[Nouveau] [Bug 92077] nouveau graphics freeze when using KDE Plasma 5; PGR engine fault

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=92077

--- Comment #24 from Andrey Mazo  ---
A similar fault and a deadlock that happened on newer kernel (4.7.2).
Looks like to be triggered by qupzilla (www-client/qupzilla-2.0.1, compiled
with KDE/Plasma 5 support) launch this time.
Could be a QtWebEngine problem though.

[838735.611662] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 000c data 003c
[838735.611680] nouveau :03:00.0: fifo: PBDMA0: 0200 [] ch 15
[001f68d000 qupzilla[12159]] subc 0 mthd 0020 data 
[838735.611694] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0030 data 003c
[838735.611707] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0034 data 20040360
[838735.611720] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0038 data 
[838735.611733] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 003c data 
[838735.611746] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0040 data 
[838735.611758] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0044 data 
[838735.611771] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0048 data 20010674
[838735.611784] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 004c data 003c
[838735.611798] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0054 data 
[838735.611811] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0058 data 00252000
[838735.611824] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 005c data 200240c7
[838735.611836] nouveau :03:00.0: fifo: PBDMA0: 0020 [ILLEGAL_MTHD] ch
15 [001f68d000 qupzilla[12159]] subc 0 mthd 0060 data 0048
[838735.611855] nouveau :03:00.0: fifo: read fault at 00 engine 07
[PFIFO] client 07 [BAR_READ] reason 02 [PAGE_NOT_PRESENT] on channel 15
[001f68d000 qupzilla[12159]]
[838735.611858] nouveau :03:00.0: fifo: fifo engine fault on channel 15,
recovering...

[838735.612309] =
[838735.612311] [ INFO: possible recursive locking detected ]
[838735.612314] 4.7.2-gentoo-20160906 #2 Not tainted
[838735.612315] -
[838735.612317] kworker/0:1/12138 is trying to acquire lock:
[838735.612319]  ((>recover.work)){+.+...}, at: []
flush_work+0x0/0x290
[838735.612329]
but task is already holding lock:
[838735.612331]  ((>recover.work)){+.+...}, at: []
process_one_work+0x14b/0x450
[838735.612336]
other info that might help us debug this:
[838735.612337]  Possible unsafe locking scenario:

[838735.612339]CPU0
[838735.612340]
[838735.612341]   lock((>recover.work));
[838735.612343]   lock((>recover.work));
[838735.612345]
 *** DEADLOCK ***

[838735.612347]  May be due to missing lock nesting notation

[838735.612349] 2 locks held by kworker/0:1/12138:
[838735.612351]  #0:  ("events"){.+.+.+}, at: []
process_one_work+0x14b/0x450
[838735.612356]  #1:  ((>recover.work)){+.+...}, at: []
process_one_work+0x14b/0x450
[838735.612360]
stack backtrace:
[838735.612363] CPU: 0 PID: 12138 Comm: kworker/0:1 Not tainted
4.7.2-gentoo-20160906 #2
[838735.612365] Hardware name: Dell Inc. Precision Tower 7910/0215PR, BIOS A06
01/19/2015
[838735.612371] Workqueue: events gf100_fifo_recover_work
[838735.612373]   8800750a7af0 b7567885
88082f22
[838735.612378]  b86fd780 8800750a7bb8 b73064c8
b86fd780
[838735.612381]  88082f220050 8800750a7b00 000268312189
0002
[838735.612385] Call Trace:
[838735.612391]  [] dump_stack+0x67/0x92
[838735.612396]  [] __lock_acquire+0x1508/0x1620
[838735.612399]  [] lock_acquire+0x60/0x80
[838735.612401]  [] ? mod_delayed_work_on+0x80/0x80
[838735.612403]  [] flush_work+0x47/0x290
[838735.612405]  [] ? mod_delayed_work_on+0x80/0x80
[838735.612408]  [] ? mark_held_locks+0x71/0x90
[838735.612414]  [] ? ktime_get+0x6a/0x130
[838735.612417]  [] gf100_fifo_fini+0x10/0x20
[838735.612420]  [] nvkm_fifo_fini+0x1a/0x30
[838735.612424]  [] nvkm_engine_fini+0x20/0x30
[838735.612429]  [] nvkm_subdev_fini+0x5a/0x160
[838735.612432]  [] gf100_fifo_recover_work+0xea/0x1f0
[838735.612434]  [] process_one_work+0x1aa/0x450
[838735.612436]  [] ? process_one_work+0x14b/0x450
[838735.612438]  [] 

[Nouveau] [Bug 97462] Graphics deadlock "ILLEGAL_MTHD" in nouveau with mesa version 11.2.2 when visiting Google Maps with firefox 49.0b5

[bugzilla-daemon]

https://bugs.freedesktop.org/show_bug.cgi?id=97462

--- Comment #10 from Paviluf  ---
By the way Chromium work well.

-- 
You are receiving this mail because:
You are the assignee for the bug.___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 2/3] drm/nouveau/led: guard against a division by 0

[Karol Herbst]

2016-09-16 11:14 GMT+02:00 Martin Peres :
> On 16/09/16 10:42, Karol Herbst wrote:
>>
>> Reviewed-by: Karol Herbst 
>>
>> 2016-09-16 9:34 GMT+02:00 Martin Peres :
>>>
>>> Signed-off-by: Martin Peres 
>>> ---
>>>   drm/nouveau/nouveau_led.c | 5 -
>>>   1 file changed, 4 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
>>> index 5e28b5f..1f731da 100644
>>> --- a/drm/nouveau/nouveau_led.c
>>> +++ b/drm/nouveau/nouveau_led.c
>>> @@ -44,7 +44,10 @@ nouveau_led_get_brightness(struct led_classdev *led)
>>>  div =  nvif_rd32(device, 0x61c880) & 0x00ff;
>>>  duty = nvif_rd32(device, 0x61c884) & 0x00ff;
>>>
>>> -   return duty * LED_FULL / div;
>>> +   if (div > 0)
>>> +   return duty * LED_FULL / div;
>>> +   else
>>> +   return 0;
>>
>> minor nitpick: you can drop the else, I don't mind though. Maybe it
>> would be clearer to do it the other way around though.
>>
>> Or to do:
>>
>> if (unlikely(div <= 0)) return 0;
>> return duty * LED_FULL / div;
>
>
> The unlikely is a good idea, not a fan of single lines if conditions though
> :s

the single line was me being lazy.

if (unlikely(div <= 0))
return 0;
return duty * LED_FULL / div;

:p
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH] volt: use kernel's 64-bit signed division function

[Karol Herbst]

yeah locks fine

Reviewed-by: Karol Herbst 

2016-09-16 11:09 GMT+02:00 Alexandre Courbot :
> Doing direct 64 bit divisions in kernel code leads to references to
> undefined symbols on 32 bit architectures. Replace such divisions with
> calls to div64_s64 to make the module usable on 32 bit archs.
>
> Signed-off-by: Alexandre Courbot 
> ---
>  drm/nouveau/nvkm/subdev/volt/base.c | 6 +++---
>  lib/include/nvif/os.h   | 1 +
>  2 files changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/drm/nouveau/nvkm/subdev/volt/base.c 
> b/drm/nouveau/nvkm/subdev/volt/base.c
> index 90d234231eed..e8569b04b55d 100644
> --- a/drm/nouveau/nvkm/subdev/volt/base.c
> +++ b/drm/nouveau/nvkm/subdev/volt/base.c
> @@ -116,9 +116,9 @@ nvkm_volt_map(struct nvkm_volt *volt, u8 id, u8 temp)
> return volt->speedo;
>
> if (ver == 0x10 || (ver == 0x20 && info.mode == 0)) {
> -   result  =  (s64)info.arg[0] / 10;
> -   result += ((s64)info.arg[1] * volt->speedo) / 10;
> -   result += ((s64)info.arg[2] * volt->speedo * 
> volt->speedo) / 10;
> +   result  = div64_s64((s64)info.arg[0], 10);
> +   result += div64_s64((s64)info.arg[1] * volt->speedo, 
> 10);
> +   result += div64_s64((s64)info.arg[2] * volt->speedo * 
> volt->speedo, 10);
> } else if (ver == 0x20) {
> switch (info.mode) {
> /* 0x0 handled above! */
> diff --git a/lib/include/nvif/os.h b/lib/include/nvif/os.h
> index b45a1862436e..2c1985d90474 100644
> --- a/lib/include/nvif/os.h
> +++ b/lib/include/nvif/os.h
> @@ -88,6 +88,7 @@ typedef dma_addr_t resource_size_t;
>  )
>  #define do_div(a,b) (a) = (a) / (b)
>  #define div_u64(a,b) (a) / (b)
> +#define div64_s64(a,b) (a) / (b)
>  #define likely(a) (a)
>  #define unlikely(a) (a)
>  #define BIT(a) (1UL << (a))
> --
> 2.9.3
>
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

[Nouveau] [PATCH] volt: use kernel's 64-bit signed division function

[Alexandre Courbot]

Doing direct 64 bit divisions in kernel code leads to references to
undefined symbols on 32 bit architectures. Replace such divisions with
calls to div64_s64 to make the module usable on 32 bit archs.

Signed-off-by: Alexandre Courbot 
---
 drm/nouveau/nvkm/subdev/volt/base.c | 6 +++---
 lib/include/nvif/os.h   | 1 +
 2 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/drm/nouveau/nvkm/subdev/volt/base.c 
b/drm/nouveau/nvkm/subdev/volt/base.c
index 90d234231eed..e8569b04b55d 100644
--- a/drm/nouveau/nvkm/subdev/volt/base.c
+++ b/drm/nouveau/nvkm/subdev/volt/base.c
@@ -116,9 +116,9 @@ nvkm_volt_map(struct nvkm_volt *volt, u8 id, u8 temp)
return volt->speedo;
 
if (ver == 0x10 || (ver == 0x20 && info.mode == 0)) {
-   result  =  (s64)info.arg[0] / 10;
-   result += ((s64)info.arg[1] * volt->speedo) / 10;
-   result += ((s64)info.arg[2] * volt->speedo * 
volt->speedo) / 10;
+   result  = div64_s64((s64)info.arg[0], 10);
+   result += div64_s64((s64)info.arg[1] * volt->speedo, 
10);
+   result += div64_s64((s64)info.arg[2] * volt->speedo * 
volt->speedo, 10);
} else if (ver == 0x20) {
switch (info.mode) {
/* 0x0 handled above! */
diff --git a/lib/include/nvif/os.h b/lib/include/nvif/os.h
index b45a1862436e..2c1985d90474 100644
--- a/lib/include/nvif/os.h
+++ b/lib/include/nvif/os.h
@@ -88,6 +88,7 @@ typedef dma_addr_t resource_size_t;
 )
 #define do_div(a,b) (a) = (a) / (b)
 #define div_u64(a,b) (a) / (b)
+#define div64_s64(a,b) (a) / (b)
 #define likely(a) (a)
 #define unlikely(a) (a)
 #define BIT(a) (1UL << (a))
-- 
2.9.3

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 2/3] drm/nouveau/led: guard against a division by 0

[Martin Peres]

On 16/09/16 10:42, Karol Herbst wrote:

Reviewed-by: Karol Herbst 

2016-09-16 9:34 GMT+02:00 Martin Peres :

Signed-off-by: Martin Peres 
---
  drm/nouveau/nouveau_led.c | 5 -
  1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
index 5e28b5f..1f731da 100644
--- a/drm/nouveau/nouveau_led.c
+++ b/drm/nouveau/nouveau_led.c
@@ -44,7 +44,10 @@ nouveau_led_get_brightness(struct led_classdev *led)
 div =  nvif_rd32(device, 0x61c880) & 0x00ff;
 duty = nvif_rd32(device, 0x61c884) & 0x00ff;

-   return duty * LED_FULL / div;
+   if (div > 0)
+   return duty * LED_FULL / div;
+   else
+   return 0;

minor nitpick: you can drop the else, I don't mind though. Maybe it
would be clearer to do it the other way around though.

Or to do:

if (unlikely(div <= 0)) return 0;
return duty * LED_FULL / div;


The unlikely is a good idea, not a fan of single lines if conditions 
though :s

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 3/3] drm/nouveau/led: abort early if the device does not have GPIOs

[Karol Herbst]

Reviewed-by: Karol Herbst 

2016-09-16 9:34 GMT+02:00 Martin Peres :
> Signed-off-by: Martin Peres 
> ---
>  drm/nouveau/nouveau_led.c | 3 +++
>  1 file changed, 3 insertions(+)
>
> diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
> index 1f731da..3e2f1b6 100644
> --- a/drm/nouveau/nouveau_led.c
> +++ b/drm/nouveau/nouveau_led.c
> @@ -82,6 +82,9 @@ nouveau_led_init(struct drm_device *dev)
> struct dcb_gpio_func logo_led;
> int ret;
>
> +   if (!gpio)
> +   return 0;
> +
> /* check that there is a GPIO controlling the logo LED */
> if (nvkm_gpio_find(gpio, 0, DCB_GPIO_LOGO_LED_PWM, 0xff, _led))
> return 0;
> --
> 2.8.0
>
> ___
> Nouveau mailing list
> Nouveau@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 3/3] drm/nouveau/led: abort early if the device does not have GPIOs

[Alexandre Courbot]

On Fri, Sep 16, 2016 at 4:34 PM, Martin Peres  wrote:
> Signed-off-by: Martin Peres 

Tested-by: Alexandre Courbot 

Fixes the crash I was seeing on Tegra, thanks!
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 2/3] drm/nouveau/led: guard against a division by 0

[Alexandre Courbot]

On Fri, Sep 16, 2016 at 4:34 PM, Martin Peres  wrote:
> Signed-off-by: Martin Peres 

Tested-by: Alexandre Courbot 

Fixes the crash I was seeing on GM206, thanks!
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

Re: [Nouveau] [PATCH 2/3] drm/nouveau/led: guard against a division by 0

[Karol Herbst]

Reviewed-by: Karol Herbst 

2016-09-16 9:34 GMT+02:00 Martin Peres :
> Signed-off-by: Martin Peres 
> ---
>  drm/nouveau/nouveau_led.c | 5 -
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
> index 5e28b5f..1f731da 100644
> --- a/drm/nouveau/nouveau_led.c
> +++ b/drm/nouveau/nouveau_led.c
> @@ -44,7 +44,10 @@ nouveau_led_get_brightness(struct led_classdev *led)
> div =  nvif_rd32(device, 0x61c880) & 0x00ff;
> duty = nvif_rd32(device, 0x61c884) & 0x00ff;
>
> -   return duty * LED_FULL / div;
> +   if (div > 0)
> +   return duty * LED_FULL / div;
> +   else
> +   return 0;

minor nitpick: you can drop the else, I don't mind though. Maybe it
would be clearer to do it the other way around though.

Or to do:

if (unlikely(div <= 0)) return 0;
return duty * LED_FULL / div;

>  }
>
>  static void
> --
> 2.8.0
>
> ___
> Nouveau mailing list
> Nouveau@lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/nouveau
___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

[Nouveau] [PATCH 3/3] drm/nouveau/led: abort early if the device does not have GPIOs

[Martin Peres]

Signed-off-by: Martin Peres 
---
 drm/nouveau/nouveau_led.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
index 1f731da..3e2f1b6 100644
--- a/drm/nouveau/nouveau_led.c
+++ b/drm/nouveau/nouveau_led.c
@@ -82,6 +82,9 @@ nouveau_led_init(struct drm_device *dev)
struct dcb_gpio_func logo_led;
int ret;
 
+   if (!gpio)
+   return 0;
+
/* check that there is a GPIO controlling the logo LED */
if (nvkm_gpio_find(gpio, 0, DCB_GPIO_LOGO_LED_PWM, 0xff, _led))
return 0;
-- 
2.8.0

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

[Nouveau] [PATCH 1/3] drm/nouveau/led: don't access led subdev if it wasn't initialized

[Martin Peres]

From: Karol Herbst 

Fixes a kernel crash on suspend/resume.
---
 drm/nouveau/nouveau_led.c | 7 ---
 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
index 9eed5a6..5e28b5f 100644
--- a/drm/nouveau/nouveau_led.c
+++ b/drm/nouveau/nouveau_led.c
@@ -107,7 +107,8 @@ nouveau_led_suspend(struct drm_device *dev)
 {
struct nouveau_drm *drm = nouveau_drm(dev);
 
-   led_classdev_suspend(>led->led);
+   if (drm->led)
+   led_classdev_suspend(>led->led);
 }
 
 void
@@ -115,8 +116,8 @@ nouveau_led_resume(struct drm_device *dev)
 {
struct nouveau_drm *drm = nouveau_drm(dev);
 
-   led_classdev_resume(>led->led);
-
+   if (drm->led)
+   led_classdev_resume(>led->led);
 }
 
 void
-- 
2.8.0

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

[Nouveau] [PATCH 2/3] drm/nouveau/led: guard against a division by 0

[Martin Peres]

Signed-off-by: Martin Peres 
---
 drm/nouveau/nouveau_led.c | 5 -
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/drm/nouveau/nouveau_led.c b/drm/nouveau/nouveau_led.c
index 5e28b5f..1f731da 100644
--- a/drm/nouveau/nouveau_led.c
+++ b/drm/nouveau/nouveau_led.c
@@ -44,7 +44,10 @@ nouveau_led_get_brightness(struct led_classdev *led)
div =  nvif_rd32(device, 0x61c880) & 0x00ff;
duty = nvif_rd32(device, 0x61c884) & 0x00ff;
 
-   return duty * LED_FULL / div;
+   if (div > 0)
+   return duty * LED_FULL / div;
+   else
+   return 0;
 }
 
 static void
-- 
2.8.0

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau

[Nouveau] [PATCH] drm/nouveau/disp: remove unused function in sorg94.c

[Baoyou Xie]

We get 1 warning when building kernel with W=1:
drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c:49:1: warning: no previous 
prototype for 'g94_sor_output_new' [-Wmissing-prototypes]

In fact, this function is called by no one and not exported,
so this patch removes it.

Signed-off-by: Baoyou Xie 
---
 drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c | 8 
 1 file changed, 8 deletions(-)

diff --git a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c 
b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c
index 1bb9d66..4510cb6 100644
--- a/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c
+++ b/drivers/gpu/drm/nouveau/nvkm/engine/disp/sorg94.c
@@ -45,14 +45,6 @@ static const struct nvkm_output_func
 g94_sor_output_func = {
 };
 
-int
-g94_sor_output_new(struct nvkm_disp *disp, int index,
-  struct dcb_output *dcbE, struct nvkm_output **poutp)
-{
-   return nvkm_output_new_(_sor_output_func, disp,
-   index, dcbE, poutp);
-}
-
 
/***
  * DisplayPort
  
**/
-- 
2.7.4

___
Nouveau mailing list
Nouveau@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/nouveau