[Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Blue Swirl]

CVSROOT:	/cvsroot/qemu
Module name:	qemu
Changes by:	Blue Swirl <blueswir1>	08/02/24 14:10:06

Modified files:
	target-sparc   : exec.h op.c op_helper.c op_mem.h translate.c 
Added files:
	target-sparc   : helper.h 
Removed files:
	target-sparc   : op_template.h 

Log message:
	 Modify Sparc32/64 to use TCG

CVSWeb URLs:
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/exec.h?cvsroot=qemu&r1=1.24&r2=1.25
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op.c?cvsroot=qemu&r1=1.47&r2=1.48
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_helper.c?cvsroot=qemu&r1=1.66&r2=1.67
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_mem.h?cvsroot=qemu&r1=1.13&r2=1.14
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/translate.c?cvsroot=qemu&r1=1.88&r2=1.89
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/helper.h?cvsroot=qemu&rev=1.1
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_template.h?cvsroot=qemu&r1=1.2&r2=0

[Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Blue Swirl]

CVSROOT:	/cvsroot/qemu
Module name:	qemu
Changes by:	Blue Swirl <blueswir1>	07/09/21 19:10:53

Modified files:
	target-sparc   : exec.h op.c op_helper.c op_mem.h translate.c 

Log message:
	 Rework ASI instructions (Aurelien Jarno)

CVSWeb URLs:
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/exec.h?cvsroot=qemu&r1=1.19&r2=1.20
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op.c?cvsroot=qemu&r1=1.39&r2=1.40
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_helper.c?cvsroot=qemu&r1=1.36&r2=1.37
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_mem.h?cvsroot=qemu&r1=1.9&r2=1.10
http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/translate.c?cvsroot=qemu&r1=1.68&r2=1.69

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 21.09.2007 um 21:10 schrieb Blue Swirl:

> CVSROOT:	/cvsroot/qemu
> Module name:	qemu
> Changes by:	Blue Swirl <blueswir1>	07/09/21 19:10:53
>
> Modified files:
> 	target-sparc   : exec.h op.c op_helper.c op_mem.h translate.c
>
> Log message:
> 	 Rework ASI instructions (Aurelien Jarno)
>
> CVSWeb URLs:
> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/exec.h? 
> cvsroot=qemu&r1=1.19&r2=1.20
> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op.c? 
> cvsroot=qemu&r1=1.39&r2=1.40
> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_helper.c? 
> cvsroot=qemu&r1=1.36&r2=1.37
> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_mem.h? 
> cvsroot=qemu&r1=1.9&r2=1.10
> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/translate.c? 
> cvsroot=qemu&r1=1.68&r2=1.69
>

These changes appear to have broken sparc32-softmmu on OS X v10.4  
ppc: The OpenBIOS screen no longer appears.

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Blue Swirl]

On 9/21/07, Andreas Färber <andreas.faerber@web.de> wrote:
> These changes appear to have broken sparc32-softmmu on OS X v10.4
> ppc: The OpenBIOS screen no longer appears.

I had exactly the same problem, but "make clean" fixed it. Some
dependencies are not correct.

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 21.09.2007 um 22:25 schrieb Blue Swirl:

> On 9/21/07, Andreas Färber <andreas.faerber@web.de> wrote:
>> These changes appear to have broken sparc32-softmmu on OS X v10.4
>> ppc: The OpenBIOS screen no longer appears.
>
> I had exactly the same problem, but "make clean" fixed it. Some
> dependencies are not correct.

Negative, not even a make distclean + configure fixes it...

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
> 
> 
> Am 21.09.2007 um 21:10 schrieb Blue Swirl:
> 
> >CVSROOT:	/cvsroot/qemu
> >Module name:	qemu
> >Changes by:	Blue Swirl <blueswir1>	07/09/21 19:10:53
> >
> >Modified files:
> >	target-sparc   : exec.h op.c op_helper.c op_mem.h translate.c
> >
> >Log message:
> >	 Rework ASI instructions (Aurelien Jarno)
> >
> >CVSWeb URLs:
> >http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/exec.h? 
> >cvsroot=qemu&r1=1.19&r2=1.20
> >http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op.c? 
> >cvsroot=qemu&r1=1.39&r2=1.40
> >http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_helper.c? 
> >cvsroot=qemu&r1=1.36&r2=1.37
> >http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_mem.h? 
> >cvsroot=qemu&r1=1.9&r2=1.10
> >http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/translate.c? 
> >cvsroot=qemu&r1=1.68&r2=1.69
> >
> 
> These changes appear to have broken sparc32-softmmu on OS X v10.4  
> ppc: The OpenBIOS screen no longer appears.
> 

Could you try to see if qemu.log contains useful debug information while
running in debug mode?

I am currently building qemu-sparc on a powerpc (GNU/Linux) to try to
reproduce the problem, but my machine isn't really fast, it will take
some time.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Aurelien Jarno a écrit :
> On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
>>
>> Am 21.09.2007 um 21:10 schrieb Blue Swirl:
>>
>>> CVSROOT:	/cvsroot/qemu
>>> Module name:	qemu
>>> Changes by:	Blue Swirl <blueswir1>	07/09/21 19:10:53
>>>
>>> Modified files:
>>> 	target-sparc   : exec.h op.c op_helper.c op_mem.h translate.c
>>>
>>> Log message:
>>> 	 Rework ASI instructions (Aurelien Jarno)
>>>
>>> CVSWeb URLs:
>>> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/exec.h? 
>>> cvsroot=qemu&r1=1.19&r2=1.20
>>> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op.c? 
>>> cvsroot=qemu&r1=1.39&r2=1.40
>>> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_helper.c? 
>>> cvsroot=qemu&r1=1.36&r2=1.37
>>> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/op_mem.h? 
>>> cvsroot=qemu&r1=1.9&r2=1.10
>>> http://cvs.savannah.gnu.org/viewcvs/qemu/target-sparc/translate.c? 
>>> cvsroot=qemu&r1=1.68&r2=1.69
>>>
>> These changes appear to have broken sparc32-softmmu on OS X v10.4  
>> ppc: The OpenBIOS screen no longer appears.
>>
> 
> Could you try to see if qemu.log contains useful debug information while
> running in debug mode?
> 
> I am currently building qemu-sparc on a powerpc (GNU/Linux) to try to
> reproduce the problem, but my machine isn't really fast, it will take
> some time.

I am able to reproduce it, so it's probably an endianness problem. I am
debugging...

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 21.09.2007 um 23:07 schrieb Aurelien Jarno:

> Aurelien Jarno a écrit :
>> On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
>>> These changes appear to have broken sparc32-softmmu on OS X v10.4
>>> ppc: The OpenBIOS screen no longer appears.
>>>
>>
>> Could you try to see if qemu.log contains useful debug information  
>> while
>> running in debug mode?
>>
>> I am currently building qemu-sparc on a powerpc (GNU/Linux) to try to
>> reproduce the problem, but my machine isn't really fast, it will take
>> some time.
>
> I am able to reproduce it, so it's probably an endianness problem.  
> I am
> debugging...

I have been unable to enter the monitor... If you need more  
information from me, please provide detailed instructions how to  
obtain the requested log file.

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

On Fri, Sep 21, 2007 at 11:20:26PM +0200, Andreas Färber wrote:
> Am 21.09.2007 um 23:07 schrieb Aurelien Jarno:
> 
> >Aurelien Jarno a écrit :
> >>On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
> >>>These changes appear to have broken sparc32-softmmu on OS X v10.4
> >>>ppc: The OpenBIOS screen no longer appears.
> >>>
> >>
> >>Could you try to see if qemu.log contains useful debug information  
> >>while
> >>running in debug mode?
> >>
> >>I am currently building qemu-sparc on a powerpc (GNU/Linux) to try to
> >>reproduce the problem, but my machine isn't really fast, it will take
> >>some time.
> >
> >I am able to reproduce it, so it's probably an endianness problem.  
> >I am
> >debugging...
> 
> I have been unable to enter the monitor... If you need more  
> information from me, please provide detailed instructions how to  
> obtain the requested log file.
> 

Unfortunately I don't have the same symptoms, so I am not sure it is the
same problem. In my case the bug has been introduced by the iommu commit
which does not consider big endian hosts.

The patch below fixes the problem in my case. Could you please try it?

Index: hw/iommu.c
===================================================================
RCS file: /sources/qemu/qemu/hw/iommu.c,v
retrieving revision 1.15
diff -u -d -p -r1.15 iommu.c
--- hw/iommu.c	20 Sep 2007 16:01:51 -0000	1.15
+++ hw/iommu.c	22 Sep 2007 11:45:55 -0000
@@ -212,7 +212,7 @@ static uint32_t iommu_page_get_flags(IOM
     addr &= ~s->iostart;
     iopte += (addr >> (PAGE_SHIFT - 2)) & ~3;
     cpu_physical_memory_read(iopte, (uint8_t *)&ret, 4);
-    bswap32s(&ret);
+    tswap32s(&ret);
     DPRINTF("get flags addr " TARGET_FMT_plx " => pte " TARGET_FMT_plx
             ", *pte = %x\n", pa, iopte, ret);
 

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Blue Swirl]

On 9/22/07, Aurelien Jarno <aurelien@aurel32.net> wrote:
> On Fri, Sep 21, 2007 at 11:20:26PM +0200, Andreas Färber wrote:
> > Am 21.09.2007 um 23:07 schrieb Aurelien Jarno:
> >
> > >Aurelien Jarno a écrit :
> > >>On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
> > >>>These changes appear to have broken sparc32-softmmu on OS X v10.4
> > >>>ppc: The OpenBIOS screen no longer appears.
> > >>>
> > >>
> > >>Could you try to see if qemu.log contains useful debug information
> > >>while
> > >>running in debug mode?
> > >>
> > >>I am currently building qemu-sparc on a powerpc (GNU/Linux) to try to
> > >>reproduce the problem, but my machine isn't really fast, it will take
> > >>some time.
> > >
> > >I am able to reproduce it, so it's probably an endianness problem.
> > >I am
> > >debugging...
> >
> > I have been unable to enter the monitor... If you need more
> > information from me, please provide detailed instructions how to
> > obtain the requested log file.
> >
>
> Unfortunately I don't have the same symptoms, so I am not sure it is the
> same problem. In my case the bug has been introduced by the iommu commit
> which does not consider big endian hosts.
>
> The patch below fixes the problem in my case. Could you please try it?

This should be it. Thanks a lot for debugging!

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 22.09.2007 um 13:49 schrieb Aurelien Jarno:

> On Fri, Sep 21, 2007 at 11:20:26PM +0200, Andreas Färber wrote:
>> Am 21.09.2007 um 23:07 schrieb Aurelien Jarno:
>>
>>> Aurelien Jarno a écrit :
>>>> On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
>>>>> These changes appear to have broken sparc32-softmmu on OS X v10.4
>>>>> ppc: The OpenBIOS screen no longer appears.
>>>>>
>>>>
>>>> Could you try to see if qemu.log contains useful debug information
>>>> while
>>>> running in debug mode?
>>>>
>>>> I am currently building qemu-sparc on a powerpc (GNU/Linux) to  
>>>> try to
>>>> reproduce the problem, but my machine isn't really fast, it will  
>>>> take
>>>> some time.
>>>
>>> I am able to reproduce it, so it's probably an endianness problem.
>>> I am
>>> debugging...
>>
>> I have been unable to enter the monitor... If you need more
>> information from me, please provide detailed instructions how to
>> obtain the requested log file.
>>
>
> Unfortunately I don't have the same symptoms, so I am not sure it  
> is the
> same problem. In my case the bug has been introduced by the iommu  
> commit
> which does not consider big endian hosts.
>
> The patch below fixes the problem in my case. Could you please try it?

I tried Blue Swirl's commit thereof and it still fails for me, same  
symptoms. The virtual CPU load is at about 80% and nothing happens,  
no graphical or textual output (after "starting qemu..."), it just  
hangs.

I can't say exactly which commit introduced the breakage but it  
must've been one this week... And x86 is not affected.

Could it be related to the slavio keyboard handling changes? I'm  
using an Apple keyboard.

Or maybe the 64-bit flag extension patch by Alexander Graf? Wildly  
guessing here...

Would recording serial output to a file help? Or does that apply only  
to boot problems with an OS? Anything else that could help?

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 22.09.2007 um 13:49 schrieb Aurelien Jarno:
> 
>> On Fri, Sep 21, 2007 at 11:20:26PM +0200, Andreas Färber wrote:
>>> Am 21.09.2007 um 23:07 schrieb Aurelien Jarno:
>>>
>>>> Aurelien Jarno a écrit :
>>>>> On Fri, Sep 21, 2007 at 09:32:00PM +0200, Andreas Färber wrote:
>>>>>> These changes appear to have broken sparc32-softmmu on OS X v10.4
>>>>>> ppc: The OpenBIOS screen no longer appears.
>>>>>>
>>>>> Could you try to see if qemu.log contains useful debug information
>>>>> while
>>>>> running in debug mode?
>>>>>
>>>>> I am currently building qemu-sparc on a powerpc (GNU/Linux) to  
>>>>> try to
>>>>> reproduce the problem, but my machine isn't really fast, it will  
>>>>> take
>>>>> some time.
>>>> I am able to reproduce it, so it's probably an endianness problem.
>>>> I am
>>>> debugging...
>>> I have been unable to enter the monitor... If you need more
>>> information from me, please provide detailed instructions how to
>>> obtain the requested log file.
>>>
>> Unfortunately I don't have the same symptoms, so I am not sure it  
>> is the
>> same problem. In my case the bug has been introduced by the iommu  
>> commit
>> which does not consider big endian hosts.
>>
>> The patch below fixes the problem in my case. Could you please try it?
> 
> I tried Blue Swirl's commit thereof and it still fails for me, same  
> symptoms. The virtual CPU load is at about 80% and nothing happens,  
> no graphical or textual output (after "starting qemu..."), it just  
> hangs.
> 
> I can't say exactly which commit introduced the breakage but it  
> must've been one this week... And x86 is not affected.
> 
> Could it be related to the slavio keyboard handling changes? I'm  
> using an Apple keyboard.
> 
> Or maybe the 64-bit flag extension patch by Alexander Graf? Wildly  
> guessing here...
> 
> Would recording serial output to a file help? Or does that apply only  
> to boot problems with an OS? Anything else that could help?
> 

The main think that would help is to find which commit has broken qemu
on OS X, or in other words the latest version that is working. Then it
would be a lot more easier to find what is wrong in the code.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 22.09.2007 um 19:50 schrieb Aurelien Jarno:

> The main think that would help is to find which commit has broken qemu
> on OS X, or in other words the latest version that is working. Then it
> would be a lot more easier to find what is wrong in the code.

Okay, so on 2007-09-20 15:30:00 UTC it still worked, that's after the  
tadd fix by Blue Swirl.

At 16:10 UTC booting from DVD was broken but the OpenBIOS screen was  
still displayed. This should be after the iommu fix by Blue Swirl  
(for which an endianness fix has been applied today).

No change at 22:50 UTC after the flags extension by A. Graf.

Again no change at 2007-09-21 19:09:45 UTC after the slavio_serial.c  
change by Blue Swirl.

I regret having to confirm this but it indeed breaks completely at  
19:11:00 UTC after your ASI patch is applied... In op_helper.c you  
appeared to change endianness handling but I don't really understand  
what the code does so it could be anything...

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 22.09.2007 um 19:50 schrieb Aurelien Jarno:
> 
>> The main think that would help is to find which commit has broken qemu
>> on OS X, or in other words the latest version that is working. Then it
>> would be a lot more easier to find what is wrong in the code.
> 
> Okay, so on 2007-09-20 15:30:00 UTC it still worked, that's after the  
> tadd fix by Blue Swirl.
> 
> At 16:10 UTC booting from DVD was broken but the OpenBIOS screen was  
> still displayed. This should be after the iommu fix by Blue Swirl  
> (for which an endianness fix has been applied today).
> 
> No change at 22:50 UTC after the flags extension by A. Graf.
> 
> Again no change at 2007-09-21 19:09:45 UTC after the slavio_serial.c  
> change by Blue Swirl.
> 
> I regret having to confirm this but it indeed breaks completely at  
> 19:11:00 UTC after your ASI patch is applied... In op_helper.c you  
> appeared to change endianness handling but I don't really understand  
> what the code does so it could be anything...
> 

I really doubt it is an endianness problem as the current CVS works
perfectly on GNU/Linux powerpc. The code in op_helper.c adds support for
target little endian access (sparc is big endian), but it is anyway not
used in openbios. We have to debug the problem and find what is wrong
with OS X:

- Which version of GCC are you using?
- Could you please try to run qemu with -nographic?
- Also could you please try to run qemu with -d in_asm,op and send us
/tmp/qemu.log?

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

[-- Attachment #1: Type: text/plain, Size: 1142 bytes --]


Am 22.09.2007 um 22:06 schrieb Aurelien Jarno:

> I really doubt it is an endianness problem as the current CVS works
> perfectly on GNU/Linux powerpc. The code in op_helper.c adds  
> support for
> target little endian access (sparc is big endian), but it is anyway  
> not
> used in openbios. We have to debug the problem and find what is wrong
> with OS X:
>
> - Which version of GCC are you using?

I have a gcc 4.0.1 installed but configure appears to choose gcc-3.3:

$ gcc-3.3 --version
gcc-3.3 (GCC) 3.3 20030304 (Apple Computer, Inc. build 1819)
Copyright (C) 2002 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There  
is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR  
PURPOSE.

> - Could you please try to run qemu with -nographic?

I already tried that, it makes no difference - it hangs after  
"starting qemu...".

> - Also could you please try to run qemu with -d in_asm,op and send us
> /tmp/qemu.log?

Of course, see attached. (CVS HEAD configured as before, with .../ 
qemu-system-sparc -boot d -cdrom ...solaris10sparc.iso -nographic -d  
in_asm,op)

[-- Attachment #2: qemu.log --]
[-- Type: application/octet-stream, Size: 1970 bytes --]

--------------
IN: 
0xffd00000:  b  0xffd040c0
0xffd00004:  nop 

OP:
0x0000: jmp_im 0xffd040c0
0x0001: movl_npc_im 0xffd040c4
0x0002: movl_T0_0
0x0003: exit_tb
0x0004: end

--------------
IN: 
0xffd040c0:  sethi  %hi(0x71200000), %g1
0xffd040c4:  lduba  [ %g1 ] #ASI_M_BYPASS, %g2
0xffd040c8:  cmp  %g2, 0x51
0xffd040cc:  bne  0xffd041a0
0xffd040d0:  inc  %g1

OP:
0x0000: movl_T0_im 0x71200000
0x0001: movl_g1_T0
0x0002: jmp_im 0xffd040c4
0x0003: movl_npc_im 0xffd040c8
0x0004: movl_T0_g1
0x0005: ld_asi 0x20 0x1 0x0
0x0006: movl_g2_T1
0x0007: movl_T0_g2
0x0008: movl_T1_sim 0x51
0x0009: sub_T1_T0_cc
0x000a: eval_bne
0x000b: movl_T0_g1
0x000c: movl_T1_sim 0x1
0x000d: add_T1_T0
0x000e: movl_g1_T0
0x000f: jz_T2_label 0x0
0x0010: goto_tb0
0x0011: jmp_im 0xffd041a0
0x0012: movl_npc_im 0xffd041a4
0x0013: movl_T0_im 0x8ae344
0x0014: exit_tb
0x0015: goto_tb1
0x0016: jmp_im 0xffd040d4
0x0017: movl_npc_im 0xffd040d8
0x0018: movl_T0_im 0x8ae345
0x0019: exit_tb
0x001a: end

--------------
IN: 
0xffd041a0:  sethi  %hi(0xf1200000), %g1
0xffd041a4:  lduba  [ %g1 ] #ASI_M_CTL, %g2
0xffd041a8:  cmp  %g2, 0x51
0xffd041ac:  bne  0xffd04198
0xffd041b0:  inc  %g1

OP:
0x0000: movl_T0_im 0xf1200000
0x0001: movl_g1_T0
0x0002: jmp_im 0xffd041a4
0x0003: movl_npc_im 0xffd041a8
0x0004: movl_T0_g1
0x0005: ld_asi 0x2f 0x1 0x0
0x0006: movl_g2_T1
0x0007: movl_T0_g2
0x0008: movl_T1_sim 0x51
0x0009: sub_T1_T0_cc
0x000a: eval_bne
0x000b: movl_T0_g1
0x000c: movl_T1_sim 0x1
0x000d: add_T1_T0
0x000e: movl_g1_T0
0x000f: jz_T2_label 0x0
0x0010: goto_tb0
0x0011: jmp_im 0xffd04198
0x0012: movl_npc_im 0xffd0419c
0x0013: movl_T0_im 0x8ae388
0x0014: exit_tb
0x0015: goto_tb1
0x0016: jmp_im 0xffd041b4
0x0017: movl_npc_im 0xffd041b8
0x0018: movl_T0_im 0x8ae389
0x0019: exit_tb
0x001a: end

--------------
IN: 
0xffd04198:  b  0xffd04198
0xffd0419c:  nop 

OP:
0x0000: goto_tb0
0x0001: jmp_im 0xffd04198
0x0002: movl_npc_im 0xffd0419c
0x0003: movl_T0_im 0x8ae3cc
0x0004: exit_tb
0x0005: end


[-- Attachment #3: Type: text/plain, Size: 52 bytes --]


Thanks for your interest, and let me know,

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 22.09.2007 um 22:06 schrieb Aurelien Jarno:
> 
>> I really doubt it is an endianness problem as the current CVS works
>> perfectly on GNU/Linux powerpc. The code in op_helper.c adds  
>> support for
>> target little endian access (sparc is big endian), but it is anyway  
>> not
>> used in openbios. We have to debug the problem and find what is wrong
>> with OS X:
>>
>> - Which version of GCC are you using?
> 
> I have a gcc 4.0.1 installed but configure appears to choose gcc-3.3:
> 
> $ gcc-3.3 --version
> gcc-3.3 (GCC) 3.3 20030304 (Apple Computer, Inc. build 1819)
> Copyright (C) 2002 Free Software Foundation, Inc.
> This is free software; see the source for copying conditions.  There  
> is NO
> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR  
> PURPOSE.
> 
>> - Could you please try to run qemu with -nographic?
> 
> I already tried that, it makes no difference - it hangs after  
> "starting qemu...".
> 
>> - Also could you please try to run qemu with -d in_asm,op and send us
>> /tmp/qemu.log?
> 
> Of course, see attached. (CVS HEAD configured as before, with .../ 
> qemu-system-sparc -boot d -cdrom ...solaris10sparc.iso -nographic -d  
> in_asm,op)
> 

Thanks, that's helpful, it looks like the "lduba" instructions is not
working correctly in your case. Now I have to find why.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

>> Am 22.09.2007 um 22:06 schrieb Aurelien Jarno:
>>
>>> I really doubt it is an endianness problem as the current CVS works
>>> perfectly on GNU/Linux powerpc.

I can confirm that OpenBIOS displays fine at CVS HEAD under Debian  
GNU/Linux ppc64 on the same machine (gcc 3.4, SDL). Maybe alignment  
differences?

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Aurelien Jarno a écrit :
> Andreas Färber a écrit :
>> Am 22.09.2007 um 22:06 schrieb Aurelien Jarno:
>>
>>> I really doubt it is an endianness problem as the current CVS works
>>> perfectly on GNU/Linux powerpc. The code in op_helper.c adds  
>>> support for
>>> target little endian access (sparc is big endian), but it is anyway  
>>> not
>>> used in openbios. We have to debug the problem and find what is wrong
>>> with OS X:
>>>
>>> - Which version of GCC are you using?
>> I have a gcc 4.0.1 installed but configure appears to choose gcc-3.3:
>>
>> $ gcc-3.3 --version
>> gcc-3.3 (GCC) 3.3 20030304 (Apple Computer, Inc. build 1819)
>> Copyright (C) 2002 Free Software Foundation, Inc.
>> This is free software; see the source for copying conditions.  There  
>> is NO
>> warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR  
>> PURPOSE.
>>
>>> - Could you please try to run qemu with -nographic?
>> I already tried that, it makes no difference - it hangs after  
>> "starting qemu...".
>>
>>> - Also could you please try to run qemu with -d in_asm,op and send us
>>> /tmp/qemu.log?
>> Of course, see attached. (CVS HEAD configured as before, with .../ 
>> qemu-system-sparc -boot d -cdrom ...solaris10sparc.iso -nographic -d  
>> in_asm,op)
>>
> 
> Thanks, that's helpful, it looks like the "lduba" instructions is not
> working correctly in your case. Now I have to find why.

I don't really understand, because that is the part of the code (at
least from the op point of view) that is almost unchanged. Looking at
the code again and again I can't see any problem.

Could you please try to find the values of T1 at the end of
helper_ld_asi (adding a printf or using gdb). That may help us to
understand what is the problem.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 22.09.2007 um 23:41 schrieb Aurelien Jarno:

> Aurelien Jarno a écrit :
>> Thanks, that's helpful, it looks like the "lduba" instructions is not
>> working correctly in your case. Now I have to find why.
>
> I don't really understand, because that is the part of the code (at
> least from the op point of view) that is almost unchanged. Looking at
> the code again and again I can't see any problem.
>
> Could you please try to find the values of T1 at the end of
> helper_ld_asi (adding a printf or using gdb). That may help us to
> understand what is the problem.

With:

printf("helper_ld_asi: T1 = %llx\n", T1); // not sure if that's correct?

I get these two:

helper_ld_asi: T1 = 5171200000
helper_ld_asi: T1 = 0

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 22.09.2007 um 23:41 schrieb Aurelien Jarno:
> 
>> Aurelien Jarno a écrit :
>>> Thanks, that's helpful, it looks like the "lduba" instructions is not
>>> working correctly in your case. Now I have to find why.
>> I don't really understand, because that is the part of the code (at
>> least from the op point of view) that is almost unchanged. Looking at
>> the code again and again I can't see any problem.
>>
>> Could you please try to find the values of T1 at the end of
>> helper_ld_asi (adding a printf or using gdb). That may help us to
>> understand what is the problem.
> 
> With:
> 
> printf("helper_ld_asi: T1 = %llx\n", T1); // not sure if that's correct?

Yes that is correct.

> I get these two:
> 
> helper_ld_asi: T1 = 5171200000

The correct value here is 51. I don't really understand why this value
is not a 8-bit one, as the memory access is done through ldub_phys().
Could you please verify that helper_ld_asi() is called with the correct
parameters. This should be 0x20 0x1 0x0.

PS: I don't know if I will be able to get internet access in the next 4
days.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 23.09.2007 um 01:11 schrieb Aurelien Jarno:

> Andreas Färber a écrit :
>> Am 22.09.2007 um 23:41 schrieb Aurelien Jarno:
>>
>>> Aurelien Jarno a écrit :
>>>> Thanks, that's helpful, it looks like the "lduba" instructions  
>>>> is not
>>>> working correctly in your case. Now I have to find why.
>>> I don't really understand, because that is the part of the code (at
>>> least from the op point of view) that is almost unchanged.  
>>> Looking at
>>> the code again and again I can't see any problem.
>>>
>>> Could you please try to find the values of T1 at the end of
>>> helper_ld_asi (adding a printf or using gdb). That may help us to
>>> understand what is the problem.
>>
>> With:
>>
>> printf("helper_ld_asi: T1 = %llx\n", T1); // not sure if that's  
>> correct?
>
> Yes that is correct.
>
>> I get these two:
>>
>> helper_ld_asi: T1 = 5171200000
>
> The correct value here is 51. I don't really understand why this value
> is not a 8-bit one, as the memory access is done through ldub_phys().
> Could you please verify that helper_ld_asi() is called with the  
> correct
> parameters. This should be 0x20 0x1 0x0.
>
> PS: I don't know if I will be able to get internet access in the  
> next 4
> days.

Take your time, I've already reverted my Q back to the last working  
state now that I've figured out that cool -D option.

Actually my printf wasn't correct: For sparc32 T1 is a uint32_t, with  
%llx raising a warning I didn't notice before; with %x it does show  
the correct value of 51.

printf("helper_ld_asi(%d, %d, %d)\n", asi, size, sign);
...
printf("helper_ld_asi: T1 = %x\n", T1);

starting qemu...
helper_ld_asi(32, 1, 0)
helper_ld_asi: T1 = 51
helper_ld_asi(47, 1, 0)
helper_ld_asi: T1 = 0

So with 0x20==32 that part looks correct.

Andreas

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 23.09.2007 um 01:11 schrieb Aurelien Jarno:

>>> Aurelien Jarno a écrit :
>>>> Thanks, that's helpful, it looks like the "lduba" instructions  
>>>> is not
>>>> working correctly in your case. Now I have to find why.
>>> I don't really understand, because that is the part of the code (at
>>> least from the op point of view) that is almost unchanged.  
>>> Looking at
>>> the code again and again I can't see any problem.
>>
> PS: I don't know if I will be able to get internet access in the  
> next 4
> days.

Just a short note that Blue Swirl's added break statements did not  
change the results previously posted.

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Aurelien Jarno]

Andreas Färber a écrit :
> Am 23.09.2007 um 01:11 schrieb Aurelien Jarno:
> 
>>>> Aurelien Jarno a écrit :
>>>>> Thanks, that's helpful, it looks like the "lduba" instructions  
>>>>> is not
>>>>> working correctly in your case. Now I have to find why.
>>>> I don't really understand, because that is the part of the code (at
>>>> least from the op point of view) that is almost unchanged.  
>>>> Looking at
>>>> the code again and again I can't see any problem.
>> PS: I don't know if I will be able to get internet access in the  
>> next 4
>> days.
> 
> Just a short note that Blue Swirl's added break statements did not  
> change the results previously posted.

Oh, I thought it was fixed by this commited.

Looking at the data you sent in your last mail, it looks like the lduba
instruction is actually working correctly, given the first returned
value is correct. But then it looks like the value is not correct in the
 cmp instruction as the wrong branch is taken.

I have currently no idea why.

-- 
  .''`.  Aurelien Jarno	            | GPG: 1024D/F1BCDB73
 : :' :  Debian developer           | Electrical Engineer
 `. `'   aurel32@debian.org         | aurelien@aurel32.net
   `-    people.debian.org/~aurel32 | www.aurel32.net

Re: [Qemu-devel] qemu/target-sparc exec.h op.c op_helper.c op_me...

[Andreas Färber]

Am 26.09.2007 um 13:05 schrieb Aurelien Jarno:

> Andreas Färber a écrit :
>> Am 23.09.2007 um 01:11 schrieb Aurelien Jarno:
>>
>>>>> Aurelien Jarno a écrit :
>>>>>> Thanks, that's helpful, it looks like the "lduba" instructions
>>>>>> is not
>>>>>> working correctly in your case. Now I have to find why.
>>>>> I don't really understand, because that is the part of the code  
>>>>> (at
>>>>> least from the op point of view) that is almost unchanged.
>>>>> Looking at
>>>>> the code again and again I can't see any problem.
>>> PS: I don't know if I will be able to get internet access in the
>>> next 4
>>> days.
>>
>> Just a short note that Blue Swirl's added break statements did not
>> change the results previously posted.
>
> Oh, I thought it was fixed by this commited.
>
> Looking at the data you sent in your last mail, it looks like the  
> lduba
> instruction is actually working correctly, given the first returned
> value is correct. But then it looks like the value is not correct  
> in the
>  cmp instruction as the wrong branch is taken.
>
> I have currently no idea why.

Okay, so this is the list of compiler warnings, configured with -- 
prefix=... --target-list=sparc-softmmu --enable-cocoa:

$ make >/dev/null
dyngen.c: In function 'get_reloc_name':
dyngen.c:1028: warning: pointer targets in passing argument 2 of  
'fetch_next_pair_value' differ in signedness
dyngen.c:1030: warning: pointer targets in passing argument 2 of  
'fetch_next_pair_value' differ in signedness
dyngen.c:1032: warning: pointer targets in passing argument 2 of  
'fetch_next_pair_value' differ in signedness
/Users/andreas/Q/myqemu/vl.c: In function `init_timer_alarm':
/Users/andreas/Q/myqemu/vl.c:1516: warning: `t' might be used  
uninitialized in this function
/Users/andreas/Q/myqemu/vl.c: At top level:
/Users/andreas/Q/myqemu/vl.c:1183: warning: `qemu_next_deadline'  
defined but not used
/Users/andreas/Q/myqemu/hw/eepro100.c:269: warning:  
`eepro100_mdi_mask' defined but not used
/Users/andreas/Q/myqemu/hw/pcnet.c:585: warning: `crctab' defined but  
not used
/Users/andreas/Q/myqemu/cocoa.m: In function `setApplicationMenu':
/Users/andreas/Q/myqemu/cocoa.m:860: warning: cannot find method `- 
setAppleMenu:'; return type `id' assumed
../dyngen-exec.h:229: warning: `__op_param1' defined but not used
../dyngen-exec.h:229: warning: `__op_param2' defined but not used
../dyngen-exec.h:229: warning: `__op_param3' defined but not used
../dyngen-exec.h:229: warning: `__op_param1' defined but not used
../dyngen-exec.h:229: warning: `__op_param2' defined but not used
../dyngen-exec.h:229: warning: `__op_param3' defined but not used

Signedness warnings appear to be quite common for cross-platform  
software on OS X and usually do not matter, the rest looks okay to me.
Just tell me if any additional information or debug output can help  
give you any idea.

Andreas