[v3,02/38] ata: add HAS_IOPORT dependencies

  In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
not being declared. We thus need to add HAS_IOPORT as dependency for
those drivers using them.

Co-developed-by: Arnd Bergmann <arnd@kernel.org>
Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
---
 drivers/ata/Kconfig | 1 +
 1 file changed, 1 insertion(+)
  

On 3/14/23 21:11, Niklas Schnelle wrote:
> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
> not being declared. We thus need to add HAS_IOPORT as dependency for
> those drivers using them.

I do not see HAS_IOPORT=y defined anywhere in 6.3-rc. Is that in linux-next ?
There is a HAS_IOPORT_MAP, but I guess it is different.

> 
> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
> ---
>  drivers/ata/Kconfig | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/ata/Kconfig b/drivers/ata/Kconfig
> index b56fba76b43f..e5e67bdc2dff 100644
> --- a/drivers/ata/Kconfig
> +++ b/drivers/ata/Kconfig
> @@ -342,6 +342,7 @@ endif # HAS_DMA
>  
>  config ATA_SFF
>  	bool "ATA SFF support (for legacy IDE and PATA)"
> +	depends on HAS_IOPORT
>  	default y
>  	help
>  	  This option adds support for ATA controllers with SFF
  

On Wed, Mar 15, 2023, at 02:23, Damien Le Moal wrote:
> On 3/14/23 21:11, Niklas Schnelle wrote:
>> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
>> not being declared. We thus need to add HAS_IOPORT as dependency for
>> those drivers using them.
>
> I do not see HAS_IOPORT=y defined anywhere in 6.3-rc. Is that in linux-next ?
> There is a HAS_IOPORT_MAP, but I guess it is different.

It's defined in patch 1 of the series, so the later patches
can't be applied into subsystem trees without that.

We can either merge patch 1 as a preparation first, or keep it
all together as a series.

      Arnd
  

On 3/15/23 15:46, Arnd Bergmann wrote:
> On Wed, Mar 15, 2023, at 02:23, Damien Le Moal wrote:
>> On 3/14/23 21:11, Niklas Schnelle wrote:
>>> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
>>> not being declared. We thus need to add HAS_IOPORT as dependency for
>>> those drivers using them.
>>
>> I do not see HAS_IOPORT=y defined anywhere in 6.3-rc. Is that in linux-next ?
>> There is a HAS_IOPORT_MAP, but I guess it is different.
> 
> It's defined in patch 1 of the series, so the later patches
> can't be applied into subsystem trees without that.
> 
> We can either merge patch 1 as a preparation first, or keep it
> all together as a series.

Got it. Either is fine with me. To allow option 2, I will send my ack.
Thanks !

> 
>       Arnd
  

On 3/14/23 21:11, Niklas Schnelle wrote:
> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
> not being declared. We thus need to add HAS_IOPORT as dependency for
> those drivers using them.
> 
> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
> ---
>  drivers/ata/Kconfig | 1 +
>  1 file changed, 1 insertion(+)
> 
> diff --git a/drivers/ata/Kconfig b/drivers/ata/Kconfig
> index b56fba76b43f..e5e67bdc2dff 100644
> --- a/drivers/ata/Kconfig
> +++ b/drivers/ata/Kconfig
> @@ -342,6 +342,7 @@ endif # HAS_DMA
>  
>  config ATA_SFF
>  	bool "ATA SFF support (for legacy IDE and PATA)"
> +	depends on HAS_IOPORT
>  	default y
>  	help
>  	  This option adds support for ATA controllers with SFF

Acked-by: Damien Le Moal <damien.lemoal@opensource.wdc.com>
  

Hi Niklas,

On Tue, Mar 14, 2023 at 1:12 PM Niklas Schnelle <schnelle@linux.ibm.com> wrote:
> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
> not being declared. We thus need to add HAS_IOPORT as dependency for
> those drivers using them.
>
> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>

Thanks for your patch!

> --- a/drivers/ata/Kconfig
> +++ b/drivers/ata/Kconfig
> @@ -342,6 +342,7 @@ endif # HAS_DMA
>
>  config ATA_SFF
>         bool "ATA SFF support (for legacy IDE and PATA)"
> +       depends on HAS_IOPORT
>         default y
>         help
>           This option adds support for ATA controllers with SFF

ATA_SFF is a dependency for lots of (S)ATA drivers.
(at least) The following don't use I/O port access:

    CONFIG_SATA_RCAR (arm/arm64)
    CONFIG_PATA_FALCON (m68k/atari and m68k/q40)
    CONFIG_PATA_GAYLE (m68k/amiga)
    CONFIG_PATA_BUDDHA (m68k/amiga)

(at least) The following can use either MMIO or I/O port accesses:

    CONFIG_PATA_PLATFORM (m68k/mac)

Gr{oetje,eeting}s,

                        Geert
  

On 3/15/23 17:39, Geert Uytterhoeven wrote:
> Hi Niklas,
> 
> On Tue, Mar 14, 2023 at 1:12 PM Niklas Schnelle <schnelle@linux.ibm.com> wrote:
>> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
>> not being declared. We thus need to add HAS_IOPORT as dependency for
>> those drivers using them.
>>
>> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
>> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
> 
> Thanks for your patch!
> 
>> --- a/drivers/ata/Kconfig
>> +++ b/drivers/ata/Kconfig
>> @@ -342,6 +342,7 @@ endif # HAS_DMA
>>
>>  config ATA_SFF
>>         bool "ATA SFF support (for legacy IDE and PATA)"
>> +       depends on HAS_IOPORT
>>         default y
>>         help
>>           This option adds support for ATA controllers with SFF
> 
> ATA_SFF is a dependency for lots of (S)ATA drivers.
> (at least) The following don't use I/O port access:
> 
>     CONFIG_SATA_RCAR (arm/arm64)
>     CONFIG_PATA_FALCON (m68k/atari and m68k/q40)
>     CONFIG_PATA_GAYLE (m68k/amiga)
>     CONFIG_PATA_BUDDHA (m68k/amiga)
> 
> (at least) The following can use either MMIO or I/O port accesses:
> 
>     CONFIG_PATA_PLATFORM (m68k/mac)

But for these arch/platforms, would there be any reason to not have HAS_IOPORT ?
It is supported right now, so we should have HAS_IOPORT for them.

> 
> Gr{oetje,eeting}s,
> 
>                         Geert
>
  

Hi Damien,

On Wed, Mar 15, 2023 at 10:12 AM Damien Le Moal
<damien.lemoal@opensource.wdc.com> wrote:
> On 3/15/23 17:39, Geert Uytterhoeven wrote:
> > On Tue, Mar 14, 2023 at 1:12 PM Niklas Schnelle <schnelle@linux.ibm.com> wrote:
> >> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
> >> not being declared. We thus need to add HAS_IOPORT as dependency for
> >> those drivers using them.
> >>
> >> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
> >> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
> >
> > Thanks for your patch!
> >
> >> --- a/drivers/ata/Kconfig
> >> +++ b/drivers/ata/Kconfig
> >> @@ -342,6 +342,7 @@ endif # HAS_DMA
> >>
> >>  config ATA_SFF
> >>         bool "ATA SFF support (for legacy IDE and PATA)"
> >> +       depends on HAS_IOPORT
> >>         default y
> >>         help
> >>           This option adds support for ATA controllers with SFF
> >
> > ATA_SFF is a dependency for lots of (S)ATA drivers.
> > (at least) The following don't use I/O port access:
> >
> >     CONFIG_SATA_RCAR (arm/arm64)
> >     CONFIG_PATA_FALCON (m68k/atari and m68k/q40)
> >     CONFIG_PATA_GAYLE (m68k/amiga)
> >     CONFIG_PATA_BUDDHA (m68k/amiga)
> >
> > (at least) The following can use either MMIO or I/O port accesses:
> >
> >     CONFIG_PATA_PLATFORM (m68k/mac)
>
> But for these arch/platforms, would there be any reason to not have HAS_IOPORT ?
> It is supported right now, so we should have HAS_IOPORT for them.

That's the point: on Amiga and Atari, HAS_IOPORT is optional, and
not related to IDE support. On Mac, it is never present.

Gr{oetje,eeting}s,

                        Geert
  

On 2023/03/15 20:36, Geert Uytterhoeven wrote:
> Hi Damien,
> 
> On Wed, Mar 15, 2023 at 10:12 AM Damien Le Moal
> <damien.lemoal@opensource.wdc.com> wrote:
>> On 3/15/23 17:39, Geert Uytterhoeven wrote:
>>> On Tue, Mar 14, 2023 at 1:12 PM Niklas Schnelle <schnelle@linux.ibm.com> wrote:
>>>> In a future patch HAS_IOPORT=n will result in inb()/outb() and friends
>>>> not being declared. We thus need to add HAS_IOPORT as dependency for
>>>> those drivers using them.
>>>>
>>>> Co-developed-by: Arnd Bergmann <arnd@kernel.org>
>>>> Signed-off-by: Niklas Schnelle <schnelle@linux.ibm.com>
>>>
>>> Thanks for your patch!
>>>
>>>> --- a/drivers/ata/Kconfig
>>>> +++ b/drivers/ata/Kconfig
>>>> @@ -342,6 +342,7 @@ endif # HAS_DMA
>>>>
>>>>  config ATA_SFF
>>>>         bool "ATA SFF support (for legacy IDE and PATA)"
>>>> +       depends on HAS_IOPORT
>>>>         default y
>>>>         help
>>>>           This option adds support for ATA controllers with SFF
>>>
>>> ATA_SFF is a dependency for lots of (S)ATA drivers.
>>> (at least) The following don't use I/O port access:
>>>
>>>     CONFIG_SATA_RCAR (arm/arm64)
>>>     CONFIG_PATA_FALCON (m68k/atari and m68k/q40)
>>>     CONFIG_PATA_GAYLE (m68k/amiga)
>>>     CONFIG_PATA_BUDDHA (m68k/amiga)
>>>
>>> (at least) The following can use either MMIO or I/O port accesses:
>>>
>>>     CONFIG_PATA_PLATFORM (m68k/mac)
>>
>> But for these arch/platforms, would there be any reason to not have HAS_IOPORT ?
>> It is supported right now, so we should have HAS_IOPORT for them.
> 
> That's the point: on Amiga and Atari, HAS_IOPORT is optional, and
> not related to IDE support. On Mac, it is never present.

Ah. OK. I see now. So indeed, applying the dependency on the entire ATA_SFF
group of drivers is very coarse.

Niklas,

Can you change this to apply the dependency per driver ?
  

On Thu, Mar 16, 2023, at 00:57, Damien Le Moal wrote:
> On 2023/03/15 20:36, Geert Uytterhoeven wrote:

> Ah. OK. I see now. So indeed, applying the dependency on the entire ATA_SFF
> group of drivers is very coarse.


> Can you change this to apply the dependency per driver ?

I think that will fail to build because of this function
on architectures that drop their non-functional
inb/outb helpers:

int ata_pci_bmdma_clear_simplex(struct pci_dev *pdev)
{
        unsigned long bmdma = pci_resource_start(pdev, 4);
        u8 simplex;
 
        if (bmdma == 0)
                return -ENOENT;
 
        simplex = inb(bmdma + 0x02); 
        outb(simplex & 0x60, bmdma + 0x02);
        simplex = inb(bmdma + 0x02);
        if (simplex & 0x80)
                return -EOPNOTSUPP;
        return 0;
}

This is only called from five pata drivers (ali, amd,
cmd64x, netcell, serverworks), so an easy workaround
would be to make sure those depend on HAS_IOPORT
and enclose the function definition in an #ifdef.

        Arnd
  

On Thu, 2023-03-16 at 16:21 +0100, Arnd Bergmann wrote:
> On Thu, Mar 16, 2023, at 00:57, Damien Le Moal wrote:
> > On 2023/03/15 20:36, Geert Uytterhoeven wrote:
> 
> > Ah. OK. I see now. So indeed, applying the dependency on the entire ATA_SFF
> > group of drivers is very coarse.
> 
> 
> > Can you change this to apply the dependency per driver ?
> 
> I think that will fail to build because of this function
> on architectures that drop their non-functional
> inb/outb helpers:
> 
> int ata_pci_bmdma_clear_simplex(struct pci_dev *pdev)
> {
>         unsigned long bmdma = pci_resource_start(pdev, 4);
>         u8 simplex;
>  
>         if (bmdma == 0)
>                 return -ENOENT;
>  
>         simplex = inb(bmdma + 0x02); 
>         outb(simplex & 0x60, bmdma + 0x02);
>         simplex = inb(bmdma + 0x02);
>         if (simplex & 0x80)
>                 return -EOPNOTSUPP;
>         return 0;
> }
> 
> This is only called from five pata drivers (ali, amd,
> cmd64x, netcell, serverworks), so an easy workaround
> would be to make sure those depend on HAS_IOPORT
> and enclose the function definition in an #ifdef.
> 
>         Arnd

There were a few additional inb()/outb() uses so a few more drivers had
to have the dependency added but for v4 it will no longer be on the
ATA_SFF option and I used your suggestion above for this function.
There was another call to it in ata_generic_init_one() that is only
done for PCI_VENDOR_ID_AL so I added an #ifdef CONFIG_PATA_ALI around
that.

Thanks,
Niklas