Re: [PATCH v5 00/27] arm64: provide pseudo NMI with GICv3

From: Julien Thierry
Date: Wed Aug 29 2018 - 08:58:22 EST




On 29/08/18 12:37, Daniel Thompson wrote:
On Tue, Aug 28, 2018 at 04:51:10PM +0100, Julien Thierry wrote:
Hi,

This series is a continuation of the work started by Daniel [1]. The goal
is to use GICv3 interrupt priorities to simulate an NMI.

The patches depend on the core API for NMIs patches [2].

As before... is there a git tree? With the NMI core API I think I might
be able to get some of my old pseudo-NMI demos working.


Here is the git:

git://linux-arm.org/linux-jt.git v4.19-nmi-pmu-public

there are also a few patches to make the PMU interrupt into an NMI.


To achieve this, set two priorities, one for standard interrupts and
another, higher priority, for NMIs. Whenever we want to disable interrupts,
we mask the standard priority instead so NMIs can still be raised. Some
corner cases though still require to actually mask all interrupts
effectively disabling the NMI.

Daniel Thompson ran some benchmarks [3] on the previous version showing a
small (<1%) performance drop when using interrupt priorities.

IMHO it is very important to disclose which micro-architecture (in this
case I think I was using C-A53 and GIC-500) the performance drop is
observed with.

We know from both micro- and macro-benchmarks that the performance delta
is deeply dependant on core implementation. In fact, my own work in this
area stalled largely because the main device that justified my spending
working-hours on pseudo-NMI was too badly impacted to see it enabled by
default.


Yes, you're right, I haven't had many boards to test on, the one I tested on had 8 C-A57. Not sure about the GIC (it implemented GICv3 architecture, that's all I know). I had similar results to yours.

I did not have the opportunity to re-run the benchmarks on real hardware for this spin however.

Thanks,

Julien


Daniel.


Currently, only PPIs and SPIs can be set as NMIs. IPIs being currently
hardcoded IRQ numbers, there isn't a generic interface to set SGIs as NMI
for now. LPIs being controlled by the ITS cannot be delivered as NMI.
When an NMI is active on a CPU, no other NMI can be triggered on the CPU.

Requirements to use this:
- Have GICv3
- SCR_EL3.FIQ is set to 1 when linux runs or have single security state
- Select Kernel Feature -> Use ICC system registers for IRQ masking


* Patches 1 to 3 aim at applying some alternatives early in the boot
process.

* Patches 4 to 7 ensure the logic of daifflags remains valid
after arch_local_irq flags use ICC_PMR_EL1.

* Patches 8 and 9 clean up GIC current priority definition to make it
easier to introduce a new priority

* Patches 10 to 16 prepare arch code for the use of priorities, saving and
restoring ICC_PMR_EL1 appropriately

* Patches 17 to 20 add the support to GICv3 driver to use priority masking
if required by the architecture

* Patches 21 to 23 make arm64 code use ICC_PMR_EL1 to enable/disable
interrupts, leaving PSR.I as often as possible

* Patches 24 to 27 add the support for NMIs to GICv3 driver


Changes since V4[4]:
* Rebased to v4.19-rc1
* Adapted GIC driver to the core NMI API
* Added option to disable priority masking on command line
* Added Daniel's Tested-by on patches related replacing PSR.I toggling with
PMR masking
* Fix scope matching for alternative features.
* Spotted some more places using PSR.I or daif and replaced with generic
interrupt functions

Changes since V3[5]:
* Big refactoring. As suggested by Marc Z., some of the bigger patches
needed to be split into smaller one.
* Try to reduce the amount of #ifdef for the new feature by introducing
an individual cpufeature for priority masking
* Do not track which alternatives have been applied (was a bit dodgy
anyway), and use an alternative for VHE cpu_enable callback
* Fix a build failure with arm by adding the correct RPR accessors
* Added Suggested-by tags for changes from comming or inspired by Daniel's
series. Do let me know if you feel I missed something and am not giving
you due credit.

Changes since V2[6]:
* Series rebase to v4.17-rc6
* Adapt pathces 1 and 2 to the rework of cpufeatures framework
* Use the group0 detection scheme in the GICv3 driver to identify
the priority view, and drop the use of a fake interrupt
* Add the case for a GIC configured in a single security state
* Use local_daif_restore instead of local_irq_enable the first time
we enable interrupts after a bp hardening in the handling of a kernel
entry. Otherwise PRS.I remains set...

Changes since V1[7]:
* Series rebased to v4.15-rc8.
* Check for arm64_early_features in this_cpu_has_cap (spotted by Suzuki).
* Fix issue where debug exception were not masked when enabling debug in
mdscr_el1.

Changes since RFC[8]:
* The series was rebased to v4.15-rc2 which implied some changes mainly
related to the work on exception entries and daif flags by James Morse.
- The first patch in the previous series was dropped because no longer
applicable.
- With the semantics James introduced of "inheriting" daif flags,
handling of PMR on exception entry is simplified as PMR is not altered
by taking an exception and already inherited from previous state.
- James pointed out that taking a PseudoNMI before reading the FAR_EL1
register should not be allowed as per the TRM (D10.2.29):
"FAR_EL1 is made UNKNOWN on an exception return from EL1."
So in this submission PSR.I bit is cleared only after FAR_EL1 is read.
* For KVM, only deal with PMR unmasking/restoring in common code, and VHE
specific code makes sure PSR.I bit is set when necessary.
* When detecting the GIC priority view (patch 5), wait for an actual
interrupt instead of trying only once.


[1] http://www.spinics.net/lists/arm-kernel/msg525077.html
[2] https://lkml.org/lkml/2018/8/28/661
[3] https://lkml.org/lkml/2018/7/20/803
[4] https://lkml.org/lkml/2018/7/24/321
[5] https://lkml.org/lkml/2018/5/21/276
[6] https://lkml.org/lkml/2018/1/17/335
[7] https://www.spinics.net/lists/arm-kernel/msg620763.html
[8] https://www.spinics.net/lists/arm-kernel/msg610736.html

Cheers,

Julien

-->

Daniel Thompson (1):
arm64: alternative: Apply alternatives early in boot process

Julien Thierry (26):
arm64: cpufeature: Set SYSREG_GIC_CPUIF as a boot system feature
arm64: cpufeature: Use alternatives for VHE cpu_enable
arm64: daifflags: Use irqflags functions for daifflags
arm64: Use daifflag_restore after bp_hardening
arm64: Delay daif masking for user return
arm64: xen: Use existing helper to check interrupt status
irqchip/gic: Unify GIC priority definitions
irqchip/gic: Lower priority of GIC interrupts
arm64: cpufeature: Add cpufeature for IRQ priority masking
arm64: Make PMR part of task context
arm64: Unmask PMR before going idle
arm/arm64: gic-v3: Add helper functions to manage IRQ priorities
arm64: kvm: Unmask PMR before entering guest
arm64: irqflags: Use ICC_PMR_EL1 for interrupt masking
arm64: daifflags: Include PMR in daifflags restore operations
irqchip/gic-v3: Factor group0 detection into functions
irqchip/gic-v3: Do not overwrite PMR value
irqchip/gic-v3: Remove acknowledge loop
irqchip/gic-v3: Switch to PMR masking after IRQ acknowledge
arm64: Switch to PMR masking when starting CPUs
arm64: Add build option for IRQ masking via priority
arm64: Handle serror in NMI context
irqchip/gic-v3: Detect current view of GIC priorities
irqchip/gic-v3: Add base support for pseudo-NMI
irqchip/gic: Add functions to access irq priorities
irqchip/gic-v3: Allow interrupts to be set as pseudo-NMI

Documentation/admin-guide/kernel-parameters.txt | 3 +
Documentation/arm64/booting.txt | 5 +
arch/arm/include/asm/arch_gicv3.h | 33 +++
arch/arm64/Kconfig | 15 ++
arch/arm64/include/asm/alternative.h | 3 +-
arch/arm64/include/asm/arch_gicv3.h | 33 +++
arch/arm64/include/asm/assembler.h | 17 +-
arch/arm64/include/asm/cpucaps.h | 3 +-
arch/arm64/include/asm/cpufeature.h | 2 +
arch/arm64/include/asm/daifflags.h | 29 ++-
arch/arm64/include/asm/efi.h | 3 +-
arch/arm64/include/asm/irqflags.h | 100 +++++++--
arch/arm64/include/asm/kvm_host.h | 12 +
arch/arm64/include/asm/processor.h | 1 +
arch/arm64/include/asm/ptrace.h | 13 +-
arch/arm64/include/asm/xen/events.h | 2 +-
arch/arm64/kernel/alternative.c | 28 ++-
arch/arm64/kernel/asm-offsets.c | 1 +
arch/arm64/kernel/cpufeature.c | 51 ++++-
arch/arm64/kernel/entry.S | 63 +++++-
arch/arm64/kernel/head.S | 35 +++
arch/arm64/kernel/process.c | 2 +
arch/arm64/kernel/smp.c | 12 +
arch/arm64/kernel/traps.c | 8 +-
arch/arm64/kvm/hyp/switch.c | 17 ++
arch/arm64/mm/fault.c | 5 +-
arch/arm64/mm/proc.S | 18 ++
drivers/irqchip/irq-gic-common.c | 10 +
drivers/irqchip/irq-gic-common.h | 2 +
drivers/irqchip/irq-gic-v3-its.c | 2 +-
drivers/irqchip/irq-gic-v3.c | 279 +++++++++++++++++++-----
include/linux/irqchip/arm-gic-common.h | 6 +
include/linux/irqchip/arm-gic.h | 5 -
33 files changed, 699 insertions(+), 119 deletions(-)

--
1.9.1

--
Julien Thierry