RE: [PATCH] [POWERPC] Xilinx: add compatibility for 'simple-bus'.

It appears that this turns out to interact badly with the probing of
PPC_UDBG_16550, which is always enabled on PPC405 (and apparently
found!) even though Virtex devices don't have them.

Steve

-----Original Message-----
From: Stephen Neuendorffer [mailto:stephen.neuendorffer@xilinx.com]
Sent: Thursday, May 08, 2008 11:25 AM
To: dwg@au1.ibm.com; jwboyer@linux.vnet.ibm.com;

grant.likely@secretlab.ca; linuxppc-dev@ozlabs.org

Cc: Stephen Neuendorffer
Subject: [PATCH] [POWERPC] Xilinx: add compatibility for 'simple-bus'.
=

ePAPR drafts propose 'simple-bus' as a generic compatibility type for
busses which cannot be probed for devices.  In addition, the Xilinx
versions of these IPs seem to be proliferating.  Hence, in the future
let's prefer to use the standard names.  I've left the old names in
for short term backward compatibility for existing device trees.
=

Signed-off-by: Stephen Neuendorffer <redacted>
---
 arch/powerpc/platforms/40x/virtex.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)
=

b/arch/powerpc/platforms/40x/virtex.c

 static struct of_device_id xilinx_of_bus_ids[] __initdata =3D {
+	{ .compatible =3D "simple-bus", },
 	{ .compatible =3D "xlnx,plb-v46-1.00.a", },
 	{ .compatible =3D "xlnx,plb-v46-1.02.a", },
 	{ .compatible =3D "xlnx,plb-v34-1.01.a", },
--
1.5.3.4
=

This email and any attachments are intended for the sole use of the named r=
ecipient(s) and contain(s) confidential information that may be proprietary=
, privileged or copyrighted under applicable law. If you are not the intend=
ed recipient, do not read, copy, or forward this email message or any attac=
hments. Delete this email message and any attachments immediately.

On Tue, Jun 3, 2008 at 3:41 PM, Stephen Neuendorffer
[off-list ref] wrote:

It appears that this turns out to interact badly with the probing of
PPC_UDBG_16550, which is always enabled on PPC405 (and apparently
found!) even though Virtex devices don't have them.

What is the symptom?

g.

Steve

quoted

-----Original Message-----
From: Stephen Neuendorffer [mailto:stephen.neuendorffer@xilinx.com]
Sent: Thursday, May 08, 2008 11:25 AM
To: dwg@au1.ibm.com; jwboyer@linux.vnet.ibm.com;

grant.likely@secretlab.ca; linuxppc-dev@ozlabs.org

quoted

Cc: Stephen Neuendorffer
Subject: [PATCH] [POWERPC] Xilinx: add compatibility for 'simple-bus'.

ePAPR drafts propose 'simple-bus' as a generic compatibility type for
busses which cannot be probed for devices.  In addition, the Xilinx
versions of these IPs seem to be proliferating.  Hence, in the future
let's prefer to use the standard names.  I've left the old names in
for short term backward compatibility for existing device trees.

Signed-off-by: Stephen Neuendorffer <redacted>
---
 arch/powerpc/platforms/40x/virtex.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/platforms/40x/virtex.c

b/arch/powerpc/platforms/40x/virtex.c

quoted

index 6c72994..b1ab7b8 100644

--- a/arch/powerpc/platforms/40x/virtex.c
+++ b/arch/powerpc/platforms/40x/virtex.c

@@ -16,6 +16,7 @@
 #include <asm/xilinx_intc.h>

 static struct of_device_id xilinx_of_bus_ids[] __initdata = {
+     { .compatible = "simple-bus", },
      { .compatible = "xlnx,plb-v46-1.00.a", },
      { .compatible = "xlnx,plb-v46-1.02.a", },
      { .compatible = "xlnx,plb-v34-1.01.a", },
--

1.5.3.4

-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

legacy_serial identifies a valid ns16550 on a simple-bus, but the
legacy_serial driver doesn't understand the shift and offset flags
necessary to get it to work, which results in no console.

I think the easiest solution is to change the Kconfig so that
PPC_UDBG_16550 is only selected based on !XILINX_VIRTEX.  I've done this
in my tree, but I've been swamped with other things at the moment, so I
haven't verified it.

Or is legacy_serial just legacy at this point and can go away entirely?
I got the impression from some of the earlier discussion that it wasn't
preferred, at least....

Steve

-----Original Message-----
From: glikely@secretlab.ca [mailto:glikely@secretlab.ca] On Behalf Of

Grant Likely

Sent: Friday, June 06, 2008 8:29 AM
To: Stephen Neuendorffer
Cc: dwg@au1.ibm.com; jwboyer@linux.vnet.ibm.com;

linuxppc-dev@ozlabs.org; git

Subject: Re: [PATCH] [POWERPC] Xilinx: add compatibility for

'simple-bus'.

=

On Tue, Jun 3, 2008 at 3:41 PM, Stephen Neuendorffer
[off-list ref] wrote:

quoted

It appears that this turns out to interact badly with the probing of
PPC_UDBG_16550, which is always enabled on PPC405 (and apparently
found!) even though Virtex devices don't have them.

=

What is the symptom?
=

g.
=

quoted

Steve

quoted

-----Original Message-----
From: Stephen Neuendorffer [mailto:stephen.neuendorffer@xilinx.com]
Sent: Thursday, May 08, 2008 11:25 AM
To: dwg@au1.ibm.com; jwboyer@linux.vnet.ibm.com;

grant.likely@secretlab.ca; linuxppc-dev@ozlabs.org

quoted

Cc: Stephen Neuendorffer
Subject: [PATCH] [POWERPC] Xilinx: add compatibility for

'simple-bus'.

quoted

quoted

ePAPR drafts propose 'simple-bus' as a generic compatibility type

for

quoted

quoted

busses which cannot be probed for devices.  In addition, the Xilinx
versions of these IPs seem to be proliferating.  Hence, in the

future

quoted

quoted

let's prefer to use the standard names.  I've left the old names in
for short term backward compatibility for existing device trees.

Signed-off-by: Stephen Neuendorffer

[off-list ref]

quoted

quoted

---
 arch/powerpc/platforms/40x/virtex.c |    1 +
 1 files changed, 1 insertions(+), 0 deletions(-)

diff --git a/arch/powerpc/platforms/40x/virtex.c

b/arch/powerpc/platforms/40x/virtex.c

quoted

index 6c72994..b1ab7b8 100644

--- a/arch/powerpc/platforms/40x/virtex.c
+++ b/arch/powerpc/platforms/40x/virtex.c

@@ -16,6 +16,7 @@
 #include <asm/xilinx_intc.h>

 static struct of_device_id xilinx_of_bus_ids[] __initdata =3D {
+     { .compatible =3D "simple-bus", },
      { .compatible =3D "xlnx,plb-v46-1.00.a", },
      { .compatible =3D "xlnx,plb-v46-1.02.a", },
      { .compatible =3D "xlnx,plb-v34-1.01.a", },
--

1.5.3.4

=

--
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

This email and any attachments are intended for the sole use of the named r=
ecipient(s) and contain(s) confidential information that may be proprietary=
, privileged or copyrighted under applicable law. If you are not the intend=
ed recipient, do not read, copy, or forward this email message or any attac=
hments. Delete this email message and any attachments immediately.

On Fri, Jun 6, 2008 at 10:16 AM, Stephen Neuendorffer
[off-list ref] wrote:

legacy_serial identifies a valid ns16550 on a simple-bus, but the
legacy_serial driver doesn't understand the shift and offset flags
necessary to get it to work, which results in no console.

I think the easiest solution is to change the Kconfig so that
PPC_UDBG_16550 is only selected based on !XILINX_VIRTEX.  I've done this
in my tree, but I've been swamped with other things at the moment, so I
haven't verified it.

This is an easy solution, but it is not a good one.  Doing so would
break UDBG on other 405 boards when building multiplatform kernels.
It would be better to teach legacy serial about the shift and offset.
Alternately, add code to add_legacy_soc_port() to skip it if the
shift/offset properties are present.

Cheers,
g.

-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.

Is there really much of a chance of that, given the differences with the bootwrappers?
Does anyone care enough about legacy_serial for this to matter?   My impression was that legacy serial was not preferred anyway...

Steve

-----Original Message-----
From: glikely@secretlab.ca on behalf of Grant Likely
Sent: Sat 6/28/2008 1:33 PM
To: Stephen Neuendorffer
Cc: dwg@au1.ibm.com; jwboyer@linux.vnet.ibm.com; linuxppc-dev@ozlabs.org; git
Subject: Re: [PATCH] [POWERPC] Xilinx: add compatibility for 'simple-bus'.
 
On Fri, Jun 6, 2008 at 10:16 AM, Stephen Neuendorffer
[off-list ref] wrote:

legacy_serial identifies a valid ns16550 on a simple-bus, but the
legacy_serial driver doesn't understand the shift and offset flags
necessary to get it to work, which results in no console.

I think the easiest solution is to change the Kconfig so that
PPC_UDBG_16550 is only selected based on !XILINX_VIRTEX.  I've done this
in my tree, but I've been swamped with other things at the moment, so I
haven't verified it.

This is an easy solution, but it is not a good one.  Doing so would
break UDBG on other 405 boards when building multiplatform kernels.
It would be better to teach legacy serial about the shift and offset.
Alternately, add code to add_legacy_soc_port() to skip it if the
shift/offset properties are present.

Cheers,
g.

-- 
Grant Likely, B.Sc., P.Eng.
Secret Lab Technologies Ltd.




This email and any attachments are intended for the sole use of the named recipient(s) and contain(s) confidential information that may be proprietary, privileged or copyrighted under applicable law. If you are not the intended recipient, do not read, copy, or forward this email message or any attachments. Delete this email message and any attachments immediately.

On Fri, 2008-06-06 at 09:16 -0700, Stephen Neuendorffer wrote:

legacy_serial identifies a valid ns16550 on a simple-bus, but the
legacy_serial driver doesn't understand the shift and offset flags
necessary to get it to work, which results in no console.

I think the easiest solution is to change the Kconfig so that
PPC_UDBG_16550 is only selected based on !XILINX_VIRTEX.  I've done this
in my tree, but I've been swamped with other things at the moment, so I
haven't verified it.

Or is legacy_serial just legacy at this point and can go away entirely?
I got the impression from some of the earlier discussion that it wasn't
preferred, at least....

No, legacy serial is useful. Just grow it an understanding of the shift
& offset bits :-)

Cheers,
Ben.