Apparent bug in sound/oss/via82cxxx_audio.c GETOPTR ioctl

[Jeremy White]

Hi,

In trying to fix Wine's dsound implementation which uses mmap and
DSP_SND_GETOPTR to write data out to the driver, I uncovered what
appears to be a bug in the via82cxx_audio.c driver.

Specifically, via_sg_offset appears to return the number of bytes
not yet consumed within the current block; we use it to adjust from blocks
to a precise block count.  However, it appears as though it returns 0
to indicate that no bytes have been consumed.

I was seeing results where I would get an OPTR return along the
lines of 0x1c00, 0x1e00, 0x2800, and then 0x2100.  Needless to say,
having info.ptr go backwards causes Wine's code to have a serious
hissy fit.

The attached patch fixes the problem for me, but I would appreciate
review by an expert on this.

Cheers,

Jeremy
--- via82cxxx_audio.c	2005-03-01 11:27:07.525265232 -0600
+++ via82cxxx_audio.c.orig	2005-03-01 11:27:24.700654176 -0600
@@ -2831,8 +2831,6 @@
 		unsigned long extra;
 		info.ptr = atomic_read (&chan->hw_ptr) * chan->frag_size;
 		extra = chan->frag_size - via_sg_offset(chan);
-                if (extra == chan->frag_size)
-                    extra = 0;
 		info.ptr += extra;
 		info.bytes += extra;
 	} else {