dma: edma: Provide granular residue accounting

As Joel pointed out, edma_read_position() uses memcpy_fromio() to read the parameter ram. That's not synchronized with the internal update as it does a byte by byte copy. We need to do a 32bit read to get a consistent value.

Further reading destination and source is pointless. In DEV_TO_MEM transfers we are only interested in the destination, in MEM_TO_DEV we care about the source. In MEM_TO_MEM it really does not matter which one you read.

Simple solution: Remove the pointers, select dest/source via a bool and return the read value.

Remove the export of this function while at it. The only potential user is the dmaengine and that's always builtin.

Signed-off-by Thomas Gleixner <[email protected]>

As Joel pointed out, edma_read_position() uses memcpy_fromio() to read the parameter ram. That's not synchronized with the internal update as it does a byte by byte copy. We need to do a 32bit read to get a consistent value.

Further reading destination and source is pointless. In DEV_TO_MEM transfers we are only interested in the destination, in MEM_TO_DEV we care about the source. In MEM_TO_MEM it really does not matter which one you read.

Simple solution: Remove the pointers, select dest/source via a bool and return the read value.

Remove the export of this function while at it. The only potential user is the dmaengine and that's always builtin.

Signed-off-by Thomas Gleixner <[email protected]>

A simpler version to provide granular residue accounting and readout for EDMA.

Delta to V1:

- Removed the double read of the address in PaRAM

- Simplified the stats update in the interrupt callback for intermediate transfers

Thanks,

tglx

Quoted 5 lines · Thomas Gleixner

A simpler version to provide granular residue accounting and readout for EDMA.

Delta to V1:

Quoted 9 lines

- Removed the double read of the address in PaRAM

- Simplified the stats update in the interrupt callback for intermediate transfers

Thanks,

tglx

Thanks for the series. I went over all the patches and it looks great.

Acked-by Joel Fernandes <[email protected]>

The patches however didn't apply and had some conflicts with my dma memcpy series and peter's cyclic series so I resolved conflicts and created a single branch based on Vinod's slave-dma next branch (commit 406efb1a745c1dc512dc9c3c859e302e7b7f907e) that Vinod can pull.

I also renamed subject line of patches in Thomas's series to be "dmaengine: edma" and documented some of the variables used.

https://github.com/joelagnel/linux-kernel.git (for-vinod branch)

Vinod, could you pull if it looks OK?

commit 770f0f3a20188b7e17db2790803b9da925dc0b94

Author: Thomas Gleixner <[email protected]>
Date:   Mon Apr 28 10:49:43 2014 +0000

dmaengine: edma: Make reading the position of active channels work

As Joel pointed out, edma_read_position() uses memcpy_fromio() to read the parameter ram. That's not synchronized with the internal update as it does a byte by byte copy. We need to do a 32bit read to get a consistent value.

Further reading destination and source is pointless. In DEV_TO_MEM transfers we are only interested in the destination, in MEM_TO_DEV we care about the source. In MEM_TO_MEM it really does not matter which one you read.

Simple solution: Remove the pointers, select dest/source via a bool and return the read value.

Remove the export of this function while at it. The only potential user is the dmaengine and that's always builtin.

Signed-off-by Thomas Gleixner <[email protected]>

You s-o-b missing in this one, also ack from Sekhar missing. Do you want to redo this or prefer me to cherry-pick patches adding acks and your s-o-b, since I already fetched your branch

Either way is fine with me...

On 04/29/2014 03:46 AM, Vinod Koul wrote:
[..]

Quoted 28 lines

commit 770f0f3a20188b7e17db2790803b9da925dc0b94

Author: Thomas Gleixner <[email protected]>
Date:   Mon Apr 28 10:49:43 2014 +0000

dmaengine: edma: Make reading the position of active channels work

As Joel pointed out, edma_read_position() uses memcpy_fromio() to read the parameter ram. That's not synchronized with the internal update as it does a byte by byte copy. We need to do a 32bit read to get a consistent value.

Further reading destination and source is pointless. In DEV_TO_MEM transfers we are only interested in the destination, in MEM_TO_DEV we care about the source. In MEM_TO_MEM it really does not matter which one you read.

Simple solution: Remove the pointers, select dest/source via a bool and return the read value.

Remove the export of this function while at it. The only potential user is the dmaengine and that's always builtin.

Signed-off-by Thomas Gleixner <[email protected]>

You s-o-b missing in this one, also ack from Sekhar missing. Do you want to redo this or prefer me to cherry-pick patches adding acks and your s-o-b, since I already fetched your branch

Quoted 2 lines

Either way is fine with me...

If its Ok with you, it would great if you could add my Ack and Sob. Thanks a lot. Let me know if you'd want me to do anything else here.

Quoted 30 lines · Joel Fernandes

On 04/29/2014 03:46 AM, Vinod Koul wrote:
[..]

Quoted 28 lines

commit 770f0f3a20188b7e17db2790803b9da925dc0b94

Author: Thomas Gleixner <[email protected]>
Date:   Mon Apr 28 10:49:43 2014 +0000

dmaengine: edma: Make reading the position of active channels work

As Joel pointed out, edma_read_position() uses memcpy_fromio() to read the parameter ram. That's not synchronized with the internal update as it does a byte by byte copy. We need to do a 32bit read to get a consistent value.

Further reading destination and source is pointless. In DEV_TO_MEM transfers we are only interested in the destination, in MEM_TO_DEV we care about the source. In MEM_TO_MEM it really does not matter which one you read.

Simple solution: Remove the pointers, select dest/source via a bool and return the read value.

Remove the export of this function while at it. The only potential user is the dmaengine and that's always builtin.

Signed-off-by Thomas Gleixner <[email protected]>

You s-o-b missing in this one, also ack from Sekhar missing. Do you want to redo this or prefer me to cherry-pick patches adding acks and your s-o-b, since I already fetched your branch

Quoted 5 lines

Quoted 2 lines

Either way is fine with me...

If its Ok with you, it would great if you could add my Ack and Sob. Thanks a lot. Let me know if you'd want me to do anything else here.

Done, The changes are getting pushed, pls do verify