Re: The failure summary report of GEN2 for linux stable v4.8

[Laurent Pinchart]

Hello Truong,

On Friday 21 Oct 2016 14:20:24 Laurent Pinchart wrote:
> On Friday 21 Oct 2016 19:53:47 Xuan Truong Nguyen wrote:
> > Hello Laurent.
> > 
> > > The document mentions that you have used shmobile_defconfig. However,
> > > that configuration in v4.8 doesn't enable CONFIG_CMA, which causes at
> > > least the DU driver to fail probing the device. I thus assume you have
> > > modified the configuration to enable CMA. The fact that
> > > shmobile_defconfig doesn't include the uvcvideo driver used by test 14
> > > seems to confirm local modifications to the configuration.
> > > Issue 8 ("Lager: bad image quality when shown on HDMI display") can't be
> > > reproduced on my system when displaying a test pattern with the modetest
> > > application. Could you please share more information on how to reproduce
> > > the problem ?
> > 
> > We retested this issue No 8 again with the CONFIG_CMA enabled but the
> > result is the same. you could see it by the error.jpg image in
> > attachments. the display device used is Panasonic TH-L19C3 (refer the
> > device.jpg). we also send you my configs file, too (lager_du.config).
> 
> Thank you for the information, I'll try to reproduce the problem here with
> your kernel configuration.

I've been able to reproduce the problem here, and at this point I'm not sure 
what goes wrong. I've tested older kernels up to the point where HDMI support 
for Lager was introduced, and the bug was present in all of them. I'll try to 
fix it, but I have other higher priority tasks at the moment.

> > just one more thing, with CONFIG_CMA enabled, the driver fails on
> > probing if we boot the board without the HDMI cable inserted in advance.
> > we have to insert the cable before booting the board. and not all the
> > display work with current HDMI driver (as our LG display will not work)
> > we are using the mainline upstream version 4.8
> 
> Interesting, I'll check that too.

I haven't had time to check this yet.

> > if you need any information to debug, please lets us know.
> 
> Could you please tell me what you're using to test the display ?

-- 
Regards,

Laurent Pinchart

Re: The failure summary report of GEN2 for linux stable v4.8

[Xuan Truong Nguyen]

Hi Shimoda-san

Thanks for your patch.

We will apply your patch and retest this issue on v4.9-rc2 then report 
you again.



Thanks and best regards
JINSO/Truong


On 2016年10月28日 17:14, Yoshihiro Shimoda wrote:

Hi Geert-san, Xuan-san,


From: linux-renesas-soc-ow...@vger.kernel.org 
[mailto:linux-renesas-soc-ow...@vger.kernel.org] On Behalf Of Geert
Uytterhoeven
Sent: Monday, October 24, 2016 6:21 PM

On Mon, Oct 24, 2016 at 11:13 AM, Xuan Truong Nguyen
 wrote:

This is with shmobile_defconfig?

yes. we also attach the configs file we used
(lager-scif-pio-v4.9-rc2.config)


Does it work better if you enable CONFIG_SERIAL_SH_SCI_DMA?

yes, it's better a little bit. the kernel does not hangs up, but the warning
message is output.
please refer lager-scif-dma.log.

we tested on v4.9-rc2. the issue is the same.

if you need any information, please let us know.
WARNING: CPU: 0 PID: 2249 at drivers/dma/sh/rcar-dmac.c:1257 
rcar_dmac_tx_status+0x128/0x4
No descriptor for cookie!
[] (rcar_dmac_tx_status) from [] (rx_timer_fn+0x48/0x148)
[] (rx_timer_fn) from [] (call_timer_fn+0x2c/0xa0)

That looks like a race condition between timeout handling and actual completion
of the DMA?

I found an issue in sh-sci.c and made a patch to resolve it.
But, I'm not sure this is correct way.
If this is correct way, we also have to fix dev_dbg() in some functions.

Best regards,
Yoshihiro Shimoda

Since I send this email using Outlook, the patch format may be not good.
---
From: Yoshihiro Shimoda 
Date: Fri, 28 Oct 2016 16:52:36 +0900
Subject: [PATCH] serial: sh-sci: remove dev_warn() to avoid double spin lock
  held

If we use serial console and CONFIG_SERIAL_SH_SCI_DMA=y, since
sci_dma_rx_push() is called with port->lock held and dev_warn() will
call serial_console_write() finally, this is possible to call
spin_lock{_irqsave}() twice.
To avoid this, this patch remove dev_warn() in sci_dma_rx_push().

Reported-by: Xuan Truong Nguyen 
Signed-off-by: Yoshihiro Shimoda 
---
  drivers/tty/serial/sh-sci.c | 5 +
  1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
index 4b26252..380b5d7 100644
--- a/drivers/tty/serial/sh-sci.c
+++ b/drivers/tty/serial/sh-sci.c
@@ -1142,11 +1142,8 @@ static int sci_dma_rx_push(struct sci_port *s, void 
*buf, size_t count)
int copied;
  
  	copied = tty_insert_flip_string(tport, buf, count);

-   if (copied < count) {
-   dev_warn(port->dev, "Rx overrun: dropping %zu bytes\n",
-count - copied);
+   if (copied < count)
port->icount.buf_overrun++;
-   }
  
  	port->icount.rx += copied;
  

RE: The failure summary report of GEN2 for linux stable v4.8

[Yoshihiro Shimoda]

Hi Geert-san, Xuan-san,

> From: linux-renesas-soc-ow...@vger.kernel.org 
> [mailto:linux-renesas-soc-ow...@vger.kernel.org] On Behalf Of Geert
> Uytterhoeven
> Sent: Monday, October 24, 2016 6:21 PM
> 
> On Mon, Oct 24, 2016 at 11:13 AM, Xuan Truong Nguyen
>  wrote:
> >> This is with shmobile_defconfig?
> >
> > yes. we also attach the configs file we used
> > (lager-scif-pio-v4.9-rc2.config)
> >
> >> Does it work better if you enable CONFIG_SERIAL_SH_SCI_DMA?
> >
> > yes, it's better a little bit. the kernel does not hangs up, but the warning
> > message is output.
> > please refer lager-scif-dma.log.
> >
> > we tested on v4.9-rc2. the issue is the same.
> >
> > if you need any information, please let us know.
> 
> > WARNING: CPU: 0 PID: 2249 at drivers/dma/sh/rcar-dmac.c:1257 
> > rcar_dmac_tx_status+0x128/0x4
> > No descriptor for cookie!
> 
> > [] (rcar_dmac_tx_status) from [] 
> > (rx_timer_fn+0x48/0x148)
> 
> > [] (rx_timer_fn) from [] (call_timer_fn+0x2c/0xa0)
> 
> That looks like a race condition between timeout handling and actual 
> completion
> of the DMA?

I found an issue in sh-sci.c and made a patch to resolve it.
But, I'm not sure this is correct way.
If this is correct way, we also have to fix dev_dbg() in some functions.

Best regards,
Yoshihiro Shimoda

Since I send this email using Outlook, the patch format may be not good.
---
From: Yoshihiro Shimoda 
Date: Fri, 28 Oct 2016 16:52:36 +0900
Subject: [PATCH] serial: sh-sci: remove dev_warn() to avoid double spin lock
 held

If we use serial console and CONFIG_SERIAL_SH_SCI_DMA=y, since
sci_dma_rx_push() is called with port->lock held and dev_warn() will
call serial_console_write() finally, this is possible to call
spin_lock{_irqsave}() twice.
To avoid this, this patch remove dev_warn() in sci_dma_rx_push().

Reported-by: Xuan Truong Nguyen 
Signed-off-by: Yoshihiro Shimoda 
---
 drivers/tty/serial/sh-sci.c | 5 +
 1 file changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/tty/serial/sh-sci.c b/drivers/tty/serial/sh-sci.c
index 4b26252..380b5d7 100644
--- a/drivers/tty/serial/sh-sci.c
+++ b/drivers/tty/serial/sh-sci.c
@@ -1142,11 +1142,8 @@ static int sci_dma_rx_push(struct sci_port *s, void 
*buf, size_t count)
int copied;
 
copied = tty_insert_flip_string(tport, buf, count);
-   if (copied < count) {
-   dev_warn(port->dev, "Rx overrun: dropping %zu bytes\n",
-count - copied);
+   if (copied < count)
port->icount.buf_overrun++;
-   }
 
port->icount.rx += copied;
 
-- 
1.9.1

Re: The failure summary report of GEN2 for linux stable v4.8

[Geert Uytterhoeven]

On Mon, Oct 24, 2016 at 11:13 AM, Xuan Truong Nguyen
 wrote:
>> This is with shmobile_defconfig?
>
> yes. we also attach the configs file we used
> (lager-scif-pio-v4.9-rc2.config)
>
>> Does it work better if you enable CONFIG_SERIAL_SH_SCI_DMA?
>
> yes, it's better a little bit. the kernel does not hangs up, but the warning
> message is output.
> please refer lager-scif-dma.log.
>
> we tested on v4.9-rc2. the issue is the same.
>
> if you need any information, please let us know.

> WARNING: CPU: 0 PID: 2249 at drivers/dma/sh/rcar-dmac.c:1257 
> rcar_dmac_tx_status+0x128/0x4
> No descriptor for cookie!

> [] (rcar_dmac_tx_status) from [] (rx_timer_fn+0x48/0x148)

> [] (rx_timer_fn) from [] (call_timer_fn+0x2c/0xa0)

That looks like a race condition between timeout handling and actual completion
of the DMA?

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- ge...@linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds

Re: The failure summary report of GEN2 for linux stable v4.8

[Laurent Pinchart]

Hello,

On Tuesday 18 Oct 2016 17:13:32 Laurent Pinchart wrote:

[snip]

> Issue 14 ("USB-CAM: Warning message appears if remove the usb-camera cable
> under using") is likely not a Renesas-specific problem, but as I'm the
> maintainer of the uvcvideo webcam driver you used for testing I'll
> investigate that.

I confirm that this issue can be reproduced on an x86 machine running v4.4.6. 
It isn't Renesas-specific.

-- 
Regards,

Laurent Pinchart

Re: The failure summary report of GEN2 for linux stable v4.8

[Wolfram Sang]

Hi all,

> We have tested the linux stable v4.8 for Gen2 Koelsch and Lager


So we would like to report the summary of the failure.


Thank you!

Is it possible to get access to the full logfiles? I am especially 
interested in the backtraces after the Kernel oops or warning.


Kind regards,

   Wolfram