diff options
author | Tom Hughes <tomhughes@chromium.org> | 2019-08-02 10:19:43 -0700 |
---|---|---|
committer | Commit Bot <commit-bot@chromium.org> | 2019-08-13 00:46:05 +0000 |
commit | 4609b59404bd6e32c4cdb267658511a46557dab2 (patch) | |
tree | fd2182960546330dd7b0d222e49d29be6a4643c1 /chip/mchp | |
parent | d09bc18d46c37070e3309b4f72a33d27134dea45 (diff) | |
download | chrome-ec-4609b59404bd6e32c4cdb267658511a46557dab2.tar.gz |
dma: separate out DMA enable status from wait_for_bytes
When wait_for_bytes returns 0 when DMA is disabled, we can't
differentiate between DMA being disabled and a transfer having completed
when it has reached the end of the requested transfer. Separating out
into separate functions lets us distinguish the two cases.
The reason we didn't hit this in the past is that the requested receive
size is generally larger than the actual amount we're sending. Since we
know the amount that we're waiting for from the header, we would stop
the transfer in software.
BRANCH=none
BUG=b:132444384
TEST=On DUTs with bloonchipper and dartmonkey:
ectool --name=cros_fp testmaxtransfer
TEST=make buildall -j
Change-Id: I885161a3e04b7a12d597d8dc8691f599990bda8b
Signed-off-by: Tom Hughes <tomhughes@chromium.org>
Reviewed-on: https://chromium-review.googlesource.com/c/chromiumos/platform/ec/+/1734010
Reviewed-by: Nicolas Boichat <drinkcat@chromium.org>
Diffstat (limited to 'chip/mchp')
-rw-r--r-- | chip/mchp/dma.c | 24 |
1 files changed, 14 insertions, 10 deletions
diff --git a/chip/mchp/dma.c b/chip/mchp/dma.c index 1c491729cd..6c9ed8dd47 100644 --- a/chip/mchp/dma.c +++ b/chip/mchp/dma.c @@ -194,28 +194,32 @@ void dma_xfr_start_rx(const struct dma_option *option, /* * Return the number of bytes transferred. - * The number of bytes transferred can be easily determinted + * The number of bytes transferred can be easily determined * from the difference in DMA memory start address register * and memory end address register. No need to look at DMA * transfer size field because the hardware increments memory - * start address by unit size on each unit tranferred. + * start address by unit size on each unit transferred. * Why is a signed integer being used for a count value? */ int dma_bytes_done(dma_chan_t *chan, int orig_count) { - int bcnt = 0; + int bcnt; - if (chan != NULL) { - if (chan->ctrl & MCHP_DMA_RUN) { - bcnt = (int)chan->mem_end; - bcnt -= (int)chan->mem_start; - bcnt = orig_count - bcnt; - } - } + if (chan == NULL) + return 0; + + bcnt = (int)chan->mem_end; + bcnt -= (int)chan->mem_start; + bcnt = orig_count - bcnt; return bcnt; } +bool dma_is_enabled(dma_chan_t *chan) +{ + return (chan->ctrl & MCHP_DMA_RUN); +} + int dma_bytes_done_chan(enum dma_channel ch, uint32_t orig_count) { uint32_t cnt; |