Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Chuck Ebbert]

In-Reply-To:<1153527274.13699.36.camel@localhost.localdomain>

On Sat, 22 Jul 2006 10:14:34 +1000, Rusty Russell wrote:
>
> Abstract sensitive instructions in assembler code, replacing them with
> macros (which currently are #defined to the native versions).  We use
> long names: assembler is case-insensitive, so if something goes wrong
> and macros do not expand, it would assemble anyway.
...
> --- working-2.6.18-rc2-hg-paravirt.orig/arch/i386/kernel/entry.S      2006-07-21 21:09:22.000000000 +1000
> +++ working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S   2006-07-22 04:32:25.000000000 +1000
> @@ -76,8 +76,15 @@
>  NT_MASK              = 0x00004000
>  VM_MASK              = 0x00020000
>  
> +/* These are replaces for paravirtualization */
> +#define DISABLE_INTERRUPTS           cli
> +#define ENABLE_INTERRUPTS            sti
> +#define ENABLE_INTERRUPTS_SYSEXIT    sti; sysexit
> +#define INTERRUPT_RETURN             iret
> +#define GET_CR0                      movl %cr0, %eax
> +
 
Could you change GET_CR0 to MOV_CR0_EAX?  GET_CR0 seems like it's
taking a reference or something.

-- 
Chuck
And did we tell you the name of the game, boy, we call it Riding the Gravy Train.

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Rusty Russell]

On Fri, 2006-07-21 at 23:23 -0400, Chuck Ebbert wrote:
> In-Reply-To:<1153527274.13699.36.camel@localhost.localdomain>
> 
> On Sat, 22 Jul 2006 10:14:34 +1000, Rusty Russell wrote:
> > +#define GET_CR0                      movl %cr0, %eax
>  
> Could you change GET_CR0 to MOV_CR0_EAX?  GET_CR0 seems like it's
> taking a reference or something.

Hi Chuck,

	Agreed: certainly eax should be mentioned.  GET_CR0_INTO_EAX?  MOV is a
bit close to describing how it's happening (which, on paravirt it might
not be) so it might lead the reader to unwarranted assumptions.

===
Abstract sensitive instructions in assembler code, replacing them with
macros (which currently are #defined to the native versions).  We use
long names: assembler is case-insensitive, so if something goes wrong
and macros do not expand, it would assemble anyway.

Resulting object files are exactly the same as before.

Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
Index: working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S
===================================================================
--- working-2.6.18-rc2-hg-paravirt.orig/arch/i386/kernel/entry.S	2006-07-21 21:09:22.000000000 +1000
+++ working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S	2006-07-22 04:32:25.000000000 +1000
@@ -76,8 +76,15 @@
 NT_MASK		= 0x00004000
 VM_MASK		= 0x00020000
 
+/* These are replaces for paravirtualization */
+#define DISABLE_INTERRUPTS		cli
+#define ENABLE_INTERRUPTS		sti
+#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
+#define INTERRUPT_RETURN		iret
+#define GET_CR0_INTO_EAX		movl %cr0, %eax
+
 #ifdef CONFIG_PREEMPT
-#define preempt_stop		cli; TRACE_IRQS_OFF
+#define preempt_stop		DISABLE_INTERRUPTS; TRACE_IRQS_OFF
 #else
 #define preempt_stop
 #define resume_kernel		restore_nocheck
@@ -234,7 +241,7 @@
 	cmpl $SEGMENT_RPL_MASK, %eax
 	jb resume_kernel		# returning to kernel or vm86-space
 ENTRY(resume_userspace)
- 	cli				# make sure we don't miss an interrupt
+ 	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	movl TI_flags(%ebp), %ecx
@@ -245,7 +252,7 @@
 
 #ifdef CONFIG_PREEMPT
 ENTRY(resume_kernel)
-	cli
+	DISABLE_INTERRUPTS
 	cmpl $0,TI_preempt_count(%ebp)	# non-zero preempt_count ?
 	jnz restore_nocheck
 need_resched:
@@ -273,7 +280,7 @@
 	 * No need to follow this irqs on/off section: the syscall
 	 * disabled irqs and here we enable it straight after entry:
 	 */
-	sti
+	ENABLE_INTERRUPTS
 	pushl $(__USER_DS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET ss, 0*/
@@ -318,7 +325,7 @@
 	jae syscall_badsys
 	call *sys_call_table(,%eax,4)
 	movl %eax,EAX(%esp)
-	cli
+	DISABLE_INTERRUPTS
 	TRACE_IRQS_OFF
 	movl TI_flags(%ebp), %ecx
 	testw $_TIF_ALLWORK_MASK, %cx
@@ -328,8 +335,7 @@
 	movl OLDESP(%esp), %ecx
 	xorl %ebp,%ebp
 	TRACE_IRQS_ON
-	sti
-	sysexit
+	ENABLE_INTERRUPTS_SYSEXIT
 	CFI_ENDPROC
 
 
@@ -354,7 +360,7 @@
 	call *sys_call_table(,%eax,4)
 	movl %eax,EAX(%esp)		# store the return value
 syscall_exit:
-	cli				# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -379,11 +385,11 @@
 	RESTORE_REGS
 	addl $4, %esp
 	CFI_ADJUST_CFA_OFFSET -4
-1:	iret
+1:	INTERRUPT_RETURN
 .section .fixup,"ax"
 iret_exc:
 	TRACE_IRQS_ON
-	sti
+	ENABLE_INTERRUPTS
 	pushl $0			# no error code
 	pushl $do_iret_error
 	jmp error_code
@@ -407,7 +413,7 @@
 	 * dosemu and wine happy. */
 	subl $8, %esp		# reserve space for switch16 pointer
 	CFI_ADJUST_CFA_OFFSET 8
-	cli
+	DISABLE_INTERRUPTS
 	TRACE_IRQS_OFF
 	movl %esp, %eax
 	/* Set up the 16bit stack frame with switch32 pointer on top,
@@ -417,7 +423,7 @@
 	TRACE_IRQS_IRET
 	RESTORE_REGS
 	lss 20+4(%esp), %esp	# switch to 16bit stack
-1:	iret
+1:	INTERRUPT_RETURN
 .section __ex_table,"a"
 	.align 4
 	.long 1b,iret_exc
@@ -432,7 +438,7 @@
 	jz work_notifysig
 work_resched:
 	call schedule
-	cli				# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -488,7 +494,7 @@
 	testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl
 	jz work_pending
 	TRACE_IRQS_ON
-	sti				# could let do_syscall_trace() call
+	ENABLE_INTERRUPTS		# could let do_syscall_trace() call
 					# schedule() instead
 	movl %esp, %eax
 	movl $1, %edx
@@ -667,7 +673,7 @@
 	pushl $-1			# mark this as an int
 	CFI_ADJUST_CFA_OFFSET 4
 	SAVE_ALL
-	movl %cr0, %eax
+	GET_CR0_INTO_EAX
 	testl $0x4, %eax		# EM (math emulation bit)
 	jne device_not_available_emulate
 	preempt_stop
@@ -797,7 +803,7 @@
 	call do_nmi
 	RESTORE_REGS
 	lss 12+4(%esp), %esp		# back to 16bit stack
-1:	iret
+1:	INTERRUPT_RETURN
 	CFI_ENDPROC
 .section __ex_table,"a"
 	.align 4
Index: working-2.6.18-rc2-hg-paravirt/include/asm-i386/spinlock.h
===================================================================
--- working-2.6.18-rc2-hg-paravirt.orig/include/asm-i386/spinlock.h	2006-07-21 20:27:59.000000000 +1000
+++ working-2.6.18-rc2-hg-paravirt/include/asm-i386/spinlock.h	2006-07-22 04:32:51.000000000 +1000
@@ -17,6 +17,9 @@
  * (the type definitions are in asm/spinlock_types.h)
  */
 
+#define CLI_STRING	"cli"
+#define STI_STRING	"sti"
+
 #define __raw_spin_is_locked(x) \
 		(*(volatile signed char *)(&(x)->slock) <= 0)
 
@@ -43,12 +46,12 @@
 	"2:\t" \
 	"testl $0x200, %1\n\t" \
 	"jz 4f\n\t" \
-	"sti\n" \
+	STI_STRING "\n" \
 	"3:\t" \
 	"rep;nop\n\t" \
 	"cmpb $0, %0\n\t" \
 	"jle 3b\n\t" \
-	"cli\n\t" \
+	CLI_STRING "\n\t" \
 	"jmp 1b\n" \
 	"4:\t" \
 	"rep;nop\n\t" \

-- 
Help! Save Australia from the worst of the DMCA: http://linux.org.au/law

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Andi Kleen]

On Sat, Jul 22, 2006 at 11:34:48PM +1000, Rusty Russell wrote:
> On Fri, 2006-07-21 at 23:23 -0400, Chuck Ebbert wrote:
> > In-Reply-To:<1153527274.13699.36.camel@localhost.localdomain>
> > 
> > On Sat, 22 Jul 2006 10:14:34 +1000, Rusty Russell wrote:
> > > +#define GET_CR0                      movl %cr0, %eax
> >  
> > Could you change GET_CR0 to MOV_CR0_EAX?  GET_CR0 seems like it's
> > taking a reference or something.
> 
> Hi Chuck,
> 
> 	Agreed: certainly eax should be mentioned.  GET_CR0_INTO_EAX?  MOV is a
> bit close to describing how it's happening (which, on paravirt it might
> not be) so it might lead the reader to unwarranted assumptions.
> 
> ===
> Abstract sensitive instructions in assembler code, replacing them with
> macros (which currently are #defined to the native versions).  We use
> long names: assembler is case-insensitive, so if something goes wrong
> and macros do not expand, it would assemble anyway.
> 
> Resulting object files are exactly the same as before.
> 
> Signed-off-by: Rusty Russell <rusty@rustcorp.com.au>
> Index: working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S
> ===================================================================
> --- working-2.6.18-rc2-hg-paravirt.orig/arch/i386/kernel/entry.S	2006-07-21 21:09:22.000000000 +1000
> +++ working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S	2006-07-22 04:32:25.000000000 +1000
> @@ -76,8 +76,15 @@
>  NT_MASK		= 0x00004000
>  VM_MASK		= 0x00020000
>  
> +/* These are replaces for paravirtualization */
> +#define DISABLE_INTERRUPTS		cli
> +#define ENABLE_INTERRUPTS		sti
> +#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
> +#define INTERRUPT_RETURN		iret
> +#define GET_CR0_INTO_EAX		movl %cr0, %eax

I would rather pass the register to the macro? If you start to
clobber registers you would need to pass in the tmp registers
too I guess.

-Andi

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Jeremy Fitzhardinge]

Andi Kleen wrote:
> I would rather pass the register to the macro? If you start to
> clobber registers you would need to pass in the tmp registers
> too I guess.
The intent is that these code sequences can be patched in place, so they 
need to have a specific abi at the machine-code level.  Some of the 
operations whose Xen implementation clobbers a register have that in 
their definition, on the assumption that 1 temp is enough, and that 
another interface which needs more can organize save/restoring registers 
for itself.

    J

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Peter Chubb]

>>>>> "Rusty" == Rusty Russell <rusty@rustcorp.com.au> writes:

Rusty> 	Agreed: certainly eax should be mentioned.  GET_CR0_INTO_EAX?
Rusty> MOV is a bit close to describing how it's happening (which, on
Rusty> paravirt it might not be) so it might lead the reader to
Rusty> unwarranted assumptions.


Aren't these ideal cases to use a tool to convert, rather than adding
manual changes to the source?  The `afterburner' approach works quite
well for this kind of thing.

Peter C

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Rusty Russell]

On Wed, 2006-07-26 at 09:52 +1000, Peter Chubb wrote:
> >>>>> "Rusty" == Rusty Russell <rusty@rustcorp.com.au> writes:
> 
> Rusty> 	Agreed: certainly eax should be mentioned.  GET_CR0_INTO_EAX?
> Rusty> MOV is a bit close to describing how it's happening (which, on
> Rusty> paravirt it might not be) so it might lead the reader to
> Rusty> unwarranted assumptions.
> 
> Aren't these ideal cases to use a tool to convert, rather than adding
> manual changes to the source?  The `afterburner' approach works quite
> well for this kind of thing.

Agreed, afterburner works well in that it's minimally invasive in the
source.  However, that is precisely the opposite of the aim of patches
like this one which are integrated: that the source code be explicit.

Cheers,
Rusty.
-- 
Help! Save Australia from the worst of the DMCA: http://linux.org.au/law

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Pavel Machek]

Hi!

> > Abstract sensitive instructions in assembler code, replacing them with
> > macros (which currently are #defined to the native versions).  We use
> > long names: assembler is case-insensitive, so if something goes wrong
> > and macros do not expand, it would assemble anyway.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

> > @@ -76,8 +76,15 @@
> >  NT_MASK		= 0x00004000
> >  VM_MASK		= 0x00020000
> >  
> > +/* These are replaces for paravirtualization */
> > +#define DISABLE_INTERRUPTS		cli
> > +#define ENABLE_INTERRUPTS		sti
> > +#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
> > +#define INTERRUPT_RETURN		iret
> 
> Could we use some less verbose names, like possibly CLI, STI,
> STI_SYSEXIT, IRET ?

Apparently I can't read, it was explained in changelog. Could we still
use something shorter, like perhaps _CLI / _STI / _IRET ?
								Pavel
-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

Re: [PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Pavel Machek]

On Sat 2006-07-22 10:14:34, Rusty Russell wrote:
> (forgot to send this to lkml before)
> 
> Abstract sensitive instructions in assembler code, replacing them with
> macros (which currently are #defined to the native versions).  We use
> long names: assembler is case-insensitive, so if something goes wrong
> and macros do not expand, it would assemble anyway.
> 
> Resulting object files are exactly the same as before.
> 
> Signed-off-by Rusty Russell <rusty@rustcorp.com.au>
> 
> Index: working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S
> ===================================================================
> --- working-2.6.18-rc2-hg-paravirt.orig/arch/i386/kernel/entry.S	2006-07-21 21:09:22.000000000 +1000
> +++ working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S	2006-07-22 04:32:25.000000000 +1000
> @@ -76,8 +76,15 @@
>  NT_MASK		= 0x00004000
>  VM_MASK		= 0x00020000
>  
> +/* These are replaces for paravirtualization */
> +#define DISABLE_INTERRUPTS		cli
> +#define ENABLE_INTERRUPTS		sti
> +#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
> +#define INTERRUPT_RETURN		iret

Could we use some less verbose names, like possibly CLI, STI,
STI_SYSEXIT, IRET ?

-- 
(english) http://www.livejournal.com/~pavelmachek
(cesky, pictures) http://atrey.karlin.mff.cuni.cz/~pavel/picture/horses/blog.html

[PATCH 5/6] Begin abstraction of sensitive instructions: asm files

[Rusty Russell]

(forgot to send this to lkml before)

Abstract sensitive instructions in assembler code, replacing them with
macros (which currently are #defined to the native versions).  We use
long names: assembler is case-insensitive, so if something goes wrong
and macros do not expand, it would assemble anyway.

Resulting object files are exactly the same as before.

Signed-off-by Rusty Russell <rusty@rustcorp.com.au>

Index: working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S
===================================================================
--- working-2.6.18-rc2-hg-paravirt.orig/arch/i386/kernel/entry.S	2006-07-21 21:09:22.000000000 +1000
+++ working-2.6.18-rc2-hg-paravirt/arch/i386/kernel/entry.S	2006-07-22 04:32:25.000000000 +1000
@@ -76,8 +76,15 @@
 NT_MASK		= 0x00004000
 VM_MASK		= 0x00020000
 
+/* These are replaces for paravirtualization */
+#define DISABLE_INTERRUPTS		cli
+#define ENABLE_INTERRUPTS		sti
+#define ENABLE_INTERRUPTS_SYSEXIT	sti; sysexit
+#define INTERRUPT_RETURN		iret
+#define GET_CR0			movl %cr0, %eax
+
 #ifdef CONFIG_PREEMPT
-#define preempt_stop		cli; TRACE_IRQS_OFF
+#define preempt_stop		DISABLE_INTERRUPTS; TRACE_IRQS_OFF
 #else
 #define preempt_stop
 #define resume_kernel		restore_nocheck
@@ -234,7 +241,7 @@
 	cmpl $SEGMENT_RPL_MASK, %eax
 	jb resume_kernel		# returning to kernel or vm86-space
 ENTRY(resume_userspace)
- 	cli				# make sure we don't miss an interrupt
+ 	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	movl TI_flags(%ebp), %ecx
@@ -245,7 +252,7 @@
 
 #ifdef CONFIG_PREEMPT
 ENTRY(resume_kernel)
-	cli
+	DISABLE_INTERRUPTS
 	cmpl $0,TI_preempt_count(%ebp)	# non-zero preempt_count ?
 	jnz restore_nocheck
 need_resched:
@@ -273,7 +280,7 @@
 	 * No need to follow this irqs on/off section: the syscall
 	 * disabled irqs and here we enable it straight after entry:
 	 */
-	sti
+	ENABLE_INTERRUPTS
 	pushl $(__USER_DS)
 	CFI_ADJUST_CFA_OFFSET 4
 	/*CFI_REL_OFFSET ss, 0*/
@@ -318,7 +325,7 @@
 	jae syscall_badsys
 	call *sys_call_table(,%eax,4)
 	movl %eax,EAX(%esp)
-	cli
+	DISABLE_INTERRUPTS
 	TRACE_IRQS_OFF
 	movl TI_flags(%ebp), %ecx
 	testw $_TIF_ALLWORK_MASK, %cx
@@ -328,8 +335,7 @@
 	movl OLDESP(%esp), %ecx
 	xorl %ebp,%ebp
 	TRACE_IRQS_ON
-	sti
-	sysexit
+	ENABLE_INTERRUPTS_SYSEXIT
 	CFI_ENDPROC
 

@@ -354,7 +360,7 @@
 	call *sys_call_table(,%eax,4)
 	movl %eax,EAX(%esp)		# store the return value
 syscall_exit:
-	cli				# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -379,11 +385,11 @@
 	RESTORE_REGS
 	addl $4, %esp
 	CFI_ADJUST_CFA_OFFSET -4
-1:	iret
+1:	INTERRUPT_RETURN
 .section .fixup,"ax"
 iret_exc:
 	TRACE_IRQS_ON
-	sti
+	ENABLE_INTERRUPTS
 	pushl $0			# no error code
 	pushl $do_iret_error
 	jmp error_code
@@ -407,7 +413,7 @@
 	 * dosemu and wine happy. */
 	subl $8, %esp		# reserve space for switch16 pointer
 	CFI_ADJUST_CFA_OFFSET 8
-	cli
+	DISABLE_INTERRUPTS
 	TRACE_IRQS_OFF
 	movl %esp, %eax
 	/* Set up the 16bit stack frame with switch32 pointer on top,
@@ -417,7 +423,7 @@
 	TRACE_IRQS_IRET
 	RESTORE_REGS
 	lss 20+4(%esp), %esp	# switch to 16bit stack
-1:	iret
+1:	INTERRUPT_RETURN
 .section __ex_table,"a"
 	.align 4
 	.long 1b,iret_exc
@@ -432,7 +438,7 @@
 	jz work_notifysig
 work_resched:
 	call schedule
-	cli				# make sure we don't miss an interrupt
+	DISABLE_INTERRUPTS		# make sure we don't miss an interrupt
 					# setting need_resched or sigpending
 					# between sampling and the iret
 	TRACE_IRQS_OFF
@@ -488,7 +494,7 @@
 	testb $(_TIF_SYSCALL_TRACE|_TIF_SYSCALL_AUDIT|_TIF_SINGLESTEP), %cl
 	jz work_pending
 	TRACE_IRQS_ON
-	sti				# could let do_syscall_trace() call
+	ENABLE_INTERRUPTS		# could let do_syscall_trace() call
 					# schedule() instead
 	movl %esp, %eax
 	movl $1, %edx
@@ -667,7 +673,7 @@
 	pushl $-1			# mark this as an int
 	CFI_ADJUST_CFA_OFFSET 4
 	SAVE_ALL
-	movl %cr0, %eax
+	GET_CR0
 	testl $0x4, %eax		# EM (math emulation bit)
 	jne device_not_available_emulate
 	preempt_stop
@@ -797,7 +803,7 @@
 	call do_nmi
 	RESTORE_REGS
 	lss 12+4(%esp), %esp		# back to 16bit stack
-1:	iret
+1:	INTERRUPT_RETURN
 	CFI_ENDPROC
 .section __ex_table,"a"
 	.align 4
Index: working-2.6.18-rc2-hg-paravirt/include/asm-i386/spinlock.h
===================================================================
--- working-2.6.18-rc2-hg-paravirt.orig/include/asm-i386/spinlock.h	2006-07-21 20:27:59.000000000 +1000
+++ working-2.6.18-rc2-hg-paravirt/include/asm-i386/spinlock.h	2006-07-22 04:32:51.000000000 +1000
@@ -17,6 +17,9 @@
  * (the type definitions are in asm/spinlock_types.h)
  */
 
+#define CLI_STRING	"cli"
+#define STI_STRING	"sti"
+
 #define __raw_spin_is_locked(x) \
 		(*(volatile signed char *)(&(x)->slock) <= 0)
 
@@ -43,12 +46,12 @@
 	"2:\t" \
 	"testl $0x200, %1\n\t" \
 	"jz 4f\n\t" \
-	"sti\n" \
+	STI_STRING "\n" \
 	"3:\t" \
 	"rep;nop\n\t" \
 	"cmpb $0, %0\n\t" \
 	"jle 3b\n\t" \
-	"cli\n\t" \
+	CLI_STRING "\n\t" \
 	"jmp 1b\n" \
 	"4:\t" \
 	"rep;nop\n\t" \

-- 
Help! Save Australia from the worst of the DMCA: http://linux.org.au/law