[Qemu-devel] Missing ARMv6 instructions?

[Qemu-devel] Missing ARMv6 instructions?

[Wolfgang Schildbach]

Hello list,

Running an ARM application in user mode emulation (qemu CVS version of 
3-15-2006), my code crashes at an SMMUL instruction (this is part of the 
ARMv6 instruction set). A brief glance at translate.c and op.c seems to 
suggest that qemu does not emulate that instruction (yet). Before I dig 
any deeper, could somebody in the know (Paul?) confirm or reject this 
suspicion?

Thanks much,
--
Wolfgang Schildbach, Senior Research Engineer
Coding Technologies GmbH

Re: [Qemu-devel] Missing ARMv6 instructions?

[Paul Brook]

On Wednesday 29 March 2006 13:33, Wolfgang Schildbach wrote:
> Hello list,
>
> Running an ARM application in user mode emulation (qemu CVS version of
> 3-15-2006), my code crashes at an SMMUL instruction (this is part of the
> ARMv6 instruction set). A brief glance at translate.c and op.c seems to
> suggest that qemu does not emulate that instruction (yet). Before I dig
> any deeper, could somebody in the know (Paul?) confirm or reject this
> suspicion?

qemu cvs only supports ARMv5TE.

The ARMv6 architecture is released under a more restrictive licence than 
ARMv5.  The Arm licencing department have explicitly prohibited the 
distribution of open source ARMv6/v7 emulators.

We're trying to get this restriction lifted, but so far to no avail.

Paul

Re: [Qemu-devel] Missing ARMv6 instructions?

[Wolfgang Schildbach]

Thanks, Paul. That explains it...

I find it strange that ARM would restrict emulation of their architecture 
-- that could hardly pose a threat to their business, I would say.

Anyhow, thanks for the note.

- Wolfgang

Paul Brook <paul@codesourcery.com> wrote on 29.03.2006 16:39:12:

> On Wednesday 29 March 2006 13:33, Wolfgang Schildbach wrote:
> > Hello list,
> >
> > Running an ARM application in user mode emulation (qemu CVS version of
> > 3-15-2006), my code crashes at an SMMUL instruction (this is part of 
the
> > ARMv6 instruction set). A brief glance at translate.c and op.c seems 
to
> > suggest that qemu does not emulate that instruction (yet). Before I 
dig
> > any deeper, could somebody in the know (Paul?) confirm or reject this
> > suspicion?
> 
> qemu cvs only supports ARMv5TE.
> 
> The ARMv6 architecture is released under a more restrictive licence than 

> ARMv5.  The Arm licencing department have explicitly prohibited the 
> distribution of open source ARMv6/v7 emulators.
> 
> We're trying to get this restriction lifted, but so far to no avail.
> 
> Paul

--
Wolfgang Schildbach, Senior Research Engineer
Coding Technologies GmbH

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jamie Lokier]

Paul Brook <paul@codesourcery.com> wrote:
>> qemu cvs only supports ARMv5TE.
>> 
>> The ARMv6 architecture is released under a more restrictive
>> licence than ARMv5.  The Arm licencing department have explicitly
>> prohibited the distribution of open source ARMv6/v7 emulators.
>>
>> > We're trying to get this restriction lifted, but so far to no avail.

Wolfgang Schildbach wrote:
> Thanks, Paul. That explains it...
> I find it strange that ARM would restrict emulation of their architecture 
> -- that could hardly pose a threat to their business, I would say.

I find it strange that they can restrict emulation, if they can.

Perhaps they can't, and maybe they like to send threatening letters to
frighten people away from doing it?

That's not so surprising for ARM.  They halted the development of an
open source ARM-compatible FPGA design too.

What claim does ARM have over software that they don't write and whose
only resemblance is that it speaks the same protocol?

It looks, and smells, like an attempt at "interface copyright".
I.e. to copyright something just for offering a similar interface.

E.g. much the same as Microsoft attempting to stop people implementing
a "task bar".  Yet, Microsoft can't do that, and nobody is afraid of
writing window managers with task bars.

I could understand a claim if someone acquired ARM's documentation
under an agreement to not produce an emulator.  But I don't see how
someone who has no connection with ARM can be restricted by ARM in
that way.

Has ARM ever actually followed through on their threats and _won_ a
case to ban compatible software from being distributed?

And if there is a possibility of that - in which countries do they
have any weight?  Dare I suggest encouraging the development of
patches they don't like to countries where they have no legal weight?

-- Jamie

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jamie Lokier]

Paul Brook <paul@codesourcery.com> wrote:
> The ARMv6 architecture is released under a more restrictive licence than 
> ARMv5.  The Arm licencing department have explicitly prohibited the 
> distribution of open source ARMv6/v7 emulators.
> 
> We're trying to get this restriction lifted, but so far to no avail.

One more thing: how about adding the above to the QEMU FAQ?

-- Jamie

Re: [Qemu-devel] Missing ARMv6 instructions?

[Paul Brook]

> I could understand a claim if someone acquired ARM's documentation
> under an agreement to not produce an emulator.  

That's exactly where the restriction comes from.

Theoretically it may be possible to reverse engineer a good proportion of 
ARMv6 from other sources (eg. gcc). However if that were done it would mean 
anyone with access to the real ARM documentation (i.e. me and probably anyone 
else with any professional/commercial interest in ARM emulation) would be 
unable to contribute to qemu.

> And if there is a possibility of that - in which countries do they
> have any weight?  Dare I suggest encouraging the development of
> patches they don't like to countries where they have no legal weight?

I don't think that's particularly helpful or practical suggestion.
Better would be to lobby ARM to allow open source emulators. 
"I'd like to use ARM hardware for <big project>, but qemu doesn't support 
ARMv7 so I'm thinking of using PowerPC instead" is a particularly good 
argument ;-)

Paul

Re: [Qemu-devel] Missing ARMv6 instructions?

[John Hogerhuis]

[-- Attachment #1: Type: text/plain, Size: 420 bytes --]

So it's a purely contractual issue as opposed to an IP issue.

Is a party to this contract allowed to write a disassembler? I can imagine a
very nice disassembler feature that would explain in detail how each
instruction it decodes works...

Of course no matter how such documentation were to escape, you're still used
up by your contractual obligations as far as implementing an emulator, I
guess.

-- John.

[-- Attachment #2: Type: text/html, Size: 438 bytes --]

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jamie Lokier]

Paul Brook wrote:
> Better would be to lobby ARM to allow open source emulators. 
> "I'd like to use ARM hardware for <big project>, but qemu doesn't support 
> ARMv7 so I'm thinking of using PowerPC instead" is a particularly good 
> argument ;-)

I suspect ARM's business model - dependent wholly on licensing the ARM
architecture in various forms - would make it very difficult for them
to agree to that argument even if it would gain them one customer.

It'll be interesting when someone develops a compiler that uses very
detailed machine descriptions - with enough detail to describe an
architecture well enough that the description can also be used (by a
different program) to emulate it to a significant degree.

They need good compilers to support the architrecture.  But a very
detailed machine description is also their business nightmare: the
basis for software, and later hardware, implementations of their
architecture.  I wonder if people such as yourself will be able to
contribute to the development of the machine description for such a
compiler.

-- Jamie

Re: [Qemu-devel] Missing ARMv6 instructions?

[Wolfgang Schildbach]

The way I understand this is that anyone who got ARM documentation with 
the license that Paul mentioned, could not contribute patches that 
implement v6 emulation. If, however, someone else (who has not signed such 
a license) were to buy documentation about the ARMv6 architecture that 
comes without a license (for example, the "ARM System Developer's Guide", 
by Sloss, Symes, and Wright, Elsevier 2004), that someone should be able 
to contribute (at least in those aspects of the ARM that are disclosed by 
the book). In that case, ARM should not have any legal hooks except 
copyright, and since at that point you are not using any of their code, 
there is no case. However, I am not a lawyer, so don't take my word for 
it.

As to the threats to ARM's business model, I could see them threatened if 
someone emulated the ARM on hardware of similar performance (without 
licensing ARM IP) -- that would undermine sales of the real ARM cores, 
depending on the price/performance point of the replacement. Software 
emulation of an ARM core on a 4 GHz Pentium is hardly an alternative to 
buying the real thing, though... I also find it hard to believe that 
building a product competitive to an ARM is possible using a description 
of the instruction set alone.

That said, I agree with Paul that lobbying ARM to change their license is 
probably a better route to go. I believe software emulators like qemu are 
really in ARMs best interest since they support ARM development and 
thereby increase the availability of software for ARM cores.

- Wolfgang

Disclaimer: The views stated above are entirely mine, and do not 
necessarily reflect the views of my employer.

Paul wrote on 30.03.2006 00:01:41:

> > I could understand a claim if someone acquired ARM's documentation
> > under an agreement to not produce an emulator. 
> 
> That's exactly where the restriction comes from.
> 
> Theoretically it may be possible to reverse engineer a good proportion 
of 
> ARMv6 from other sources (eg. gcc). However if that were done it would 
mean 
> anyone with access to the real ARM documentation (i.e. me and probably 
anyone 
> else with any professional/commercial interest in ARM emulation) would 
be 
> unable to contribute to qemu.
> 
> > And if there is a possibility of that - in which countries do they
> > have any weight?  Dare I suggest encouraging the development of
> > patches they don't like to countries where they have no legal weight?
> 
> I don't think that's particularly helpful or practical suggestion.
> Better would be to lobby ARM to allow open source emulators. 
> "I'd like to use ARM hardware for <big project>, but qemu doesn't 
support 
> ARMv7 so I'm thinking of using PowerPC instead" is a particularly good 
> argument ;-)
> 
> Paul
> 
> 
> _______________________________________________
> Qemu-devel mailing list
> Qemu-devel@nongnu.org
> http://lists.nongnu.org/mailman/listinfo/qemu-devel

--
Wolfgang Schildbach, Senior Research Engineer
Coding Technologies GmbH

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jamie Lokier]

Wolfgang Schildbach wrote:
> The way I understand this is that anyone who got ARM documentation with 
> the license that Paul mentioned, could not contribute patches that 
> implement v6 emulation. If, however, someone else (who has not signed such 
> a license) were to buy documentation about the ARMv6 architecture that 
> comes without a license (for example, the "ARM System Developer's Guide", 
> by Sloss, Symes, and Wright, Elsevier 2004), that someone should be able 
> to contribute (at least in those aspects of the ARM that are disclosed by 
> the book). In that case, ARM should not have any legal hooks except 
> copyright, and since at that point you are not using any of their code, 
> there is no case. However, I am not a lawyer, so don't take my word for 
> it.

I wonder about patents (and their validity).

A MIPS hardware implementation that I worked with had all the basic
MIPS integer instructions except one small group, on the ground that
those instructions were covered by a MIPS patent and so they dared not
include it.  It was the instruction used for unaligned memory word
access.  For that reason, we used a modified GCC specially for that
chip.

> I also find it hard to believe that building a product competitive
> to an ARM is possible using a description of the instruction set alone.

(I meant a description of the architectural behaviour... not just the
instruction set.  Compilers are moving beyond merely implementing
"user space" code, to optimising paths between security levels,
analysing memory orderings between threads, etc.  And for that matter,
so are decompilers.)

It's not possible.. yet.  Believe me, there are people working on it.

And it doesn't have to be competitive for ARM to have a reaction... as
we're seeing with software emulatation.

> That said, I agree with Paul that lobbying ARM to change their license is 
> probably a better route to go. I believe software emulators like qemu are 
> really in ARMs best interest since they support ARM development and 
> thereby increase the availability of software for ARM cores.

That makes sense to me.

-- Jamie

Re: [Qemu-devel] Missing ARMv6 instructions?

[Lennert Buytenhek]

On Thu, Mar 30, 2006 at 03:30:30PM +0100, Jamie Lokier wrote:

> I wonder about patents (and their validity).
> 
> A MIPS hardware implementation that I worked with had all the basic
> MIPS integer instructions except one small group, on the ground that
> those instructions were covered by a MIPS patent and so they dared not
> include it.  It was the instruction used for unaligned memory word
> access.  For that reason, we used a modified GCC specially for that
> chip.

Note that the new ARM ABI being pushed by ARM ltd, which is called
EABI, mandates using a thumb-related instruction (bx) in the function
exit path.

In other words, if the arm-compatible core that you wrote for your
FPGA board to play with does not implement that instruction (possibly
because you want to avoid infringing on ARM patents), you won't be
able to run any EABI-using ARM distros on it.  


--L

Re: [Qemu-devel] Missing ARMv6 instructions?

[Paul Brook]

> Note that the new ARM ABI being pushed by ARM ltd, which is called
> EABI, mandates using a thumb-related instruction (bx) in the function
> exit path.

No it doesn't. The EABI mandates that code be interworking safe.
It's possible to implement this is an way that only uses bx when actually 
returning to Thumb code, ie when you already know your hardware supports bx.

In practice many distros may assume Thumb capable hardware in the same way 
many x86 distros assume at least i486 or i686, but that's an entirely 
different issue.

Paul

Re: [Qemu-devel] Missing ARMv6 instructions?

[Chris Wilson]

Hi all,

On Wed, 2006-03-29 at 18:00 +0200, Wolfgang Schildbach wrote:

> I find it strange that ARM would restrict emulation of their architecture 
> -- that could hardly pose a threat to their business, I would say.

Unfortunately, I don't. ARM probably makes quite a lot of money from
their development tools (compilers, debuggers, simulators) as well as
actual chips, and they have been an extensive user of software patents,
even in the EU where they are probably not valid.

How about we target OpenCores instead of ARM?

Cheers, Chris.
-- 
  ___ __     _
 / __/ / ,__(_)_  | Chris Wilson <0000 at qwirx.com> - Cambs UK |
/ (_/ ,\/ _/ /_ \ | Security/C/C++/Java/Perl/SQL/HTML Developer |
\ _/_/_/_//_/___/ | We are GNU-free your mind-and your software |

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jonas Maebe]

On 01 Apr 2006, at 22:51, Chris Wilson wrote:

> and they have been an extensive user of software patents,

And how:
   http://www.patent.gov.uk/patent/legal/summaries/2004/o29204.htm

"The invention in this case involves locating all of the input  
registers in one data storage area, and all of the output registers  
in another data storage area. Then the simulator only has to switch  
the two storage areas around (eg. by exchanging two pointer values)  
in order to effectively COPY all of the output registers to the  
corresponding input registers of the next stage in a single  
operation. The Hearing Officer concluded that this invention did  
involve a technical contribution — not simply because it produced a  
faster simulator, but because the fundamental construction of the  
simulator had been modified."


So yes, using

struct registers *a, *b;
a = b;

instead of

struct registers a, b;
memcpy(&a,&b,sizeof(b));

has been patented by ARM in the UK. In fact, wouldn't surprise me if  
Qemu violates this patent.


Jonas

Re: [Qemu-devel] Missing ARMv6 instructions?

[Kevin F. Quinn]

[-- Attachment #1: Type: text/plain, Size: 2495 bytes --]

On Sat, 1 Apr 2006 23:06:07 +0300
Jonas Maebe <jonas.maebe@elis.ugent.be> wrote:

> On 01 Apr 2006, at 22:51, Chris Wilson wrote:
> 
> > and they have been an extensive user of software patents,
> 
> And how:
>    http://www.patent.gov.uk/patent/legal/summaries/2004/o29204.htm
> 
> "The invention in this case involves locating all of the input  
> registers in one data storage area, and all of the output registers  
> in another data storage area. Then the simulator only has to switch  
> the two storage areas around (eg. by exchanging two pointer values)  
> in order to effectively COPY all of the output registers to the  
> corresponding input registers of the next stage in a single  
> operation. The Hearing Officer concluded that this invention did  
> involve a technical contribution — not simply because it produced a  
> faster simulator, but because the fundamental construction of the  
> simulator had been modified."

I've got a better idea.  In an imaginary 3-stage pipeline, instead of
doing:

i1[*] -> <stage1> -> o1[*]
i2[*] -> <stage2> -> o2[*]
i3[*] -> <stage3> -> o3[*]
i2[*] = o1[*]
i3[*] = o2[*]
<commit o3[*]>

which the patent optimises to:

i1[*] -> <stage1> -> o1[*]
i2[*] -> <stage2> -> o2[*]
i3[*] -> <stage3> -> o3[*]
i2,o1 = o1,i2
i3,o2 = o2,i3
<commit o3[*]>

how about this:

o2i3[*] -> <stage3> -> o3[*]
o1i2[*] -> <stage2> -> o2i3[*]
i1[*] -> <stage1> -> o1i2[*]
<commit o3[*]>

which saves mucking around with pointers completely (requires the
stages are implemented sequentially not in pararllel, but since we're
talking about a software simulation that's likely to be the case).

Doesn't break the patent (well, the summary at least) and is a fraction
quicker :)

To be honest both optimisations are clearly obvious to anyone
sufficiently skilled in the field - maybe not so much to hardware
engineers who tend to think in parallel but certainly to software
engineers who tend to think in series.

> In fact, wouldn't surprise me if Qemu violates this patent.

Qemu doesn't simulate the pipeline as such as it emulates each
instruction completely before starting the next one.  Makes things much
simpler - if you don't simulate the pipeline, you don't have to
simulate flushing it etc :). The patent is talking about much deeper
simulation, probably for use in simulating the core in relation to
other components, or for use in accurate timing simulations.

-- 
Kevin F. Quinn

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 191 bytes --]

Re: [Qemu-devel] Missing ARMv6 instructions?

[Jamie Lokier]

Chris Wilson wrote:
> > I find it strange that ARM would restrict emulation of their architecture 
> > -- that could hardly pose a threat to their business, I would say.
> 
> Unfortunately, I don't. ARM probably makes quite a lot of money from
> their development tools (compilers, debuggers, simulators) as well as
> actual chips, and they have been an extensive user of software patents,
> even in the EU where they are probably not valid.
> 
> How about we target OpenCores instead of ARM?

I like the idea, but do you know of anyone using OpenCores devices
implemented in silicon?  It seems to me the motivation for ARM
emulation is to be able to simulate embedded devices that people may
feasibly end up using.

-- Jamie

Re: [Qemu-devel] Missing ARMv6 instructions?

[Chris Wilson]

Hi Jamie,

> I like the idea, but do you know of anyone using OpenCores devices
> implemented in silicon?  It seems to me the motivation for ARM
> emulation is to be able to simulate embedded devices that people may
> feasibly end up using.

I'm no expert, but it appears that OpenCores have a working core that
runs most of the MIPS instruction set. MIPS is a very well known, tried
and trusted architecture. My cable modem has a MIPS-compatible processor
made by Toshiba. It seems to me that MIPS is just as realistic and
usable platform as ARM. But I would be very interested to hear from
anyone who knows better.

Now, I wonder when Qemu will support MIPS emulation? :-)

Cheers, Chris.
-- 
  ___ __     _
 / __/ / ,__(_)_  | Chris Wilson <0000 at qwirx.com> - Cambs UK |
/ (_/ ,\/ _/ /_ \ | Security/C/C++/Java/Perl/SQL/HTML Developer |
\ _/_/_/_//_/___/ | We are GNU-free your mind-and your software |

Re: [Qemu-devel] Missing ARMv6 instructions?

[Karel Gardas]

On Sun, 2 Apr 2006, Jamie Lokier wrote:

> Chris Wilson wrote:
>>> I find it strange that ARM would restrict emulation of their architecture
>>> -- that could hardly pose a threat to their business, I would say.
>>
>> Unfortunately, I don't. ARM probably makes quite a lot of money from
>> their development tools (compilers, debuggers, simulators) as well as
>> actual chips, and they have been an extensive user of software patents,
>> even in the EU where they are probably not valid.
>>
>> How about we target OpenCores instead of ARM?
>
> I like the idea, but do you know of anyone using OpenCores devices
> implemented in silicon?  It seems to me the motivation for ARM
> emulation is to be able to simulate embedded devices that people may
> feasibly end up using.

IMHO majority of embedded devices is still ARM7TDMI based and this 
instruction set (ARMv4 IIRC) should be emulated well in Qemu. Also please 
have a look at www.skyeye.org project to see how other projects do 
emulation for actual embedded devices.

Cheers,
Karel
--
Karel Gardas                  kgardas@objectsecurity.com
ObjectSecurity Ltd.           http://www.objectsecurity.com

Re: [Qemu-devel] Missing ARMv6 instructions?

[M. Warner Losh]

In message: <Pine.LNX.4.63.0604030840070.419@silence.gardas.net>
            Karel Gardas <kgardas@objectsecurity.com> writes:
: On Sun, 2 Apr 2006, Jamie Lokier wrote:
: 
: > Chris Wilson wrote:
: >>> I find it strange that ARM would restrict emulation of their architecture
: >>> -- that could hardly pose a threat to their business, I would say.
: >>
: >> Unfortunately, I don't. ARM probably makes quite a lot of money from
: >> their development tools (compilers, debuggers, simulators) as well as
: >> actual chips, and they have been an extensive user of software patents,
: >> even in the EU where they are probably not valid.
: >>
: >> How about we target OpenCores instead of ARM?
: >
: > I like the idea, but do you know of anyone using OpenCores devices
: > implemented in silicon?  It seems to me the motivation for ARM
: > emulation is to be able to simulate embedded devices that people may
: > feasibly end up using.
: 
: IMHO majority of embedded devices is still ARM7TDMI based and this 
: instruction set (ARMv4 IIRC) should be emulated well in Qemu. Also please 
: have a look at www.skyeye.org project to see how other projects do 
: emulation for actual embedded devices.

You mean badly emulate the embedded devices, and only few of them to
boot?

Warner

Re: [Qemu-devel] Missing ARMv6 instructions?

[Paul Brook]

> IMHO majority of embedded devices is still ARM7TDMI based and this
> instruction set (ARMv4 IIRC) should be emulated well in Qemu.

Depends which market segments you're aiming at.

If you're aiming at well established legacy or cheap, low-end hardware then 
yes armv4t is fine.

If you want to use qemu for development/prototyping of new products then you 
need an emulator that knows about modern systems, not something two 
generations out of date.

Paul

Re: [Qemu-devel] Missing ARMv6 instructions?

[Daniel Jacobowitz]

On Sun, Apr 02, 2006 at 04:42:42PM +0000, Chris Wilson wrote:
> Hi Jamie,
> 
> > I like the idea, but do you know of anyone using OpenCores devices
> > implemented in silicon?  It seems to me the motivation for ARM
> > emulation is to be able to simulate embedded devices that people may
> > feasibly end up using.
> 
> I'm no expert, but it appears that OpenCores have a working core that
> runs most of the MIPS instruction set. MIPS is a very well known, tried
> and trusted architecture. My cable modem has a MIPS-compatible processor
> made by Toshiba. It seems to me that MIPS is just as realistic and
> usable platform as ARM. But I would be very interested to hear from
> anyone who knows better.
> 
> Now, I wonder when Qemu will support MIPS emulation? :-)

Good question.  How about... last year?

-- 
Daniel Jacobowitz
CodeSourcery

Re: [Qemu-devel] Missing ARMv6 instructions?

[Thiemo Seufer]

On Sun, Apr 02, 2006 at 04:42:42PM +0000, Chris Wilson wrote:
> Hi Jamie,
> 
> > I like the idea, but do you know of anyone using OpenCores devices
> > implemented in silicon?  It seems to me the motivation for ARM
> > emulation is to be able to simulate embedded devices that people may
> > feasibly end up using.
> 
> I'm no expert, but it appears that OpenCores have a working core that
> runs most of the MIPS instruction set. MIPS is a very well known, tried
> and trusted architecture. My cable modem has a MIPS-compatible processor
> made by Toshiba. It seems to me that MIPS is just as realistic and
> usable platform as ARM. But I would be very interested to hear from
> anyone who knows better.
> 
> Now, I wonder when Qemu will support MIPS emulation? :-)

When somebody implements more of the necessary support. The specs
are available from mips.com (free registration needed).


Thiemo