[PATCH] usb: dwc2: Read GNPTXFSIZ when in forced HOST mode.

The documentation for GNPTXFSIZ says that "For host mode, this field
is always valid."  Since we're already switching to host mode for
HPTXFSIZ, let's also read GNPTXFSIZ in host mode.

On an rk3288 SoC, without this change we see this at bootup:
  dwc2 ff580000.usb: gnptxfsiz=00100400
  dwc2 ff580000.usb: 128 invalid for host_nperio_tx_fifo_size. Check HW configuration.

After this change we see:
  dwc2 ff580000.usb: gnptxfsiz=04000400

Signed-off-by: Doug Anderson <dianders@chromium.org>
---
 drivers/usb/dwc2/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
index 27d2c9b..c184ed43 100644
--- a/drivers/usb/dwc2/core.c
+++ b/drivers/usb/dwc2/core.c

@@ -2674,23 +2674,23 @@ int dwc2_get_hwparams(struct dwc2_hsotg *hsotg)
 	hwcfg2 = readl(hsotg->regs + GHWCFG2);
 	hwcfg3 = readl(hsotg->regs + GHWCFG3);
 	hwcfg4 = readl(hsotg->regs + GHWCFG4);
-	gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
 	grxfsiz = readl(hsotg->regs + GRXFSIZ);
 
 	dev_dbg(hsotg->dev, "hwcfg1=%08x\n", hwcfg1);
 	dev_dbg(hsotg->dev, "hwcfg2=%08x\n", hwcfg2);
 	dev_dbg(hsotg->dev, "hwcfg3=%08x\n", hwcfg3);
 	dev_dbg(hsotg->dev, "hwcfg4=%08x\n", hwcfg4);
-	dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
 	dev_dbg(hsotg->dev, "grxfsiz=%08x\n", grxfsiz);
 
-	/* Force host mode to get HPTXFSIZ exact power on value */
+	/* Force host mode to get HPTXFSIZ / GNPTXFSIZ exact power on value */
 	gusbcfg = readl(hsotg->regs + GUSBCFG);
 	gusbcfg |= GUSBCFG_FORCEHOSTMODE;
 	writel(gusbcfg, hsotg->regs + GUSBCFG);
 	usleep_range(100000, 150000);
 
+	gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
 	hptxfsiz = readl(hsotg->regs + HPTXFSIZ);
+	dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
 	dev_dbg(hsotg->dev, "hptxfsiz=%08x\n", hptxfsiz);
 	gusbcfg = readl(hsotg->regs + GUSBCFG);
 	gusbcfg &= ~GUSBCFG_FORCEHOSTMODE;

-- 
2.0.0.526.g5318336

Nice! I wonder if this is a bug in the original driver, and they
actually meant to read this register instead of HPTXFSIZ? Well, it
doesn't really matter I guess.

Acked-by: Paul Zimmerman <redacted>

You may want to resend this to Greg after -rc1 is out and he reopens
his usb-next tree.

-- 
Paul

Paul,

On Thu, Aug 7, 2014 at 1:18 PM, Paul Zimmerman
[off-list ref] wrote:

quoted

From: Doug Anderson [mailto:dianders at chromium.org]
Sent: Thursday, August 07, 2014 12:48 PM

The documentation for GNPTXFSIZ says that "For host mode, this field
is always valid."  Since we're already switching to host mode for
HPTXFSIZ, let's also read GNPTXFSIZ in host mode.

On an rk3288 SoC, without this change we see this at bootup:
  dwc2 ff580000.usb: gnptxfsiz=00100400
  dwc2 ff580000.usb: 128 invalid for host_nperio_tx_fifo_size. Check HW configuration.

After this change we see:
  dwc2 ff580000.usb: gnptxfsiz=04000400

Signed-off-by: Doug Anderson <dianders@chromium.org>
---
 drivers/usb/dwc2/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
index 27d2c9b..c184ed43 100644
--- a/drivers/usb/dwc2/core.c
+++ b/drivers/usb/dwc2/core.c

@@ -2674,23 +2674,23 @@ int dwc2_get_hwparams(struct dwc2_hsotg *hsotg)
      hwcfg2 = readl(hsotg->regs + GHWCFG2);
      hwcfg3 = readl(hsotg->regs + GHWCFG3);
      hwcfg4 = readl(hsotg->regs + GHWCFG4);
-     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      grxfsiz = readl(hsotg->regs + GRXFSIZ);

      dev_dbg(hsotg->dev, "hwcfg1=%08x\n", hwcfg1);
      dev_dbg(hsotg->dev, "hwcfg2=%08x\n", hwcfg2);
      dev_dbg(hsotg->dev, "hwcfg3=%08x\n", hwcfg3);
      dev_dbg(hsotg->dev, "hwcfg4=%08x\n", hwcfg4);
-     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "grxfsiz=%08x\n", grxfsiz);

-     /* Force host mode to get HPTXFSIZ exact power on value */
+     /* Force host mode to get HPTXFSIZ / GNPTXFSIZ exact power on value */
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg |= GUSBCFG_FORCEHOSTMODE;
      writel(gusbcfg, hsotg->regs + GUSBCFG);
      usleep_range(100000, 150000);

+     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      hptxfsiz = readl(hsotg->regs + HPTXFSIZ);
+     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "hptxfsiz=%08x\n", hptxfsiz);
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg &= ~GUSBCFG_FORCEHOSTMODE;

Nice! I wonder if this is a bug in the original driver, and they
actually meant to read this register instead of HPTXFSIZ? Well, it
doesn't really matter I guess.

Acked-by: Paul Zimmerman <redacted>

You may want to resend this to Greg after -rc1 is out and he reopens
his usb-next tree.

Thanks!

...do you think this is a fix that needs to go in for 3.17?  Is it
affecting anyone there?

-Doug

From: dianders at google.com [mailto:dianders at google.com] On Behalf Of Doug Anderson
Sent: Thursday, August 07, 2014 5:12 PM

On Thu, Aug 7, 2014 at 1:18 PM, Paul Zimmerman
[off-list ref] wrote:

quoted

quoted

From: Doug Anderson [mailto:dianders at chromium.org]
Sent: Thursday, August 07, 2014 12:48 PM

The documentation for GNPTXFSIZ says that "For host mode, this field
is always valid."  Since we're already switching to host mode for
HPTXFSIZ, let's also read GNPTXFSIZ in host mode.

On an rk3288 SoC, without this change we see this at bootup:
  dwc2 ff580000.usb: gnptxfsiz=00100400
  dwc2 ff580000.usb: 128 invalid for host_nperio_tx_fifo_size. Check HW configuration.

After this change we see:
  dwc2 ff580000.usb: gnptxfsiz=04000400

Signed-off-by: Doug Anderson <dianders@chromium.org>
---
 drivers/usb/dwc2/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
index 27d2c9b..c184ed43 100644
--- a/drivers/usb/dwc2/core.c
+++ b/drivers/usb/dwc2/core.c

@@ -2674,23 +2674,23 @@ int dwc2_get_hwparams(struct dwc2_hsotg *hsotg)
      hwcfg2 = readl(hsotg->regs + GHWCFG2);
      hwcfg3 = readl(hsotg->regs + GHWCFG3);
      hwcfg4 = readl(hsotg->regs + GHWCFG4);
-     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      grxfsiz = readl(hsotg->regs + GRXFSIZ);

      dev_dbg(hsotg->dev, "hwcfg1=%08x\n", hwcfg1);
      dev_dbg(hsotg->dev, "hwcfg2=%08x\n", hwcfg2);
      dev_dbg(hsotg->dev, "hwcfg3=%08x\n", hwcfg3);
      dev_dbg(hsotg->dev, "hwcfg4=%08x\n", hwcfg4);
-     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "grxfsiz=%08x\n", grxfsiz);

-     /* Force host mode to get HPTXFSIZ exact power on value */
+     /* Force host mode to get HPTXFSIZ / GNPTXFSIZ exact power on value */
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg |= GUSBCFG_FORCEHOSTMODE;
      writel(gusbcfg, hsotg->regs + GUSBCFG);
      usleep_range(100000, 150000);

+     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      hptxfsiz = readl(hsotg->regs + HPTXFSIZ);
+     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "hptxfsiz=%08x\n", hptxfsiz);
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg &= ~GUSBCFG_FORCEHOSTMODE;

Nice! I wonder if this is a bug in the original driver, and they
actually meant to read this register instead of HPTXFSIZ? Well, it
doesn't really matter I guess.

Acked-by: Paul Zimmerman <redacted>

You may want to resend this to Greg after -rc1 is out and he reopens
his usb-next tree.

Thanks!

...do you think this is a fix that needs to go in for 3.17?  Is it
affecting anyone there?

I don't think so. It only has an effect if the controller starts up in
peripheral mode and then later switches to host mode, right? I think
yours is the first dwc2 platform in the kernel with that capability.

-- 
Paul

Doug:

On 08/08/2014 03:48 AM, Doug Anderson wrote:

The documentation for GNPTXFSIZ says that "For host mode, this field
is always valid."  Since we're already switching to host mode for
HPTXFSIZ, let's also read GNPTXFSIZ in host mode.

On an rk3288 SoC, without this change we see this at bootup:
   dwc2 ff580000.usb: gnptxfsiz=00100400
   dwc2 ff580000.usb: 128 invalid for host_nperio_tx_fifo_size. Check HW configuration.

After this change we see:
   dwc2 ff580000.usb: gnptxfsiz=04000400

Yeap, that is the problem cause the log you shown in rk3288-evb and 
further more
cause fifo setting fail.

I was plan to commit this patch just the same as you did.
It's great that you also find out the problem and send this patch.

There may be a potential problem still need to fix, the grxfsiz may have 
being changed,
the bootrom and uboot will change this value if they use the dwc2 
controller.
The way we get the register value here can not make sure this is the 
power-on
value which we actually need.

Let me do more test for that, and maybe we need another patch.

Anyway, this patch works and reasonable.

Reviewed-by: Kever Yang <redacted>

Greg,

On Thu, Aug 7, 2014 at 1:18 PM, Paul Zimmerman
[off-list ref] wrote:

quoted

From: Doug Anderson [mailto:dianders at chromium.org]
Sent: Thursday, August 07, 2014 12:48 PM

The documentation for GNPTXFSIZ says that "For host mode, this field
is always valid."  Since we're already switching to host mode for
HPTXFSIZ, let's also read GNPTXFSIZ in host mode.

On an rk3288 SoC, without this change we see this at bootup:
  dwc2 ff580000.usb: gnptxfsiz=00100400
  dwc2 ff580000.usb: 128 invalid for host_nperio_tx_fifo_size. Check HW configuration.

After this change we see:
  dwc2 ff580000.usb: gnptxfsiz=04000400

Signed-off-by: Doug Anderson <dianders@chromium.org>
---
 drivers/usb/dwc2/core.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/usb/dwc2/core.c b/drivers/usb/dwc2/core.c
index 27d2c9b..c184ed43 100644
--- a/drivers/usb/dwc2/core.c
+++ b/drivers/usb/dwc2/core.c

@@ -2674,23 +2674,23 @@ int dwc2_get_hwparams(struct dwc2_hsotg *hsotg)
      hwcfg2 = readl(hsotg->regs + GHWCFG2);
      hwcfg3 = readl(hsotg->regs + GHWCFG3);
      hwcfg4 = readl(hsotg->regs + GHWCFG4);
-     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      grxfsiz = readl(hsotg->regs + GRXFSIZ);

      dev_dbg(hsotg->dev, "hwcfg1=%08x\n", hwcfg1);
      dev_dbg(hsotg->dev, "hwcfg2=%08x\n", hwcfg2);
      dev_dbg(hsotg->dev, "hwcfg3=%08x\n", hwcfg3);
      dev_dbg(hsotg->dev, "hwcfg4=%08x\n", hwcfg4);
-     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "grxfsiz=%08x\n", grxfsiz);

-     /* Force host mode to get HPTXFSIZ exact power on value */
+     /* Force host mode to get HPTXFSIZ / GNPTXFSIZ exact power on value */
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg |= GUSBCFG_FORCEHOSTMODE;
      writel(gusbcfg, hsotg->regs + GUSBCFG);
      usleep_range(100000, 150000);

+     gnptxfsiz = readl(hsotg->regs + GNPTXFSIZ);
      hptxfsiz = readl(hsotg->regs + HPTXFSIZ);
+     dev_dbg(hsotg->dev, "gnptxfsiz=%08x\n", gnptxfsiz);
      dev_dbg(hsotg->dev, "hptxfsiz=%08x\n", hptxfsiz);
      gusbcfg = readl(hsotg->regs + GUSBCFG);
      gusbcfg &= ~GUSBCFG_FORCEHOSTMODE;

Nice! I wonder if this is a bug in the original driver, and they
actually meant to read this register instead of HPTXFSIZ? Well, it
doesn't really matter I guess.

Acked-by: Paul Zimmerman <redacted>

You may want to resend this to Greg after -rc1 is out and he reopens
his usb-next tree.

Is this something you want me to resend with Paul's Acked-by and
Kever's Reviewed-by?  Please let me know and I'm happy to repost.

-Doug