[RFC] libata IORDY handling

[Sergei Shtylyov <sshtylyov@ru.mvista.com>]

Hello.

Sergei Shtylyov wrote:

> Fix PIO mode 1 overclocked taskfile transfers -- probably a typo carried over
> from drivers/ide/pci/siimage.c where I've found it by documentation check...

>  drivers/ata/pata_sil680.c |    2 +-
>  1 files changed, 1 insertion(+), 1 deletion(-)
> 
> Index: linux-2.6/drivers/ata/pata_sil680.c
> ===================================================================
> --- linux-2.6.orig/drivers/ata/pata_sil680.c
> +++ linux-2.6/drivers/ata/pata_sil680.c
> @@ -135,7 +135,7 @@ static void sil680_error_handler(struct 
>  static void sil680_set_piomode(struct ata_port *ap, struct ata_device *adev)

    It's sad to say but there's another bug in this function (even a 
regression from drivers/ide/pci/siimage.c) -- the 16-bit IORDY is not enabled 
when setting PIO mode (there's code that twiddles IORDY enable but that's 
actually only for *taskfile* accesses, 16-bit IORDY is controlled by the same 
PCI  config. registers 80h/84h that enable DMA/UDMA transfer on SiI 680).
    I looked into fixing this but had a feeling that the thing wasn't right 
from the very start, including ata_pio_need_iordy().  In my understanding of 
the ANSI T13 stadrads, when one issues Set Features subcommand Set Transfer 
Mode with sector count register of 0x8 thru 0xC this means that IORDY *must* 
be enabled. That's what the ATA/ATAPI-6 says, for example:

         Table 45 - Transfer mode values
Mode                             Bits (7:3) Bits (2:0)
PIO default mode                 00000b     000b
PIO default mode, disable IORDY  00000b     001b
PIO *flow control* transfer mode 00001b     mode

    Comments?

MBR, Sergei