On Fri, Oct 17, 2008 at 08:06:55PM +0200, Joern Nettingsmeier wrote:
> excuse my chiming in here, i'm not really much of a c programmer... but
> how can this
> - rb->read_ptr += n1;
> - rb->read_ptr &= rb->size_mask;
> + rb->read_ptr = (rb->read_ptr + n1) & rb->size_mask;
> fix anything?
>
> iiuc, both versions are equivalent. a context switch could happen just
> as well after the parenthesis has been computed..!?
> putting stuff on one line doesn't make it atomic. maybe you are now
> getting another compiler optimization that helps to hide the bug?
The idea is that it is very unlikely that the compiler
would store the intermediate result in rb->read_ptr,
and so this value is updated only once.
I agree it is fragile, there is nothing stopping
the compiler from doing it wrong. Unless the
the pointers are made volatile again.
One real solution is to *not* use the size_mask
at this point, but apply it only when read_ptr
and write_ptr are read. So only a single addition
remains. The pointers are allowed to grow, they
will wrap around at the word size, but since the
buffer size is a power of two they are still
correct when that happens.
This is what is done in the C++ class I
posted earlier.
Ciao,
-- FA Laboratorio di Acustica ed Elettroacustica Parma, Italia Lascia la spina, cogli la rosa. _______________________________________________ Linux-audio-user mailing list Linux-audio-user@email-addr-hidden http://lists.linuxaudio.org/mailman/listinfo/linux-audio-userReceived on Sat Oct 18 00:15:03 2008
This archive was generated by hypermail 2.1.8 : Sat Oct 18 2008 - 00:15:03 EEST