Tree:
https://github.com/rth7680/qemu.git tgt-arm-vhe-5
Testcase:
qemu-test:~rth/linux/initramfs-min.cpio.gz
The host kernel could be anything, but I've been using
the same Image.gz that is inside the cpio archive.
./aarch64-softmmu/qemu-system-aarch64 -m 4G \
-M virt,virtualization=on,gic-version=max -cpu max \
-kernel Image.gz -initrd initramfs-min.cpio.gz
At the shell prompt, ./test will run a guest kernel with kvm.
As momentarily discussed with PMM in the hallway:
As soon as the guest kernel enables interrupts,
arch_timer_starting_cpu
enable_percpu_irq
irq_percpu_enable
gic_unmask_irq
-- Incorrect exception delivery.
the GTIMER_PHYS interrupt is delivered to EL2 (seems to be ok), the host kernel
does something (haven't dug into what exactly, bug presumably setting bits that
are supposed to pass the virq to the guest), and immediately another interrupt
is delivered to EL2. Repeat.
Whether this is incorrect routing of the virq interrupt, or incorrect
masking/acking of the hard irq interrupt at EL2, I do not yet know.
PMM: I don't know the answer to either (a) or (b) as asked on hangouts. I
think (b) is correct, but I can't be sure. I'm trying to understand how (a) is
supposed to work now. In particular, I can't find any code that sets
HCR_EL2.{VI,VF}, only tests them.
r~
# Progress #
o Ramp up
* Concluded.
o Annual Review
* Attended discussions.
o Upstream GDB
* Patch reviews on gerrit.
* Answered questions.
* The state of ARM/AArch64 GDB upstream seems to be reasonable, with
a few failures here and there. Numerous failures on problematic racy
tests (gdb.threads).
o GDB:
* GNU-170 - GDB BZ #21221 - gdb hangs while stepping an empty loop
- Spent some more time on this and improve the patch further,
covering most problematic cases for "for", "while" and "do/while" loops.
* Read some documentation on setting up Fast Models for testing MVE
(Helium).
* Inspected various aspects of ARM support in GDB, like SVE, PAC and
ACLE.
# Plan #
o Annual Review
* Conclude.
o GDB:
* GNU-170 - GDB BZ #21221 - gdb hangs while stepping an empty loop
- Discuss with gcc@ a bit more about my proposed solution.
* Attempt to setup system QEMU and/or Fast Model for testing ACLE SVE
and, maybe, MVE.
== Progress ==
* Out of office 1 day
* Buildbot monitoring
- Moved the buildbots to pull from github
* Trying to setup a build environment on ex40-01
- Gave up on the tcwg-sq-01/2 boards because they seemed too unstable
* Still no access to Morello docs
* Playing with lldb python scripting
- Got a script that intercepts all calls to
VectorType::getNumElements that don't come from a getElementCount
(since that likely means that they won't be preserving the 'scalable'
property)
- This should help figure out problems spotted by the fuzzer
- Likely needs a bit more refining
== Plan ==
* More of the same
* Out of office on Friday (1 November)
== Progress ==
* GCC:
- -mpure-code on v6m: no feedback yet
* FDPIC/GDB
- problems with the board I used, it hangs shortly after or during
boot. None of the workaround/fixes suggested to me worked. Having an
stm32 qemu config would help.
* GCC upstream validation:
- reported several issues
* misc:
- infra fixes / troubleshooting / reviews
- sent 2 small qemu patches (fix vmrs support for m-profile, and add cortex-m7)
- confirmed that gcc LTO profiled bootstrap works on arm with recent
trunk, although it takes ages. Will need to try on a more powerful
board
== Next ==
* Holidays next week, back Nov 4th
* FDPIC: resume work on GDB: check the various qemu forks with stm32
board support.
Add FDPIC configuration in the GCC trunk validation.
* GCC: pure-code/v6m, handle feedback
* Binutils: support non-contiguous memory regions in linker
Progress:
* VIRT-65 [QEMU upstream maintainership]
+ lots of wrangling of patches and pulls since I'm away
next week and it's also going to be softfreeze
+ preparation for KVM Forum next week
thanks
-- PMM
Progress:
[VIRT-344 # ARMv8.5-MemTag, Memory Tagging Extension ]
Updates for user-only.
Merge bug fixes from eugeni.stepanov(a)gmail.com.
[VIRT-349 # QEMU SVE2 Supprt ]
Convert neon pmul helpers to a form that will be usable for sve2.
[VIRT-327 # Richard's upstream QEMU work ]
Pull for tcg-next.
Review plugins v5.
Update for capstone submodule.
Started reviewing multi-phase reset v5.
[Kernel]
Hacked up a patch for ARMv8.5-RNG.
r~
