[PATCH] libata: fix boot panic with SATAPI devices on non-SFF HBAs

[PATCH] libata: fix boot panic with SATAPI devices on non-SFF HBAs

[James Bottomley]

Current 2.6.25 git head now panics reliably on boot if you have a SATAPI
device connected.

The problem was introduced by the libata merge trying to pull out all
the SFF code into a separate module.  Unfortunately, if you're a satapi
device you usually need to call atapi_request_sense, which has a bare
invocation of a SFF callback which is NULL on non-SFF HBAs.  Fix this by
making the call conditional.

Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>

---

diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index 784ac41..f3c69a8 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -2390,7 +2390,8 @@ static void atapi_request_sense(struct ata_queued_cmd *qc)
 	memset(cmd->sense_buffer, 0, SCSI_SENSE_BUFFERSIZE);
 
 #ifdef CONFIG_ATA_SFF
-	ap->ops->sff_tf_read(ap, &qc->tf);
+	if (ap->ops->sff_tf_read)
+		ap->ops->sff_tf_read(ap, &qc->tf);
 #endif
 
 	/* fill these in, for the case where they are -not- overwritten */

Re: [PATCH] libata: fix boot panic with SATAPI devices on non-SFF HBAs

[walt]

James Bottomley wrote:
> Current 2.6.25 git head now panics reliably on boot if you have a SATAPI
> device connected.
>
> The problem was introduced by the libata merge trying to pull out all
> the SFF code into a separate module...

Hi James,

When I saw your commit I thought it would fix the panic introduced by Tejun's
commit <305d2a1a>, but no luck.

I have a pata drive connected to a sata controller using a plugin adapter.
After Tejun's commit the kernel panics with unknown-block (8,9) instead of
mounting /dev/sda9.  Is this the scenario your patch is intended to fix?

Re: [PATCH] libata: fix boot panic with SATAPI devices on non-SFF HBAs

[James Bottomley]

On Sat, 2008-04-19 at 06:34 -0700, walt wrote:
> James Bottomley wrote:
> > Current 2.6.25 git head now panics reliably on boot if you have a SATAPI
> > device connected.
> >
> > The problem was introduced by the libata merge trying to pull out all
> > the SFF code into a separate module...
> 
> Hi James,
> 
> When I saw your commit I thought it would fix the panic introduced by Tejun's
> commit <305d2a1a>, but no luck.
> 
> I have a pata drive connected to a sata controller using a plugin adapter.
> After Tejun's commit the kernel panics with unknown-block (8,9) instead of
> mounting /dev/sda9.  Is this the scenario your patch is intended to fix?

If it's a panic with a SATA drive, then no it wouldn't ... the problem I
fixed was in SATAPI only (That's CD/DVD/Tape devices).  If you actually
post the boot log and panic, linux-ide and Tejun might be able to help.
Your brief snippet indicates the controller isn't finding sda, I think.

James