[PATCH] suspend-to-{RAM,disk}

Here's suspend-to-{RAM,disk} combined patch for
2.5.17. Suspend-to-disk is pretty stable and was tested in
2.4-ac. Suspend-to-RAM is little more experimental, but works for me,
and is certainly better than disk-eating version currently in kernel.

Major parts are: process stopper, S3 specific code, S4 specific
code.

Documentation/driver-model.txt

@@ -52,7 +52,8 @@ appropriate callbacks that each device (or bridge) of that type share.
 Each bus layer should implement the callbacks for these drivers. It then
 forwards the calls on to the device-specific callbacks. This means that
 device-specific drivers must still implement callbacks for each operation.
-But, they are not called from the top level driver layer.
+But, they are not called from the top level driver layer. [So for example
+PCI devices will not call device_register but pci_device_register.]
 
 This does add another layer of indirection for calling one of these functions,
 but there are benefits that are believed to outweigh this slowdown.
@@ -225,7 +226,6 @@ platform_data:
 	It also allows the platform driver (e.g. ACPI) to a driver without the driver
 	having to have explicit knowledge of (atrocities like) ACPI.
 
-
 current_state:
 	Current power state of the device. For PCI and other modern devices, this is
 	0-3, though it's not necessarily limited to those values.
@@ -251,18 +251,24 @@ struct device_driver {
 }
 
 probe:
-	Check for device existence and associate driver with it.
+	Check for device existence and associate driver with it. In case of device 
+	insertion, *all* drivers are called. Struct device has parent and bus_id 
+	valid at this point. probe() may only be called from process context. Returns
+	0 if it handles that device, -ESRCH if this driver does not know how to handle
+	this device, valid error otherwise.
 
 remove:
 	Dissociate driver with device. Releases device so that it could be used by
 	another driver. Also, if it is a hotplug device (hotplug PCI, Cardbus), an
-	ejection event could take place here.
+	ejection event could take place here. remove() can be called from interrupt 
+	context. [Fixme: Is that good?] Returns 0 on success. [Can we recover from
+	failed remove or should I define that remove() never fails?]
 
 suspend:
-	Perform one step of the device suspend process.
+	Perform one step of the device suspend process. Returns 0 on success.
 
 resume:
-	Perform one step of the device resume process.
+	Perform one step of the device resume process. Returns 0 on success.
 
 The probe() and remove() callbacks are intended to be much simpler than the
 current PCI correspondents.
@@ -275,7 +281,7 @@ probe() should do the following only:
 
 Some device initialisation was done in probe(). This should not be the case
 anymore. All initialisation should take place in the open() call for the
-device.
+device. [FIXME: How do you "open" uhci?]
 
 Breaking initialisation code out must also be done for the resume() callback,
 as most devices will have to be completely reinitialised when coming back from
@@ -324,6 +330,7 @@ the stage:
 
 enum{
 	SUSPEND_NOTIFY,
+	SUSPEND_DISABLE,
 	SUSPEND_SAVE_STATE,
 	SUSPEND_POWER_DOWN,
 };
@@ -331,6 +338,7 @@ enum{
 enum {
 	RESUME_POWER_ON,
 	RESUME_RESTORE_STATE,
+	RESUME_ENABLE,
 };
 
 
@@ -352,9 +360,9 @@ shutting down the device(s) you want to save state to.
 Instead, the walking of the device tree has been moved to userspace. When a
 user requests the system to suspend, it will walk the device tree, as exported
 via driverfs, and tell each device to go to sleep. It will do this multiple
-times based on what the system policy is.
-
-[ FIXME: URL pointer to the corresponding utility is missing here! ]
+times based on what the system policy is. [Not possible. Take ACPI enabled 
+system, with battery critically low. In such state, you want to suspend-to-disk,
+*fast*. User maybe is not even running powerd (think system startup)!]
 
 Device resume should happen in the same manner when the system awakens.
 
@@ -366,22 +374,25 @@ This level to notify the driver that it is going to sleep. If it knows that it
 cannot resume the hardware from the requested level, or it feels that it is
 too important to be put to sleep, it should return an error from this function.
 
-It does not have to stop I/O requests or actually save state at this point.
+It does not have to stop I/O requests or actually save state at this point. Called
+from process context.
 
 SUSPEND_DISABLE:
 
 The driver should stop taking I/O requests at this stage. Because the save
 state stage happens afterwards, the driver may not want to physically disable
-the device; only mark itself unavailable if possible.
+the device; only mark itself unavailable if possible. Called from process 
+context.
 
 SUSPEND_SAVE_STATE:
 
 The driver should allocate memory and save any device state that is relevant
-for the state it is going to enter.
+for the state it is going to enter. Called from process context.
 
 SUSPEND_POWER_DOWN:
 
-The driver should place the device in the power state requested.
+The driver should place the device in the power state requested. May be called
+from interrupt context.
 
 
 For resume, the stages are defined as follows:
@@ -389,25 +400,27 @@ For resume, the stages are defined as follows:
 RESUME_POWER_ON:
 
 Devices should be powered on and reinitialised to some known working state.
+Called from process context.
 
 RESUME_RESTORE_STATE:
 
 The driver should restore device state to its pre-suspend state and free any
-memory allocated for its saved state.
+memory allocated for its saved state. Called from process context.
 
 RESUME_ENABLE:
 
-The device should start taking I/O requests again.
+The device should start taking I/O requests again. Called from process context.
 
 
 Each driver does not have to implement each stage. But, it if it does
-implemente a stage, it should do what is described above. It should not assume
+implement a stage, it should do what is described above. It should not assume
 that it performed any stage previously, or that it will perform any stage
-later.
+later. [Really? It makes sense to support SAVE_STATE only after DISABLE].
 
 It is quite possible that a driver can fail during the suspend process, for
 whatever reason. In this event, the calling process must gracefully recover
 and restore everything to their states before the suspend transition began. 
+[Suspend may not fail, think battery low.]
 
 If a driver knows that it cannot suspend or resume properly, it should fail
 during the notify stage. Properly implemented power management schemes should

Documentation/swsusp.txt

+From kernel/suspend.c:
+
+ * BIG FAT WARNING *********************************************************
+ *
+ * If you have unsupported (*) devices using DMA...
+ *				...say goodbye to your data.
+ *
+ * If you touch anything on disk between suspend and resume...
+ *				...kiss your data goodbye.
+ *
+ * If your disk driver does not support suspend... (IDE does)
+ *				...you'd better find out how to get along
+ *				   without your data.
+ *
+ * (*) pm interface support is needed to make it safe.
+
+You need to append resume=/dev/your_swap_partition to kernel command
+line. Then you suspend by echo 4 > /proc/acpi/sleep.
+
+[Notice. Rest docs is pretty outdated (see date!) It should be safe to
+use swsusp on ext3/reiserfs these days.]
+
+
+Article about goals and implementation of Software Suspend for Linux
+Author: G‚ábor Kuti
+Last revised: 2002-04-08
+
+Idea and goals to achieve
+
+Nowadays it is common in several laptops that they have a suspend button. It
+saves the state of the machine to a filesystem or to a partition and switches
+to standby mode. Later resuming the machine the saved state is loaded back to
+ram and the machine can continue its work. It has two real benefits. First we
+save ourselves the time machine goes down and later boots up, energy costs
+real high when running from batteries. The other gain is that we don't have to
+interrupt our programs so processes that are calculating something for a long
+time shouldn't need to be written interruptible.
+
+On desk machines the power saving function isn't as important as it is in
+laptops but we really may benefit from the second one. Nowadays the number of
+desk machines supporting suspend function in their APM is going up but there
+are (and there will still be for a long time) machines that don't even support
+APM of any kind. On the other hand it is reported that using APM's suspend
+some irqs (e.g. ATA disk irq) is lost and it is annoying for the user until
+the Linux kernel resets the device.
+
+So I started thinking about implementing Software Suspend which doesn't need
+any APM support and - since it uses pretty near only high-level routines - is
+supposed to be architecture independent code.
+
+Using the code
+
+The code is experimental right now - testers, extra eyes are welcome. To
+compile this support into the kernel, you need CONFIG_EXPERIMENTAL, 
+and then CONFIG_SOFTWARE_SUSPEND in menu General Setup to be  enabled. It
+cannot be used as a module and I don't think it will ever be needed.
+
+You have two ways to use this code. The first one is if you've compiled in
+sysrq support then you may press Sysrq-D to request suspend. The other way
+is with a patched SysVinit (my patch is against 2.76 and available at my
+home page). You might call 'swsusp' or 'shutdown -z <time>'. Next way is to
+echo 4 > /proc/acpi/sleep.
+
+Either way it saves the state of the machine into active swaps and then
+reboots. You must explicitly specify the swap partition to resume from with ``resume=''
+kernel option. If signature is found it loads and restores saved state. If the
+option ``noresume'' is specified as a boot parameter, it skips the resuming.
+Warning! Look at section ``Things to implement'' to see what isn't yet
+implemented.  Also I strongly suggest you to list all active swaps in
+/etc/fstab. Firstly because you don't have to specify anything to resume and
+secondly if you have more than one swap area you can't decide which one has the
+'root' signature. 
+
+In the meantime while the system is suspended you should not touch any of the
+hardware!
+
+About the code
+Goals reached
+
+The code can be downloaded from
+http://falcon.sch.bme.hu/~seasons/linux/. It mainly works but there are still
+some of XXXs, TODOs, FIXMEs in the code which seem not to be too important. It
+should work all right except for the problems listed in ``Things to
+implement''. Notes about the code are really welcome.
+
+How the code works
+
+When suspending is triggered it immediately wakes up process bdflush. Bdflush
+checks whether we have anything in our run queue tq_bdflush. Since we queued up
+function do_software_suspend, it is called. Here we shrink everything including
+dcache, inodes, buffers and memory (here mainly processes are swapped out). We
+count how many pages we need to duplicate (we have to be atomical!) then we
+create an appropiate sized page directory. It will point to the original and
+the new (copied) address of the page. We get the free pages by
+__get_free_pages() but since it changes state we have to be able to track it
+later so it also flips in a bit in page's flags (a new Nosave flag). We
+duplicate pages and then mark them as used (so atomicity is ensured). After
+this we write out the image to swaps, do another sync and the machine may
+reboot. We also save registers to stack.
+
+By resuming an ``inverse'' method is executed. The image if exists is loaded,
+loadling is either triggered by ``resume='' kernel option.  We
+change our task to bdflush (it is needed because if we don't do this init does
+an oops when it is waken up later) and then pages are copied back to their
+original location. We restore registers, free previously allocated memory,
+activate memory context and task information. Here we should restore hardware
+state but even without this the machine is restored and processes are continued
+to work. I think hardware state should be restored by some list (using
+notify_chain) and probably by some userland program (run-parts?) for users'
+pleasure. Check out my patch at the same location for the sysvinit patch.
+
+WARNINGS!
+- It does not like pcmcia cards. And this is logical: pcmcia cards need cardmgr to be
+  initialized. they are not initialized during singleuser boot, but "resumed" kernel does
+  expect them to be initialized. That leads to armagedon. You should eject any pcmcia cards
+  before suspending.
+
+Things to implement
+- SMP support. I've done an SMP support but since I don't have access to a kind
+  of this one I cannot test it. Please SMP people test it.  .. Tested it,
+  doesn't work. Had no time to figure out why. There is some mess with
+  interrupts AFAIK..
+- We should only make a copy of data related to kernel segment, since any
+  process data won't be changed.
+- By copying pages back to their original position, copy_page caused General
+  Protection Fault. Why?
+- Hardware state restoring.  Now there's support for notifying via the notify
+  chain, event handlers are welcome. Some devices may have microcodes loaded
+  into them. We should have event handlers for them aswell.
+- We should support other architectures (There are really only some arch
+  related functions..)
+- We should also restore original state of swaps if the ``noresume'' kernel
+  option is specified.. Or do we need such a feature to save state for some
+  other time? Do we need some kind of ``several saved states''?  (Linux-HA
+  people?). There's been some discussion about checkpointing on linux-future.
+- Should make more sanity checks. Or are these enough?
+
+Not so important ideas for implementing
+
+- If a real time process is running then don't suspend the machine.
+- Is there any sense in compressing the outwritten pages?
+- Support for power.conf file as in Solaris, autoshutdown, special
+  devicetypes support, maybe in sysctl.
+- Introduce timeout for SMP locking. But first locking ought to work :O
+- Pre-detect if we don't have enough swap space or free it instead of
+  calling panic.
+- Support for adding/removing hardware while suspended?
+- We should not free pages at the beginning so aggressively, most of them
+  go there anyway..
+- If X is active while suspending then by resuming calling svgatextmode
+  corrupts the virtual console of X.. (Maybe this has been fixed AFAIK).
+
+Any other idea you might have tell me!
+
+Contacting the author
+If you have any question or any patch that solves the above or detected
+problems please contact me at seasons@falcon.sch.bme.hu. I might delay
+answering, sorry about that.
+

MAINTAINERS

@@ -1446,6 +1446,14 @@ M:	neilb@cse.unsw.edu.au
 L:	linux-raid@vger.kernel.org
 S:	Maintained
 
+SOFTWARE SUSPEND:
+P:	Gabor Kuti
+M:	seasons@falcon.sch.bme.hu
+M:	seasons@makosteszta.sote.hu
+L:	http://lister.fornax.hu/mailman/listinfo/swsusp
+W:	http://falcon.sch.bme.hu/~seasons/linux
+S:	Maintained
+
 SONIC NETWORK DRIVER
 P:	Thomas Bogendoerfer
 M:	tsbogend@alpha.franken.de

arch/i386/Config.help

@@ -940,3 +940,28 @@ CONFIG_DEBUG_BUGVERBOSE
 CONFIG_DEBUG_OBSOLETE
   Say Y here if you want to reduce the chances of the tree compiling,
   and are prepared to dig into driver internals to fix compile errors.
+
+Software Suspend
+CONFIG_SOFTWARE_SUSPEND
+  Enable the possibilty of suspendig machine. It doesn't need APM.
+  You may suspend your machine by either pressing Sysrq-d or with
+  'swsusp' or 'shutdown -z <time>' (patch for sysvinit needed). It
+  creates an image which is saved in your active swaps. By the next
+  booting the kernel detects the saved image, restores the memory from
+  it and then it continues to run as before you've suspended.
+  If you don't want the previous state to continue use the 'noresume'
+  kernel option. However note that your partitions will be fsck'd and
+  you must re-mkswap your swap partitions/files.
+
+  Right now you may boot without resuming and then later resume but
+  in meantime you cannot use those swap partitions/files which were
+  involved in suspending. Also in this case there is a risk that buffers
+  on disk won't match with saved ones.
+
+  SMP is supported ``as-is''. There's a code for it but doesn't work.
+  There have been problems reported relating SCSI.
+  
+  This option is about getting stable. However there is still some
+  absence of features.
+
+  For more information take a look at Documentation/swsusp.txt.

arch/i386/boot/video.S

@@ -438,6 +438,7 @@ setalias:
 
 # Setting of user mode (AX=mode ID) => CF=success
 mode_set:
+	movw	%ax, %fs:(0x01fa)		# Store mode for use in acpi_wakeup.S
 	movw	%ax, %bx
 	cmpb	$0xff, %ah
 	jz	setalias

arch/i386/config.in

@@ -396,6 +396,9 @@ source net/bluetooth/Config.in
 
 mainmenu_option next_comment
 comment 'Kernel hacking'
+if [ "$CONFIG_EXPERIMENTAL" = "y" ]; then
+   dep_bool 'Software Suspend' CONFIG_SOFTWARE_SUSPEND $CONFIG_PM
+fi
 
 bool 'Kernel debugging' CONFIG_DEBUG_KERNEL
 if [ "$CONFIG_DEBUG_KERNEL" != "n" ]; then

arch/i386/defconfig

@@ -119,6 +119,7 @@ CONFIG_BINFMT_ELF=y
 CONFIG_BINFMT_MISC=y
 CONFIG_PM=y
 # CONFIG_APM is not set
+# CONFIG_SOFTWARE_SUSPEND is not set
 
 #
 # Memory Technology Devices (MTD)

arch/i386/kernel/acpi.c

@@ -44,6 +44,8 @@
 #include <asm/pgalloc.h>
 #include <asm/io_apic.h>
 #include <asm/tlbflush.h>
+#define ACPI_C
+#include <asm/suspend.h>
 
 
 #define PREFIX			"ACPI: "
@@ -623,5 +625,33 @@ void __init acpi_reserve_bootmem(void)
 	printk(KERN_DEBUG "ACPI: have wakeup address 0x%8.8lx\n", acpi_wakeup_address);
 }
 
+/*
+ * (KG): Since we affect stack here, we make this function as flat and easy
+ * as possible in order to not provoke gcc to use local variables on the stack.
+ * Note that on resume, all (expect nosave) variables will have the state from
+ * the time of writing (suspend_save_image) and the registers (including the
+ * stack pointer, but excluding the instruction pointer) will be loaded with 
+ * the values saved at save_processor_context() time.
+ */
+void do_suspend_magic(int resume)
+{
+	/* DANGER WILL ROBINSON!
+	 *
+	 * If this function is too difficult for gcc to optimize, it will crash and burn!
+	 * see above.
+	 *
+	 * DO NOT TOUCH.
+	 */
+	if (!resume) {
+		save_processor_context();
+		acpi_save_register_state((unsigned long)&&acpi_sleep_done);
+		acpi_enter_sleep_state(3);
+		return;
+	}
+acpi_sleep_done:
+	restore_processor_context();
+	printk("CPU context restored...\n");
+}
+
 #endif /*CONFIG_ACPI_SLEEP*/
 

arch/i386/kernel/acpi_wakeup.S

@@ -3,6 +3,7 @@
 #include <linux/linkage.h>
 #include <asm/segment.h>
 
+# Do we need to deal with A20?
 
 ALIGN
 wakeup_start:
@@ -10,9 +11,17 @@ wakeup_code:
 	wakeup_code_start = .
 	.code16
 
+ 	movw	$0xb800, %ax
+	movw	%ax,%fs
+	movw	$0x0e00 + 'L', %fs:(0x10)
 	cli
 	cld
 	  
+# setup video mode
+#	movw	$0x4117, %bx		# 0x4000 for linear framebuffer
+#	movw	$0x4f02, %ax
+#	int	$0x10
+
 	# setup data segment
 	movw	%cs, %ax
 
@@ -20,6 +29,14 @@ wakeup_code:
 	movw	%ax, %ds
 	movw	%ax, %ss
 	mov	$(wakeup_stack - wakeup_data), %sp		# Private stack is needed for ASUS board
+	movw	$0x0e00 + 'S', %fs:(0x12)	
+
+	movl	real_magic - wakeup_data, %eax
+	cmpl	$0x12345678, %eax
+	jne	bogus_real_magic
+
+	mov	video_mode - wakeup_data, %ax
+	call	mode_set
 
 	# set up page table
 	movl	(real_save_cr3 - wakeup_data), %eax
@@ -28,40 +45,114 @@ wakeup_code:
 	# make sure %cr4 is set correctly (features, etc)
 	movl	(real_save_cr4 - wakeup_data), %eax
 	movl	%eax, %cr4
+	movw	$0xb800, %ax
+	movw	%ax,%fs
+	movw	$0x0e00 + 'i', %fs:(0x12)
 
 	# need a gdt
 	lgdt	real_save_gdt - wakeup_data
 
-	movl	%cr0, %eax
-	orl     $0x80000001, %eax
+	movl	(real_save_cr0 - wakeup_data), %eax
 	movl	%eax, %cr0
+	movw	$0x0e00 + 'n', %fs:(0x14)
+
+	movl	real_magic - wakeup_data, %eax
+	cmpl	$0x12345678, %eax
+	jne	bogus_real_magic
 
 	ljmpl	$__KERNEL_CS,$wakeup_pmode_return
 
+bogus_real_magic:
+	movw	$0x0e00 + 'B', %fs:(0x12)
+	jmp bogus_real_magic
+
+/* This code uses an extended set of video mode numbers. These include:
+ * Aliases for standard modes
+ *	NORMAL_VGA (-1)
+ *	EXTENDED_VGA (-2)
+ *	ASK_VGA (-3)
+ * Video modes numbered by menu position -- NOT RECOMMENDED because of lack
+ * of compatibility when extending the table. These are between 0x00 and 0xff.
+ */
+#define VIDEO_FIRST_MENU 0x0000
+
+/* Standard BIOS video modes (BIOS number + 0x0100) */
+#define VIDEO_FIRST_BIOS 0x0100
+
+/* VESA BIOS video modes (VESA number + 0x0200) */
+#define VIDEO_FIRST_VESA 0x0200
+
+/* Video7 special modes (BIOS number + 0x0900) */
+#define VIDEO_FIRST_V7 0x0900
+
+# Setting of user mode (AX=mode ID) => CF=success
+mode_set:
+	movw	%ax, %bx
+#if 0
+	cmpb	$0xff, %ah
+	jz	setalias
+
+	testb	$VIDEO_RECALC>>8, %ah
+	jnz	_setrec
+
+	cmpb	$VIDEO_FIRST_RESOLUTION>>8, %ah
+	jnc	setres
+	
+	cmpb	$VIDEO_FIRST_SPECIAL>>8, %ah
+	jz	setspc
+
+	cmpb	$VIDEO_FIRST_V7>>8, %ah
+	jz	setv7
+#endif
+	
+	cmpb	$VIDEO_FIRST_VESA>>8, %ah
+	jnc	check_vesa
+#if 0	
+	orb	%ah, %ah
+	jz	setmenu
+#endif
+	
+	decb	%ah
+#	jz	setbios				  Add bios modes later
+
+setbad:	clc
+	ret
+
+check_vesa:
+	subb	$VIDEO_FIRST_VESA>>8, %bh
+	orw	$0x4000, %bx			# Use linear frame buffer
+	movw	$0x4f02, %ax			# VESA BIOS mode set call
+	int	$0x10
+	cmpw	$0x004f, %ax			# AL=4f if implemented
+	jnz	_setbad				# AH=0 if OK
+
+	stc
+	ret
+
+_setbad: jmp setbad
+
 	.code32
 	ALIGN
 
-.org	0x100
+.org	0x300
 wakeup_data:
 		.word 0
 real_save_gdt:	.word 0
 		.long 0
+real_save_cr0:	.long 0
 real_save_cr3:	.long 0
 real_save_cr4:	.long 0
+real_magic:	.long 0
+video_mode:	.long 0
 
-.org	0x300
+.org	0x500
 wakeup_stack:
 wakeup_end:
 
 wakeup_pmode_return:
-	# restore data segment
 	movl	$__KERNEL_DS, %eax
-	movw	%ax, %ds
-	movw	%ax, %es
-
-	# and restore the stack
-	movw	%ax, %ss
-	movl	saved_esp, %esp
+	movl	%eax, %ds
+	movw	$0x0e00 + 'u', %ds:(0xb8016)
 
 	# restore other segment registers
 	xorl	%eax, %eax
@@ -72,6 +163,30 @@ wakeup_pmode_return:
 	lgdt	saved_gdt
 	lidt	saved_idt
 	lldt	saved_ldt
+	ljmp	$(__KERNEL_CS),$1f
+1:
+	movl	%cr3, %eax
+	movl	%eax, %cr3
+	wbinvd
+
+	# and restore the stack ... but you need gdt for this to work
+	movl	$__KERNEL_DS, %eax
+	movw	%ax, %ss
+	movw	%ax, %ds
+	movw	%ax, %es
+	movw	%ax, %fs
+	movw	%ax, %gs
+	movl	saved_esp, %esp
+
+	movw	$0x0e00 + 'W', %ds:(0xb8018)
+	movl	$(1024*1024*3), %ecx
+	movl	$0, %esi
+	rep	lodsb
+	movw	$0x0e00 + 'O', %ds:(0xb8018)
+
+	movl	%cs:saved_magic2, %eax
+	cmpl	$0x12345678, %eax
+	jne	bogus_magic
 
 	# restore the other general registers
 	movl	saved_ebx, %ebx
@@ -81,8 +196,21 @@ wakeup_pmode_return:
 
 	# jump to place where we left off
 	movl	saved_eip,%eax
+	movw	$0x0e00 + 'x', %ds:(0xb8018)
+	pushl	%eax
+	popl	%eax
+	movw	$0x0e00 + '!', %ds:(0xb801a)
 	jmp	*%eax
 
+bogus_magic:
+	movw	$0x0e00 + 'B', %ds:(0xb8018)
+	jmp bogus_magic
+
+bogus_magic2:
+	movw	$0x0e00 + '2', %ds:(0xb8018)
+	jmp bogus_magic2
+		
+
 ##
 # acpi_copy_wakeup_routine
 #
@@ -113,23 +241,22 @@ ENTRY(acpi_copy_wakeup_routine)
 	movl    %edx, real_save_cr3 - wakeup_start (%eax)
 	movl    %cr4, %edx
 	movl    %edx, real_save_cr4 - wakeup_start (%eax)
+	movl	%cr0, %edx
+	movl	%edx, real_save_cr0 - wakeup_start (%eax)
 	sgdt    real_save_gdt - wakeup_start (%eax)
 
+	movl	saved_videomode, %edx
+	movl	%edx, video_mode - wakeup_start (%eax)
+	movl	$0x12345678, real_magic - wakeup_start (%eax)
+	movl	$0x12345678, saved_magic2
+
 	# restore the regs we used
 	popl	%edi
 	popl	%esi
 	ret
 
-
 .data
 ALIGN
-# saved registers
-saved_gdt:	.long	0,0
-saved_idt:	.long	0,0
-saved_ldt:	.long	0
-saved_tss:	.long	0
-saved_cr0:	.long	0
-
 ENTRY(saved_ebp)	.long	0
 ENTRY(saved_esi)	.long	0
 ENTRY(saved_edi)	.long	0
@@ -137,3 +264,16 @@ ENTRY(saved_ebx)	.long	0
 
 ENTRY(saved_eip)	.long	0
 ENTRY(saved_esp)	.long	0
+
+ENTRY(saved_magic)	.long	0
+ENTRY(saved_magic2)	.long	0	
+ENTRY(saved_videomode)	.long	0
+
+ALIGN
+# saved registers
+saved_gdt:	.long	0,0
+saved_idt:	.long	0,0
+saved_ldt:	.long	0
+saved_tss:	.long	0
+saved_cr0:	.long	0
+

arch/i386/kernel/apm.c

@@ -1667,6 +1667,7 @@ static int apm(void *unused)
 	daemonize();
 
 	strcpy(current->comm, "kapmd");
+	current->flags |= PF_IOTHREAD;
 	sigfillset(&current->blocked);
 
 	if (apm_info.connection_version == 0) {

arch/i386/kernel/i8259.c

@@ -11,6 +11,7 @@
 #include <linux/smp_lock.h>
 #include <linux/init.h>
 #include <linux/kernel_stat.h>
+#include <linux/device.h>
 
 #include <asm/atomic.h>
 #include <asm/system.h>
@@ -237,7 +238,19 @@ void mask_and_ack_8259A(unsigned int irq)
 	}
 }
 
-void __init init_8259A(int auto_eoi)
+static struct device device_i8259A = {
+	name:	       	"i8259A",
+	bus_id:		"0020",
+};
+
+static void __init init_8259A_devicefs(void)
+{
+	register_sys_device(&device_i8259A);
+}
+
+__initcall(init_8259A_devicefs);
+
+void init_8259A(int auto_eoi)
 {
 	unsigned long flags;
 

arch/i386/kernel/setup.c

@@ -168,6 +168,8 @@ void __init visws_get_board_type_and_rev(void);
 static int disable_x86_serial_nr __initdata = 1;
 static int disable_x86_fxsr __initdata = 0;
 
+extern unsigned long saved_videomode;
+
 /*
  * This is set up by the setup-routine at boot-time
  */
@@ -182,6 +184,7 @@ static int disable_x86_fxsr __initdata = 0;
 #define SYS_DESC_TABLE (*(struct sys_desc_table_struct*)(PARAM+0xa0))
 #define MOUNT_ROOT_RDONLY (*(unsigned short *) (PARAM+0x1F2))
 #define RAMDISK_FLAGS (*(unsigned short *) (PARAM+0x1F8))
+#define VIDEO_MODE (*(unsigned short *) (PARAM+0x1FA))
 #define ORIG_ROOT_DEV (*(unsigned short *) (PARAM+0x1FC))
 #define AUX_DEVICE_INFO (*(unsigned char *) (PARAM+0x1FF))
 #define LOADER_TYPE (*(unsigned char *) (PARAM+0x210))
@@ -681,6 +684,8 @@ void __init setup_arch(char **cmdline_p)
  	drive_info = DRIVE_INFO;
  	screen_info = SCREEN_INFO;
 	apm_info.bios = APM_BIOS_INFO;
+	saved_videomode = VIDEO_MODE;
+	printk("Video mode to be used for restore is %lx\n", saved_videomode);
 	if( SYS_DESC_TABLE.length != 0 ) {
 		MCA_bus = SYS_DESC_TABLE.table[3] &0x2;
 		machine_id = SYS_DESC_TABLE.table[0];

arch/i386/kernel/signal.c

@@ -21,6 +21,7 @@
 #include <linux/tty.h>
 #include <linux/personality.h>
 #include <linux/binfmts.h>
+#include <linux/suspend.h>
 #include <asm/ucontext.h>
 #include <asm/uaccess.h>
 #include <asm/i387.h>
@@ -594,6 +595,11 @@ int do_signal(struct pt_regs *regs, sigset_t *oldset)
 	if ((regs->xcs & 3) != 3)
 		return 1;
 
+	if (current->flags & PF_FREEZE) {
+		refrigerator(0);
+		goto no_signal;
+	}
+
 	if (!oldset)
 		oldset = &current->blocked;
 
@@ -701,6 +707,7 @@ int do_signal(struct pt_regs *regs, sigset_t *oldset)
 		return 1;
 	}
 
+ no_signal:
 	/* Did we come from a system call? */
 	if (regs->orig_eax >= 0) {
 		/* Restart the system call - no handlers present */

arch/i386/kernel/time.c

@@ -42,6 +42,7 @@
 #include <linux/init.h>
 #include <linux/smp.h>
 #include <linux/module.h>
+#include <linux/device.h>
 
 #include <asm/io.h>
 #include <asm/smp.h>
@@ -636,6 +637,17 @@ static unsigned long __init calibrate_tsc(void)
 	return 0;
 }
 
+static struct device device_i8253;
+
+static void time_init_driverfs(void)
+{
+	strcpy(device_i8253.name, "i8253");
+	strcpy(device_i8253.bus_id, "0040");
+	register_sys_device(&device_i8253);
+}
+
+__initcall(time_init_driverfs);
+
 void __init time_init(void)
 {
 	extern int x86_udelay_tsc;

arch/i386/vmlinux.lds

@@ -65,6 +65,12 @@ SECTIONS
   . = ALIGN(4096);
   __init_end = .;
 
+  . = ALIGN(4096);
+  __nosave_begin = .;
+  .data_nosave : { *(.data.nosave) }
+  . = ALIGN(4096);
+  __nosave_end = .;
+
   . = ALIGN(4096);
   .data.page_aligned : { *(.data.idt) }
 

drivers/acpi/acpi_system.c

@@ -33,6 +33,7 @@
 #include <linux/delay.h>
 #include <linux/sysrq.h>
 #include <linux/pm.h>
+#include <linux/device.h>
 #include <asm/uaccess.h>
 #include <asm/acpi.h>
 #include "acpi_bus.h"
@@ -128,6 +129,12 @@ acpi_system_restore_state (
 	/* restore device context */
 	device_resume(RESUME_RESTORE_STATE);
 #endif
+
+	if (dmi_broken & BROKEN_INIT_AFTER_S1) {
+		printk("Broken toshiba laptop -> kicking interrupts\n");
+		init_8259A(0);
+	}
+
 	return AE_OK;
 }
 
@@ -254,23 +261,21 @@ acpi_system_suspend(
 	switch (state)
 	{
 	case ACPI_STATE_S1:
-		/* do nothing */
+		barrier();
+		status = acpi_enter_sleep_state(state);
 		break;
 
 	case ACPI_STATE_S2:
 	case ACPI_STATE_S3:
-		acpi_save_register_state((unsigned long)&&acpi_sleep_done);
+		do_suspend_magic(0);
 		break;
 	}
 
-	barrier();
-	status = acpi_enter_sleep_state(state);
-
-acpi_sleep_done:
-
+	printk("acpi_restore_register_state...");
 	acpi_restore_register_state();
 	restore_flags(flags);
 
+	printk("acpi returning...");
 	return status;
 }
 
@@ -290,6 +295,8 @@ acpi_suspend (
 	if (state < ACPI_STATE_S1 || state > ACPI_STATE_S5)
 		return AE_ERROR;
 
+	freeze_processes();
+
 	/* do we have a wakeup address for S2 and S3? */
 	if (state == ACPI_STATE_S2 || state == ACPI_STATE_S3) {
 		if (!acpi_wakeup_address)
@@ -315,7 +322,9 @@ acpi_suspend (
 	 * no matter what.
 	 */
 	acpi_system_restore_state(state);
+	printk("acpi_leave_sleep_state...");
 	acpi_leave_sleep_state(state);
+	printk("ook\n");
 
 	/* make sure interrupts are enabled */
 	ACPI_ENABLE_IRQS();
@@ -323,6 +332,8 @@ acpi_suspend (
 	/* reset firmware waking vector */
 	acpi_set_firmware_waking_vector((ACPI_PHYSICAL_ADDRESS) 0);
 
+	thaw_processes();
+
 	return status;
 }
 
@@ -701,7 +712,16 @@ acpi_system_write_sleep (
 	if (!system->states[state])
 		return_VALUE(-ENODEV);
 
+	
+#ifdef CONFIG_SOFTWARE_SUSPEND
+	if (state == 4) {
+		/* We are working from process context, that's why we may call it directly. */ 
+		do_software_suspend();
+		return_VALUE(count);
+	}
+#endif
 	status = acpi_suspend(state);
+
 	if (ACPI_FAILURE(status))
 		return_VALUE(-ENODEV);
 	
@@ -1180,6 +1200,10 @@ acpi_system_add (
 		}
 	}
 	printk(")\n");
+#ifdef CONFIG_SOFTWARE_SUSPEND
+	printk(KERN_INFO "Software suspend => we can do S4.");
+	system->states[4] = 1;
+#endif
 
 #ifdef CONFIG_PM
 	/* Install the soft-off (S5) handler. */

drivers/block/floppy.c

@@ -173,6 +173,7 @@ static int print_unex=1;
 #include <linux/interrupt.h>
 #include <linux/init.h>
 #include <linux/devfs_fs_kernel.h>
+#include <linux/device.h>
 
 /*
  * PS/2 floppies have much slower step rates than regular floppies.
@@ -4178,11 +4179,15 @@ static int __init floppy_setup(char *str)
 
 static int have_no_fdc= -ENODEV;
 
+static struct device device_floppy;
 
 int __init floppy_init(void)
 {
 	int i,unit,drive;
 
+	strcpy(device_floppy.name, "floppy");
+	strcpy(device_floppy.bus_id, "03?0");
+	register_sys_device(&device_floppy);
 
 	raw_cmd = NULL;
 

drivers/block/ll_rw_blk.c

@@ -1418,7 +1418,9 @@ static int __make_request(request_queue_t *q, struct bio *bio)
 	req->buffer = bio_data(bio);	/* see ->buffer comment above */
 	req->waiting = NULL;
 	req->bio = req->biotail = bio;
+	if (bio->bi_bdev)
 		req->rq_dev = to_kdev_t(bio->bi_bdev->bd_dev);
+	else	req->rq_dev = NODEV;
 	add_request(q, req, insert_here);
 out:
 	if (freereq)

drivers/block/loop.c

@@ -71,11 +71,11 @@
 #include <linux/smp_lock.h>
 #include <linux/swap.h>
 #include <linux/slab.h>
+#include <linux/loop.h>
+#include <linux/suspend.h>
 
 #include <asm/uaccess.h>
 
-#include <linux/loop.h>		
-
 #define MAJOR_NR LOOP_MAJOR
 
 static int max_loop = 8;
@@ -534,6 +534,7 @@ static int loop_thread(void *data)
 	daemonize();
 
 	sprintf(current->comm, "loop%d", lo->lo_number);
+	current->flags |= PF_IOTHREAD;
 
 	spin_lock_irq(&current->sigmask_lock);
 	sigfillset(&current->blocked);

drivers/char/sysrq.c

@@ -27,6 +27,7 @@
 #include <linux/quotaops.h>
 #include <linux/smp_lock.h>
 #include <linux/module.h>
+#include <linux/suspend.h>
 
 #include <linux/spinlock.h>
 
@@ -317,6 +318,22 @@ static struct sysrq_key_op sysrq_kill_op = {
 	action_msg:	"Kill All Tasks",
 };
 
+#ifdef CONFIG_SOFTWARE_SUSPEND
+static void sysrq_handle_swsusp(int key, struct pt_regs *pt_regs,
+		struct kbd_struct *kbd, struct tty_struct *tty) {
+        if(!software_suspend_enabled) {
+		printk("Software Suspend is not possible now\n");
+		return;
+	}
+	software_suspend();
+}
+static struct sysrq_key_op sysrq_swsusp_op = {
+	handler:	sysrq_handle_swsusp,
+	help_msg:	"suspenD",
+	action_msg:	"Software suspend\n",
+};
+#endif
+
 /* END SIGNAL SYSRQ HANDLERS BLOCK */
 
 
@@ -339,7 +356,11 @@ static struct sysrq_key_op *sysrq_key_table[SYSRQ_KEY_TABLE_LENGTH] = {
 		 and will never arive */
 /* b */	&sysrq_reboot_op,
 /* c */	NULL,
+#ifdef CONFIG_SOFTWARE_SUSPEND
+/* d */	&sysrq_swsusp_op,
+#else
 /* d */	NULL,
+#endif
 /* e */	&sysrq_term_op,
 /* f */	NULL,
 /* g */	NULL,

drivers/ide/ide-disk.c

@@ -27,6 +27,7 @@
 #include <linux/slab.h>
 #include <linux/delay.h>
 #include <linux/ide.h>
+#include <linux/suspend.h>
 
 #include <asm/byteorder.h>
 #include <asm/irq.h>
@@ -562,6 +563,8 @@ static int idedisk_suspend(struct device *dev, u32 state, u32 level)
 	 * already been done...
 	 */
 
+	BUG_ON(in_interrupt());
+
 	if (level != SUSPEND_SAVE_STATE)
 		return 0;
 

drivers/pci/power.c

@@ -110,7 +110,7 @@ static int pci_pm_save_state(u32 state)
 	return error;
 }
 
-static int pci_pm_suspend(u32 state)
+int pci_pm_suspend(u32 state)
 {
 	struct list_head *list;
 	struct pci_bus *bus;
@@ -123,7 +123,7 @@ static int pci_pm_suspend(u32 state)
 	return 0;
 }
 
-static int pci_pm_resume(void)
+int pci_pm_resume(void)
 {
 	struct list_head *list;
 	struct pci_bus *bus;

drivers/usb/storage/usb.c

@@ -323,6 +323,7 @@ static int usb_stor_control_thread(void * __us)
 	/* avoid getting signals */
 	spin_lock_irq(&current->sigmask_lock);
 	flush_signals(current);
+	current->flags |= PF_IOTHREAD;
 	sigfillset(&current->blocked);
 	recalc_sigpending();
 	spin_unlock_irq(&current->sigmask_lock);

fs/buffer.c

@@ -32,6 +32,7 @@
 #include <linux/writeback.h>
 #include <linux/mempool.h>
 #include <linux/hash.h>
+#include <linux/suspend.h>
 #include <asm/bitops.h>
 
 #define BH_ENTRY(list) list_entry((list), struct buffer_head, b_assoc_buffers)
@@ -120,6 +121,8 @@ void unlock_buffer(struct buffer_head *bh)
 	wake_up_buffer(bh);
 }
 
+DECLARE_TASK_QUEUE(tq_bdflush);
+
 /*
  * Block until a buffer comes unlocked.  This doesn't stop it
  * from becoming locked again - you have to lock it yourself

fs/jbd/journal.c

@@ -32,6 +32,7 @@
 #include <linux/init.h>
 #include <linux/mm.h>
 #include <linux/slab.h>
+#include <linux/suspend.h>
 #include <linux/pagemap.h>
 #include <asm/uaccess.h>
 #include <linux/proc_fs.h>
@@ -225,6 +226,7 @@ int kjournald(void *arg)
 			journal->j_commit_interval / HZ);
 	list_add(&journal->j_all_journals, &all_journals);
 
+	current->flags |= PF_KERNTHREAD;
 	/* And now, wait forever for commit wakeup events. */
 	while (1) {
 		if (journal->j_flags & JFS_UNMOUNT)
@@ -245,6 +247,14 @@ int kjournald(void *arg)
 		}
 
 		wake_up(&journal->j_wait_done_commit);
+		if (current->flags & PF_FREEZE) { /* The simpler the better. Flushing journal isn't a
+						     good idea, because that depends on threads that
+						     may be already stopped. */
+			jbd_debug(1, "Now suspending kjournald\n");
+			refrigerator(PF_IOTHREAD);
+			jbd_debug(1, "Resuming kjournald\n");						
+		} else		/* we assume on resume that commits are already there,
+				   so we don't sleep */
 			interruptible_sleep_on(&journal->j_wait_commit);
 
 		jbd_debug(1, "kjournald wakes\n");

fs/reiserfs/journal.c

@@ -57,6 +57,7 @@
 #include <linux/stat.h>
 #include <linux/string.h>
 #include <linux/smp_lock.h>
+#include <linux/suspend.h> 
 
 /* the number of mounted filesystems.  This is used to decide when to
 ** start and kill the commit thread
@@ -1886,6 +1887,7 @@ static int reiserfs_journal_commit_thread(void *nullp) {
   spin_unlock_irq(&current->sigmask_lock);
 
   sprintf(current->comm, "kreiserfsd") ;
+  current->flags |= PF_KERNTHREAD;
   lock_kernel() ;
   while(1) {
 
@@ -1899,6 +1901,11 @@ static int reiserfs_journal_commit_thread(void *nullp) {
       break ;
     }
     wake_up(&reiserfs_commit_thread_done) ;
+#ifdef CONFIG_SOFTWARE_SUSPEND
+    if (current->flags & PF_FREEZE) {
+	    refrigerator(PF_IOTHREAD);
+    } else
+#endif
 	    interruptible_sleep_on_timeout(&reiserfs_commit_thread_wait, 5 * HZ) ;
   }
   unlock_kernel() ;

include/asm-generic/bitops.h

@@ -51,6 +51,12 @@ extern __inline__ int test_bit(int nr, long * addr)
 	return ((mask & *addr) != 0);
 }
 
+/*
+ * fls: find last bit set.
+ */
+
+#define fls(x) generic_fls(x)
+
 #ifdef __KERNEL__
 
 /*

include/asm-i386/bitops.h

@@ -414,6 +414,12 @@ static __inline__ unsigned long __ffs(unsigned long word)
 	return word;
 }
 
+/*
+ * fls: find last bit set.
+ */
+
+#define fls(x) generic_fls(x)
+
 #ifdef __KERNEL__
 
 /*

include/asm-i386/suspend.h

+#ifndef __ASM_I386_SUSPEND_H
+#define __ASM_I386_SUSPEND_H
+#endif
+
+/*
+ * Copyright 2001-2002 Pavel Machek <pavel@suse.cz>
+ * Based on code
+ * Copyright 2001 Patrick Mochel <mochel@osdl.org>
+ */
+#if defined(SUSPEND_C) || defined(ACPI_C)
+#include <asm/desc.h>
+#include <asm/i387.h>
+
+static inline void
+arch_prepare_suspend(void)
+{
+	if (!cpu_has_pse)
+		panic("pse required");
+}
+
+/* image of the saved processor state */
+struct saved_context {
+	u32 eax, ebx, ecx, edx;
+	u32 esp, ebp, esi, edi;
+	u16 es, fs, gs, ss;
+	u32 cr0, cr2, cr3, cr4;
+	u16 gdt_pad;
+	u16 gdt_limit;
+	u32 gdt_base;
+	u16 idt_pad;
+	u16 idt_limit;
+	u32 idt_base;
+	u16 ldt;
+	u16 tss;
+	u32 tr;
+	u32 safety;
+	u32 return_address;
+	u32 eflags;
+} __attribute__((packed));
+
+static struct saved_context saved_context;
+
+#define loaddebug(thread,register) \
+               __asm__("movl %0,%%db" #register  \
+                       : /* no output */ \
+                       :"r" ((thread)->debugreg[register]))
+
+ 
+/*
+ * save_processor_context
+ * 
+ * Save the state of the processor before we go to sleep.
+ *
+ * return_stack is the value of the stack pointer (%esp) as the caller sees it.
+ * A good way could not be found to obtain it from here (don't want to make _too_
+ * many assumptions about the layout of the stack this far down.) Also, the 
+ * handy little __builtin_frame_pointer(level) where level > 0, is blatantly 
+ * buggy - it returns the value of the stack at the proper location, not the 
+ * location, like it should (as of gcc 2.91.66)
+ * 
+ * Note that the context and timing of this function is pretty critical.
+ * With a minimal amount of things going on in the caller and in here, gcc
+ * does a good job of being just a dumb compiler.  Watch the assembly output
+ * if anything changes, though, and make sure everything is going in the right
+ * place. 
+ */
+static inline void save_processor_context (void)
+{
+	kernel_fpu_begin();
+
+	/*
+	 * descriptor tables
+	 */
+	asm volatile ("sgdt (%0)" : "=m" (saved_context.gdt_limit));
+	asm volatile ("sidt (%0)" : "=m" (saved_context.idt_limit));
+	asm volatile ("sldt (%0)" : "=m" (saved_context.ldt));
+	asm volatile ("str (%0)"  : "=m" (saved_context.tr));
+
+	/*
+	 * save the general registers.
+	 * note that gcc has constructs to specify output of certain registers,
+	 * but they're not used here, because it assumes that you want to modify
+	 * those registers, so it tries to be smart and save them beforehand.
+	 * It's really not necessary, and kinda fishy (check the assembly output),
+	 * so it's avoided. 
+	 */
+	asm volatile ("movl %%esp, (%0)" : "=m" (saved_context.esp));
+	asm volatile ("movl %%eax, (%0)" : "=m" (saved_context.eax));
+	asm volatile ("movl %%ebx, (%0)" : "=m" (saved_context.ebx));
+	asm volatile ("movl %%ecx, (%0)" : "=m" (saved_context.ecx));
+	asm volatile ("movl %%edx, (%0)" : "=m" (saved_context.edx));
+	asm volatile ("movl %%ebp, (%0)" : "=m" (saved_context.ebp));
+	asm volatile ("movl %%esi, (%0)" : "=m" (saved_context.esi));
+	asm volatile ("movl %%edi, (%0)" : "=m" (saved_context.edi));
+
+	/*
+	 * segment registers
+	 */
+	asm volatile ("movw %%es, %0" : "=r" (saved_context.es));
+	asm volatile ("movw %%fs, %0" : "=r" (saved_context.fs));
+	asm volatile ("movw %%gs, %0" : "=r" (saved_context.gs));
+	asm volatile ("movw %%ss, %0" : "=r" (saved_context.ss));
+
+	/*
+	 * control registers 
+	 */
+	asm volatile ("movl %%cr0, %0" : "=r" (saved_context.cr0));
+	asm volatile ("movl %%cr2, %0" : "=r" (saved_context.cr2));
+	asm volatile ("movl %%cr3, %0" : "=r" (saved_context.cr3));
+	asm volatile ("movl %%cr4, %0" : "=r" (saved_context.cr4));
+
+	/*
+	 * eflags
+	 */
+	asm volatile ("pushfl ; popl (%0)" : "=m" (saved_context.eflags));
+}
+
+static void fix_processor_context(void)
+{
+	int nr = smp_processor_id();
+	struct tss_struct * t = &init_tss[nr];
+
+	set_tss_desc(nr,t);	/* This just modifies memory; should not be neccessary. But... This is neccessary, because 386 hardware has concept of busy tsc or some similar stupidity. */
+        gdt_table[__TSS(nr)].b &= 0xfffffdff;
+
+	load_TR(nr);		/* This does ltr */
+
+	load_LDT(&current->mm->context);	/* This does lldt */
+
+	/*
+	 * Now maybe reload the debug registers
+	 */
+	if (current->thread.debugreg[7]){
+                loaddebug(&current->thread, 0);
+                loaddebug(&current->thread, 1);
+                loaddebug(&current->thread, 2);
+                loaddebug(&current->thread, 3);
+                /* no 4 and 5 */
+                loaddebug(&current->thread, 6);
+                loaddebug(&current->thread, 7);
+	}
+
+}
+
+static void
+do_fpu_end(void)
+{
+        /* restore FPU regs if necessary */
+	/* Do it out of line so that gcc does not move cr0 load to some stupid place */
+        kernel_fpu_end();
+}
+
+/*
+ * restore_processor_context
+ * 
+ * Restore the processor context as it was before we went to sleep
+ * - descriptor tables
+ * - control registers
+ * - segment registers
+ * - flags
+ * 
+ * Note that it is critical that this function is declared inline.  
+ * It was separated out from restore_state to make that function
+ * a little clearer, but it needs to be inlined because we won't have a
+ * stack when we get here (so we can't push a return address).
+ */
+static inline void restore_processor_context (void)
+{
+	/*
+	 * first restore %ds, so we can access our data properly
+	 */
+	asm volatile (".align 4");
+	asm volatile ("movw %0, %%ds" :: "r" ((u16)__KERNEL_DS));
+
+
+	/*
+	 * control registers
+	 */
+	asm volatile ("movl %0, %%cr4" :: "r" (saved_context.cr4));
+	asm volatile ("movl %0, %%cr3" :: "r" (saved_context.cr3));
+	asm volatile ("movl %0, %%cr2" :: "r" (saved_context.cr2));
+	asm volatile ("movl %0, %%cr0" :: "r" (saved_context.cr0));
+	
+	/*
+	 * segment registers
+	 */
+	asm volatile ("movw %0, %%es" :: "r" (saved_context.es));
+	asm volatile ("movw %0, %%fs" :: "r" (saved_context.fs));
+	asm volatile ("movw %0, %%gs" :: "r" (saved_context.gs));
+	asm volatile ("movw %0, %%ss" :: "r" (saved_context.ss));
+
+	/*
+	 * the other general registers
+	 *
+	 * note that even though gcc has constructs to specify memory 
+	 * input into certain registers, it will try to be too smart
+	 * and save them at the beginning of the function.  This is esp.
+	 * bad since we don't have a stack set up when we enter, and we 
+	 * want to preserve the values on exit. So, we set them manually.
+	 */
+	asm volatile ("movl %0, %%esp" :: "m" (saved_context.esp));
+	asm volatile ("movl %0, %%ebp" :: "m" (saved_context.ebp));
+	asm volatile ("movl %0, %%eax" :: "m" (saved_context.eax));
+	asm volatile ("movl %0, %%ebx" :: "m" (saved_context.ebx));
+	asm volatile ("movl %0, %%ecx" :: "m" (saved_context.ecx));
+	asm volatile ("movl %0, %%edx" :: "m" (saved_context.edx));
+	asm volatile ("movl %0, %%esi" :: "m" (saved_context.esi));
+	asm volatile ("movl %0, %%edi" :: "m" (saved_context.edi));
+
+	/*
+	 * now restore the descriptor tables to their proper values
+	 */
+	asm volatile ("lgdt (%0)" :: "m" (saved_context.gdt_limit));
+	asm volatile ("lidt (%0)" :: "m" (saved_context.idt_limit));
+	asm volatile ("lldt (%0)" :: "m" (saved_context.ldt));
+
+#if 0
+	asm volatile ("ltr (%0)"  :: "m" (saved_context.tr));
+#endif
+
+	fix_processor_context();
+
+	/*
+	 * the flags
+	 */
+	asm volatile ("pushl %0 ; popfl" :: "m" (saved_context.eflags));
+
+	do_fpu_end();
+}
+
+#endif
+#ifdef SUSPEND_C
+#if 1
+/* Local variables for do_magic */
+static int loop __nosavedata = 0;
+static int loop2 __nosavedata = 0;
+
+/*
+ * (KG): Since we affect stack here, we make this function as flat and easy
+ * as possible in order to not provoke gcc to use local variables on the stack.
+ * Note that on resume, all (expect nosave) variables will have the state from
+ * the time of writing (suspend_save_image) and the registers (including the
+ * stack pointer, but excluding the instruction pointer) will be loaded with 
+ * the values saved at save_processor_context() time.
+ */
+static void do_magic(int resume)
+{
+	/* DANGER WILL ROBINSON!
+	 *
+	 * If this function is too difficult for gcc to optimize, it will crash and burn!
+	 * see above.
+	 *
+	 * DO NOT TOUCH.
+	 */
+
+	if (!resume) {
+		do_magic_suspend_1();
+		save_processor_context();	/* We need to capture registers and memory at "same time" */
+		do_magic_suspend_2();		/* If everything goes okay, this function does not return */
+		return;
+	}
+
+	/* We want to run from swapper_pg_dir, since swapper_pg_dir is stored in constant
+	 * place in memory 
+	 */
+
+        __asm__( "movl %%ecx,%%cr3\n" ::"c"(__pa(swapper_pg_dir)));
+
+/*
+ * Final function for resuming: after copying the pages to their original
+ * position, it restores the register state.
+ */
+
+	do_magic_resume_1();
+
+	/* Critical section here: noone should touch memory from now */
+	/* This works, because nr_copy_pages, pagedir_nosave, loop and loop2 are nosavedata */
+	for (loop=0; loop < nr_copy_pages; loop++) {
+		/* You may not call something (like copy_page) here:
+		   We may absolutely not use stack at this point */
+		for (loop2=0; loop2 < PAGE_SIZE; loop2++) {
+			*(((char *)((pagedir_nosave+loop)->orig_address))+loop2) =
+				*(((char *)((pagedir_nosave+loop)->address))+loop2);
+			__flush_tlb();
+		}
+	}
+/* FIXME: What about page tables? Writing data pages may toggle
+   accessed/dirty bits in our page tables. That should be no problems
+   with 4MB page tables. That's why we require have_pse. */
+
+/* Danger: previous loop probably destroyed our current stack. Better hope it did not use
+   any stack space, itself.
+
+   When this function is entered at resume time, we move stack to _old_ place.
+   This is means that this function must use no stack and no local variables in registers.
+*/
+	restore_processor_context();
+/* Ahah, we now run with our old stack, and with registers copied from suspend time */
+
+	do_magic_resume_2();
+}
+#endif
+#endif 
+

include/linux/bitops.h

 #ifndef _LINUX_BITOPS_H
 #define _LINUX_BITOPS_H
-
+#include <asm/bitops.h>
 
 /*
  * ffs: find first bit set. This is defined the same way as
@@ -37,6 +37,47 @@ static inline int generic_ffs(int x)
 	return r;
 }
 
+/*
+ * fls: find last bit set.
+ */
+
+extern __inline__ int generic_fls(int x)
+{
+	int r = 32;
+
+	if (!x)
+		return 0;
+	if (!(x & 0xffff0000)) {
+		x <<= 16;
+		r -= 16;
+	}
+	if (!(x & 0xff000000)) {
+		x <<= 8;
+		r -= 8;
+	}
+	if (!(x & 0xf0000000)) {
+		x <<= 4;
+		r -= 4;
+	}
+	if (!(x & 0xc0000000)) {
+		x <<= 2;
+		r -= 2;
+	}
+	if (!(x & 0x80000000)) {
+		x <<= 1;
+		r -= 1;
+	}
+	return r;
+}
+
+extern __inline__ int get_bitmask_order(unsigned int count)
+{
+	int order;
+	
+	order = fls(count);
+	return order;	/* We could be slightly more clever with -1 here... */
+}
+
 /*
  * hweightN: returns the hamming weight (i.e. the number
  * of bits set) of a N-bit word

include/linux/init.h

@@ -169,6 +169,9 @@ typedef void (*__cleanup_module_func_t)(void);
 
 #endif
 
+/* Data marked not to be saved by software_suspend() */
+#define __nosavedata __attribute__ ((__section__ (".data.nosave")))
+
 #ifdef CONFIG_HOTPLUG
 #define __devinit
 #define __devinitdata

include/linux/page-flags.h

@@ -64,6 +64,7 @@
 
 #define PG_private		12	/* Has something at ->private */
 #define PG_writeback		13	/* Page is under writeback */
+#define PG_nosave		15	/* Used for system suspend/resume */
 
 /*
  * Global page accounting.  One instance per CPU.
@@ -207,6 +208,12 @@ extern void get_page_state(struct page_state *ret);
 		ret;							\
 	})
 
+#define PageNosave(page)	test_bit(PG_nosave, &(page)->flags)
+#define SetPageNosave(page)	set_bit(PG_nosave, &(page)->flags)
+#define TestSetPageNosave(page)	test_and_set_bit(PG_nosave, &(page)->flags)
+#define ClearPageNosave(page)		clear_bit(PG_nosave, &(page)->flags)
+#define TestClearPageNosave(page)	test_and_clear_bit(PG_nosave, &(page)->flags)
+
 /*
  * The PageSwapCache predicate doesn't use a PG_flag at this time,
  * but it may again do so one day.

include/linux/reboot.h

@@ -20,6 +20,7 @@
  * CAD_OFF     Ctrl-Alt-Del sequence sends SIGINT to init task.
  * POWER_OFF   Stop OS and remove all power from system, if possible.
  * RESTART2    Restart system using given command string.
+ * SW_SUSPEND  Suspend system using Software Suspend if compiled in
  */
 
 #define	LINUX_REBOOT_CMD_RESTART	0x01234567
@@ -28,6 +29,7 @@
 #define	LINUX_REBOOT_CMD_CAD_OFF	0x00000000
 #define	LINUX_REBOOT_CMD_POWER_OFF	0x4321FEDC
 #define	LINUX_REBOOT_CMD_RESTART2	0xA1B2C3D4
+#define	LINUX_REBOOT_CMD_SW_SUSPEND	0xD000FCE2
 
 
 #ifdef __KERNEL__
@@ -46,6 +48,13 @@ extern void machine_restart(char *cmd);
 extern void machine_halt(void);
 extern void machine_power_off(void);
 
+/*
+ * Architecture-independent suspend facility
+ */
+
+extern void software_suspend(void);
+extern unsigned char software_suspend_enabled;
+
 #endif
 
 #endif /* _LINUX_REBOOT_H */

include/linux/sched.h

@@ -388,6 +388,11 @@ do { if (atomic_dec_and_test(&(tsk)->usage)) __put_task_struct(tsk); } while(0)
 #define PF_FLUSHER	0x00004000	/* responsible for disk writeback */
 #define PF_RADIX_TREE	0x00008000	/* debug: performing radix tree alloc */
 
+#define PF_FREEZE	0x00010000	/* this task should be frozen for suspend */
+#define PF_IOTHREAD	0x00020000	/* this thread is needed for doing I/O to swap */
+#define PF_KERNTHREAD	0x00040000	/* this thread is a kernel thread that cannot be sent signals to */
+#define PF_FROZEN	0x00080000	/* frozen for system suspend */
+
 /*
  * Ptrace flags
  */

include/linux/suspend.h

+#ifndef _LINUX_SWSUSP_H
+#define _LINUX_SWSUSP_H
+
+#include <asm/suspend.h>
+#include <linux/swap.h>
+#include <linux/notifier.h>
+#include <linux/config.h>
+
+extern unsigned char software_suspend_enabled;
+
+#define NORESUME	 1
+#define RESUME_SPECIFIED 2
+
+#ifdef CONFIG_SOFTWARE_SUSPEND
+/* page backup entry */
+typedef struct pbe {
+	unsigned long address;		/* address of the copy */
+	unsigned long orig_address;	/* original address of page */
+	swp_entry_t swap_address;	
+	swp_entry_t dummy;		/* we need scratch space at 
+					 * end of page (see link, diskpage)
+					 */
+} suspend_pagedir_t;
+
+#define SWAP_FILENAME_MAXLENGTH	32
+
+struct suspend_header {
+	__u32 version_code;
+	unsigned long num_physpages;
+	char machine[8];
+	char version[20];
+	int num_cpus;
+	int page_size;
+	unsigned long suspend_pagedir;
+	unsigned int num_pbes;
+	struct swap_location {
+		char filename[SWAP_FILENAME_MAXLENGTH];
+	} swap_location[MAX_SWAPFILES];
+};
+
+#define SUSPEND_PD_PAGES(x)     (((x)*sizeof(struct pbe))/PAGE_SIZE+1)
+   
+extern struct tq_struct suspend_tq;
+
+/* mm/vmscan.c */
+extern int shrink_mem(void);
+
+/* kernel/suspend.c */
+extern void software_suspend(void);
+extern void software_resume(void);
+extern int resume_setup(char *str);
+
+extern int register_suspend_notifier(struct notifier_block *);
+extern int unregister_suspend_notifier(struct notifier_block *);
+extern void refrigerator(unsigned long);
+
+#else
+#define software_suspend()		do { } while(0)
+#define software_resume()		do { } while(0)
+#define register_suspend_notifier(a)	do { } while(0)
+#define unregister_suspend_notifier(a)	do { } while(0)
+#define refrigerator(a)			do { BUG(); } while(0)
+#endif
+
+#endif /* _LINUX_SWSUSP_H */

include/linux/tqueue.h

@@ -66,7 +66,7 @@ typedef struct list_head task_queue;
 #define DECLARE_TASK_QUEUE(q)	LIST_HEAD(q)
 #define TQ_ACTIVE(q)		(!list_empty(&q))
 
-extern task_queue tq_timer, tq_immediate, tq_disk;
+extern task_queue tq_timer, tq_immediate, tq_disk, tq_bdflush;
 
 /*
  * To implement your own list of active bottom halfs, use the following

init/do_mounts.c

@@ -10,6 +10,7 @@
 #include <linux/fd.h>
 #include <linux/tty.h>
 #include <linux/init.h>
+#include <linux/suspend.h>
 
 #include <linux/nfs_fs.h>
 #include <linux/nfs_fs_sb.h>
@@ -825,6 +826,11 @@ void prepare_namespace(void)
 #endif
 
 	create_dev("/dev/root", ROOT_DEV, NULL);
+
+	/* This has to be before mounting root, because even readonly mount of reiserfs would replay
+	   log corrupting stuff */
+	software_resume();
+
 	if (mount_initrd) {
 		if (initrd_load() && !kdev_same(ROOT_DEV, mk_kdev(RAMDISK_MAJOR, 0))) {
 			handle_initrd();

kernel/Makefile

@@ -10,7 +10,7 @@
 O_TARGET := kernel.o
 
 export-objs = signal.o sys.o kmod.o context.o ksyms.o pm.o exec_domain.o \
-		printk.o platform.o 
+		printk.o platform.o suspend.o
 
 obj-y     = sched.o dma.o fork.o exec_domain.o panic.o printk.o \
 	    module.o exit.o itimer.o info.o time.o softirq.o resource.o \
@@ -21,6 +21,7 @@ obj-$(CONFIG_UID16) += uid16.o
 obj-$(CONFIG_MODULES) += ksyms.o
 obj-$(CONFIG_PM) += pm.o
 obj-$(CONFIG_BSD_PROCESS_ACCT) += acct.o
+obj-$(CONFIG_SOFTWARE_SUSPEND) += suspend.o
 
 ifneq ($(CONFIG_IA64),y)
 # According to Alan Modra <alan@linuxcare.com.au>, the -fno-omit-frame-pointer is

kernel/context.c

@@ -72,6 +72,7 @@ static int context_thread(void *startup)
 
 	daemonize();
 	strcpy(curtask->comm, "keventd");
+	current->flags |= PF_IOTHREAD;
 	keventd_running = 1;
 	keventd_task = curtask;
 

kernel/signal.c

@@ -493,7 +493,7 @@ static int send_signal(int sig, struct siginfo *info, struct sigpending *signals
  * No need to set need_resched since signal event passing
  * goes through ->blocked
  */
-static inline void signal_wake_up(struct task_struct *t)
+inline void signal_wake_up(struct task_struct *t)
 {
 	set_tsk_thread_flag(t,TIF_SIGPENDING);
 

kernel/softirq.c

@@ -366,6 +366,7 @@ static int ksoftirqd(void * __bind_cpu)
 
 	daemonize();
 	set_user_nice(current, 19);
+	current->flags |= PF_IOTHREAD;
 	sigfillset(&current->blocked);
 
 	/* Migrate to the right CPU */

kernel/sys.c

@@ -4,6 +4,7 @@
  *  Copyright (C) 1991, 1992  Linus Torvalds
  */
 
+#include <linux/config.h>
 #include <linux/module.h>
 #include <linux/mm.h>
 #include <linux/utsname.h>
@@ -347,6 +348,16 @@ asmlinkage long sys_reboot(int magic1, int magic2, unsigned int cmd, void * arg)
 		machine_restart(buffer);
 		break;
 
+#ifdef CONFIG_SOFTWARE_SUSPEND
+	case LINUX_REBOOT_CMD_SW_SUSPEND:
+		if(!software_suspend_enabled)
+			return -EAGAIN;
+		
+		software_suspend();
+		do_exit(0);
+		break;
+#endif
+
 	default:
 		unlock_kernel();
 		return -EINVAL;

mm/page_alloc.c

@@ -22,6 +22,7 @@
 #include <linux/slab.h>
 #include <linux/compiler.h>
 #include <linux/module.h>
+#include <linux/suspend.h>
 
 unsigned long totalram_pages;
 unsigned long totalhigh_pages;
@@ -247,6 +248,46 @@ static struct page * rmqueue(zone_t *zone, unsigned int order)
 	return NULL;
 }
 
+#ifdef CONFIG_SOFTWARE_SUSPEND
+int is_head_of_free_region(struct page *p)
+{
+	pg_data_t *pgdat = pgdat_list;
+	unsigned type;
+	unsigned long flags;
+
+	for (type=0;type < MAX_NR_ZONES; type++) {
+		zone_t *zone = pgdat->node_zones + type;
+		int order = MAX_ORDER - 1;
+		free_area_t *area;
+		struct list_head *head, *curr;
+		spin_lock_irqsave(&zone->lock, flags);	/* Should not matter as we need quiescent system for suspend anyway, but... */
+
+		do {
+			area = zone->free_area + order;
+			head = &area->free_list;
+			curr = head;
+
+			for(;;) {
+				if(!curr) {
+//					printk("FIXME: this should not happen but it does!!!");
+					break;
+				}
+				if(p != memlist_entry(curr, struct page, list)) {
+					curr = memlist_next(curr);
+					if (curr == head)
+						break;
+					continue;
+				}
+				return 1 << order;
+			}
+		} while(order--);
+		spin_unlock_irqrestore(&zone->lock, flags);
+
+	}
+	return 0;
+}
+#endif /* CONFIG_SOFTWARE_SUSPEND */
+
 #ifndef CONFIG_DISCONTIGMEM
 struct page *_alloc_pages(unsigned int gfp_mask, unsigned int order)
 {

mm/page_io.c

@@ -86,12 +86,16 @@ static int rw_swap_page_base(int rw, swp_entry_t entry, struct page *page)
  *  - it's marked as being swap-cache
  *  - it's associated with the swap inode
  */
+extern long suspend_device;
 void rw_swap_page(int rw, struct page *page)
 {
 	swp_entry_t entry;
 
 	entry.val = page->index;
 
+	if (suspend_device)
+		panic("I refuse to corrupt memory/swap.");
+
 	if (!PageLocked(page))
 		PAGE_BUG(page);
 	if (!PageSwapCache(page))

mm/pdflush.c

@@ -14,6 +14,7 @@
 #include <linux/gfp.h>
 #include <linux/init.h>
 #include <linux/module.h>
+#include <linux/suspend.h>
 
 
 /*
@@ -96,7 +97,7 @@ static int __pdflush(struct pdflush_work *my_work)
 	recalc_sigpending();
 	spin_unlock_irq(&current->sigmask_lock);
 
-	current->flags |= PF_FLUSHER;
+	current->flags |= PF_FLUSHER | PF_KERNTHREAD;
 	my_work->fn = NULL;
 	my_work->who = current;
 
@@ -107,14 +108,20 @@ static int __pdflush(struct pdflush_work *my_work)
 	for ( ; ; ) {
 		struct pdflush_work *pdf;
 
+#ifdef CONFIG_SOFTWARE_SUSPEND
+		run_task_queue(&tq_bdflush);
+#endif
 		list_add(&my_work->list, &pdflush_list);
 		my_work->when_i_went_to_sleep = jiffies;
 		set_current_state(TASK_INTERRUPTIBLE);
 		spin_unlock_irq(&pdflush_lock);
 
+		if (current->flags & PF_FREEZE)
+			refrigerator(PF_IOTHREAD);
 		schedule();
 
 		preempt_enable();
+		if (my_work->fn)
 			(*my_work->fn)(my_work->arg0);
 		preempt_disable();
 
@@ -146,6 +153,7 @@ static int __pdflush(struct pdflush_work *my_work)
 			pdf->when_i_went_to_sleep = jiffies;	/* Limit exit rate */
 			break;					/* exeunt */
 		}
+		my_work->fn = NULL;
 	}
 	nr_pdflush_threads--;
 //	printk("pdflush %d [%d] ends\n", nr_pdflush_threads, current->pid);

mm/vmscan.c

@@ -23,6 +23,7 @@
 #include <linux/file.h>
 #include <linux/writeback.h>
 #include <linux/compiler.h>
+#include <linux/suspend.h>
 
 #include <asm/pgalloc.h>
 #include <asm/tlbflush.h>
@@ -781,18 +782,22 @@ int kswapd(void *unused)
 	 * us from recursively trying to free more memory as we're
 	 * trying to free the first piece of memory in the first place).
 	 */
-	tsk->flags |= PF_MEMALLOC;
+	tsk->flags |= PF_MEMALLOC | PF_KERNTHREAD;
 
 	/*
 	 * Kswapd main loop.
 	 */
 	for (;;) {
+		if (current->flags & PF_FREEZE)
+			refrigerator(PF_IOTHREAD);
 		__set_current_state(TASK_INTERRUPTIBLE);
 		add_wait_queue(&kswapd_wait, &wait);
 
 		mb();
-		if (kswapd_can_sleep())
+		if (kswapd_can_sleep()) {
 			schedule();
+		}
+		
 
 		__set_current_state(TASK_RUNNING);
 		remove_wait_queue(&kswapd_wait, &wait);