From: Alexandru Elisei <alexandru.elisei@arm.com>
To: Andrew Jones <drjones@redhat.com>
Cc: lvivier@redhat.com, linux-s390@vger.kernel.org, thuth@redhat.com,
frankja@linux.ibm.com, kvm@vger.kernel.org, david@redhat.com,
andre.przywara@arm.com, cohuck@redhat.com,
kvm-ppc@vger.kernel.org, vivek.gautam@arm.com, maz@kernel.org,
pbonzini@redhat.com, imbrenda@linux.ibm.com,
kvmarm@lists.cs.columbia.edu
Subject: Re: [kvm-unit-tests RFC PATCH 4/5] scripts: Generate kvmtool standalone tests
Date: Thu, 9 Sep 2021 15:42:54 +0100 [thread overview]
Message-ID: <e8560cdb-532c-0320-420d-c57d14cdae18@arm.com> (raw)
In-Reply-To: <20210909135429.dqreodxr7elpvmfm@gator>
Hi Drew,
On 9/9/21 2:54 PM, Andrew Jones wrote:
> On Thu, Sep 09, 2021 at 02:47:57PM +0100, Alexandru Elisei wrote:
>> Hi Drew,
>>
>> On 9/9/21 2:05 PM, Andrew Jones wrote:
>>> On Thu, Sep 09, 2021 at 12:11:52PM +0100, Alexandru Elisei wrote:
>>>> Hi Drew,
>>>>
>>>> On 9/8/21 5:07 PM, Andrew Jones wrote:
>>>>> On Wed, Sep 08, 2021 at 04:37:39PM +0100, Alexandru Elisei wrote:
>>>>>> Hi Drew,
>>>>>>
>>>>>> On 9/7/21 11:21 AM, Andrew Jones wrote:
>>>>>>> On Fri, Jul 02, 2021 at 05:31:21PM +0100, Alexandru Elisei wrote:
>>>>>>>> Add support for the standalone target when running kvm-unit-tests under
>>>>>>>> kvmtool.
>>>>>>>>
>>>>>>>> Example command line invocation:
>>>>>>>>
>>>>>>>> $ ./configure --target=kvmtool
>>>>>>>> $ make clean && make standalone
>>>>>>>>
>>>>>>>> Signed-off-by: Alexandru Elisei <alexandru.elisei@arm.com>
>>>>>>>> ---
>>>>>>>> scripts/mkstandalone.sh | 14 +++++++-------
>>>>>>>> 1 file changed, 7 insertions(+), 7 deletions(-)
>>>>>>>>
>>>>>>>> diff --git a/scripts/mkstandalone.sh b/scripts/mkstandalone.sh
>>>>>>>> index 16f461c06842..d84bdb7e278c 100755
>>>>>>>> --- a/scripts/mkstandalone.sh
>>>>>>>> +++ b/scripts/mkstandalone.sh
>>>>>>>> @@ -44,6 +44,10 @@ generate_test ()
>>>>>>>> config_export ARCH_NAME
>>>>>>>> config_export PROCESSOR
>>>>>>>>
>>>>>>>> + if [ "$ARCH" = "arm64" ] || [ "$ARCH" = "arm" ]; then
>>>>>>>> + config_export TARGET
>>>>>>>> + fi
>>>>>>> Should export unconditionally, since we'll want TARGET set
>>>>>>> unconditionally.
>>>>>> Yes, will do.
>>>>>>
>>>>>>>> +
>>>>>>>> echo "echo BUILD_HEAD=$(cat build-head)"
>>>>>>>>
>>>>>>>> if [ ! -f $kernel ]; then
>>>>>>>> @@ -59,7 +63,7 @@ generate_test ()
>>>>>>>> echo 'export FIRMWARE'
>>>>>>>> fi
>>>>>>>>
>>>>>>>> - if [ "$ENVIRON_DEFAULT" = "yes" ] && [ "$ERRATATXT" ]; then
>>>>>>>> + if [ "$TARGET" != "kvmtool" ] && [ "$ENVIRON_DEFAULT" = "yes" ] && [ "$ERRATATXT" ]; then
>>>>>>> I think it would be better to ensure that ENVIRON_DEFAULT is "no" for
>>>>>>> TARGET=kvmtool in configure.
>>>>>> From looking at the code, it is my understanding that with ENVIRON_DEFAULT=yes, an
>>>>>> initrd file is generated with the contents of erratatxt and other information, in
>>>>>> a key=value pair format. This initrd is then passed on to the test (please correct
>>>>>> me if I'm wrong). With ENVIRON_DEFAULT=no (set via ./configure
>>>>>> --disable-default-environ), this initrd is not generated.
>>>>>>
>>>>>> kvmtool doesn't have support for passing an initrd when loading firmware, so yes,
>>>>>> I believe the default should be no.
>>>>>>
>>>>>> However, I have two questions:
>>>>>>
>>>>>> 1. What happens when the user specifically enables the default environ via
>>>>>> ./configure --enable-default-environ --target=kvmtool? In my opinion, that should
>>>>>> be an error because the user wants something that is not possible with kvmtool
>>>>>> (loading an image with --firmware in kvmtool means that the initrd image it not
>>>>>> loaded into the guest memory and no node is generated for it in the dtb), but I
>>>>>> would like to hear your thoughts about it.
>>>>> As part of the forcing ENVIRON_DEFAULT to "no" for kvmtool in configure an
>>>>> error should be generated if a user tries to explicitly enable it.
>>>>>
>>>>>> 2. If the default environment is disabled, is it still possible for an user to
>>>>>> pass an initrd via other means? I couldn't find where that is implemented, so I'm
>>>>>> guessing it's not possible.
>>>>> Yes, a user could have a KVM_UNIT_TESTS_ENV environment variable set when
>>>>> they launch the tests. If that variable points to a file then it will get
>>>>> passed as an initrd. I guess you should also report a warning in arm/run
>>>>> if KVM_UNIT_TESTS_ENV is set which states that the environment file will
>>>>> be ignored when running with kvmtool.
>>>> Thank you for explaining it, I had looked at
>>>> scripts/arch-run.bash::initrd_create(), but it didn't click that setting the
>>>> KVM_UNIT_TESTS_ENV environment variable is enough to generate and use the initrd.
>>>>
>>>> After looking at the code some more, in the logs the -initrd argument is shown as
>>>> a comment, instead of an actual argument that is passed to qemu:
>>>>
>>>> timeout -k 1s --foreground 90s /usr/bin/qemu-system-aarch64 -nodefaults -machine
>>>> virt,gic-version=host,accel=kvm -cpu host -device virtio-serial-device -device
>>>> virtconsole,chardev=ctd -chardev testdev,id=ctd -device pci-testdev -display none
>>>> -serial stdio -kernel arm/cache.flat -smp 1 # -initrd /tmp/tmp.rUIZ3h9KLJ
>>>> QEMU_ACCEL = kvm
>>>> INFO: IDC-DIC: dcache clean to PoU required
>>>> INFO: IDC-DIC: icache invalidation to PoU required
>>>> PASS: IDC-DIC: code generation
>>>> SUMMARY: 1 tests
>>>>
>>>> This is done intentionally in scripts/arch-run.bash::run_qemu(). I don't
>>>> understand the reason for that. When I first looked at the logs, I was sure that
>>>> no initrd is passed to the test. I had to go dig through the scripts to figure out
>>>> that the "#" sign (which marks the beginning of a comment) is not present in the
>>>> qemu invocation.
>>> It's commented out because if you want to copy+paste the command line to
>>> use it again it'll fail to run because the temp file will be gone. Of
>>> course somebody depending on the environment for their test run will have
>>> other problems when it's gone, but those people can use the
>>> KVM_UNIT_TESTS_ENV variable to specify a non-temp file which includes the
>>> default environment and then configure without the default environment.
>>> The command line won't get the # in that case.
>> Hmm... wouldn't it make more sense then to generate the initrd in the logs
>> directory, and keep it there? To ensure the test runs can be reproduced manually,
>> if needed?
> Well, there's no logs directory for standalone tests, but I do like the
> idea of capturing the environment when possible. Possibly the best thing
> to do is to provide an option that, when enabled, says to dump the
> environment into the log before executing the test. That would be similar
> to how BUILD_HEAD is output first when running the tests standalone.
> Anyway, this is a good idea, but probably outside the scope of your
> kvmtool work unless the initrd thing is blocking you and you need to
> rework it anyway.
I don't need to change anything about how initrd works in kvm-unit-tests for my
kvmtool series, I was just curious to understand more about it. Thank you for the
explanations!
Thanks,
Alex
>
> Thanks,
> drew
>
>> Thanks,
>>
>> Alex
>>
>>> Thanks,
>>> drew
>>>
>>>> Thanks,
>>>>
>>>> Alex
>>>>
>>>>> There aren't currently any other ways to invoke the addition of the
>>>>> -initrd command line option, because so far we only support passing a
>>>>> single file to test (the environment "file"). If we ever want to pass
>>>>> more files, then we'd need to create a simple file system on the initrd
>>>>> and make it possible to add -initrd even when no environment is desired.
>>>>> But, that may never happen.
>>>>>
>>>>> Thanks,
>>>>> drew
>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> Alex
>>>>>>
>>>>>>>> temp_file ERRATATXT "$ERRATATXT"
>>>>>>>> echo 'export ERRATATXT'
>>>>>>>> fi
>>>>>>>> @@ -95,12 +99,8 @@ function mkstandalone()
>>>>>>>> echo Written $standalone.
>>>>>>>> }
>>>>>>>>
>>>>>>>> -if [ "$TARGET" = "kvmtool" ]; then
>>>>>>>> - echo "Standalone tests not supported with kvmtool"
>>>>>>>> - exit 2
>>>>>>>> -fi
>>>>>>>> -
>>>>>>>> -if [ "$ENVIRON_DEFAULT" = "yes" ] && [ "$ERRATATXT" ] && [ ! -f "$ERRATATXT" ]; then
>>>>>>>> +if [ "$TARGET" != "kvmtool" ] && [ "$ENVIRON_DEFAULT" = "yes" ] && \
>>>>>>>> + [ "$ERRATATXT" ] && [ ! -f "$ERRATATXT" ]; then
>>>>>>>> echo "$ERRATATXT not found. (ERRATATXT=$ERRATATXT)" >&2
>>>>>>>> exit 2
>>>>>>>> fi
>>>>>>>> --
>>>>>>>> 2.32.0
>>>>>>>>
>>>>>>> Thanks,
>>>>>>> drew
>>>>>>>
_______________________________________________
kvmarm mailing list
kvmarm@lists.cs.columbia.edu
https://lists.cs.columbia.edu/mailman/listinfo/kvmarm
next prev parent reply other threads:[~2021-09-09 14:41 UTC|newest]
Thread overview: 36+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-07-02 16:31 [kvm-unit-tests RFC PATCH 0/5] arm: Add kvmtool to the runner script Alexandru Elisei
2021-07-02 16:31 ` [kvm-unit-tests RFC PATCH 1/5] lib: arm: Print test exit status on exit if chr-testdev is not available Alexandru Elisei
2021-07-12 16:36 ` Andrew Jones
2021-09-06 10:20 ` Alexandru Elisei
2021-09-06 10:58 ` Andrew Jones
2021-09-06 11:06 ` Alexandru Elisei
2021-07-12 16:51 ` Andre Przywara
2021-07-12 17:07 ` Andrew Jones
2021-07-12 17:12 ` Nadav Amit
2021-07-02 16:31 ` [kvm-unit-tests RFC PATCH 2/5] scripts: Rename run_qemu_status -> run_test_status Alexandru Elisei
2021-07-12 16:37 ` Andrew Jones
2021-07-13 7:45 ` Thomas Huth
2021-07-02 16:31 ` [kvm-unit-tests RFC PATCH 3/5] run_tests.sh: Add kvmtool support Alexandru Elisei
2021-07-12 16:52 ` Andre Przywara
2021-09-06 10:28 ` Alexandru Elisei
2021-09-06 11:01 ` Andrew Jones
2021-09-06 11:07 ` Alexandru Elisei
2021-09-07 10:17 ` Andrew Jones
2021-09-08 14:33 ` Alexandru Elisei
2021-09-08 15:09 ` Andrew Jones
2021-09-08 15:46 ` Alexandru Elisei
2021-09-08 15:49 ` Andrew Jones
2021-09-09 11:33 ` Alexandru Elisei
2021-09-09 12:49 ` Andrew Jones
2021-07-02 16:31 ` [kvm-unit-tests RFC PATCH 4/5] scripts: Generate kvmtool standalone tests Alexandru Elisei
2021-09-07 10:21 ` Andrew Jones
2021-09-08 15:37 ` Alexandru Elisei
2021-09-08 16:07 ` Andrew Jones
2021-09-09 11:11 ` Alexandru Elisei
2021-09-09 13:05 ` Andrew Jones
2021-09-09 13:47 ` Alexandru Elisei
2021-09-09 13:54 ` Andrew Jones
2021-09-09 14:42 ` Alexandru Elisei [this message]
2021-07-02 16:31 ` [kvm-unit-tests RFC PATCH 5/5] configure: Ignore --erratatxt when --target=kvmtool Alexandru Elisei
2021-09-07 10:25 ` Andrew Jones
2021-09-08 16:13 ` Alexandru Elisei
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=e8560cdb-532c-0320-420d-c57d14cdae18@arm.com \
--to=alexandru.elisei@arm.com \
--cc=andre.przywara@arm.com \
--cc=cohuck@redhat.com \
--cc=david@redhat.com \
--cc=drjones@redhat.com \
--cc=frankja@linux.ibm.com \
--cc=imbrenda@linux.ibm.com \
--cc=kvm-ppc@vger.kernel.org \
--cc=kvm@vger.kernel.org \
--cc=kvmarm@lists.cs.columbia.edu \
--cc=linux-s390@vger.kernel.org \
--cc=lvivier@redhat.com \
--cc=maz@kernel.org \
--cc=pbonzini@redhat.com \
--cc=thuth@redhat.com \
--cc=vivek.gautam@arm.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).