Dropping "depends on SMP" for HAVE_ARM_TWD -- take 2
[ take 1 was sent on 2015-03-26 ]
Hello everyone,
In http://thread.gmane.org/gmane.linux.ports.arm.kernel/389931/focus=392348
Quoted 2 lines · Stephen Boyd
I don't see any problem with the TWD dropping the dependency on SMP. The code should work the same on a UP configuration [...]
Quoted 2 lines · And Arnd recently
I think this has come up before and should be fixed. Could you send a patch that allows using TWD in uniprocessor configurations?
Basically, this means reverting Shawn Guo's 904464b91eca patch. and removing "depends on SMP" for HAVE_ARM_TWD.
However, Shawn's patch fixed an issue, therefore it seems likely that simply reverting is not the proper solution?
What should I do?
Regards.
[ Adding original reporter ]
Quoted 15 lines · Mason
[ take 1 was sent on 2015-03-26 ]
Hello everyone,
In http://thread.gmane.org/gmane.linux.ports.arm.kernel/389931/focus=392348
Quoted 2 lines · Stephen Boyd
I don't see any problem with the TWD dropping the dependency on SMP. The code should work the same on a UP configuration [...]
Quoted 2 lines · And Arnd recently
I think this has come up before and should be fixed. Could you send a patch that allows using TWD in uniprocessor configurations?
Quoted 7 lines
Basically, this means reverting Shawn Guo's
904464b91ecapatch. and removing "depends on SMP" forHAVE_ARM_TWD.However, Shawn's patch fixed an issue, therefore it seems likely that simply reverting is not the proper solution?
What should I do?
For reference, the warning used to be:
------------[ cut here ]------------
WARNING: at arch/arm/kernel/smp_twd.c:345
twd_local_timer_of_register+0x7c/0x90()
twd_local_timer_of_register failed (-6)
Modules linked in:
Backtrace:
[<80011f14>] (dump_backtrace+0x0/0x10c) from [<8044dd30>]
(dump_stack+0x18/0x1c)
r7:805e9f58 r6:805ba84c r5:80539331 r4:00000159
[<8044dd18>] (dump_stack+0x0/0x1c) from [<80020fbc>]
(warn_slowpath_common+0x54/0x6c)
[<80020f68>] (warn_slowpath_common+0x0/0x6c) from [<80021078>]
(warn_slowpath_fmt+0x38/0x40)
r9:412fc09a r8:8fffffff r7:ffffffff r6:00000001 r5:80633b8c
r4:80b32da8
[<80021040>] (warn_slowpath_fmt+0x0/0x40) from [<805ba84]
(twd_local_timer_of_register+0x7c/0x90)
r3:fffffffa r2:8053934b
[<805ba7d0>] (twd_local_timer_of_register+0x0/0x90) from [<805c0bec>]
(imx6q_timer_init+0x18/0x4c)
r5:80633800 r4:8053b701
[<805c0bd4>] (imx6q_timer_init+0x0/0x4c) from [<805ba4e8>]
(time_init+0x28/0x38)
r5:80633800 r4:805dc0f4
[<805ba4c0>] (time_init+0x0/0x38) from [<805b6854>]
(start_kernel+0x1a0/0x310)
[<805b66b4>] (start_kernel+0x0/0x310) from [<10008044>] (0x10008044)
r8:1000406a r7:805f3f8c r6:805dc0c4 r5:805f0518 r4:10c5387d
---[ end trace 1b75b31a2719ed1c ]---I cannot reproduce on v4.2 + my platform...
Booting with the following command line:
ip=dhcp root=/dev/nfs rdinit=/none console=ttyS0,115200 mem=640M debug nosmp
produces the attached boot log (no run-time warning).
Perhaps CONFIG_SMP_ON_UP matters? (I have it enabled.) Nope, I don't get any warning with CONFIG_SMP_ON_UP disabled.
Running out of ideas...
What does nosmp actually do?
static int __init nosmp(char *str)
{
setup_max_cpus = 0;
arch_disable_smp_support();
return 0;
}I don't think arch/arm overrides arch_disable_smp_support, so it's just a NOP. So we're just setting setup_max_cpus to 0...
There have been several updates to arch/arm/kernel/smp_twd.c since Shawn's patch, perhaps one of them unknowingly fixed the problem?
$ git ls arch/arm/kernel/smp_twd.c
4ed89f222806 ARM: convert printk(KERN_* to pr_*
06b96c8beb94 arm: Replace __this_cpu_ptr with raw_cpu_ptr
0b443ead714f cpufreq: remove unused notifier: CPUFREQ_{SUSPENDCHANGE|RESUMECHANGE}
2e874ea34214 ARM: twd: data endian fix
47dcd3563e45 Merge tag 'remove-local-timers' of git://git.kernel.org/pub/scm/linux/kernel/git/davidb/linux-msm into next/cleanup
8bd26e3a7e49 arm: delete __cpuinit/__CPUINIT usage from all ARM users
cbbe6f82b489 ARM: 7778/1: smp_twd: twd_update_frequency need be run on all online CPUs
a894fcc2d01a ARM: smp_twd: Divorce smp_twd from local timer API
da4a686a2cfb ARM: smp_twd: convert to use CLKSRC_OF init
904464b91eca ARM: 7655/1: smp_twd: make twd_local_timer_of_register() no-op for nosmp"smp_twd: Divorce smp_twd from local timer API" might be a likely candidate. (What do you think, Stephen?)
Regards.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: boot.log
Type: text/x-log
Size: 7666 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20151002/85eb1a49/attachment-0001.bin>
Quoted 60 lines · Mason
[ Adding original reporter ]
Quoted 15 lines · Mason
[ take 1 was sent on 2015-03-26 ]
Hello everyone,
In http://thread.gmane.org/gmane.linux.ports.arm.kernel/389931/focus=392348
Quoted 2 lines · Stephen Boyd
I don't see any problem with the TWD dropping the dependency on SMP. The code should work the same on a UP configuration [...]
Quoted 2 lines · And Arnd recently
I think this has come up before and should be fixed. Could you send a patch that allows using TWD in uniprocessor configurations?
Quoted 7 lines
Basically, this means reverting Shawn Guo's
904464b91ecapatch. and removing "depends on SMP" forHAVE_ARM_TWD.However, Shawn's patch fixed an issue, therefore it seems likely that simply reverting is not the proper solution?
What should I do?
For reference, the warning used to be:
------------[ cut here ]------------ WARNING: at arch/arm/kernel/smp_twd.c:345 twd_local_timer_of_register+0x7c/0x90() twd_local_timer_of_register failed (-6) Modules linked in: Backtrace: [<80011f14>] (dump_backtrace+0x0/0x10c) from [<8044dd30>] (dump_stack+0x18/0x1c) r7:805e9f58 r6:805ba84c r5:80539331 r4:00000159 [<8044dd18>] (dump_stack+0x0/0x1c) from [<80020fbc>] (warn_slowpath_common+0x54/0x6c) [<80020f68>] (warn_slowpath_common+0x0/0x6c) from [<80021078>] (warn_slowpath_fmt+0x38/0x40) r9:412fc09a r8:8fffffff r7:ffffffff r6:00000001 r5:80633b8c r4:80b32da8 [<80021040>] (warn_slowpath_fmt+0x0/0x40) from [<805ba84] (twd_local_timer_of_register+0x7c/0x90) r3:fffffffa r2:8053934b [<805ba7d0>] (twd_local_timer_of_register+0x0/0x90) from [<805c0bec>] (imx6q_timer_init+0x18/0x4c) r5:80633800 r4:8053b701 [<805c0bd4>] (imx6q_timer_init+0x0/0x4c) from [<805ba4e8>] (time_init+0x28/0x38) r5:80633800 r4:805dc0f4 [<805ba4c0>] (time_init+0x0/0x38) from [<805b6854>] (start_kernel+0x1a0/0x310) [<805b66b4>] (start_kernel+0x0/0x310) from [<10008044>] (0x10008044) r8:1000406a r7:805f3f8c r6:805dc0c4 r5:805f0518 r4:10c5387d ---[ end trace 1b75b31a2719ed1c ]---
Quoted 1 line
I cannot reproduce on v4.2 + my platform...
The warning has been removed in commit 5028090d1da1 (ARM: 8434/1:
Revert "7655/1: smp_twd: make twd_local_timer_of_register() no-op for nosmp", 2015-09-14) sitting in linux-next. Oddly, that commit doesn't remove the depends on SMP for the Kconfig. So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.
---8<----
arch/arm/Kconfig +0 -1
@@ -1389,7 +1389,6 @@ config HAVE_ARM_ARCH_TIMER config HAVE_ARM_TWD bool - depends on SMP select CLKSRC_OF if OF help This options enables support for the ARM timer and watchdog unit
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Hi,
Quoted 63 lines · Stephen Boyd
Quoted 60 lines · Mason
[ Adding original reporter ]
Quoted 15 lines · Mason
[ take 1 was sent on 2015-03-26 ]
Hello everyone,
In http://thread.gmane.org/gmane.linux.ports.arm.kernel/389931/focus=392348
Quoted 2 lines · Stephen Boyd
I don't see any problem with the TWD dropping the dependency on SMP. The code should work the same on a UP configuration [...]
Quoted 2 lines · And Arnd recently
I think this has come up before and should be fixed. Could you send a patch that allows using TWD in uniprocessor configurations?
Quoted 7 lines
Basically, this means reverting Shawn Guo's
904464b91ecapatch. and removing "depends on SMP" forHAVE_ARM_TWD.However, Shawn's patch fixed an issue, therefore it seems likely that simply reverting is not the proper solution?
What should I do?
For reference, the warning used to be:
------------[ cut here ]------------ WARNING: at arch/arm/kernel/smp_twd.c:345 twd_local_timer_of_register+0x7c/0x90() twd_local_timer_of_register failed (-6) Modules linked in: Backtrace: [<80011f14>] (dump_backtrace+0x0/0x10c) from [<8044dd30>] (dump_stack+0x18/0x1c) r7:805e9f58 r6:805ba84c r5:80539331 r4:00000159 [<8044dd18>] (dump_stack+0x0/0x1c) from [<80020fbc>] (warn_slowpath_common+0x54/0x6c) [<80020f68>] (warn_slowpath_common+0x0/0x6c) from [<80021078>] (warn_slowpath_fmt+0x38/0x40) r9:412fc09a r8:8fffffff r7:ffffffff r6:00000001 r5:80633b8c r4:80b32da8 [<80021040>] (warn_slowpath_fmt+0x0/0x40) from [<805ba84] (twd_local_timer_of_register+0x7c/0x90) r3:fffffffa r2:8053934b [<805ba7d0>] (twd_local_timer_of_register+0x0/0x90) from [<805c0bec>] (imx6q_timer_init+0x18/0x4c) r5:80633800 r4:8053b701 [<805c0bd4>] (imx6q_timer_init+0x0/0x4c) from [<805ba4e8>] (time_init+0x28/0x38) r5:80633800 r4:805dc0f4 [<805ba4c0>] (time_init+0x0/0x38) from [<805b6854>] (start_kernel+0x1a0/0x310) [<805b66b4>] (start_kernel+0x0/0x310) from [<10008044>] (0x10008044) r8:1000406a r7:805f3f8c r6:805dc0c4 r5:805f0518 r4:10c5387d ---[ end trace 1b75b31a2719ed1c ]---Quoted 1 line
I cannot reproduce on v4.2 + my platform...
Quoted 7 lines
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next. Oddly, that commit doesn't remove the depends on SMP for theKconfig. So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.
heh, sorry about that. I forgot omap2plus_defconfig enables everything. I could refresh that patch if Russell is still okay taking a newer version of it. Russell ?
Signature · 8 lines
balbi -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 819 bytes Desc: not available URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20151002/5962cbea/attachment.sig>
Quoted 3 lines · Felipe Balbi
heh, sorry about that. I forgot
omap2plus_defconfigenables everything. I could refresh that patch if Russell is still okay taking a newer version of it. Russell ?
Ok.
Signature · 2 lines
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up according to speedtest.net.
Quoted 6 lines · Russell King - ARM Linux
Quoted 3 lines · Felipe Balbi
heh, sorry about that. I forgot
omap2plus_defconfigenables everything. I could refresh that patch if Russell is still okay taking a newer version of it. Russell ?Ok.
thanks, patch coming shortly. If anybody is against refreshing original patch, let me know now :-)
Signature · 8 lines
balbi -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 819 bytes Desc: not available URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20151002/e70ea9f0/attachment.sig>
Quoted 61 lines · Stephen Boyd
Quoted 60 lines · Mason
[ Adding original reporter ]
Quoted 15 lines · Mason
[ take 1 was sent on 2015-03-26 ]
Hello everyone,
In http://thread.gmane.org/gmane.linux.ports.arm.kernel/389931/focus=392348
Quoted 2 lines · Stephen Boyd
I don't see any problem with the TWD dropping the dependency on SMP. The code should work the same on a UP configuration [...]
Quoted 2 lines · And Arnd recently
I think this has come up before and should be fixed. Could you send a patch that allows using TWD in uniprocessor configurations?
Quoted 7 lines
Basically, this means reverting Shawn Guo's
904464b91ecapatch. and removing "depends on SMP" forHAVE_ARM_TWD.However, Shawn's patch fixed an issue, therefore it seems likely that simply reverting is not the proper solution?
What should I do?
For reference, the warning used to be:
------------[ cut here ]------------ WARNING: at arch/arm/kernel/smp_twd.c:345 twd_local_timer_of_register+0x7c/0x90() twd_local_timer_of_register failed (-6) Modules linked in: Backtrace: [<80011f14>] (dump_backtrace+0x0/0x10c) from [<8044dd30>] (dump_stack+0x18/0x1c) r7:805e9f58 r6:805ba84c r5:80539331 r4:00000159 [<8044dd18>] (dump_stack+0x0/0x1c) from [<80020fbc>] (warn_slowpath_common+0x54/0x6c) [<80020f68>] (warn_slowpath_common+0x0/0x6c) from [<80021078>] (warn_slowpath_fmt+0x38/0x40) r9:412fc09a r8:8fffffff r7:ffffffff r6:00000001 r5:80633b8c r4:80b32da8 [<80021040>] (warn_slowpath_fmt+0x0/0x40) from [<805ba84] (twd_local_timer_of_register+0x7c/0x90) r3:fffffffa r2:8053934b [<805ba7d0>] (twd_local_timer_of_register+0x0/0x90) from [<805c0bec>] (imx6q_timer_init+0x18/0x4c) r5:80633800 r4:8053b701 [<805c0bd4>] (imx6q_timer_init+0x0/0x4c) from [<805ba4e8>] (time_init+0x28/0x38) r5:80633800 r4:805dc0f4 [<805ba4c0>] (time_init+0x0/0x38) from [<805b6854>] (start_kernel+0x1a0/0x310) [<805b66b4>] (start_kernel+0x0/0x310) from [<10008044>] (0x10008044) r8:1000406a r7:805f3f8c r6:805dc0c4 r5:805f0518 r4:10c5387d ---[ end trace 1b75b31a2719ed1c ]---
Quoted 5 lines
Quoted 1 line
I cannot reproduce on v4.2 + my platform...
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next.
Does that mean it will automatically make it into v4.3?
Quoted 1 line
Oddly, that commit doesn't remove the depends on SMP for the
Kconfig.
Perhaps because the commit is just a revert, and the depends statement has been here since the beginning (f32f4ce257452)
Quoted 3 lines
So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.
My port requires the TWD to function. So I'm using select HAVE_ARM_TWD (no "if SMP") to have it work even on MULTIPLATFORM UP configs.
arch/arm/Kconfig config HAVE_ARM_ARCH_TIMER · line 1394
Quoted 13 lines
---8<----
arch/arm/Kconfig +0
-1@@ -1389,7 +1389,6 @@ config HAVE_ARM_ARCH_TIMER config HAVE_ARM_TWD bool - depends on SMP select CLKSRC_OF if OF help This options enables support for the ARM timer and watchdog unit
Regards.
Quoted 11 lines · Mason
Quoted 8 lines · Stephen Boyd
Quoted 3 lines · Mason
I cannot reproduce on v4.2 + my platform...
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next.Does that mean it will automatically make it into v4.3?
No. It will most likely be in v4.4
Quoted 7 lines
Quoted 3 lines
So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.
So you probably get some Kconfig warning about unmet dependencies.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.
You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
Signature · 2 lines
M?ns Rullg?rd mans at mansr.com
Quoted 8 lines · On Sat, 3 Oct 2015 00:34:02 +0100 M?ns Rullg?rd <[email protected]> wrote:
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
I have no idea what your code does, but using the timer that is integrated into the core (just like on any other A9 implementation) feels like the right thing to do, unless the HW is too broken to use TWD.
Another timer is a nice thing to have (it probably comes in handy for suspend/resume, and it may have a much better resolution), but this seems like icing on the cake at this point. Let's see the cake first.
Thanks,
M.
Signature · 1 line
Jazz is not dead. It just smells funny.
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · On Sat, 3 Oct 2015 00:34:02 +0100 M?ns Rullg?rd wrote:
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
Quoted 8 lines
I have no idea what your code does, but using the timer that is integrated into the core (just like on any other A9 implementation) feels like the right thing to do, unless the HW is too broken to use TWD.
Another timer is a nice thing to have (it probably comes in handy for suspend/resume, and it may have a much better resolution), but this seems like icing on the cake at this point. Let's see the cake first.
Marc,
I have a quick question for you :-)
As I stated several months ago, my goal is to produce the simplest port possible, which is why I chose the TWD instead of platform timers. At one point, I was also considering the global timer. (I have a Cortex A9 MPCore.) But as far as I could tell, the code for the global timer does not handle CPU frequency changes (while smp_twd.c does), is that correct?
AFAIU, TWD timers and global timer "tick" at the frequency of PERIPHCLK (which is CPUCLK/2 in my implementation). Therefore, code needs to be specifically written to handle cpufreq events. Is that a correct understanding?
As for you comment about resolution, nominal PERIPHCLK is 600 MHz. Hard to beat that kind of resolution, I think.
Have a nice week end.
Regards.
Quoted 23 lines · On Sat, 3 Oct 2015 11:49:41 +0200 Mason <[email protected]> wrote:
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · On Sat, 3 Oct 2015 00:34:02 +0100 M?ns Rullg?rd wrote:
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
Quoted 8 lines
I have no idea what your code does, but using the timer that is integrated into the core (just like on any other A9 implementation) feels like the right thing to do, unless the HW is too broken to use TWD.
Another timer is a nice thing to have (it probably comes in handy for suspend/resume, and it may have a much better resolution), but this seems like icing on the cake at this point. Let's see the cake first.
Marc,
Quoted 8 lines
I have a quick question for you :-)
As I stated several months ago, my goal is to produce the simplest port possible, which is why I chose the TWD instead of platform timers. At one point, I was also considering the global timer. (I have a Cortex A9 MPCore.) But as far as I could tell, the code for the global timer does not handle CPU frequency changes (while
smp_twd.cdoes), is that correct?
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
Quoted 4 lines
AFAIU, TWD timers and global timer "tick" at the frequency of PERIPHCLK (which is CPUCLK/2 in my implementation). Therefore, code needs to be specifically written to handle cpufreq events. Is that a correct understanding?
I don't have the A9 TRM handy (nor the desire to read it on a sunny Saturday morning), so I can't really tell. But yes, this is likely to require some clock change handling.
M.
Signature · 1 line
Jazz is not dead. It just smells funny.
Quoted 35 lines · Marc Zyngier
Quoted 23 lines · On Sat, 3 Oct 2015 11:49:41 +0200 Mason <[email protected]> wrote:
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · On Sat, 3 Oct 2015 00:34:02 +0100 M?ns Rullg?rd wrote:
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
Quoted 8 lines
I have no idea what your code does, but using the timer that is integrated into the core (just like on any other A9 implementation) feels like the right thing to do, unless the HW is too broken to use TWD.
Another timer is a nice thing to have (it probably comes in handy for suspend/resume, and it may have a much better resolution), but this seems like icing on the cake at this point. Let's see the cake first.
Marc,
Quoted 8 lines
I have a quick question for you :-)
As I stated several months ago, my goal is to produce the simplest port possible, which is why I chose the TWD instead of platform timers. At one point, I was also considering the global timer. (I have a Cortex A9 MPCore.) But as far as I could tell, the code for the global timer does not handle CPU frequency changes (while
smp_twd.cdoes), is that correct?
Quoted 4 lines
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
I think the TWD only provides a clock_event device. Clocksource and schedclock would have to be provided by something else.
Quoted 9 lines
Quoted 5 lines
AFAIU, TWD timers and global timer "tick" at the frequency of PERIPHCLK (which is CPUCLK/2 in my implementation). Therefore, code needs to be specifically written to handle cpufreq events. Is that a correct understanding?
I don't have the A9 TRM handy (nor the desire to read it on a sunny Saturday morning), so I can't really tell. But yes, this is likely to require some clock change handling.
I looked at that once. IIRC, the problems are schedclock and clocksource. Other than a clockevent device which can be adjusted for frequency changes, there is (at least was) no such mechanism for clocksources and schedclock. Those are required to run at a stable frequency at all times
(https://www.kernel.org/doc/Documentation/timers/timekeeping.txt).
On Zynq I had the same problem and created a construct of notifiers for our clocksource to deal with this problem
(https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/clocksource/cadence_ttc_timer.c)
I never really liked that "solution" much, though.
That approach would probably not work with the GT because, last time I checked, it runs at the maximum frequency possible, i.e. dividers at the minimum (as a timer should), which means, if cpufreq throttles your CPU speed down, you'd have to decrease the clock dividers which isn't possible anymore (I could do that on the TTC because we have to run it a at very low speeds with high dividers to avoid constantly overflowing the 16-bit counter).
S?ren
Quoted 34 lines · S?ren Brinkmann
Quoted 33 lines · Marc Zyngier
Quoted 29 lines · On Sat, 3 Oct 2015 11:49:41 +0200 Mason <[email protected]> wrote:
Quoted 11 lines · Marc Zyngier
Quoted 8 lines · On Sat, 3 Oct 2015 00:34:02 +0100 M?ns Rullg?rd wrote:
Quoted 3 lines · Mason
My port requires the TWD to function. So I'm using select
HAVE_ARM_TWD(no "if SMP") to have it work even on MULTIPLATFORM UP configs.You could just let it use another timer in the non-smp case if you didn't have that weird aversion towards using my code.
Quoted 8 lines
I have no idea what your code does, but using the timer that is integrated into the core (just like on any other A9 implementation) feels like the right thing to do, unless the HW is too broken to use TWD.
Another timer is a nice thing to have (it probably comes in handy for suspend/resume, and it may have a much better resolution), but this seems like icing on the cake at this point. Let's see the cake first.
I have a quick question for you :-)
As I stated several months ago, my goal is to produce the simplest port possible, which is why I chose the TWD instead of platform timers. At one point, I was also considering the global timer. (I have a Cortex A9 MPCore.) But as far as I could tell, the code for the global timer does not handle CPU frequency changes (while
smp_twd.cdoes), is that correct?
Quoted 7 lines
Quoted 4 lines
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
I think the TWD only provides a
clock_eventdevice. Clocksource and schedclock would have to be provided by something else.
That is correct AFAIU. But my platform provides a simple 32-bit 27 MHz tick counter, which makes it trivial to implement clock source, sched clock and delay timer.
http://article.gmane.org/gmane.linux.kernel/2049558
I have identified two problems with using the TWD in Linux:
1) the TWD is only available on SMP kernels
=> that issue seems close to being resolved
2) I was told that some platforms "turn off" the TWD block in low-power modes (like during WFI), and therefore must be marked as unreliable in "C3" (FEAT_C3STOP).
I have a patch to selectively not mark the TWD as unreliable. (I will make a formal submission today.)
Quoted 20 lines
Quoted 8 lines
Quoted 4 lines
AFAIU, TWD timers and global timer "tick" at the frequency of PERIPHCLK (which is CPUCLK/2 in my implementation). Therefore, code needs to be specifically written to handle cpufreq events. Is that a correct understanding?
I don't have the A9 TRM handy (nor the desire to read it on a sunny Saturday morning), so I can't really tell. But yes, this is likely to require some clock change handling.
I looked at that once. IIRC, the problems are schedclock and clocksource. Other than a clockevent device which can be adjusted for frequency changes, there is (at least was) no such mechanism for clocksources and schedclock. Those are required to run at a stable frequency at all times
(https://www.kernel.org/doc/Documentation/timers/timekeeping.txt).On Zynq I had the same problem and created a construct of notifiers for our clocksource to deal with this problem
(https://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/tree/drivers/clocksource/cadence_ttc_timer.c)
I never really liked that "solution" much, though.
Quoted 7 lines
That approach would probably not work with the GT because, last time I checked, it runs at the maximum frequency possible, i.e. dividers at the minimum (as a timer should), which means, if cpufreq throttles your CPU speed down, you'd have to decrease the clock dividers which isn't possible anymore (I could do that on the TTC because we have to run it a at very low speeds with high dividers to avoid constantly overflowing the 16-bit counter).
Thanks for sharing your analysis.
Regards.
Quoted 6 lines · Mason
2) I was told that some platforms "turn off" the TWD block in low-power modes (like during WFI), and therefore must be marked as unreliable in "C3" (
FEAT_C3STOP).I have a patch to selectively not mark the TWD as unreliable. (I will make a formal submission today.)
This sounds like something that should be done with a device tree bool. Don't forget to patch Documentation/devicetree/bindings/arm/twd.txt
Yours,
Linus Walleij
Hi,
Quoted 9 lines · S?ren Brinkmann
Quoted 4 lines · Marc Zyngier
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
I think the TWD only provides a
clock_eventdevice. Clocksource and schedclock would have to be provided by something else.
If no clocksource, sched clock is provided, default jiffies based ones would be sufficient for single core, no ?, though not a preferred one.
Regards
afzal
Quoted 4 lines
I looked at that once. IIRC, the problems are schedclock and clocksource. Other than a clockevent device which can be adjusted for frequency changes, there is (at least was) no such mechanism for clocksources and schedclock. Those are required to run at a stable frequency at all times
Quoted 10 lines · Afzal Mohammed
Hi,
Quoted 7 lines · S?ren Brinkmann
Quoted 4 lines · Marc Zyngier
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
Quoted 5 lines
Quoted 2 lines
I think the TWD only provides a
clock_eventdevice. Clocksource and schedclock would have to be provided by something else.If no clocksource, sched clock is provided, default jiffies based ones would be sufficient for single core, no ?, though not a preferred one.
Yes that is correct. My point was more, that twd+cpufreq is a rather easy case since it doesn't provide a clocksource/schedclock. A driver that provides those needs quite some more work if it depends on the CPU frequency and needs to be usable with cpufreq.
S?ren
Quoted 14 lines · S?ren Brinkmann
Quoted 13 lines · Afzal Mohammed
Hi,
Quoted 9 lines · S?ren Brinkmann
Quoted 4 lines · Marc Zyngier
Indeed, I cannot see any code that does that in the GT driver. But if you have an A9 MP, you probably want to stick to TWD, which gives you a per-cpu timer instead of a global timer that will require IPIs to other CPUs.
I think the TWD only provides a
clock_eventdevice. Clocksource and schedclock would have to be provided by something else.
Quoted 7 lines
Quoted 2 lines
If no clocksource, sched clock is provided, default jiffies based ones would be sufficient for single core, no ?, though not a preferred one.
Yes that is correct. My point was more, that twd+cpufreq is a rather easy case since it doesn't provide a clocksource/schedclock. A driver that provides those needs quite some more work if it depends on the CPU frequency and needs to be usable with cpufreq.
_And_ you need to take the decision that you don't care about time keeping accuracy - because each and every time you change the clocksource's frequency, you will never be able to know exactly when the frequency change took effect in comparison with the counter value.
You could do a very tight "read counter before, write clock rate, read counter after" and use that to calculate the number of timer ticks at the old rate, and re-set the timing parameters with the new initial counter value after the change, but even that's not perfect as you don't actually know when the write hits the hardware compared to the reads.
Signature · 2 lines
FTTC broadband for 0.8mile line: currently at 9.6Mbps down 400kbps up according to speedtest.net.
Quoted 14 lines · Stephen Boyd
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next. Oddly, that commit doesn't remove the depends on SMP for theKconfig. So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.---8<----
arch/arm/Kconfig +0
-0
arch/arm/Kconfig config HAVE_ARM_ARCH_TIMER · line 1394
Quoted 7 lines
arch/arm/Kconfig +0
-1@@ -1389,7 +1389,6 @@ config HAVE_ARM_ARCH_TIMER config HAVE_ARM_TWD bool - depends on SMP select CLKSRC_OF if OF help This options enables support for the ARM timer and watchdog unit
Hello Stephen,
Should I make a formal submission for the above patch to be accepted?
(Did you see my "twd: Don't set CLOCK_EVT_FEAT_C3STOP unconditionally" patch? What should I do now for it to be accepted?)
Regards.
Quoted 26 lines · Mason
Quoted 14 lines · Stephen Boyd
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next. Oddly, that commit doesn't remove the depends on SMP for theKconfig. So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.---8<----
arch/arm/Kconfig +0
-0arch/arm/Kconfig config HAVE_ARM_ARCH_TIMER · line 1394
Quoted 7 lines
arch/arm/Kconfig +0
-1@@ -1389,7 +1389,6 @@ config HAVE_ARM_ARCH_TIMER config HAVE_ARM_TWD bool - depends on SMP select CLKSRC_OF if OF help This options enables support for the ARM timer and watchdog unitHello Stephen,
Quoted 1 line
Should I make a formal submission for the above patch to be accepted?
No. I believe Felipe is going to update commit 5028090d1da1 to remove the depends on.
Quoted 4 lines
(Did you see my "twd: Don't set
CLOCK_EVT_FEAT_C3STOPunconditionally" patch? What should I do now for it to be accepted?)
Assuming you sent it to the correct maintainer from the MAINTAINERS file, I would expect them to apply it directly.
Signature · 2 lines
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum, a Linux Foundation Collaborative Project
Quoted 27 lines · Stephen Boyd
Quoted 26 lines · Mason
Quoted 14 lines · Stephen Boyd
The warning has been removed in commit
5028090d1da1(ARM: 8434/1:
Revert "7655/1:smp_twd: maketwd_local_timer_of_register()no-op for nosmp", 2015-09-14) sitting in linux-next. Oddly, that commit doesn't remove the depends on SMP for theKconfig. So it seems that we can apply this patch and everyone is happy? We could also drop the "if SMP" part of most of the platform selects if anyone actually cares.---8<----
arch/arm/Kconfig +0
-0arch/arm/Kconfig config HAVE_ARM_ARCH_TIMER · line 1394
Quoted 7 lines
arch/arm/Kconfig +0
-1@@ -1389,7 +1389,6 @@ config HAVE_ARM_ARCH_TIMER config HAVE_ARM_TWD bool - depends on SMP select CLKSRC_OF if OF help This options enables support for the ARM timer and watchdog unitHello Stephen,
Quoted 4 lines
Quoted 1 line
Should I make a formal submission for the above patch to be accepted?
No. I believe Felipe is going to update commit
5028090d1da1to remove the depends on.
Already did it. Should be in next shortly.
Signature · 8 lines
balbi -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 818 bytes Desc: not available URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20151006/5823d60b/attachment.sig>
Quoted 7 lines · Stephen Boyd
Quoted 2 lines · Mason
Did you see my "twd: Don't set
CLOCK_EVT_FEAT_C3STOPunconditionally" patch? What should I do now for it to be accepted?Assuming you sent it to the correct maintainer from the
MAINTAINERSfile, I would expect them to apply it directly.
<confused>
The patch touches arch/arm/kernel/smp_twd.c
I do not see any entry for arch/arm/kernel in MAINTAINERS. This means I should fall back to the generic arch/arm entry, right?
ARM PORT
M: Russell King <[email protected]>
L: linux-arm-kernel at lists.infradead.org (moderated for non-subscribers)
W: http://www.arm.linux.org.uk/
S: Maintained
F: arch/arm/
AFAIU, there is a separate process for submitting patches for arch/arm:
http://www.arm.linux.org.uk/developer/patches/
Russell, Mark Rutland wrote: "Otherwise this looks ok." Does that mean I should now submit via the web form?
Regards.
Hi,
Quoted 23 lines · Mason
Quoted 7 lines · Stephen Boyd
Quoted 2 lines · Mason
Did you see my "twd: Don't set
CLOCK_EVT_FEAT_C3STOPunconditionally" patch? What should I do now for it to be accepted?Assuming you sent it to the correct maintainer from the
MAINTAINERSfile, I would expect them to apply it directly.<confused>
The patch touches
arch/arm/kernel/smp_twd.c
I do not see any entry forarch/arm/kernelinMAINTAINERS. This means I should fall back to the genericarch/armentry, right?ARM PORT
M: Russell King <[email protected]>
L: linux-arm-kernel at lists.infradead.org (moderated for non-subscribers)
W: http://www.arm.linux.org.uk/
S: Maintained
F:arch/arm/
Quoted 5 lines
AFAIU, there is a separate process for submitting patches for
arch/arm:
http://www.arm.linux.org.uk/developer/patches/Russell, Mark Rutland wrote: "Otherwise this looks ok." Does that mean I should now submit via the web form?
Mason, please have a read at Russell's patch info [1]. You don't even need to use the web form.
[1] http://www.arm.linux.org.uk/developer/patches/info.php
Signature · 8 lines
balbi -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 818 bytes Desc: not available URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20151006/0580d2ad/attachment.sig>