Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > linux.kernel > #1674750 > unrolled thread

[PATCH 17/20] dt-bindings: serial: stm32: add dma using note

Back to article view | Back to linux.kernel

This discussion starts older than the indexed window; earlier articles aren't shown. The article labeled Started by below is the oldest one visible, not the original post.

Contents

  [PATCH 17/20] dt-bindings: serial: stm32: add dma using note Bich HEMON <bich.hemon@st.com> - 2017-06-26 15:00 +0200
    Re: [PATCH 17/20] dt-bindings: serial: stm32: add dma using note Rob Herring <robh@kernel.org> - 2017-06-29 00:30 +0200
    Re: [PATCH 17/20] dt-bindings: serial: stm32: add dma using note Greg Kroah-Hartman <gregkh@linuxfoundation.org> - 2017-06-29 17:00 +0200

#1674750 — [PATCH 17/20] dt-bindings: serial: stm32: add dma using note

From: Bich Hemon <bich.hemon@st.com>

Signed-off-by: Gerald Baeza <gerald.baeza@st.com>
---
 .../devicetree/bindings/serial/st,stm32-usart.txt  | 22 ++++++++++++++++++++++
 1 file changed, 22 insertions(+)

diff --git a/Documentation/devicetree/bindings/serial/st,stm32-usart.txt b/Documentation/devicetree/bindings/serial/st,stm32-usart.txt
index 3b42138..6c1cfc8 100644
--- a/Documentation/devicetree/bindings/serial/st,stm32-usart.txt
+++ b/Documentation/devicetree/bindings/serial/st,stm32-usart.txt
@@ -18,6 +18,28 @@ Optional properties:
 - dmas: phandle(s) to DMA controller node(s). Refer to stm32-dma.txt
 - dma-names: "rx" and/or "tx"
 
+Note for dma using:
+- "tx" dma can be used without any constraint since it uses single
+dma transfers.
+- "rx" dma using requires some attention:
+   1) if you cannot anticipate the length of your received packets
+   and if your usart device embeds an internal fifo, then DON'T use
+   dma mode.
+   2) if you enable dma mode WITHOUT mdma intermediate copy (cf.
+   stm32-dma.txt), then the availability of the received data will
+   depend on the dma driver policy and it may be delayed until dma
+   internal fifo is full. The usart driver will see this checking
+   the dma residue when rx interrupt (RXNE or RTO) occurs.
+   3) if you enable dma mode WITH mdma intermediate copy (cf.
+   stm32-dma.txt) then the usart driver will never see the dma
+   residue becoming smaller than RX_BUF_P but it will get its
+   rx dma complete callback called when the cyclic transfer period
+   (RX_BUF_P) is reached.
+The three possibilities above are ordered from the most cpu time
+consuming one to the least one. The counterpart of this optimisation
+is the reception granularity achievable by the usart driver, from
+one byte up to RX_BUF_P.
+
 Note: Each usart controller should have an alias correctly numbered
 in "aliases" node.
 
-- 
1.9.1

[toc] | [next] | [standalone]

#1677240

On Mon, Jun 26, 2017 at 12:49:16PM +0000, Bich HEMON wrote:
> From: Bich Hemon <bich.hemon@st.com>
> 
> Signed-off-by: Gerald Baeza <gerald.baeza@st.com>
> ---
>  .../devicetree/bindings/serial/st,stm32-usart.txt  | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)

Sounds like broken DMA to me.

Acked-by: Rob Herring <robh@kernel.org>

[toc] | [prev] | [next] | [standalone]

#1677835

On Mon, Jun 26, 2017 at 12:49:16PM +0000, Bich HEMON wrote:
> From: Bich Hemon <bich.hemon@st.com>
> 
> Signed-off-by: Gerald Baeza <gerald.baeza@st.com>

I can't take patches without any changelog text at all, sorry.

Also, you didn't sign off on this?  Why not???

[toc] | [prev] | [standalone]

Back to top | Article view | linux.kernel

csiph-web