Skip to content
mtip32xx.c 103 KiB
Newer Older
	}
	num_command_slots = dd->slot_groups * 32;

	hba_setup(dd);

	tasklet_init(&dd->tasklet, mtip_tasklet, (unsigned long)dd);

	dd->port = kzalloc(sizeof(struct mtip_port), GFP_KERNEL);
	if (!dd->port) {
		dev_err(&dd->pdev->dev,
			"Memory allocation: port structure\n");
		return -ENOMEM;
	}

	/* Counting semaphore to track command slot usage */
	sema_init(&dd->port->cmd_slot, num_command_slots - 1);

	/* Spinlock to prevent concurrent issue */
	spin_lock_init(&dd->port->cmd_issue_lock);

	/* Set the port mmio base address. */
	dd->port->mmio	= dd->mmio + PORT_OFFSET;
	dd->port->dd	= dd;

	/* Allocate memory for the command list. */
	dd->port->command_list =
		dmam_alloc_coherent(&dd->pdev->dev,
			HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 4),
			&dd->port->command_list_dma,
			GFP_KERNEL);
	if (!dd->port->command_list) {
		dev_err(&dd->pdev->dev,
			"Memory allocation: command list\n");
		rv = -ENOMEM;
		goto out1;
	}

	/* Clear the memory we have allocated. */
	memset(dd->port->command_list,
		0,
		HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 4));

	/* Setup the addresse of the RX FIS. */
	dd->port->rxfis	    = dd->port->command_list + HW_CMD_SLOT_SZ;
	dd->port->rxfis_dma = dd->port->command_list_dma + HW_CMD_SLOT_SZ;

	/* Setup the address of the command tables. */
	dd->port->command_table	  = dd->port->rxfis + AHCI_RX_FIS_SZ;
	dd->port->command_tbl_dma = dd->port->rxfis_dma + AHCI_RX_FIS_SZ;

	/* Setup the address of the identify data. */
	dd->port->identify     = dd->port->command_table +
					HW_CMD_TBL_AR_SZ;
	dd->port->identify_dma = dd->port->command_tbl_dma +
					HW_CMD_TBL_AR_SZ;

	/* Setup the address of the sector buffer - for some non-ncq cmds */
	dd->port->sector_buffer	= (void *) dd->port->identify + ATA_SECT_SIZE;
	dd->port->sector_buffer_dma = dd->port->identify_dma + ATA_SECT_SIZE;

	/* Setup the address of the log buf - for read log command */
	dd->port->log_buf = (void *)dd->port->sector_buffer  + ATA_SECT_SIZE;
	dd->port->log_buf_dma = dd->port->sector_buffer_dma + ATA_SECT_SIZE;

	/* Setup the address of the smart buf - for smart read data command */
	dd->port->smart_buf = (void *)dd->port->log_buf  + ATA_SECT_SIZE;
	dd->port->smart_buf_dma = dd->port->log_buf_dma + ATA_SECT_SIZE;


	/* Point the command headers at the command tables. */
	for (i = 0; i < num_command_slots; i++) {
		dd->port->commands[i].command_header =
					dd->port->command_list +
					(sizeof(struct mtip_cmd_hdr) * i);
		dd->port->commands[i].command_header_dma =
					dd->port->command_list_dma +
					(sizeof(struct mtip_cmd_hdr) * i);

		dd->port->commands[i].command =
			dd->port->command_table + (HW_CMD_TBL_SZ * i);
		dd->port->commands[i].command_dma =
			dd->port->command_tbl_dma + (HW_CMD_TBL_SZ * i);

		if (readl(dd->mmio + HOST_CAP) & HOST_CAP_64)
			dd->port->commands[i].command_header->ctbau =
			__force_bit2int cpu_to_le32(
			(dd->port->commands[i].command_dma >> 16) >> 16);
		dd->port->commands[i].command_header->ctba =
			__force_bit2int cpu_to_le32(
			dd->port->commands[i].command_dma & 0xFFFFFFFF);

		/*
		 * If this is not done, a bug is reported by the stock
		 * FC11 i386. Due to the fact that it has lots of kernel
		 * debugging enabled.
		 */
		sg_init_table(dd->port->commands[i].sg, MTIP_MAX_SG);

		/* Mark all commands as currently inactive.*/
		atomic_set(&dd->port->commands[i].active, 0);
	}

	/* Setup the pointers to the extended s_active and CI registers. */
	for (i = 0; i < dd->slot_groups; i++) {
		dd->port->s_active[i] =
			dd->port->mmio + i*0x80 + PORT_SCR_ACT;
		dd->port->cmd_issue[i] =
			dd->port->mmio + i*0x80 + PORT_COMMAND_ISSUE;
		dd->port->completed[i] =
			dd->port->mmio + i*0x80 + PORT_SDBV;
	}

	timetaken = jiffies;
	timeout = jiffies + msecs_to_jiffies(30000);
	while (((readl(dd->port->mmio + PORT_SCR_STAT) & 0x0F) != 0x03) &&
		 time_before(jiffies, timeout)) {
		mdelay(100);
	}
	if (unlikely(mtip_check_surprise_removal(dd->pdev))) {
		timetaken = jiffies - timetaken;
		dev_warn(&dd->pdev->dev,
			"Surprise removal detected at %u ms\n",
			jiffies_to_msecs(timetaken));
		rv = -ENODEV;
		goto out2 ;
	}
	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag))) {
		timetaken = jiffies - timetaken;
		dev_warn(&dd->pdev->dev,
			"Removal detected at %u ms\n",
			jiffies_to_msecs(timetaken));
		rv = -EFAULT;
	/* Conditionally reset the HBA. */
	if (!(readl(dd->mmio + HOST_CAP) & HOST_CAP_NZDMA)) {
		if (mtip_hba_reset(dd) < 0) {
			dev_err(&dd->pdev->dev,
				"Card did not reset within timeout\n");
			rv = -EIO;
			goto out2;
		}
	} else {
		/* Clear any pending interrupts on the HBA */
		writel(readl(dd->mmio + HOST_IRQ_STAT),
			dd->mmio + HOST_IRQ_STAT);
	}

	mtip_init_port(dd->port);
	mtip_start_port(dd->port);

	/* Setup the ISR and enable interrupts. */
	rv = devm_request_irq(&dd->pdev->dev,
				dd->pdev->irq,
				mtip_irq_handler,
				IRQF_SHARED,
				dev_driver_string(&dd->pdev->dev),
				dd);

	if (rv) {
		dev_err(&dd->pdev->dev,
			"Unable to allocate IRQ %d\n", dd->pdev->irq);
		goto out2;
	}

	/* Enable interrupts on the HBA. */
	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
					dd->mmio + HOST_CTL);

	init_timer(&dd->port->cmd_timer);
	init_waitqueue_head(&dd->port->svc_wait);

	dd->port->cmd_timer.data = (unsigned long int) dd->port;
	dd->port->cmd_timer.function = mtip_timeout_function;
	mod_timer(&dd->port->cmd_timer,
		jiffies + msecs_to_jiffies(MTIP_TIMEOUT_CHECK_PERIOD));

	if (test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)) {
	if (mtip_get_identify(dd->port, NULL) < 0) {
		rv = -EFAULT;
		goto out3;
	}

	if (*(dd->port->identify + MTIP_FTL_REBUILD_OFFSET) ==
		MTIP_FTL_REBUILD_MAGIC) {
		set_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags);
		return MTIP_FTL_REBUILD_MAGIC;
	mtip_dump_identify(dd->port);

	/* check write protect, over temp and rebuild statuses */
	rv = mtip_read_log_page(dd->port, ATA_LOG_SATA_NCQ,
				dd->port->log_buf,
				dd->port->log_buf_dma, 1);
	if (rv) {
		dev_warn(&dd->pdev->dev,
			"Error in READ LOG EXT (10h) command\n");
		/* non-critical error, don't fail the load */
	} else {
		buf = (unsigned char *)dd->port->log_buf;
		if (buf[259] & 0x1) {
			dev_info(&dd->pdev->dev,
				"Write protect bit is set.\n");
			set_bit(MTIP_DDF_WRITE_PROTECT_BIT, &dd->dd_flag);
		}
		if (buf[288] == 0xF7) {
			dev_info(&dd->pdev->dev,
				"Exceeded Tmax, drive in thermal shutdown.\n");
			set_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag);
		}
		if (buf[288] == 0xBF) {
			dev_info(&dd->pdev->dev,
				"Drive indicates rebuild has failed.\n");
			/* TODO */
		}
	}

	/* get write protect progess */
	memset(&attr242, 0, sizeof(struct smart_attr));
	if (mtip_get_smart_attr(dd->port, 242, &attr242))
		dev_warn(&dd->pdev->dev,
				"Unable to check write protect progress\n");
	else
		dev_info(&dd->pdev->dev,
				"Write protect progress: %u%% (%u blocks)\n",
				attr242.cur, le32_to_cpu(attr242.data));
	return rv;

out3:
	del_timer_sync(&dd->port->cmd_timer);

	/* Disable interrupts on the HBA. */
	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
			dd->mmio + HOST_CTL);

	/*Release the IRQ. */
	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);

out2:
	mtip_deinit_port(dd->port);

	/* Free the command/command header memory. */
	dmam_free_coherent(&dd->pdev->dev,
				HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 4),
				dd->port->command_list,
				dd->port->command_list_dma);
out1:
	/* Free the memory allocated for the for structure. */
	kfree(dd->port);

	return rv;
}

/*
 * Called to deinitialize an interface.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0
 */
static int mtip_hw_exit(struct driver_data *dd)
{
	/*
	 * Send standby immediate (E0h) to the drive so that it
	 * saves its state.
	 */
	if (!test_bit(MTIP_DDF_CLEANUP_BIT, &dd->dd_flag)) {
		if (!test_bit(MTIP_PF_REBUILD_BIT, &dd->port->flags))
			if (mtip_standby_immediate(dd->port))
				dev_warn(&dd->pdev->dev,
					"STANDBY IMMEDIATE failed\n");

		/* de-initialize the port. */
		mtip_deinit_port(dd->port);

		/* Disable interrupts on the HBA. */
		writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
				dd->mmio + HOST_CTL);
	}

	del_timer_sync(&dd->port->cmd_timer);

	/* Release the IRQ. */
	devm_free_irq(&dd->pdev->dev, dd->pdev->irq, dd);

	/* Stop the bottom half tasklet. */
	tasklet_kill(&dd->tasklet);

	/* Free the command/command header memory. */
	dmam_free_coherent(&dd->pdev->dev,
			HW_PORT_PRIV_DMA_SZ + (ATA_SECT_SIZE * 4),
			dd->port->command_list,
			dd->port->command_list_dma);
	/* Free the memory allocated for the for structure. */
	kfree(dd->port);

	return 0;
}

/*
 * Issue a Standby Immediate command to the device.
 *
 * This function is called by the Block Layer just before the
 * system powers off during a shutdown.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0
 */
static int mtip_hw_shutdown(struct driver_data *dd)
{
	/*
	 * Send standby immediate (E0h) to the drive so that it
	 * saves its state.
	 */
	mtip_standby_immediate(dd->port);

	return 0;
}

/*
 * Suspend function
 *
 * This function is called by the Block Layer just before the
 * system hibernates.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0	Suspend was successful
 *	-EFAULT Suspend was not successful
 */
static int mtip_hw_suspend(struct driver_data *dd)
{
	/*
	 * Send standby immediate (E0h) to the drive
	 * so that it saves its state.
	 */
	if (mtip_standby_immediate(dd->port) != 0) {
		dev_err(&dd->pdev->dev,
			"Failed standby-immediate command\n");
		return -EFAULT;
	}

	/* Disable interrupts on the HBA.*/
	writel(readl(dd->mmio + HOST_CTL) & ~HOST_IRQ_EN,
			dd->mmio + HOST_CTL);
	mtip_deinit_port(dd->port);

	return 0;
}

/*
 * Resume function
 *
 * This function is called by the Block Layer as the
 * system resumes.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0	Resume was successful
 *      -EFAULT Resume was not successful
 */
static int mtip_hw_resume(struct driver_data *dd)
{
	/* Perform any needed hardware setup steps */
	hba_setup(dd);

	/* Reset the HBA */
	if (mtip_hba_reset(dd) != 0) {
		dev_err(&dd->pdev->dev,
			"Unable to reset the HBA\n");
		return -EFAULT;
	}

	/*
	 * Enable the port, DMA engine, and FIS reception specific
	 * h/w in controller.
	 */
	mtip_init_port(dd->port);
	mtip_start_port(dd->port);

	/* Enable interrupts on the HBA.*/
	writel(readl(dd->mmio + HOST_CTL) | HOST_IRQ_EN,
			dd->mmio + HOST_CTL);

	return 0;
}

/*
 * Helper function for reusing disk name
 * upon hot insertion.
 */
static int rssd_disk_name_format(char *prefix,
				 int index,
				 char *buf,
				 int buflen)
{
	const int base = 'z' - 'a' + 1;
	char *begin = buf + strlen(prefix);
	char *end = buf + buflen;
	char *p;
	int unit;

	p = end - 1;
	*p = '\0';
	unit = base;
	do {
		if (p == begin)
			return -EINVAL;
		*--p = 'a' + (index % unit);
		index = (index / unit) - 1;
	} while (index >= 0);

	memmove(begin, p, end - p);
	memcpy(buf, prefix, strlen(prefix));

	return 0;
}

/*
 * Block layer IOCTL handler.
 *
 * @dev Pointer to the block_device structure.
 * @mode ignored
 * @cmd IOCTL command passed from the user application.
 * @arg Argument passed from the user application.
 *
 * return value
 *	0        IOCTL completed successfully.
 *	-ENOTTY  IOCTL not supported or invalid driver data
 *                 structure pointer.
 */
static int mtip_block_ioctl(struct block_device *dev,
			    fmode_t mode,
			    unsigned cmd,
			    unsigned long arg)
{
	struct driver_data *dd = dev->bd_disk->private_data;

	if (!capable(CAP_SYS_ADMIN))
		return -EACCES;

	if (!dd)
		return -ENOTTY;

	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
	switch (cmd) {
	case BLKFLSBUF:
		return -ENOTTY;
		return mtip_hw_ioctl(dd, cmd, arg);
#ifdef CONFIG_COMPAT
/*
 * Block layer compat IOCTL handler.
 *
 * @dev Pointer to the block_device structure.
 * @mode ignored
 * @cmd IOCTL command passed from the user application.
 * @arg Argument passed from the user application.
 *
 * return value
 *	0        IOCTL completed successfully.
 *	-ENOTTY  IOCTL not supported or invalid driver data
 *                 structure pointer.
 */
static int mtip_block_compat_ioctl(struct block_device *dev,
			    fmode_t mode,
			    unsigned cmd,
			    unsigned long arg)
{
	struct driver_data *dd = dev->bd_disk->private_data;

	if (!capable(CAP_SYS_ADMIN))
		return -EACCES;

	if (!dd)
		return -ENOTTY;

	if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT, &dd->dd_flag)))
	switch (cmd) {
	case BLKFLSBUF:
		return -ENOTTY;
	case HDIO_DRIVE_TASKFILE: {
		struct mtip_compat_ide_task_request_s __user *compat_req_task;
		ide_task_request_t req_task;
		int compat_tasksize, outtotal, ret;

		compat_tasksize =
			sizeof(struct mtip_compat_ide_task_request_s);

		compat_req_task =
			(struct mtip_compat_ide_task_request_s __user *) arg;

		if (copy_from_user(&req_task, (void __user *) arg,
			compat_tasksize - (2 * sizeof(compat_long_t))))
			return -EFAULT;

		if (get_user(req_task.out_size, &compat_req_task->out_size))
			return -EFAULT;

		if (get_user(req_task.in_size, &compat_req_task->in_size))
			return -EFAULT;

		outtotal = sizeof(struct mtip_compat_ide_task_request_s);

		ret = exec_drive_taskfile(dd, (void __user *) arg,
						&req_task, outtotal);

		if (copy_to_user((void __user *) arg, &req_task,
				compat_tasksize -
				(2 * sizeof(compat_long_t))))
			return -EFAULT;

		if (put_user(req_task.out_size, &compat_req_task->out_size))
			return -EFAULT;

		if (put_user(req_task.in_size, &compat_req_task->in_size))
			return -EFAULT;

		return ret;
	}
		return mtip_hw_ioctl(dd, cmd, arg);

/*
 * Obtain the geometry of the device.
 *
 * You may think that this function is obsolete, but some applications,
 * fdisk for example still used CHS values. This function describes the
 * device as having 224 heads and 56 sectors per cylinder. These values are
 * chosen so that each cylinder is aligned on a 4KB boundary. Since a
 * partition is described in terms of a start and end cylinder this means
 * that each partition is also 4KB aligned. Non-aligned partitions adversely
 * affects performance.
 *
 * @dev Pointer to the block_device strucutre.
 * @geo Pointer to a hd_geometry structure.
 *
 * return value
 *	0       Operation completed successfully.
 *	-ENOTTY An error occurred while reading the drive capacity.
 */
static int mtip_block_getgeo(struct block_device *dev,
				struct hd_geometry *geo)
{
	struct driver_data *dd = dev->bd_disk->private_data;
	sector_t capacity;

	if (!dd)
		return -ENOTTY;

	if (!(mtip_hw_get_capacity(dd, &capacity))) {
		dev_warn(&dd->pdev->dev,
			"Could not get drive capacity.\n");
		return -ENOTTY;
	}

	geo->heads = 224;
	geo->sectors = 56;
	sector_div(capacity, (geo->heads * geo->sectors));
	geo->cylinders = capacity;
	return 0;
}

/*
 * Block device operation function.
 *
 * This structure contains pointers to the functions required by the block
 * layer.
 */
static const struct block_device_operations mtip_block_ops = {
	.ioctl		= mtip_block_ioctl,
#ifdef CONFIG_COMPAT
	.compat_ioctl	= mtip_block_compat_ioctl,
	.getgeo		= mtip_block_getgeo,
	.owner		= THIS_MODULE
};

/*
 * Block layer make request function.
 *
 * This function is called by the kernel to process a BIO for
 * the P320 device.
 *
 * @queue Pointer to the request queue. Unused other than to obtain
 *              the driver data structure.
 * @bio   Pointer to the BIO.
 *
 */
static void mtip_make_request(struct request_queue *queue, struct bio *bio)
{
	struct driver_data *dd = queue->queuedata;
	struct scatterlist *sg;
	struct bio_vec *bvec;
	int nents = 0;
	int tag = 0;

	if (unlikely(dd->dd_flag & MTIP_DDF_STOP_IO)) {
		if (unlikely(test_bit(MTIP_DDF_REMOVE_PENDING_BIT,
							&dd->dd_flag))) {
			bio_endio(bio, -ENXIO);
			return;
		}
		if (unlikely(test_bit(MTIP_DDF_OVER_TEMP_BIT, &dd->dd_flag))) {
			bio_endio(bio, -ENODATA);
			return;
		}
		if (unlikely(test_bit(MTIP_DDF_WRITE_PROTECT_BIT,
							&dd->dd_flag) &&
				bio_data_dir(bio))) {
			bio_endio(bio, -ENODATA);
			return;
		}
		if (unlikely(test_bit(MTIP_DDF_SEC_LOCK_BIT, &dd->dd_flag))) {
			bio_endio(bio, -ENODATA);
			return;
		}
	if (unlikely(!bio_has_data(bio))) {
		blk_queue_flush(queue, 0);
		bio_endio(bio, 0);
	}

	sg = mtip_hw_get_scatterlist(dd, &tag);
	if (likely(sg != NULL)) {
		blk_queue_bounce(queue, &bio);

		if (unlikely((bio)->bi_vcnt > MTIP_MAX_SG)) {
			dev_warn(&dd->pdev->dev,
				"Maximum number of SGL entries exceeded\n");
			bio_io_error(bio);
			mtip_hw_release_scatterlist(dd, tag);
		}

		/* Create the scatter list for this bio. */
		bio_for_each_segment(bvec, bio, nents) {
			sg_set_page(&sg[nents],
					bvec->bv_page,
					bvec->bv_len,
					bvec->bv_offset);
		}

		/* Issue the read/write. */
		mtip_hw_submit_io(dd,
				bio->bi_sector,
				bio_sectors(bio),
				nents,
				tag,
				bio_endio,
				bio,
				bio_data_dir(bio));
		bio_io_error(bio);
}

/*
 * Block layer initialization function.
 *
 * This function is called once by the PCI layer for each P320
 * device that is connected to the system.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0 on success else an error code.
 */
static int mtip_block_initialize(struct driver_data *dd)
	int rv = 0, wait_for_rebuild = 0;
	sector_t capacity;
	unsigned int index = 0;
	struct kobject *kobj;
	unsigned char thd_name[16];
	if (dd->disk)
		goto skip_create_disk; /* hw init done, before rebuild */

	/* Initialize the protocol layer. */
	wait_for_rebuild = mtip_hw_init(dd);
	if (wait_for_rebuild < 0) {
		dev_err(&dd->pdev->dev,
			"Protocol layer initialization failed\n");
		rv = -EINVAL;
		goto protocol_init_error;
	}

	dd->disk = alloc_disk(MTIP_MAX_MINORS);
	if (dd->disk  == NULL) {
		dev_err(&dd->pdev->dev,
			"Unable to allocate gendisk structure\n");
		rv = -EINVAL;
		goto alloc_disk_error;
	}

	/* Generate the disk name, implemented same as in sd.c */
	do {
		if (!ida_pre_get(&rssd_index_ida, GFP_KERNEL))
			goto ida_get_error;

		spin_lock(&rssd_index_lock);
		rv = ida_get_new(&rssd_index_ida, &index);
		spin_unlock(&rssd_index_lock);
	} while (rv == -EAGAIN);

	if (rv)
		goto ida_get_error;

	rv = rssd_disk_name_format("rssd",
				index,
				dd->disk->disk_name,
				DISK_NAME_LEN);
	if (rv)
		goto disk_index_error;

	dd->disk->driverfs_dev	= &dd->pdev->dev;
	dd->disk->major		= dd->major;
	dd->disk->first_minor	= dd->instance * MTIP_MAX_MINORS;
	dd->disk->fops		= &mtip_block_ops;
	dd->disk->private_data	= dd;
	dd->index		= index;

	/*
	 * if rebuild pending, start the service thread, and delay the block
	 * queue creation and add_disk()
	 */
	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
		goto start_service_thread;

skip_create_disk:
	/* Allocate the request queue. */
	dd->queue = blk_alloc_queue(GFP_KERNEL);
	if (dd->queue == NULL) {
		dev_err(&dd->pdev->dev,
			"Unable to allocate request queue\n");
		rv = -ENOMEM;
		goto block_queue_alloc_init_error;
	}

	/* Attach our request function to the request queue. */
	blk_queue_make_request(dd->queue, mtip_make_request);

	dd->disk->queue		= dd->queue;
	dd->queue->queuedata	= dd;

	/* Set device limits. */
	set_bit(QUEUE_FLAG_NONROT, &dd->queue->queue_flags);
	blk_queue_max_segments(dd->queue, MTIP_MAX_SG);
	blk_queue_physical_block_size(dd->queue, 4096);
	blk_queue_max_hw_sectors(dd->queue, 0xffff);
	blk_queue_max_segment_size(dd->queue, 0x400000);
	blk_queue_io_min(dd->queue, 4096);
	/*
	 * write back cache is not supported in the device. FUA depends on
	 * write back cache support, hence setting flush support to zero.
	 */
	blk_queue_flush(dd->queue, 0);

	/* Set the capacity of the device in 512 byte sectors. */
	if (!(mtip_hw_get_capacity(dd, &capacity))) {
		dev_warn(&dd->pdev->dev,
			"Could not read drive capacity\n");
		rv = -EIO;
		goto read_capacity_error;
	}
	set_capacity(dd->disk, capacity);

	/* Enable the block device and add it to /dev */
	add_disk(dd->disk);

	/*
	 * Now that the disk is active, initialize any sysfs attributes
	 * managed by the protocol layer.
	 */
	kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
	if (kobj) {
		mtip_hw_sysfs_init(dd, kobj);
		kobject_put(kobj);
	}
	mtip_hw_debugfs_init(dd);
	if (dd->mtip_svc_handler) {
		set_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag);
		return rv; /* service thread created for handling rebuild */
	sprintf(thd_name, "mtip_svc_thd_%02d", index);

	dd->mtip_svc_handler = kthread_run(mtip_service_thread,
						dd, thd_name);

	if (IS_ERR(dd->mtip_svc_handler)) {
		dev_err(&dd->pdev->dev, "service thread failed to start\n");
		dd->mtip_svc_handler = NULL;
		rv = -EFAULT;
		goto kthread_run_error;
	if (wait_for_rebuild == MTIP_FTL_REBUILD_MAGIC)
		rv = wait_for_rebuild;

	/* Delete our gendisk. This also removes the device from /dev */
read_capacity_error:
	blk_cleanup_queue(dd->queue);

block_queue_alloc_init_error:
disk_index_error:
	spin_lock(&rssd_index_lock);
	ida_remove(&rssd_index_ida, index);
	spin_unlock(&rssd_index_lock);

ida_get_error:
	put_disk(dd->disk);

alloc_disk_error:
	mtip_hw_exit(dd); /* De-initialize the protocol layer. */

protocol_init_error:
	return rv;
}

/*
 * Block layer deinitialization function.
 *
 * Called by the PCI layer as each P320 device is removed.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0
 */
static int mtip_block_remove(struct driver_data *dd)

	if (dd->mtip_svc_handler) {
		set_bit(MTIP_PF_SVC_THD_STOP_BIT, &dd->port->flags);
		wake_up_interruptible(&dd->port->svc_wait);
		kthread_stop(dd->mtip_svc_handler);
	}

	/* Clean up the sysfs attributes, if created */
	if (test_bit(MTIP_DDF_INIT_DONE_BIT, &dd->dd_flag)) {
		kobj = kobject_get(&disk_to_dev(dd->disk)->kobj);
		if (kobj) {
			mtip_hw_sysfs_exit(dd, kobj);
			kobject_put(kobj);
		}
	mtip_hw_debugfs_exit(dd);

	/*
	 * Delete our gendisk structure. This also removes the device
	 * from /dev
	 */
	del_gendisk(dd->disk);
Asai Thambi S P's avatar
Asai Thambi S P committed

	spin_lock(&rssd_index_lock);
	ida_remove(&rssd_index_ida, dd->index);
	spin_unlock(&rssd_index_lock);

	blk_cleanup_queue(dd->queue);
	dd->disk  = NULL;
	dd->queue = NULL;

	/* De-initialize the protocol layer. */
	mtip_hw_exit(dd);

	return 0;
}

/*
 * Function called by the PCI layer when just before the
 * machine shuts down.
 *
 * If a protocol layer shutdown function is present it will be called
 * by this function.
 *
 * @dd Pointer to the driver data structure.
 *
 * return value
 *	0
 */
static int mtip_block_shutdown(struct driver_data *dd)
{
	dev_info(&dd->pdev->dev,
		"Shutting down %s ...\n", dd->disk->disk_name);

	/* Delete our gendisk structure, and cleanup the blk queue. */
	del_gendisk(dd->disk);
Asai Thambi S P's avatar
Asai Thambi S P committed

	spin_lock(&rssd_index_lock);
	ida_remove(&rssd_index_ida, dd->index);
	spin_unlock(&rssd_index_lock);

	blk_cleanup_queue(dd->queue);
	dd->disk  = NULL;
	dd->queue = NULL;

	mtip_hw_shutdown(dd);
	return 0;
}

static int mtip_block_suspend(struct driver_data *dd)
{
	dev_info(&dd->pdev->dev,
		"Suspending %s ...\n", dd->disk->disk_name);
	mtip_hw_suspend(dd);
	return 0;
}

static int mtip_block_resume(struct driver_data *dd)
{
	dev_info(&dd->pdev->dev, "Resuming %s ...\n",
		dd->disk->disk_name);
	mtip_hw_resume(dd);
	return 0;
}

/*
 * Called for each supported PCI device detected.
 *
 * This function allocates the private data structure, enables the
 * PCI device and then calls the block layer initialization function.
 *
 * return value
 *	0 on success else an error code.
 */
static int mtip_pci_probe(struct pci_dev *pdev,
			const struct pci_device_id *ent)
{
	int rv = 0;
	struct driver_data *dd = NULL;

	/* Allocate memory for this devices private data. */
	dd = kzalloc(sizeof(struct driver_data), GFP_KERNEL);
	if (dd == NULL) {
		dev_err(&pdev->dev,
			"Unable to allocate memory for driver data\n");
		return -ENOMEM;
	}

	/* Attach the private data to this PCI device.  */
	pci_set_drvdata(pdev, dd);

	rv = pcim_enable_device(pdev);
	if (rv < 0) {
		dev_err(&pdev->dev, "Unable to enable device\n");
		goto iomap_err;
	}

	/* Map BAR5 to memory. */
	rv = pcim_iomap_regions(pdev, 1 << MTIP_ABAR, MTIP_DRV_NAME);
	if (rv < 0) {
		dev_err(&pdev->dev, "Unable to map regions\n");
		goto iomap_err;
	}

	if (!pci_set_dma_mask(pdev, DMA_BIT_MASK(64))) {
		rv = pci_set_consistent_dma_mask(pdev, DMA_BIT_MASK(64));

		if (rv) {
			rv = pci_set_consistent_dma_mask(pdev,
						DMA_BIT_MASK(32));
			if (rv) {
				dev_warn(&pdev->dev,
					"64-bit DMA enable failed\n");