RE: PATCH for disabling write cache on SATA drives

RE: PATCH for disabling write cache on SATA drives

[Edwards, Scott (MED, Kelly IT Resouces)]

On Monday, June 28, 2004 at 6:12 PM, Jeff Garzik wrote:

> On Fri, Jun 25, 2004 at 06:21:24PM -0500, Edwards, Scott (MED, Kelly
IT Resouces) wrote:
>> 
>> I don't know if this is useful for anyone or if anyone even cares,
but just
>> in case someone does, this is the patch I made to disable the write
cache on
>> the Serial ATA drives.  It doesn't make the write cache controllable
from
>> the user level, it just ALWAYS disables it.  (I tried to make one
that could
>> be sent via an ioctl, but there were far too many things that I
didn't
>> understand and it didn't work.)  Since I have never done this before
I don't
>> know if there is some protocol that I missed, if so I'm sorry.  If
there are
>> glaring stupid mistakes (very likely) please let me know.
>
> In principle your patch is fine, but I would really like to find a
> better way of allowing the user to do this.
>
> My immediate guess would be to use the standard SCSI method of
disabling
> write cache, and have that initiate wbcache disable on the SATA side.
>
> Long term, this would be augmented by an "ATA passthru" opcode or
ioctl.
>
>	Jeff

I first tried to make a real scsi ioctl command in libata-scsi.c, so I
could just use the scsiinfo program to disable the write cache.  I  made
a xlat_mode_select command but it didn't work and management didn't want
me
spending more time trying to figure it out.  And really for it to work
correctly there would have to be a xlat_mode_sense because now it is
just 
simulated (in ata_scsi_simulate).

I'm willing to work on it on my own time as soon as I get a SATA drive, 
if someone would be willing to advise me on how to send the ATA commands
to the drive (I think this is where it failed before).

Thanks
  -Scott

Re: PATCH for disabling write cache on SATA drives

[Jeff Garzik]

Edwards, Scott (MED, Kelly IT Resouces) wrote:
> On Monday, June 28, 2004 at 6:12 PM, Jeff Garzik wrote:
> 
> 
>>On Fri, Jun 25, 2004 at 06:21:24PM -0500, Edwards, Scott (MED, Kelly
> 
> IT Resouces) wrote:
> 
>>>I don't know if this is useful for anyone or if anyone even cares,
> 
> but just
> 
>>>in case someone does, this is the patch I made to disable the write
> 
> cache on
> 
>>>the Serial ATA drives.  It doesn't make the write cache controllable
> 
> from
> 
>>>the user level, it just ALWAYS disables it.  (I tried to make one
> 
> that could
> 
>>>be sent via an ioctl, but there were far too many things that I
> 
> didn't
> 
>>>understand and it didn't work.)  Since I have never done this before
> 
> I don't
> 
>>>know if there is some protocol that I missed, if so I'm sorry.  If
> 
> there are
> 
>>>glaring stupid mistakes (very likely) please let me know.
>>
>>In principle your patch is fine, but I would really like to find a
>>better way of allowing the user to do this.
>>
>>My immediate guess would be to use the standard SCSI method of
> 
> disabling
> 
>>write cache, and have that initiate wbcache disable on the SATA side.
>>
>>Long term, this would be augmented by an "ATA passthru" opcode or
> 
> ioctl.
> 
>>	Jeff
> 
> 
> I first tried to make a real scsi ioctl command in libata-scsi.c, so I
> could just use the scsiinfo program to disable the write cache.  I  made
> a xlat_mode_select command but it didn't work and management didn't want
> me
> spending more time trying to figure it out.  And really for it to work
> correctly there would have to be a xlat_mode_sense because now it is
> just 
> simulated (in ata_scsi_simulate).
> 
> I'm willing to work on it on my own time as soon as I get a SATA drive, 
> if someone would be willing to advise me on how to send the ATA commands
> to the drive (I think this is where it failed before).

Basically, it involves the translation of MODE SELECT, something I've 
been meaning to add, but is low priority.  If you wanted to work on 
that, that would be quite useful.

The following is a proposal for translating ATA to SCSI:
http://www.t10.org/ftp/t10/document.04/04-196r0.pdf

In that proposal, it describes the use of MODE SELECT Caching Mode Page 
(8h).

Note that, in addition to actually issuing the SET FEATURES command, to 
disable/enable writeback caching, you must update libata's internal copy 
of the IDENTIFY DEVICE page.  This involves either re-issuing IDENTIFY 
DEVICE (difficult), or simply updating the 'enabled' bit in the 
internally cached data (easy).

	Jeff

Re: PATCH for disabling write cache on SATA drives

[J. Scott Edwards]

Jeff Garzik wrote:

>
> Basically, it involves the translation of MODE SELECT, something I've been 
> meaning to add, but is low priority.  If you wanted to work on that, that 
> would be quite useful.

This is what I did, but now I'm not sure if I still have the code.  We had 
a couple of discs get trashed during testing.  I hope the code wasn't on 
one of them. ;-)  I can probably do it again anyway.

> The following is a proposal for translating ATA to SCSI:
> http://www.t10.org/ftp/t10/document.04/04-196r0.pdf
>
> In that proposal, it describes the use of MODE SELECT Caching Mode Page (8h).

I will give it a look, some nice bedtime reading :)

> Note that, in addition to actually issuing the SET FEATURES command, to 
> disable/enable writeback caching, you must update libata's internal copy of 
> the IDENTIFY DEVICE page.  This involves either re-issuing IDENTIFY DEVICE 
> (difficult), or simply updating the 'enabled' bit in the internally cached 
> data (easy).
>

I have already given this some thought.  I agree that re-issuing the 
IDENTIFY DEVICE would be difficult, but I would feel more confident that 
the SET FEATURES actually worked if it was read back.  Maybe I could do it 
in two stages, do the easy one first and just update the cached data. 
Then later wrestle with re-issuing the IDENTIFY DEVICE?

If I'm not installing my server this weekend I will buy a SATA drive and 
take a shot at it again.  If I end up installing the server this weekend 
it will be a couple of weeks before I will have time (we are moving).

Thanks
   -Scott

Re: PATCH for disabling write cache on SATA drives

[Jeff Garzik]

J. Scott Edwards wrote:
> I have already given this some thought.  I agree that re-issuing the 
> IDENTIFY DEVICE would be difficult, but I would feel more confident that 
> the SET FEATURES actually worked if it was read back.  Maybe I could do 
> it in two stages, do the easy one first and just update the cached data. 
> Then later wrestle with re-issuing the IDENTIFY DEVICE?


I would definitely suggest two stages, as I haven't yet added the 
infrastructure to re-issue commands (such as IDENTIFY DEVICE) yet :)

	Jeff

PATCH for disabling write cache on SATA drives

[Edwards, Scott (MED, Kelly IT Resouces)]

I don't know if this is useful for anyone or if anyone even cares, but just
in case someone does, this is the patch I made to disable the write cache on
the Serial ATA drives.  It doesn't make the write cache controllable from
the user level, it just ALWAYS disables it.  (I tried to make one that could
be sent via an ioctl, but there were far too many things that I didn't
understand and it didn't work.)  Since I have never done this before I don't
know if there is some protocol that I missed, if so I'm sorry.  If there are
glaring stupid mistakes (very likely) please let me know.

Thanks
  -Scott


diff -ru linux-2.6.7/drivers/scsi/libata-core.c
linux-2.6.7-dwc/drivers/scsi/libata-core.c
--- linux-2.6.7/drivers/scsi/libata-core.c	2004-06-15
23:18:38.000000000 -0600
+++ linux-2.6.7-dwc/drivers/scsi/libata-core.c	2004-06-25
16:39:16.000000000 -0600
@@ -55,6 +55,7 @@
 static void ata_host_set_udma(struct ata_port *ap);
 static void ata_dev_set_pio(struct ata_port *ap, unsigned int device);
 static void ata_dev_set_udma(struct ata_port *ap, unsigned int device);
+static void ata_dev_disable_wcache(struct ata_port *ap, unsigned int
device);
 static void ata_set_mode(struct ata_port *ap);
 static int ata_qc_issue_prot(struct ata_queued_cmd *qc);
 
@@ -1111,14 +1112,20 @@
 			dev->n_sectors = ata_id_u32(dev, 60);
 		}
 
+		if (dev->id[85] & (1 << 5)) {
+			dev->flags |= ATA_DFLAG_WCACHE;
+		}
+
 		ap->host->max_cmd_len = 16;
 
 		/* print device info to dmesg */
-		printk(KERN_INFO "ata%u: dev %u ATA, max %s, %Lu
sectors:%s\n",
+		printk(KERN_INFO 
+		       "ata%u: dev %u ATA, max %s, %Lu sectors:%s,
wcache:%s\n",
 		       ap->id, device,
 		       ata_udma_string(udma_modes),
 		       (unsigned long long)dev->n_sectors,
-		       dev->flags & ATA_DFLAG_LBA48 ? " lba48" : "");
+		       dev->flags & ATA_DFLAG_LBA48 ? " lba48" : "",
+		       dev->flags & ATA_DFLAG_WCACHE ? "on" : "off");
 	}
 
 	/* ATAPI-specific feature tests */
@@ -1297,6 +1304,12 @@
 		ata_dev_set_udma(ap, 1);
 	}
 
+	if (ap->device[0]->flags & ATA_DFLAG_WCACHE)
+		ata_dev_disable_wcache(ap, 0);
+
+	if (ap->device[1]->flags & ATA_DFLAG_WCACHE)
+		ata_dev_disable_wcache(ap, 1);
+
 	if (ap->flags & ATA_FLAG_PORT_DISABLED)
 		return;
 
@@ -1680,6 +1693,43 @@
 	ap->ops->port_disable(ap);
 }
 
+static void ata_dev_disable_wcache(struct ata_port *ap, unsigned int
device)
+{
+	struct ata_taskfile tf;
+
+	struct ata_device *dev = &ap->device[device];
+
+	if (!ata_dev_present(dev) || (ap->flags & ATA_FLAG_PORT_DISABLED))
+		return;
+
+	printk(KERN_INFO "disable write cache on dev: %u\n", device);
+
+	/* set up set-features taskfile */
+	DPRINTK("set features - disable write cache\n");
+	ata_tf_init(ap, &tf, dev->devno);
+	tf.ctl |= ATA_NIEN;
+	tf.command = ATA_CMD_SET_FEATURES;
+	tf.feature = SETFEATURES_XFER;
+	tf.flags |= ATA_TFLAG_ISADDR | ATA_TFLAG_DEVICE;
+	tf.protocol = ATA_PROT_NODATA;
+	tf.nsect = DISABLE_WRITE_CACHE;
+
+	/* do bus reset */
+	ata_tf_to_host(ap, &tf);
+
+	/* crazy ATAPI devices... */
+	if (dev->class == ATA_DEV_ATAPI)
+		msleep(150);
+
+	ata_busy_sleep(ap, ATA_TMOUT_BOOT_QUICK, ATA_TMOUT_BOOT);
+
+	ata_irq_on(ap);	/* re-enable interrupts */
+
+	ata_wait_idle(ap);
+
+	DPRINTK("EXIT\n");
+}
+
 /**
  *	ata_dev_set_xfermode - Issue SET FEATURES - XFER MODE command
  *	@ap: Port associated with device @dev
diff -ru linux-2.6.7/include/linux/ata.h linux-2.6.7-dwc/include/linux/ata.h
--- linux-2.6.7/include/linux/ata.h	2004-06-15 23:19:36.000000000 -0600
+++ linux-2.6.7-dwc/include/linux/ata.h	2004-06-25 16:19:14.000000000 -0600
@@ -133,6 +133,8 @@
 	XFER_UDMA_0		= 0x40,
 	XFER_PIO_4		= 0x0C,
 	XFER_PIO_3		= 0x0B,
+	ENABLE_WRITE_CACHE      = 0x02,
+	DISABLE_WRITE_CACHE     = 0x82,
 
 	/* ATAPI stuff */
 	ATAPI_PKT_DMA		= (1 << 0),

Re: PATCH for disabling write cache on SATA drives

[Jeff Garzik]

On Fri, Jun 25, 2004 at 06:21:24PM -0500, Edwards, Scott (MED, Kelly IT Resouces) wrote:
> 
> I don't know if this is useful for anyone or if anyone even cares, but just
> in case someone does, this is the patch I made to disable the write cache on
> the Serial ATA drives.  It doesn't make the write cache controllable from
> the user level, it just ALWAYS disables it.  (I tried to make one that could
> be sent via an ioctl, but there were far too many things that I didn't
> understand and it didn't work.)  Since I have never done this before I don't
> know if there is some protocol that I missed, if so I'm sorry.  If there are
> glaring stupid mistakes (very likely) please let me know.

In principle your patch is fine, but I would really like to find a
better way of allowing the user to do this.

My immediate guess would be to use the standard SCSI method of disabling
write cache, and have that initiate wbcache disable on the SATA side.

Long term, this would be augmented by an "ATA passthru" opcode or ioctl.

	Jeff