[PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Michael Schmitz]

Add Kconfig option, module parameter and PCMCIA reset code
required to support 100 Mbit PCMCIA ethernet cards on Amiga.

10 Mbit and 100 Mbit mode are supported by the same module.
A module parameter switches Amiga ISA IO accessors to word
access by changing isa_type at runtime. Additional code to
reset the PCMCIA hardware is also added to the driver probe.

Patch modified after patch "[PATCH RFC net-next] Amiga PCMCIA
100 MBit card support" submitted to netdev 2018/09/16 by Alex
Kazik <alex@kazik.de>.

CC: netdev@vger.kernel.org
Tested-by: Alex Kazik <alex@kazik.de>
Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>

--
Changes from v1:

- fix module parameter name in Kconfig help text

Alex Kazik:
- change module parameter type to bool, fix module parameter
  permission

Changes from RFC:

Geert Uytterhoeven:
- change APNE_100MBIT to depend on APNE
- change '---help---' to 'help' (former no longer supported)
- fix whitespace errors
- fix module_param_named() arg count
- protect all added code by #ifdef CONFIG_APNE_100MBIT
---
 drivers/net/ethernet/8390/Kconfig | 12 ++++++++++++
 drivers/net/ethernet/8390/apne.c  | 21 +++++++++++++++++++++
 2 files changed, 33 insertions(+)

diff --git a/drivers/net/ethernet/8390/Kconfig b/drivers/net/ethernet/8390/Kconfig
index 9f4b302..6e4db63 100644
--- a/drivers/net/ethernet/8390/Kconfig
+++ b/drivers/net/ethernet/8390/Kconfig
@@ -143,6 +143,18 @@ config APNE
 	  To compile this driver as a module, choose M here: the module
 	  will be called apne.
 
+config APNE100MBIT
+	bool "PCMCIA NE2000 100MBit support"
+	depends on APNE
+	default n
+	help
+	  This changes the driver to support 10/100Mbit cards (e.g. Netgear
+	  FA411, CNet Singlepoint). 10 MBit cards and 100 MBit cards are
+	  supported by the same driver.
+
+	  To activate 100 Mbit support at runtime or from the kernel
+	  command line, use the apne.100mbit module parameter.
+
 config PCMCIA_PCNET
 	tristate "NE2000 compatible PCMCIA support"
 	depends on PCMCIA
diff --git a/drivers/net/ethernet/8390/apne.c b/drivers/net/ethernet/8390/apne.c
index fe6c834..59e41ad 100644
--- a/drivers/net/ethernet/8390/apne.c
+++ b/drivers/net/ethernet/8390/apne.c
@@ -120,6 +120,12 @@ static u32 apne_msg_enable;
 module_param_named(msg_enable, apne_msg_enable, uint, 0444);
 MODULE_PARM_DESC(msg_enable, "Debug message level (see linux/netdevice.h for bitmap)");
 
+#ifdef CONFIG_APNE100MBIT
+static bool apne_100_mbit;
+module_param_named(apne_100_mbit_msg, apne_100_mbit, bool, 0444);
+MODULE_PARM_DESC(apne_100_mbit_msg, "Enable 100 Mbit support");
+#endif
+
 struct net_device * __init apne_probe(int unit)
 {
 	struct net_device *dev;
@@ -139,6 +145,11 @@ struct net_device * __init apne_probe(int unit)
 	if ( !(AMIGAHW_PRESENT(PCMCIA)) )
 		return ERR_PTR(-ENODEV);
 
+#ifdef CONFIG_APNE100MBIT
+	if (apne_100_mbit)
+		isa_type = ISA_TYPE_AG16;
+#endif
+
 	pr_info("Looking for PCMCIA ethernet card : ");
 
 	/* check if a card is inserted */
@@ -590,6 +601,16 @@ static int init_pcmcia(void)
 #endif
 	u_long offset;
 
+#ifdef CONFIG_APNE100MBIT
+	/* reset card (idea taken from CardReset by Artur Pogoda) */
+	{
+		u_char  tmp = gayle.intreq;
+
+		gayle.intreq = 0xff;    mdelay(1);
+		gayle.intreq = tmp;     mdelay(300);
+	}
+#endif
+
 	pcmcia_reset();
 	pcmcia_program_voltage(PCMCIA_0V);
 	pcmcia_access_speed(PCMCIA_SPEED_250NS);
-- 
2.7.4

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Finn Thain]

On Thu, 17 Jun 2021, Michael Schmitz wrote:

> Add Kconfig option, module parameter and PCMCIA reset code
> required to support 100 Mbit PCMCIA ethernet cards on Amiga.
> 
> 10 Mbit and 100 Mbit mode are supported by the same module.
> A module parameter switches Amiga ISA IO accessors to word
> access by changing isa_type at runtime. Additional code to
> reset the PCMCIA hardware is also added to the driver probe.
> 
> Patch modified after patch "[PATCH RFC net-next] Amiga PCMCIA
> 100 MBit card support" submitted to netdev 2018/09/16 by Alex
> Kazik <alex@kazik.de>.
> 
> CC: netdev@vger.kernel.org
> Tested-by: Alex Kazik <alex@kazik.de>
> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> 
> --
> Changes from v1:
> 
> - fix module parameter name in Kconfig help text
> 
> Alex Kazik:
> - change module parameter type to bool, fix module parameter
>   permission
> 
> Changes from RFC:
> 
> Geert Uytterhoeven:
> - change APNE_100MBIT to depend on APNE
> - change '---help---' to 'help' (former no longer supported)
> - fix whitespace errors
> - fix module_param_named() arg count
> - protect all added code by #ifdef CONFIG_APNE_100MBIT
> ---
>  drivers/net/ethernet/8390/Kconfig | 12 ++++++++++++
>  drivers/net/ethernet/8390/apne.c  | 21 +++++++++++++++++++++
>  2 files changed, 33 insertions(+)
> 
> diff --git a/drivers/net/ethernet/8390/Kconfig b/drivers/net/ethernet/8390/Kconfig
> index 9f4b302..6e4db63 100644
> --- a/drivers/net/ethernet/8390/Kconfig
> +++ b/drivers/net/ethernet/8390/Kconfig
> @@ -143,6 +143,18 @@ config APNE
>  	  To compile this driver as a module, choose M here: the module
>  	  will be called apne.
>  
> +config APNE100MBIT
> +	bool "PCMCIA NE2000 100MBit support"
> +	depends on APNE
> +	default n
> +	help
> +	  This changes the driver to support 10/100Mbit cards (e.g. Netgear
> +	  FA411, CNet Singlepoint). 10 MBit cards and 100 MBit cards are
> +	  supported by the same driver.
> +
> +	  To activate 100 Mbit support at runtime or from the kernel
> +	  command line, use the apne.100mbit module parameter.
> +
>  config PCMCIA_PCNET
>  	tristate "NE2000 compatible PCMCIA support"
>  	depends on PCMCIA
> diff --git a/drivers/net/ethernet/8390/apne.c b/drivers/net/ethernet/8390/apne.c
> index fe6c834..59e41ad 100644
> --- a/drivers/net/ethernet/8390/apne.c
> +++ b/drivers/net/ethernet/8390/apne.c
> @@ -120,6 +120,12 @@ static u32 apne_msg_enable;
>  module_param_named(msg_enable, apne_msg_enable, uint, 0444);
>  MODULE_PARM_DESC(msg_enable, "Debug message level (see linux/netdevice.h for bitmap)");
>  
> +#ifdef CONFIG_APNE100MBIT
> +static bool apne_100_mbit;
> +module_param_named(apne_100_mbit_msg, apne_100_mbit, bool, 0444);
> +MODULE_PARM_DESC(apne_100_mbit_msg, "Enable 100 Mbit support");
> +#endif
> +
>  struct net_device * __init apne_probe(int unit)
>  {
>  	struct net_device *dev;
> @@ -139,6 +145,11 @@ struct net_device * __init apne_probe(int unit)
>  	if ( !(AMIGAHW_PRESENT(PCMCIA)) )
>  		return ERR_PTR(-ENODEV);
>  
> +#ifdef CONFIG_APNE100MBIT
> +	if (apne_100_mbit)
> +		isa_type = ISA_TYPE_AG16;
> +#endif
> +

I think isa_type has to be assigned unconditionally otherwise it can't be 
reset for 10 mbit cards. Therefore, the AMIGAHW_PRESENT(PCMCIA) logic in 
arch/m68k/kernel/setup_mm.c probably should move here.

>  	pr_info("Looking for PCMCIA ethernet card : ");
>  
>  	/* check if a card is inserted */
> @@ -590,6 +601,16 @@ static int init_pcmcia(void)
>  #endif
>  	u_long offset;
>  
> +#ifdef CONFIG_APNE100MBIT
> +	/* reset card (idea taken from CardReset by Artur Pogoda) */
> +	{
> +		u_char  tmp = gayle.intreq;
> +
> +		gayle.intreq = 0xff;    mdelay(1);
> +		gayle.intreq = tmp;     mdelay(300);
> +	}
> +#endif
> +

The indentation/alignment here doesn't conform to the kernel coding style. 

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Michael Schmitz]

Hi Finn,

thanks for your review!

On 17/06/21 6:51 pm, Finn Thain wrote:
> On Thu, 17 Jun 2021, Michael Schmitz wrote:
>
>> Add Kconfig option, module parameter and PCMCIA reset code
>> required to support 100 Mbit PCMCIA ethernet cards on Amiga.
>>
>> 10 Mbit and 100 Mbit mode are supported by the same module.
>> A module parameter switches Amiga ISA IO accessors to word
>> access by changing isa_type at runtime. Additional code to
>> reset the PCMCIA hardware is also added to the driver probe.
>>
>> Patch modified after patch "[PATCH RFC net-next] Amiga PCMCIA
>> 100 MBit card support" submitted to netdev 2018/09/16 by Alex
>> Kazik <alex@kazik.de>.
>>
>> CC: netdev@vger.kernel.org
>> Tested-by: Alex Kazik <alex@kazik.de>
>> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
>>
>> --
>> Changes from v1:
>>
>> - fix module parameter name in Kconfig help text
>>
>> Alex Kazik:
>> - change module parameter type to bool, fix module parameter
>>    permission
>>
>> Changes from RFC:
>>
>> Geert Uytterhoeven:
>> - change APNE_100MBIT to depend on APNE
>> - change '---help---' to 'help' (former no longer supported)
>> - fix whitespace errors
>> - fix module_param_named() arg count
>> - protect all added code by #ifdef CONFIG_APNE_100MBIT
>> ---
>>   drivers/net/ethernet/8390/Kconfig | 12 ++++++++++++
>>   drivers/net/ethernet/8390/apne.c  | 21 +++++++++++++++++++++
>>   2 files changed, 33 insertions(+)
>>
>> diff --git a/drivers/net/ethernet/8390/Kconfig b/drivers/net/ethernet/8390/Kconfig
>> index 9f4b302..6e4db63 100644
>> --- a/drivers/net/ethernet/8390/Kconfig
>> +++ b/drivers/net/ethernet/8390/Kconfig
>> @@ -143,6 +143,18 @@ config APNE
>>   	  To compile this driver as a module, choose M here: the module
>>   	  will be called apne.
>>   
>> +config APNE100MBIT
>> +	bool "PCMCIA NE2000 100MBit support"
>> +	depends on APNE
>> +	default n
>> +	help
>> +	  This changes the driver to support 10/100Mbit cards (e.g. Netgear
>> +	  FA411, CNet Singlepoint). 10 MBit cards and 100 MBit cards are
>> +	  supported by the same driver.
>> +
>> +	  To activate 100 Mbit support at runtime or from the kernel
>> +	  command line, use the apne.100mbit module parameter.
>> +
>>   config PCMCIA_PCNET
>>   	tristate "NE2000 compatible PCMCIA support"
>>   	depends on PCMCIA
>> diff --git a/drivers/net/ethernet/8390/apne.c b/drivers/net/ethernet/8390/apne.c
>> index fe6c834..59e41ad 100644
>> --- a/drivers/net/ethernet/8390/apne.c
>> +++ b/drivers/net/ethernet/8390/apne.c
>> @@ -120,6 +120,12 @@ static u32 apne_msg_enable;
>>   module_param_named(msg_enable, apne_msg_enable, uint, 0444);
>>   MODULE_PARM_DESC(msg_enable, "Debug message level (see linux/netdevice.h for bitmap)");
>>   
>> +#ifdef CONFIG_APNE100MBIT
>> +static bool apne_100_mbit;
>> +module_param_named(apne_100_mbit_msg, apne_100_mbit, bool, 0444);
>> +MODULE_PARM_DESC(apne_100_mbit_msg, "Enable 100 Mbit support");
>> +#endif
>> +
>>   struct net_device * __init apne_probe(int unit)
>>   {
>>   	struct net_device *dev;
>> @@ -139,6 +145,11 @@ struct net_device * __init apne_probe(int unit)
>>   	if ( !(AMIGAHW_PRESENT(PCMCIA)) )
>>   		return ERR_PTR(-ENODEV);
>>   
>> +#ifdef CONFIG_APNE100MBIT
>> +	if (apne_100_mbit)
>> +		isa_type = ISA_TYPE_AG16;
>> +#endif
>> +
> I think isa_type has to be assigned unconditionally otherwise it can't be
> reset for 10 mbit cards. Therefore, the AMIGAHW_PRESENT(PCMCIA) logic in
> arch/m68k/kernel/setup_mm.c probably should move here.

Good catch! I am uncertain though as to whether replacing a 100 Mbit 
card by a 10 Mbit one at run time is a common use case (or even 
possible, given constraints of the Amiga PCMCIA interface?), but it 
ought to work even if rarely used.

The comment there says isa_type must be set as early as possible, so I'd 
rather leave that alone, and add an 'else' clause here.

This of course raise the question whether we ought to move the entire 
isa_type handling into arch code instead - make it a generic 
amiga_pcmcia_16bit option settable via sysfs. There may be other 16 bit 
cards that require the same treatment, and duplicating PCMCIA mode 
switching all over the place could be avoided. Opinions?

>
>>   	pr_info("Looking for PCMCIA ethernet card : ");
>>   
>>   	/* check if a card is inserted */
>> @@ -590,6 +601,16 @@ static int init_pcmcia(void)
>>   #endif
>>   	u_long offset;
>>   
>> +#ifdef CONFIG_APNE100MBIT
>> +	/* reset card (idea taken from CardReset by Artur Pogoda) */
>> +	{
>> +		u_char  tmp = gayle.intreq;
>> +
>> +		gayle.intreq = 0xff;    mdelay(1);
>> +		gayle.intreq = tmp;     mdelay(300);
>> +	}
>> +#endif
>> +
> The indentation/alignment here doesn't conform to the kernel coding style.

Good one. Checkpatch missed that for some reason...

I'll respin ...

Cheers,

     Michael

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Geert Uytterhoeven]

Hi Michael,

On Thu, Jun 17, 2021 at 9:33 PM Michael Schmitz <schmitzmic@gmail.com> wrote:
> On 17/06/21 6:51 pm, Finn Thain wrote:
> > On Thu, 17 Jun 2021, Michael Schmitz wrote:
> >> Add Kconfig option, module parameter and PCMCIA reset code
> >> required to support 100 Mbit PCMCIA ethernet cards on Amiga.
> >>
> >> 10 Mbit and 100 Mbit mode are supported by the same module.
> >> A module parameter switches Amiga ISA IO accessors to word
> >> access by changing isa_type at runtime. Additional code to
> >> reset the PCMCIA hardware is also added to the driver probe.
> >>
> >> Patch modified after patch "[PATCH RFC net-next] Amiga PCMCIA
> >> 100 MBit card support" submitted to netdev 2018/09/16 by Alex
> >> Kazik <alex@kazik.de>.
> >>
> >> CC: netdev@vger.kernel.org
> >> Tested-by: Alex Kazik <alex@kazik.de>
> >> Signed-off-by: Michael Schmitz <schmitzmic@gmail.com>
> >> --- a/drivers/net/ethernet/8390/apne.c
> >> +++ b/drivers/net/ethernet/8390/apne.c
> >> @@ -120,6 +120,12 @@ static u32 apne_msg_enable;
> >>   module_param_named(msg_enable, apne_msg_enable, uint, 0444);
> >>   MODULE_PARM_DESC(msg_enable, "Debug message level (see linux/netdevice.h for bitmap)");
> >>
> >> +#ifdef CONFIG_APNE100MBIT
> >> +static bool apne_100_mbit;
> >> +module_param_named(apne_100_mbit_msg, apne_100_mbit, bool, 0444);
> >> +MODULE_PARM_DESC(apne_100_mbit_msg, "Enable 100 Mbit support");
> >> +#endif
> >> +
> >>   struct net_device * __init apne_probe(int unit)
> >>   {
> >>      struct net_device *dev;
> >> @@ -139,6 +145,11 @@ struct net_device * __init apne_probe(int unit)
> >>      if ( !(AMIGAHW_PRESENT(PCMCIA)) )
> >>              return ERR_PTR(-ENODEV);
> >>
> >> +#ifdef CONFIG_APNE100MBIT
> >> +    if (apne_100_mbit)
> >> +            isa_type = ISA_TYPE_AG16;
> >> +#endif
> >> +
> > I think isa_type has to be assigned unconditionally otherwise it can't be
> > reset for 10 mbit cards. Therefore, the AMIGAHW_PRESENT(PCMCIA) logic in
> > arch/m68k/kernel/setup_mm.c probably should move here.
>
> Good catch! I am uncertain though as to whether replacing a 100 Mbit
> card by a 10 Mbit one at run time is a common use case (or even
> possible, given constraints of the Amiga PCMCIA interface?), but it
> ought to work even if rarely used.

Given it's PCMCIA, I guess that's a possibility.
Furthermore, always setting isa_type means the user can recover from
a mistake by unloading the module, and modprobe'ing again with the
correct parameter.
For the builtin-case, that needs a s/0444/0644/ change, though.

> The comment there says isa_type must be set as early as possible, so I'd
> rather leave that alone, and add an 'else' clause here.
>
> This of course raise the question whether we ought to move the entire
> isa_type handling into arch code instead - make it a generic
> amiga_pcmcia_16bit option settable via sysfs. There may be other 16 bit
> cards that require the same treatment, and duplicating PCMCIA mode
> switching all over the place could be avoided. Opinions?

Indeed.

Still, can we autodetect in the driver?

I'm wondering how this is handled on PCs with PCMCIA, or if there
really is something special about Amiga PCMCIA hardware...

And I'd really like to get rid of the CONFIG_APNE100MBIT option,
i.e. always include the support, if possible.

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Michael Schmitz]

Hi Geert,

Am 18.06.2021 um 19:16 schrieb Geert Uytterhoeven:
>>>> +#ifdef CONFIG_APNE100MBIT
>>>> +    if (apne_100_mbit)
>>>> +            isa_type = ISA_TYPE_AG16;
>>>> +#endif
>>>> +
>>> I think isa_type has to be assigned unconditionally otherwise it can't be
>>> reset for 10 mbit cards. Therefore, the AMIGAHW_PRESENT(PCMCIA) logic in
>>> arch/m68k/kernel/setup_mm.c probably should move here.
>>
>> Good catch! I am uncertain though as to whether replacing a 100 Mbit
>> card by a 10 Mbit one at run time is a common use case (or even
>> possible, given constraints of the Amiga PCMCIA interface?), but it
>> ought to work even if rarely used.
>
> Given it's PCMCIA, I guess that's a possibility.
> Furthermore, always setting isa_type means the user can recover from
> a mistake by unloading the module, and modprobe'ing again with the
> correct parameter.
> For the builtin-case, that needs a s/0444/0644/ change, though.

How does re-probing after a card change for a builtin driver work? 
Changing the permission bits is a minor issue.

>
>> The comment there says isa_type must be set as early as possible, so I'd
>> rather leave that alone, and add an 'else' clause here.
>>
>> This of course raise the question whether we ought to move the entire
>> isa_type handling into arch code instead - make it a generic
>> amiga_pcmcia_16bit option settable via sysfs. There may be other 16 bit
>> cards that require the same treatment, and duplicating PCMCIA mode
>> switching all over the place could be avoided. Opinions?
>
> Indeed.

The only downside I can see is that setting isa_type needs to be done 
ahead of modprobe, through sysfs. That might be a little error prone.

> Still, can we autodetect in the driver?

Guess we'll have to find out how the 16 bit cards behave if first poked 
in 8 bit mode, attempting to force a reset of the 8390 chip, and 
switching to 16 bit mode if this fails. That's normally done in 
apne_probe1() which runs after init_pcmcia(), so we can't rely on the 
result of a 8390 reset autoprobe to do the PCMCIA software reset there.

The 8390 reset part does not rely on anything else in apne_probe1(), so 
that code can be lifted out of apne_probe1() and run early in 
apne_probe() (after the check for an inserted PCMCIA card). I'll try and 
prepare a patch for Alex to test that method.

>
> I'm wondering how this is handled on PCs with PCMCIA, or if there
> really is something special about Amiga PCMCIA hardware...

What's special about Amiga PCMCIA hardware is that the card reset isn't 
connected for those 16 bit cards, so pcmcia_reset() does not work.

Whether the software reset workaround hurts for 8 bit cards is something 
I don't know and cannot test. But

> And I'd really like to get rid of the CONFIG_APNE100MBIT option,
> i.e. always include the support, if possible.

I can't see why that wouldn't be possible - the only downside is that we 
force MULTI_ISA=1 always for Amiga, and lose the optimizations done for 
MUTLI_ISA=0 in io_mm.h. Unless we autoprobe, we can use isa_type to 
guard against running a software reset on 8 bit cards ...

Cheers,

	Michael

> Gr{oetje,eeting}s,
>
>                         Geert
>
> --
> Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org
>
> In personal conversations with technical people, I call myself a hacker. But
> when I'm talking to journalists I just say "programmer" or something like that.
>                                 -- Linus Torvalds
>

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Geert Uytterhoeven]

Hi Michael,

On Fri, Jun 18, 2021 at 10:06 AM Michael Schmitz <schmitzmic@gmail.com> wrote:
> Am 18.06.2021 um 19:16 schrieb Geert Uytterhoeven:
> >>>> +#ifdef CONFIG_APNE100MBIT
> >>>> +    if (apne_100_mbit)
> >>>> +            isa_type = ISA_TYPE_AG16;
> >>>> +#endif
> >>>> +
> >>> I think isa_type has to be assigned unconditionally otherwise it can't be
> >>> reset for 10 mbit cards. Therefore, the AMIGAHW_PRESENT(PCMCIA) logic in
> >>> arch/m68k/kernel/setup_mm.c probably should move here.
> >>
> >> Good catch! I am uncertain though as to whether replacing a 100 Mbit
> >> card by a 10 Mbit one at run time is a common use case (or even
> >> possible, given constraints of the Amiga PCMCIA interface?), but it
> >> ought to work even if rarely used.
> >
> > Given it's PCMCIA, I guess that's a possibility.
> > Furthermore, always setting isa_type means the user can recover from
> > a mistake by unloading the module, and modprobe'ing again with the
> > correct parameter.
> > For the builtin-case, that needs a s/0444/0644/ change, though.
>
> How does re-probing after a card change for a builtin driver work?
> Changing the permission bits is a minor issue.

Oh right, this driver predates the driver framework, and doesn't support
PCMCIA hotplug.  So auto-unregister on removal doesn't work.
Even using unbind/bind in sysfs won't work.

So rmmod/modprobe is the only thing that has a chance to work...

> >> The comment there says isa_type must be set as early as possible, so I'd
> >> rather leave that alone, and add an 'else' clause here.
> >>
> >> This of course raise the question whether we ought to move the entire
> >> isa_type handling into arch code instead - make it a generic
> >> amiga_pcmcia_16bit option settable via sysfs. There may be other 16 bit
> >> cards that require the same treatment, and duplicating PCMCIA mode
> >> switching all over the place could be avoided. Opinions?
> >
> > Indeed.
>
> The only downside I can see is that setting isa_type needs to be done
> ahead of modprobe, through sysfs. That might be a little error prone.
>
> > Still, can we autodetect in the driver?
>
> Guess we'll have to find out how the 16 bit cards behave if first poked
> in 8 bit mode, attempting to force a reset of the 8390 chip, and
> switching to 16 bit mode if this fails. That's normally done in
> apne_probe1() which runs after init_pcmcia(), so we can't rely on the
> result of a 8390 reset autoprobe to do the PCMCIA software reset there.
>
> The 8390 reset part does not rely on anything else in apne_probe1(), so
> that code can be lifted out of apne_probe1() and run early in
> apne_probe() (after the check for an inserted PCMCIA card). I'll try and
> prepare a patch for Alex to test that method.
>
> > I'm wondering how this is handled on PCs with PCMCIA, or if there
> > really is something special about Amiga PCMCIA hardware...
>
> What's special about Amiga PCMCIA hardware is that the card reset isn't
> connected for those 16 bit cards, so pcmcia_reset() does not work.

I was mostly thinking about the difference between 8-bit and 16-bit
accesses.

> Whether the software reset workaround hurts for 8 bit cards is something
> I don't know and cannot test. But
>
> > And I'd really like to get rid of the CONFIG_APNE100MBIT option,
> > i.e. always include the support, if possible.
>
> I can't see why that wouldn't be possible - the only downside is that we
> force MULTI_ISA=1 always for Amiga, and lose the optimizations done for
> MUTLI_ISA=0 in io_mm.h. Unless we autoprobe, we can use isa_type to
> guard against running a software reset on 8 bit cards ...

The latter sounds like a neat trick...

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds

Re: [PATCH net-next v3 2/2] net/8390: apne.c - add 100 Mbit support to apne.c driver

[Michael Schmitz]

Hi Geert,

Am 18.06.2021 um 20:13 schrieb Geert Uytterhoeven:
>>
>> How does re-probing after a card change for a builtin driver work?
>> Changing the permission bits is a minor issue.
>
> Oh right, this driver predates the driver framework, and doesn't support
> PCMCIA hotplug.  So auto-unregister on removal doesn't work.
> Even using unbind/bind in sysfs won't work.
>
> So rmmod/modprobe is the only thing that has a chance to work...
>
>>>> The comment there says isa_type must be set as early as possible, so I'd
>>>> rather leave that alone, and add an 'else' clause here.
>>>>
>>>> This of course raise the question whether we ought to move the entire
>>>> isa_type handling into arch code instead - make it a generic
>>>> amiga_pcmcia_16bit option settable via sysfs. There may be other 16 bit
>>>> cards that require the same treatment, and duplicating PCMCIA mode
>>>> switching all over the place could be avoided. Opinions?
>>>
>>> Indeed.
>>
>> The only downside I can see is that setting isa_type needs to be done
>> ahead of modprobe, through sysfs. That might be a little error prone.
>>
>>> Still, can we autodetect in the driver?
>>
>> Guess we'll have to find out how the 16 bit cards behave if first poked
>> in 8 bit mode, attempting to force a reset of the 8390 chip, and
>> switching to 16 bit mode if this fails. That's normally done in
>> apne_probe1() which runs after init_pcmcia(), so we can't rely on the
>> result of a 8390 reset autoprobe to do the PCMCIA software reset there.
>>
>> The 8390 reset part does not rely on anything else in apne_probe1(), so
>> that code can be lifted out of apne_probe1() and run early in
>> apne_probe() (after the check for an inserted PCMCIA card). I'll try and
>> prepare a patch for Alex to test that method.

I just realized that might not work - ínit_pcmcia() enables the PCMCIA 
interface for us, so the early 8390 reset may not go through at all... 
The module parameter may have to stay as a fallback option at least.

>>
>>> I'm wondering how this is handled on PCs with PCMCIA, or if there
>>> really is something special about Amiga PCMCIA hardware...
>>
>> What's special about Amiga PCMCIA hardware is that the card reset isn't
>> connected for those 16 bit cards, so pcmcia_reset() does not work.
>
> I was mostly thinking about the difference between 8-bit and 16-bit
> accesses.

No idea. I've never even seen an 8 bit PCMCIA card - I have a few old 
16/32 bit ones around that were great for crashing my PowerBook, nothing 
else...

>>> And I'd really like to get rid of the CONFIG_APNE100MBIT option,
>>> i.e. always include the support, if possible.
>>
>> I can't see why that wouldn't be possible - the only downside is that we
>> force MULTI_ISA=1 always for Amiga, and lose the optimizations done for
>> MUTLI_ISA=0 in io_mm.h. Unless we autoprobe, we can use isa_type to
>> guard against running a software reset on 8 bit cards ...
>
> The latter sounds like a neat trick...

Yes, we can at least get rid of the APNE100MBIT option that way. I'll 
have to think about the autoprobe a bit more.

Cheers,

	Michael

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