| Message ID | 52293763.5070206@intel.com (mailing list archive) |
|---|---|
| State | Not Applicable, archived |
| Headers | show |
On Fri, 6 Sep 2013, Aaron Lu wrote: > Migrate SCSI Optical Disk Drive(ODD) driver sr to make use of block > layer runtime PM. Accordingly, the SCSI bus layer runtime PM callback is > simplified as all SCSI devices that implement runtime PM are now request > based. > > Note that due to ODD will be polled every 2 seconds, for suspend to > actually happen, the autosuspend_delay can not be set to more than 2 > seconds or the polling interval has to be increased. > > Signed-off-by: Aaron Lu <aaron.lu@intel.com> Acked-by: Alan Stern <stern@rowland.harvard.edu> -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 2013-09-06 at 10:01 +0800, Aaron Lu wrote: > Migrate SCSI Optical Disk Drive(ODD) I'm not very keen on this description because it's not quite accurate. sr stands for SCSI ROM. You could say optical SCSI ROM perhaps, but Magneto Optical disks are handled by sd not sr. > driver sr to make use of block I'd just say "Migrate sr to make use of .." > layer runtime PM. Accordingly, the SCSI bus layer runtime PM callback is > simplified as all SCSI devices that implement runtime PM are now request > based. OK, let's now try for a descriptive changelog. All SCSI devices (including sr) are "request based". I think what you mean is "all SCSI devices which implement runtime PM have an exposed block device"? > Note that due to ODD will be polled every 2 seconds, Since the device will be polled every 2 seconds > for suspend to > actually happen, the autosuspend_delay can not be set to more than 2 > seconds or the polling interval has to be increased. Is this true? What about event driven devices? Supposing a distro has a different interval. How about If your Distribution polls the device, the autosuspend interval cannot be set to longer than the polling interval otherwise the device will never suspend. James -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Fri, 6 Sep 2013, James Bottomley wrote: > > layer runtime PM. Accordingly, the SCSI bus layer runtime PM callback is > > simplified as all SCSI devices that implement runtime PM are now request > > based. > > OK, let's now try for a descriptive changelog. All SCSI devices > (including sr) are "request based". I think what you mean is "all SCSI > devices which implement runtime PM have an exposed block device"? Actually Aaron means "All SCSI drivers implementing runtime PM now use the block layer's request-based mechanism." > > Note that due to ODD will be polled every 2 seconds, > > Since the device will be polled every 2 seconds > > > for suspend to > > actually happen, the autosuspend_delay can not be set to more than 2 > > seconds or the polling interval has to be increased. > > Is this true? What about event driven devices? Supposing a distro has > a different interval. > > How about > > If your Distribution polls the device, the autosuspend interval cannot > be set to longer than the polling interval otherwise the device will > never suspend. The default polling done by the kernel uses 2-second intervals. Of course, distributions and users can change this or disable it entirely. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/06/2013 11:00 PM, James Bottomley wrote: > On Fri, 2013-09-06 at 10:01 +0800, Aaron Lu wrote: >> Migrate SCSI Optical Disk Drive(ODD) > > I'm not very keen on this description because it's not quite accurate. > sr stands for SCSI ROM. You could say optical SCSI ROM perhaps, but > Magneto Optical disks are handled by sd not sr. > >> driver sr to make use of block > > I'd just say "Migrate sr to make use of .." OK, will update in v3, thanks for the correction. Thanks, Aaron -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On 09/06/2013 11:56 PM, Alan Stern wrote: > On Fri, 6 Sep 2013, James Bottomley wrote: > >>> layer runtime PM. Accordingly, the SCSI bus layer runtime PM callback is >>> simplified as all SCSI devices that implement runtime PM are now request >>> based. >> >> OK, let's now try for a descriptive changelog. All SCSI devices >> (including sr) are "request based". I think what you mean is "all SCSI >> devices which implement runtime PM have an exposed block device"? > > Actually Aaron means "All SCSI drivers implementing runtime PM now use > the block layer's request-based mechanism." Exactly, thanks for the clarification Alan. I'll update changlog using the above words in v3. > >>> Note that due to ODD will be polled every 2 seconds, >> >> Since the device will be polled every 2 seconds >> >>> for suspend to >>> actually happen, the autosuspend_delay can not be set to more than 2 >>> seconds or the polling interval has to be increased. >> >> Is this true? What about event driven devices? Supposing a distro has >> a different interval. >> >> How about >> >> If your Distribution polls the device, the autosuspend interval cannot >> be set to longer than the polling interval otherwise the device will >> never suspend. > > The default polling done by the kernel uses 2-second intervals. Of > course, distributions and users can change this or disable it entirely. What about: Note that due to the device will be polled by kernel at a constant interval specified at /sys/module/block/parameters/events_dfl_poll_msecs, the default value is 2 seconds, the autosuspend delay cannot be set to longer than the polling interval otherwise the device will never suspend. Thanks, Aaron -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
On Sat, 7 Sep 2013, Aaron Lu wrote: > >>> Note that due to ODD will be polled every 2 seconds, > >> > >> Since the device will be polled every 2 seconds > >> > >>> for suspend to > >>> actually happen, the autosuspend_delay can not be set to more than 2 > >>> seconds or the polling interval has to be increased. > >> > >> Is this true? What about event driven devices? Supposing a distro has > >> a different interval. > >> > >> How about > >> > >> If your Distribution polls the device, the autosuspend interval cannot > >> be set to longer than the polling interval otherwise the device will > >> never suspend. > > > > The default polling done by the kernel uses 2-second intervals. Of > > course, distributions and users can change this or disable it entirely. > > What about: > Note that due to the device will be polled by kernel at a constant > interval specified at /sys/module/block/parameters/events_dfl_poll_msecs, > the default value is 2 seconds, the autosuspend delay cannot be set to > longer than the polling interval otherwise the device will never suspend. Well, I wouldn't say "cannot" -- people can set the delay to whatever they want. Just: If the autosuspend delay is set longer than the polling interval then the device will never suspend. Also, the actual polling interval is controlled by /sys/block/sdX/events_poll_msec. The file you mentioned merely contains the default polling interval used for new devices. Alan Stern -- To unsubscribe from this list: send the line "unsubscribe linux-pm" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/drivers/scsi/scsi_pm.c b/drivers/scsi/scsi_pm.c index 4c5aabe..bb31fc9 100644 --- a/drivers/scsi/scsi_pm.c +++ b/drivers/scsi/scsi_pm.c @@ -16,6 +16,8 @@ #include "scsi_priv.h" +#ifdef CONFIG_PM_SLEEP + static int scsi_dev_type_suspend(struct device *dev, int (*cb)(struct device *)) { int err; @@ -43,8 +45,6 @@ static int scsi_dev_type_resume(struct device *dev, int (*cb)(struct device *)) return err; } -#ifdef CONFIG_PM_SLEEP - static int scsi_bus_suspend_common(struct device *dev, int (*cb)(struct device *)) { @@ -144,38 +144,22 @@ static int scsi_bus_restore(struct device *dev) #ifdef CONFIG_PM_RUNTIME -static int sdev_blk_runtime_suspend(struct scsi_device *sdev, - int (*cb)(struct device *)) +static int sdev_runtime_suspend(struct device *dev) { + const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL; + struct scsi_device *sdev = to_scsi_device(dev); int err; err = blk_pre_runtime_suspend(sdev->request_queue); if (err) return err; - if (cb) - err = cb(&sdev->sdev_gendev); + if (pm && pm->runtime_suspend) + err = pm->runtime_suspend(dev); blk_post_runtime_suspend(sdev->request_queue, err); return err; } -static int sdev_runtime_suspend(struct device *dev) -{ - const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL; - int (*cb)(struct device *) = pm ? pm->runtime_suspend : NULL; - struct scsi_device *sdev = to_scsi_device(dev); - int err; - - if (sdev->request_queue->dev) - return sdev_blk_runtime_suspend(sdev, cb); - - err = scsi_dev_type_suspend(dev, cb); - if (err == -EAGAIN) - pm_schedule_suspend(dev, jiffies_to_msecs( - round_jiffies_up_relative(HZ/10))); - return err; -} - static int scsi_runtime_suspend(struct device *dev) { int err = 0; @@ -189,31 +173,20 @@ static int scsi_runtime_suspend(struct device *dev) return err; } -static int sdev_blk_runtime_resume(struct scsi_device *sdev, - int (*cb)(struct device *)) +static int sdev_runtime_resume(struct device *dev) { + struct scsi_device *sdev = to_scsi_device(dev); + const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL; int err = 0; blk_pre_runtime_resume(sdev->request_queue); - if (cb) - err = cb(&sdev->sdev_gendev); + if (pm && pm->runtime_resume) + err = pm->runtime_resume(dev); blk_post_runtime_resume(sdev->request_queue, err); return err; } -static int sdev_runtime_resume(struct device *dev) -{ - struct scsi_device *sdev = to_scsi_device(dev); - const struct dev_pm_ops *pm = dev->driver ? dev->driver->pm : NULL; - int (*cb)(struct device *) = pm ? pm->runtime_resume : NULL; - - if (sdev->request_queue->dev) - return sdev_blk_runtime_resume(sdev, cb); - else - return scsi_dev_type_resume(dev, cb); -} - static int scsi_runtime_resume(struct device *dev) { int err = 0; @@ -234,14 +207,11 @@ static int scsi_runtime_idle(struct device *dev) /* Insert hooks here for targets, hosts, and transport classes */ if (scsi_is_sdev_device(dev)) { - struct scsi_device *sdev = to_scsi_device(dev); - - if (sdev->request_queue->dev) { - pm_runtime_mark_last_busy(dev); - pm_runtime_autosuspend(dev); - return -EBUSY; - } + pm_runtime_mark_last_busy(dev); + pm_runtime_autosuspend(dev); + return -EBUSY; } + return 0; } diff --git a/drivers/scsi/sr.c b/drivers/scsi/sr.c index 119d67f..40d8592 100644 --- a/drivers/scsi/sr.c +++ b/drivers/scsi/sr.c @@ -161,14 +161,10 @@ static inline struct scsi_cd *scsi_cd_get(struct gendisk *disk) goto out; cd = scsi_cd(disk); kref_get(&cd->kref); - if (scsi_device_get(cd->device)) - goto out_put; - if (!scsi_autopm_get_device(cd->device)) - goto out; - - out_put: - kref_put(&cd->kref, sr_kref_release); - cd = NULL; + if (scsi_device_get(cd->device)) { + kref_put(&cd->kref, sr_kref_release); + cd = NULL; + } out: mutex_unlock(&sr_ref_mutex); return cd; @@ -180,7 +176,6 @@ static void scsi_cd_put(struct scsi_cd *cd) mutex_lock(&sr_ref_mutex); kref_put(&cd->kref, sr_kref_release); - scsi_autopm_put_device(sdev); scsi_device_put(sdev); mutex_unlock(&sr_ref_mutex); } @@ -558,8 +553,6 @@ static int sr_block_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, void __user *argp = (void __user *)arg; int ret; - scsi_autopm_get_device(cd->device); - mutex_lock(&sr_mutex); /* @@ -591,7 +584,6 @@ static int sr_block_ioctl(struct block_device *bdev, fmode_t mode, unsigned cmd, out: mutex_unlock(&sr_mutex); - scsi_autopm_put_device(cd->device); return ret; } @@ -599,17 +591,11 @@ static unsigned int sr_block_check_events(struct gendisk *disk, unsigned int clearing) { struct scsi_cd *cd = scsi_cd(disk); - unsigned int ret; - if (atomic_read(&cd->device->disk_events_disable_depth) == 0) { - scsi_autopm_get_device(cd->device); - ret = cdrom_check_events(&cd->cdi, clearing); - scsi_autopm_put_device(cd->device); - } else { - ret = 0; - } + if (atomic_read(&cd->device->disk_events_disable_depth)) + return 0; - return ret; + return cdrom_check_events(&cd->cdi, clearing); } static int sr_block_revalidate_disk(struct gendisk *disk) @@ -617,8 +603,6 @@ static int sr_block_revalidate_disk(struct gendisk *disk) struct scsi_cd *cd = scsi_cd(disk); struct scsi_sense_hdr sshdr; - scsi_autopm_get_device(cd->device); - /* if the unit is not ready, nothing more to do */ if (scsi_test_unit_ready(cd->device, SR_TIMEOUT, MAX_RETRIES, &sshdr)) goto out; @@ -626,7 +610,6 @@ static int sr_block_revalidate_disk(struct gendisk *disk) sr_cd_check(&cd->cdi); get_sectorsize(cd); out: - scsi_autopm_put_device(cd->device); return 0; } @@ -747,6 +730,12 @@ static int sr_probe(struct device *dev) if (register_cdrom(&cd->cdi)) goto fail_put; + /* + * Initialize block layer runtime PM stuffs before the + * periodic event checking request gets started in add_disk. + */ + blk_pm_runtime_init(sdev->request_queue, dev); + dev_set_drvdata(dev, cd); disk->flags |= GENHD_FL_REMOVABLE; add_disk(disk);
Migrate SCSI Optical Disk Drive(ODD) driver sr to make use of block layer runtime PM. Accordingly, the SCSI bus layer runtime PM callback is simplified as all SCSI devices that implement runtime PM are now request based. Note that due to ODD will be polled every 2 seconds, for suspend to actually happen, the autosuspend_delay can not be set to more than 2 seconds or the polling interval has to be increased. Signed-off-by: Aaron Lu <aaron.lu@intel.com> --- v2: scsi_dev_type_suspend/resume should be covered by CONFIG_PM_SLEEP, suggested by Alan Stern. Make sure to use the util-linux utility of version 2.23.2 or later if a ATA host is in use, as there is a bug fix for eject command to correctly update the ODD's locked state. If the fix is not applied, after the ODD is ejected with the eject command, it will not be able to enter runtime suspend state any more due to SCSI EH code will submit a lock door command for the ODD right after its parent ATA port is runtime suspended. https://git.kernel.org/cgit/utils/util-linux/util-linux.git/commit/?id=12272030e563201e0b06732d8a87d8cea7157a04 drivers/scsi/scsi_pm.c | 62 +++++++++++++------------------------------------- drivers/scsi/sr.c | 37 +++++++++++------------------- 2 files changed, 29 insertions(+), 70 deletions(-)