libata-sff: Fix oops reported in kerneloops.org for pnp devices with no ctl

- Make ata_sff_altstatus private so nobody uses it by mistake
- Drop the 400nS delay from it

Add

ata_sff_irq_status	-	encapsulates the IRQ check logic

This function keeps the existing behaviour for altstatus using devices. I
actually suspect the logic was wrong before the changes but -rc isn't the
time to play with that

ata_sff_sync		-	ensure writes hit the device

Really we want an io* operation for 'is posted' eg ioisposted(ioaddr) so
that we can fix the nasty delay this causes on most systems.

- ata_sff_pause		-	400nS delay

Ensure the command hit the device and delay 400nS

- ata_sff_dma_pause

Ensure the I/O hit the device and enforce an HDMA1:0 transition delay.
Requires altstatus register exists, BUG if not so we don't risk
corruption in MWDMA modes. (UDMA the checksum will save your backside in
theory)

The only other complication then is devices with their own handlers.
rb532 can use dma_pause but scc needs to access its own altstatus
register for internal errata workarounds so directly call the drivers own
altstatus function.

Signed-off-by: Alan Cox <alan@redhat.com>
Signed-off-by: Jeff Garzik <jgarzik@redhat.com>

1 files changed, 3 insertions, 1 deletions

diff --git a/drivers/ata/pata_rb532_cf.c b/drivers/ata/pata_rb532_cf.c
index a108d259f19..f8b3ffc8ae9 100644
--- a/drivers/ata/pata_rb532_cf.c
+++ b/drivers/ata/pata_rb532_cf.c
@@ -57,7 +57,9 @@ static inline void rb532_pata_finish_io(struct ata_port *ap)
 	struct ata_host *ah = ap->host;
 	struct rb532_cf_info *info = ah->private_data;
 
-	ata_sff_altstatus(ap);
+	/* FIXME: Keep previous delay. If this is merely a fence then
+	   ata_sff_sync might be sufficient. */
+	ata_sff_dma_pause(ap);
 	ndelay(RB500_CF_IO_DELAY);
 
 	set_irq_type(info->irq, IRQ_TYPE_LEVEL_HIGH);