Patchwork rseq/s390: choosing code signature

login
register
mail settings
Submitter Martin Schwidefsky
Date April 10, 2019, 10:32 a.m.
Message ID <20190410123258.37f182cf@mschwideX1>
Download mbox | patch
Permalink /patch/32245/
State New
Headers show

Comments

Martin Schwidefsky - April 10, 2019, 10:32 a.m.
On Tue, 9 Apr 2019 15:32:22 -0400 (EDT)
Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:

> Hi,
> 
> We are about to include the code signature required prior to restartable
> sequences abort handlers into glibc, which will make this ABI choice final.
> We need architecture maintainer input on that signature value.
> 
> That code signature is placed before each abort handler, so the kernel can
> validate that it is indeed jumping to an abort handler (and not some
> arbitrary attacker-chosen code). The signature is never executed.
> 
> The current discussion thread on the glibc mailing list leads us towards
> using a trap with uncommon immediate operand, which simplifies integration
> with disassemblers, emulators, makes it easier to debug if the control
> flow gets redirected there by mistake, and is nicer for some architecture's
> speculative execution.
> 
> We can have different signatures for each sub-architecture, as long as they
> don't have to co-exist within the same process. We can special-case with
> #ifdef for each sub-architecture and endianness if need be. If the architecture
> has instruction set extensions that can co-exist with the architecture
> instruction set within the same process, we need to take into account to which
> instruction the chosen signature value would map (and possibly decide if we
> need to extend rseq to support many signatures).
> 
> Here is an example of rseq signature definition template:
> 
> /*
>  * TODO: document trap instruction objdump output on each sub-architecture
>  * instruction sets, as well as instruction set extensions.
>  */
> #define RSEQ_SIG 0x########
> 
> Ideally we'd need a patch on top of the Linux kernel
> tools/testing/selftests/rseq/rseq-s390.h file that updates
> the signature value, so I can then pick it up for the glibc
> patchset.

The trap4 instruction is a suitable one. The patch would look like this
--
commit 2ee28f6d1de968a71f074ab150384b90b4121216
Author: Martin Schwidefsky <schwidefsky@de.ibm.com>
Date:   Wed Apr 10 12:28:41 2019 +0200

    s390/rseq: use trap4 for RSEQ_SIG
    
    Use trap4 as the guard instruction for the restartable sequence abort
    handler.
    
    Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Mathieu Desnoyers - April 10, 2019, 3:50 p.m.
----- On Apr 10, 2019, at 6:32 AM, schwidefsky schwidefsky@de.ibm.com wrote:

> On Tue, 9 Apr 2019 15:32:22 -0400 (EDT)
> Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> 
>> Hi,
>> 
>> We are about to include the code signature required prior to restartable
>> sequences abort handlers into glibc, which will make this ABI choice final.
>> We need architecture maintainer input on that signature value.
>> 
>> That code signature is placed before each abort handler, so the kernel can
>> validate that it is indeed jumping to an abort handler (and not some
>> arbitrary attacker-chosen code). The signature is never executed.
>> 
>> The current discussion thread on the glibc mailing list leads us towards
>> using a trap with uncommon immediate operand, which simplifies integration
>> with disassemblers, emulators, makes it easier to debug if the control
>> flow gets redirected there by mistake, and is nicer for some architecture's
>> speculative execution.
>> 
>> We can have different signatures for each sub-architecture, as long as they
>> don't have to co-exist within the same process. We can special-case with
>> #ifdef for each sub-architecture and endianness if need be. If the architecture
>> has instruction set extensions that can co-exist with the architecture
>> instruction set within the same process, we need to take into account to which
>> instruction the chosen signature value would map (and possibly decide if we
>> need to extend rseq to support many signatures).
>> 
>> Here is an example of rseq signature definition template:
>> 
>> /*
>>  * TODO: document trap instruction objdump output on each sub-architecture
>>  * instruction sets, as well as instruction set extensions.
>>  */
>> #define RSEQ_SIG 0x########
>> 
>> Ideally we'd need a patch on top of the Linux kernel
>> tools/testing/selftests/rseq/rseq-s390.h file that updates
>> the signature value, so I can then pick it up for the glibc
>> patchset.
> 
> The trap4 instruction is a suitable one. The patch would look like this

Great! I'm picking it up into my rseq tree if that's OK with you.

Thanks,

Mathieu


> --
> commit 2ee28f6d1de968a71f074ab150384b90b4121216
> Author: Martin Schwidefsky <schwidefsky@de.ibm.com>
> Date:   Wed Apr 10 12:28:41 2019 +0200
> 
>    s390/rseq: use trap4 for RSEQ_SIG
>    
>    Use trap4 as the guard instruction for the restartable sequence abort
>    handler.
>    
>    Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
> 
> diff --git a/tools/testing/selftests/rseq/rseq-s390.h
> b/tools/testing/selftests/rseq/rseq-s390.h
> index 1069e85258ce..d4c8e1147d86 100644
> --- a/tools/testing/selftests/rseq/rseq-s390.h
> +++ b/tools/testing/selftests/rseq/rseq-s390.h
> @@ -1,6 +1,13 @@
> /* SPDX-License-Identifier: LGPL-2.1 OR MIT */
> 
> -#define RSEQ_SIG	0x53053053
> +/*
> + * RSEQ_SIG uses the trap4 instruction. As Linux does not make use of the
> + * access-register mode nor the linkage stack this instruction will always
> + * cause a special-operation exception (the trap-enabled bit in the DUCT
> + * is and will stay 0). The instruction pattern is
> + *	b2 ff 0f ff	trap4   4095(%r0)
> + */
> +#define RSEQ_SIG	0xB2FF0FFF
> 
> #define rseq_smp_mb()	__asm__ __volatile__ ("bcr 15,0" ::: "memory")
> #define rseq_smp_rmb()	rseq_smp_mb()
> --
> blue skies,
>   Martin.
> 
> "Reality continues to ruin my life." - Calvin.
Martin Schwidefsky - April 10, 2019, 3:52 p.m.
On Wed, 10 Apr 2019 11:50:39 -0400 (EDT)
Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:

> ----- On Apr 10, 2019, at 6:32 AM, schwidefsky schwidefsky@de.ibm.com wrote:
> 
> > On Tue, 9 Apr 2019 15:32:22 -0400 (EDT)
> > Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> >   
> >> Hi,
> >> 
> >> We are about to include the code signature required prior to restartable
> >> sequences abort handlers into glibc, which will make this ABI choice final.
> >> We need architecture maintainer input on that signature value.
> >> 
> >> That code signature is placed before each abort handler, so the kernel can
> >> validate that it is indeed jumping to an abort handler (and not some
> >> arbitrary attacker-chosen code). The signature is never executed.
> >> 
> >> The current discussion thread on the glibc mailing list leads us towards
> >> using a trap with uncommon immediate operand, which simplifies integration
> >> with disassemblers, emulators, makes it easier to debug if the control
> >> flow gets redirected there by mistake, and is nicer for some architecture's
> >> speculative execution.
> >> 
> >> We can have different signatures for each sub-architecture, as long as they
> >> don't have to co-exist within the same process. We can special-case with
> >> #ifdef for each sub-architecture and endianness if need be. If the architecture
> >> has instruction set extensions that can co-exist with the architecture
> >> instruction set within the same process, we need to take into account to which
> >> instruction the chosen signature value would map (and possibly decide if we
> >> need to extend rseq to support many signatures).
> >> 
> >> Here is an example of rseq signature definition template:
> >> 
> >> /*
> >>  * TODO: document trap instruction objdump output on each sub-architecture
> >>  * instruction sets, as well as instruction set extensions.
> >>  */
> >> #define RSEQ_SIG 0x########
> >> 
> >> Ideally we'd need a patch on top of the Linux kernel
> >> tools/testing/selftests/rseq/rseq-s390.h file that updates
> >> the signature value, so I can then pick it up for the glibc
> >> patchset.  
> > 
> > The trap4 instruction is a suitable one. The patch would look like this  
> 
> Great! I'm picking it up into my rseq tree if that's OK with you.

Just added the patch to s390/linux:features for the next merge window as well.
Mathieu Desnoyers - April 10, 2019, 3:57 p.m.
----- On Apr 10, 2019, at 11:52 AM, schwidefsky schwidefsky@de.ibm.com wrote:

> On Wed, 10 Apr 2019 11:50:39 -0400 (EDT)
> Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> 
>> ----- On Apr 10, 2019, at 6:32 AM, schwidefsky schwidefsky@de.ibm.com wrote:
>> 
>> > On Tue, 9 Apr 2019 15:32:22 -0400 (EDT)
>> > Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
>> >   
>> >> Hi,
>> >> 
>> >> We are about to include the code signature required prior to restartable
>> >> sequences abort handlers into glibc, which will make this ABI choice final.
>> >> We need architecture maintainer input on that signature value.
>> >> 
>> >> That code signature is placed before each abort handler, so the kernel can
>> >> validate that it is indeed jumping to an abort handler (and not some
>> >> arbitrary attacker-chosen code). The signature is never executed.
>> >> 
>> >> The current discussion thread on the glibc mailing list leads us towards
>> >> using a trap with uncommon immediate operand, which simplifies integration
>> >> with disassemblers, emulators, makes it easier to debug if the control
>> >> flow gets redirected there by mistake, and is nicer for some architecture's
>> >> speculative execution.
>> >> 
>> >> We can have different signatures for each sub-architecture, as long as they
>> >> don't have to co-exist within the same process. We can special-case with
>> >> #ifdef for each sub-architecture and endianness if need be. If the architecture
>> >> has instruction set extensions that can co-exist with the architecture
>> >> instruction set within the same process, we need to take into account to which
>> >> instruction the chosen signature value would map (and possibly decide if we
>> >> need to extend rseq to support many signatures).
>> >> 
>> >> Here is an example of rseq signature definition template:
>> >> 
>> >> /*
>> >>  * TODO: document trap instruction objdump output on each sub-architecture
>> >>  * instruction sets, as well as instruction set extensions.
>> >>  */
>> >> #define RSEQ_SIG 0x########
>> >> 
>> >> Ideally we'd need a patch on top of the Linux kernel
>> >> tools/testing/selftests/rseq/rseq-s390.h file that updates
>> >> the signature value, so I can then pick it up for the glibc
>> >> patchset.
>> > 
>> > The trap4 instruction is a suitable one. The patch would look like this
>> 
>> Great! I'm picking it up into my rseq tree if that's OK with you.
> 
> Just added the patch to s390/linux:features for the next merge window as well.

Sounds good! I'll carry it in my tree to have a comprehensive up-to-date list of
rseq signatures for all architectures in a single tree. Worse-case the exact same
change will be pulled from both architecture and rseq trees, which I don't think
should be an issue, right ?

Thanks,

Mathieu

> 
> --
> blue skies,
>   Martin.
> 
> "Reality continues to ruin my life." - Calvin.
Martin Schwidefsky - April 10, 2019, 4:04 p.m.
On Wed, 10 Apr 2019 11:57:36 -0400 (EDT)
Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:

> ----- On Apr 10, 2019, at 11:52 AM, schwidefsky schwidefsky@de.ibm.com wrote:
> 
> > On Wed, 10 Apr 2019 11:50:39 -0400 (EDT)
> > Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> >   
> >> ----- On Apr 10, 2019, at 6:32 AM, schwidefsky schwidefsky@de.ibm.com wrote:
> >>   
> >> > On Tue, 9 Apr 2019 15:32:22 -0400 (EDT)
> >> > Mathieu Desnoyers <mathieu.desnoyers@efficios.com> wrote:
> >> >     
> >> >> Hi,
> >> >> 
> >> >> We are about to include the code signature required prior to restartable
> >> >> sequences abort handlers into glibc, which will make this ABI choice final.
> >> >> We need architecture maintainer input on that signature value.
> >> >> 
> >> >> That code signature is placed before each abort handler, so the kernel can
> >> >> validate that it is indeed jumping to an abort handler (and not some
> >> >> arbitrary attacker-chosen code). The signature is never executed.
> >> >> 
> >> >> The current discussion thread on the glibc mailing list leads us towards
> >> >> using a trap with uncommon immediate operand, which simplifies integration
> >> >> with disassemblers, emulators, makes it easier to debug if the control
> >> >> flow gets redirected there by mistake, and is nicer for some architecture's
> >> >> speculative execution.
> >> >> 
> >> >> We can have different signatures for each sub-architecture, as long as they
> >> >> don't have to co-exist within the same process. We can special-case with
> >> >> #ifdef for each sub-architecture and endianness if need be. If the architecture
> >> >> has instruction set extensions that can co-exist with the architecture
> >> >> instruction set within the same process, we need to take into account to which
> >> >> instruction the chosen signature value would map (and possibly decide if we
> >> >> need to extend rseq to support many signatures).
> >> >> 
> >> >> Here is an example of rseq signature definition template:
> >> >> 
> >> >> /*
> >> >>  * TODO: document trap instruction objdump output on each sub-architecture
> >> >>  * instruction sets, as well as instruction set extensions.
> >> >>  */
> >> >> #define RSEQ_SIG 0x########
> >> >> 
> >> >> Ideally we'd need a patch on top of the Linux kernel
> >> >> tools/testing/selftests/rseq/rseq-s390.h file that updates
> >> >> the signature value, so I can then pick it up for the glibc
> >> >> patchset.  
> >> > 
> >> > The trap4 instruction is a suitable one. The patch would look like this  
> >> 
> >> Great! I'm picking it up into my rseq tree if that's OK with you.  
> > 
> > Just added the patch to s390/linux:features for the next merge window as well.  
> 
> Sounds good! I'll carry it in my tree to have a comprehensive up-to-date list of
> rseq signatures for all architectures in a single tree. Worse-case the exact same
> change will be pulled from both architecture and rseq trees, which I don't think
> should be an issue, right ?

Should be fine, the worst that can happen is a minor merge conflict.

Patch

diff --git a/tools/testing/selftests/rseq/rseq-s390.h b/tools/testing/selftests/rseq/rseq-s390.h
index 1069e85258ce..d4c8e1147d86 100644
--- a/tools/testing/selftests/rseq/rseq-s390.h
+++ b/tools/testing/selftests/rseq/rseq-s390.h
@@ -1,6 +1,13 @@ 
 /* SPDX-License-Identifier: LGPL-2.1 OR MIT */
 
-#define RSEQ_SIG	0x53053053
+/*
+ * RSEQ_SIG uses the trap4 instruction. As Linux does not make use of the
+ * access-register mode nor the linkage stack this instruction will always
+ * cause a special-operation exception (the trap-enabled bit in the DUCT
+ * is and will stay 0). The instruction pattern is
+ *	b2 ff 0f ff	trap4   4095(%r0)
+ */
+#define RSEQ_SIG	0xB2FF0FFF
 
 #define rseq_smp_mb()	__asm__ __volatile__ ("bcr 15,0" ::: "memory")
 #define rseq_smp_rmb()	rseq_smp_mb()