Thread (26 messages) flat view 26 messages, 2 authors, 2011-08-23
STALE5486d

[PATCH v6 04/15] DMA: PL330: Add DMA_CYCLIC capability

From: Boojin Kim <hidden>
Date: 2011-08-23 07:08:27
Also in: linux-samsung-soc

Jassi Brar [mailto:jassisinghbrar at gmail.com]
Sent: Tuesday, August 23, 2011 2:42 PM
To: Boojin Kim
Cc: linux-arm-kernel at lists.infradead.org; linux-samsung-
soc at vger.kernel.org; Vinod Koul; Kukjin Kim; Dan Williams; Mark Brown;
Grant Likely; Russell King
Subject: Re: [PATCH v6 04/15] DMA: PL330: Add DMA_CYCLIC capability

On Mon, Aug 22, 2011 at 5:33 PM, Boojin Kim [off-list ref]
wrote:
quoted
quoted
quoted
+static struct dma_async_tx_descriptor *pl330_prep_dma_cyclic(
+               struct dma_chan *chan, dma_addr_t dma_addr, size_t
len,
quoted
quoted
quoted
+               size_t period_len, enum dma_data_direction
direction)
quoted
quoted
quoted
+{
+       struct dma_pl330_desc *desc;
+       struct dma_pl330_chan *pch = to_pchan(chan);
+       dma_addr_t dst;
+       dma_addr_t src;
+
+       desc = pl330_get_desc(pch);
+       if (!desc) {
+               dev_err(pch->dmac->pif.dev, "%s:%d Unable to fetch
desc\n",
quoted
+                       __func__, __LINE__);
+               return NULL;
+       }
+
+       switch (direction) {
+       case DMA_TO_DEVICE:
+               desc->rqcfg.src_inc = 1;
+               desc->rqcfg.dst_inc = 0;
+               src = dma_addr;
+               dst = pch->fifo_addr;
+               break;
+       case DMA_FROM_DEVICE:
+               desc->rqcfg.src_inc = 0;
+               desc->rqcfg.dst_inc = 1;
+               src = pch->fifo_addr;
+               dst = dma_addr;
+               break;
+       default:
+               dev_err(pch->dmac->pif.dev, "%s:%d Invalid dma
direction\n",
quoted
+               __func__, __LINE__);
+               return NULL;
+       }
+
+       desc->rqcfg.brst_size = pch->burst_sz;
+       desc->rqcfg.brst_len = 1;
+
+       if (!pch->cyclic)
+               pch->cyclic = CYCLIC_PREP;
The need for check here seems suspicious.
Is it really needed? If not, please remove it.
It's required because Cyclic operation is passed from CYCLIC_PREP to
CYCLIC_TRIGGER.
I don't see why you need to differentiate between PREP and TRIGGER in
cyclic mode. I think, you should simply have 'bool cyclic' instead of
enum.
On cyclic mode, Pl330_tasklet() operation is changed according to the value of 
'cyclic'.
In case of CYCLIC_PREP, Pl330_tasklet()operation is same with normal 
operation.
In case of CYCLIC_TRIGGER, pl330_tasklet()reloads the done data into 
work_list.
So, 'Cyclic' needs two status(CYCLIC_PREP and CYCLIC_TRIGGER) to distinguish 
the operation on pl330_tasklet().
The sequence is following.
device_prep_dma_cyclic() -> set CYCLIC_PREP -> device_issue_pending() -> 
Pl330_tasklet() with CYCLIC_PREP -> set CYCLIC_TRIGGER -> PL330 IRQ -> 
Pl330_tasklet() with CYCLIC_TRIGGER.

Thanks
Boojin
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help