Re: [PATCH] Document Linux's memory barriers [try #4]

[Sergei Organov <osv@javad.com>]

"linux-os \(Dick Johnson\)" <linux-os@analogic.com> writes:

> On Tue, 14 Mar 2006, David Howells wrote:
>
>> Sergei Organov <osv@javad.com> wrote:
>>
>>> "You can prevent an `asm' instruction from being deleted by writing the
>>> keyword `volatile' after the `asm'. [...]
>>> The `volatile' keyword indicates that the instruction has important
>>> side-effects.  GCC will not delete a volatile `asm' if it is reachable.
>>> (The instruction can still be deleted if GCC can prove that
>>> control-flow will never reach the location of the instruction.)  *Note
>>> that even a volatile `asm' instruction can be moved relative to other
>>> code, including across jump instructions.*"
>>
>> Ummm... If "asm volatile" statements don't form compiler barriers, then how do
>> you specify a compiler barrier? Or is that what the "memory" bit in:
>>
>> 	#define barrier() __asm__ __volatile__("": : :"memory")
>>
>> does?
>>
>> David
>
> Yeh. This is the problem (restated) that I mentioned the other
> day when you must do a dummy read of the PCI/Bus to flush all
> the writes, to some variable that gcc can't decide isn't
> important. That's why (void)readl(PCI_STATUS) won't work
> (with gcc 3.3.3 anyway).

If it indeed doesn't, then it's a bug in GCC. GCC shouldn't throw
away volatile accesses. I've already quoted the GCC manual for you:

" Less obvious expressions are where something which looks like an access
is used in a void context.  An example would be,

     volatile int *src = SOMEVALUE;
     *src;

 With C, such expressions are rvalues, and as rvalues cause a read of
the object, GCC interprets this as a read of the volatile being pointed
to. "

-- Sergei.