[openib-general] Re: [PATCH 53 of 53] ipath - add memory barrier when waiting for writes

[ralphc at pathscale.com]

> On Mon, 2006-05-15 at 08:57 -0700, Roland Dreier wrote:
>>  >  static void i2c_wait_for_writes(struct ipath_devdata *dd)
>>  >  {
>>  > +	mb();
>>  >  	(void)ipath_read_kreg32(dd, dd->ipath_kregs->kr_scratch);
>>  >  }
>>
>> This needs a comment explaining why it's needed.  A memory barrier
>> before a readl() looks very strange since readl() should be ordered
>> anyway.
>
> Yeah.  It's actually working around what appears to be a gcc bug if the
> kernel is compiled with -Os.  Ralph knows the details; he can give a
> more complete answer.
>
> 	<b

I don't have a lot to add to this other than I looked at the
assembly code output for -Os and -O3 and both looked OK.
I put the mb() in to be sure the writes were complete and
I found this to work by experimentation.
Without it, the driver fails to read the EEPROM correctly.