Skip to contentSkip to conversation

AM335x board with disabled RTC crashes

DISCUSSION Stefan Roese 3 participants 7 messages Nov 20, 2013 – Nov 21, 2013 1 message not in the archive Device tree bindings

lore
Share & tools
lore
#0 lore raw

Hi!

I'm currently working on a custom AM335x based board, where the SoC RTC is disabled. As described in this page in the "RTC feature disabled" column:

http://processors.wiki.ti.com/index.php/AM335x_Schematic_Checklist#RTC

On this board the registers of the RTC are not accessible. But the OMAP hwmod code tries to read the sysc register unconditionally from the RTC. And this leads to this crash (using v3.12):

...

[    0.179491] CPU: Testing write buffer coherency: ok
[    0.187702] Setting up static identity map for 0xc04dfa90 - 0xc04dfb00
[    0.204665] devtmpfs: initialized
[    0.219513] VFP support v0.3: implementor 41 architecture 3 part 30 variant c rev 3
[    0.237462] *** _update_sysc_cache (263): name=emif
[    0.246247] omap_hwmod: debugss: _wait_target_disable failed
[    0.252290] *** _update_sysc_cache (263): name=uart1
[    0.257668] *** _update_sysc_cache (263): name=rtc
[    0.262885] Unhandled fault: external abort on non-linefetch (0x1028) at 0xf9e3e078
[    0.270893] Internal error: : 1028 [#1] ARM
[    0.275302] Modules linked in:
[    0.278572] CPU: 0 PID: 1 Comm: swapper Not tainted 3.12.0-dirty #21
[    0.285222] task: c7047340 ti: c7048000 task.ti: c7048000
[    0.290909] PC is at _update_sysc_cache+0x44/0xb0
[    0.295861] LR is at _update_sysc_cache+0x1c/0xb0

...

Disabling the RTC in the DT doesn't help (of course). So what's the recommended way to disable this hwmod access to the RTC registers?

Thanks,
Stefan

#1 lore raw
Quoted 28 lines · Stefan Roese

Hi!

I'm currently working on a custom AM335x based board, where the SoC RTC is disabled. As described in this page in the "RTC feature disabled" column:

http://processors.wiki.ti.com/index.php/AM335x_Schematic_Checklist#RTC

On this board the registers of the RTC are not accessible. But the OMAP hwmod code tries to read the sysc register unconditionally from the RTC. And this leads to this crash (using v3.12):

...

[    0.179491] CPU: Testing write buffer coherency: ok
[    0.187702] Setting up static identity map for 0xc04dfa90 - 0xc04dfb00
[    0.204665] devtmpfs: initialized
[    0.219513] VFP support v0.3: implementor 41 architecture 3 part 30 variant c rev 3
[    0.237462] *** _update_sysc_cache (263): name=emif
[    0.246247] omap_hwmod: debugss: _wait_target_disable failed
[    0.252290] *** _update_sysc_cache (263): name=uart1
[    0.257668] *** _update_sysc_cache (263): name=rtc
[    0.262885] Unhandled fault: external abort on non-linefetch (0x1028) at 0xf9e3e078
[    0.270893] Internal error: : 1028 [#1] ARM
[    0.275302] Modules linked in:
[    0.278572] CPU: 0 PID: 1 Comm: swapper Not tainted 3.12.0-dirty #21
[    0.285222] task: c7047340 ti: c7048000 task.ti: c7048000
[    0.290909] PC is at _update_sysc_cache+0x44/0xb0
[    0.295861] LR is at _update_sysc_cache+0x1c/0xb0
Quoted 4 lines

...

Disabling the RTC in the DT doesn't help (of course). So what's the recommended way to disable this hwmod access to the RTC registers?

http://www.spinics.net/lists/linux-omap/msg98207.html
might help you - we had thought it might get queued for 3.12, but it was queued for 3.13 instead..

Signature · 2 lines
Regards,
Nishanth Menon

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

#3 lore raw
Quoted 41 lines · Nishanth Menon

Paul, Tony, Rajendra,

Quoted 38 lines · Nishanth Menon
Quoted 28 lines · Stefan Roese

Hi!

I'm currently working on a custom AM335x based board, where the SoC RTC is disabled. As described in this page in the "RTC feature disabled" column:

http://processors.wiki.ti.com/index.php/AM335x_Schematic_Checklist#RTC

On this board the registers of the RTC are not accessible. But the OMAP hwmod code tries to read the sysc register unconditionally from the RTC. And this leads to this crash (using v3.12):

...

[    0.179491] CPU: Testing write buffer coherency: ok
[    0.187702] Setting up static identity map for 0xc04dfa90 - 0xc04dfb00
[    0.204665] devtmpfs: initialized
[    0.219513] VFP support v0.3: implementor 41 architecture 3 part 30 variant c rev 3
[    0.237462] *** _update_sysc_cache (263): name=emif
[    0.246247] omap_hwmod: debugss: _wait_target_disable failed
[    0.252290] *** _update_sysc_cache (263): name=uart1
[    0.257668] *** _update_sysc_cache (263): name=rtc
[    0.262885] Unhandled fault: external abort on non-linefetch (0x1028) at 0xf9e3e078
[    0.270893] Internal error: : 1028 [#1] ARM
[    0.275302] Modules linked in:
[    0.278572] CPU: 0 PID: 1 Comm: swapper Not tainted 3.12.0-dirty #21
[    0.285222] task: c7047340 ti: c7048000 task.ti: c7048000
[    0.290909] PC is at _update_sysc_cache+0x44/0xb0
[    0.295861] LR is at _update_sysc_cache+0x1c/0xb0
Quoted 4 lines

...

Disabling the RTC in the DT doesn't help (of course). So what's the recommended way to disable this hwmod access to the RTC registers?

http://www.spinics.net/lists/linux-omap/msg98207.html
might help you - we had thought it might get queued for 3.12, but it was queued for 3.13 instead..

Unfortunately this patch doesn't fix the crash for me (yet). With this patch applied and the RTC DT node completely removed from am33xx.dtsi it works. But only disabling it (via status = "disabled" in the board dts file) doesn't work.

I'll dig a bit deeper today...

Thanks,
Stefan

#4 lore raw
Quoted 8 lines · Stefan Roese
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

Disabling the RTC in the DT doesn't help (of course). So what's the recommended way to disable this hwmod access to the RTC registers?

http://www.spinics.net/lists/linux-omap/msg98207.html
might help you - we had thought it might get queued for 3.12, but it was queued for 3.13 instead..

Quoted 6 lines

Unfortunately this patch doesn't fix the crash for me (yet). With this patch applied and the RTC DT node completely removed from am33xx.dtsi it works. But only disabling it (via status = "disabled" in the board dts file) doesn't work.

I'll dig a bit deeper today...

This (hacky) patch works, but I'm not sure if this is acceptable upstream:

am335x-board_foo.dts:

...

&rtc {
	reg = <0x0 0x0>;
};

Thanks,
Stefan

#5 lore raw
Quoted 18 lines · Stefan Roese
Quoted 8 lines · Stefan Roese
Quoted 7 lines
Quoted 7 lines
Quoted 2 lines

Disabling the RTC in the DT doesn't help (of course). So what's the recommended way to disable this hwmod access to the RTC registers?

http://www.spinics.net/lists/linux-omap/msg98207.html
might help you - we had thought it might get queued for 3.12, but it was queued for 3.13 instead..

Quoted 6 lines

Unfortunately this patch doesn't fix the crash for me (yet). With this patch applied and the RTC DT node completely removed from am33xx.dtsi it works. But only disabling it (via status = "disabled" in the board dts file) doesn't work.

I'll dig a bit deeper today...

This (hacky) patch works, but I'm not sure if this is acceptable upstream:

Quoted 8 lines

am335x-board_foo.dts:

...

&rtc {
	reg = <0x0 0x0>;
};

You should be able to achieve the same effect as following (example from BBB) - though I dont see this defined in Documentation/devicetree/bindings/arm/omap/omap.txt

I will allow the wisdom of others to comment better here :)

arch/arm/boot/dts/am335x-boneblack.dts +5 -0
index 6b71ad9..a734ef4 100644
@@ -66,6 +66,11 @@
 	status = "okay";
 };
 
+&am335xrtc {
+	status = "disabled";
+	ti,hwmods="disabled";
+};
+
 / {
 	hdmi {
 		compatible = "ti,tilcdc,slave";
arch/arm/boot/dts/am33xx.dtsi +1 -1
index f6d8ffe..71bf2ec 100644
@@ -373,7 +373,7 @@
 			ti,timer-pwm;
 		};
 
-		rtc@44e3e000 {
+		am335xrtc: rtc@44e3e000 {
 			compatible = "ti,da830-rtc";
 			reg = <0x44e3e000 0x1000>;
 			interrupts = <75
Signature · 2 lines
Regards,
Nishanth Menon
#6 lore raw

Thanks Nishanth!

Quoted 20 lines · Nishanth Menon
Quoted 10 lines

This (hacky) patch works, but I'm not sure if this is acceptable upstream:

am335x-board_foo.dts:

...

&rtc {
	reg = <0x0 0x0>;
};

You should be able to achieve the same effect as following (example from BBB) - though I dont see this defined in Documentation/devicetree/bindings/arm/omap/omap.txt

I will allow the wisdom of others to comment better here :)

arch/arm/boot/dts/am335x-boneblack.dts +0 -0
index 6b71ad9..a734ef4 100644

arch/arm/boot/dts/am335x-boneblack.dts line 72

Quoted 7 lines
arch/arm/boot/dts/am335x-boneblack.dts +4 -0
@@ -66,6 +66,11 @@
 	status = "okay";
 };
 
+&am335xrtc {
+	status = "disabled";
+	ti,hwmods="disabled";
+};

Yes, this works too. Thanks.

Which leaves only the quite ugly WARN() resulting from Suman's patch:

[    0.230270] omap_hwmod: rtc: Could not ioremap
[    0.234962] ------------[ cut here ]------------
[    0.239936] WARNING: CPU: 0 PID: 1 at arch/arm/mach-omap2/omap_hwmod.c:2434 _init+0x144/0x310()
[    0.249054] omap_hwmod: rtc: doesn't have mpu register target base
[    0.255526] Modules linked in:
[    0.258845] CPU: 0 PID: 1 Comm: swapper Not tainted 3.12.0-00004-gfcb6c2c-dirty #31
[    0.266938] [<c00197f4>] (unwind_backtrace+0x0/0xf0) from [<c0016ee8>] (show_stack+0x10/0x14)
[    0.275916] [<c0016ee8>] (show_stack+0x10/0x14) from [<c003be68>] (warn_slowpath_common+0x6c/0x8c)
[    0.285284] [<c003be68>] (warn_slowpath_common+0x6c/0x8c) from [<c003bf1c>] (warn_slowpath_fmt+0x30/0x40)
[    0.295323] [<c003bf1c>] (warn_slowpath_fmt+0x30/0x40) from [<c06fdac8>] (_init+0x144/0x310)
[    0.304200] [<c06fdac8>] (_init+0x144/0x310) from [<c0029e40>] (omap_hwmod_for_each+0x34/0x5c)
[    0.313250] [<c0029e40>] (omap_hwmod_for_each+0x34/0x5c) from [<c06fe1fc>] (__omap_hwmod_setup_all+0x24/0x40)
[    0.323643] [<c06fe1fc>] (__omap_hwmod_setup_all+0x24/0x40) from [<c00087e8>] (do_one_initcall+0x34/0x160)
[    0.333774] [<c00087e8>] (do_one_initcall+0x34/0x160) from [<c06f3af0>] (kernel_init_freeable+0xe8/0x1b4)
[    0.343824] [<c06f3af0>] (kernel_init_freeable+0xe8/0x1b4) from [<c04d0ff4>] (kernel_init+0x8/0xe4)
[    0.353338] [<c04d0ff4>] (kernel_init+0x8/0xe4) from [<c00139e8>] (ret_from_fork+0x14/0x2c)
[    0.362428] ---[ end trace 1b75b31a2719ed1c ]---

Do we really need this WARN() here? Or is the "Could not ioremap" line enough?

Thanks,
Stefan

#7 lore raw
Quoted 53 lines · Stefan Roese

Thanks Nishanth!

Quoted 20 lines · Nishanth Menon
Quoted 10 lines

This (hacky) patch works, but I'm not sure if this is acceptable upstream:

am335x-board_foo.dts:

...

&rtc {
	reg = <0x0 0x0>;
};

You should be able to achieve the same effect as following (example from BBB) - though I dont see this defined in Documentation/devicetree/bindings/arm/omap/omap.txt

I will allow the wisdom of others to comment better here :)

arch/arm/boot/dts/am335x-boneblack.dts +0 -0
index 6b71ad9..a734ef4 100644

arch/arm/boot/dts/am335x-boneblack.dts line 72

Quoted 7 lines
arch/arm/boot/dts/am335x-boneblack.dts +4 -0
@@ -66,6 +66,11 @@
 	status = "okay";
 };
 
+&am335xrtc {
+	status = "disabled";
+	ti,hwmods="disabled";
+};

Yes, this works too. Thanks.

Which leaves only the quite ugly WARN() resulting from Suman's patch:

[    0.230270] omap_hwmod: rtc: Could not ioremap
[    0.234962] ------------[ cut here ]------------
[    0.239936] WARNING: CPU: 0 PID: 1 at arch/arm/mach-omap2/omap_hwmod.c:2434 _init+0x144/0x310()
[    0.249054] omap_hwmod: rtc: doesn't have mpu register target base
[    0.255526] Modules linked in:
[    0.258845] CPU: 0 PID: 1 Comm: swapper Not tainted 3.12.0-00004-gfcb6c2c-dirty #31
[    0.266938] [<c00197f4>] (unwind_backtrace+0x0/0xf0) from [<c0016ee8>] (show_stack+0x10/0x14)
[    0.275916] [<c0016ee8>] (show_stack+0x10/0x14) from [<c003be68>] (warn_slowpath_common+0x6c/0x8c)
[    0.285284] [<c003be68>] (warn_slowpath_common+0x6c/0x8c) from [<c003bf1c>] (warn_slowpath_fmt+0x30/0x40)
[    0.295323] [<c003bf1c>] (warn_slowpath_fmt+0x30/0x40) from [<c06fdac8>] (_init+0x144/0x310)
[    0.304200] [<c06fdac8>] (_init+0x144/0x310) from [<c0029e40>] (omap_hwmod_for_each+0x34/0x5c)
[    0.313250] [<c0029e40>] (omap_hwmod_for_each+0x34/0x5c) from [<c06fe1fc>] (__omap_hwmod_setup_all+0x24/0x40)
[    0.323643] [<c06fe1fc>] (__omap_hwmod_setup_all+0x24/0x40) from [<c00087e8>] (do_one_initcall+0x34/0x160)
[    0.333774] [<c00087e8>] (do_one_initcall+0x34/0x160) from [<c06f3af0>] (kernel_init_freeable+0xe8/0x1b4)
[    0.343824] [<c06f3af0>] (kernel_init_freeable+0xe8/0x1b4) from [<c04d0ff4>] (kernel_init+0x8/0xe4)
[    0.353338] [<c04d0ff4>] (kernel_init+0x8/0xe4) from [<c00139e8>] (ret_from_fork+0x14/0x2c)
[    0.362428] ---[ end trace 1b75b31a2719ed1c ]---
Quoted 2 lines

Do we really need this WARN() here? Or is the "Could not ioremap" line enough?

I had seen this too :) - There is a bit of history around this:
- we are in the process of moving functionality from hwmod into dts since both are meant to describe the hardware and dt is the way to go.
- at this point - hwmod is supposed to accurately represent SoC
modules and dts is supposed to match up.
- as part of this conversion, one of the first things to move is to memory range description from hwmod to dts (reg description).
- the WARN() was introduced to catch such instances where hwmod contain nodes and dts entries dont contain the description.

What we did miss is a valid condition where the generic definitions for the presence of a SoC peripheral depends on a pin pull on the board for the same SoC.

What should happen: if maintainers think that 'ti,hwmods="disabled";' is an sufficient way of describing this current scenario, then we can intelligently parse the same and avoid the warning for these cases.

Signature · 2 lines
Regards,
Nishanth Menon