Re: [PATCH] Wistron laptop button driver

[Miloslav Trmac <mitr@volny.cz>]

Andrew Morton wrote:
> Miloslav Trmac <mitr@volny.cz> wrote:
> 
>>+static void call_bios(struct regs *regs)
>> +{
>> +	unsigned long flags;
>> +
>> +	preempt_disable();
>> +	local_irq_save(flags);
>> +	asm volatile ("pushl %%ebp;"
>> +		      "movl %[data], %%ebp;"
>> +		      "call *%[routine];"
>> +		      "popl %%ebp"
>> +		      : "=a" (regs->eax), "=b" (regs->ebx), "=c" (regs->ecx)
>> +		      : "0" (regs->eax), "1" (regs->ebx), "2" (regs->ecx),
>> +			[routine] "m" (bios_entry_point),
>> +			[data] "m" (bios_data_map_base)
>> +		      : "edx", "edi", "esi", "memory");
>> +	local_irq_restore(flags);
>> +	preempt_enable();
>> +}
>> +
> 
> gcc-2.95.x spits the dummy over this [routine] stuff.  What compiler does
> this require?
(info gcc) says gcc >= 3.1.

> Is it necessary to open-code the BIOS call in the driver?  Does it make
> sense to have some library function to do this?
A general library function (handling all registers) would have to be
written in assembler because gcc has trouble allocating registers
otherwise; considering that we should ideally avoid BIOS calls and that
such a function wasn't needed in the last 14 years, I don't think it is
worth the effort.
	Mirek