[PATCH] Allow initrd_load() before software_resume()

[PATCH] Allow initrd_load() before software_resume()

[Pascal Brisset]

[-- Attachment #1: message body text --]
[-- Type: text/plain, Size: 770 bytes --]

This patch adds a boot parameter "resume_initrd".
If present, init will load the initrd before trying to resume.

This makes it posssible to resume from an encrypted suspend image.
The initrd should insmod cryptoloop.o or loop-AES.o and perform
losetup -e so that resume=/dev/loopX makes sense.
Note: software_resume() should not be allowed to complete if
initrd has altered disks (e.g. by flushing journals).

/initrd
|-- bin
|   |-- ash
|   |-- insmod
|   `-- losetup
|-- dev
|   |-- console
|   |-- hdaX
|   |-- loopX
|   |-- null
|   `-- tty
|-- linuxrc
|-- loop.o
`-- lost+found

Resuming works, but suspension seems to fail more frequently when
the swap is encrypted. I am using loop-AES-v1.7d + patch for 2.6.

Is it safe to suspend to loop devices ?

-- Pascal


[-- Attachment #2: resume_initrd.diff --]
[-- Type: text/plain, Size: 1814 bytes --]

diff -ur linux-2.6.0-test1.orig/Documentation/kernel-parameters.txt linux-2.6.0-test1/Documentation/kernel-parameters.txt
--- linux-2.6.0-test1.orig/Documentation/kernel-parameters.txt	2003-07-14 05:39:36.000000000 +0200
+++ linux-2.6.0-test1/Documentation/kernel-parameters.txt	2003-08-01 01:19:46.000000000 +0200
@@ -816,6 +816,8 @@
 
 	resume=		[SWSUSP] Specify the partition device for software suspension
 
+	resume_initrd   [SWSUSP] Run initrd before resuming from software suspension
+
 	riscom8=	[HW,SERIAL]
 			Format: <io_board1>[,<io_board2>[,...<io_boardN>]]
 
diff -ur linux-2.6.0-test1.orig/init/do_mounts.c linux-2.6.0-test1/init/do_mounts.c
--- linux-2.6.0-test1.orig/init/do_mounts.c	2003-07-14 05:32:44.000000000 +0200
+++ linux-2.6.0-test1/init/do_mounts.c	2003-08-01 01:21:44.000000000 +0200
@@ -49,6 +49,15 @@
 __setup("ro", readonly);
 __setup("rw", readwrite);
 
+static int resume_initrd = 0;
+static int __init set_resume_initrd(char *str)
+{
+	resume_initrd = 1;
+	return 1;
+}
+
+__setup("resume_initrd", set_resume_initrd);
+
 static dev_t __init try_name(char *name, int part)
 {
 	char path[64];
@@ -365,12 +374,21 @@
 
 	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 
-	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
-	   log corrupting stuff */
-	software_resume();
+	/* software_resume() has to be before mounting root, because even
+	   readonly mount of reiserfs would replay log corrupting stuff.
+	   However, users may still want to run initrd first. */
+	if (resume_initrd) {
+		if (initrd_load()) {
+			software_resume();
+			goto out;
+		}
+	}
+	else {
+		software_resume();
 
-	if (initrd_load())
-		goto out;
+		if (initrd_load())
+			goto out;
+	}
 
 	if (is_floppy && rd_doload && rd_load_disk(0))
 		ROOT_DEV = Root_RAM0;

Re: [PATCH] Allow initrd_load() before software_resume()

[Nigel Cunningham]

Hi.

I'll try to apply it to the 2.4 version and include it in the upcoming
2.6 version. (Yes, upcoming means it's finally on the way).

Regards,

Nigel

On Fri, 2003-08-01 at 12:29, Pascal Brisset wrote:
> This patch adds a boot parameter "resume_initrd".
> If present, init will load the initrd before trying to resume.
> 
> This makes it posssible to resume from an encrypted suspend image.
> The initrd should insmod cryptoloop.o or loop-AES.o and perform
> losetup -e so that resume=/dev/loopX makes sense.
> Note: software_resume() should not be allowed to complete if
> initrd has altered disks (e.g. by flushing journals).
> 
> /initrd
> |-- bin
> |   |-- ash
> |   |-- insmod
> |   `-- losetup
> |-- dev
> |   |-- console
> |   |-- hdaX
> |   |-- loopX
> |   |-- null
> |   `-- tty
> |-- linuxrc
> |-- loop.o
> `-- lost+found
> 
> Resuming works, but suspension seems to fail more frequently when
> the swap is encrypted. I am using loop-AES-v1.7d + patch for 2.6.
> 
> Is it safe to suspend to loop devices ?
> 
> -- Pascal
> 
> 
> ______________________________________________________________________
> 
> diff -ur linux-2.6.0-test1.orig/Documentation/kernel-parameters.txt linux-2.6.0-test1/Documentation/kernel-parameters.txt
> --- linux-2.6.0-test1.orig/Documentation/kernel-parameters.txt	2003-07-14 05:39:36.000000000 +0200
> +++ linux-2.6.0-test1/Documentation/kernel-parameters.txt	2003-08-01 01:19:46.000000000 +0200
> @@ -816,6 +816,8 @@
>  
>  	resume=		[SWSUSP] Specify the partition device for software suspension
>  
> +	resume_initrd   [SWSUSP] Run initrd before resuming from software suspension
> +
>  	riscom8=	[HW,SERIAL]
>  			Format: <io_board1>[,<io_board2>[,...<io_boardN>]]
>  
> diff -ur linux-2.6.0-test1.orig/init/do_mounts.c linux-2.6.0-test1/init/do_mounts.c
> --- linux-2.6.0-test1.orig/init/do_mounts.c	2003-07-14 05:32:44.000000000 +0200
> +++ linux-2.6.0-test1/init/do_mounts.c	2003-08-01 01:21:44.000000000 +0200
> @@ -49,6 +49,15 @@
>  __setup("ro", readonly);
>  __setup("rw", readwrite);
>  
> +static int resume_initrd = 0;
> +static int __init set_resume_initrd(char *str)
> +{
> +	resume_initrd = 1;
> +	return 1;
> +}
> +
> +__setup("resume_initrd", set_resume_initrd);
> +
>  static dev_t __init try_name(char *name, int part)
>  {
>  	char path[64];
> @@ -365,12 +374,21 @@
>  
>  	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
>  
> -	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
> -	   log corrupting stuff */
> -	software_resume();
> +	/* software_resume() has to be before mounting root, because even
> +	   readonly mount of reiserfs would replay log corrupting stuff.
> +	   However, users may still want to run initrd first. */
> +	if (resume_initrd) {
> +		if (initrd_load()) {
> +			software_resume();
> +			goto out;
> +		}
> +	}
> +	else {
> +		software_resume();
>  
> -	if (initrd_load())
> -		goto out;
> +		if (initrd_load())
> +			goto out;
> +	}
>  
>  	if (is_floppy && rd_doload && rd_load_disk(0))
>  		ROOT_DEV = Root_RAM0;
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Pascal Brisset]

[-- Attachment #1: message body text --]
[-- Type: text/plain, Size: 563 bytes --]

[Sorry, new version already... I hadn't noticed that mount_root() is
 also called from within initrd_load(). resume() must run before.]


This patch adds a boot parameter "resume_initrd".  If present,
init will load the initrd before trying to resume from swsusp.

This makes it posssible to resume from an encrypted suspend image.
The initrd should insmod cryptoloop.o or loop-AES.o and perform
losetup -e so that resume=/dev/loopX makes sense.
Note: software_resume() should not be allowed to complete if
initrd has altered disks (e.g. by flushing journals).



[-- Attachment #2: resume_initrd-2.diff --]
[-- Type: text/plain, Size: 2909 bytes --]

diff -ur linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt linux-2.6.0-test2/Documentation/kernel-parameters.txt
--- linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt	2003-07-27 19:12:45.000000000 +0200
+++ linux-2.6.0-test2/Documentation/kernel-parameters.txt	2003-08-01 11:28:10.000000000 +0200
@@ -816,6 +816,8 @@
 
 	resume=		[SWSUSP] Specify the partition device for software suspension
 
+	resume_initrd	[SWSUSP] Run initrd before resuming from software suspension
+
 	riscom8=	[HW,SERIAL]
 			Format: <io_board1>[,<io_board2>[,...<io_boardN>]]
 
diff -ur linux-2.6.0-test2.orig/init/do_mounts.c linux-2.6.0-test2/init/do_mounts.c
--- linux-2.6.0-test2.orig/init/do_mounts.c	2003-07-27 19:00:37.000000000 +0200
+++ linux-2.6.0-test2/init/do_mounts.c	2003-08-01 11:31:17.000000000 +0200
@@ -15,6 +15,7 @@
 extern int get_filesystem_list(char * buf);
 
 int __initdata rd_doload;	/* 1 = load RAM disk, 0 = don't load */
+unsigned char resume_initrd = 0; /* Run initrd before resuming from swsusp */
 
 int root_mountflags = MS_RDONLY | MS_VERBOSE;
 char * __initdata root_device_name;
@@ -49,6 +50,13 @@
 __setup("ro", readonly);
 __setup("rw", readwrite);
 
+static int __init set_resume_initrd(char *str)
+{
+	resume_initrd = 1;
+	return 1;
+}
+__setup("resume_initrd", set_resume_initrd);
+
 static dev_t __init try_name(char *name, int part)
 {
 	char path[64];
@@ -365,9 +373,11 @@
 
 	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
 
-	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
-	   log corrupting stuff */
-	software_resume();
+	/* software_resume() has to be before mounting root, because even
+	   readonly mount of reiserfs would replay log corrupting stuff.
+	   However, users may want to run a special initrd first. */
+	if (!resume_initrd)
+		software_resume();
 
 	if (initrd_load())
 		goto out;
diff -ur linux-2.6.0-test2.orig/init/do_mounts.h linux-2.6.0-test2/init/do_mounts.h
--- linux-2.6.0-test2.orig/init/do_mounts.h	2003-07-27 19:04:19.000000000 +0200
+++ linux-2.6.0-test2/init/do_mounts.h	2003-08-01 11:31:41.000000000 +0200
@@ -28,6 +28,7 @@
 void  mount_root(void);
 extern int root_mountflags;
 extern char *root_device_name;
+extern unsigned char resume_initrd;
 
 #ifdef CONFIG_DEVFS_FS
 
diff -ur linux-2.6.0-test2.orig/init/do_mounts_initrd.c linux-2.6.0-test2/init/do_mounts_initrd.c
--- linux-2.6.0-test2.orig/init/do_mounts_initrd.c	2003-07-27 18:57:13.000000000 +0200
+++ linux-2.6.0-test2/init/do_mounts_initrd.c	2003-08-01 11:33:36.000000000 +0200
@@ -6,6 +6,7 @@
 #include <linux/romfs_fs.h>
 #include <linux/initrd.h>
 #include <linux/sched.h>
+#include <linux/suspend.h>
 
 #include "do_mounts.h"
 
@@ -74,6 +75,10 @@
 		return;
 	}
 
+	/* Must resume from swsusp before mounting a journalling root fs */
+	if (resume_initrd)
+		software_resume();
+
 	ROOT_DEV = real_root_dev;
 	mount_root();
 

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Nigel Cunningham]

Okay. I hadn't tried it yet. I'll happily take up the barrow for you and
push it to Pavel and Linus with the rest, if you like.

Regards,

Nigel

On Fri, 2003-08-01 at 22:32, Pascal Brisset wrote:
> [Sorry, new version already... I hadn't noticed that mount_root() is
>  also called from within initrd_load(). resume() must run before.]
> 
> 
> This patch adds a boot parameter "resume_initrd".  If present,
> init will load the initrd before trying to resume from swsusp.
> 
> This makes it posssible to resume from an encrypted suspend image.
> The initrd should insmod cryptoloop.o or loop-AES.o and perform
> losetup -e so that resume=/dev/loopX makes sense.
> Note: software_resume() should not be allowed to complete if
> initrd has altered disks (e.g. by flushing journals).
> 
> 
> 
> ______________________________________________________________________
> 
> diff -ur linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt linux-2.6.0-test2/Documentation/kernel-parameters.txt
> --- linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt	2003-07-27 19:12:45.000000000 +0200
> +++ linux-2.6.0-test2/Documentation/kernel-parameters.txt	2003-08-01 11:28:10.000000000 +0200
> @@ -816,6 +816,8 @@
>  
>  	resume=		[SWSUSP] Specify the partition device for software suspension
>  
> +	resume_initrd	[SWSUSP] Run initrd before resuming from software suspension
> +
>  	riscom8=	[HW,SERIAL]
>  			Format: <io_board1>[,<io_board2>[,...<io_boardN>]]
>  
> diff -ur linux-2.6.0-test2.orig/init/do_mounts.c linux-2.6.0-test2/init/do_mounts.c
> --- linux-2.6.0-test2.orig/init/do_mounts.c	2003-07-27 19:00:37.000000000 +0200
> +++ linux-2.6.0-test2/init/do_mounts.c	2003-08-01 11:31:17.000000000 +0200
> @@ -15,6 +15,7 @@
>  extern int get_filesystem_list(char * buf);
>  
>  int __initdata rd_doload;	/* 1 = load RAM disk, 0 = don't load */
> +unsigned char resume_initrd = 0; /* Run initrd before resuming from swsusp */
>  
>  int root_mountflags = MS_RDONLY | MS_VERBOSE;
>  char * __initdata root_device_name;
> @@ -49,6 +50,13 @@
>  __setup("ro", readonly);
>  __setup("rw", readwrite);
>  
> +static int __init set_resume_initrd(char *str)
> +{
> +	resume_initrd = 1;
> +	return 1;
> +}
> +__setup("resume_initrd", set_resume_initrd);
> +
>  static dev_t __init try_name(char *name, int part)
>  {
>  	char path[64];
> @@ -365,9 +373,11 @@
>  
>  	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
>  
> -	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
> -	   log corrupting stuff */
> -	software_resume();
> +	/* software_resume() has to be before mounting root, because even
> +	   readonly mount of reiserfs would replay log corrupting stuff.
> +	   However, users may want to run a special initrd first. */
> +	if (!resume_initrd)
> +		software_resume();
>  
>  	if (initrd_load())
>  		goto out;
> diff -ur linux-2.6.0-test2.orig/init/do_mounts.h linux-2.6.0-test2/init/do_mounts.h
> --- linux-2.6.0-test2.orig/init/do_mounts.h	2003-07-27 19:04:19.000000000 +0200
> +++ linux-2.6.0-test2/init/do_mounts.h	2003-08-01 11:31:41.000000000 +0200
> @@ -28,6 +28,7 @@
>  void  mount_root(void);
>  extern int root_mountflags;
>  extern char *root_device_name;
> +extern unsigned char resume_initrd;
>  
>  #ifdef CONFIG_DEVFS_FS
>  
> diff -ur linux-2.6.0-test2.orig/init/do_mounts_initrd.c linux-2.6.0-test2/init/do_mounts_initrd.c
> --- linux-2.6.0-test2.orig/init/do_mounts_initrd.c	2003-07-27 18:57:13.000000000 +0200
> +++ linux-2.6.0-test2/init/do_mounts_initrd.c	2003-08-01 11:33:36.000000000 +0200
> @@ -6,6 +6,7 @@
>  #include <linux/romfs_fs.h>
>  #include <linux/initrd.h>
>  #include <linux/sched.h>
> +#include <linux/suspend.h>
>  
>  #include "do_mounts.h"
>  
> @@ -74,6 +75,10 @@
>  		return;
>  	}
>  
> +	/* Must resume from swsusp before mounting a journalling root fs */
> +	if (resume_initrd)
> +		software_resume();
> +
>  	ROOT_DEV = real_root_dev;
>  	mount_root();
>  
-- 
Nigel Cunningham
495 St Georges Road South, Hastings 4201, New Zealand

You see, at just the right time, when we were still powerless,
Christ died for the ungodly.
	-- Romans 5:6, NIV.

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Pavel Machek]

Hi!

> Okay. I hadn't tried it yet. I'll happily take up the barrow for you and
> push it to Pavel and Linus with the rest, if you like.

Don't even think about that.

It is not safe to run userspace *before* doing resume. You don't want
to see problems this would bring in. Forget it.
								Pavel

> > [Sorry, new version already... I hadn't noticed that mount_root() is
> >  also called from within initrd_load(). resume() must run before.]
> > 
> > 
> > This patch adds a boot parameter "resume_initrd".  If present,
> > init will load the initrd before trying to resume from swsusp.
> > 
> > This makes it posssible to resume from an encrypted suspend image.
> > The initrd should insmod cryptoloop.o or loop-AES.o and perform
> > losetup -e so that resume=/dev/loopX makes sense.
> > Note: software_resume() should not be allowed to complete if
> > initrd has altered disks (e.g. by flushing journals).
> > 
> > 
> > 
> > ______________________________________________________________________
> > 
> > diff -ur linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt linux-2.6.0-test2/Documentation/kernel-parameters.txt
> > --- linux-2.6.0-test2.orig/Documentation/kernel-parameters.txt	2003-07-27 19:12:45.000000000 +0200
> > +++ linux-2.6.0-test2/Documentation/kernel-parameters.txt	2003-08-01 11:28:10.000000000 +0200
> > @@ -816,6 +816,8 @@
> >  
> >  	resume=		[SWSUSP] Specify the partition device for software suspension
> >  
> > +	resume_initrd	[SWSUSP] Run initrd before resuming from software suspension
> > +
> >  	riscom8=	[HW,SERIAL]
> >  			Format: <io_board1>[,<io_board2>[,...<io_boardN>]]
> >  
> > diff -ur linux-2.6.0-test2.orig/init/do_mounts.c linux-2.6.0-test2/init/do_mounts.c
> > --- linux-2.6.0-test2.orig/init/do_mounts.c	2003-07-27 19:00:37.000000000 +0200
> > +++ linux-2.6.0-test2/init/do_mounts.c	2003-08-01 11:31:17.000000000 +0200
> > @@ -15,6 +15,7 @@
> >  extern int get_filesystem_list(char * buf);
> >  
> >  int __initdata rd_doload;	/* 1 = load RAM disk, 0 = don't load */
> > +unsigned char resume_initrd = 0; /* Run initrd before resuming from swsusp */
> >  
> >  int root_mountflags = MS_RDONLY | MS_VERBOSE;
> >  char * __initdata root_device_name;
> > @@ -49,6 +50,13 @@
> >  __setup("ro", readonly);
> >  __setup("rw", readwrite);
> >  
> > +static int __init set_resume_initrd(char *str)
> > +{
> > +	resume_initrd = 1;
> > +	return 1;
> > +}
> > +__setup("resume_initrd", set_resume_initrd);
> > +
> >  static dev_t __init try_name(char *name, int part)
> >  {
> >  	char path[64];
> > @@ -365,9 +373,11 @@
> >  
> >  	is_floppy = MAJOR(ROOT_DEV) == FLOPPY_MAJOR;
> >  
> > -	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
> > -	   log corrupting stuff */
> > -	software_resume();
> > +	/* software_resume() has to be before mounting root, because even
> > +	   readonly mount of reiserfs would replay log corrupting stuff.
> > +	   However, users may want to run a special initrd first. */
> > +	if (!resume_initrd)
> > +		software_resume();
> >  
> >  	if (initrd_load())
> >  		goto out;
> > diff -ur linux-2.6.0-test2.orig/init/do_mounts.h linux-2.6.0-test2/init/do_mounts.h
> > --- linux-2.6.0-test2.orig/init/do_mounts.h	2003-07-27 19:04:19.000000000 +0200
> > +++ linux-2.6.0-test2/init/do_mounts.h	2003-08-01 11:31:41.000000000 +0200
> > @@ -28,6 +28,7 @@
> >  void  mount_root(void);
> >  extern int root_mountflags;
> >  extern char *root_device_name;
> > +extern unsigned char resume_initrd;
> >  
> >  #ifdef CONFIG_DEVFS_FS
> >  
> > diff -ur linux-2.6.0-test2.orig/init/do_mounts_initrd.c linux-2.6.0-test2/init/do_mounts_initrd.c
> > --- linux-2.6.0-test2.orig/init/do_mounts_initrd.c	2003-07-27 18:57:13.000000000 +0200
> > +++ linux-2.6.0-test2/init/do_mounts_initrd.c	2003-08-01 11:33:36.000000000 +0200
> > @@ -6,6 +6,7 @@
> >  #include <linux/romfs_fs.h>
> >  #include <linux/initrd.h>
> >  #include <linux/sched.h>
> > +#include <linux/suspend.h>
> >  
> >  #include "do_mounts.h"
> >  
> > @@ -74,6 +75,10 @@
> >  		return;
> >  	}
> >  
> > +	/* Must resume from swsusp before mounting a journalling root fs */
> > +	if (resume_initrd)
> > +		software_resume();
> > +
> >  	ROOT_DEV = real_root_dev;
> >  	mount_root();
> >  

-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 494 bytes --]

On Wed, 2003-08-06 at 13:30, Pavel Machek wrote:
> Hi!
> 
> > Okay. I hadn't tried it yet. I'll happily take up the barrow for you and
> > push it to Pavel and Linus with the rest, if you like.
> 
> Don't even think about that.
> 
> It is not safe to run userspace *before* doing resume. You don't want
> to see problems this would bring in. Forget it.
> 			
so how do you resume from a partition on a device mapper volume?

(and yes I basically agree with your sentiment though)

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Pavel Machek]

Hi!

> > > Okay. I hadn't tried it yet. I'll happily take up the barrow for you and
> > > push it to Pavel and Linus with the rest, if you like.
> > 
> > Don't even think about that.
> > 
> > It is not safe to run userspace *before* doing resume. You don't want
> > to see problems this would bring in. Forget it.
> > 			
> so how do you resume from a partition on a device mapper volume?
> 
> (and yes I basically agree with your sentiment though)

I know very little about DM, its very well possible that resume from it is not supported.

-- 
				Pavel
Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Måns Rullgård]

Pavel Machek <pavel@ucw.cz> writes:

>> > > Okay. I hadn't tried it yet. I'll happily take up the barrow for you and
>> > > push it to Pavel and Linus with the rest, if you like.
>> > 
>> > Don't even think about that.
>> > 
>> > It is not safe to run userspace *before* doing resume. You don't want
>> > to see problems this would bring in. Forget it.
>> > 			
>> so how do you resume from a partition on a device mapper volume?
>> 
>> (and yes I basically agree with your sentiment though)
>
> I know very little about DM, its very well possible that resume from
> it is not supported.

Since DM requires some userspace program to set up the mappings, it
seems to me that it wouldn't work to resume from a DM volume.  I'd
much appreciate if it would work, somehow.

-- 
Måns Rullgård
mru@users.sf.net

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Rob Landley]

On Wednesday 06 August 2003 09:16, Måns Rullgård wrote:
> Pavel Machek <pavel@ucw.cz> writes:
> >> > > Okay. I hadn't tried it yet. I'll happily take up the barrow for you
> >> > > and push it to Pavel and Linus with the rest, if you like.
> >> >
> >> > Don't even think about that.
> >> >
> >> > It is not safe to run userspace *before* doing resume. You don't want
> >> > to see problems this would bring in. Forget it.
> >>
> >> so how do you resume from a partition on a device mapper volume?
> >>
> >> (and yes I basically agree with your sentiment though)
> >
> > I know very little about DM, its very well possible that resume from
> > it is not supported.
>
> Since DM requires some userspace program to set up the mappings, it
> seems to me that it wouldn't work to resume from a DM volume.  I'd
> much appreciate if it would work, somehow.

Er, query:

At some point in the vague nebulous future, after initrd has become initramfs, 
the partition detection code is scheduled to be ripped out, correct?  And 
replaced with a userspace thing run out of initrd ala hotplug and udev and 
all that?

So at that point, it's not just device mapper that's going to need something 
else to run in userspace to attach block devices to partitions.  Everything 
will.

So are you saying that swsusp is a short-term thing that will be dropped in 
2.8 because it can't be made to work?  Or that we WILL have to deal with this 
at some point, just not yet?

Rob

Re: [Swsusp-devel] Re: [PATCH] Allow initrd_load() before software_resume() (version 2)

[Pavel Machek]

Hi!

> > >> > > Okay. I hadn't tried it yet. I'll happily take up the barrow for you
> > >> > > and push it to Pavel and Linus with the rest, if you like.
> > >> >
> > >> > Don't even think about that.
> > >> >
> > >> > It is not safe to run userspace *before* doing resume. You don't want
> > >> > to see problems this would bring in. Forget it.
> > >>
> > >> so how do you resume from a partition on a device mapper volume?
> > >>
> > >> (and yes I basically agree with your sentiment though)
> > >
> > > I know very little about DM, its very well possible that resume from
> > > it is not supported.
> >
> > Since DM requires some userspace program to set up the mappings, it
> > seems to me that it wouldn't work to resume from a DM volume.  I'd
> > much appreciate if it would work, somehow.
> 
> Er, query:
> 
> At some point in the vague nebulous future, after initrd has become initramfs, 
> the partition detection code is scheduled to be ripped out, correct?  And 
> replaced with a userspace thing run out of initrd ala hotplug and udev and 
> all that?
> 
> So at that point, it's not just device mapper that's going to need something 
> else to run in userspace to attach block devices to partitions.  Everything 
> will.
> 
> So are you saying that swsusp is a short-term thing that will be dropped in 
> 2.8 because it can't be made to work?  Or that we WILL have to deal with this 
> at some point, just not yet?

At that point we'll have to carefully audit all userland code that can
run before software resume.
									Pavel
-- 
When do you have a heart between your knees?
[Johanka's followup: and *two* hearts?]