by Zhangfei Gao

[permalink] [raw]

Hi,

On Thu, Oct 24, 2013 at 11:43 PM, Guennadi Liakhovetski
<[email protected]> wrote:
> On Fri, 25 Oct 2013, Vinod Koul wrote:
>
>> On Fri, Oct 25, 2013 at 08:32:12AM +0200, Guennadi Liakhovetski wrote:
>> > Hi Vinod
>> >
>> > On Fri, 25 Oct 2013, Vinod Koul wrote:
>> >
>> > > On Thu, Oct 24, 2013 at 11:28:29PM +0200, Guennadi Liakhovetski wrote:
>> > > > Hi Vinod
>> > > >
>> > > > On Thu, 17 Oct 2013, Vinod Koul wrote:
>> > > > > Yes i missed it in first place update the patch to fix that
>> > > >
>> > > > Are you planning to post a fixed version of this patch or you just fix it
>> > > > internally? Would be good to have it posted to be able to ack it and other
>> > > > relevant patches.
>> > > looks like you missed it... I had posted updated patch [1] in this thread here
>> > > and I posted 29th patch as removal one [2]
>> >
>> > No, I didn't miss those, but as Sebastian pointed out and as I commented
>> > too, also that v2 version wasn't correct, so, a fixed v3 was needed.
>> > Consider this:
>> >
>> > In patch 1 you do:
>> >
>> > diff --git a/include/linux/dmaengine.h b/include/linux/dmaengine.h
>> > index 0bc7275..683c380 100644
>> > --- a/include/linux/dmaengine.h
>> > +++ b/include/linux/dmaengine.h
>> > @@ -45,16 +45,17 @@ static inline int dma_submit_error(dma_cookie_t cookie)
>> >
>> > /**
>> > * enum dma_status - DMA transaction status
>> > - * @DMA_SUCCESS: transaction completed successfully
>> > + * @DMA_COMPLETE: transaction completed
>> > * @DMA_IN_PROGRESS: transaction not yet processed
>> > * @DMA_PAUSED: transaction is paused
>> > * @DMA_ERROR: transaction failed
>> > */
>> > enum dma_status {
>> > - DMA_SUCCESS,
>> > + DMA_COMPLETE,
>> > DMA_IN_PROGRESS,
>> > DMA_PAUSED,
>> > DMA_ERROR,
>> > + DMA_SUCCESS,
>> > };
>> >
>> > /**
>> >
>> > and then in a couple of places
>> >
>> > - return DMA_SUCCESS;
>> > + return DMA_COMPLETE;
>> >
>> > So, after that your patch dmaengine would be returning DMA_COMPLETE in
>> > case of success, i.e. 0. But all the DMAC and user drivers would still be
>> > checking for
>> >
>> > if (status != DMA_COMPLETE) {
>> >
>> > i.e. comparing status with 4 and thus detecting false errors, until your
>> > further 28 patches fix them. That's why, as Sebastian pointed out it was
>> > important to define DMA_COMPLETE and DMA_SUCCESS with the _same_ numerical
>> > value in your patch 1.
>> Sure, I will fix this up now as suggested
>>
>> enum dma_status {
>> - DMA_SUCCESS,
>> + DMA_COMPLETE = 0, DMA_SUCCESS = 0,
>>
>> Thanks for poiting out.
>
> Great! In my reply to Sebastian I proposed an equivlent but slightly
> different version:
>
> enum dma_status {
> - DMA_SUCCESS,
> + DMA_COMPLETE,
> DMA_IN_PROGRESS,
> DMA_PAUSED,
> DMA_ERROR,
> };
>
> + #define DMA_SUCCESS DMA_COMPLETE
>
> but it doesn't really matter. Feel free to use whichever version you
> prefer.

This patch broke some TI platforms (using edma) due to namespace
conflicts. The error is:

../../include/linux/dmaengine.h:54:2: error: expected identifier
before numeric constant (for DMA_COMPLETE)

.. it's because they define DMA_COMPLETE in include/platform_data/edma.h.

The edma driver seems to overall lack prefixes and thus pollutes
global namespaces, so it should be fixed up.

-Olof

2013-10-30 13:58:02

by Vinod Koul

[permalink] [raw]

Subject: Re: [PATCH 01/28] dmaengine: use DMA_COMPLETE for dma completion status

On Tue, Oct 29, 2013 at 09:47:55AM -0700, Olof Johansson wrote:
> This patch broke some TI platforms (using edma) due to namespace
> conflicts. The error is:
>
> ../../include/linux/dmaengine.h:54:2: error: expected identifier
> before numeric constant (for DMA_COMPLETE)
>
> .. it's because they define DMA_COMPLETE in include/platform_data/edma.h.
>
> The edma driver seems to overall lack prefixes and thus pollutes
> global namespaces, so it should be fixed up.
Yup this needs some work.

I fixed up the error with below, can you pls verify. Pls note this is compile
tested only!

-----><8---------------------------

From: Vinod Koul <[email protected]>
Date: Wed, 30 Oct 2013 18:22:30 +0530
Subject: [PATCH] dmanengine: fix edma driver to not define DMA_COMPLETE

edma header defines DMA_COMPLETE, this causes issues as commit adfedd9a32e4 move
DMA_SUCCESS to DMA_COMPLETE. edma should properly namespace its defines and
needs a future fix

Reported-by: Olof Johansson <[email protected]>
Signed-off-by: Vinod Koul <[email protected]>
---
arch/arm/common/edma.c | 4 ++--
drivers/dma/edma.c | 4 ++--
include/linux/platform_data/edma.h | 8 ++++----
3 files changed, 8 insertions(+), 8 deletions(-)

diff --git a/arch/arm/common/edma.c b/arch/arm/common/edma.c
index 8e1a024..41bca32 100644
--- a/arch/arm/common/edma.c
+++ b/arch/arm/common/edma.c
@@ -404,7 +404,7 @@ static irqreturn_t dma_irq_handler(int irq, void *data)
BIT(slot));
if (edma_cc[ctlr]->intr_data[channel].callback)
edma_cc[ctlr]->intr_data[channel].callback(
- channel, DMA_COMPLETE,
+ channel, EDMA_DMA_COMPLETE,
edma_cc[ctlr]->intr_data[channel].data);
}
} while (sh_ipr);
@@ -459,7 +459,7 @@ static irqreturn_t dma_ccerr_handler(int irq, void *data)
callback) {
edma_cc[ctlr]->intr_data[k].
callback(k,
- DMA_CC_ERROR,
+ EDMA_DMA_CC_ERROR,
edma_cc[ctlr]->intr_data
[k].data);
}
diff --git a/drivers/dma/edma.c b/drivers/dma/edma.c
index 29fa358..cb69f3a 100644
--- a/drivers/dma/edma.c
+++ b/drivers/dma/edma.c
@@ -407,7 +407,7 @@ static void edma_callback(unsigned ch_num, u16 ch_status, void *data)
edma_pause(echan->ch_num);

switch (ch_status) {
- case DMA_COMPLETE:
+ case EDMA_DMA_COMPLETE:
spin_lock_irqsave(&echan->vchan.lock, flags);

edesc = echan->edesc;
@@ -426,7 +426,7 @@ static void edma_callback(unsigned ch_num, u16 ch_status, void *data)
spin_unlock_irqrestore(&echan->vchan.lock, flags);

break;
- case DMA_CC_ERROR:
+ case EDMA_DMA_CC_ERROR:
spin_lock_irqsave(&echan->vchan.lock, flags);

edma_read_slot(EDMA_CHAN_SLOT(echan->slot[0]), &p);
diff --git a/include/linux/platform_data/edma.h b/include/linux/platform_data/edma.h
index 179fb91..f50821c 100644
--- a/include/linux/platform_data/edma.h
+++ b/include/linux/platform_data/edma.h
@@ -67,10 +67,10 @@ struct edmacc_param {
#define ITCCHEN BIT(23)

/*ch_status paramater of callback function possible values*/
-#define DMA_COMPLETE 1
-#define DMA_CC_ERROR 2
-#define DMA_TC1_ERROR 3
-#define DMA_TC2_ERROR 4
+#define EDMA_DMA_COMPLETE 1
+#define EDMA_DMA_CC_ERROR 2
+#define EDMA_DMA_TC1_ERROR 3
+#define EDMA_DMA_TC2_ERROR 4

enum address_mode {
INCR = 0,
--
1.7.0.4

--
~Vinod