From: Doug Anderson <dianders@chromium.org> Date: 2014-09-11 15:52:20
Some 32-bit (ARMv7) systems are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset between the
virtual and physical counters. Each core gets a different random
offset.
On systems like the above, it doesn't make sense to use the virtual
counter. There's nobody managing the offset and each time a core goes
down and comes back up it will get reinitialized to some other random
value.
Let's add a property to the device tree to say that we shouldn't use
the virtual timer. Firmware could potentially remove this property
before passing the device tree to the kernel if it really wants the
kernel to use a virtual timer.
Note that it's been said that ARM64 (ARMv8) systems the firmware and
kernel really can't be architected as described above. That means
using the physical timer like this really only makes sense for ARMv7
systems.
In order for this patch to do anything useful, we also need Sonny's
patch at <https://patchwork.kernel.org/patch/4790921/>
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Sonny Rao <redacted>
---
Documentation/devicetree/bindings/arm/arch_timer.txt | 6 ++++++
drivers/clocksource/arm_arch_timer.c | 3 +++
2 files changed, 9 insertions(+)
@@ -22,6 +22,12 @@ to deliver its interrupts via SPIs. - always-on : a boolean property. If present, the timer is powered through an always-on power domain, therefore it never loses context.+** Optional properties:++- arm,use-physical-timer : Don't ever use the virtual timer, just use the+ physical one. Not supported for ARM64.++ Example: timer {
From: Christopher Covington <hidden> Date: 2014-09-11 15:58:09
Hi Doug,
On 09/11/2014 11:52 AM, Doug Anderson wrote:
quoted hunk
Some 32-bit (ARMv7) systems are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset between the
virtual and physical counters. Each core gets a different random
offset.
On systems like the above, it doesn't make sense to use the virtual
counter. There's nobody managing the offset and each time a core goes
down and comes back up it will get reinitialized to some other random
value.
Let's add a property to the device tree to say that we shouldn't use
the virtual timer. Firmware could potentially remove this property
before passing the device tree to the kernel if it really wants the
kernel to use a virtual timer.
Note that it's been said that ARM64 (ARMv8) systems the firmware and
kernel really can't be architected as described above. That means
using the physical timer like this really only makes sense for ARMv7
systems.
In order for this patch to do anything useful, we also need Sonny's
patch at <https://patchwork.kernel.org/patch/4790921/>
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Sonny Rao <redacted>
---
Documentation/devicetree/bindings/arm/arch_timer.txt | 6 ++++++
drivers/clocksource/arm_arch_timer.c | 3 +++
2 files changed, 9 insertions(+)
@@ -22,6 +22,12 @@ to deliver its interrupts via SPIs. - always-on : a boolean property. If present, the timer is powered through an always-on power domain, therefore it never loses context.+** Optional properties:++- arm,use-physical-timer : Don't ever use the virtual timer, just use the+ physical one. Not supported for ARM64.++ Example: timer {
How's the VDSO supposed to deal with this? It currently does:
cycle_now = arch_counter_get_cntvct()
Christopher
--
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by the Linux Foundation.
From: Doug Anderson <dianders@chromium.org> Date: 2014-09-11 16:04:16
Christopher,
On Thu, Sep 11, 2014 at 8:58 AM, Christopher Covington
[off-list ref] wrote:
Hi Doug,
On 09/11/2014 11:52 AM, Doug Anderson wrote:
quoted
Some 32-bit (ARMv7) systems are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset between the
virtual and physical counters. Each core gets a different random
offset.
On systems like the above, it doesn't make sense to use the virtual
counter. There's nobody managing the offset and each time a core goes
down and comes back up it will get reinitialized to some other random
value.
Let's add a property to the device tree to say that we shouldn't use
the virtual timer. Firmware could potentially remove this property
before passing the device tree to the kernel if it really wants the
kernel to use a virtual timer.
Note that it's been said that ARM64 (ARMv8) systems the firmware and
kernel really can't be architected as described above. That means
using the physical timer like this really only makes sense for ARMv7
systems.
In order for this patch to do anything useful, we also need Sonny's
patch at <https://patchwork.kernel.org/patch/4790921/>
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Sonny Rao <redacted>
---
Documentation/devicetree/bindings/arm/arch_timer.txt | 6 ++++++
drivers/clocksource/arm_arch_timer.c | 3 +++
2 files changed, 9 insertions(+)
@@ -22,6 +22,12 @@ to deliver its interrupts via SPIs. - always-on : a boolean property. If present, the timer is powered through an always-on power domain, therefore it never loses context.+** Optional properties:++- arm,use-physical-timer : Don't ever use the virtual timer, just use the+ physical one. Not supported for ARM64.++ Example: timer {
How's the VDSO supposed to deal with this? It currently does:
cycle_now = arch_counter_get_cntvct()
I don't see that line of code anywhere when I do a "git grep" on linux
or linuxnext. Can you give me a clearer pointer?
Also, I just want to make sure you saw the part of the patch
description that says:
In order for this patch to do anything useful, we also need Sonny's
patch at <https://patchwork.kernel.org/patch/4790921/>
-Doug
How's the VDSO supposed to deal with this? It currently does:
cycle_now = arch_counter_get_cntvct()
Note the VDSO for 32-bit ARM is still out of tree so there's no mainline
regression in that respect.
I think as long as there's some way for the vdso initialization code to
discover that it can't use the virtual counter, it should be tractable.
If that involves adding an API to the arch timer code, that can be
added along with the vdso code.
From: Christopher Covington <hidden> Date: 2014-09-11 16:20:27
On 09/11/2014 12:04 PM, Doug Anderson wrote:
Christopher,
On Thu, Sep 11, 2014 at 8:58 AM, Christopher Covington
[off-list ref] wrote:
quoted
Hi Doug,
On 09/11/2014 11:52 AM, Doug Anderson wrote:
quoted
Some 32-bit (ARMv7) systems are architected like this:
* The firmware doesn't know and doesn't care about hypervisor mode and
we don't want to add the complexity of hypervisor there.
* The firmware isn't involved in SMP bringup or resume.
* The ARCH timer come up with an uninitialized offset between the
virtual and physical counters. Each core gets a different random
offset.
On systems like the above, it doesn't make sense to use the virtual
counter. There's nobody managing the offset and each time a core goes
down and comes back up it will get reinitialized to some other random
value.
Let's add a property to the device tree to say that we shouldn't use
the virtual timer. Firmware could potentially remove this property
before passing the device tree to the kernel if it really wants the
kernel to use a virtual timer.
Note that it's been said that ARM64 (ARMv8) systems the firmware and
kernel really can't be architected as described above. That means
using the physical timer like this really only makes sense for ARMv7
systems.
In order for this patch to do anything useful, we also need Sonny's
patch at <https://patchwork.kernel.org/patch/4790921/>
Signed-off-by: Doug Anderson <dianders@chromium.org>
Signed-off-by: Sonny Rao <redacted>
---
Documentation/devicetree/bindings/arm/arch_timer.txt | 6 ++++++
drivers/clocksource/arm_arch_timer.c | 3 +++
2 files changed, 9 insertions(+)
@@ -22,6 +22,12 @@ to deliver its interrupts via SPIs. - always-on : a boolean property. If present, the timer is powered through an always-on power domain, therefore it never loses context.+** Optional properties:++- arm,use-physical-timer : Don't ever use the virtual timer, just use the+ physical one. Not supported for ARM64.++ Example: timer {
How's the VDSO supposed to deal with this? It currently does:
cycle_now = arch_counter_get_cntvct()
I don't see that line of code anywhere when I do a "git grep" on linux
or linuxnext. Can you give me a clearer pointer?
As Nathan mentioned, the code on the 32 bit side isn't merged yet.
http://www.spinics.net/lists/arm-kernel/msg356336.html
Christopher
--
Employee of Qualcomm Innovation Center, Inc.
Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
hosted by the Linux Foundation.