SH7760: early_printk build broken

SH7760: early_printk build broken

[Manuel Lauss]

Hello Iwamatsu-san,

Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):

  CC      arch/sh/kernel/early_printk.o
/mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
/mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'

I don't know immediately how to fix it though, could you please take a look?

Thank you,
	Manuel Lauss

Re: SH7760: early_printk build broken

[Paul Mundt]

On Tue, Jul 29, 2008 at 04:27:43PM +0200, Manuel Lauss wrote:
> Hello Iwamatsu-san,
> 
> Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
> broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):
> 
>   CC      arch/sh/kernel/early_printk.o
> /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
> /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'
> 
> I don't know immediately how to fix it though, could you please take a look?
> 
I see what happened. Proof that this file is such a ridiculous
clusterfuck it's hard to know where to even begin.

---

diff --git a/drivers/serial/sh-sci.h b/drivers/serial/sh-sci.h
index cd728df..e8d3633 100644
--- a/drivers/serial/sh-sci.h
+++ b/drivers/serial/sh-sci.h
@@ -463,6 +463,8 @@ SCIF_FNS(SCLSR,				0,  0, 0x28, 16)
 SCIF_FNS(SCFDR,				0,  0, 0x1C, 16)
 SCIF_FNS(SCSPTR2,			0,  0, 0x20, 16)
 SCIF_FNS(SCLSR2, 			0,  0, 0x24, 16)
+#else
+SCIF_FNS(SCFDR,			     0x0e, 16, 0x1C, 16)
 #endif /* CONFIG_CPU_SUBTYPE_SH7763 */
 #else
 SCIF_FNS(SCFDR,                      0x0e, 16, 0x1C, 16)

Re: SH7760: early_printk build broken

[Paul Mundt]

On Wed, Jul 30, 2008 at 12:23:10AM +0900, Paul Mundt wrote:
> On Tue, Jul 29, 2008 at 04:27:43PM +0200, Manuel Lauss wrote:
> > Hello Iwamatsu-san,
> > 
> > Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
> > broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):
> > 
> >   CC      arch/sh/kernel/early_printk.o
> > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
> > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'
> > 
> > I don't know immediately how to fix it though, could you please take a look?
> > 
> I see what happened. Proof that this file is such a ridiculous
> clusterfuck it's hard to know where to even begin.
> 
Here's a slightly more coherent version. It's really not worth trying to
shoe-horn 7763 in with 7760/7780/7785. If it has any differences, it's
always best to separate it out completely, or we are very quickly reduced
to impossible to follow ifdef hell.

With the generic kgdbts stuff in now and the sh-bios users falling by the
way side, sh-sci is long overdue for a total rewrite anyways. It's only
been 6+ years since the last time I rewrote it anyways, when I had
planned to get rid of sh-sci.h also.

The biggest mess comes from having it split between sh64/sh/h8300 (and
probably RX in the not so distant future), and most of this sharing
happened pre driver-model. Unfortunately these subtle variations across
the blocks make supporting it a total pain. Now that we have totally
different port types on newer parts it's necessary to rework a bit
anyways (presently we shove SCIFA parts from SH-MobileR2 in as SCI
ports), and the DMA auto requests start to matter on SMP parts (which of
course has nothing in common with the previous SMP extensions that SH-2
implemented back in the day to get around the fact that bus accesses were
mutually exclusive..). ;-)

Re: SH7760: early_printk build broken

[Paul Mundt]

On Wed, Jul 30, 2008 at 12:45:39AM +0900, Paul Mundt wrote:
> On Wed, Jul 30, 2008 at 12:23:10AM +0900, Paul Mundt wrote:
> > On Tue, Jul 29, 2008 at 04:27:43PM +0200, Manuel Lauss wrote:
> > > Hello Iwamatsu-san,
> > > 
> > > Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
> > > broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):
> > > 
> > >   CC      arch/sh/kernel/early_printk.o
> > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
> > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'
> > > 
> > > I don't know immediately how to fix it though, could you please take a look?
> > > 
> > I see what happened. Proof that this file is such a ridiculous
> > clusterfuck it's hard to know where to even begin.
> > 
> Here's a slightly more coherent version. It's really not worth trying to
> shoe-horn 7763 in with 7760/7780/7785. If it has any differences, it's
> always best to separate it out completely, or we are very quickly reduced
> to impossible to follow ifdef hell.
> 
> With the generic kgdbts stuff in now and the sh-bios users falling by the
> way side, sh-sci is long overdue for a total rewrite anyways. It's only
> been 6+ years since the last time I rewrote it anyways, when I had
> planned to get rid of sh-sci.h also.
> 
> The biggest mess comes from having it split between sh64/sh/h8300 (and
> probably RX in the not so distant future), and most of this sharing
> happened pre driver-model. Unfortunately these subtle variations across
> the blocks make supporting it a total pain. Now that we have totally
> different port types on newer parts it's necessary to rework a bit
> anyways (presently we shove SCIFA parts from SH-MobileR2 in as SCI
> ports), and the DMA auto requests start to matter on SMP parts (which of
> course has nothing in common with the previous SMP extensions that SH-2
> implemented back in the day to get around the fact that bus accesses were
> mutually exclusive..). ;-)

---

diff --git a/drivers/serial/sh-sci.h b/drivers/serial/sh-sci.h
index cd728df..8a0749e 100644
--- a/drivers/serial/sh-sci.h
+++ b/drivers/serial/sh-sci.h
@@ -451,19 +451,21 @@ SCIx_FNS(SCxSR,  0x08,  8, 0x10,  8, 0x08, 16, 0x10, 16, 0x04,  8)
 SCIx_FNS(SCxRDR, 0x0a,  8, 0x14,  8, 0x0A,  8, 0x14,  8, 0x05,  8)
 SCIF_FNS(SCFCR,                      0x0c,  8, 0x18, 16)
 #if defined(CONFIG_CPU_SUBTYPE_SH7760) || \
-    defined(CONFIG_CPU_SUBTYPE_SH7763) || \
     defined(CONFIG_CPU_SUBTYPE_SH7780) || \
     defined(CONFIG_CPU_SUBTYPE_SH7785)
+SCIF_FNS(SCFDR,			     0x0e, 16, 0x1C, 16)
 SCIF_FNS(SCTFDR,		     0x0e, 16, 0x1C, 16)
 SCIF_FNS(SCRFDR,		     0x0e, 16, 0x20, 16)
 SCIF_FNS(SCSPTR,			0,  0, 0x24, 16)
 SCIF_FNS(SCLSR,				0,  0, 0x28, 16)
-#if defined(CONFIG_CPU_SUBTYPE_SH7763)
-/* SH7763 SCIF2 */
+#elif defined(CONFIG_CPU_SUBTYPE_SH7763)
 SCIF_FNS(SCFDR,				0,  0, 0x1C, 16)
 SCIF_FNS(SCSPTR2,			0,  0, 0x20, 16)
-SCIF_FNS(SCLSR2, 			0,  0, 0x24, 16)
-#endif /* CONFIG_CPU_SUBTYPE_SH7763 */
+SCIF_FNS(SCLSR2,			0,  0, 0x24, 16)
+SCIF_FNS(SCTFDR,		     0x0e, 16, 0x1C, 16)
+SCIF_FNS(SCRFDR,		     0x0e, 16, 0x20, 16)
+SCIF_FNS(SCSPTR,			0,  0, 0x24, 16)
+SCIF_FNS(SCLSR,				0,  0, 0x28, 16)
 #else
 SCIF_FNS(SCFDR,                      0x0e, 16, 0x1C, 16)
 #if defined(CONFIG_CPU_SUBTYPE_SH7722)

Re: SH7760: early_printk build broken

[Manuel Lauss]

> > > > Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
> > > > broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):
> > > > 
> > > >   CC      arch/sh/kernel/early_printk.o
> > > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
> > > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'
> > > > 
> > > > I don't know immediately how to fix it though, could you please take a look?
> > > > 
> > > I see what happened. Proof that this file is such a ridiculous
> > > clusterfuck it's hard to know where to even begin.
> > >

Heh, it's an interesting maze of Hitachi serial port silicon history... ;-)


> > With the generic kgdbts stuff in now and the sh-bios users falling by the
> > way side, sh-sci is long overdue for a total rewrite anyways. It's only
> > been 6+ years since the last time I rewrote it anyways, when I had
> > planned to get rid of sh-sci.h also.

Have fun ;-)  If you are busy with more interesting stuff, I may be able
to take a shot at it next week.

Thank you.
	Manuel Lauss

Re: SH7760: early_printk build broken

[Nobuhiro Iwamatsu]

Hi, Manuel and Paul.

2008/7/30 Manuel Lauss <mano@roarinelk.homelinux.net>:
>
>> > > > Commit c63847a3621d2bac054f5709783860ecabd0ee7e ("sh: Add SCIF2 support for SH7763.")
>> > > > broke build with CONFIG_EARLY_PRINTK enabled for me (SH7760):
>> > > >
>> > > >   CC      arch/sh/kernel/early_printk.o
>> > > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c: In function 'scif_sercon_putc':
>> > > > /mnt/work/sh7760/kernel/linux-2.6.git/arch/sh/kernel/early_printk.c:84: error: implicit declaration of function 'sci_SCFDR_in'
>> > > >
>> > > > I don't know immediately how to fix it though, could you please take a look?
>> > > >
>> > > I see what happened. Proof that this file is such a ridiculous
>> > > clusterfuck it's hard to know where to even begin.
>> > >
Sorry, This is a problem of my patch.
I forgot a movement check of early_printk.
Because there is not SCFDR in SH7760 etc.
Of cource this problem has SH7760, SH7780 and SH7785.

I confirmed that I worked without a problem with the patch which
Paul wrote.

Thank you.

Nobuhiro