Skip to contentSkip to conversation

IMX6 flexcan: Change serial clock_feq source from IPG clock to PER clock

PATCH New Belser Florian 3 participants 3 messages Nov 28, 2013 – Nov 29, 2013 1 message not in the archive CAN network drivers

lore
Share & tools
lore

A message that is not in the archive (sent off-list or not yet received): [email protected]

#1 lore raw

PATCH 1/1 IMX6 flexcan: Change serial clock_feq source from IPG clock to PER clock

Hi,

think I found a bug in the flexcan driver. Currently the serial clock frequency for bitrate calculation is derived from IPG clock.
According to the imx6d/q reference manual, the CAN serial clock is derived from can_root clock (PER clock in dtb). The patch below fixes this issue. However, I'm not sure if the patch below will also work for the rest of the imx family. Does anybody know if the CAN serial clock is derived from IPG clock in other imx processors? Is there a case distinction required to keep backward compatibility?

Be aware that getting the flexcan module running on the imx6 also requires patch "[PATCH] IMX6: Add static /8 prescaler to can_root clock configuration"

Here is my patch:

drivers/net/can/flexcan.c 2013-11-28 09:34:01.539469212 +0100 renamed from drivers/net/can/flexcan.c.orig 2013-11-20 21:28:01.000000000 +0100 +1 -1
@@ -1031,7 +1031,6 @@ static int flexcan_probe(struct platform
             err = PTR_ERR(clk_ipg);
             goto failed_clock;
         }
-        clock_freq = clk_get_rate(clk_ipg);
 
         clk_per = devm_clk_get(&pdev->dev, "per");
         if (IS_ERR(clk_per)) {
@@ -1039,6 +1038,7 @@ static int flexcan_probe(struct platform
             err = PTR_ERR(clk_per);
             goto failed_clock;
         }
+        clock_freq = clk_get_rate(clk_per);
     }
 
     mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);

----------------------------------------

In the same file there is another bug which only occurs in combination with an imx6 processor.
Within the function "static int flexcan_chip_start(struct net_device *dev)" all message boxes will be initialized.
Think that isn't correct for the IMX6 processor.
In the newest kernel v3.12.r1 the problem is already fixed and I'm not sure if the fix will be automatically added to the LTE kernel version 3.10.20 ? If not, shall I create a patch for it ?

Best regards
Florian Belser

#2 lore raw

PATCH 1/1 IMX6 flexcan: Change serial clock_feq source from IPG clock to PER clock

Quoted 4 lines · Belser Florian

Hi,

think I found a bug in the flexcan driver. Currently the serial clock frequency for bitrate calculation is derived from IPG clock. According to the imx6d/q reference manual, the CAN serial clock is derived from can_root clock (PER clock in dtb). The patch below fixes this issue.

Marc already had a patch for it.

http://www.spinics.net/lists/arm-kernel/msg288971.html

Shawn

Quoted 37 lines

However, I'm not sure if the patch below will also work for the rest of the imx family. Does anybody know if the CAN serial clock is derived from IPG clock in other imx processors? Is there a case distinction required to keep backward compatibility?

Be aware that getting the flexcan module running on the imx6 also requires patch "[PATCH] IMX6: Add static /8 prescaler to can_root clock configuration"

Here is my patch:

drivers/net/can/flexcan.c 2013-11-28 09:34:01.539469212 +0100 renamed from drivers/net/can/flexcan.c.orig 2013-11-20 21:28:01.000000000 +0100 +1 -2
@@ -1031,7 +1031,6 @@ static int flexcan_probe(struct platform
             err = PTR_ERR(clk_ipg);
             goto failed_clock;
         }
-        clock_freq = clk_get_rate(clk_ipg);
 
         clk_per = devm_clk_get(&pdev->dev, "per");
         if (IS_ERR(clk_per)) {
@@ -1039,6 +1038,7 @@ static int flexcan_probe(struct platform
             err = PTR_ERR(clk_per);
             goto failed_clock;
         }
+        clock_freq = clk_get_rate(clk_per);
     }
 
     mem = platform_get_resource(pdev, IORESOURCE_MEM, 0);
 
----------------------------------------

In the same file there is another bug which only occurs in combination with an imx6 processor.
Within the function "static int flexcan_chip_start(struct net_device *dev)" all message boxes will be initialized.
Think that isn't correct for the IMX6 processor.
In the newest kernel v3.12.r1 the problem is already fixed and I'm not sure if the fix will be automatically added to the LTE kernel version 3.10.20 ? If not, shall I create a patch for it ?

Best regards
Florian Belser

#3 lore raw

PATCH 1/1 IMX6 flexcan: Change serial clock_feq source from IPG clock to PER clock

Quoted 9 lines · Shawn Guo
Quoted 5 lines

think I found a bug in the flexcan driver. Currently the serial clock frequency for bitrate calculation is derived from IPG clock. According to the imx6d/q reference manual, the CAN serial clock is derived from can_root clock (PER clock in dtb). The patch below fixes this issue.

Marc already had a patch for it.

http://www.spinics.net/lists/arm-kernel/msg288971.html

Ack, it's already in David Miller's net/master and will hit Linus tree soon, stable in on Cc.

Quoted 4 lines
Quoted 4 lines

However, I'm not sure if the patch below will also work for the rest of the imx family. Does anybody know if the CAN serial clock is derived from IPG clock in other imx processors? Is there a case distinction required to keep backward compatibility?

All other processors use the same rate, for now, for both clocks. However this will change soon on the mx53, once my patch in mainline.

Quoted 3 lines
Quoted 3 lines

Be aware that getting the flexcan module running on the imx6 also requires patch "[PATCH] IMX6: Add static /8 prescaler to can_root clock configuration"

That problem has already been fixed with:

9b3d423 ARM: i.MX6q: fix the wrong parent of can_root clock

and I've asked Greg to queue this patch for stable, too:

http://www.spinics.net/lists/stable/msg26282.html

Quoted 7 lines
Quoted 7 lines

In the same file there is another bug which only occurs in combination with an imx6 processor. Within the function "static int flexcan_chip_start(struct net_device *dev)" all message boxes will be initialized. Think that isn't correct for the IMX6 processor. In the newest kernel v3.12.r1 the problem is already fixed and I'm not sure if the fix will be automatically added to the LTE kernel version 3.10.20 ? If not, shall I create a patch for it ?

As there already is a patch for the problem, I think you mean:

0d1862e can: flexcan: fix flexcan_chip_start() on imx6

There was a bit of confusion about this patch, as it creates a regression that was fixes by the later patch:

d5a7b40 can: flexcan: flexcan_chip_start: fix regression, mark one
MB for TX and abort pending TX

It seems that d5a7b40 got applied but 0d1862e not. I'll ask Greg to queue the missing patch ("0d1862e") to stable.

Marc

Signature · 12 lines
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 259 bytes
Desc: OpenPGP digital signature
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20131129/b6b45f8e/attachment.sig>