[U-Boot-Users] [RFC] raw go command

[U-Boot-Users] [RFC] raw go command

[Kumar Gala]

This patch is just a quick work up to see how to handle a particular
problem I've got.  The idea is that I want finer control over what happens
when I 'go' to piece of code.  The idea is it looks more like a native
function call than the current go which is more main() like.

The calling convention/num of args is specific to the PPC ABI.

I'm looking for suggestions, better names, etc.

thanks

- k

diff --git a/common/cmd_boot.c b/common/cmd_boot.c
index e68f16f..5b0bbe4 100644
--- a/common/cmd_boot.c
+++ b/common/cmd_boot.c
@@ -32,6 +32,42 @@
 DECLARE_GLOBAL_DATA_PTR;
 #endif

+int do_go_raw (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
+{
+	ulong	addr, i;
+	ulong	r[8] = { 0 };
+
+	void	(*img)(ulong, ulong, ulong, ulong,
+		       ulong, ulong, ulong, ulong);
+
+	if ((argc < 2) || (argc > 9)) {
+		printf ("Usage:\n%s\n", cmdtp->usage);
+		return 1;
+	}
+
+	addr = simple_strtoul(argv[1], NULL, 16);
+
+	img = (void (*)(ulong, ulong, ulong, ulong,
+			ulong, ulong, ulong, ulong)) addr;
+
+	for (i = 2; i < argc; i++)
+		r[i-2] = simple_strtoul(argv[i], NULL, 16);
+
+	(*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
+	/* does not return */
+
+	return 0;
+}
+
+/* -------------------------------------------------------------------- */
+
+U_BOOT_CMD(
+	go_raw, CFG_MAXARGS, 1,	do_go_raw,
+	"go_raw  - start application at address 'addr'\n",
+	"addr [arg ...]\n    - start application@address 'addr'\n"
+	"      passing 'arg' as arguments\n"
+);
+
 int do_go (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
 {
 	ulong	addr, rc;

[U-Boot-Users] [RFC] raw go command

[Wolfgang Denk]

In message <Pine.LNX.4.64.0802131746070.29374@blarg.am.freescale.net> you wrote:
> This patch is just a quick work up to see how to handle a particular
> problem I've got.  The idea is that I want finer control over what happens
> when I 'go' to piece of code.  The idea is it looks more like a native
> function call than the current go which is more main() like.

To me it looks different, but not more native.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
You see things; and you say ``Why?'' But I dream  things  that  never
were; and I say ``Why not?''
       - George Bernard Shaw _Back to Methuselah_ (1921) pt. 1, act 1

[U-Boot-Users] [RFC] raw go command

[Kumar Gala]

On Feb 13, 2008, at 6:40 PM, Wolfgang Denk wrote:

> In message <Pine.LNX. 
> 4.64.0802131746070.29374 at blarg.am.freescale.net> you wrote:
>> This patch is just a quick work up to see how to handle a particular
>> problem I've got.  The idea is that I want finer control over what  
>> happens
>> when I 'go' to piece of code.  The idea is it looks more like a  
>> native
>> function call than the current go which is more main() like.
>
> To me it looks different, but not more native.

Maybe my descriptions not the best, but its intended to provide a  
mechanism similar to how we boot a kernel that the user can specify  
the values in the registers.

- k

[U-Boot-Users] [RFC] raw go command

[Haavard Skinnemoen]

On Wed, 13 Feb 2008 17:50:20 -0600 (CST)
Kumar Gala <galak@kernel.crashing.org> wrote:

> The calling convention/num of args is specific to the PPC ABI.

> diff --git a/common/cmd_boot.c b/common/cmd_boot.c
> index e68f16f..5b0bbe4 100644
> --- a/common/cmd_boot.c
> +++ b/common/cmd_boot.c

So what's it doing under common/ with a generic-sounding name?

Haavard

[U-Boot-Users] [RFC] raw go command

[Kumar Gala]

On Feb 14, 2008, at 2:12 AM, Haavard Skinnemoen wrote:

> On Wed, 13 Feb 2008 17:50:20 -0600 (CST)
> Kumar Gala <galak@kernel.crashing.org> wrote:
>
>> The calling convention/num of args is specific to the PPC ABI.
>
>> diff --git a/common/cmd_boot.c b/common/cmd_boot.c
>> index e68f16f..5b0bbe4 100644
>> --- a/common/cmd_boot.c
>> +++ b/common/cmd_boot.c
>
> So what's it doing under common/ with a generic-sounding name?

there is no reason it couldn't be implemented by other arch's but I  
get your point.  Any suggestions on a better name?

I hadn't thought about putting this in lib_ppc, thus the RFC on the  
email :)

- k

[U-Boot-Users] [PATCH] Add call command on PPC

[Kumar Gala]

The call command tries to mimic a function call in that the 'arg's to
the command are passed in registers according to the PPC ABI.

The prototype that call mimics is some variation of the following
depending on how many arguments are passed to the command:

void func(void)
void func(unsigned long r3)
void func(unsigned long r3, unsigned long r4)
...
void func(unsigned long r3, unsigned long r4, ... unsigned long r10)

The maximum number of 'arg's is 8.  There are no arguments passed on
a stack, no floating point or vector arguments.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
---

This is attempt to at the concept I was trying to get a w/the raw go
thread.

 lib_ppc/Makefile   |    4 ++-
 lib_ppc/cmd_call.c |   64 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 67 insertions(+), 1 deletions(-)
 create mode 100644 lib_ppc/cmd_call.c

diff --git a/lib_ppc/Makefile b/lib_ppc/Makefile
index afbd5ca..a776052 100644
--- a/lib_ppc/Makefile
+++ b/lib_ppc/Makefile
@@ -25,9 +25,11 @@ include $(TOPDIR)/config.mk

 LIB	= $(obj)lib$(ARCH).a

+COBJS-$(CONFIG_CMD_CALL) += cmd_call.o
+
 SOBJS	= ppccache.o ppcstring.o ticks.o

-COBJS	= board.o \
+COBJS	= board.o $(COBJS-y) \
 	  bat_rw.o cache.o extable.o kgdb.o time.o interrupts.o

 SRCS 	:= $(SOBJS:.o=.S) $(COBJS:.o=.c)
diff --git a/lib_ppc/cmd_call.c b/lib_ppc/cmd_call.c
new file mode 100644
index 0000000..5476f27
--- /dev/null
+++ b/lib_ppc/cmd_call.c
@@ -0,0 +1,64 @@
+/*
+ * Copyright 2008 Freescale Semiconductor, Inc.
+ *
+ * (C) Copyright 2000-2003
+ * Wolfgang Denk, DENX Software Engineering, wd at denx.de.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#include <common.h>
+#include <command.h>
+
+int do_call (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
+{
+	ulong	addr, i;
+	ulong	r[8] = { 0 };
+
+	void	(*img)(ulong, ulong, ulong, ulong,
+		       ulong, ulong, ulong, ulong);
+
+	if ((argc < 2) || (argc > 10)) {
+		printf ("Usage:\n%s\n", cmdtp->usage);
+		return 1;
+	}
+
+	addr = simple_strtoul(argv[1], NULL, 16);
+
+	img = (void (*)(ulong, ulong, ulong, ulong,
+			ulong, ulong, ulong, ulong)) addr;
+
+	for (i = 2; i < argc; i++)
+		r[i-2] = simple_strtoul(argv[i], NULL, 16);
+
+	(*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
+
+	/* may not return */
+
+	return 0;
+}
+
+/* -------------------------------------------------------------------- */
+
+U_BOOT_CMD(
+	call, CFG_MAXARGS, 1,	do_call,
+	"call    - jump to address 'addr'\n",
+	"addr [arg ...]\n    - jump to address 'addr' passing 'arg's"
+	" as if it were a function call\n"
+);
-- 
1.5.3.8

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Kumar Gala wrote:
> --- /dev/null
> +++ b/lib_ppc/cmd_call.c

there's nothing ppc specific about this

> +	if ((argc < 2) || (argc > 10)) {
> ...
> +U_BOOT_CMD(
> +	call, CFG_MAXARGS, 1,	do_call,

considering the u-boot infrastructure for commands will do the max checking 
for you, there's no need to check the upper bound of argc.  it's also out of 
sync: in one place you have a random "10" but in another you 
have "CFG_MAXARGS".

> +	(*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
> +
> +	/* may not return */
> +
> +	return 0;

why not ?  you might as well not arbitrarily limit things and have the 
prototype return an int and have it return that instead of "0":
	return (*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080215/6ceb1be8/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Kumar Gala]

On Feb 14, 2008, at 11:54 PM, Mike Frysinger wrote:

> On Friday 15 February 2008, Kumar Gala wrote:
>> --- /dev/null
>> +++ b/lib_ppc/cmd_call.c
>
> there's nothing ppc specific about this

I'm not familiar w/other architecture ABIs to say if that's true or  
not.  Clearly the upper limit may vary based on the differences in the  
ABIs.

>> +	if ((argc < 2) || (argc > 10)) {
>> ...
>> +U_BOOT_CMD(
>> +	call, CFG_MAXARGS, 1,	do_call,
>
> considering the u-boot infrastructure for commands will do the max  
> checking
> for you, there's no need to check the upper bound of argc.  it's  
> also out of
> sync: in one place you have a random "10" but in another you
> have "CFG_MAXARGS".

agreed.  At first I was going for something more generic.

>> +	(*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
>> +
>> +	/* may not return */
>> +
>> +	return 0;
>
> why not ?  you might as well not arbitrarily limit things and have the
> prototype return an int and have it return that instead of "0":
> 	return (*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);

will do.

- k

[U-Boot-Users] [PATCH v2] Add call command on PPC

[Kumar Gala]

The call command tries to mimic a function call in that the 'arg's to
the command are passed in registers according to the PPC ABI.

The prototype that call mimics is some variation of the following
depending on how many arguments are passed to the command:

void func(void)
void func(unsigned long r3)
void func(unsigned long r3, unsigned long r4)
...
void func(unsigned long r3, unsigned long r4, ... unsigned long r10)

The maximum number of 'arg's is 8.  There are no arguments passed on
a stack, no floating point or vector arguments.

Signed-off-by: Kumar Gala <galak@kernel.crashing.org>
---

Incorporate Mike's feedback on using u-boot max arg checking and allow the
function to return.

 lib_ppc/Makefile   |    4 ++-
 lib_ppc/cmd_call.c |   60 ++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 63 insertions(+), 1 deletions(-)
 create mode 100644 lib_ppc/cmd_call.c

diff --git a/lib_ppc/Makefile b/lib_ppc/Makefile
index afbd5ca..a776052 100644
--- a/lib_ppc/Makefile
+++ b/lib_ppc/Makefile
@@ -25,9 +25,11 @@ include $(TOPDIR)/config.mk

 LIB	= $(obj)lib$(ARCH).a

+COBJS-$(CONFIG_CMD_CALL) += cmd_call.o
+
 SOBJS	= ppccache.o ppcstring.o ticks.o

-COBJS	= board.o \
+COBJS	= board.o $(COBJS-y) \
 	  bat_rw.o cache.o extable.o kgdb.o time.o interrupts.o

 SRCS 	:= $(SOBJS:.o=.S) $(COBJS:.o=.c)
diff --git a/lib_ppc/cmd_call.c b/lib_ppc/cmd_call.c
new file mode 100644
index 0000000..83a1764
--- /dev/null
+++ b/lib_ppc/cmd_call.c
@@ -0,0 +1,60 @@
+/*
+ * Copyright 2008 Freescale Semiconductor, Inc.
+ *
+ * (C) Copyright 2000-2003
+ * Wolfgang Denk, DENX Software Engineering, wd at denx.de.
+ *
+ * See file CREDITS for list of people who contributed to this
+ * project.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License as
+ * published by the Free Software Foundation; either version 2 of
+ * the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston,
+ * MA 02111-1307 USA
+ */
+
+#include <common.h>
+#include <command.h>
+
+int do_call (cmd_tbl_t *cmdtp, int flag, int argc, char *argv[])
+{
+	ulong	addr, i;
+	ulong	r[8] = { 0 };
+
+	int	(*img)(ulong, ulong, ulong, ulong,
+		       ulong, ulong, ulong, ulong);
+
+	if (argc < 2) {
+		printf ("Usage:\n%s\n", cmdtp->usage);
+		return 1;
+	}
+
+	addr = simple_strtoul(argv[1], NULL, 16);
+
+	img = (int (*)(ulong, ulong, ulong, ulong,
+			ulong, ulong, ulong, ulong)) addr;
+
+	for (i = 2; i < argc; i++)
+		r[i-2] = simple_strtoul(argv[i], NULL, 16);
+
+	return (*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
+}
+
+/* -------------------------------------------------------------------- */
+
+U_BOOT_CMD(
+	call, 10, 1,	do_call,
+	"call    - jump to address 'addr'\n",
+	"addr [arg ...]\n    - jump to address 'addr' passing 'arg's"
+	" as if it were a function call\n"
+);
-- 
1.5.3.8

[U-Boot-Users] [PATCH v2] Add call command on PPC

[Wolfgang Denk]

In message <Pine.LNX.4.64.0802150001460.1512@blarg.am.freescale.net> you wrote:
> The call command tries to mimic a function call in that the 'arg's to
> the command are passed in registers according to the PPC ABI.
> 
> The prototype that call mimics is some variation of the following
> depending on how many arguments are passed to the command:
> 
> void func(void)
> void func(unsigned long r3)
> void func(unsigned long r3, unsigned long r4)
> ...
> void func(unsigned long r3, unsigned long r4, ... unsigned long r10)
> 
> The maximum number of 'arg's is 8.  There are no arguments passed on
> a stack, no floating point or vector arguments.

This really makes zero sense to me.

> +	img = (int (*)(ulong, ulong, ulong, ulong,
> +			ulong, ulong, ulong, ulong)) addr;
> +
> +	for (i = 2; i < argc; i++)
> +		r[i-2] = simple_strtoul(argv[i], NULL, 16);
> +
> +	return (*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);

This looks not clean. Either  this  is  a  function  with  exactly  8
arguments,  or  it  isn't. And where is the code that makes sure that
``the 'arg's to the command are passed in  registers''  ?  I  see  no
difference between this call and what "go" does. 

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
"More software projects have gone awry for lack of calendar time than
for all other causes combined."
                         - Fred Brooks, Jr., _The Mythical Man Month_

[U-Boot-Users] [PATCH v2] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Wolfgang Denk wrote:
> > +	return (*img)(r[0], r[1], r[2], r[3], r[4], r[5], r[6], r[7]);
>
> I see no difference between this call and what "go" does.

well there's an obvious difference.  go passes argc/argv[] as parameters where 
as this passes everything as parameters.  as for whether you think it's 
needed at all, that's a different matter ;)
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080216/4011dc80/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Kumar Gala wrote:
> On Feb 14, 2008, at 11:54 PM, Mike Frysinger wrote:
> > On Friday 15 February 2008, Kumar Gala wrote:
> >> --- /dev/null
> >> +++ b/lib_ppc/cmd_call.c
> >
> > there's nothing ppc specific about this
>
> I'm not familiar w/other architecture ABIs to say if that's true or
> not.  Clearly the upper limit may vary based on the differences in the
> ABIs.

if the code were written in assembly, i'd agree.  you've written it in pure C 
though, so the compiler worries about the ABI.  if your ABI has different 
semantics for function pointers than normal function calls, it's a piece of 
crap.  i dont think any ports of u-boot need worry about this.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080215/eea233b2/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Kumar Gala]

On Feb 15, 2008, at 12:32 AM, Mike Frysinger wrote:

> On Friday 15 February 2008, Kumar Gala wrote:
>> On Feb 14, 2008, at 11:54 PM, Mike Frysinger wrote:
>>> On Friday 15 February 2008, Kumar Gala wrote:
>>>> --- /dev/null
>>>> +++ b/lib_ppc/cmd_call.c
>>>
>>> there's nothing ppc specific about this
>>
>> I'm not familiar w/other architecture ABIs to say if that's true or
>> not.  Clearly the upper limit may vary based on the differences in  
>> the
>> ABIs.
>
> if the code were written in assembly, i'd agree.  you've written it  
> in pure C
> though, so the compiler worries about the ABI.  if your ABI has  
> different
> semantics for function pointers than normal function calls, it's a  
> piece of
> crap.  i dont think any ports of u-boot need worry about this.

agreed there.

The main reason for me doing this was to provide a "generic" mechanism  
that mimic how we boot kernels on PPC.

The other issue w/providing this on all arch's is that the number of  
params passed via regs may vary and I don't know what is for arm,  
avr32, blackfin, i386, m68k, microblaze, mips, nios, nios2, and sh.

and each variation will require a different prototype.

if someone can tell me what the number of args are for each arch we  
can see about providing this everywhere.

- k

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Kumar Gala wrote:
> On Feb 15, 2008, at 12:32 AM, Mike Frysinger wrote:
> > On Friday 15 February 2008, Kumar Gala wrote:
> >> On Feb 14, 2008, at 11:54 PM, Mike Frysinger wrote:
> >>> On Friday 15 February 2008, Kumar Gala wrote:
> >>>> --- /dev/null
> >>>> +++ b/lib_ppc/cmd_call.c
> >>>
> >>> there's nothing ppc specific about this
> >>
> >> I'm not familiar w/other architecture ABIs to say if that's true or
> >> not.  Clearly the upper limit may vary based on the differences in
> >> the
> >> ABIs.
> >
> > if the code were written in assembly, i'd agree.  you've written it
> > in pure C
> > though, so the compiler worries about the ABI.  if your ABI has
> > different
> > semantics for function pointers than normal function calls, it's a
> > piece of
> > crap.  i dont think any ports of u-boot need worry about this.
>
> agreed there.
>
> The main reason for me doing this was to provide a "generic" mechanism
> that mimic how we boot kernels on PPC.
>
> The other issue w/providing this on all arch's is that the number of
> params passed via regs may vary and I don't know what is for arm,
> avr32, blackfin, i386, m68k, microblaze, mips, nios, nios2, and sh.
>
> and each variation will require a different prototype.
>
> if someone can tell me what the number of args are for each arch we
> can see about providing this everywhere.

are you sure you actually need to worry about such things ?  i'm pretty sure 
any relevant architecture says that the first X arguments go via registers 
and the rest go on the stack.  let the compiler worry about the exact value 
for X.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080215/05dde53e/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Haavard Skinnemoen]

On Fri, 15 Feb 2008 02:08:23 -0500
Mike Frysinger <vapier@gentoo.org> wrote:

> > if someone can tell me what the number of args are for each arch we
> > can see about providing this everywhere.  
> 
> are you sure you actually need to worry about such things ?  i'm pretty sure 
> any relevant architecture says that the first X arguments go via registers 
> and the rest go on the stack.  let the compiler worry about the exact value 
> for X.

I guess the real question is whether it's acceptable to pass arguments
on the stack to such subprograms. If the answer is yes, why not simply
use varargs?

If the answer is no: avr32 can pass up to 5 arguments through registers.

Haavard

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Haavard Skinnemoen wrote:
> Mike Frysinger <vapier@gentoo.org> wrote:
> > > if someone can tell me what the number of args are for each arch we
> > > can see about providing this everywhere.
> >
> > are you sure you actually need to worry about such things ?  i'm pretty
> > sure any relevant architecture says that the first X arguments go via
> > registers and the rest go on the stack.  let the compiler worry about the
> > exact value for X.
>
> I guess the real question is whether it's acceptable to pass arguments
> on the stack to such subprograms. If the answer is yes, why not simply
> use varargs?

if the goal is to keep things in registers, "call" and its description are way 
too vague.  the description makes me think this is meant to be a "go" for 
functions.

> If the answer is no: avr32 can pass up to 5 arguments through registers.

x86 does 0 while Blackfin does 3.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080215/90d02255/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Kumar Gala]

On Feb 15, 2008, at 1:29 AM, Mike Frysinger wrote:

> On Friday 15 February 2008, Haavard Skinnemoen wrote:
>> Mike Frysinger <vapier@gentoo.org> wrote:
>>>> if someone can tell me what the number of args are for each arch we
>>>> can see about providing this everywhere.
>>>
>>> are you sure you actually need to worry about such things ?  i'm  
>>> pretty
>>> sure any relevant architecture says that the first X arguments go  
>>> via
>>> registers and the rest go on the stack.  let the compiler worry  
>>> about the
>>> exact value for X.
>>
>> I guess the real question is whether it's acceptable to pass  
>> arguments
>> on the stack to such subprograms. If the answer is yes, why not  
>> simply
>> use varargs?
>
> if the goal is to keep things in registers, "call" and its  
> description are way
> too vague.  the description makes me think this is meant to be a  
> "go" for
> functions.

As per Wolfgang's desires the help is terse, the commit message was  
suppose to convey more details.  But I can see where the confusion  
comes from.  I'm open to better wording and will try to think of some  
myself.

>> If the answer is no: avr32 can pass up to 5 arguments through  
>> registers.
>
> x86 does 0 while Blackfin does 3.

3 arch's down, 6 to go.

- k

[U-Boot-Users] [PATCH] Add call command on PPC

[Haavard Skinnemoen]

On Fri, 15 Feb 2008 02:29:39 -0500
Mike Frysinger <vapier@gentoo.org> wrote:

> x86 does 0 while Blackfin does 3.

x86 can do more (at least 3) if you use the regparm attribute. But you
obviously need to keep both sides in sync.

Haavard

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Haavard Skinnemoen wrote:
> Mike Frysinger <vapier@gentoo.org> wrote:
> > x86 does 0 while Blackfin does 3.
>
> x86 can do more (at least 3) if you use the regparm attribute. But you
> obviously need to keep both sides in sync.

i'm aware of all the neat tricks you can do with gcc, but we were talking 
about the ABI standard (which the gcc tricks violate)
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080215/1ee98017/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Kumar Gala]

On Feb 15, 2008, at 1:21 AM, Haavard Skinnemoen wrote:

> On Fri, 15 Feb 2008 02:08:23 -0500
> Mike Frysinger <vapier@gentoo.org> wrote:
>
>>> if someone can tell me what the number of args are for each arch we
>>> can see about providing this everywhere.
>>
>> are you sure you actually need to worry about such things ?  i'm  
>> pretty sure
>> any relevant architecture says that the first X arguments go via  
>> registers
>> and the rest go on the stack.  let the compiler worry about the  
>> exact value
>> for X.
>
> I guess the real question is whether it's acceptable to pass arguments
> on the stack to such subprograms. If the answer is yes, why not simply
> use varargs?

for my purposes the answer is no.
>
> If the answer is no: avr32 can pass up to 5 arguments through  
> registers.

thanks

- k

[U-Boot-Users] [PATCH] Add call command on PPC

[Wolfgang Denk]

In message <19411A28-BCCB-46D1-A80A-6C31657757D2@kernel.crashing.org> you wrote:
> 
> > I guess the real question is whether it's acceptable to pass arguments
> > on the stack to such subprograms. If the answer is yes, why not simply
> > use varargs?
> 
> for my purposes the answer is no.

Then you cannot use C code. Or you rely on the ABI, and then you can
use "go" as well.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Generally speaking, there are other ways to accomplish whatever it is
that you think you need ...                               - Doug Gwyn

[U-Boot-Users] [PATCH] Add call command on PPC

[Wolfgang Denk]

In message <ECB2C6CF-0AB8-44AE-AD35-769B741B0528@kernel.crashing.org> you wrote:
> 
> The main reason for me doing this was to provide a "generic" mechanism  
> that mimic how we boot kernels on PPC.

May I ask why exactly this is needed at all? Kernels should be booted
using the bootm command - if somethingis missing there, it should  be
added  /  fixed  in  this  contexzt  instead of adding another highly
specialized command.

And *if* we add this, then please let's make it general enough so  we
can use it on all architectures without adding a lot of #ifdef mess.

> The other issue w/providing this on all arch's is that the number of  
> params passed via regs may vary and I don't know what is for arm,  
> avr32, blackfin, i386, m68k, microblaze, mips, nios, nios2, and sh.
> and each variation will require a different prototype.

No, just change the design, then. Make it a varargs function and pass
the number of registers as one argument.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
Death. Destruction. Disease. Horror. That's what war  is  all  about.
That's what makes it a thing to be avoided.
	-- Kirk, "A Taste of Armageddon", stardate 3193.0

[U-Boot-Users] [PATCH] Add call command on PPC

[Mike Frysinger]

On Friday 15 February 2008, Kumar Gala wrote:
> The main reason for me doing this was to provide a "generic" mechanism
> that mimic how we boot kernels on PPC.

is that the only reason ?  what format is the kernel that you're attempting to 
boot ?  assuming it's an ELF ...

at the moment, the Blackfin code base maintained on blackfin.uclinux.org has 
hacked the "bootelf" command to assume it is always booting a kernel.  this 
way we can load/boot the ELF image of our kernels in memory and have the 
$(bootargs) and such actually get passed properly (dont yell at me for this, 
the change was before my time, and i agree it's crap :p).

in other words, a common function for booting a linux ELF kernel image would 
be great.  especially considering the fact we have a "bootvx" that takes an 
ELF image, sets up the bootargs, and then boots it properly.  seems messed up 
that booting vxworks sees better support than booting linux ;)
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 827 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20080216/9472ea25/attachment.pgp 

[U-Boot-Users] [PATCH] Add call command on PPC

[Wolfgang Denk]

In message <Pine.LNX.4.64.0802142341250.31987@blarg.am.freescale.net> you wrote:
> The call command tries to mimic a function call in that the 'arg's to
> the command are passed in registers according to the PPC ABI.
> 
> The prototype that call mimics is some variation of the following
> depending on how many arguments are passed to the command:
> 
> void func(void)
> void func(unsigned long r3)
> void func(unsigned long r3, unsigned long r4)
> ...
> void func(unsigned long r3, unsigned long r4, ... unsigned long r10)
> 
> The maximum number of 'arg's is 8.  There are no arguments passed on
> a stack, no floating point or vector arguments.

Sorry, but this looks still broken to me.

Best regards,

Wolfgang Denk

-- 
DENX Software Engineering GmbH,     MD: Wolfgang Denk & Detlev Zundel
HRB 165235 Munich, Office: Kirchenstr.5, D-82194 Groebenzell, Germany
Phone: (+49)-8142-66989-10 Fax: (+49)-8142-66989-80 Email: wd at denx.de
A Puritan is someone who is deathly afraid that  someone,  somewhere,
is having fun.