Re: [PATCH 1/1] libata: Handle ATAPI interrupt before CDB is sent

[Bartlomiej Zolnierkiewicz <bzolnier@gmail.com>]

Hi Albert,
Hi Jeff,

On 6/8/05, Albert Lee <albertcc@tw.ibm.com> wrote:

> @@ -3457,6 +3461,32 @@
>  {
>         u8 status, host_stat;
> 
> +       /* ATAPI: Handle the interrupt before CDB is sent */
> +       if (is_atapi_taskfile(&qc->tf) &&
> +           ap->pio_task_state != PIO_ST_CDB_SENT) {
> +               /* check whether it's our irq */
> +               host_stat = ap->ops->bmdma_status(ap);
> +
> +               if (host_stat & ATA_DMA_INTR) {
> +                       /* Some pre-ATAPI-4 devices assert INTRQ here
> +                        * if nIEN is zero. qc->dev->id[0] bits 5-6 can
> +                        * be used to identify such devices.
> +                        */
> +                       DPRINTK("ata%u: atapi interrupt handled\n", ap->id);

This looks wrong for ATA_PROT_ATAPI_[NODATA],
we can't expect host to support DMA.

> +
> +                       /* clear INTRQ */
> +                       status = ata_chk_status(ap);
> +
> +                       /* clear bmdma status irq bit */
> +                       ap->ops->irq_clear(ap);
> +
> +                       return 1;       /* irq handled */
> +               } else {
> +                       DPRINTK("ata%u: not my atapi interrupt\n", ap->id);
> +                       goto idle_irq;
> +               }
> +       }
> +
>         switch (qc->tf.protocol) {
> 
>         case ATA_PROT_DMA:

	case ATA_PROT_DMA:
	case ATA_PROT_ATAPI_DMA:
	case ATA_PROT_ATAPI:
		/* check status of DMA engine */
		host_stat = ap->ops->bmdma_status(ap);
		VPRINTK("ata%u: host_stat 0x%X\n", ap->id, host_stat);

		/* if it's not our irq... */
		if (!(host_stat & ATA_DMA_INTR))
			goto idle_irq;

		/* before we do anything else, clear DMA-Start bit */
		ap->ops->bmdma_stop(ap);

		/* fall through */

ditto (vanilla libata-core.c)

Bartlomiej