libata-eh.c

来自「linux 内核源代码」· C语言 代码 · 共 2,561 行 · 第 1/5 页

 *	Called just before performing EH actions to clear related bits
 *	in @link->eh_info such that eh actions are not unnecessarily
 *	repeated.
 *
 *	LOCKING:
 *	None.
 */
void ata_eh_about_to_do(struct ata_link *link, struct ata_device *dev,
			unsigned int action)
{
	struct ata_port *ap = link->ap;
	struct ata_eh_info *ehi = &link->eh_info;
	struct ata_eh_context *ehc = &link->eh_context;
	unsigned long flags;

	spin_lock_irqsave(ap->lock, flags);

	/* Reset is represented by combination of actions and EHI
	 * flags.  Suck in all related bits before clearing eh_info to
	 * avoid losing requested action.
	 */
	if (action & ATA_EH_RESET_MASK) {
		ehc->i.action |= ehi->action & ATA_EH_RESET_MASK;
		ehc->i.flags |= ehi->flags & ATA_EHI_RESET_MODIFIER_MASK;

		/* make sure all reset actions are cleared & clear EHI flags */
		action |= ATA_EH_RESET_MASK;
		ehi->flags &= ~ATA_EHI_RESET_MODIFIER_MASK;
	}

	ata_eh_clear_action(link, dev, ehi, action);

	if (!(ehc->i.flags & ATA_EHI_QUIET))
		ap->pflags |= ATA_PFLAG_RECOVERED;

	spin_unlock_irqrestore(ap->lock, flags);
}

/**
 *	ata_eh_done - EH action complete
*	@ap: target ATA port
 *	@dev: target ATA dev for per-dev action (can be NULL)
 *	@action: action just completed
 *
 *	Called right after performing EH actions to clear related bits
 *	in @link->eh_context.
 *
 *	LOCKING:
 *	None.
 */
void ata_eh_done(struct ata_link *link, struct ata_device *dev,
		 unsigned int action)
{
	struct ata_eh_context *ehc = &link->eh_context;

	/* if reset is complete, clear all reset actions & reset modifier */
	if (action & ATA_EH_RESET_MASK) {
		action |= ATA_EH_RESET_MASK;
		ehc->i.flags &= ~ATA_EHI_RESET_MODIFIER_MASK;
	}

	ata_eh_clear_action(link, dev, &ehc->i, action);
}

/**
 *	ata_err_string - convert err_mask to descriptive string
 *	@err_mask: error mask to convert to string
 *
 *	Convert @err_mask to descriptive string.  Errors are
 *	prioritized according to severity and only the most severe
 *	error is reported.
 *
 *	LOCKING:
 *	None.
 *
 *	RETURNS:
 *	Descriptive string for @err_mask
 */
static const char *ata_err_string(unsigned int err_mask)
{
	if (err_mask & AC_ERR_HOST_BUS)
		return "host bus error";
	if (err_mask & AC_ERR_ATA_BUS)
		return "ATA bus error";
	if (err_mask & AC_ERR_TIMEOUT)
		return "timeout";
	if (err_mask & AC_ERR_HSM)
		return "HSM violation";
	if (err_mask & AC_ERR_SYSTEM)
		return "internal error";
	if (err_mask & AC_ERR_MEDIA)
		return "media error";
	if (err_mask & AC_ERR_INVALID)
		return "invalid argument";
	if (err_mask & AC_ERR_DEV)
		return "device error";
	return "unknown error";
}

/**
 *	ata_read_log_page - read a specific log page
 *	@dev: target device
 *	@page: page to read
 *	@buf: buffer to store read page
 *	@sectors: number of sectors to read
 *
 *	Read log page using READ_LOG_EXT command.
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 *
 *	RETURNS:
 *	0 on success, AC_ERR_* mask otherwise.
 */
static unsigned int ata_read_log_page(struct ata_device *dev,
				      u8 page, void *buf, unsigned int sectors)
{
	struct ata_taskfile tf;
	unsigned int err_mask;

	DPRINTK("read log page - page %d\n", page);

	ata_tf_init(dev, &tf);
	tf.command = ATA_CMD_READ_LOG_EXT;
	tf.lbal = page;
	tf.nsect = sectors;
	tf.hob_nsect = sectors >> 8;
	tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_LBA48 | ATA_TFLAG_DEVICE;
	tf.protocol = ATA_PROT_PIO;

	err_mask = ata_exec_internal(dev, &tf, NULL, DMA_FROM_DEVICE,
				     buf, sectors * ATA_SECT_SIZE, 0);

	DPRINTK("EXIT, err_mask=%x\n", err_mask);
	return err_mask;
}

/**
 *	ata_eh_read_log_10h - Read log page 10h for NCQ error details
 *	@dev: Device to read log page 10h from
 *	@tag: Resulting tag of the failed command
 *	@tf: Resulting taskfile registers of the failed command
 *
 *	Read log page 10h to obtain NCQ error details and clear error
 *	condition.
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 *
 *	RETURNS:
 *	0 on success, -errno otherwise.
 */
static int ata_eh_read_log_10h(struct ata_device *dev,
			       int *tag, struct ata_taskfile *tf)
{
	u8 *buf = dev->link->ap->sector_buf;
	unsigned int err_mask;
	u8 csum;
	int i;

	err_mask = ata_read_log_page(dev, ATA_LOG_SATA_NCQ, buf, 1);
	if (err_mask)
		return -EIO;

	csum = 0;
	for (i = 0; i < ATA_SECT_SIZE; i++)
		csum += buf[i];
	if (csum)
		ata_dev_printk(dev, KERN_WARNING,
			       "invalid checksum 0x%x on log page 10h\n", csum);

	if (buf[0] & 0x80)
		return -ENOENT;

	*tag = buf[0] & 0x1f;

	tf->command = buf[2];
	tf->feature = buf[3];
	tf->lbal = buf[4];
	tf->lbam = buf[5];
	tf->lbah = buf[6];
	tf->device = buf[7];
	tf->hob_lbal = buf[8];
	tf->hob_lbam = buf[9];
	tf->hob_lbah = buf[10];
	tf->nsect = buf[12];
	tf->hob_nsect = buf[13];

	return 0;
}

/**
 *	atapi_eh_request_sense - perform ATAPI REQUEST_SENSE
 *	@dev: device to perform REQUEST_SENSE to
 *	@sense_buf: result sense data buffer (SCSI_SENSE_BUFFERSIZE bytes long)
 *
 *	Perform ATAPI REQUEST_SENSE after the device reported CHECK
 *	SENSE.  This function is EH helper.
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 *
 *	RETURNS:
 *	0 on success, AC_ERR_* mask on failure
 */
static unsigned int atapi_eh_request_sense(struct ata_queued_cmd *qc)
{
	struct ata_device *dev = qc->dev;
	unsigned char *sense_buf = qc->scsicmd->sense_buffer;
	struct ata_port *ap = dev->link->ap;
	struct ata_taskfile tf;
	u8 cdb[ATAPI_CDB_LEN];

	DPRINTK("ATAPI request sense\n");

	/* FIXME: is this needed? */
	memset(sense_buf, 0, SCSI_SENSE_BUFFERSIZE);

	/* initialize sense_buf with the error register,
	 * for the case where they are -not- overwritten
	 */
	sense_buf[0] = 0x70;
	sense_buf[2] = qc->result_tf.feature >> 4;

	/* some devices time out if garbage left in tf */
	ata_tf_init(dev, &tf);

	memset(cdb, 0, ATAPI_CDB_LEN);
	cdb[0] = REQUEST_SENSE;
	cdb[4] = SCSI_SENSE_BUFFERSIZE;

	tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
	tf.command = ATA_CMD_PACKET;

	/* is it pointless to prefer PIO for "safety reasons"? */
	if (ap->flags & ATA_FLAG_PIO_DMA) {
		tf.protocol = ATA_PROT_ATAPI_DMA;
		tf.feature |= ATAPI_PKT_DMA;
	} else {
		tf.protocol = ATA_PROT_ATAPI;
		tf.lbam = SCSI_SENSE_BUFFERSIZE;
		tf.lbah = 0;
	}

	return ata_exec_internal(dev, &tf, cdb, DMA_FROM_DEVICE,
				 sense_buf, SCSI_SENSE_BUFFERSIZE, 0);
}

/**
 *	ata_eh_analyze_serror - analyze SError for a failed port
 *	@link: ATA link to analyze SError for
 *
 *	Analyze SError if available and further determine cause of
 *	failure.
 *
 *	LOCKING:
 *	None.
 */
static void ata_eh_analyze_serror(struct ata_link *link)
{
	struct ata_eh_context *ehc = &link->eh_context;
	u32 serror = ehc->i.serror;
	unsigned int err_mask = 0, action = 0;
	u32 hotplug_mask;

	if (serror & SERR_PERSISTENT) {
		err_mask |= AC_ERR_ATA_BUS;
		action |= ATA_EH_HARDRESET;
	}
	if (serror &
	    (SERR_DATA_RECOVERED | SERR_COMM_RECOVERED | SERR_DATA)) {
		err_mask |= AC_ERR_ATA_BUS;
		action |= ATA_EH_SOFTRESET;
	}
	if (serror & SERR_PROTOCOL) {
		err_mask |= AC_ERR_HSM;
		action |= ATA_EH_SOFTRESET;
	}
	if (serror & SERR_INTERNAL) {
		err_mask |= AC_ERR_SYSTEM;
		action |= ATA_EH_HARDRESET;
	}

	/* Determine whether a hotplug event has occurred.  Both
	 * SError.N/X are considered hotplug events for enabled or
	 * host links.  For disabled PMP links, only N bit is
	 * considered as X bit is left at 1 for link plugging.
	 */
	hotplug_mask = 0;

	if (!(link->flags & ATA_LFLAG_DISABLED) || ata_is_host_link(link))
		hotplug_mask = SERR_PHYRDY_CHG | SERR_DEV_XCHG;
	else
		hotplug_mask = SERR_PHYRDY_CHG;

	if (serror & hotplug_mask)
		ata_ehi_hotplugged(&ehc->i);

	ehc->i.err_mask |= err_mask;
	ehc->i.action |= action;
}

/**
 *	ata_eh_analyze_ncq_error - analyze NCQ error
 *	@link: ATA link to analyze NCQ error for
 *
 *	Read log page 10h, determine the offending qc and acquire
 *	error status TF.  For NCQ device errors, all LLDDs have to do
 *	is setting AC_ERR_DEV in ehi->err_mask.  This function takes
 *	care of the rest.
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 */
static void ata_eh_analyze_ncq_error(struct ata_link *link)
{
	struct ata_port *ap = link->ap;
	struct ata_eh_context *ehc = &link->eh_context;
	struct ata_device *dev = link->device;
	struct ata_queued_cmd *qc;
	struct ata_taskfile tf;
	int tag, rc;

	/* if frozen, we can't do much */
	if (ap->pflags & ATA_PFLAG_FROZEN)
		return;

	/* is it NCQ device error? */
	if (!link->sactive || !(ehc->i.err_mask & AC_ERR_DEV))
		return;

	/* has LLDD analyzed already? */
	for (tag = 0; tag < ATA_MAX_QUEUE; tag++) {
		qc = __ata_qc_from_tag(ap, tag);

		if (!(qc->flags & ATA_QCFLAG_FAILED))
			continue;

		if (qc->err_mask)
			return;
	}

	/* okay, this error is ours */
	rc = ata_eh_read_log_10h(dev, &tag, &tf);
	if (rc) {
		ata_link_printk(link, KERN_ERR, "failed to read log page 10h "
				"(errno=%d)\n", rc);
		return;
	}

	if (!(link->sactive & (1 << tag))) {
		ata_link_printk(link, KERN_ERR, "log page 10h reported "
				"inactive tag %d\n", tag);
		return;
	}

	/* we've got the perpetrator, condemn it */
	qc = __ata_qc_from_tag(ap, tag);
	memcpy(&qc->result_tf, &tf, sizeof(tf));
	qc->err_mask |= AC_ERR_DEV | AC_ERR_NCQ;
	ehc->i.err_mask &= ~AC_ERR_DEV;
}

/**
 *	ata_eh_analyze_tf - analyze taskfile of a failed qc
 *	@qc: qc to analyze
 *	@tf: Taskfile registers to analyze
 *
 *	Analyze taskfile of @qc and further determine cause of
 *	failure.  This function also requests ATAPI sense data if
 *	avaliable.
 *
 *	LOCKING:
 *	Kernel thread context (may sleep).
 *
 *	RETURNS:
 *	Determined recovery action
 */
static unsigned int ata_eh_analyze_tf(struct ata_queued_cmd *qc,
				      const struct ata_taskfile *tf)
{
	unsigned int tmp, action = 0;
	u8 stat = tf->command, err = tf->feature;

	if ((stat & (ATA_BUSY | ATA_DRQ | ATA_DRDY)) != ATA_DRDY) {
		qc->err_mask |= AC_ERR_HSM;
		return ATA_EH_SOFTRESET;
	}

	if (stat & (ATA_ERR | ATA_DF))
		qc->err_mask |= AC_ERR_DEV;
	else
		return 0;

	switch (qc->dev->class) {
	case ATA_DEV_ATA:
		if (err & ATA_ICRC)
			qc->err_mask |= AC_ERR_ATA_BUS;
		if (err & ATA_UNC)
			qc->err_mask |= AC_ERR_MEDIA;
		if (err & ATA_IDNF)
			qc->err_mask |= AC_ERR_INVALID;
		break;

	case ATA_DEV_ATAPI:
		if (!(qc->ap->pflags & ATA_PFLAG_FROZEN)) {
			tmp = atapi_eh_request_sense(qc);
			if (!tmp) {
				/* ATA_QCFLAG_SENSE_VALID is used to
				 * tell atapi_qc_complete() that sense
				 * data is already valid.
				 *
				 * TODO: interpret sense data and set
				 * appropriate err_mask.
				 */
				qc->flags |= ATA_QCFLAG_SENSE_VALID;
			} else
				qc->err_mask |= tmp;
		}
	}

	if (qc->err_mask & (AC_ERR_HSM | AC_ERR_TIMEOUT | AC_ERR_ATA_BUS))
		action |= ATA_EH_SOFTRESET;

	return action;
}

static int ata_eh_categorize_error(int is_io, unsigned int err_mask)
{
	if (err_mask & AC_ERR_ATA_BUS)
		return 1;

	if (err_mask & AC_ERR_TIMEOUT)
		return 2;

	if (is_io) {
		if (err_mask & AC_ERR_HSM)
			return 2;
		if ((err_mask &
		     (AC_ERR_DEV|AC_ERR_MEDIA|AC_ERR_INVALID)) == AC_ERR_DEV)
			return 3;
	}

	return 0;
}

struct speed_down_verdict_arg {
	u64 since;
	int nr_errors[4];
};

static int speed_down_verdict_cb(struct ata_ering_entry *ent, void *void_arg)
{
	struct speed_down_verdict_arg *arg = void_arg;
	int cat = ata_eh_categorize_error(ent->is_io, ent->err_mask);

	if (ent->timestamp < arg->since)
		return -1;

	arg->nr_errors[cat]++;
	return 0;
}

/**
 *	ata_eh_speed_down_verdict - Determine speed down verdict
 *	@dev: Device of interest
 *
 *	This function examines error ring of @dev and determines
 *	whether NCQ needs to be turned off, transfer speed should be
 *	stepped down, or falling back to PIO is necessary.
 *
 *	Cat-1 is ATA_BUS error for any command.
 *
 *	Cat-2 is TIMEOUT for any command or HSM violation for known
 *	supported commands.
 *
 *	Cat-3 is is unclassified DEV error for known supported
 *	command.
 *
 *	NCQ needs to be turned off if there have been more than 3
 *	Cat-2 + Cat-3 errors during last 10 minutes.
 *
 *	Speed down is necessary if there have been more than 3 Cat-1 +
 *	Cat-2 errors or 10 Cat-3 errors during last 10 minutes.
 *
 *	Falling back to PIO mode is necessary if there have been more
 *	than 10 Cat-1 + Cat-2 + Cat-3 errors during last 5 minutes.
 *
 *	LOCKING:
 *	Inherited from caller.
 *
 *	RETURNS:
 *	OR of ATA_EH_SPDN_* flags.
 */
static unsigned int ata_eh_speed_down_verdict(struct ata_device *dev)
{
	const u64 j5mins = 5LLU * 60 * HZ, j10mins = 10LLU * 60 * HZ;
	u64 j64 = get_jiffies_64();
	struct speed_down_verdict_arg arg;
	unsigned int verdict = 0;

	/* scan past 10 mins of error history */
	memset(&arg, 0, sizeof(arg));
	arg.since = j64 - min(j64, j10mins);
	ata_ering_map(&dev->ering, speed_down_verdict_cb, &arg);

	if (arg.nr_errors[2] + arg.nr_errors[3] > 3)
		verdict |= ATA_EH_SPDN_NCQ_OFF;
	if (arg.nr_errors[1] + arg.nr_errors[2] > 3 || arg.nr_errors[3] > 10)
		verdict |= ATA_EH_SPDN_SPEED_DOWN;

	/* scan past 3 mins of error history */
	memset(&arg, 0, sizeof(arg));