[PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Todd E Brandt]

Part 2 of the hard disk resume optimization patch, this one applies
to the scsi subsystem.


Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>

 drivers/scsi/sd.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 81 insertions(+), 1 deletion(-)

diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
index 86fcf2c..d4bf784 100644
--- a/drivers/scsi/sd.c
+++ b/drivers/scsi/sd.c
@@ -107,6 +107,7 @@ static int  sd_remove(struct device *);
 static void sd_shutdown(struct device *);
 static int sd_suspend(struct device *);
 static int sd_resume(struct device *);
+static int sd_resume_async(struct device *);
 static void sd_rescan(struct device *);
 static int sd_done(struct scsi_cmnd *);
 static int sd_eh_action(struct scsi_cmnd *, unsigned char *, int, int);
@@ -484,7 +485,7 @@ static struct class sd_disk_class = {
 
 static const struct dev_pm_ops sd_pm_ops = {
 	.suspend		= sd_suspend,
-	.resume			= sd_resume,
+	.resume			= sd_resume_async,
 	.poweroff		= sd_suspend,
 	.restore		= sd_resume,
 	.runtime_suspend	= sd_suspend,
@@ -3137,6 +3138,85 @@ done:
 	return ret;
 }
 
+static void sd_resume_async_end(struct request *rq, int error)
+{
+	struct scsi_sense_hdr sshdr;
+	struct scsi_disk *sdkp = rq->end_io_data;
+	char *sense = rq->sense;
+
+	if (error) {
+		sd_printk(KERN_WARNING, sdkp, "START FAILED\n");
+		sd_print_result(sdkp, error);
+		if (sense && (driver_byte(error) & DRIVER_SENSE)) {
+			scsi_normalize_sense(sense,
+				SCSI_SENSE_BUFFERSIZE, &sshdr);
+			sd_print_sense_hdr(sdkp, &sshdr);
+		}
+	} else
+		sd_printk(KERN_NOTICE, sdkp, "START SUCCESS\n");
+
+	kfree(sense);
+	rq->sense = NULL;
+	rq->end_io_data = NULL;
+	__blk_put_request(rq->q, rq);
+	scsi_disk_put(sdkp);
+}
+
+static int sd_resume_async(struct device *dev)
+{
+	unsigned char cmd[6] = { START_STOP };
+	struct scsi_disk *sdkp = scsi_disk_get_from_dev(dev);
+	struct request *req;
+	char *sense = NULL;
+	int ret = 0;
+
+	if (!sdkp->device->manage_start_stop)
+		goto error;
+
+	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
+
+	cmd[4] |= 1;
+
+	if (sdkp->device->start_stop_pwr_cond)
+		cmd[4] |= 1 << 4;	/* Active or Standby */
+
+	if (!scsi_device_online(sdkp->device)) {
+		ret = -ENODEV;
+		goto error;
+	}
+
+	req = blk_get_request(sdkp->device->request_queue, 0, __GFP_WAIT);
+	if (!req) {
+		ret = -ENOMEM;
+		goto error;
+	}
+
+	sense = kzalloc(SCSI_SENSE_BUFFERSIZE, GFP_NOIO);
+	if (!sense) {
+		ret = -ENOMEM;
+		goto error_sense;
+	}
+
+	req->cmd_len = COMMAND_SIZE(cmd[0]);
+	memcpy(req->cmd, cmd, req->cmd_len);
+	req->sense = sense;
+	req->sense_len = 0;
+	req->retries = SD_MAX_RETRIES;
+	req->timeout = SD_TIMEOUT;
+	req->cmd_type = REQ_TYPE_BLOCK_PC;
+	req->cmd_flags |= REQ_PM | REQ_QUIET | REQ_PREEMPT;
+
+	req->end_io_data = sdkp;
+	blk_execute_rq_nowait(req->q, NULL, req, 1, sd_resume_async_end);
+	return 0;
+
+ error_sense:
+	__blk_put_request(req->q, req);
+ error:
+	scsi_disk_put(sdkp);
+	return ret;
+}
+
 /**
  *	init_sd - entry point for this driver (both when built in or when
  *	a module).

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Sergei Shtylyov]

Hello.

On 06-09-2013 4:44, Todd E Brandt wrote:

> Part 2 of the hard disk resume optimization patch, this one applies
> to the scsi subsystem.

    Don't give the same name to both patches. In this case, you could prefix 
the patch name with e.g. "sd: ".

> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
> Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>

>   drivers/scsi/sd.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>   1 file changed, 81 insertions(+), 1 deletion(-)

> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index 86fcf2c..d4bf784 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
[...]
> @@ -3137,6 +3138,85 @@ done:
>   	return ret;
>   }
>
> +static void sd_resume_async_end(struct request *rq, int error)
> +{
> +	struct scsi_sense_hdr sshdr;
> +	struct scsi_disk *sdkp = rq->end_io_data;
> +	char *sense = rq->sense;
> +
> +	if (error) {
> +		sd_printk(KERN_WARNING, sdkp, "START FAILED\n");
> +		sd_print_result(sdkp, error);
> +		if (sense && (driver_byte(error) & DRIVER_SENSE)) {
> +			scsi_normalize_sense(sense,
> +				SCSI_SENSE_BUFFERSIZE, &sshdr);
> +			sd_print_sense_hdr(sdkp, &sshdr);
> +		}
> +	} else
> +		sd_printk(KERN_NOTICE, sdkp, "START SUCCESS\n");

    According to Documentation/CodingStyle, both arms of the *if* statement 
should have {} when one arm has them.

WBR, Sergei

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[James Bottomley]

On Thu, 2013-09-05 at 17:44 -0700, Todd E Brandt wrote:
> Part 2 of the hard disk resume optimization patch, this one applies
> to the scsi subsystem.

This is a horrible patch description, it won't tell the next reader
anything about what is going on and why.  Give the patches two separate
summaries and refer back by summary explaining what this one does.

Also, having two code paths do the same thing is always a bad idea:
someone will update one and not the other.  sd_resume() should be come
sd_resume_async(); wait for async event

James

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Bartlomiej Zolnierkiewicz]

Hi,

On Thursday, September 05, 2013 05:44:25 PM Todd E Brandt wrote:
> Part 2 of the hard disk resume optimization patch, this one applies
> to the scsi subsystem.

Please update the patch description to say what the patch does (and why).

> Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
> Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
> 
>  drivers/scsi/sd.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 81 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> index 86fcf2c..d4bf784 100644
> --- a/drivers/scsi/sd.c
> +++ b/drivers/scsi/sd.c
> @@ -107,6 +107,7 @@ static int  sd_remove(struct device *);
>  static void sd_shutdown(struct device *);
>  static int sd_suspend(struct device *);
>  static int sd_resume(struct device *);
> +static int sd_resume_async(struct device *);
>  static void sd_rescan(struct device *);
>  static int sd_done(struct scsi_cmnd *);
>  static int sd_eh_action(struct scsi_cmnd *, unsigned char *, int, int);
> @@ -484,7 +485,7 @@ static struct class sd_disk_class = {
>  
>  static const struct dev_pm_ops sd_pm_ops = {
>  	.suspend		= sd_suspend,
> -	.resume			= sd_resume,
> +	.resume			= sd_resume_async,

Same comment as with patch #1.

When the command execution later fails the error will be now lost
(->resume returned 0 already and nothing can be done to fix it).

Sorry but this looks like a wrong approach and you should fix
the PM layer to do async ->resume when possible instead.

>  	.poweroff		= sd_suspend,
>  	.restore		= sd_resume,
>  	.runtime_suspend	= sd_suspend,
> @@ -3137,6 +3138,85 @@ done:
>  	return ret;
>  }
>  
> +static void sd_resume_async_end(struct request *rq, int error)
> +{
> +	struct scsi_sense_hdr sshdr;
> +	struct scsi_disk *sdkp = rq->end_io_data;
> +	char *sense = rq->sense;
> +
> +	if (error) {
> +		sd_printk(KERN_WARNING, sdkp, "START FAILED\n");
> +		sd_print_result(sdkp, error);
> +		if (sense && (driver_byte(error) & DRIVER_SENSE)) {
> +			scsi_normalize_sense(sense,
> +				SCSI_SENSE_BUFFERSIZE, &sshdr);
> +			sd_print_sense_hdr(sdkp, &sshdr);
> +		}
> +	} else
> +		sd_printk(KERN_NOTICE, sdkp, "START SUCCESS\n");
> +
> +	kfree(sense);
> +	rq->sense = NULL;
> +	rq->end_io_data = NULL;
> +	__blk_put_request(rq->q, rq);
> +	scsi_disk_put(sdkp);
> +}
> +
> +static int sd_resume_async(struct device *dev)
> +{
> +	unsigned char cmd[6] = { START_STOP };
> +	struct scsi_disk *sdkp = scsi_disk_get_from_dev(dev);
> +	struct request *req;
> +	char *sense = NULL;
> +	int ret = 0;
> +
> +	if (!sdkp->device->manage_start_stop)
> +		goto error;
> +
> +	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
> +
> +	cmd[4] |= 1;
> +
> +	if (sdkp->device->start_stop_pwr_cond)
> +		cmd[4] |= 1 << 4;	/* Active or Standby */
> +
> +	if (!scsi_device_online(sdkp->device)) {
> +		ret = -ENODEV;
> +		goto error;
> +	}
> +
> +	req = blk_get_request(sdkp->device->request_queue, 0, __GFP_WAIT);
> +	if (!req) {
> +		ret = -ENOMEM;
> +		goto error;
> +	}
> +
> +	sense = kzalloc(SCSI_SENSE_BUFFERSIZE, GFP_NOIO);
> +	if (!sense) {
> +		ret = -ENOMEM;
> +		goto error_sense;
> +	}
> +
> +	req->cmd_len = COMMAND_SIZE(cmd[0]);
> +	memcpy(req->cmd, cmd, req->cmd_len);
> +	req->sense = sense;
> +	req->sense_len = 0;
> +	req->retries = SD_MAX_RETRIES;
> +	req->timeout = SD_TIMEOUT;
> +	req->cmd_type = REQ_TYPE_BLOCK_PC;
> +	req->cmd_flags |= REQ_PM | REQ_QUIET | REQ_PREEMPT;
> +
> +	req->end_io_data = sdkp;
> +	blk_execute_rq_nowait(req->q, NULL, req, 1, sd_resume_async_end);
> +	return 0;
> +
> + error_sense:
> +	__blk_put_request(req->q, req);
> + error:
> +	scsi_disk_put(sdkp);
> +	return ret;
> +}
> +
>  /**
>   *	init_sd - entry point for this driver (both when built in or when
>   *	a module).

Best regards,
--
Bartlomiej Zolnierkiewicz
Samsung R&D Institute Poland
Samsung Electronics

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Todd E Brandt]

Sorry for the missing description, I've submitted it several times so I
just assumed that the previous entries in the thread were enough. I'll
resubmit with a full description.

As for replacing sd_resume altogether, the reason I didn't is because
there are three callbacks which use sd_resume. I've only tested this
functionality for the system resume call, not restore or runtime_resume,
and I didn't know if the other two made sense as asynchronous since
it doesn't really buy anything in runtime_resume, but I understand your
point that it adds unnecessary complexity. I'll fix it.

I'm sure there are no issues but I'll be sure and test all the new scenerios 
before I resubmit.

So, with fixes to those two issues, do you think this technique is something 
you'd accept? I've tried it two different ways, one where I changed the pm
code to allow nonblocking resume callbacks, which just finalized system 
resume without waiting for the ata and scsi drivers to finish. But that 
didn't fly well with Rafael as it violates some implicit assumptions 
about how the callbacks function. I worked with Rafael to come up with a 
"skip_resume" version which just skipped the system resume of the disks, 
changed them to runtime suspended, and let them runtime resume whenever 
they were accessed, but it proved to be impractical in a real OS environment 
because the OS tends to access everything at once when it comes back online.

This approach is really the only remaining way (that I can think of) to 
remove the ata-port wakeup delay in software, so given the massive 
performance benefit I hope you'll consider it. Thanks for the reply!

On Fri, Sep 06, 2013 at 08:10:34AM -0700, James Bottomley wrote:
> On Thu, 2013-09-05 at 17:44 -0700, Todd E Brandt wrote:
> > Part 2 of the hard disk resume optimization patch, this one applies
> > to the scsi subsystem.
> 
> This is a horrible patch description, it won't tell the next reader
> anything about what is going on and why.  Give the patches two separate
> summaries and refer back by summary explaining what this one does.
> 
> Also, having two code paths do the same thing is always a bad idea:
> someone will update one and not the other.  sd_resume() should be come
> sd_resume_async(); wait for async event
> 
> James
> 
> 

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Todd E Brandt]

I'll fix those, thanks for the feedback. 

On Fri, Sep 06, 2013 at 04:37:00PM +0400, Sergei Shtylyov wrote:
> Hello.
> 
> On 06-09-2013 4:44, Todd E Brandt wrote:
> 
> >Part 2 of the hard disk resume optimization patch, this one applies
> >to the scsi subsystem.
> 
>    Don't give the same name to both patches. In this case, you could
> prefix the patch name with e.g. "sd: ".
> 
> >Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
> >Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
> 
> >  drivers/scsi/sd.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> >  1 file changed, 81 insertions(+), 1 deletion(-)
> 
> >diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> >index 86fcf2c..d4bf784 100644
> >--- a/drivers/scsi/sd.c
> >+++ b/drivers/scsi/sd.c
> [...]
> >@@ -3137,6 +3138,85 @@ done:
> >  	return ret;
> >  }
> >
> >+static void sd_resume_async_end(struct request *rq, int error)
> >+{
> >+	struct scsi_sense_hdr sshdr;
> >+	struct scsi_disk *sdkp = rq->end_io_data;
> >+	char *sense = rq->sense;
> >+
> >+	if (error) {
> >+		sd_printk(KERN_WARNING, sdkp, "START FAILED\n");
> >+		sd_print_result(sdkp, error);
> >+		if (sense && (driver_byte(error) & DRIVER_SENSE)) {
> >+			scsi_normalize_sense(sense,
> >+				SCSI_SENSE_BUFFERSIZE, &sshdr);
> >+			sd_print_sense_hdr(sdkp, &sshdr);
> >+		}
> >+	} else
> >+		sd_printk(KERN_NOTICE, sdkp, "START SUCCESS\n");
> 
>    According to Documentation/CodingStyle, both arms of the *if*
> statement should have {} when one arm has them.
> 
> WBR, Sergei
> 

Re: [PATCH/RESEND 2/2] Hard disk S3 resume time optimization

[Todd E Brandt]

I'll resubmit with a better description. As for the return value, if you look
at the mailing list history you'll see that I already tried that approach (and 
was actually advised to do it this way):

http://marc.info/?l=linux-scsi&m=136881989601276&w=2

On Fri, Sep 06, 2013 at 07:03:26PM +0200, Bartlomiej Zolnierkiewicz wrote:
> 
> Hi,
> 
> On Thursday, September 05, 2013 05:44:25 PM Todd E Brandt wrote:
> > Part 2 of the hard disk resume optimization patch, this one applies
> > to the scsi subsystem.
> 
> Please update the patch description to say what the patch does (and why).
> 
> > Signed-off-by: Todd Brandt <todd.e.brandt@intel.com>
> > Signed-off-by: Arjan van de Ven <arjan@linux.intel.com>
> > 
> >  drivers/scsi/sd.c | 82 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
> >  1 file changed, 81 insertions(+), 1 deletion(-)
> > 
> > diff --git a/drivers/scsi/sd.c b/drivers/scsi/sd.c
> > index 86fcf2c..d4bf784 100644
> > --- a/drivers/scsi/sd.c
> > +++ b/drivers/scsi/sd.c
> > @@ -107,6 +107,7 @@ static int  sd_remove(struct device *);
> >  static void sd_shutdown(struct device *);
> >  static int sd_suspend(struct device *);
> >  static int sd_resume(struct device *);
> > +static int sd_resume_async(struct device *);
> >  static void sd_rescan(struct device *);
> >  static int sd_done(struct scsi_cmnd *);
> >  static int sd_eh_action(struct scsi_cmnd *, unsigned char *, int, int);
> > @@ -484,7 +485,7 @@ static struct class sd_disk_class = {
> >  
> >  static const struct dev_pm_ops sd_pm_ops = {
> >  	.suspend		= sd_suspend,
> > -	.resume			= sd_resume,
> > +	.resume			= sd_resume_async,
> 
> Same comment as with patch #1.
> 
> When the command execution later fails the error will be now lost
> (->resume returned 0 already and nothing can be done to fix it).
> 
> Sorry but this looks like a wrong approach and you should fix
> the PM layer to do async ->resume when possible instead.
> 
> >  	.poweroff		= sd_suspend,
> >  	.restore		= sd_resume,
> >  	.runtime_suspend	= sd_suspend,
> > @@ -3137,6 +3138,85 @@ done:
> >  	return ret;
> >  }
> >  
> > +static void sd_resume_async_end(struct request *rq, int error)
> > +{
> > +	struct scsi_sense_hdr sshdr;
> > +	struct scsi_disk *sdkp = rq->end_io_data;
> > +	char *sense = rq->sense;
> > +
> > +	if (error) {
> > +		sd_printk(KERN_WARNING, sdkp, "START FAILED\n");
> > +		sd_print_result(sdkp, error);
> > +		if (sense && (driver_byte(error) & DRIVER_SENSE)) {
> > +			scsi_normalize_sense(sense,
> > +				SCSI_SENSE_BUFFERSIZE, &sshdr);
> > +			sd_print_sense_hdr(sdkp, &sshdr);
> > +		}
> > +	} else
> > +		sd_printk(KERN_NOTICE, sdkp, "START SUCCESS\n");
> > +
> > +	kfree(sense);
> > +	rq->sense = NULL;
> > +	rq->end_io_data = NULL;
> > +	__blk_put_request(rq->q, rq);
> > +	scsi_disk_put(sdkp);
> > +}
> > +
> > +static int sd_resume_async(struct device *dev)
> > +{
> > +	unsigned char cmd[6] = { START_STOP };
> > +	struct scsi_disk *sdkp = scsi_disk_get_from_dev(dev);
> > +	struct request *req;
> > +	char *sense = NULL;
> > +	int ret = 0;
> > +
> > +	if (!sdkp->device->manage_start_stop)
> > +		goto error;
> > +
> > +	sd_printk(KERN_NOTICE, sdkp, "Starting disk\n");
> > +
> > +	cmd[4] |= 1;
> > +
> > +	if (sdkp->device->start_stop_pwr_cond)
> > +		cmd[4] |= 1 << 4;	/* Active or Standby */
> > +
> > +	if (!scsi_device_online(sdkp->device)) {
> > +		ret = -ENODEV;
> > +		goto error;
> > +	}
> > +
> > +	req = blk_get_request(sdkp->device->request_queue, 0, __GFP_WAIT);
> > +	if (!req) {
> > +		ret = -ENOMEM;
> > +		goto error;
> > +	}
> > +
> > +	sense = kzalloc(SCSI_SENSE_BUFFERSIZE, GFP_NOIO);
> > +	if (!sense) {
> > +		ret = -ENOMEM;
> > +		goto error_sense;
> > +	}
> > +
> > +	req->cmd_len = COMMAND_SIZE(cmd[0]);
> > +	memcpy(req->cmd, cmd, req->cmd_len);
> > +	req->sense = sense;
> > +	req->sense_len = 0;
> > +	req->retries = SD_MAX_RETRIES;
> > +	req->timeout = SD_TIMEOUT;
> > +	req->cmd_type = REQ_TYPE_BLOCK_PC;
> > +	req->cmd_flags |= REQ_PM | REQ_QUIET | REQ_PREEMPT;
> > +
> > +	req->end_io_data = sdkp;
> > +	blk_execute_rq_nowait(req->q, NULL, req, 1, sd_resume_async_end);
> > +	return 0;
> > +
> > + error_sense:
> > +	__blk_put_request(req->q, req);
> > + error:
> > +	scsi_disk_put(sdkp);
> > +	return ret;
> > +}
> > +
> >  /**
> >   *	init_sd - entry point for this driver (both when built in or when
> >   *	a module).
> 
> Best regards,
> --
> Bartlomiej Zolnierkiewicz
> Samsung R&D Institute Poland
> Samsung Electronics
>