[deliverable/linux.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)");

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 (mlx4_is_master(dev))
		mlx4_report_internal_err_comm_event(dev);

	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;
}

static int mlx4_reset_slave(struct mlx4_dev *dev)
{
#define COM_CHAN_RST_REQ_OFFSET 0x10
#define COM_CHAN_RST_ACK_OFFSET 0x08

	u32 comm_flags;
	u32 rst_req;
	u32 rst_ack;
	unsigned long end;
	struct mlx4_priv *priv = mlx4_priv(dev);

	if (pci_channel_offline(dev->persist->pdev))
		return 0;

	comm_flags = swab32(readl((__iomem char *)priv->mfunc.comm +
				  MLX4_COMM_CHAN_FLAGS));
	if (comm_flags == 0xffffffff) {
		mlx4_err(dev, "VF reset is not needed\n");
		return 0;
	}

	if (!(dev->caps.vf_caps & MLX4_VF_CAP_FLAG_RESET)) {
		mlx4_err(dev, "VF reset is not supported\n");
		return -EOPNOTSUPP;
	}

	rst_req = (comm_flags & (u32)(1 << COM_CHAN_RST_REQ_OFFSET)) >>
		COM_CHAN_RST_REQ_OFFSET;
	rst_ack = (comm_flags & (u32)(1 << COM_CHAN_RST_ACK_OFFSET)) >>
		COM_CHAN_RST_ACK_OFFSET;
	if (rst_req != rst_ack) {
		mlx4_err(dev, "Communication channel isn't sync, fail to send reset\n");
		return -EIO;
	}

	rst_req ^= 1;
	mlx4_warn(dev, "VF is sending reset request to Firmware\n");
	comm_flags = rst_req << COM_CHAN_RST_REQ_OFFSET;
	__raw_writel((__force u32)cpu_to_be32(comm_flags),
		     (__iomem char *)priv->mfunc.comm + MLX4_COMM_CHAN_FLAGS);
	/* Make sure that our comm channel write doesn't
	 * get mixed in with writes from another CPU.
	 */
	mmiowb();

	end = msecs_to_jiffies(MLX4_COMM_TIME) + jiffies;
	while (time_before(jiffies, end)) {
		comm_flags = swab32(readl((__iomem char *)priv->mfunc.comm +
					  MLX4_COMM_CHAN_FLAGS));
		rst_ack = (comm_flags & (u32)(1 << COM_CHAN_RST_ACK_OFFSET)) >>
			COM_CHAN_RST_ACK_OFFSET;

		/* Reading rst_req again since the communication channel can
		 * be reset at any time by the PF and all its bits will be
		 * set to zero.
		 */
		rst_req = (comm_flags & (u32)(1 << COM_CHAN_RST_REQ_OFFSET)) >>
			COM_CHAN_RST_REQ_OFFSET;

		if (rst_ack == rst_req) {
			mlx4_warn(dev, "VF Reset succeed\n");
			return 0;
		}
		cond_resched();
	}
	mlx4_err(dev, "Fail to send reset over the communication channel\n");
	return -ETIMEDOUT;
}

static int mlx4_comm_internal_err(u32 slave_read)
{
	return (u32)COMM_CHAN_EVENT_INTERNAL_ERR ==
		(slave_read & (u32)COMM_CHAN_EVENT_INTERNAL_ERR) ? 1 : 0;
}

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");
	if (mlx4_is_slave(dev))
		err = mlx4_reset_slave(dev);
	else
		err = mlx4_reset_master(dev);

	if (!err) {
		mlx4_err(dev, "device was reset successfully\n");
	} else {
		/* EEH could have disabled the PCI channel during reset. That's
		 * recoverable and the PCI error flow will handle it.
		 */
		if (!pci_channel_offline(dev->persist->pdev))
			BUG_ON(1);
	}
	dev->persist->state |= MLX4_DEVICE_STATE_INTERNAL_ERROR;
	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);
	u32 slave_read;

	if (mlx4_is_slave(dev)) {
		slave_read = swab32(readl(&priv->mfunc.comm->slave_read));
		if (mlx4_comm_internal_err(slave_read)) {
			mlx4_warn(dev, "Internal error detected on the communication channel\n");
			goto internal_err;
		}
	} else 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;

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

	if (!mlx4_is_slave(dev)) {
		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,
55ad3592	48	"Reset device on internal errors if non-zero (default 1)");
ee49bd93	49
f6bc11e4 YH	50	static int read_vendor_id(struct mlx4_dev *dev)
	51	{
	52	u16 vendor_id = 0;
	53	int ret;
	54
	55	ret = pci_read_config_word(dev->persist->pdev, 0, &vendor_id);
	56	if (ret) {
	57	mlx4_err(dev, "Failed to read vendor ID, ret=%d\n", ret);
	58	return ret;
	59	}
	60
	61	if (vendor_id == 0xffff) {
	62	mlx4_err(dev, "PCI can't be accessed to read vendor id\n");
	63	return -EINVAL;
	64	}
	65
	66	return 0;
	67	}
	68
	69	static int mlx4_reset_master(struct mlx4_dev *dev)
	70	{
	71	int err = 0;
	72
55ad3592 YH	73	if (mlx4_is_master(dev))
	74	mlx4_report_internal_err_comm_event(dev);
	75
f6bc11e4 YH	76	if (!pci_channel_offline(dev->persist->pdev)) {
	77	err = read_vendor_id(dev);
	78	/* If PCI can't be accessed to read vendor ID we assume that its
	79	* link was disabled and chip was already reset.
	80	*/
	81	if (err)
	82	return 0;
	83
	84	err = mlx4_reset(dev);
	85	if (err)
	86	mlx4_err(dev, "Fail to reset HCA\n");
	87	}
	88
	89	return err;
	90	}
	91
55ad3592 YH	92	static int mlx4_reset_slave(struct mlx4_dev *dev)
	93	{
	94	#define COM_CHAN_RST_REQ_OFFSET 0x10
	95	#define COM_CHAN_RST_ACK_OFFSET 0x08
	96
	97	u32 comm_flags;
	98	u32 rst_req;
	99	u32 rst_ack;
	100	unsigned long end;
	101	struct mlx4_priv *priv = mlx4_priv(dev);
	102
	103	if (pci_channel_offline(dev->persist->pdev))
	104	return 0;
	105
	106	comm_flags = swab32(readl((__iomem char *)priv->mfunc.comm +
	107	MLX4_COMM_CHAN_FLAGS));
	108	if (comm_flags == 0xffffffff) {
	109	mlx4_err(dev, "VF reset is not needed\n");
	110	return 0;
	111	}
	112
	113	if (!(dev->caps.vf_caps & MLX4_VF_CAP_FLAG_RESET)) {
	114	mlx4_err(dev, "VF reset is not supported\n");
	115	return -EOPNOTSUPP;
	116	}
	117
	118	rst_req = (comm_flags & (u32)(1 << COM_CHAN_RST_REQ_OFFSET)) >>
	119	COM_CHAN_RST_REQ_OFFSET;
	120	rst_ack = (comm_flags & (u32)(1 << COM_CHAN_RST_ACK_OFFSET)) >>
	121	COM_CHAN_RST_ACK_OFFSET;
	122	if (rst_req != rst_ack) {
	123	mlx4_err(dev, "Communication channel isn't sync, fail to send reset\n");
	124	return -EIO;
	125	}
	126
	127	rst_req ^= 1;
	128	mlx4_warn(dev, "VF is sending reset request to Firmware\n");
	129	comm_flags = rst_req << COM_CHAN_RST_REQ_OFFSET;
	130	__raw_writel((__force u32)cpu_to_be32(comm_flags),
	131	(__iomem char *)priv->mfunc.comm + MLX4_COMM_CHAN_FLAGS);
	132	/* Make sure that our comm channel write doesn't
	133	* get mixed in with writes from another CPU.
	134	*/
	135	mmiowb();
	136
	137	end = msecs_to_jiffies(MLX4_COMM_TIME) + jiffies;
	138	while (time_before(jiffies, end)) {
	139	comm_flags = swab32(readl((__iomem char *)priv->mfunc.comm +
	140	MLX4_COMM_CHAN_FLAGS));
	141	rst_ack = (comm_flags & (u32)(1 << COM_CHAN_RST_ACK_OFFSET)) >>
	142	COM_CHAN_RST_ACK_OFFSET;
	143
	144	/* Reading rst_req again since the communication channel can
	145	* be reset at any time by the PF and all its bits will be
	146	* set to zero.
	147	*/
	148	rst_req = (comm_flags & (u32)(1 << COM_CHAN_RST_REQ_OFFSET)) >>
	149	COM_CHAN_RST_REQ_OFFSET;
	150
	151	if (rst_ack == rst_req) {
	152	mlx4_warn(dev, "VF Reset succeed\n");
	153	return 0;
	154	}
	155	cond_resched();
156	}
157	mlx4_err(dev, "Fail to send reset over the communication channel\n");
158	return -ETIMEDOUT;
159	}
160
161	static int mlx4_comm_internal_err(u32 slave_read)
162	{
163	return (u32)COMM_CHAN_EVENT_INTERNAL_ERR ==
164	(slave_read & (u32)COMM_CHAN_EVENT_INTERNAL_ERR) ? 1 : 0;
165	}
166
f6bc11e4 YH	167	void mlx4_enter_error_state(struct mlx4_dev_persistent *persist)
	168	{
	169	int err;
	170	struct mlx4_dev *dev;
	171
f5aef5aa	172	if (!mlx4_internal_err_reset)
f6bc11e4 YH	173	return;
	174
	175	mutex_lock(&persist->device_state_mutex);
	176	if (persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR)
	177	goto out;
	178
	179	dev = persist->dev;
	180	mlx4_err(dev, "device is going to be reset\n");
55ad3592 YH	181	if (mlx4_is_slave(dev))
	182	err = mlx4_reset_slave(dev);
	183	else
	184	err = mlx4_reset_master(dev);
f6bc11e4	185
22e3817e DJ	186	if (!err) {
	187	mlx4_err(dev, "device was reset successfully\n");
	188	} else {
	189	/* EEH could have disabled the PCI channel during reset. That's
	190	* recoverable and the PCI error flow will handle it.
	191	*/
	192	if (!pci_channel_offline(dev->persist->pdev))
	193	BUG_ON(1);
	194	}
f6bc11e4	195	dev->persist->state \|= MLX4_DEVICE_STATE_INTERNAL_ERROR;
f6bc11e4 YH	196	mutex_unlock(&persist->device_state_mutex);
	197
	198	/* At that step HW was already reset, now notify clients */
	199	mlx4_dispatch_event(dev, MLX4_DEV_EVENT_CATASTROPHIC_ERROR, 0);
f5aef5aa	200	mlx4_cmd_wake_completions(dev);
f6bc11e4 YH	201	return;
	202
	203	out:
	204	mutex_unlock(&persist->device_state_mutex);
	205	}
	206
	207	static void mlx4_handle_error_state(struct mlx4_dev_persistent *persist)
	208	{
	209	int err = 0;
	210
	211	mlx4_enter_error_state(persist);
c69453e2 YH	212	mutex_lock(&persist->interface_state_mutex);
	213	if (persist->interface_state & MLX4_INTERFACE_STATE_UP &&
	214	!(persist->interface_state & MLX4_INTERFACE_STATE_DELETION)) {
	215	err = mlx4_restart_one(persist->pdev);
	216	mlx4_info(persist->dev, "mlx4_restart_one was ended, ret=%d\n",
	217	err);
	218	}
	219	mutex_unlock(&persist->interface_state_mutex);
f6bc11e4 YH	220	}
f6bc11e4 YH	221
ee49bd93	222	static void dump_err_buf(struct mlx4_dev *dev)
225c7b1f RD	223	{
	224	struct mlx4_priv *priv = mlx4_priv(dev);
	225
	226	int i;
	227
ee49bd93	228	mlx4_err(dev, "Internal error detected:\n");
225c7b1f RD	229	for (i = 0; i < priv->fw.catas_size; ++i)
	230	mlx4_err(dev, " buf[%02x]: %08x\n",
	231	i, swab32(readl(priv->catas_err.map + i)));
ee49bd93	232	}
225c7b1f	233
ee49bd93 JM	234	static void poll_catas(unsigned long dev_ptr)
	235	{
	236	struct mlx4_dev dev = (struct mlx4_dev ) dev_ptr;
	237	struct mlx4_priv *priv = mlx4_priv(dev);
55ad3592 YH	238	u32 slave_read;
	239
	240	if (mlx4_is_slave(dev)) {
	241	slave_read = swab32(readl(&priv->mfunc.comm->slave_read));
	242	if (mlx4_comm_internal_err(slave_read)) {
	243	mlx4_warn(dev, "Internal error detected on the communication channel\n");
	244	goto internal_err;
	245	}
	246	} else if (readl(priv->catas_err.map)) {
f6bc11e4 YH	247	dump_err_buf(dev);
	248	goto internal_err;
	249	}
	250
	251	if (dev->persist->state & MLX4_DEVICE_STATE_INTERNAL_ERROR) {
	252	mlx4_warn(dev, "Internal error mark was detected on device\n");
	253	goto internal_err;
	254	}
	255
	256	mod_timer(&priv->catas_err.timer,
	257	round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL));
	258	return;
	259
	260	internal_err:
f5aef5aa	261	if (mlx4_internal_err_reset)
f6bc11e4	262	queue_work(dev->persist->catas_wq, &dev->persist->catas_work);
225c7b1f RD	263	}
225c7b1f RD	264
ee49bd93 JM	265	static void catas_reset(struct work_struct *work)
ee49bd93 JM	266	{
ad9a0bf0 YH	267	struct mlx4_dev_persistent *persist =
	268	container_of(work, struct mlx4_dev_persistent,
	269	catas_work);
57dbf29a	270
f6bc11e4	271	mlx4_handle_error_state(persist);
ee49bd93 JM	272	}
	273
	274	void mlx4_start_catas_poll(struct mlx4_dev *dev)
225c7b1f RD	275	{
225c7b1f RD	276	struct mlx4_priv *priv = mlx4_priv(dev);
4979d18f	277	phys_addr_t addr;
225c7b1f	278
ee49bd93 JM	279	INIT_LIST_HEAD(&priv->catas_err.list);
	280	init_timer(&priv->catas_err.timer);
	281	priv->catas_err.map = NULL;
	282
55ad3592 YH	283	if (!mlx4_is_slave(dev)) {
	284	addr = pci_resource_start(dev->persist->pdev,
	285	priv->fw.catas_bar) +
	286	priv->fw.catas_offset;
	287
	288	priv->catas_err.map = ioremap(addr, priv->fw.catas_size * 4);
	289	if (!priv->catas_err.map) {
	290	mlx4_warn(dev, "Failed to map internal error buffer at 0x%llx\n",
	291	(unsigned long long)addr);
	292	return;
	293	}
ee49bd93	294	}
225c7b1f	295
ee49bd93 JM	296	priv->catas_err.timer.data = (unsigned long) dev;
	297	priv->catas_err.timer.function = poll_catas;
	298	priv->catas_err.timer.expires =
	299	round_jiffies(jiffies + MLX4_CATAS_POLL_INTERVAL);
	300	add_timer(&priv->catas_err.timer);
225c7b1f RD	301	}
225c7b1f RD	302
ee49bd93	303	void mlx4_stop_catas_poll(struct mlx4_dev *dev)
225c7b1f RD	304	{
	305	struct mlx4_priv *priv = mlx4_priv(dev);
	306
ee49bd93 JM	307	del_timer_sync(&priv->catas_err.timer);
ee49bd93 JM	308
ad9a0bf0	309	if (priv->catas_err.map) {
225c7b1f	310	iounmap(priv->catas_err.map);
ad9a0bf0 YH	311	priv->catas_err.map = NULL;
ad9a0bf0 YH	312	}
c69453e2 YH	313
	314	if (dev->persist->interface_state & MLX4_INTERFACE_STATE_DELETION)
	315	flush_workqueue(dev->persist->catas_wq);
ad9a0bf0	316	}
ee49bd93	317
ad9a0bf0 YH	318	int mlx4_catas_init(struct mlx4_dev *dev)
	319	{
	320	INIT_WORK(&dev->persist->catas_work, catas_reset);
	321	dev->persist->catas_wq = create_singlethread_workqueue("mlx4_health");
	322	if (!dev->persist->catas_wq)
	323	return -ENOMEM;
	324
	325	return 0;
ee49bd93 JM	326	}
ee49bd93 JM	327
ad9a0bf0	328	void mlx4_catas_end(struct mlx4_dev *dev)
ee49bd93	329	{
ad9a0bf0 YH	330	if (dev->persist->catas_wq) {
	331	destroy_workqueue(dev->persist->catas_wq);
	332	dev->persist->catas_wq = NULL;
	333	}
225c7b1f	334	}