Re: [PATCH] ide: Fix drive's DWORD-IO handling

Re: [PATCH] ide: Fix drive's DWORD-IO handling

[Bartlomiej Zolnierkiewicz]

On Monday 15 December 2008, Mario Schwalbe wrote:
> According the documentation, id[ATA_ID_DWORD_IO] is non-zero if
> the drive supports dword IO, while the code disables support by
> setting IDE_DFLAG_NO_IO_32BIT. In addition, this word has been
> reused by the ATA8 specification. This patch fixes both cases.
> ---
>  drivers/ide/ide-probe.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
> index c55bdbd..859b0e5 100644
> --- a/drivers/ide/ide-probe.c
> +++ b/drivers/ide/ide-probe.c
> @@ -856,7 +856,7 @@ static void ide_port_tune_devices(ide_hwif_t *hwif)
>  		ide_drive_t *drive = &hwif->drives[unit];
> 
>  		if ((hwif->host_flags & IDE_HFLAG_NO_IO_32BIT) ||
> -		    drive->id[ATA_ID_DWORD_IO])
> +		    !ata_id_has_dword_io(drive->id))
>  			drive->dev_flags |= IDE_DFLAG_NO_IO_32BIT;
>  		else
>  			drive->dev_flags &= ~IDE_DFLAG_NO_IO_32BIT;

This change is OK and I would happily apply the patch but...

ata_id_has_dword_io() itself has bugs that need fixing (which would
have a nice side-effect of fixing some libata drivers that happen to
use it directly).  ATA_ID_DWORD_IO word is defined only in ATA-1 spec
and devices confirming to the later versions of spec support dword IO
just fine despite using zero value for the said word.

Thus could you please fix ata_id_has_dword_io() first (preferrably in
a separate patch) and than re-submit this patch (btw please remember to
include "Signed-off-by" line in your patches).

Thanks,
Bart

Re: [PATCH] ide: Fix drive's DWORD-IO handling

[Bartlomiej Zolnierkiewicz]

On Thursday 18 December 2008, Mario Schwalbe wrote:
> Resubmitted on request.
> 
> According the documentation, id[ATA_ID_DWORD_IO] is non-zero if
> the drive supports dword IO, while the code disables support by
> setting IDE_DFLAG_NO_IO_32BIT. In addition, this word has been
> reused by the ATA8 specification. This patch fixes both cases.
> 
> Signed-off-by: Mario Schwalbe <schwalbe@inf.tu-dresden.de>
> ---
>  drivers/ide/ide-probe.c |    2 +-
>  1 files changed, 1 insertions(+), 1 deletions(-)
> 
> diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
> index c55bdbd..859b0e5 100644
> --- a/drivers/ide/ide-probe.c
> +++ b/drivers/ide/ide-probe.c
> @@ -856,7 +856,7 @@ static void ide_port_tune_devices(ide_hwif_t *hwif)
>  		ide_drive_t *drive = &hwif->drives[unit];
> 
>  		if ((hwif->host_flags & IDE_HFLAG_NO_IO_32BIT) ||
> -		    drive->id[ATA_ID_DWORD_IO])
> +		    !ata_id_has_dword_io(drive->id))
>  			drive->dev_flags |= IDE_DFLAG_NO_IO_32BIT;
>  		else
>  			drive->dev_flags &= ~IDE_DFLAG_NO_IO_32BIT;

Since it seems that libata developers need more time to fully analyze
issues with ata_id_has_dword_io() and I don't want to delay this fix
because if it could you please modify the patch according to Sergei's
suggestion of removing id[ATA_ID_DWORD_IO] check altogether?

[ It is a safe thing to do since the old code has been enabling 32-bit
  I/O _unless_ the ATA-1 dword I/O bit for VU compatibility was set... ]

Thanks,
Bart

Re: [PATCH] ide: Fix drive's DWORD-IO handling

[Bartlomiej Zolnierkiewicz]

On Friday 19 December 2008, Bartlomiej Zolnierkiewicz wrote:
> On Thursday 18 December 2008, Mario Schwalbe wrote:
> > Resubmitted on request.
> > 
> > According the documentation, id[ATA_ID_DWORD_IO] is non-zero if
> > the drive supports dword IO, while the code disables support by
> > setting IDE_DFLAG_NO_IO_32BIT. In addition, this word has been
> > reused by the ATA8 specification. This patch fixes both cases.
> > 
> > Signed-off-by: Mario Schwalbe <schwalbe@inf.tu-dresden.de>
> > ---
> >  drivers/ide/ide-probe.c |    2 +-
> >  1 files changed, 1 insertions(+), 1 deletions(-)
> > 
> > diff --git a/drivers/ide/ide-probe.c b/drivers/ide/ide-probe.c
> > index c55bdbd..859b0e5 100644
> > --- a/drivers/ide/ide-probe.c
> > +++ b/drivers/ide/ide-probe.c
> > @@ -856,7 +856,7 @@ static void ide_port_tune_devices(ide_hwif_t *hwif)
> >  		ide_drive_t *drive = &hwif->drives[unit];
> > 
> >  		if ((hwif->host_flags & IDE_HFLAG_NO_IO_32BIT) ||
> > -		    drive->id[ATA_ID_DWORD_IO])
> > +		    !ata_id_has_dword_io(drive->id))
> >  			drive->dev_flags |= IDE_DFLAG_NO_IO_32BIT;
> >  		else
> >  			drive->dev_flags &= ~IDE_DFLAG_NO_IO_32BIT;
> 
> Since it seems that libata developers need more time to fully analyze
> issues with ata_id_has_dword_io() and I don't want to delay this fix
> because if it could you please modify the patch according to Sergei's
> suggestion of removing id[ATA_ID_DWORD_IO] check altogether?
> 
> [ It is a safe thing to do since the old code has been enabling 32-bit
>   I/O _unless_ the ATA-1 dword I/O bit for VU compatibility was set... ]

I merged the patch and did the modification while at it:

From: Mario Schwalbe <schwalbe@inf.tu-dresden.de>
Subject: [PATCH] ide: Fix drive's DWORD-IO handling

According the documentation, id[ATA_ID_DWORD_IO] is non-zero if
the drive supports dword IO, while the code disables support by
setting IDE_DFLAG_NO_IO_32BIT. In addition, this word has been
reused by the ATA8 specification. This patch fixes both cases.

Signed-off-by: Mario Schwalbe <schwalbe@inf.tu-dresden.de>
Cc: Sergei Shtylyov <sshtylyov@ru.mvista.com>
[bart: remove id[ATA_ID_DWORD_IO] check altogether per Sergei's suggestion]
Signed-off-by: Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>
---
 drivers/ide/ide-probe.c |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

Index: b/drivers/ide/ide-probe.c
===================================================================
--- a/drivers/ide/ide-probe.c
+++ b/drivers/ide/ide-probe.c
@@ -849,8 +849,7 @@ static void ide_port_tune_devices(ide_hw
 	}
 
 	ide_port_for_each_dev(i, drive, hwif) {
-		if ((hwif->host_flags & IDE_HFLAG_NO_IO_32BIT) ||
-		    drive->id[ATA_ID_DWORD_IO])
+		if (hwif->host_flags & IDE_HFLAG_NO_IO_32BIT)
 			drive->dev_flags |= IDE_DFLAG_NO_IO_32BIT;
 		else
 			drive->dev_flags &= ~IDE_DFLAG_NO_IO_32BIT;
\0