[PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[Huang, Ying]

This patch add a field of 64-bit physical pointer to NULL terminated
single linked list of struct setup_data to real-mode kernel
header. This is used as a more extensible boot parameters passing
mechanism.

This patch has been tested against 2.6.23-rc6-mm1 kernel on x86_64. It
is based on the proposal of Peter Anvin.


Known Issues:

1. Where is safe to place the linked list of setup_data?
Because the length of the linked list of setup_data is variable, it
can not be copied into BSS segment of kernel as that of "zero
page". We must find a safe place for it, where it will not be
overwritten by kernel during booting up. The i386 kernel will
overwrite some pages after _end. The x86_64 kernel will overwrite some
pages from 0x1000 on.


ChangeLog:

-- v2 --

- Increase the boot protocol version number.
- Check version number before parsing setup_data.


Signed-off-by: Huang Ying <ying.huang@intel.com>

---

 arch/i386/Kconfig            |    3 ---
 arch/i386/boot/header.S      |    8 +++++++-
 arch/i386/kernel/setup.c     |   22 ++++++++++++++++++++++
 arch/x86_64/kernel/setup.c   |   21 +++++++++++++++++++++
 include/asm-i386/bootparam.h |   15 +++++++++++++++
 include/asm-i386/io.h        |    7 +++++++
 6 files changed, 72 insertions(+), 4 deletions(-)

Index: linux-2.6.23-rc6/include/asm-i386/bootparam.h
===================================================================
--- linux-2.6.23-rc6.orig/include/asm-i386/bootparam.h	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/include/asm-i386/bootparam.h	2007-09-19 16:41:57.000000000 +0800
@@ -9,6 +9,17 @@
 #include <asm/ist.h>
 #include <video/edid.h>
 
+/* setup data types */
+#define SETUP_NONE			0
+
+/* extensible setup data list node */
+struct setup_data {
+	u64 next;
+	u32 type;
+	u32 len;
+	u8 data[0];
+} __attribute__((packed));
+
 struct setup_header {
 	u8	setup_sects;
 	u16	root_flags;
@@ -41,6 +52,10 @@
 	u32	initrd_addr_max;
 	u32	kernel_alignment;
 	u8	relocatable_kernel;
+	u8	_pad2[3];
+	u32	cmdline_size;
+	u32	_pad3;
+	u64	setup_data;
 } __attribute__((packed));
 
 struct sys_desc_table {
Index: linux-2.6.23-rc6/arch/i386/boot/header.S
===================================================================
--- linux-2.6.23-rc6.orig/arch/i386/boot/header.S	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/arch/i386/boot/header.S	2007-09-19 10:47:34.000000000 +0800
@@ -119,7 +119,7 @@
 	# Part 2 of the header, from the old setup.S
 
 		.ascii	"HdrS"		# header signature
-		.word	0x0206		# header version number (>= 0x0105)
+		.word	0x0207		# header version number (>= 0x0105)
 					# or else old loadlin-1.5 will fail)
 		.globl realmode_swtch
 realmode_swtch:	.word	0, 0		# default_switch, SETUPSEG
@@ -214,6 +214,12 @@
                                                 #added with boot protocol
                                                 #version 2.06
 
+pad4:			.long 0
+
+setup_data:		.quad 0			# 64-bit physical pointer to
+						# single linked list of
+						# struct setup_data
+
 # End of setup header #####################################################
 
 	.section ".inittext", "ax"
Index: linux-2.6.23-rc6/arch/x86_64/kernel/setup.c
===================================================================
--- linux-2.6.23-rc6.orig/arch/x86_64/kernel/setup.c	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/arch/x86_64/kernel/setup.c	2007-09-19 16:41:57.000000000 +0800
@@ -221,6 +221,25 @@
 		ebda_size = 64*1024;
 }
 
+void __init parse_setup_data(void)
+{
+	struct setup_data *setup_data;
+	unsigned long pa_setup_data;
+
+	if (boot_params.hdr.version < 0x0207)
+		return;
+	pa_setup_data = boot_params.hdr.setup_data;
+	while (pa_setup_data) {
+		setup_data = early_ioremap(pa_setup_data, PAGE_SIZE);
+		switch (setup_data->type) {
+		default:
+			break;
+		}
+		pa_setup_data = setup_data->next;
+		early_iounmap(setup_data, PAGE_SIZE);
+	}
+}
+
 void __init setup_arch(char **cmdline_p)
 {
 	printk(KERN_INFO "Command line: %s\n", boot_command_line);
@@ -256,6 +275,8 @@
 	strlcpy(command_line, boot_command_line, COMMAND_LINE_SIZE);
 	*cmdline_p = command_line;
 
+	parse_setup_data();
+
 	parse_early_param();
 
 	finish_e820_parsing();
Index: linux-2.6.23-rc6/arch/i386/kernel/setup.c
===================================================================
--- linux-2.6.23-rc6.orig/arch/i386/kernel/setup.c	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/arch/i386/kernel/setup.c	2007-09-19 10:47:34.000000000 +0800
@@ -496,6 +496,25 @@
 	return machine_specific_memory_setup();
 }
 
+void __init parse_setup_data(void)
+{
+	struct setup_data *setup_data;
+	unsigned long pa_setup_data, pa_next;
+
+	if (boot_params.hdr.version < 0x0207)
+		return;
+	pa_setup_data = boot_params.hdr.setup_data;
+	while (pa_setup_data) {
+		setup_data = boot_ioremap(pa_setup_data, PAGE_SIZE);
+		pa_next = setup_data->next;
+		switch (setup_data->type) {
+		default:
+			break;
+		}
+		pa_setup_data = pa_next;
+	}
+}
+
 /*
  * Determine if we were loaded by an EFI loader.  If so, then we have also been
  * passed the efi memmap, systab, etc., so we should use these data structures
@@ -544,6 +563,9 @@
 	rd_prompt = ((boot_params.hdr.ram_size & RAMDISK_PROMPT_FLAG) != 0);
 	rd_doload = ((boot_params.hdr.ram_size & RAMDISK_LOAD_FLAG) != 0);
 #endif
+
+	parse_setup_data();
+
 	ARCH_SETUP
 	if (efi_enabled)
 		efi_init();
Index: linux-2.6.23-rc6/include/asm-i386/io.h
===================================================================
--- linux-2.6.23-rc6.orig/include/asm-i386/io.h	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/include/asm-i386/io.h	2007-09-19 10:47:34.000000000 +0800
@@ -126,6 +126,13 @@
 extern void iounmap(volatile void __iomem *addr);
 
 /*
+ * boot_ioremap is for temporary early boot-time mappings, before
+ * paging_init(), when the boot-time page tables are still in use.
+ * The mapping is currently limited to at most 4 pages.
+ */
+extern void *boot_ioremap(unsigned long offset, unsigned long size);
+
+/*
  * bt_ioremap() and bt_iounmap() are for temporary early boot-time
  * mappings, before the real ioremap() is functional.
  * A boot-time mapping is currently limited to at most 16 pages.
Index: linux-2.6.23-rc6/arch/i386/Kconfig
===================================================================
--- linux-2.6.23-rc6.orig/arch/i386/Kconfig	2007-09-19 10:22:02.000000000 +0800
+++ linux-2.6.23-rc6/arch/i386/Kconfig	2007-09-19 10:47:34.000000000 +0800
@@ -773,11 +773,8 @@
  	  The default yes will allow the kernel to do irq load balancing.
 	  Saying no will keep the kernel from doing irq load balancing.
 
-# turning this on wastes a bunch of space.
-# Summit needs it only when NUMA is on
 config BOOT_IOREMAP
 	bool
-	depends on (((X86_SUMMIT || X86_GENERICARCH) && NUMA) || (X86 && EFI))
 	default y
 
 config SECCOMP

Re: [PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[H. Peter Anvin]

Huang, Ying wrote:
> 
> Known Issues:
> 
> 1. Where is safe to place the linked list of setup_data?
> Because the length of the linked list of setup_data is variable, it
> can not be copied into BSS segment of kernel as that of "zero
> page". We must find a safe place for it, where it will not be
> overwritten by kernel during booting up. The i386 kernel will
> overwrite some pages after _end. The x86_64 kernel will overwrite some
> pages from 0x1000 on.
> 

The latter is definitely not safe, since the space below 640K is the
documented place to put the command line (and presumably where the
bootloader would put other auxilliary chunks.)

I'll try to do a full review of this later today.  Haven't had time yet
to look at this anything than but piecemeal.

	-hpa

Re: [PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[Jeremy Fitzhardinge]

Huang, Ying wrote:
> This patch add a field of 64-bit physical pointer to NULL terminated
> single linked list of struct setup_data to real-mode kernel
> header. This is used as a more extensible boot parameters passing
> mechanism.
>
> This patch has been tested against 2.6.23-rc6-mm1 kernel on x86_64. It
> is based on the proposal of Peter Anvin.
>
>
> Known Issues:
>
> 1. Where is safe to place the linked list of setup_data?
> Because the length of the linked list of setup_data is variable, it
> can not be copied into BSS segment of kernel as that of "zero
> page". We must find a safe place for it, where it will not be
> overwritten by kernel during booting up. The i386 kernel will
> overwrite some pages after _end. The x86_64 kernel will overwrite some
> pages from 0x1000 on.
>   

Do you actually need a linked list of data?  This is similar to the
changes to bzImage to support booting bzImage a paravirt environment,
but I just proposed a pointer to a single info structure, along with a
field to identify the boot environment (ie, native/lguest/xen etc).  It
would be easy to extend this to handle EFI as just another boot
environment, and you could hang a list of structures off the pointer.


    J

Re: [PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[H. Peter Anvin]

Jeremy Fitzhardinge wrote:
> 
> Do you actually need a linked list of data?  This is similar to the
> changes to bzImage to support booting bzImage a paravirt environment,
> but I just proposed a pointer to a single info structure, along with a
> field to identify the boot environment (ie, native/lguest/xen etc).  It
> would be easy to extend this to handle EFI as just another boot
> environment, and you could hang a list of structures off the pointer.
> 

*He* doesn't, but *we* do.  We have already run into at least one case
where the existing structure is insufficient (EDD overhaul), and so we
need to do something extensible.

	-hpa

Re: [PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[Huang, Ying]

Hi, Peter,

On Wed, 2007-09-19 at 09:04 -0700, H. Peter Anvin wrote:
> Huang, Ying wrote:
> > 
> > Known Issues:
> > 
> > 1. Where is safe to place the linked list of setup_data?
> > Because the length of the linked list of setup_data is variable, it
> > can not be copied into BSS segment of kernel as that of "zero
> > page". We must find a safe place for it, where it will not be
> > overwritten by kernel during booting up. The i386 kernel will
> > overwrite some pages after _end. The x86_64 kernel will overwrite some
> > pages from 0x1000 on.
> > 
> 
> The latter is definitely not safe, since the space below 640K is the
> documented place to put the command line (and presumably where the
> bootloader would put other auxilliary chunks.)
> 
> I'll try to do a full review of this later today.  Haven't had time yet
> to look at this anything than but piecemeal.

Do you think this patch and the 32-bit boot protocol patch are ready to
merge for -mm? If not, I can revise them.

Best Regards,
Huang Ying

Re: [PATCH -mm -v3 1/2] i386/x86_64 boot: setup data

[H. Peter Anvin]

Huang, Ying wrote:
> Hi, Peter,
> 
> On Wed, 2007-09-19 at 09:04 -0700, H. Peter Anvin wrote:
>> Huang, Ying wrote:
>>> Known Issues:
>>>
>>> 1. Where is safe to place the linked list of setup_data?
>>> Because the length of the linked list of setup_data is variable, it
>>> can not be copied into BSS segment of kernel as that of "zero
>>> page". We must find a safe place for it, where it will not be
>>> overwritten by kernel during booting up. The i386 kernel will
>>> overwrite some pages after _end. The x86_64 kernel will overwrite some
>>> pages from 0x1000 on.
>>>
>> The latter is definitely not safe, since the space below 640K is the
>> documented place to put the command line (and presumably where the
>> bootloader would put other auxilliary chunks.)
>>
>> I'll try to do a full review of this later today.  Haven't had time yet
>> to look at this anything than but piecemeal.
> 
> Do you think this patch and the 32-bit boot protocol patch are ready to
> merge for -mm? If not, I can revise them.
> 

Sorry, haven't had a chance to look at it in proper detail yet, mostly
due to debugging, but one thing I'd like to see is both the boot_params
structure as well as all the chained information pointers exported into
sysfs.  The experience with the x86 setup code has shown that it would
help immensely with debugging, not to mention being available to tools
like kexec.

	-hpa