Re: [PATCH] x86,setup: add serial_console_port_base in boot_params

Re: [PATCH] x86,setup: add serial_console_port_base in boot_params

[Yinghai Lu]

On 07/31/2010 02:02 PM, H. Peter Anvin wrote:
> On 07/31/2010 11:32 AM, Cyrill Gorcunov wrote:
>>>
>>> No, this is the internal part of the boot protocol, so it's not an issue.
>>>
>>
>> Peter, I didn't mean any issue here, I meant that bootloaders don't know about
>> this field yet and they will have to update own sources to pass port value
>> at proper place of boot params. Or I miss something?
>>
> 
> Boot loaders that use the 16-bit entry point are unaffected.
> 
> Boot loaders which use the 32-bit entry point but properly clears the
> zero page simply will not have the feature.
> 
> Boot loaders which use the 32-bit entry point but doesn't clear the zero
> page are broken.
> 
can you if this one is right for kexec path?

Thanks

Yinghai

[PATCH] kexec-tools: pass serial console base for early console in kernel

when kexec is used with "console_serial", set serial_console_port_base in boot_params.

Signed-off-by: Yinghai Lu <yinghai@kernel.org>

---
 doc/linux-i386-zero-page.txt      |    1 +
 include/x86/x86-linux.h           |    3 ++-
 kexec/arch/i386/kexec-x86.h       |    1 +
 kexec/arch/i386/x86-linux-setup.c |    4 ++++
 kexec/arch/x86_64/kexec-x86_64.c  |    9 +--------
 5 files changed, 9 insertions(+), 9 deletions(-)

Index: kexec-tools/doc/linux-i386-zero-page.txt
===================================================================
--- kexec-tools.orig/doc/linux-i386-zero-page.txt
+++ kexec-tools/doc/linux-i386-zero-page.txt
@@ -22,6 +22,7 @@ Offset	Type		Description
 			  0x90000 + contents of CL_OFFSET
 			(only taken, when CL_MAGIC = 0xA33F)
  0x40	20 bytes	struct apm_bios_info, APM_BIOS_INFO
+ 0x54   2 bytes		serial console base
  0x60	16 bytes	Intel SpeedStep (IST) BIOS support information
  0x80	16 bytes	hd0-disk-parameter from intvector 0x41
  0x90	16 bytes	hd1-disk-parameter from intvector 0x46
Index: kexec-tools/include/x86/x86-linux.h
===================================================================
--- kexec-tools.orig/include/x86/x86-linux.h
+++ kexec-tools/include/x86/x86-linux.h
@@ -40,7 +40,6 @@ struct apm_bios_info {
 	uint16_t cseg_len;	/* 0x4e */
 	uint16_t cseg_16_len;	/* 0x50 */
 	uint16_t dseg_len;	/* 0x52 */
-	uint8_t  reserved[44];	/* 0x54 */
 };
 
 /*
@@ -105,6 +104,8 @@ struct x86_linux_param_header {
 	uint8_t  reserved4[12];			/* 0x34 -- 0x3f reserved for future expansion */
 
 	struct apm_bios_info apm_bios_info;	/* 0x40 */
+	uint16_t  serial_console_port_base;	/* 0x54 */
+	uint8_t  reserved41[42];		/* 0x56 */
 	struct drive_info_struct drive_info;	/* 0x80 */
 	struct sys_desc_table sys_desc_table;	/* 0xa0 */
 	uint32_t alt_mem_k;			/* 0x1e0 */
Index: kexec-tools/kexec/arch/i386/x86-linux-setup.c
===================================================================
--- kexec-tools.orig/kexec/arch/i386/x86-linux-setup.c
+++ kexec-tools/kexec/arch/i386/x86-linux-setup.c
@@ -100,6 +100,9 @@ void setup_linux_bootloader_parameters(
 	cmdline_ptr = ((char *)real_mode) + cmdline_offset;
 	memcpy(cmdline_ptr, cmdline, cmdline_len);
 	cmdline_ptr[cmdline_len - 1] = '\0';
+
+	if (arch_options.console_serial)
+		real_mode->serial_console_port_base = arch_options.serial_base;
 }
 
 int setup_linux_vesafb(struct x86_linux_param_header *real_mode)
@@ -422,6 +425,7 @@ void setup_linux_system_parameters(struc
 
 	/* Default APM info */
 	memset(&real_mode->apm_bios_info, 0, sizeof(real_mode->apm_bios_info));
+
 	/* Default drive info */
 	memset(&real_mode->drive_info, 0, sizeof(real_mode->drive_info));
 	/* Default sysdesc table */
Index: kexec-tools/kexec/arch/i386/kexec-x86.h
===================================================================
--- kexec-tools.orig/kexec/arch/i386/kexec-x86.h
+++ kexec-tools/kexec/arch/i386/kexec-x86.h
@@ -51,6 +51,7 @@ struct arch_options_t {
 	uint8_t  	console_serial;
 	enum coretype	core_header_type;
 };
+extern struct arch_options_t arch_options;
 
 int multiboot_x86_probe(const char *buf, off_t len);
 int multiboot_x86_load(int argc, char **argv, const char *buf, off_t len,
Index: kexec-tools/kexec/arch/x86_64/kexec-x86_64.c
===================================================================
--- kexec-tools.orig/kexec/arch/x86_64/kexec-x86_64.c
+++ kexec-tools/kexec/arch/x86_64/kexec-x86_64.c
@@ -55,14 +55,7 @@ void arch_usage(void)
 		);
 }
 
-static struct {
-	uint8_t  reset_vga;
-	uint16_t serial_base;
-	uint32_t serial_baud;
-	uint8_t  console_vga;
-	uint8_t  console_serial;
-	int core_header_type;
-} arch_options = {
+struct arch_options_t arch_options = {
 	.reset_vga   = 0,
 	.serial_base = 0x3f8,
 	.serial_baud = 0,

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [PATCH] x86,setup: add serial_console_port_base in boot_params

[Eric W. Biederman]

Yinghai Lu <yinghai@kernel.org> writes:

> On 07/31/2010 02:02 PM, H. Peter Anvin wrote:
>> On 07/31/2010 11:32 AM, Cyrill Gorcunov wrote:
>>>>
>>>> No, this is the internal part of the boot protocol, so it's not an issue.
>>>>
>>>
>>> Peter, I didn't mean any issue here, I meant that bootloaders don't know about
>>> this field yet and they will have to update own sources to pass port value
>>> at proper place of boot params. Or I miss something?
>>>
>> 
>> Boot loaders that use the 16-bit entry point are unaffected.
>> 
>> Boot loaders which use the 32-bit entry point but properly clears the
>> zero page simply will not have the feature.
>> 
>> Boot loaders which use the 32-bit entry point but doesn't clear the zero
>> page are broken.
>> 
> can you if this one is right for kexec path?

I am walking out the door, but this seems like nonsense to me.

The kexec console_serial option today is a debugging feature for
when you to debug the purgatory code.  Only once in a bluemoon
should it be useful, so I don't see why we would add it here.
The kexec purgatory is silent by default.

Further I don't see why we would add something to the zero page
when we have a perfectly good way to pass this information via
the kernel command line.  strstr and strtoul are trivial little
functions so I don't see why anything would need to parse anything
other than console= or early_printk=.  The difference in code size
is negligible.

This option in the zero page appears very fragile. There are a lot of
dimensions that we are skipping.  I admit we can detect most of them
by reading the serial port.  But we still have the assumption that the
serial port is an 8259 serial port in i/o space and not mmio space.

*runs out the door before another silly email comes*

Eric
>
> [PATCH] kexec-tools: pass serial console base for early console in kernel
>
> when kexec is used with "console_serial", set serial_console_port_base in boot_params.
>
> Signed-off-by: Yinghai Lu <yinghai@kernel.org>
>
> ---
>  doc/linux-i386-zero-page.txt      |    1 +
>  include/x86/x86-linux.h           |    3 ++-
>  kexec/arch/i386/kexec-x86.h       |    1 +
>  kexec/arch/i386/x86-linux-setup.c |    4 ++++
>  kexec/arch/x86_64/kexec-x86_64.c  |    9 +--------
>  5 files changed, 9 insertions(+), 9 deletions(-)
>
> Index: kexec-tools/doc/linux-i386-zero-page.txt
> ===================================================================
> --- kexec-tools.orig/doc/linux-i386-zero-page.txt
> +++ kexec-tools/doc/linux-i386-zero-page.txt
> @@ -22,6 +22,7 @@ Offset	Type		Description
>  			  0x90000 + contents of CL_OFFSET
>  			(only taken, when CL_MAGIC = 0xA33F)
>   0x40	20 bytes	struct apm_bios_info, APM_BIOS_INFO
> + 0x54   2 bytes		serial console base
>   0x60	16 bytes	Intel SpeedStep (IST) BIOS support information
>   0x80	16 bytes	hd0-disk-parameter from intvector 0x41
>   0x90	16 bytes	hd1-disk-parameter from intvector 0x46
> Index: kexec-tools/include/x86/x86-linux.h
> ===================================================================
> --- kexec-tools.orig/include/x86/x86-linux.h
> +++ kexec-tools/include/x86/x86-linux.h
> @@ -40,7 +40,6 @@ struct apm_bios_info {
>  	uint16_t cseg_len;	/* 0x4e */
>  	uint16_t cseg_16_len;	/* 0x50 */
>  	uint16_t dseg_len;	/* 0x52 */
> -	uint8_t  reserved[44];	/* 0x54 */
>  };
>  
>  /*
> @@ -105,6 +104,8 @@ struct x86_linux_param_header {
>  	uint8_t  reserved4[12];			/* 0x34 -- 0x3f reserved for future expansion */
>  
>  	struct apm_bios_info apm_bios_info;	/* 0x40 */
> +	uint16_t  serial_console_port_base;	/* 0x54 */
> +	uint8_t  reserved41[42];		/* 0x56 */
>  	struct drive_info_struct drive_info;	/* 0x80 */
>  	struct sys_desc_table sys_desc_table;	/* 0xa0 */
>  	uint32_t alt_mem_k;			/* 0x1e0 */
> Index: kexec-tools/kexec/arch/i386/x86-linux-setup.c
> ===================================================================
> --- kexec-tools.orig/kexec/arch/i386/x86-linux-setup.c
> +++ kexec-tools/kexec/arch/i386/x86-linux-setup.c
> @@ -100,6 +100,9 @@ void setup_linux_bootloader_parameters(
>  	cmdline_ptr = ((char *)real_mode) + cmdline_offset;
>  	memcpy(cmdline_ptr, cmdline, cmdline_len);
>  	cmdline_ptr[cmdline_len - 1] = '\0';
> +
> +	if (arch_options.console_serial)
> +		real_mode->serial_console_port_base = arch_options.serial_base;
>  }
>  
>  int setup_linux_vesafb(struct x86_linux_param_header *real_mode)
> @@ -422,6 +425,7 @@ void setup_linux_system_parameters(struc
>  
>  	/* Default APM info */
>  	memset(&real_mode->apm_bios_info, 0, sizeof(real_mode->apm_bios_info));
> +
>  	/* Default drive info */
>  	memset(&real_mode->drive_info, 0, sizeof(real_mode->drive_info));
>  	/* Default sysdesc table */
> Index: kexec-tools/kexec/arch/i386/kexec-x86.h
> ===================================================================
> --- kexec-tools.orig/kexec/arch/i386/kexec-x86.h
> +++ kexec-tools/kexec/arch/i386/kexec-x86.h
> @@ -51,6 +51,7 @@ struct arch_options_t {
>  	uint8_t  	console_serial;
>  	enum coretype	core_header_type;
>  };
> +extern struct arch_options_t arch_options;
>  
>  int multiboot_x86_probe(const char *buf, off_t len);
>  int multiboot_x86_load(int argc, char **argv, const char *buf, off_t len,
> Index: kexec-tools/kexec/arch/x86_64/kexec-x86_64.c
> ===================================================================
> --- kexec-tools.orig/kexec/arch/x86_64/kexec-x86_64.c
> +++ kexec-tools/kexec/arch/x86_64/kexec-x86_64.c
> @@ -55,14 +55,7 @@ void arch_usage(void)
>  		);
>  }
>  
> -static struct {
> -	uint8_t  reset_vga;
> -	uint16_t serial_base;
> -	uint32_t serial_baud;
> -	uint8_t  console_vga;
> -	uint8_t  console_serial;
> -	int core_header_type;
> -} arch_options = {
> +struct arch_options_t arch_options = {
>  	.reset_vga   = 0,
>  	.serial_base = 0x3f8,
>  	.serial_baud = 0,

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [PATCH] x86,setup: add serial_console_port_base in boot_params

[Yinghai Lu]

On 07/31/2010 07:42 PM, Eric W. Biederman wrote:
> Yinghai Lu <yinghai@kernel.org> writes:
> 
>> On 07/31/2010 02:02 PM, H. Peter Anvin wrote:
>>> On 07/31/2010 11:32 AM, Cyrill Gorcunov wrote:
>>>>>
>>>>> No, this is the internal part of the boot protocol, so it's not an issue.
>>>>>
>>>>
>>>> Peter, I didn't mean any issue here, I meant that bootloaders don't know about
>>>> this field yet and they will have to update own sources to pass port value
>>>> at proper place of boot params. Or I miss something?
>>>>
>>>
>>> Boot loaders that use the 16-bit entry point are unaffected.
>>>
>>> Boot loaders which use the 32-bit entry point but properly clears the
>>> zero page simply will not have the feature.
>>>
>>> Boot loaders which use the 32-bit entry point but doesn't clear the zero
>>> page are broken.
>>>
>> can you if this one is right for kexec path?
> 
> I am walking out the door, but this seems like nonsense to me.
> 
> Further I don't see why we would add something to the zero page
> when we have a perfectly good way to pass this information via
> the kernel command line.  strstr and strtoul are trivial little
> functions so I don't see why anything would need to parse anything
> other than console= or early_printk=.  The difference in code size
> is negligible.
> 
so you prefer to check command line for console info in arch/x86/boot/compressed/misc.c again?

that commandline is analyzed in arch/x86/boot/tty.c already.

Yinghai

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec

Re: [PATCH] x86,setup: add serial_console_port_base in boot_params

[Eric W. Biederman]

Yinghai Lu <yinghai@kernel.org> writes:

> On 07/31/2010 07:42 PM, Eric W. Biederman wrote:
>> Yinghai Lu <yinghai@kernel.org> writes:
>> 
>>> On 07/31/2010 02:02 PM, H. Peter Anvin wrote:
>>>> On 07/31/2010 11:32 AM, Cyrill Gorcunov wrote:
>>>>>>
>>>>>> No, this is the internal part of the boot protocol, so it's not an issue.
>>>>>>
>>>>>
>>>>> Peter, I didn't mean any issue here, I meant that bootloaders don't know about
>>>>> this field yet and they will have to update own sources to pass port value
>>>>> at proper place of boot params. Or I miss something?
>>>>>
>>>>
>>>> Boot loaders that use the 16-bit entry point are unaffected.
>>>>
>>>> Boot loaders which use the 32-bit entry point but properly clears the
>>>> zero page simply will not have the feature.
>>>>
>>>> Boot loaders which use the 32-bit entry point but doesn't clear the zero
>>>> page are broken.
>>>>
>>> can you if this one is right for kexec path?
>> 
>> I am walking out the door, but this seems like nonsense to me.
>> 
>> Further I don't see why we would add something to the zero page
>> when we have a perfectly good way to pass this information via
>> the kernel command line.  strstr and strtoul are trivial little
>> functions so I don't see why anything would need to parse anything
>> other than console= or early_printk=.  The difference in code size
>> is negligible.
>> 
> so you prefer to check command line for console info in arch/x86/boot/compressed/misc.c again?
>
> that commandline is analyzed in arch/x86/boot/tty.c already.

Instead of changing 2 or 3 bootloaders, waiting years for the change
to propagate, and then trying to change users habits.

I definitely do.

Eric

_______________________________________________
kexec mailing list
kexec@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/kexec