libata-eh.c

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

/*
 *  libata-eh.c - libata error handling
 *
 *  Maintained by:  Jeff Garzik <[e-mail]>
 *    		    Please ALWAYS copy [e-mail]
 *		    on emails.
 *
 *  Copyright 2006 Tejun Heo <[e-mail]>
 *
 *
 *  This program is free software; you can redistribute it and/or
 *  modify it under the terms of the GNU General Public License as
 *  published by the Free Software Foundation; either version 2, or
 *  (at your option) any later version.
 *
 *  This program is distributed in the hope that it will be useful,
 *  but WITHOUT ANY WARRANTY; without even the implied warranty of
 *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 *  General Public License for more details.
 *
 *  You should have received a copy of the GNU General Public License
 *  along with this program; see the file COPYING.  If not, write to
 *  the Free Software Foundation, 675 Mass Ave, Cambridge, MA 02139,
 *  USA.
 *
 *
 *  libata documentation is available via 'make {ps|pdf}docs',
 *  as Documentation/DocBook/libata.*
 *
 *  Hardware documentation available from http://www.t13.org/ and
 *  http://www.sata-io.org/
 *
 */

#include <linux/kernel.h>
#include <linux/pci.h>
#include <scsi/scsi.h>
#include <scsi/scsi_host.h>
#include <scsi/scsi_eh.h>
#include <scsi/scsi_device.h>
#include <scsi/scsi_cmnd.h>
#include "../scsi/scsi_transport_api.h"

#include <linux/libata.h>

#include "libata.h"

enum {
	ATA_EH_SPDN_NCQ_OFF		= (1 << 0),
	ATA_EH_SPDN_SPEED_DOWN		= (1 << 1),
	ATA_EH_SPDN_FALLBACK_TO_PIO	= (1 << 2),
};

/* Waiting in ->prereset can never be reliable.  It's sometimes nice
 * to wait there but it can't be depended upon; otherwise, we wouldn't
 * be resetting.  Just give it enough time for most drives to spin up.
 */
enum {
	ATA_EH_PRERESET_TIMEOUT		= 10 * HZ,
	ATA_EH_FASTDRAIN_INTERVAL	= 3 * HZ,
};

/* The following table determines how we sequence resets.  Each entry
 * represents timeout for that try.  The first try can be soft or
 * hardreset.  All others are hardreset if available.  In most cases
 * the first reset w/ 10sec timeout should succeed.  Following entries
 * are mostly for error handling, hotplug and retarded devices.
 */
static const unsigned long ata_eh_reset_timeouts[] = {
	10 * HZ,	/* most drives spin up by 10sec */
	10 * HZ,	/* > 99% working drives spin up before 20sec */
	35 * HZ,	/* give > 30 secs of idleness for retarded devices */
	5 * HZ,		/* and sweet one last chance */
	/* > 1 min has elapsed, give up */
};

static void __ata_port_freeze(struct ata_port *ap);
#ifdef CONFIG_PM
static void ata_eh_handle_port_suspend(struct ata_port *ap);
static void ata_eh_handle_port_resume(struct ata_port *ap);
#else /* CONFIG_PM */
static void ata_eh_handle_port_suspend(struct ata_port *ap)
{ }

static void ata_eh_handle_port_resume(struct ata_port *ap)
{ }
#endif /* CONFIG_PM */

static void __ata_ehi_pushv_desc(struct ata_eh_info *ehi, const char *fmt,
				 va_list args)
{
	ehi->desc_len += vscnprintf(ehi->desc + ehi->desc_len,
				     ATA_EH_DESC_LEN - ehi->desc_len,
				     fmt, args);
}

/**
 *	__ata_ehi_push_desc - push error description without adding separator
 *	@ehi: target EHI
 *	@fmt: printf format string
 *
 *	Format string according to @fmt and append it to @ehi->desc.
 *
 *	LOCKING:
 *	spin_lock_irqsave(host lock)
 */
void __ata_ehi_push_desc(struct ata_eh_info *ehi, const char *fmt, ...)
{
	va_list args;

	va_start(args, fmt);
	__ata_ehi_pushv_desc(ehi, fmt, args);
	va_end(args);
}

/**
 *	ata_ehi_push_desc - push error description with separator
 *	@ehi: target EHI
 *	@fmt: printf format string
 *
 *	Format string according to @fmt and append it to @ehi->desc.
 *	If @ehi->desc is not empty, ", " is added in-between.
 *
 *	LOCKING:
 *	spin_lock_irqsave(host lock)
 */
void ata_ehi_push_desc(struct ata_eh_info *ehi, const char *fmt, ...)
{
	va_list args;

	if (ehi->desc_len)
		__ata_ehi_push_desc(ehi, ", ");

	va_start(args, fmt);
	__ata_ehi_pushv_desc(ehi, fmt, args);
	va_end(args);
}

/**
 *	ata_ehi_clear_desc - clean error description
 *	@ehi: target EHI
 *
 *	Clear @ehi->desc.
 *
 *	LOCKING:
 *	spin_lock_irqsave(host lock)
 */
void ata_ehi_clear_desc(struct ata_eh_info *ehi)
{
	ehi->desc[0] = '\0';
	ehi->desc_len = 0;
}

/**
 *	ata_port_desc - append port description
 *	@ap: target ATA port
 *	@fmt: printf format string
 *
 *	Format string according to @fmt and append it to port
 *	description.  If port description is not empty, " " is added
 *	in-between.  This function is to be used while initializing
 *	ata_host.  The description is printed on host registration.
 *
 *	LOCKING:
 *	None.
 */
void ata_port_desc(struct ata_port *ap, const char *fmt, ...)
{
	va_list args;

	WARN_ON(!(ap->pflags & ATA_PFLAG_INITIALIZING));

	if (ap->link.eh_info.desc_len)
		__ata_ehi_push_desc(&ap->link.eh_info, " ");

	va_start(args, fmt);
	__ata_ehi_pushv_desc(&ap->link.eh_info, fmt, args);
	va_end(args);
}

#ifdef CONFIG_PCI

/**
 *	ata_port_pbar_desc - append PCI BAR description
 *	@ap: target ATA port
 *	@bar: target PCI BAR
 *	@offset: offset into PCI BAR
 *	@name: name of the area
 *
 *	If @offset is negative, this function formats a string which
 *	contains the name, address, size and type of the BAR and
 *	appends it to the port description.  If @offset is zero or
 *	positive, only name and offsetted address is appended.
 *
 *	LOCKING:
 *	None.
 */
void ata_port_pbar_desc(struct ata_port *ap, int bar, ssize_t offset,
			const char *name)
{
	struct pci_dev *pdev = to_pci_dev(ap->host->dev);
	char *type = "";
	unsigned long long start, len;

	if (pci_resource_flags(pdev, bar) & IORESOURCE_MEM)
		type = "m";
	else if (pci_resource_flags(pdev, bar) & IORESOURCE_IO)
		type = "i";

	start = (unsigned long long)pci_resource_start(pdev, bar);
	len = (unsigned long long)pci_resource_len(pdev, bar);

	if (offset < 0)
		ata_port_desc(ap, "%s %s%llu@0x%llx", name, type, len, start);
	else
		ata_port_desc(ap, "%s 0x%llx", name, start + offset);
}

#endif /* CONFIG_PCI */

static void ata_ering_record(struct ata_ering *ering, int is_io,
			     unsigned int err_mask)
{
	struct ata_ering_entry *ent;

	WARN_ON(!err_mask);

	ering->cursor++;
	ering->cursor %= ATA_ERING_SIZE;

	ent = &ering->ring[ering->cursor];
	ent->is_io = is_io;
	ent->err_mask = err_mask;
	ent->timestamp = get_jiffies_64();
}

static void ata_ering_clear(struct ata_ering *ering)
{
	memset(ering, 0, sizeof(*ering));
}

static int ata_ering_map(struct ata_ering *ering,
			 int (*map_fn)(struct ata_ering_entry *, void *),
			 void *arg)
{
	int idx, rc = 0;
	struct ata_ering_entry *ent;

	idx = ering->cursor;
	do {
		ent = &ering->ring[idx];
		if (!ent->err_mask)
			break;
		rc = map_fn(ent, arg);
		if (rc)
			break;
		idx = (idx - 1 + ATA_ERING_SIZE) % ATA_ERING_SIZE;
	} while (idx != ering->cursor);

	return rc;
}

static unsigned int ata_eh_dev_action(struct ata_device *dev)
{
	struct ata_eh_context *ehc = &dev->link->eh_context;

	return ehc->i.action | ehc->i.dev_action[dev->devno];
}

static void ata_eh_clear_action(struct ata_link *link, struct ata_device *dev,
				struct ata_eh_info *ehi, unsigned int action)
{
	struct ata_device *tdev;

	if (!dev) {
		ehi->action &= ~action;
		ata_link_for_each_dev(tdev, link)
			ehi->dev_action[tdev->devno] &= ~action;
	} else {
		/* doesn't make sense for port-wide EH actions */
		WARN_ON(!(action & ATA_EH_PERDEV_MASK));

		/* break ehi->action into ehi->dev_action */
		if (ehi->action & action) {
			ata_link_for_each_dev(tdev, link)
				ehi->dev_action[tdev->devno] |=
					ehi->action & action;
			ehi->action &= ~action;
		}

		/* turn off the specified per-dev action */
		ehi->dev_action[dev->devno] &= ~action;
	}
}

/**
 *	ata_scsi_timed_out - SCSI layer time out callback
 *	@cmd: timed out SCSI command
 *
 *	Handles SCSI layer timeout.  We race with normal completion of
 *	the qc for @cmd.  If the qc is already gone, we lose and let
 *	the scsi command finish (EH_HANDLED).  Otherwise, the qc has
 *	timed out and EH should be invoked.  Prevent ata_qc_complete()
 *	from finishing it by setting EH_SCHEDULED and return
 *	EH_NOT_HANDLED.
 *
 *	TODO: kill this function once old EH is gone.
 *
 *	LOCKING:
 *	Called from timer context
 *
 *	RETURNS:
 *	EH_HANDLED or EH_NOT_HANDLED
 */
enum scsi_eh_timer_return ata_scsi_timed_out(struct scsi_cmnd *cmd)
{
	struct Scsi_Host *host = cmd->device->host;
	struct ata_port *ap = ata_shost_to_port(host);
	unsigned long flags;
	struct ata_queued_cmd *qc;
	enum scsi_eh_timer_return ret;

	DPRINTK("ENTER\n");

	if (ap->ops->error_handler) {
		ret = EH_NOT_HANDLED;
		goto out;
	}

	ret = EH_HANDLED;
	spin_lock_irqsave(ap->lock, flags);
	qc = ata_qc_from_tag(ap, ap->link.active_tag);
	if (qc) {
		WARN_ON(qc->scsicmd != cmd);
		qc->flags |= ATA_QCFLAG_EH_SCHEDULED;
		qc->err_mask |= AC_ERR_TIMEOUT;
		ret = EH_NOT_HANDLED;
	}
	spin_unlock_irqrestore(ap->lock, flags);

 out:
	DPRINTK("EXIT, ret=%d\n", ret);
	return ret;
}

/**
 *	ata_scsi_error - SCSI layer error handler callback
 *	@host: SCSI host on which error occurred
 *
 *	Handles SCSI-layer-thrown error events.
 *
 *	LOCKING:
 *	Inherited from SCSI layer (none, can sleep)
 *
 *	RETURNS:
 *	Zero.
 */
void ata_scsi_error(struct Scsi_Host *host)
{
	struct ata_port *ap = ata_shost_to_port(host);
	int i;
	unsigned long flags;

	DPRINTK("ENTER\n");

	/* synchronize with port task */
	ata_port_flush_task(ap);

	/* synchronize with host lock and sort out timeouts */

	/* For new EH, all qcs are finished in one of three ways -
	 * normal completion, error completion, and SCSI timeout.
	 * Both cmpletions can race against SCSI timeout.  When normal
	 * completion wins, the qc never reaches EH.  When error
	 * completion wins, the qc has ATA_QCFLAG_FAILED set.
	 *
	 * When SCSI timeout wins, things are a bit more complex.
	 * Normal or error completion can occur after the timeout but
	 * before this point.  In such cases, both types of
	 * completions are honored.  A scmd is determined to have
	 * timed out iff its associated qc is active and not failed.
	 */
	if (ap->ops->error_handler) {
		struct scsi_cmnd *scmd, *tmp;
		int nr_timedout = 0;

		spin_lock_irqsave(ap->lock, flags);

		list_for_each_entry_safe(scmd, tmp, &host->eh_cmd_q, eh_entry) {
			struct ata_queued_cmd *qc;

			for (i = 0; i < ATA_MAX_QUEUE; i++) {
				qc = __ata_qc_from_tag(ap, i);
				if (qc->flags & ATA_QCFLAG_ACTIVE &&
				    qc->scsicmd == scmd)
					break;
			}

			if (i < ATA_MAX_QUEUE) {
				/* the scmd has an associated qc */
				if (!(qc->flags & ATA_QCFLAG_FAILED)) {
					/* which hasn't failed yet, timeout */
					qc->err_mask |= AC_ERR_TIMEOUT;
					qc->flags |= ATA_QCFLAG_FAILED;
					nr_timedout++;
				}
			} else {
				/* Normal completion occurred after
				 * SCSI timeout but before this point.
				 * Successfully complete it.
				 */
				scmd->retries = scmd->allowed;
				scsi_eh_finish_cmd(scmd, &ap->eh_done_q);
			}
		}

		/* If we have timed out qcs.  They belong to EH from
		 * this point but the state of the controller is
		 * unknown.  Freeze the port to make sure the IRQ
		 * handler doesn't diddle with those qcs.  This must
		 * be done atomically w.r.t. setting QCFLAG_FAILED.
		 */
		if (nr_timedout)
			__ata_port_freeze(ap);

		spin_unlock_irqrestore(ap->lock, flags);

		/* initialize eh_tries */
		ap->eh_tries = ATA_EH_MAX_TRIES;
	} else
		spin_unlock_wait(ap->lock);

 repeat:
	/* invoke error handler */
	if (ap->ops->error_handler) {
		struct ata_link *link;

		/* kill fast drain timer */
		del_timer_sync(&ap->fastdrain_timer);

		/* process port resume request */
		ata_eh_handle_port_resume(ap);

		/* fetch & clear EH info */
		spin_lock_irqsave(ap->lock, flags);

		__ata_port_for_each_link(link, ap) {
			memset(&link->eh_context, 0, sizeof(link->eh_context));
			link->eh_context.i = link->eh_info;
			memset(&link->eh_info, 0, sizeof(link->eh_info));
		}

		ap->pflags |= ATA_PFLAG_EH_IN_PROGRESS;
		ap->pflags &= ~ATA_PFLAG_EH_PENDING;
		ap->excl_link = NULL;	/* don't maintain exclusion over EH */

		spin_unlock_irqrestore(ap->lock, flags);

		/* invoke EH, skip if unloading or suspended */
		if (!(ap->pflags & (ATA_PFLAG_UNLOADING | ATA_PFLAG_SUSPENDED)))
			ap->ops->error_handler(ap);
		else
			ata_eh_finish(ap);

		/* process port suspend request */
		ata_eh_handle_port_suspend(ap);

		/* Exception might have happend after ->error_handler
		 * recovered the port but before this point.  Repeat
		 * EH in such case.
		 */
		spin_lock_irqsave(ap->lock, flags);

		if (ap->pflags & ATA_PFLAG_EH_PENDING) {
			if (--ap->eh_tries) {
				spin_unlock_irqrestore(ap->lock, flags);
				goto repeat;
			}
			ata_port_printk(ap, KERN_ERR, "EH pending after %d "
					"tries, giving up\n", ATA_EH_MAX_TRIES);
			ap->pflags &= ~ATA_PFLAG_EH_PENDING;
		}

		/* this run is complete, make sure EH info is clear */
		__ata_port_for_each_link(link, ap)
			memset(&link->eh_info, 0, sizeof(link->eh_info));

		/* Clear host_eh_scheduled while holding ap->lock such
		 * that if exception occurs after this point but
		 * before EH completion, SCSI midlayer will
		 * re-initiate EH.
		 */
		host->host_eh_scheduled = 0;

		spin_unlock_irqrestore(ap->lock, flags);
	} else {
		WARN_ON(ata_qc_from_tag(ap, ap->link.active_tag) == NULL);
		ap->ops->eng_timeout(ap);
	}

	/* finish or retry handled scmd's and clean up */
	WARN_ON(host->host_failed || !list_empty(&host->eh_cmd_q));

	scsi_eh_flush_done_q(&ap->eh_done_q);

	/* clean up */
	spin_lock_irqsave(ap->lock, flags);

	if (ap->pflags & ATA_PFLAG_LOADING)
		ap->pflags &= ~ATA_PFLAG_LOADING;
	else if (ap->pflags & ATA_PFLAG_SCSI_HOTPLUG)
		queue_delayed_work(ata_aux_wq, &ap->hotplug_task, 0);