x86_64 built-in command line

PATCH In review Matt Mackall 6 participants 18 messages Jul 31, 2006 – Aug 1, 2006

lore

Share & tools

lore

Matt Mackall

author

2006-07-31 17:16 UTC #0 lore raw

PATCH 1/1 x86_64 built-in command line

Allow setting a command line at build time on x86_64. Compiled but not tested.

Signed-off-by Matt Mackall <[email protected]>

arch/x86_64/Kconfig +14 -0

Index: linux/arch/x86_64/Kconfig
===================================================================
@@ -558,6 +558,20 @@ config K8_NB
 	def_bool y
 	depends on AGP_AMD64 || IOMMU || (PCI && NUMA)
 
+config CMDLINE_BOOL
+	bool "Default bootloader kernel arguments" if EMBEDDED
+
+config CMDLINE
+	string "Initial kernel command string" if EMBEDDED
+	depends on CMDLINE_BOOL
+	default "root=/dev/hda1 ro"
+	help
+	  On some systems, there is no way for the boot loader to pass
+	  arguments to the kernel. For these platforms, you can supply
+	  some command-line options at build time by entering them
+	  here. In most cases you will need to specify the root device
+	  here.
+
 endmenu
 
 #

arch/x86_64/kernel/setup.c +4 -0

Index: linux/arch/x86_64/kernel/setup.c
===================================================================
@@ -289,6 +289,10 @@ static __init void parse_cmdline_early (
 	int len = 0;
 	int userdef = 0;
 
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
+
 	for (;;) {
 		if (c != ' ') 
 			goto next_char;

Signature · 1 line

Mathematics is the supreme nostalgia of our time.

linux-os (Dick Johnson)

2006-07-31 17:47 UTC #1 lore raw

Quoted 47 lines · Matt Mackall

Allow setting a command line at build time on x86_64. Compiled but not tested.

Signed-off-by Matt Mackall <[email protected]>

Index: linux/arch/x86_64/Kconfig
===================================================================

arch/x86_64/Kconfig +14 -0

@@ -558,6 +558,20 @@ config K8_NB
 	def_bool y
 	depends on AGP_AMD64 || IOMMU || (PCI && NUMA)
 
+config CMDLINE_BOOL
+	bool "Default bootloader kernel arguments" if EMBEDDED
+
+config CMDLINE
+	string "Initial kernel command string" if EMBEDDED
+	depends on CMDLINE_BOOL
+	default "root=/dev/hda1 ro"
+	help
+	  On some systems, there is no way for the boot loader to pass
+	  arguments to the kernel. For these platforms, you can supply
+	  some command-line options at build time by entering them
+	  here. In most cases you will need to specify the root device
+	  here.
+

endmenu

#
Index: linux/arch/x86_64/kernel/setup.c
===================================================================

arch/x86_64/kernel/setup.c +4 -1

@@ -289,6 +289,10 @@ static __init void parse_cmdline_early (
 	int len = 0;
 	int userdef = 0;
 
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
+
 	for (;;) {
 		if (c != ' ')
 			goto next_char;
 
--

Quoted 2 lines

Mathematics is the supreme nostalgia of our time.
-

But this just makes it nice for __your__ embedded system. I suggest you use:

arch/x86_64/kernel/setup.c

Quoted 1 line

arch/x86_64/kernel/setup.c +1 -0
+	default "root=/dev/root ro"

The boot setup code makes a symlink to whatever your specific setup requires.

Cheers,
Dick Johnson
Penguin : Linux version 2.6.16.24 on an i686 machine (5592.62 BogoMips). New book: http://www.AbominableFirebug.com/
_

****************************************************************
The information transmitted in this message is confidential and may be privileged. Any review, retransmission, dissemination, or other use of this information by persons or entities other than the intended recipient is prohibited. If you are not the intended recipient, please notify Analogic Corporation immediately - by replying to this message or by sending an email to [email protected] - and destroy all copies of this information, including any attachments, without reading or disclosing them.

Thank you.

Andi Kleen

2006-07-31 20:08 UTC #2 lore raw

In reply to Matt Mackall

arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294

Quoted 4 lines

arch/x86_64/kernel/setup.c +3 -0

+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif

I think I would prefer a strcat.

Also you should describe the exact behaviour (override/append) in Kconfig help.

-Andi

H. Peter Anvin

2006-07-31 20:41 UTC #3 lore raw

Quoted 9 lines · Andi Kleen

arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294
Quoted 4 lines
arch/x86_64/kernel/setup.c +3 -0
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
I think I would prefer a strcat.
Also you should describe the exact behaviour (override/append) in Kconfig help.

In the i386 thread, Matt described having a firmware bootloader which passes bogus parameters. For that case, it would make sense to have a non-default CONFIG option to have override rather than conjoined (and I maintain that the built-in command line should be prepended.)

-hpa

Andi Kleen

2006-07-31 22:17 UTC #4 lore raw

Quoted 8 lines · H. Peter Anvin

Quoted 7 lines · Andi Kleen
arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294
Quoted 4 lines
arch/x86_64/kernel/setup.c +3 -0
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
I think I would prefer a strcat.

Quoted 7 lines

Quoted 2 lines
Also you should describe the exact behaviour (override/append) in Kconfig help.
In the i386 thread, Matt described having a firmware bootloader which passes bogus parameters. For that case, it would make sense to have a non-default CONFIG option to have override rather than conjoined (and I maintain that the built-in command line should be prepended.)

Is that boot loader common? What's its name?
If not I would prefer that he keeps the one liner patch to deal with that private.

For generic semantics strcat (or possible prepend) is probably better.

-Andi

Matt Mackall

author

2006-08-01 01:44 UTC #5 lore raw

Quoted 17 lines · Andi Kleen

Quoted 8 lines · H. Peter Anvin
Quoted 7 lines · Andi Kleen
arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294
Quoted 4 lines
arch/x86_64/kernel/setup.c +3 -0
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
I think I would prefer a strcat.
Quoted 7 lines
Quoted 2 lines
Also you should describe the exact behaviour (override/append) in Kconfig help.
In the i386 thread, Matt described having a firmware bootloader which passes bogus parameters. For that case, it would make sense to have a non-default CONFIG option to have override rather than conjoined (and I maintain that the built-in command line should be prepended.)

Quoted 5 lines

Is that boot loader common? What's its name?
If not I would prefer that he keeps the one liner patch to deal with that private.
For generic semantics strcat (or possible prepend) is probably better.

No, it doesn't work for numerous kernel options that can't be negated.

Signature · 1 line

Mathematics is the supreme nostalgia of our time.

H. Peter Anvin

2006-08-01 02:29 UTC #6 lore raw

Quoted 11 lines · Matt Mackall

Quoted 10 lines · Andi Kleen
Quoted 9 lines · H. Peter Anvin
Quoted 8 lines · Andi Kleen
arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294
Quoted 4 lines
arch/x86_64/kernel/setup.c +3 -0
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
I think I would prefer a strcat.
Also you should describe the exact behaviour (override/append) in Kconfig help.

Quoted 8 lines

Quoted 5 lines
Quoted 1 line
In the i386 thread, Matt described having a firmware bootloader which passes bogus parameters. For that case, it would make sense to have a non-default CONFIG option to have override rather than conjoined (and I maintain that the built-in command line should be prepended.)
Is that boot loader common? What's its name? If not I would prefer that he keeps the one liner patch to deal with that private.
For generic semantics strcat (or possible prepend) is probably better.
No, it doesn't work for numerous kernel options that can't be negated.

How about we fix the real problem, then?

-hpa

H. Peter Anvin

2006-08-01 02:33 UTC #7 lore raw

In reply to Matt Mackall

Quoted 5 lines · Matt Mackall

Quoted 2 lines
For generic semantics strcat (or possible prepend) is probably better.
No, it doesn't work for numerous kernel options that can't be negated.

Anyway, there are four possible behaviours: builtin only, bootloader only, buildin+bootloader, and bootloader+builtin. I believe the last is the most appropriate default, but there really is no reason why this can't be a four-way configuration option.

-hpa

H. Peter Anvin

2006-08-01 02:38 UTC #8 lore raw

In reply to Matt Mackall

Actually, the best thing to do might be to designate a symbol (say &, like in automount) as "insert the boot loader command line here."

That way you can specify things in the builtin command line that are both prepended and appended to the boot loader command, and if you wish, you can emit it completely.

The default would be just "&".

-hpa

Matt Mackall

author

2006-08-01 02:42 UTC #9 lore raw

Quoted 8 lines · H. Peter Anvin

Actually, the best thing to do might be to designate a symbol (say &, like in automount) as "insert the boot loader command line here."
That way you can specify things in the builtin command line that are both prepended and appended to the boot loader command, and if you wish, you can emit it completely.
The default would be just "&".

That idea doesn't suck. I'll take a look at it.

We still have a problem with overriding things like console=tty, for which there's no obvious fix that doesn't break the world.

Signature · 1 line

Mathematics is the supreme nostalgia of our time.

Andi Kleen

2006-08-01 02:44 UTC #10 lore raw

Quoted 11 lines · Matt Mackall

Quoted 8 lines · H. Peter Anvin
Actually, the best thing to do might be to designate a symbol (say &, like in automount) as "insert the boot loader command line here."
That way you can specify things in the builtin command line that are both prepended and appended to the boot loader command, and if you wish, you can emit it completely.
The default would be just "&".
That idea doesn't suck. I'll take a look at it.

With %s it would be much less code to write.

-Andi

H. Peter Anvin

2006-08-01 02:46 UTC #11 lore raw

Quoted 11 lines · Andi Kleen

Quoted 7 lines · Matt Mackall
Quoted 5 lines · H. Peter Anvin
Actually, the best thing to do might be to designate a symbol (say &, like in automount) as "insert the boot loader command line here."
That way you can specify things in the builtin command line that are both prepended and appended to the boot loader command, and if you wish, you can emit it completely.
The default would be just "&".
That idea doesn't suck. I'll take a look at it.
With %s it would be much less code to write.

True; it would just mean that the buildin command line would have to double %'s, but that's not really a problem.

On the other hand, a single-character substitution loop is hardly complex.

-hpa

Andi Kleen

2006-08-01 02:47 UTC #12 lore raw

Quoted 3 lines

True; it would just mean that the buildin command line would have to double %'s, but that's not really a problem.

Similar for &s

-Andi

H. Peter Anvin

2006-08-01 02:47 UTC #13 lore raw

In reply to Matt Mackall

Quoted 6 lines · Matt Mackall

That idea doesn't suck. I'll take a look at it.
We still have a problem with overriding things like console=tty, for which there's no obvious fix that doesn't break the world.

One could define "console=" (empty argument) as a command to reset the list. It probably would be useful.

-hpa

Andrew Morton

2006-08-01 02:49 UTC #14 lore raw

In reply to Matt Mackall

Quoted 18 lines · On Mon, 31 Jul 2006 20:43:19 -0500 Matt Mackall <[email protected]> wrote:

Quoted 17 lines · Andi Kleen
Quoted 15 lines · H. Peter Anvin
Quoted 9 lines · Andi Kleen
arch/x86_64/kernel/setup.c static __init void parse_cmdline_early ( · line 294
Quoted 4 lines
arch/x86_64/kernel/setup.c +3 -0
+#ifdef CONFIG_CMDLINE_BOOL
+	strlcpy(saved_command_line, CONFIG_CMDLINE, COMMAND_LINE_SIZE);
+#endif
I think I would prefer a strcat.
Also you should describe the exact behaviour (override/append) in Kconfig help.
In the i386 thread, Matt described having a firmware bootloader which passes bogus parameters. For that case, it would make sense to have a non-default CONFIG option to have override rather than conjoined (and I maintain that the built-in command line should be prepended.)

Quoted 7 lines

Quoted 5 lines
Is that boot loader common? What's its name?
If not I would prefer that he keeps the one liner patch to deal with that private.
For generic semantics strcat (or possible prepend) is probably better.
No, it doesn't work for numerous kernel options that can't be negated.

I think what I'm hearing is that we need three options: append, prepend or replace.

Deepak Saxena

2006-08-01 01:05 UTC #15 lore raw

In reply to Matt Mackall

On Jul 31 2006, at 12:14, Matt Mackall was caught saying:

Quoted 2 lines

Allow setting a command line at build time on x86_64. Compiled but not tested.

Can we just make this into a generic option and put the relevant strcpy (strcat) in init/main.c. We've supported a default in-kernel command line on ARM for sometime now and I think it would be best to just have a single implementation.

~Deepak

Signature · 3 lines

Deepak Saxena - [email protected] - http://www.plexity.net

"An open heart has no possessions, only experiences" - Matt Bibbeau

Andi Kleen

2006-08-01 01:18 UTC #16 lore raw

Quoted 8 lines · Deepak Saxena

On Jul 31 2006, at 12:14, Matt Mackall was caught saying:
Quoted 2 lines
Allow setting a command line at build time on x86_64. Compiled but not tested.
Can we just make this into a generic option and put the relevant strcpy (strcat) in init/main.c. We've supported a default in-kernel command line on ARM for sometime now and I think it would be best to just have a single implementation.

One problem is that some architectures (like i386/x86-64) have early boot arguments that are processed before init/main

-Andi

H. Peter Anvin

2006-08-01 03:47 UTC #17 lore raw

In reply to Deepak Saxena

Quoted 9 lines · Deepak Saxena

On Jul 31 2006, at 12:14, Matt Mackall was caught saying:
Quoted 2 lines
Allow setting a command line at build time on x86_64. Compiled but not tested.
Can we just make this into a generic option and put the relevant strcpy (strcat) in init/main.c. We've supported a default in-kernel command line on ARM for sometime now and I think it would be best to just have a single implementation.

Yes on the generic option, no on putting it in init/main.c... on a lot of architectures, some parsing of the command line happens *very* early, sometimes even before the C code is invoked.

-hpa