Message-ID: <CA+CK2bCwqCHr6oecmL6R65qjgXZNuzU7SJ9diVkYX442tUdU+Q@mail.gmail.com>
Date: Thu, 13 Nov 2025 14:59:29 -0500
From: Pasha Tatashin <pasha.tatashin@...een.com>
To: "Yanjun.Zhu" <yanjun.zhu@...ux.dev>
Cc: Pratyush Yadav <pratyush@...nel.org>, akpm@...ux-foundation.org, brauner@...nel.org, 
	corbet@....net, graf@...zon.com, jgg@...pe.ca, linux-kernel@...r.kernel.org, 
	linux-kselftest@...r.kernel.org, linux-mm@...ck.org, masahiroy@...nel.org, 
	ojeda@...nel.org, rdunlap@...radead.org, rppt@...nel.org, tj@...nel.org
Subject: Re: [PATCH v9 1/9] kho: make debugfs interface optional

On Wed, Nov 12, 2025 at 9:11 PM Yanjun.Zhu <yanjun.zhu@...ux.dev> wrote:
>
>
> On 11/12/25 5:41 PM, Yanjun.Zhu wrote:
> >
> > On 11/11/25 7:26 AM, Pasha Tatashin wrote:
> >> On Mon, Nov 10, 2025 at 11:11 PM Zhu Yanjun <yanjun.zhu@...ux.dev>
> >> wrote:
> >>> In FC42, when I run "./luo_multi_session"
> >>>
> >>> # ./luo_multi_session
> >>> # [STAGE 1] Starting pre-kexec setup for multi-session test...
> >>> # [STAGE 1] Creating state file for next stage (2)...
> >>> # [STAGE 1] Creating empty sessions 'multi-test-empty-1' and
> >>> 'multi-test-empty-2'...
> >>> # [STAGE 1] Creating session 'multi-test-files-1' with one memfd...
> >>> # [STAGE 1] Creating session 'multi-test-files-2' with two memfds...
> >>> # [STAGE 1] Executing kexec...
> >>>
> >>> Then the system hang. And via virt-viewer, a calltrace will appear.
> >> Looks like mountroot fails, are you passing the same kernel parameters
> >> as the during cold boot?
> >> i.e. kexec -l -s --reuse-cmdline --initrd=[initramfs] [kernel]
> >
> >
> > Thanks a lot. It can work now.  The logs are as below:
> >
> > "
> >
> > # [STAGE 2] Starting post-kexec verification...
> > # [STAGE 2] Retrieving all sessions...
> > # [STAGE 2] Verifying contents of session 'multi-test-files-1'...
> > # [STAGE 2] Verifying contents of session 'multi-test-files-2'...
> > # [STAGE 2] Test data verified successfully.
> > # [STAGE 2] Finalizing all test sessions...
> > # [STAGE 2] Finalizing state session...
> > #
> > --- MULTI-SESSION KEXEC TEST PASSED ---
> > "
> >
> > Yanjun.Zhu
>
>
> Hi, Pasha
>
> In my tests, I found that luo_kexec_simple and luo_multi_session
> currently depend on the glibc-static library.
> If this library is not installed, build errors occur.
> By making the following changes, luo_kexec_simple and luo_multi_session
> would no longer depend on glibc-static, reducing external library
> dependencies.
> I am not sure whether you agree with these proposed changes.
>
> diff --git a/tools/testing/selftests/liveupdate/Makefile
> b/tools/testing/selftests/liveupdate/Makefile
> index 6ee6efeec62d..b226b9976150 100644
> --- a/tools/testing/selftests/liveupdate/Makefile
> +++ b/tools/testing/selftests/liveupdate/Makefile
> @@ -3,7 +3,6 @@
>   KHDR_INCLUDES ?= -I../../../../usr/include
>   CFLAGS += -Wall -O2 -Wno-unused-function
>   CFLAGS += $(KHDR_INCLUDES)
> -LDFLAGS += -static

Hi Yanjun,

Thank you for testing. I prefer to keep the '-static' flag because
these self-test files are not executed in the same environment where
they are compiled but in a VM which might have a different userspace.

Thank you,
Pasha

>   OUTPUT ?= .
>
>   # --- Test Configuration (Edit this section when adding new tests) ---
>
> Yanjun.Zhu
>
> >
> >
> >>
> >> Pasha
> >>
> >>> The call trace is in the attachment.
> >>>
> >>> Yanjun.Zhu
> >>>
> >>> 在 2025/11/10 7:26, Pasha Tatashin 写道:
> >>>> On Mon, Nov 10, 2025 at 8:16 AM Pratyush Yadav
> >>>> <pratyush@...nel.org> wrote:
> >>>>> On Sun, Nov 09 2025, Zhu Yanjun wrote:
> >>>>>
> >>>>>> 在 2025/11/8 10:13, Pasha Tatashin 写道:
> >>>>>>> On Fri, Nov 7, 2025 at 6:36 PM Yanjun.Zhu <yanjun.zhu@...ux.dev>
> >>>>>>> wrote:
> >>>>>>>> On 11/7/25 4:02 AM, Pasha Tatashin wrote:
> >>>>>>>>> On Fri, Nov 7, 2025 at 7:00 AM Pasha Tatashin
> >>>>>>>>> <pasha.tatashin@...een.com> wrote:
> >>>>>>>>>>> Hi, Pasha
> >>>>>>>>>>>
> >>>>>>>>>>> In our previous discussion, we talked about counting the
> >>>>>>>>>>> number of times
> >>>>>>>>>>> the kernel is rebooted via kexec. At that time, you
> >>>>>>>>>>> suggested adding a
> >>>>>>>>>>> variable in debugfs to keep track of this count.
> >>>>>>>>>>> However, since debugfs is now optional, where would be an
> >>>>>>>>>>> appropriate
> >>>>>>>>>>> place to store this variable?
> >>>>>>>>>> It is an optional config and can still be enabled if the live
> >>>>>>>>>> update
> >>>>>>>>>> reboot number value needs to be accessed through debugfs.
> >>>>>>>>>> However,
> >>>>>>>>>> given that debugfs does not guarantee a stable interface,
> >>>>>>>>>> tooling
> >>>>>>>>>> should not be built to require these interfaces.
> >>>>>>>>>>
> >>>>>>>>>> In the WIP LUO [1] I have, I pr_info() the live update number
> >>>>>>>>>> during
> >>>>>>>>>> boot and also store it in the incoming LUO FDT tree, which
> >>>>>>>>>> can also be
> >>>>>>>>>> accessed through this optional debugfs interface.
> >>>>>>>>>>
> >>>>>>>>>> The pr_info message appears like this during boot:
> >>>>>>>>>> [    0.000000] luo: Retrieved live update data, liveupdate
> >>>>>>>>>> number: 17
> >>>>>>>>>>
> >>>>>>>>>> Pasha
> >>>>>>>>> Forgot to add link to WIP LUOv5:
> >>>>>>>>> [1] https://github.com/soleen/linux/tree/luo/v5rc04
> >>>>>>>> Thanks a lot. I’ve carefully read this commit:
> >>>>>>>> https://github.com/soleen/linux/commit/60205b9a95c319dc9965f119303a1d83f0ff08fa.
> >>>>>>>>
> >>>>>>>>
> >>>>>>>> To be honest, I’d like to run some tests with who/luo,
> >>>>>>>> including the
> >>>>>>>> selftest for kho/luo. Could you please share the steps with me?
> >>>>>>>>
> >>>>>>>> If the testing steps have already been documented somewhere,
> >>>>>>>> could you
> >>>>>>>> please share the link?
> >>>>>>> Currently the test performs in-kernel tests for FLB data, it
> >>>>>>> creates a
> >>>>>>> number of FLB for every registered LUO file-handler, which at the
> >>>>>>> moment is only memfd.
> >>>>>>>
> >>>>>>> It works together with any of the kexec based live update tests. In
> >>>>>>> v5, I introduce two tests:
> >>>>>>> luo_kexec_simple
> >>>>>>> luo_multi_session
> >>>>>>>
> >>>>>>> For example, with luo_multi_session:
> >>>>>> Hi, Pasha
> >>>>>>
> >>>>>> I enabled "CONFIG_LIVEUPDATE=y"
> >>>>>>
> >>>>>> # ./luo_multi_session
> >>>>>> 1..0 # SKIP Failed to open /dev/liveupdate. Is the luo module
> >>>>>> loaded?
> >>>>>>
> >>>>>> # ls /dev/liveupdate
> >>>>>> ls: cannot access '/dev/liveupdate': No such file or directory
> >>>>>>
> >>>>>> # grep "LIVEUPDATE" -inrHI /boot/config-`uname -r`
> >>>>>> /boot/config-next-20251107-luo+:349:CONFIG_LIVEUPDATE=y
> >>>>>> /boot/config-next-20251107-luo+:11985:CONFIG_LIVEUPDATE_TEST=y
> >>>>>>
> >>>>>> I made tests on FC42. But /dev/liveupdate is missing.
> >>>>> You need to add liveupdate=1 to your kernel cmdline to enable LUO and
> >>>>> get /dev/liveupdate.
> >>>> +1, kernel parameters require: kho=1 liveupdate=1
> >>>>
> >>>>> Pasha, your LUO series doesn't add the liveupdate parameter to
> >>>>> kernel-parameters.txt. I think it should be done in the next
> >>>>> version to
> >>>>> this parameter is discoverable.
> >>>> Yeah, that is missing, I will update that in a standalone patch, or in
> >>>> a next version.
> >>>>
> >>>> Thanks,
> >>>> Pasha
> >>> --
> >>> Best Regards,
> >>> Yanjun.Zhu

Powered by blists - more mailing lists