Re: [Qemu-devel] [PATCH] isapc: Shadow ISA BIOS by default

[Anthony Liguori <aliguori@us.ibm.com>]

Jan Kiszka <jan.kiszka@siemens.com> writes:

> On 2012-10-12 15:41, Anthony Liguori wrote:
>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>> 
>>> On 2012-10-08 20:52, Anthony Liguori wrote:
>>>> Jan Kiszka <jan.kiszka@siemens.com> writes:
>>>>
>>>>> On 2012-09-11 17:53, Jan Kiszka wrote:
>>>>>> Our one and only BIOS depends on a writable shadowed BIOS in the ISA
>>>>>> range. As we have no interface to control the write property, make that
>>>>>> region writable by default.
>>>>>>
>>>>>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>>>>>> ---
>>>>>>
>>>>>> This unbreaks isapc for TCG, and keep it working for KVM once it starts
>>>>>> supporting read-only memslots.
>>>>>>
>>>>>>  hw/pc_sysfw.c |   13 +++++++++----
>>>>>>  1 files changed, 9 insertions(+), 4 deletions(-)
>>>>>>
>>>>>> diff --git a/hw/pc_sysfw.c b/hw/pc_sysfw.c
>>>>>> index b45f0ac..027d98a 100644
>>>>>> --- a/hw/pc_sysfw.c
>>>>>> +++ b/hw/pc_sysfw.c
>>>>>> @@ -136,6 +136,7 @@ static void old_pc_system_rom_init(MemoryRegion *rom_memory)
>>>>>>  {
>>>>>>      char *filename;
>>>>>>      MemoryRegion *bios, *isa_bios;
>>>>>> +    void *isa_bios_ptr;
>>>>>>      int bios_size, isa_bios_size;
>>>>>>      int ret;
>>>>>>  
>>>>>> @@ -167,19 +168,23 @@ static void old_pc_system_rom_init(MemoryRegion *rom_memory)
>>>>>>          g_free(filename);
>>>>>>      }
>>>>>>  
>>>>>> -    /* map the last 128KB of the BIOS in ISA space */
>>>>>> +    /* Shadow the last 128KB of the BIOS in ISA space as RAM -
>>>>>> +     * Seabios depends on this */
>>>>>>      isa_bios_size = bios_size;
>>>>>>      if (isa_bios_size > (128 * 1024)) {
>>>>>>          isa_bios_size = 128 * 1024;
>>>>>>      }
>>>>>>      isa_bios = g_malloc(sizeof(*isa_bios));
>>>>>> -    memory_region_init_alias(isa_bios, "isa-bios", bios,
>>>>>> -                             bios_size - isa_bios_size, isa_bios_size);
>>>>>> +    memory_region_init_ram(isa_bios, "isa-bios", isa_bios_size);
>>>>>> +    vmstate_register_ram_global(isa_bios);
>>>>>>      memory_region_add_subregion_overlap(rom_memory,
>>>>>>                                          0x100000 - isa_bios_size,
>>>>>>                                          isa_bios,
>>>>>>                                          1);
>>>>>> -    memory_region_set_readonly(isa_bios, true);
>>>>>> +
>>>>>> +    /* copy ISA rom image from top of the ROM */
>>>>>> +    isa_bios_ptr = memory_region_get_ram_ptr(isa_bios);
>>>>>> +    rom_copy(isa_bios_ptr, (uint32_t)(-isa_bios_size), isa_bios_size);
>>>>>>  
>>>>>>      /* map all the bios at the top of memory */
>>>>>>      memory_region_add_subregion(rom_memory,
>>>>>>
>>>>>
>>>>> Ping. Or already queued?
>>>>
>>>> I've got it queued now.  Thanks.
>>>
>>> I'm withdrawing it: This breaks Windows 95 booting. For unknown reasons,
>>> this nice OS decided to overwrite the F-segment during boot. That is
>>> fine as long as it is properly protected. But it breaks under current
>>> KVM and with the patch above for the isapc. So we need a firmware
>>> interface to enable/disable write protection for this segment in isapc
>>> mode, specifically as that machine targets these old OSes.
>> 
>> Ah, if it wasn't for a build break caused by one of the pull requests, I
>> would have pushed last night.  Thanks for the heads up, I'll remove it
>> from my queue.
>> 
>> Is fw_cfg the right interface?  I presume this is i440fx specific?  How
>> does q35 handle this?
>
> No, there is no i440fx or q35 in that case. There are discrete chips
> and wiring on an undefined ISA motherboard. As Seabios depends on a
> writable E&F-segments (maybe only on E, still need to find out) for a
> certain period, we need to invent a pv channel (probably via fw_cfg) to
> provide the necessary control knob.

I see, I thought this was primarily for shadowing.  But it's a
SeaBIOS-ism.  fw_cfg is the right answer.

Regards,

Anthony Liguori

>
>>  Presumably there's a second window for the BIOS
>> mapping.  There's got to be some way to do shadowing of it I would
>> think.
>
> Not sure what you mean here. This is only about shadowing the top 128K
> of the BIOS into the E/F-segment and providing a write-enable knob for it.
>
> Jan
>
> -- 
> Siemens AG, Corporate Technology, CT RTC ITP SDP-DE
> Corporate Competence Center Embedded Linux