atapi_dmadir: why do we not (also) auto-detect it from IDENTIFY PACKET ??

atapi_dmadir: why do we not (also) auto-detect it from IDENTIFY PACKET ??

[Mark Lord]

Albert / Jeff,

Way back in 2.6.17-rc2, you guys signed-off on 95de719adc94392a95c3c4d0a2d6b8b1ea39d236,
which added a libata module parameter for atapi_dmadir.

That's nice, but many modern drives (or bridge chips) flag this in
their IDENTIFY PACKET response, as bit-15 of word-62 of the returned data.

So for those at lease, we should be automatically setting atapi_dmadir=1.
But on a per-device basis, not globally as done now.

Albert: do you want to cook up a bug fix for this?

Cheers

[PATCH] libata-scsi: automatically use DMADIR if drive/bridge requires it

[Mark Lord]

Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.

That's nice, but most SATA devices which need it will tell us about it
in their IDENTIFY PACKET response, as bit-15 of word-62 of the
returned data (as per ATA7, ATA8 specifications).

So for those which specify it, we should automatically use the DMADIR bit.
Otherwise, disc writing will fail by default on many SATA-ATAPI drives.

Signed-off-by: Mark Lord <mlord@pobox.com>

--- old/drivers/ata/libata-scsi.c	2008-02-18 14:45:08.000000000 -0500
+++ linux/drivers/ata/libata-scsi.c	2008-02-18 14:53:02.000000000 -0500
@@ -2543,9 +2543,10 @@
 		qc->tf.protocol = ATAPI_PROT_DMA;
 		qc->tf.feature |= ATAPI_PKT_DMA;
 
-		if (atapi_dmadir && (scmd->sc_data_direction != DMA_TO_DEVICE))
-			/* some SATA bridges need us to indicate data xfer direction */
-			qc->tf.feature |= ATAPI_DMADIR;
+		/* some SATA bridges need us to indicate data xfer direction */
+		if (atapi_dmadir || (dev->id[62] & 0x8000))
+			if (scmd->sc_data_direction != DMA_TO_DEVICE)
+				qc->tf.feature |= ATAPI_DMADIR;
 	}
 
 

Re: [PATCH] libata-scsi: automatically use DMADIR if drive/bridge requires it

[Alan Cox]

On Mon, Feb 18, 2008 at 03:53:46PM -0500, Mark Lord wrote:
> +		/* some SATA bridges need us to indicate data xfer direction 
> */
> +		if (atapi_dmadir || (dev->id[62] & 0x8000))

The rest of the code uses ata_ inlines/defines in ata.h for things like this
and religiously checks validity bits and ATA versions.

Re: [PATCH] libata-scsi: automatically use DMADIR if drive/bridge requires it

[Mark Lord]

Alan Cox wrote:
> On Mon, Feb 18, 2008 at 03:53:46PM -0500, Mark Lord wrote:
>> +		/* some SATA bridges need us to indicate data xfer direction 
>> */
>> +		if (atapi_dmadir || (dev->id[62] & 0x8000))
> 
> The rest of the code uses ata_ inlines/defines in ata.h for things like this
> and religiously checks validity bits and ATA versions.
..

Feel free.  I don't actually have a device like that here to test on,
so I'm not particularly motivated about doing more than the above right now.

Perhaps you or Albert may be, though.  :)

That particular bit has always been zero (reserved), from ATA1 through present,
except when used as a DMADIR bit, beginning somewhere around ATA7.

Cheers

Re: [PATCH] libata-scsi: automatically use DMADIR if drive/bridge requires it

[Jeff Garzik]

Mark Lord wrote:
> Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.
> 
> That's nice, but most SATA devices which need it will tell us about it
> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
> returned data (as per ATA7, ATA8 specifications).
> 
> So for those which specify it, we should automatically use the DMADIR bit.
> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
> 
> Signed-off-by: Mark Lord <mlord@pobox.com>
> 
> --- old/drivers/ata/libata-scsi.c    2008-02-18 14:45:08.000000000 -0500
> +++ linux/drivers/ata/libata-scsi.c    2008-02-18 14:53:02.000000000 -0500
> @@ -2543,9 +2543,10 @@
>         qc->tf.protocol = ATAPI_PROT_DMA;
>         qc->tf.feature |= ATAPI_PKT_DMA;
> 
> -        if (atapi_dmadir && (scmd->sc_data_direction != DMA_TO_DEVICE))
> -            /* some SATA bridges need us to indicate data xfer 
> direction */
> -            qc->tf.feature |= ATAPI_DMADIR;
> +        /* some SATA bridges need us to indicate data xfer direction */
> +        if (atapi_dmadir || (dev->id[62] & 0x8000))
> +            if (scmd->sc_data_direction != DMA_TO_DEVICE)
> +                qc->tf.feature |= ATAPI_DMADIR;

Regardless of the historical usage (or lack thereof) of that I.D. word, 
I would prefer

1) an ata version check

2) a simple wrapper in linux/ata.h, similar to the many that already exist

So, conditional ACK...

Re: [PATCH] libata-scsi: automatically use DMADIR if drive/bridge requires it

[Mark Lord]

Jeff Garzik wrote:
> Mark Lord wrote:
>> Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.
>>
>> That's nice, but most SATA devices which need it will tell us about it
>> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
>> returned data (as per ATA7, ATA8 specifications).
>>
>> So for those which specify it, we should automatically use the DMADIR 
>> bit.
>> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
>>
>> Signed-off-by: Mark Lord <mlord@pobox.com>
>>
>> --- old/drivers/ata/libata-scsi.c    2008-02-18 14:45:08.000000000 -0500
>> +++ linux/drivers/ata/libata-scsi.c    2008-02-18 14:53:02.000000000 
>> -0500
>> @@ -2543,9 +2543,10 @@
>>         qc->tf.protocol = ATAPI_PROT_DMA;
>>         qc->tf.feature |= ATAPI_PKT_DMA;
>>
>> -        if (atapi_dmadir && (scmd->sc_data_direction != DMA_TO_DEVICE))
>> -            /* some SATA bridges need us to indicate data xfer 
>> direction */
>> -            qc->tf.feature |= ATAPI_DMADIR;
>> +        /* some SATA bridges need us to indicate data xfer direction */
>> +        if (atapi_dmadir || (dev->id[62] & 0x8000))
>> +            if (scmd->sc_data_direction != DMA_TO_DEVICE)
>> +                qc->tf.feature |= ATAPI_DMADIR;
> 
> Regardless of the historical usage (or lack thereof) of that I.D. word, 
> I would prefer
> 
> 1) an ata version check
> 
> 2) a simple wrapper in linux/ata.h, similar to the many that already exist
> 
> So, conditional ACK...
..

Good.  As I said, I don't expect to be doing this,
so if anyone else wants to fix this issue in libata
they are more than welcome to take the above and build on it.

Cheers 

[PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Tejun Heo]

Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.

That's nice, but most SATA devices which need it will tell us about it
in their IDENTIFY PACKET response, as bit-15 of word-62 of the
returned data (as per ATA7, ATA8 specifications).

So for those which specify it, we should automatically use the DMADIR bit.
Otherwise, disc writing will fail by default on many SATA-ATAPI drives.

This patch adds ATA_DFLAG_DMADIR and make ata_dev_configure() set it
if atapi_dmadir is set or identify data indicates DMADIR is necessary.
atapi_xlat() is converted to check ATA_DFLAG_DMADIR before setting
DMADIR.

Original patch is from Mark Lord.

Signed-off-by: Tejun Heo <htejun@gmail.com>
Cc: Mark Lord <mlord@pobox.com>
---
I don't have a bridge which sets DMADIR but so only checked atapi_dmadir
parameter.  Thanks.

 drivers/ata/libata-core.c |   11 +++++++++--
 drivers/ata/libata-scsi.c |    3 ++-
 include/linux/ata.h       |    5 +++++
 include/linux/libata.h    |    1 +
 4 files changed, 17 insertions(+), 3 deletions(-)

diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index 3011919..d38f113 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -2200,6 +2200,7 @@ int ata_dev_configure(struct ata_device *dev)
 	else if (dev->class == ATA_DEV_ATAPI) {
 		const char *cdb_intr_string = "";
 		const char *atapi_an_string = "";
+		const char *dma_dir_string = "";
 		u32 sntf;
 
 		rc = atapi_cdb_len(id);
@@ -2240,13 +2241,19 @@ int ata_dev_configure(struct ata_device *dev)
 			cdb_intr_string = ", CDB intr";
 		}
 
+		if (atapi_dmadir || atapi_id_dmadir(dev->id)) {
+			dev->flags |= ATA_DFLAG_DMADIR;
+			dma_dir_string = ", DMADIR";
+		}
+
 		/* print device info to dmesg */
 		if (ata_msg_drv(ap) && print_info)
 			ata_dev_printk(dev, KERN_INFO,
-				       "ATAPI: %s, %s, max %s%s%s\n",
+				       "ATAPI: %s, %s, max %s%s%s%s\n",
 				       modelbuf, fwrevbuf,
 				       ata_mode_string(xfer_mask),
-				       cdb_intr_string, atapi_an_string);
+				       cdb_intr_string, atapi_an_string,
+				       dma_dir_string);
 	}
 
 	/* determine max_sectors */
diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index c02c490..f50b96b 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -2543,7 +2543,8 @@ static unsigned int atapi_xlat(struct ata_queued_cmd *qc)
 		qc->tf.protocol = ATAPI_PROT_DMA;
 		qc->tf.feature |= ATAPI_PKT_DMA;
 
-		if (atapi_dmadir && (scmd->sc_data_direction != DMA_TO_DEVICE))
+		if ((dev->flags & ATA_DFLAG_DMADIR) &&
+		    (scmd->sc_data_direction != DMA_TO_DEVICE))
 			/* some SATA bridges need us to indicate data xfer direction */
 			qc->tf.feature |= ATAPI_DMADIR;
 	}
diff --git a/include/linux/ata.h b/include/linux/ata.h
index 78bbaca..1c622e2 100644
--- a/include/linux/ata.h
+++ b/include/linux/ata.h
@@ -659,6 +659,11 @@ static inline int atapi_command_packet_set(const u16 *dev_id)
 	return (dev_id[0] >> 8) & 0x1f;
 }
 
+static inline int atapi_id_dmadir(const u16 *dev_id)
+{
+	return ata_id_major_version(dev_id) >= 7 && (dev_id[62] & 0x8000);
+}
+
 static inline int is_multi_taskfile(struct ata_taskfile *tf)
 {
 	return (tf->command == ATA_CMD_READ_MULTI) ||
diff --git a/include/linux/libata.h b/include/linux/libata.h
index bc5a8d0..b942e90 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -138,6 +138,7 @@ enum {
 	ATA_DFLAG_AN		= (1 << 7), /* AN configured */
 	ATA_DFLAG_HIPM		= (1 << 8), /* device supports HIPM */
 	ATA_DFLAG_DIPM		= (1 << 9), /* device supports DIPM */
+	ATA_DFLAG_DMADIR	= (1 << 10), /* device requires DMADIR */
 	ATA_DFLAG_CFG_MASK	= (1 << 12) - 1,
 
 	ATA_DFLAG_PIO		= (1 << 12), /* device limited to PIO mode */

Re: [PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Albert Lee]

Tejun Heo wrote:
> Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.
> 
> That's nice, but most SATA devices which need it will tell us about it
> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
> returned data (as per ATA7, ATA8 specifications).
> 
> So for those which specify it, we should automatically use the DMADIR bit.
> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
> 
> This patch adds ATA_DFLAG_DMADIR and make ata_dev_configure() set it
> if atapi_dmadir is set or identify data indicates DMADIR is necessary.
> atapi_xlat() is converted to check ATA_DFLAG_DMADIR before setting
> DMADIR.
> 
> Original patch is from Mark Lord.
> 
> Signed-off-by: Tejun Heo <htejun@gmail.com>
> Cc: Mark Lord <mlord@pobox.com>
> ---
> I don't have a bridge which sets DMADIR but so only checked atapi_dmadir
> parameter.  Thanks.
> 

The patch looks good. However, it seems there is no realworld IDE-to-SATA
bridge that requires DMADIR and also mangles IDENTIFY PACKET bit-15 of
word-62 to indicate its presence.

>From the previous test of the IDE-to-SATA bridges, only the Sil 3611
requires the host software to hint on ATAPI DMADIR. But Sil 3611 doesn't
mangle IDENTIFY PACKET word-62:
http://www.spinics.net/lists/linux-ide/msg01514.html

Maybe we could use ata_dev_knobble() instead? i.e. If it's an ATAPI
device && ata_dev_knobble() then we turn on DMADIR (regardless the
bridge chip used since setting DMADIR won't hurt those don't need it.)
--
albert

Re: [PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Mark Lord]

Albert Lee wrote:
> Tejun Heo wrote:
>> Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.
>>
>> That's nice, but most SATA devices which need it will tell us about it
>> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
>> returned data (as per ATA7, ATA8 specifications).
>>
>> So for those which specify it, we should automatically use the DMADIR bit.
>> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
>>
>> This patch adds ATA_DFLAG_DMADIR and make ata_dev_configure() set it
>> if atapi_dmadir is set or identify data indicates DMADIR is necessary.
>> atapi_xlat() is converted to check ATA_DFLAG_DMADIR before setting
>> DMADIR.
>>
>> Original patch is from Mark Lord.
>>
>> Signed-off-by: Tejun Heo <htejun@gmail.com>
>> Cc: Mark Lord <mlord@pobox.com>
>> ---
>> I don't have a bridge which sets DMADIR but so only checked atapi_dmadir
>> parameter.  Thanks.
>>
> 
> The patch looks good. However, it seems there is no realworld IDE-to-SATA
> bridge that requires DMADIR and also mangles IDENTIFY PACKET bit-15 of
> word-62 to indicate its presence.
> 
>>From the previous test of the IDE-to-SATA bridges, only the Sil 3611
> requires the host software to hint on ATAPI DMADIR. But Sil 3611 doesn't
> mangle IDENTIFY PACKET word-62:
> http://www.spinics.net/lists/linux-ide/msg01514.html
..

Ahh, great!  We have "history" here!

But there are more SATA bridges out there than the ones Alber has there
(SiI 3611/3811, Marvell 88i8030/88SA8040, Acard ARC770, JMicron JM20330).

There's the Marvell 88SA8050, and I have a funky looking thing here
that says "SataLink SPiF223A" on it (????).  I wonder what it does?
And how many other ones are there?

The Word62 patch (from Tejun) is simple enough that the "complexity"
point from before really doesn't hold water.

To be ATA compliant, we really should add it in.
But yes, it would also be good to see it in action.

That's more likely to happen in the general user community,
than in our limited circle of kernel folk with limited hardware.

Cheers

Re: [PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Mark Lord]

Mark Lord wrote:
> I have a funky looking thing here
> that says "SataLink SPiF223A" on it (????).  I wonder what it does?
..

Well, after some experimentation, this one appears to be ATA-only,
with no ATAPI support at all.

Cheers

Re: [PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Jeff Garzik]

Mark Lord wrote:
> Albert Lee wrote:
>> Tejun Heo wrote:
>>> Back in 2.6.17-rc2, a libata module parameter was added for 
>>> atapi_dmadir.
>>>
>>> That's nice, but most SATA devices which need it will tell us about it
>>> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
>>> returned data (as per ATA7, ATA8 specifications).
>>>
>>> So for those which specify it, we should automatically use the DMADIR 
>>> bit.
>>> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
>>>
>>> This patch adds ATA_DFLAG_DMADIR and make ata_dev_configure() set it
>>> if atapi_dmadir is set or identify data indicates DMADIR is necessary.
>>> atapi_xlat() is converted to check ATA_DFLAG_DMADIR before setting
>>> DMADIR.
>>>
>>> Original patch is from Mark Lord.
>>>
>>> Signed-off-by: Tejun Heo <htejun@gmail.com>
>>> Cc: Mark Lord <mlord@pobox.com>
>>> ---
>>> I don't have a bridge which sets DMADIR but so only checked atapi_dmadir
>>> parameter.  Thanks.
>>>
>>
>> The patch looks good. However, it seems there is no realworld IDE-to-SATA
>> bridge that requires DMADIR and also mangles IDENTIFY PACKET bit-15 of
>> word-62 to indicate its presence.
>>
>>> From the previous test of the IDE-to-SATA bridges, only the Sil 3611
>> requires the host software to hint on ATAPI DMADIR. But Sil 3611 doesn't
>> mangle IDENTIFY PACKET word-62:
>> http://www.spinics.net/lists/linux-ide/msg01514.html
> ..
> 
> Ahh, great!  We have "history" here!
> 
> But there are more SATA bridges out there than the ones Alber has there
> (SiI 3611/3811, Marvell 88i8030/88SA8040, Acard ARC770, JMicron JM20330).
> 
> There's the Marvell 88SA8050, and I have a funky looking thing here
> that says "SataLink SPiF223A" on it (????).  I wonder what it does?
> And how many other ones are there?
> 
> The Word62 patch (from Tejun) is simple enough that the "complexity"
> point from before really doesn't hold water.

I'm going to apply Tejun's patch.  It's fine (as I noted I conditionally 
ACK'd your original patch).

My opinion, though, it that this is merely coding to the spec, rather 
than making a bunch of real-world hardware work.  All the SATA<->PATA 
bridges I have, that support ATAPI, do not flag DMADIR.

	Jeff

Re: [PATCH] libata: automatically use DMADIR if drive/bridge requires it

[Jeff Garzik]

Tejun Heo wrote:
> Back in 2.6.17-rc2, a libata module parameter was added for atapi_dmadir.
> 
> That's nice, but most SATA devices which need it will tell us about it
> in their IDENTIFY PACKET response, as bit-15 of word-62 of the
> returned data (as per ATA7, ATA8 specifications).
> 
> So for those which specify it, we should automatically use the DMADIR bit.
> Otherwise, disc writing will fail by default on many SATA-ATAPI drives.
> 
> This patch adds ATA_DFLAG_DMADIR and make ata_dev_configure() set it
> if atapi_dmadir is set or identify data indicates DMADIR is necessary.
> atapi_xlat() is converted to check ATA_DFLAG_DMADIR before setting
> DMADIR.
> 
> Original patch is from Mark Lord.
> 
> Signed-off-by: Tejun Heo <htejun@gmail.com>
> Cc: Mark Lord <mlord@pobox.com>
> ---
> I don't have a bridge which sets DMADIR but so only checked atapi_dmadir
> parameter.  Thanks.
> 
>  drivers/ata/libata-core.c |   11 +++++++++--
>  drivers/ata/libata-scsi.c |    3 ++-
>  include/linux/ata.h       |    5 +++++
>  include/linux/libata.h    |    1 +
>  4 files changed, 17 insertions(+), 3 deletions(-)

applied