdev);
	129	mlx4_info(persist->dev, "mlx4_restart_one was ended, ret=%d\n",
	130	err);
	131	}
	132	mutex_unlock(&persist->interface_state_mutex);
f6bc11e4 YH	133	}
f6bc11e4 YH	134
ee49bd93	135	static void dump_err_buf(struct mlx4_dev *dev)
225c7b1f RD	136	{
	137	struct mlx4_priv *priv = mlx4_priv(dev);
	138
	139	int i;
	140
ee49bd93	141	mlx4_err(dev, "Internal error detected:\n");
225c7b1f RD	142	for (i = 0; i < priv->fw.catas_size; ++i)
	143	mlx4_err(dev, " buf[%02x]: %08x\n",
	144	i, swab32(readl(priv->catas_err.map + i)));
ee49bd93	145	}
225c7b1f	146
ee49bd93 JM	147	static void poll_catas(unsigned long dev_ptr)
	148	{
	149	struct mlx4_dev dev = (struct mlx4_dev ) dev_ptr;
	150	struct mlx4_priv *priv = mlx4_priv(dev);
	151
	152	if (readl(priv->catas_err.map)) {
f6bc11e4 YH	153	dump_err_buf(dev);
	154	goto internal_err;
	155	}
	156
	157	if (dev->persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR) {
	158	mlx4_warn(dev, "Internal error mark was detected on device\n");
	159	goto internal_err;
	160	}
	161
	162	mod_timer(&priv->catas_err.timer,
	163	round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL));
	164	return;
	165
	166	internal_err:
f5aef5aa	167	if (mlx4_internal_err_reset)
f6bc11e4	168	queue_work(dev->persist->catas_wq, &dev->persist->catas_work);
225c7b1f RD	169	}
225c7b1f RD	170
ee49bd93 JM	171	static void catas_reset(struct work_struct *work)
ee49bd93 JM	172	{
ad9a0bf0 YH	173	struct mlx4_dev_persistent *persist =
	174	container_of(work, struct mlx4_dev_persistent,
	175	catas_work);
57dbf29a	176
f6bc11e4	177	mlx4_handle_error_state(persist);
ee49bd93 JM	178	}
	179
	180	void mlx4_start_catas_poll(struct mlx4_dev *dev)
225c7b1f RD	181	{
225c7b1f RD	182	struct mlx4_priv *priv = mlx4_priv(dev);
4979d18f	183	phys_addr_t addr;
225c7b1f	184
d81c7186 JM	185	/If we are in SRIOV the default of the module param must be 0/
d81c7186 JM	186	if (mlx4_is_mfunc(dev))
f5aef5aa	187	mlx4_internal_err_reset = 0;
d81c7186	188
ee49bd93 JM	189	INIT_LIST_HEAD(&priv->catas_err.list);
	190	init_timer(&priv->catas_err.timer);
	191	priv->catas_err.map = NULL;
	192
872bf2fb	193	addr = pci_resource_start(dev->persist->pdev, priv->fw.catas_bar) +
225c7b1f RD	194	priv->fw.catas_offset;
	195
	196	priv->catas_err.map = ioremap(addr, priv->fw.catas_size * 4);
ee49bd93	197	if (!priv->catas_err.map) {
4979d18f RD	198	mlx4_warn(dev, "Failed to map internal error buffer at 0x%llx\n",
4979d18f RD	199	(unsigned long long) addr);
ee49bd93 JM	200	return;
ee49bd93 JM	201	}
225c7b1f	202
ee49bd93 JM	203	priv->catas_err.timer.data = (unsigned long) dev;
	204	priv->catas_err.timer.function = poll_catas;
	205	priv->catas_err.timer.expires =
	206	round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL);
	207	add_timer(&priv->catas_err.timer);
225c7b1f RD	208	}
225c7b1f RD	209
ee49bd93	210	void mlx4_stop_catas_poll(struct mlx4_dev *dev)
225c7b1f RD	211	{
	212	struct mlx4_priv *priv = mlx4_priv(dev);
	213
ee49bd93 JM	214	del_timer_sync(&priv->catas_err.timer);
ee49bd93 JM	215
ad9a0bf0	216	if (priv->catas_err.map) {
225c7b1f	217	iounmap(priv->catas_err.map);
ad9a0bf0 YH	218	priv->catas_err.map = NULL;
ad9a0bf0 YH	219	}
c69453e2 YH	220
	221	if (dev->persist->interface_state & MLX4_INTERFACE_STATE_DELETION)
	222	flush_workqueue(dev->persist->catas_wq);
ad9a0bf0	223	}
ee49bd93	224
ad9a0bf0 YH	225	int mlx4_catas_init(struct mlx4_dev *dev)
	226	{
	227	INIT_WORK(&dev->persist->catas_work, catas_reset);
	228	dev->persist->catas_wq = create_singlethread_workqueue("mlx4_health");
	229	if (!dev->persist->catas_wq)
	230	return -ENOMEM;
	231
	232	return 0;
ee49bd93 JM	233	}
ee49bd93 JM	234
ad9a0bf0	235	void mlx4_catas_end(struct mlx4_dev *dev)
ee49bd93	236	{
ad9a0bf0 YH	237	if (dev->persist->catas_wq) {
	238	destroy_workqueue(dev->persist->catas_wq);
	239	dev->persist->catas_wq = NULL;
	240	}
225c7b1f	241	}