A couple of fields in a data structure, which is used by the driver only,
were not initialized properly during the driver's setup.
The primary issue with this bug was that channel->wr_buf_size remained zero,
so calls to dma_sync_single_for_cpu() took place with zero size, and
consequently did nothing.
This had a rather minimal practical impact, because
(a) these calls are NOPs on Intel/AMD platforms, as well as other platforms
with coherent cache, and
(b) it's extremely rare that any cache line would survive between two reads
from a given DMA buffer
Hence no significant practical difference is expected with this patch.
Signed-off-by: Eli Billauer <eli.billauer@gmail.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
channel->log2_element_size = ((format > 2) ?
2 : format);
- bytebufsize = channel->rd_buf_size = bufsize *
+ bytebufsize = bufsize *
(1 << channel->log2_element_size);
buffers = devm_kcalloc(dev, bufnum,
if (!is_writebuf) {
channel->num_rd_buffers = bufnum;
+ channel->rd_buf_size = bytebufsize;
channel->rd_allow_partial = allowpartial;
channel->rd_synchronous = synchronous;
channel->rd_exclusive_open = exclusive_open;
bufnum, bytebufsize);
} else if (channelnum > 0) {
channel->num_wr_buffers = bufnum;
+ channel->wr_buf_size = bytebufsize;
channel->seekable = seekable;
channel->wr_supports_nonempty = supports_nonempty;