[SCSI] implement runtime Power Management

This patch (as1398b) adds runtime PM support to the SCSI layer. Only
the machanism is provided; use of it is up to the various high-level
drivers, and the patch doesn't change any of them. Except for sg --
the patch expicitly prevents a device from being runtime-suspended
while its sg device file is open.

The implementation is simplistic. In general, hosts and targets are
automatically suspended when all their children are asleep, but for
them the runtime-suspend code doesn't actually do anything. (A host's
runtime PM status is propagated up the device tree, though, so a
runtime-PM-aware lower-level driver could power down the host adapter
hardware at the appropriate times.) There are comments indicating
where a transport class might be notified or some other hooks added.

LUNs are runtime-suspended by calling the drivers' existing suspend
handlers (and likewise for runtime-resume). Somewhat arbitrarily, the
implementation delays for 100 ms before suspending an eligible LUN.
This is because there typically are occasions during bootup when the
same device file is opened and closed several times in quick
succession.

The way this all works is that the SCSI core increments a device's
PM-usage count when it is registered. If a high-level driver does
nothing then the device will not be eligible for runtime-suspend
because of the elevated usage count. If a high-level driver wants to
use runtime PM then it can call scsi_autopm_put_device() in its probe
routine to decrement the usage count and scsi_autopm_get_device() in
its remove routine to restore the original count.

Hosts, targets, and LUNs are not suspended while they are being probed
or removed, or while the error handler is running. In fact, a fairly
large part of the patch consists of code to make sure that things
aren't suspended at such times.

[jejb: fix up compile issues in PM config variations]
Signed-off-by: Alan Stern <[e-mail]>
Signed-off-by: James Bottomley <[e-mail]>

drivers/scsi/hosts.c

···
 #include <linux/completion.h>
 #include <linux/transport_class.h>
 #include <linux/platform_device.h>
+#include <linux/pm_runtime.h>
 
 #include <scsi/scsi_device.h>
 #include <scsi/scsi_host.h>
···
 void scsi_remove_host(struct Scsi_Host *shost)
 {
 	unsigned long flags;
+
 	mutex_lock(&shost->scan_mutex);
 	spin_lock_irqsave(shost->host_lock, flags);
 	if (scsi_host_set_state(shost, SHOST_CANCEL))
···
 			return;
 		}
 	spin_unlock_irqrestore(shost->host_lock, flags);
+
+	scsi_autopm_get_host(shost);
 	scsi_forget_host(shost);
 	mutex_unlock(&shost->scan_mutex);
 	scsi_proc_host_rm(shost);
···
 		shost->shost_gendev.parent = dev ? dev : &platform_bus;
 	shost->dma_dev = dma_dev;
 
-	device_enable_async_suspend(&shost->shost_gendev);
-
 	error = device_add(&shost->shost_gendev);
 	if (error)
 		goto out;
+
+	pm_runtime_set_active(&shost->shost_gendev);
+	pm_runtime_enable(&shost->shost_gendev);
+	device_enable_async_suspend(&shost->shost_gendev);
 
 	scsi_host_set_state(shost, SHOST_RUNNING);
 	get_device(shost->shost_gendev.parent);

drivers/scsi/scsi_error.c

···
 		 * what we need to do to get it up and online again (if we can).
 		 * If we fail, we end up taking the thing offline.
 		 */
+		if (scsi_autopm_get_host(shost) != 0) {
+			SCSI_LOG_ERROR_RECOVERY(1,
+				printk(KERN_ERR "Error handler scsi_eh_%d "
+						"unable to autoresume\n",
+						shost->host_no));
+			continue;
+		}
+
 		if (shost->transportt->eh_strategy_handler)
 			shost->transportt->eh_strategy_handler(shost);
 		else
···
 		 * which are still online.
 		 */
 		scsi_restart_operations(shost);
+		scsi_autopm_put_host(shost);
 		set_current_state(TASK_INTERRUPTIBLE);
 	}
 	__set_current_state(TASK_RUNNING);
···
 int
 scsi_reset_provider(struct scsi_device *dev, int flag)
 {
-	struct scsi_cmnd *scmd = scsi_get_command(dev, GFP_KERNEL);
+	struct scsi_cmnd *scmd;
 	struct Scsi_Host *shost = dev->host;
 	struct request req;
 	unsigned long flags;
 	int rtn;
 
+	if (scsi_autopm_get_host(shost) < 0)
+		return FAILED;
+
+	scmd = scsi_get_command(dev, GFP_KERNEL);
 	blk_rq_init(NULL, &req);
 	scmd->request = &req;
 
···
 	scsi_run_host_queues(shost);
 
 	scsi_next_command(scmd);
+	scsi_autopm_put_host(shost);
 	return rtn;
 }
 EXPORT_SYMBOL(scsi_reset_provider);

drivers/scsi/scsi_pm.c

···
 
 	if (scsi_is_sdev_device(dev))
 		err = scsi_dev_type_resume(dev);
+
+	if (err == 0) {
+		pm_runtime_disable(dev);
+		pm_runtime_set_active(dev);
+		pm_runtime_enable(dev);
+	}
 	return err;
 }
 
···
 
 #endif /* CONFIG_PM_SLEEP */
 
+#ifdef CONFIG_PM_RUNTIME
+
+static int scsi_runtime_suspend(struct device *dev)
+{
+	int err = 0;
+
+	dev_dbg(dev, "scsi_runtime_suspend\n");
+	if (scsi_is_sdev_device(dev)) {
+		err = scsi_dev_type_suspend(dev, PMSG_AUTO_SUSPEND);
+		if (err == -EAGAIN)
+			pm_schedule_suspend(dev, jiffies_to_msecs(
+				round_jiffies_up_relative(HZ/10)));
+	}
+
+	/* Insert hooks here for targets, hosts, and transport classes */
+
+	return err;
+}
+
+static int scsi_runtime_resume(struct device *dev)
+{
+	int err = 0;
+
+	dev_dbg(dev, "scsi_runtime_resume\n");
+	if (scsi_is_sdev_device(dev))
+		err = scsi_dev_type_resume(dev);
+
+	/* Insert hooks here for targets, hosts, and transport classes */
+
+	return err;
+}
+
+static int scsi_runtime_idle(struct device *dev)
+{
+	int err;
+
+	dev_dbg(dev, "scsi_runtime_idle\n");
+
+	/* Insert hooks here for targets, hosts, and transport classes */
+
+	if (scsi_is_sdev_device(dev))
+		err = pm_schedule_suspend(dev, 100);
+	else
+		err = pm_runtime_suspend(dev);
+	return err;
+}
+
+int scsi_autopm_get_device(struct scsi_device *sdev)
+{
+	int	err;
+
+	err = pm_runtime_get_sync(&sdev->sdev_gendev);
+	if (err < 0)
+		pm_runtime_put_sync(&sdev->sdev_gendev);
+	else if (err > 0)
+		err = 0;
+	return err;
+}
+EXPORT_SYMBOL_GPL(scsi_autopm_get_device);
+
+void scsi_autopm_put_device(struct scsi_device *sdev)
+{
+	pm_runtime_put_sync(&sdev->sdev_gendev);
+}
+EXPORT_SYMBOL_GPL(scsi_autopm_put_device);
+
+void scsi_autopm_get_target(struct scsi_target *starget)
+{
+	pm_runtime_get_sync(&starget->dev);
+}
+
+void scsi_autopm_put_target(struct scsi_target *starget)
+{
+	pm_runtime_put_sync(&starget->dev);
+}
+
+int scsi_autopm_get_host(struct Scsi_Host *shost)
+{
+	int	err;
+
+	err = pm_runtime_get_sync(&shost->shost_gendev);
+	if (err < 0)
+		pm_runtime_put_sync(&shost->shost_gendev);
+	else if (err > 0)
+		err = 0;
+	return err;
+}
+
+void scsi_autopm_put_host(struct Scsi_Host *shost)
+{
+	pm_runtime_put_sync(&shost->shost_gendev);
+}
+
+#else
+
+#define scsi_runtime_suspend	NULL
+#define scsi_runtime_resume	NULL
+#define scsi_runtime_idle	NULL
+
+#endif /* CONFIG_PM_RUNTIME */
+
 const struct dev_pm_ops scsi_bus_pm_ops = {
 	.suspend =		scsi_bus_suspend,
 	.resume =		scsi_bus_resume_common,
···
 	.thaw =			scsi_bus_resume_common,
 	.poweroff =		scsi_bus_poweroff,
 	.restore =		scsi_bus_resume_common,
+	.runtime_suspend =	scsi_runtime_suspend,
+	.runtime_resume =	scsi_runtime_resume,
+	.runtime_idle =		scsi_runtime_idle,
 };

drivers/scsi/scsi_priv.h

···
 struct request;
 struct scsi_cmnd;
 struct scsi_device;
+struct scsi_target;
 struct scsi_host_template;
 struct Scsi_Host;
 struct scsi_nl_hdr;
···
 /* scsi_pm.c */
 #ifdef CONFIG_PM_OPS
 extern const struct dev_pm_ops scsi_bus_pm_ops;
-#else
+#else /* CONFIG_PM_OPS */
 #define scsi_bus_pm_ops		(*NULL)
 #endif
+#ifdef CONFIG_PM_RUNTIME
+extern void scsi_autopm_get_target(struct scsi_target *);
+extern void scsi_autopm_put_target(struct scsi_target *);
+extern int scsi_autopm_get_host(struct Scsi_Host *);
+extern void scsi_autopm_put_host(struct Scsi_Host *);
+#else
+static inline void scsi_autopm_get_target(struct scsi_target *t) {}
+static inline void scsi_autopm_put_target(struct scsi_target *t) {}
+static inline int scsi_autopm_get_host(struct Scsi_Host *h) { return 0; }
+static inline void scsi_autopm_put_host(struct Scsi_Host *h) {}
+#endif /* CONFIG_PM_RUNTIME */
 
 /* 
  * internal scsi timeout functions: for use by mid-layer and transport

drivers/scsi/scsi_scan.c

···
 	starget = scsi_alloc_target(parent, channel, id);
 	if (!starget)
 		return ERR_PTR(-ENOMEM);
+	scsi_autopm_get_target(starget);
 
 	mutex_lock(&shost->scan_mutex);
 	if (!shost->async_scan)
 		scsi_complete_async_scans();
 
-	if (scsi_host_scan_allowed(shost))
+	if (scsi_host_scan_allowed(shost) && scsi_autopm_get_host(shost) == 0) {
 		scsi_probe_and_add_lun(starget, lun, NULL, &sdev, 1, hostdata);
+		scsi_autopm_put_host(shost);
+	}
 	mutex_unlock(&shost->scan_mutex);
+	scsi_autopm_put_target(starget);
 	scsi_target_reap(starget);
 	put_device(&starget->dev);
 
···
 	starget = scsi_alloc_target(parent, channel, id);
 	if (!starget)
 		return;
+	scsi_autopm_get_target(starget);
 
 	if (lun != SCAN_WILD_CARD) {
 		/*
···
 	}
 
  out_reap:
+	scsi_autopm_put_target(starget);
 	/* now determine if the target has any children at all
 	 * and if not, nuke it */
 	scsi_target_reap(starget);
···
 	if (!shost->async_scan)
 		scsi_complete_async_scans();
 
-	if (scsi_host_scan_allowed(shost))
+	if (scsi_host_scan_allowed(shost) && scsi_autopm_get_host(shost) == 0) {
 		__scsi_scan_target(parent, channel, id, lun, rescan);
+		scsi_autopm_put_host(shost);
+	}
 	mutex_unlock(&shost->scan_mutex);
 }
 EXPORT_SYMBOL(scsi_scan_target);
···
 	if (!shost->async_scan)
 		scsi_complete_async_scans();
 
-	if (scsi_host_scan_allowed(shost)) {
+	if (scsi_host_scan_allowed(shost) && scsi_autopm_get_host(shost) == 0) {
 		if (channel == SCAN_WILD_CARD)
 			for (channel = 0; channel <= shost->max_channel;
 			     channel++)
···
 						  rescan);
 		else
 			scsi_scan_channel(shost, channel, id, lun, rescan);
+		scsi_autopm_put_host(shost);
 	}
 	mutex_unlock(&shost->scan_mutex);
 
···
 static int do_scan_async(void *_data)
 {
 	struct async_scan_data *data = _data;
-	do_scsi_scan_host(data->shost);
+	struct Scsi_Host *shost = data->shost;
+
+	do_scsi_scan_host(shost);
 	scsi_finish_async_scan(data);
+	scsi_autopm_put_host(shost);
 	return 0;
 }
 
···
 
 	if (strncmp(scsi_scan_type, "none", 4) == 0)
 		return;
+	if (scsi_autopm_get_host(shost) < 0)
+		return;
 
 	data = scsi_prep_async_scan(shost);
 	if (!data) {
 		do_scsi_scan_host(shost);
+		scsi_autopm_put_host(shost);
 		return;
 	}
 
 	p = kthread_run(do_scan_async, data, "scsi_scan_%d", shost->host_no);
 	if (IS_ERR(p))
 		do_scan_async(data);
+	/* scsi_autopm_put_host(shost) is called in do_scan_async() */
 }
 EXPORT_SYMBOL(scsi_scan_host);
 

drivers/scsi/scsi_sysfs.c

···
 #include <linux/init.h>
 #include <linux/blkdev.h>
 #include <linux/device.h>
+#include <linux/pm_runtime.h>
 
 #include <scsi/scsi.h>
 #include <scsi/scsi_device.h>
···
 	if (starget->state != STARGET_CREATED)
 		return 0;
 
-	device_enable_async_suspend(&starget->dev);
-
 	error = device_add(&starget->dev);
 	if (error) {
 		dev_err(&starget->dev, "target device_add failed, error %d\n", error);
···
 	}
 	transport_add_device(&starget->dev);
 	starget->state = STARGET_RUNNING;
+
+	pm_runtime_set_active(&starget->dev);
+	pm_runtime_enable(&starget->dev);
+	device_enable_async_suspend(&starget->dev);
 
 	return 0;
 }
···
 		return error;
 
 	transport_configure_device(&starget->dev);
+
 	device_enable_async_suspend(&sdev->sdev_gendev);
+	scsi_autopm_get_target(starget);
+	pm_runtime_set_active(&sdev->sdev_gendev);
+	pm_runtime_forbid(&sdev->sdev_gendev);
+	pm_runtime_enable(&sdev->sdev_gendev);
+	scsi_autopm_put_target(starget);
+
+	/* The following call will keep sdev active indefinitely, until
+	 * its driver does a corresponding scsi_autopm_pm_device().  Only
+	 * drivers supporting autosuspend will do this.
+	 */
+	scsi_autopm_get_device(sdev);
+
 	error = device_add(&sdev->sdev_gendev);
 	if (error) {
 		printk(KERN_INFO "error 1\n");

drivers/scsi/sg.c

···
 	if (retval)
 		goto sg_put;
 
+	retval = scsi_autopm_get_device(sdp->device);
+	if (retval)
+		goto sdp_put;
+
 	if (!((flags & O_NONBLOCK) ||
 	      scsi_block_when_processing_errors(sdp->device))) {
 		retval = -ENXIO;
···
 	}
 	retval = 0;
 error_out:
-	if (retval)
+	if (retval) {
+		scsi_autopm_put_device(sdp->device);
+sdp_put:
 		scsi_device_put(sdp->device);
+	}
 sg_put:
 	if (sdp)
 		sg_put_dev(sdp);
···
 	sdp->exclude = 0;
 	wake_up_interruptible(&sdp->o_excl_wait);
 
+	scsi_autopm_put_device(sdp->device);
 	kref_put(&sfp->f_ref, sg_remove_sfp);
 	return 0;
 }

include/scsi/scsi_device.h

···
 			    struct scsi_sense_hdr *, int timeout, int retries,
 			    int *resid);
 
+#ifdef CONFIG_PM_RUNTIME
+extern int scsi_autopm_get_device(struct scsi_device *);
+extern void scsi_autopm_put_device(struct scsi_device *);
+#else
+static inline int scsi_autopm_get_device(struct scsi_device *d) { return 0; }
+static inline void scsi_autopm_put_device(struct scsi_device *d) {}
+#endif /* CONFIG_PM_RUNTIME */
+
 static inline int __must_check scsi_device_reprobe(struct scsi_device *sdev)
 {
 	return device_reprobe(&sdev->sdev_gendev);