dmaengine Archive on lore.kernel.org
 help / color / Atom feed
From: Peter Ujfalusi <peter.ujfalusi@ti.com>
To: <vkoul@kernel.org>
Cc: <dan.j.williams@intel.com>, <dmaengine@vger.kernel.org>,
	<linux-arm-kernel@lists.infradead.org>,
	<linux-omap@vger.kernel.org>
Subject: [PATCH v5 2/3] dmaengine: ti: edma: Correct the residue calculation (fix for memcpy)
Date: Tue, 16 Jul 2019 11:26:54 +0300
Message-ID: <20190716082655.1620-3-peter.ujfalusi@ti.com> (raw)
In-Reply-To: <20190716082655.1620-1-peter.ujfalusi@ti.com>

For memcpy we never stored the start address of the transfer for the pset
which rendered the memcpy residue calculation completely broken.

In the edma_residue() function we also need to to some correction for the
calculations:
Instead waiting for all EDMA channels to be idle (in a busy system it can
take few iteration to hit a point when all queues are idle) wait for the
event pending on the given channel (SH_ER for hw synchronized channels,
SH_ESR for manually triggered channels).

If the position returned by EMDA is 0 it imiplies that the last paRAM set
has been consumed and we are at the closing dummy set, thus we can conclude
that the transfer is completed and we can return 0 as residue.

Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
---
 drivers/dma/ti/edma.c | 31 +++++++++++++++++++++++--------
 1 file changed, 23 insertions(+), 8 deletions(-)

diff --git a/drivers/dma/ti/edma.c b/drivers/dma/ti/edma.c
index a39f817b3888..5b8cbd6d7610 100644
--- a/drivers/dma/ti/edma.c
+++ b/drivers/dma/ti/edma.c
@@ -1026,6 +1026,7 @@ static int edma_config_pset(struct dma_chan *chan, struct edma_pset *epset,
 		src_cidx = cidx;
 		dst_bidx = acnt;
 		dst_cidx = cidx;
+		epset->addr = src_addr;
 	} else {
 		dev_err(dev, "%s: direction not implemented yet\n", __func__);
 		return -EINVAL;
@@ -1736,7 +1737,11 @@ static u32 edma_residue(struct edma_desc *edesc)
 	int loop_count = EDMA_MAX_TR_WAIT_LOOPS;
 	struct edma_chan *echan = edesc->echan;
 	struct edma_pset *pset = edesc->pset;
-	dma_addr_t done, pos;
+	dma_addr_t done, pos, pos_old;
+	int channel = EDMA_CHAN_SLOT(echan->ch_num);
+	int idx = EDMA_REG_ARRAY_INDEX(channel);
+	int ch_bit = EDMA_CHANNEL_BIT(channel);
+	int event_reg;
 	int i;
 
 	/*
@@ -1749,16 +1754,20 @@ static u32 edma_residue(struct edma_desc *edesc)
 	 * "pos" may represent a transfer request that is still being
 	 * processed by the EDMACC or EDMATC. We will busy wait until
 	 * any one of the situations occurs:
-	 *   1. the DMA hardware is idle
-	 *   2. a new transfer request is setup
+	 *   1. while and event is pending for the channel
+	 *   2. a position updated
 	 *   3. we hit the loop limit
 	 */
-	while (edma_read(echan->ecc, EDMA_CCSTAT) & EDMA_CCSTAT_ACTV) {
-		/* check if a new transfer request is setup */
-		if (edma_get_position(echan->ecc,
-				      echan->slot[0], dst) != pos) {
+	if (is_slave_direction(edesc->direction))
+		event_reg = SH_ER;
+	else
+		event_reg = SH_ESR;
+
+	pos_old = pos;
+	while (edma_shadow0_read_array(echan->ecc, event_reg, idx) & ch_bit) {
+		pos = edma_get_position(echan->ecc, echan->slot[0], dst);
+		if (pos != pos_old)
 			break;
-		}
 
 		if (!--loop_count) {
 			dev_dbg_ratelimited(echan->vchan.chan.device->dev,
@@ -1783,6 +1792,12 @@ static u32 edma_residue(struct edma_desc *edesc)
 		return edesc->residue_stat;
 	}
 
+	/*
+	 * If the position is 0, then EDMA loaded the closing dummy slot, the
+	 * transfer is completed
+	 */
+	if (!pos)
+		return 0;
 	/*
 	 * For SG operation we catch up with the last processed
 	 * status.
-- 
Peter

Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki


  parent reply index

Thread overview: 10+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2019-07-16  8:26 [PATCH v5 0/3] dmaengine: ti: edma: Polled completion support Peter Ujfalusi
2019-07-16  8:26 ` [PATCH v5 1/3] dmaengine: ti: edma: Clean up the 2x32bit array register accesses Peter Ujfalusi
2019-07-16  8:26 ` Peter Ujfalusi [this message]
2019-07-25 13:46   ` [PATCH v5 2/3] dmaengine: ti: edma: Correct the residue calculation (fix for memcpy) Vinod Koul
2019-07-16  8:26 ` [PATCH v5 3/3] dmaengine: ti: edma: Support for polled (memcpy) completion Peter Ujfalusi
2019-07-29  6:42 ` [PATCH v5 0/3] dmaengine: ti: edma: Polled completion support Vinod Koul
2019-07-29  7:22   ` Peter Ujfalusi
2019-07-29  7:33     ` Vinod Koul
2019-07-29  7:24   ` Peter Ujfalusi
2019-07-29  7:33     ` Vinod Koul

Reply instructions:

You may reply publically to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20190716082655.1620-3-peter.ujfalusi@ti.com \
    --to=peter.ujfalusi@ti.com \
    --cc=dan.j.williams@intel.com \
    --cc=dmaengine@vger.kernel.org \
    --cc=linux-arm-kernel@lists.infradead.org \
    --cc=linux-omap@vger.kernel.org \
    --cc=vkoul@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link

dmaengine Archive on lore.kernel.org

Archives are clonable:
	git clone --mirror https://lore.kernel.org/dmaengine/0 dmaengine/git/0.git

	# If you have public-inbox 1.1+ installed, you may
	# initialize and index your mirror using the following commands:
	public-inbox-init -V2 dmaengine dmaengine/ https://lore.kernel.org/dmaengine \
		dmaengine@vger.kernel.org
	public-inbox-index dmaengine

Example config snippet for mirrors

Newsgroup available over NNTP:
	nntp://nntp.lore.kernel.org/org.kernel.vger.dmaengine


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git