[PATCH] mtd: onenand: omap2: Pass correct flags for prep_dma_memcpy

[PATCH] mtd: onenand: omap2: Pass correct flags for prep_dma_memcpy

[Peter Ujfalusi]

The commit converting the driver to DMAengine was missing the flags for
the memcpy prepare call.
It went unnoticed since the omap-dma drive was ignoring them.

Fixes: 3ed6a4d1de2c5 (" mtd: onenand: omap2: Convert to use dmaengine for memcp")
Reported-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
---
Hi,

Aaro reported [1] a failure on omap2-onenand pointing to
4689d35c765c696bdf0535486a990038b242a26b. It looks like the root cause is the
conversion of omap2-onenand to DMAengine which missed the flags.

Basically the client is waiting for a callback without asking for it. This
certainly causes timeout.

I have not tested the patch, but it should fix the issue.

[1] https://lore.kernel.org/lkml/20200103081726.GD15023@darkstar.musicnaut.iki.fi/

 drivers/mtd/nand/onenand/omap2.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/mtd/nand/onenand/omap2.c b/drivers/mtd/nand/onenand/omap2.c
index edf94ee54ec7..71a632b815aa 100644
--- a/drivers/mtd/nand/onenand/omap2.c
+++ b/drivers/mtd/nand/onenand/omap2.c
@@ -328,7 +328,8 @@ static inline int omap2_onenand_dma_transfer(struct omap2_onenand *c,
 	struct dma_async_tx_descriptor *tx;
 	dma_cookie_t cookie;
 
-	tx = dmaengine_prep_dma_memcpy(c->dma_chan, dst, src, count, 0);
+	tx = dmaengine_prep_dma_memcpy(c->dma_chan, dst, src, count,
+				       DMA_CTRL_ACK | DMA_PREP_INTERRUPT);
 	if (!tx) {
 		dev_err(&c->pdev->dev, "Failed to prepare DMA memcpy\n");
 		return -EIO;
-- 
Peter

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

Re: [PATCH] mtd: onenand: omap2: Pass correct flags for prep_dma_memcpy

[H. Nikolaus Schaller]

> Am 04.01.2020 um 08:34 schrieb Peter Ujfalusi <peter.ujfalusi@ti.com>:
> 
> The commit converting the driver to DMAengine was missing the flags for
> the memcpy prepare call.
> It went unnoticed since the omap-dma drive was ignoring them.

s/drive/driver/

> 
> Fixes: 3ed6a4d1de2c5 (" mtd: onenand: omap2: Convert to use dmaengine for memcp")
> Reported-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
> ---
> Hi,
> 
> Aaro reported [1] a failure on omap2-onenand pointing to
> 4689d35c765c696bdf0535486a990038b242a26b. It looks like the root cause is the
> conversion of omap2-onenand to DMAengine which missed the flags.
> 
> Basically the client is waiting for a callback without asking for it. This
> certainly causes timeout.
> 
> I have not tested the patch, but it should fix the issue.

makes the gta04a5one.dts work again.

Tested-by: H. Nikolaus Schaller <hns@goldelico.com>

BR and thanks!
Nikolaus

> 
> [1] https://lore.kernel.org/lkml/20200103081726.GD15023@darkstar.musicnaut.iki.fi/
> 
> drivers/mtd/nand/onenand/omap2.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/mtd/nand/onenand/omap2.c b/drivers/mtd/nand/onenand/omap2.c
> index edf94ee54ec7..71a632b815aa 100644
> --- a/drivers/mtd/nand/onenand/omap2.c
> +++ b/drivers/mtd/nand/onenand/omap2.c
> @@ -328,7 +328,8 @@ static inline int omap2_onenand_dma_transfer(struct omap2_onenand *c,
> 	struct dma_async_tx_descriptor *tx;
> 	dma_cookie_t cookie;
> 
> -	tx = dmaengine_prep_dma_memcpy(c->dma_chan, dst, src, count, 0);
> +	tx = dmaengine_prep_dma_memcpy(c->dma_chan, dst, src, count,
> +				       DMA_CTRL_ACK | DMA_PREP_INTERRUPT);
> 	if (!tx) {
> 		dev_err(&c->pdev->dev, "Failed to prepare DMA memcpy\n");
> 		return -EIO;
> -- 
> Peter
> 
> Texas Instruments Finland Oy, Porkkalankatu 22, 00180 Helsinki.
> Y-tunnus/Business ID: 0615521-4. Kotipaikka/Domicile: Helsinki
> 

Re: [PATCH] mtd: onenand: omap2: Pass correct flags for prep_dma_memcpy

[Aaro Koskinen]

Hi,

On Sat, Jan 04, 2020 at 09:34:53AM +0200, Peter Ujfalusi wrote:
> The commit converting the driver to DMAengine was missing the flags for
> the memcpy prepare call.
> It went unnoticed since the omap-dma drive was ignoring them.
> 
> Fixes: 3ed6a4d1de2c5 (" mtd: onenand: omap2: Convert to use dmaengine for memcp")
> Reported-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> Signed-off-by: Peter Ujfalusi <peter.ujfalusi@ti.com>
> ---
> Hi,
> 
> Aaro reported [1] a failure on omap2-onenand pointing to
> 4689d35c765c696bdf0535486a990038b242a26b. It looks like the root cause is the
> conversion of omap2-onenand to DMAengine which missed the flags.
> 
> Basically the client is waiting for a callback without asking for it. This
> certainly causes timeout.
> 
> I have not tested the patch, but it should fix the issue.

This fixes the issue on Nokia boards.

Tested-by: Aaro Koskinen <aaro.koskinen@iki.fi>

Thanks,

A.