* question about using DMA for davinci SPI controller
@ 2010-05-10 8:17 Bi Junxiao
0 siblings, 0 replies; only message in thread
From: Bi Junxiao @ 2010-05-10 8:17 UTC (permalink / raw
To: davinci-linux-open-source; +Cc: linux-kernel
Hi all,
I am trying to fix a bug about full duplex DMA transfer for SPI
controller hang on ti-omapl138 ref borad.
I read the DMA setting code of davinci SPI controller. It's in
drivers/spi/davinci_spi.c:davinci_spi_bufs_dma().
There is a code snippet which I don't understand. List below:
--------------------------------------------------------------------------------
if (t->tx_buf) {
t->tx_dma = dma_map_single(&spi->dev, (void *)t->tx_buf, count,
DMA_TO_DEVICE);
if (dma_mapping_error(&spi->dev, t->tx_dma)) {
dev_dbg(sdev, "Unable to DMA map a %d bytes"
" TX buffer\n", count);
return -ENOMEM;
}
temp_count = count;
} else {
/* We need TX clocking for RX transaction */
t->tx_dma = dma_map_single(&spi->dev,
(void *)davinci_spi->tmp_buf, count + 1,
DMA_TO_DEVICE);
if (dma_mapping_error(&spi->dev, t->tx_dma)) {
dev_dbg(sdev, "Unable to DMA map a %d bytes"
" TX tmp buffer\n", count);
return -ENOMEM;
}
temp_count = count + 1;
}
----------------------------------------------------------------------------------
My question is why the transfer count in else logic is count+1 but in if
logic it's only count?
When I changed it from count＋1 to count and use a test program to read
a spi flash, I found the DMA rx interrupt can not be triggered.
The test program hung for ever.
Would anybody have any idea about this?
Thanks.
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2010-05-10 8:19 UTC | newest]
Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2010-05-10 8:17 question about using DMA for davinci SPI controller Bi Junxiao
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).