Linaro
- On buildbot monitoring duty, relatively quiet week with just a
couple of fairly simple to diagnose problems to report.
Morello
- Dynamic linking progressing albeit slowly.
-- Trying to work out the requirements from existing documents and
implementation.
-- Have some simple cases doing mostly the right thing and have
written some tests.
-- Will need to rewrite to move calculations earlier in the link-step.
-- Morello is quite different from Cheri in this regard so I have had
to diverge much more from the implementation.
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~
# Progress #
o Ramp up
* Credentials, machine access and LDAP updates done.
o Qualcomm Landing Team sunsetting
* Returned Qualcomm's Laptop.
o Upstream GDB
* Ramping up on reviews.
* Gathering data on the current state of GDB on ARM.
o GDB:
* GNU-170 - GDB BZ #21221 - gdb hangs while stepping an empty loop
- Came up with a little hack/proof-of-concept to get this fixed.
Though ugly, it seems fixing this in the front-end may make more sense,
as the information i need (source line) is easily accessible in there.
- Discussion ongoing with gcc@. GDB clearly needs the compiler to
provide more information.
* Created JIRA cards for all known pending ARM tasks for GDB, based
on Alan's and Joey's input. TODO-ed all of them for the time being.
# Plan #
o GDB:
* GNU-170 - GDB BZ #21221 - gdb hangs while stepping an empty loop
- Continue pursuing a fix.
* Prioritize GDB JIRA cards and start work on them.
[Morello]
- Got static linking support to the point that I can successfully link
with LLD the coremark, dhrystone and EEMBC from the arran-toolchain.
Not got any outstanding failures to investigate.
- Altered LLD so a linker script is no longer necessary for newlib.
- Started the process of rebasing and adding tests for all the
fixes/hacks I needed to make to the linker work.
- Aligned the base and limit of capabilities according to the incoming
CHERI concentrate scheme. Interesting question of what should a linker
do when alignment requirements on the base and limit cross section
boundaries, and what are the responsibilities for an object producer
when creating a section when the length of the capability is known at
compile time.
Planned absences:
Holiday Thursday, Friday (24th, 25th October)
Progress:
* VIRT-65 [QEMU upstream maintainership]
+ finishing off the ptimer API transition work
+ review of rth's "speed up calculation of tbflags" patchset
+ put together and sent an arm pullreq
* VIRT-350 [Update Arm KVM support in QEMU]
+ the patchset for hotpluggable RAM support is now upstream
thanks
-- PMM
== Progress ==
* GCC:
- Work on -mpure-code on v6m. Patches sent for upstream review.
* GCC upstream validation:
- reported several issues
* misc:
- infra fixes / troubleshooting / reviews
== Next ==
* FDPIC: resume work on GDB
* GCC: pure-code/v6m
* Binutils: support non-contiguous memory regions in linker
== Progress ==
* clang-tidy workshop (and associated prep)
- I think this went really well, we got good feedback from some of
the participants
* Trying to setup a build environment on tcwg-sq-02.tcwglab
- Mostly so I can deploy the SVE fuzzer there, but maybe for other things too
- All the compilers that I've tried so far are ICE-ing at some point
or another while building llvm
- Still looking into it but I'm starting to suspect there's
something fishy about this board
* Still no access to Morello docs
* Read a bit more about LLDB
* Finished annual review
== Plan ==
* Maybe THIS time I'll get access to the Morello docs next week
* If not, more SVE fuzzer
* One day off
o LLVM:
* Buildbots babysitting:
- Various breakage on the bots and in the kernel build
* Machine Outliner:
- preparing upstream submission
o Misc
* Various meetings and discussions.
== Progress ==
* GCC:
- Work on -mpure-code on v6m
* GCC upstream validation:
- reported several issues
* misc:
- infra fixes / troubleshooting / reviews
- watched a couple of GNU Cauldron presentations
== Next ==
* FDPIC: resume work on GDB
* GCC: pure-code/v6m
* Binutils: support non-contiguous memory regions in linker
[VIRT-263 # ARMv8.1-VHE Virtual Host Extensions ]
Still need to think of more test cases...
[VIRT-344 # ARMv8.5-MemTag, Memory Tagging Extension ]
Posted v5 of the system-only patch set,
with testing help from Alex.
[VIRT-327 # Richard's upstream QEMU work ]
Catching up on patch review
- arm semihosting
- tcg profiler
- ptimer transactions
- s390 mvcl interrupt
- started on v2 of dave martin's bti kernel patch set.
Posted v6 of my arm hflags patch set.
r~
== Progress ==
* Support Morello fat pointers in LLDB [LLVM-597]
- Read an intro to Cheri (research project that Morello is based on)
- Read more LLDB docs
* Started writing annual review
* Setup VM for a clang-tidy workshop that I'm co-organizing as part of
the Stockholm LLVM socials
== Plan ==
* Hopefully will get access to Morello docs next week so I can start actual work
* Rebase and play more with the SVE IR fuzzer [LLVM-586]
[VIRT-263 # ARMv8.1-VHE Virtual Host Extensions ]
Lots of work with Alex trying to produce a reduced test case.
We are now unit testing entry and exit from EL0 (EL2&0),
EL1 and EL0-in-EL1 (EL1&0).
Next would be to test the various memory access faults.
[VIRT-327 # Richard's upstream QEMU work ]
Patch review for SVE in KVM, S390 interrupt handling during MVCL.
r~
Short week (2 days off)
== Progress ==
* GCC:
- looked at what's needed to enable -mexecute-only on v6m
* GCC upstream validation:
- reported a couple of issues.
* misc:
- infra fixes / troubleshooting / reviews
- catching up after Connect (internal debrief, ...)
- started looking at GNU Cauldron presentations
== Next ==
* FDPIC: resume work on GDB
* GCC: execute-only/v6m
* Binutils: support non-contiguous memory regions in linker
Slightly overlong this time as it covers pre&post connect weeks.
QEMU Tooling ([VIRT-252])
=========================
[VIRT-252] https://projects.linaro.org/browse/VIRT-252
[VIRT-280] https://projects.linaro.org/browse/VIRT-280
QEMU plugin support ([VIRT-280])
- sporadic work on the [v5 branch]
- posted {PATCH v6 0/6} semihosting cleanups (plus minor tests/tcg
tweak) Message-Id: <20190913151845.12582-1-alex.bennee(a)linaro.org>
- now merged - delta down a little ;-)
[VIRT-280] https://projects.linaro.org/browse/VIRT-280
[v5 branch] https://github.com/stsquad/qemu/tree/plugins/plugins-v5
GSoC Mentoring Afermath ([VIRT-348])
- more work preparing [subset for final list review]
- stats, CONFIG_PROFILER and perf integration
- dropped coverset and dot diagram as a bit too rough
- however should form a good basis going forward
[VIRT-348] https://projects.linaro.org/browse/VIRT-384
[subset for final list review]
https://github.com/stsquad/qemu/tree/tcg/tbstats-and-perf
ARMv8.1 VHE Extensions ([VIRT-263])
===================================
- worked with rth to get a minimal testcase
- very messy [wip branch]
[VIRT-263] https://projects.linaro.org/browse/VIRT-263
[wip branch] https://github.com/rth7680/qemu/tree/test-vhe
Upstream Work ([VIRT-109])
==========================
- posted {PATCH v3 00/33} testing/next (docker,tcg, alpha ;-)
Message-Id: <20190924210106.27117-1-alex.bennee(a)linaro.org>
- posted {RFC PATCH} configure: deprecate 32 bit build hosts
Message-Id: <20190925233013.6449-1-alex.bennee(a)linaro.org> mostly to
stimulate discussion of our modest proposal
- posted {PULL 00/28} testing updates (docker,podman,tcg,alpha)
Message-Id: <20190926183553.13895-1-alex.bennee(a)linaro.org>
- posted {PATCH} accel/kvm: ensure ret always set Message-Id:
<20191002102212.6100-1-alex.bennee(a)linaro.org>
[VIRT-109] https://projects.linaro.org/browse/VIRT-109
Other Activities
================
- More Connect travel administrava (airport transfer booked now)
- Connect itself
- many interesting talks
- discussions with bemi w.r.t migration and command line opts
- discussions with rth w.r.t 64/32, VHE and PR process
- discussions with FutureWei w.r.t scaling QEMU emulation
- KVM Forum administrava
Completed Reviews [2/2]
=======================
{PATCH} configure: Remove s390 (31-bit mode) from the list of supported CPUs
Message-Id: <20190928190334.6897-1-thuth(a)redhat.com>
{PATCH v2} s390x/tcg: MVCL: Exit to main loop if requested
Message-Id: <20191002082636.7739-1-david(a)redhat.com>
Absences
========
- KVM Forum Oct 29th-Nov 1st
Current Review Queue
====================
* {PATCH v2 00/15} target/arm: Implement semihosting v2.0
Message-Id: <20190916141544.17540-1-peter.maydell(a)linaro.org>
Added: <2019-10-03 Thu>
* {PATCH 00/19} hw/arm/raspi: Improve Raspberry Pi 2/3 reliability
Message-Id: <20190926173428.10713-1-f4bug(a)amsat.org>
Added: <2019-09-27 Fri>
* {PATCH RFC} docker: automatic dependencies for dockerfiles
Message-Id: <20190920001823.23279-1-jsnow(a)redhat.com>
Added: <2019-09-24 Tue>
* {PATCH v4 0/9} target/arm/kvm: enable SVE in guests
Message-Id: <20190924113105.19076-1-drjones(a)redhat.com>
Added: <2019-09-24 Tue>
--
Alex Bennée
On buildbot duty
- Several buildbot failures to investigate along with one linux kernel
regression
- Didn't manage to get the libcxx buildbot failures resolved on time.
I think the community are still trying to fix it (Script needs to be
Python2/Python3
Started work on LLD support for Morello
- Getting familiar with the toolchain
- First target is sufficient support to statically link the Howdy
"hello world" equivalent.
- Made the first couple of local patches to implement the easiest set
of static relocations.
Some upstream LLD patch review.
