KVM on ARM currently masks the timer interrupt on the timer side when
delivering an interrupt to the guest. This itself is a workaround for an
issue where the interrupt is reraised and trapped by the host as soon as
the guest is entered, resulting in the guest being starved.
Work around this by calling ArmGenericTimerEnable () after servicing each
interrupt. The virtual version of ArmGenericTimerCounterLib will then
make sure to unmask the interrupt again.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-By: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@16144 6f19259b-4bc3-4df7-8a09-765794883524
The PCD gArmTokenSpaceGuid.PcdArmArchTimerFreqInHz is used in the
SEC phase (if applicable) to write the platform's counter frequency
to the CNTFRQ system register, as this needs to be done by the highest
exception level implemented.
Under virtualization, we should be able to rely on the host to have
initialized this register to a sane value, as we run at EL1 and only
use the virtual timer, so the PcdArmArchTimerFreqInHz PCD has little
meaning here.
So in either case, by the time we enter the DXE phase, we can use the
CNTFRQ system register to read the frequency instead of looking at the
PCD.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-by: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@16090 6f19259b-4bc3-4df7-8a09-765794883524
Move TimerDxe and ArmArchTimerLib to ArmGenericTimerCounterLib, and update all
platforms to select the physical counter instance they have been using
implicitly all along.
Contributed-under: TianoCore Contribution Agreement 1.0
Acked-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Signed-off-By: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@16078 6f19259b-4bc3-4df7-8a09-765794883524
The ArmArchTimerLib.h include file is not directly related to the TimerLib
instance ArmArchTimerLib, so the name is confusing. Rename to ArmArchTimer.h
instead.
Contributed-under: TianoCore Contribution Agreement 1.0
Acked-by: Laszlo Ersek <lersek@redhat.com>
Signed-off-by: Ard Biesheuvel <ard.biesheuvel@linaro.org>
Reviewed-By: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@16073 6f19259b-4bc3-4df7-8a09-765794883524
Prior to this change, the TimerPeriod was re-initialize at the
end of the interrupt handling with the value of the period.
It means the real timer period was: Timer Interrupt Processing
time + Timer Period
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15923 6f19259b-4bc3-4df7-8a09-765794883524
Prior to this change the period was restored to the default period.
This change restores the latest 'set period'.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15922 6f19259b-4bc3-4df7-8a09-765794883524
Prior to this change the frequency was rounded to 1Mhz.
This change rounds the timer frequency to 1KHz.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@15921 6f19259b-4bc3-4df7-8a09-765794883524
- Registering a interrupt handler implicitly enables said interrupt. This
is in the UEFI Spec. No need to enable the interrupts a second time.
- Make sure the Timer is completely disabled before configuring it. Only
enable after configuration is complete.
Contributed-under: TianoCore Contribution Agreement 1.0
Signed-off-by: Olivier Martin <olivier.martin@arm.com>
git-svn-id: https://svn.code.sf.net/p/edk2/code/trunk/edk2@14433 6f19259b-4bc3-4df7-8a09-765794883524