[OpenBIOS] Solaris anyone?

[Blue Swirl]

On Tue, Jun 14, 2011 at 8:19 AM, Bob Breuer <breuerr at mc.net> wrote:
> Blue Swirl wrote:
>> On Mon, Jun 13, 2011 at 2:06 AM, Bob Breuer <breuerr at mc.net> wrote:
>>
>>> Nathan Kunkee wrote:
>>>
>>>>> Date: Mon, 6 Jun 2011 15:54:39 -0400
>>>>> From: tarl-b2 at tarl.net
>>>>> To: openbios at openbios.org
>>>>> Subject: Re: [OpenBIOS] Solaris anyone?
>>>>>
>>>>>
>>>>>> Hmmm it looks from reading the source code that QEMU doesn't actually
>>>>>> implement a proper audio device for SS-5 at all but merely provides a
>>>>>> dummy device with an empty mapped region. In that case it may be
>>>>>> possible to provide a simple mapping in OpenBIOS to match the one in
>>>>>> QEMU which may be enough to persuade Solaris to boot.
>>>>>>
>>>>> Ugh. I recall that audio chip being non-trivial...
>>>>>
>>>>> It might be worth deleting that audio device from the device tree and
>>>>> see if Solaris really does get further.
>>>>>
>>>>>
>>>> I've tried the same CD image with the SS-10 QEMU machine, which seems
>>>> to lack the audio device, and I still hang. I get the first part of
>>>> the kernel, no extra logging, and then the spinner to infinity...
>>>>
>>>> nathan at xanth:/usr/local/build/qemu$ sparc-softmmu/qemu-system-sparc
>>>> -bios openbios-builtin.elf -M SS-10 -cdrom
>>>> /local/storage/ISO/solaris-9-sparc-d1.iso -nographic
>>>>
>>>> Configuration device id QEMU version 1 machine id 64
>>>> CPUs: 1 x TI,TMS390Z55
>>>> UUID: 00000000-0000-0000-0000-000000000000
>>>> Welcome to OpenBIOS v1.0 built on Apr 30 2011 02:11
>>>>   Type 'help' for detailed information
>>>> Trying disk...
>>>> No valid state has been set by load or init-program
>>>>
>>>> 0 > boot cd:d -k -v Not a bootable ELF image
>>>> Loading a.out image...
>>>> Loaded 7680 bytes
>>>> entry point is 0x4000
>>>> bootpath: /iommu/sbus/espdma/esp/sd at 2,0:d
>>>>
>>>> Jumping to entry point 00004000 for type 00000005...
>>>> switching to new context:
>>>> Size: 0x45c9f+0xdaf1+0x1d6a7 Bytes
>>>> SunOS Release 5.9 Version Generic_112233-12 32-bit
>>>> Copyright 1983-2003 Sun Microsystems, Inc.  All rights reserved.
>>>> Use is subject to license terms.
>>>> Ethernet address = 52:54:0:12:34:56
>>>> Using default device instance data
>>>> \
>>>>
>>>>
>>>> Does anyone know what loads on a real SS after the audio device?
>>>>
>>>> Nathan
>>>>
>>> For this hang, Solaris is stuck trying to set and verify a bit which is
>>> currently un-settable in qemu.  It's the MMU breakpoint action register
>>> (asi 0x4c), and if bit 12 is set, then it should continue on just like
>>> with SS-5.  After fcp0, the next thing Solaris should initialize is the
>>> network:
>>>
>>> pseudo-device: fcp0
>>> fcp0 is /pseudo/fcp at 0
>>> Using RPC Bootparams for network configuration information.
>>> le0: No carrier - cable disconnected or hub link test disabled?
>>> le0: No carrier - cable disconnected or hub link test disabled?
>>> Skipping interface le0
>>>
>>
>> Would this (untested) patch to QEMU help?
>>
> It does work, but ...
>> diff --git a/target-sparc/cpu.h b/target-sparc/cpu.h
>> index 320530e..b93fdfd 100644
>> --- a/target-sparc/cpu.h
>> +++ b/target-sparc/cpu.h
>> @@ -403,6 +403,7 @@ typedef struct CPUSPARCState {
>>      uint32_t mmuregs[32];
>>      uint64_t mxccdata[4];
>>      uint64_t mxccregs[8];
>> +    uint64_t mmubpaction;
>>
> Looks like it's less than 64 bits:
>
> ok 0 0 0 4c spaced!
> ok 0 4c spaced@ . .
> 0 0
> ok 0 not dup 0 4c spaced!
> ok 0 4c spaced@ . .
> 0 1fff
> ok 0 4c spacel@ . 4 4c spacel@ .
> 1fff 1fff
>
> Would it make sense to declare it as 32-bit, or just mask out the extra
> bits?  or not even care since almost nobody writes to that register?

Better use the correct size.

>>      uint64_t mmubpregs[4];
>>      uint64_t prom_addr;
>>  #endif
>> diff --git a/target-sparc/op_helper.c b/target-sparc/op_helper.c
>> index b38691e..1023870 100644
>> --- a/target-sparc/op_helper.c
>> +++ b/target-sparc/op_helper.c
>> @@ -1941,7 +1941,7 @@ uint64_t helper_ld_asi(target_ulong addr, int
>> asi, int size, int sign)
>>      case 0x32: // Turbosparc page table descriptor diagnostic
>>      case 0x39: /* data cache diagnostic register */
>>      case 0x4c: /* SuperSPARC MMU Breakpoint Action register */
>> -        ret = 0;
>> +        ret = env->mmubpaction;
>>          break;
>>
> And here, 0x39 and above should still have the ret = 0.

OK.

>>      case 0x38: /* SuperSPARC MMU Breakpoint Control Registers */
>>          {
>> @@ -2304,7 +2304,9 @@ void helper_st_asi(target_ulong addr, uint64_t
>> val, int asi, int size)
>>                 // descriptor diagnostic
>>      case 0x36: /* I-cache flash clear */
>>      case 0x37: /* D-cache flash clear */
>> +        break;
>>      case 0x4c: /* breakpoint action */
>> +        env->mmubpaction = val;
>>          break;
>>      case 0x38: /* SuperSPARC MMU Breakpoint Control Registers*/
>>          {
>>
> By the way, what would it take to get the space[cwLd] commands in
> OpenBIOS?  They come in handy for verifying these asi accesses.

Shouldn't be very difficult. Since on Sparc32 it's not possible to
give ASI in a register, a switch or hand assembly would be needed.