mbox series

[RFC,v9,00/23] target/arm: Implement FEAT_NMI and FEAT_GICv3_NMI

Message ID 20240321130812.2983113-1-ruanjinjie@huawei.com (mailing list archive)
Headers show
Series target/arm: Implement FEAT_NMI and FEAT_GICv3_NMI | expand

Message

Jinjie Ruan March 21, 2024, 1:07 p.m. UTC
This patch set implements FEAT_NMI and FEAT_GICv3_NMI for armv8. These
introduce support for a new category of interrupts in the architecture
which we can use to provide NMI like functionality.

There are two modes for using this FEAT_NMI. When PSTATE.ALLINT or
PSTATE.SP & SCTLR_ELx.SCTLR_SPINTMASK is set, any entry to ELx causes all
interrupts including those with superpriority to be masked on entry to ELn
until the mask is explicitly removed by software or hardware. PSTATE.ALLINT
can be managed by software using the new register control ALLINT.ALLINT.
Independent controls are provided for this feature at each EL, usage at EL1
should not disrupt EL2 or EL3.

I have tested it with the following linux patches which try to support
FEAT_NMI in linux kernel:

	https://lore.kernel.org/linux-arm-kernel/Y4sH5qX5bK9xfEBp@lpieralisi/T/#mb4ba4a2c045bf72c10c2202c1dd1b82d3240dc88

In the test, SGI, PPI and SPI interrupts can all be set to have super priority
to be converted to a hardware NMI interrupt. The SGI is tested with kernel
IPI as NMI framework, softlockup, hardlockup and kgdb test cases, and the PPI
interrupt is tested with "perf top" command with hardware NMI enabled, and
the SPI interrupt is tested with a custom test module, in which NMI interrupts
can be received and sent normally.

And the Virtual NMI(VNMI) SGI, PPI and SPI interrupt has also been tested
by accessing the GICD_INMIR, GICR_INMIR0, GICR_ISPENDR0 and GICD_ISPENDR
registers with devmem command.

         +-------------------------------------------------+
         |               Distributor                       |
         +-------------------------------------------------+
             SPI |  NMI                         |  NMI
                \/                            \/
            +--------+                     +-------+
            | Redist |                     | Redist|
            +--------+                     +-------+
            SGI  | NMI                     PPI | NMI
                \/                            \/
          +-------------+             +---------------+
          |CPU Interface|   ...       | CPU Interface |
          +-------------+             +---------------+
               | NMI                         | NMI
              \/                            \/
            +-----+                       +-----+
            |  PE |                       |  PE |
            +-----+                       +-----+

Changes in v9:
- Move nmi_reginfo and related functions inside an existing ifdef
  TARGET_AARCH64 to solve the --target-list=aarch64-softmmu,arm-softmmu
  compilation problem.
- Check 'isread' when writing to ALLINT.
- Update the GPIOs passed in the arm_cpu_kvm_set_irq, and update the comment.
- Definitely not merge VINMI and VFNMI into EXCP_VNMI.
- ARM_CPU_VNMI -> ARM_CPU_VINMI, CPU_INTERRUPT_VNMI -> CPU_INTERRUPT_VINMI.
- Update VINMI and VFNMI when writing HCR_EL2 or HCRX_EL2.
- Update the commit subject and message, VNMI -> VINMI.
- Handle CPSR_F and ISR_FS according to CPU_INTERRUPT_VFNMI instead of
  CPU_INTERRUPT_VFIQ and HCRX_EL2.VFNMI.
- Not check SCTLR_NMI in arm_cpu_do_interrupt_aarch64().
- Correct the INTID_NMI logic.
- Declare cpu variable to reuse latter.

Changes in v8:
- Fix the rcu stall after sending a VNMI in qemu VM.
- Fix an unexpected interrupt bug when sending VNMI by running qemu VM.
- Test the VNMI interrupt.
- Update the commit message.
- Add Reviewed-by.

Changes in v7:
- env->cp15.hcrx_el2 -> arm_hcrx_el2_eff().
- Reorder the irqbetter() code for clarity.
- Eliminate the has_superprio local variable for gicv3_get_priority().
- false -> cs->hpplpi.superprio in gicv3_redist_update_noirqset().
- 0x0 -> false in arm_gicv3_common_reset_hold().
- Clear superprio in several places for hppi, hpplpi and hppvlpi.
- Add Reviewed-by.

Changes in v6:
- Fix DISAS_TOO_MANY to DISAS_UPDATE_EXIT for ALLINT MSR (immediate).
- Verify that HCR_EL2.VF is set before checking VFNMI.
- env->cp15.hcr_el2 -> arm_hcr_el2_eff().
- env->cp15.hcrx_el2 -> arm_hcrx_el2_eff().
- Not combine VFNMI with CPU_INTERRUPT_VNMI.
- Implement icv_nmiar1_read().
- Put the "extract superprio info" code into gicv3_get_priority().
- Update the comment in irqbetter().
- Reset the cs->hppi.superprio to 0x0.
- Set hppi.superprio to false for LPI.
- Add Reviewed-by.

Changes in v5:
- Remove the comment for ALLINT in cpu.h.
- Merge allint_check() to msr_i_allint to clear the ALLINT MSR (immediate)
  implementation.
- Rename msr_i_allint() to msr_set_allint_el1() to make it clearer.
- Drop the & 1 in trans_MSR_i_ALLINT().
- Add Reviewed-by.

Changes in v4:
- Handle VNMI within the CPU and the GIC.
- Keep PSTATE.ALLINT in env->pstate but not env->allint.
- Fix the ALLINT MSR (immediate) decodetree implementation.
- Accept NMI unconditionally for arm_cpu_has_work() but add comment.
- Improve nmi mask in arm_excp_unmasked().
- Make the GICR_INMIR0 and GICD_INMIR implementation more clearer.
- Improve ICC_NMIAR1_EL1 implementation
- Extract gicv3_get_priority() to avoid priority code repetition.
- Add Reviewed-by.

Changes in v3:
- Remove the FIQ NMI.
- Adjust the patches Sequence.
- Reomve the patch "Set pstate.ALLINT in arm_cpu_reset_hold".
- Check whether support FEAT_NMI and FEAT_GICv3 for FEAT_GICv3_NMI.
- With CPU_INTERRUPT_NMI, both CPSR_I and ISR_IS must be set.
- Not include NMI logic when FEAT_NMI or SCTLR_ELx.NMI not enabled.
- Refator nmi mask in arm_excp_unmasked().
- Add VNMI definitions, add HCRX_VINMI and HCRX_VFNMI support in HCRX_EL2.
- Add Reviewed-by and Acked-by.
- Update the commit message.

Changes in v2:
- Break up the patches so that each one does only one thing.
- Remove the command line option and just implement it in "max" cpu.

Jinjie Ruan (23):
  target/arm: Handle HCR_EL2 accesses for bits introduced with FEAT_NMI
  target/arm: Add PSTATE.ALLINT
  target/arm: Add support for FEAT_NMI, Non-maskable Interrupt
  target/arm: Implement ALLINT MSR (immediate)
  target/arm: Support MSR access to ALLINT
  target/arm: Add support for Non-maskable Interrupt
  target/arm: Add support for NMI in arm_phys_excp_target_el()
  target/arm: Handle IS/FS in ISR_EL1 for NMI, VINMI and VFNMI
  target/arm: Handle PSTATE.ALLINT on taking an exception
  hw/arm/virt: Wire NMI and VINMI irq lines from GIC to CPU
  hw/intc/arm_gicv3: Add external IRQ lines for NMI
  target/arm: Handle NMI in arm_cpu_do_interrupt_aarch64()
  hw/intc/arm_gicv3: Add irq superpriority information
  hw/intc/arm_gicv3_redist: Implement GICR_INMIR0
  hw/intc/arm_gicv3: Implement GICD_INMIR
  hw/intc: Enable FEAT_GICv3_NMI Feature
  hw/intc/arm_gicv3: Add NMI handling CPU interface registers
  hw/intc/arm_gicv3: Handle icv_nmiar1_read() for icc_nmiar1_read()
  hw/intc/arm_gicv3: Implement NMI interrupt prioirty
  hw/intc/arm_gicv3: Report the NMI interrupt in gicv3_cpuif_update()
  hw/intc/arm_gicv3: Report the VINMI interrupt
  target/arm: Add FEAT_NMI to max
  hw/arm/virt: Add FEAT_GICv3_NMI feature support in virt GIC

 docs/system/arm/emulation.rst      |   1 +
 hw/arm/virt.c                      |  25 +++++-
 hw/intc/arm_gicv3.c                |  69 ++++++++++++++--
 hw/intc/arm_gicv3_common.c         |  10 +++
 hw/intc/arm_gicv3_cpuif.c          | 127 +++++++++++++++++++++++++++--
 hw/intc/arm_gicv3_dist.c           |  36 ++++++++
 hw/intc/arm_gicv3_redist.c         |  22 +++++
 hw/intc/gicv3_internal.h           |   8 ++
 hw/intc/trace-events               |   2 +
 include/hw/intc/arm_gic_common.h   |   2 +
 include/hw/intc/arm_gicv3_common.h |   7 ++
 target/arm/cpu-features.h          |   5 ++
 target/arm/cpu-qom.h               |   5 +-
 target/arm/cpu.c                   | 124 ++++++++++++++++++++++++++--
 target/arm/cpu.h                   |   9 ++
 target/arm/helper.c                | 102 +++++++++++++++++++++--
 target/arm/internals.h             |  21 +++++
 target/arm/tcg/a64.decode          |   1 +
 target/arm/tcg/cpu64.c             |   1 +
 target/arm/tcg/helper-a64.c        |  12 +++
 target/arm/tcg/helper-a64.h        |   1 +
 target/arm/tcg/translate-a64.c     |  19 +++++
 22 files changed, 575 insertions(+), 34 deletions(-)

Comments

Peter Maydell March 21, 2024, 4:40 p.m. UTC | #1
On Thu, 21 Mar 2024 at 13:10, Jinjie Ruan <ruanjinjie@huawei.com> wrote:
>
> This patch set implements FEAT_NMI and FEAT_GICv3_NMI for armv8. These
> introduce support for a new category of interrupts in the architecture
> which we can use to provide NMI like functionality.
>
> There are two modes for using this FEAT_NMI. When PSTATE.ALLINT or
> PSTATE.SP & SCTLR_ELx.SCTLR_SPINTMASK is set, any entry to ELx causes all
> interrupts including those with superpriority to be masked on entry to ELn
> until the mask is explicitly removed by software or hardware. PSTATE.ALLINT
> can be managed by software using the new register control ALLINT.ALLINT.
> Independent controls are provided for this feature at each EL, usage at EL1
> should not disrupt EL2 or EL3.
>
> I have tested it with the following linux patches which try to support
> FEAT_NMI in linux kernel:
>
>         https://lore.kernel.org/linux-arm-kernel/Y4sH5qX5bK9xfEBp@lpieralisi/T/#mb4ba4a2c045bf72c10c2202c1dd1b82d3240dc88
>
> In the test, SGI, PPI and SPI interrupts can all be set to have super priority
> to be converted to a hardware NMI interrupt. The SGI is tested with kernel
> IPI as NMI framework, softlockup, hardlockup and kgdb test cases, and the PPI
> interrupt is tested with "perf top" command with hardware NMI enabled, and
> the SPI interrupt is tested with a custom test module, in which NMI interrupts
> can be received and sent normally.

It looks like your changes to the GIC have missed the handling
of NMIs in the active priority registers and the running
priority registers (both ICC and ICV versions). The way this
works is that the ICH_AP1R0 and ICC_AP1R0 registers get an
extra bit for NMI status in bit 63 (luckily we had the foresight
to make these struct fields be uint64_t). When we activate
an NMI IRQ, instead of setting the ICC_APR bit according to
its priority, we set the NMI bit instead. Similarly, on
deactivate we clear the NMI bit, not the priority-related bit.
The ICC_RPR register also has new bits in bit 63 and 62 for
whether there's an active NMI. On read of the RPR we figure
out the values for these bits based on the NMI bits in
ICC_AP1R_EL1S.NMI and ICC_AP1R_EL1NS.NMI (you might find the
pseudocode functions for ICC_RPR_EL1 in chapter 13 helpful to
look at here). The icc_highest_active_prio() likely also needs
changes to handle NMI. Similarly for all the ICV versions of
these registers and functions.

thanks
-- PMM
Peter Maydell March 22, 2024, 11:01 a.m. UTC | #2
On Thu, 21 Mar 2024 at 13:10, Jinjie Ruan <ruanjinjie@huawei.com> wrote:
>
> This patch set implements FEAT_NMI and FEAT_GICv3_NMI for armv8. These
> introduce support for a new category of interrupts in the architecture
> which we can use to provide NMI like functionality.

By the way, when you send the next version of this you can
drop the "RFC" tag -- I think this patchset is in basically
good shape and we're just into the process of finding minor
bugs and missing pieces.

thanks
-- PMM