* Best practices to write CIP kernel details in a SBOM?
@ 2026-11-20 1:29 Dominique Martinet
2026-11-20 12:01 ` Christoph Steiger
0 siblings, 1 reply; 4+ messages in thread
From: Dominique Martinet @ 2026-11-20 1:29 UTC (permalink / raw)
To: cip-dev; +Cc: Akihito Irie, Dominique Martinet
Hi,
I asked about SBOM on IRC a couple of days ago and got told "mailing
list" would be better - I hope I got the right one.. :)
(also note: my org (atmark-techno.com) is not a CIP member as far as I
know, we're not using CIP kernels yet but we do maintain a 5.10 kernel
and I intend to participate at least in reviews/testing starting next
month when upstream 5.10 goes EOL, as I've been doing with upstream 5.10
stable-rc tests. If it's a pre-requisite for discussion please let me
know)
Now for some more useful background and questions...
We're working on improving our SBOM, as it becomes mandatory in various
regulations, and in particular the linux kernel has been difficult to
describe in a useful manner so our users (we ship an OS) can act on the
SBOM.
As long as we stick to upstream kernel stable releases, we can
apparently write PURL/CPE tags as follow:
- pkg:generic/linux-kernel@[version]
- cpe:2.3:o:linux:linux_kernel:[version]:::::::*
and since the linux upstream CVEs are fed to NVD/OSV databases then
scanners will pick up the many CVEs still affecting the releases, so one
can triage them and export the result as VEX as seems to be the current
best practices; but if we don't provide a VEX there doesn't seem to be
any way to describe the kernel version as e.g. v5.10.272-cip79 with all
the fixes covered by CIP automatically covered.
I've found about cip-kernel-sec[1] and kernel-cve-triage[2] and we can
use it to automatically check boxes in a VEX, but would it make sense to
publish this data as well somewhere so that "standard" scanners
(dependency-track/grype/osv-scanner/etc) automatically do it in the
first place?
From what I've seen it's possible to setup a data source for OSV.dev
that they'd pick it up[3], I'm not sure if it'd be useful for CIP as a
project or if it's something that'd gather interest?
[1] https://gitlab.com/cip-project/cip-kernel/cip-kernel-sec
[2] https://gitlab.com/cip-project/cip-kernel/kernel-cve-triage
[3] https://google.github.io/osv.dev/data/
Just to be clear I'm not asking for free lunch here:
- My main question is, if people here publish a SBOM, what do they use
to describe their kernel to others, is there any recommendation
somewhere I didn't see?
(I saw https://github.com/siemens/debsbom from cip-documents, but
there doesn't seem to be anything about the kernel there)
- Are there any plans going forward for kernel? If so is there anything
we can help with?
Thanks!
--
Dominique Martinet | Asmadeus
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: Best practices to write CIP kernel details in a SBOM?
2026-11-20 1:29 Best practices to write CIP kernel details in a SBOM? Dominique Martinet
@ 2026-11-20 12:01 ` Christoph Steiger
2026-11-20 12:56 ` [cip-dev] " Dominique Martinet
0 siblings, 1 reply; 4+ messages in thread
From: Christoph Steiger @ 2026-11-20 12:01 UTC (permalink / raw)
To: Dominique Martinet, cip-dev; +Cc: Akihito Irie, Dominique Martinet, Jan Kiszka
Hi Dominique,
On 11/20/26 2:29 AM, Dominique Martinet wrote:
> Hi,
>
> I asked about SBOM on IRC a couple of days ago and got told "mailing
> list" would be better - I hope I got the right one.. :)
> (also note: my org (atmark-techno.com) is not a CIP member as far as I
> know, we're not using CIP kernels yet but we do maintain a 5.10 kernel
> and I intend to participate at least in reviews/testing starting next
> month when upstream 5.10 goes EOL, as I've been doing with upstream 5.10
> stable-rc tests. If it's a pre-requisite for discussion please let me
> know)
>
No worries, you do not need to be a member to join the discussion ;-)
> Now for some more useful background and questions...
> We're working on improving our SBOM, as it becomes mandatory in various
> regulations, and in particular the linux kernel has been difficult to
> describe in a useful manner so our users (we ship an OS) can act on the
> SBOM.
>
> As long as we stick to upstream kernel stable releases, we can
> apparently write PURL/CPE tags as follow:
> - pkg:generic/linux-kernel@[version]
> - cpe:2.3:o:linux:linux_kernel:[version]:::::::*
>
> and since the linux upstream CVEs are fed to NVD/OSV databases then
> scanners will pick up the many CVEs still affecting the releases, so one
> can triage them and export the result as VEX as seems to be the current
> best practices; but if we don't provide a VEX there doesn't seem to be
> any way to describe the kernel version as e.g. v5.10.272-cip79 with all
> the fixes covered by CIP automatically covered.
>
It is true there currently is no globally understood way to reference
CIP kernels. It would be simple enough to use something similar to the
kernel e.g. pkg:generic/linux-cip-kernel@<version>. The second step
would then be to start integration with all the scanners. I dont think
there are currently any efforts in that direction though.
> I've found about cip-kernel-sec[1] and kernel-cve-triage[2] and we can
> use it to automatically check boxes in a VEX, but would it make sense to
> publish this data as well somewhere so that "standard" scanners
> (dependency-track/grype/osv-scanner/etc) automatically do it in the
> first place?
Is there any information you are missing from the data published in
cip-kernel-sec? It is not integrated with all the scanners, but I think
everything that we need should already be public.
>>From what I've seen it's possible to setup a data source for OSV.dev
> that they'd pick it up[3], I'm not sure if it'd be useful for CIP as a
> project or if it's something that'd gather interest?
>
> [1] https://gitlab.com/cip-project/cip-kernel/cip-kernel-sec
> [2] https://gitlab.com/cip-project/cip-kernel/kernel-cve-triage
> [3] https://google.github.io/osv.dev/data/
>
>
> Just to be clear I'm not asking for free lunch here:
> - My main question is, if people here publish a SBOM, what do they use
> to describe their kernel to others, is there any recommendation
> somewhere I didn't see?
I actually posted a lengthy idea about including kernel SBOMs as part of
standard CIP kernel builds on this mailing list yesterday, but that is
stuck in limbo somewhere.
Since v7.2 the mainline kernel provides scripts to generate kernel SBOMs
[1]. That is as close as we can get to a standard way of describing your
kernel, and it is conveniently already in the correct format for all of
the compliance stuff. My suggestion is to backport the SBOM generation
to all CIP SLTS kernels and generate them as part of a standard build.
The kernel SBOM can also be used for CVE triaging, very similar to what
is done in kernel-cve-triage. Yocto for example is using their kernel
SBOMs for exactly that purpose.
> (I saw https://github.com/siemens/debsbom from cip-documents, but
> there doesn't seem to be anything about the kernel there)
I maintain debsbom, so I can give you some insight on what we are
currently planning to do.
The kernels in debsbom are currently only referenced by the Debian
package that provides them. If we have a kernel SBOM associated with
that kernel, we could reference it or directly include it. That way
everything would be connected and you have a clear description of the
kernel that was used in the distribution.
> - Are there any plans going forward for kernel? If so is there anything
> we can help with?
I hoped to start a discussion about this exact topic with my post
yesterday. @Jan can we do anything to get it unstuck? I can repost too
if it is not recoverable for whatever reason.
>
>
> Thanks!
[1] https://docs.kernel.org/tools/sbom/sbom.html
Christoph
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [cip-dev] Best practices to write CIP kernel details in a SBOM?
2026-11-20 12:01 ` Christoph Steiger
@ 2026-11-20 12:56 ` Dominique Martinet
2026-11-20 13:35 ` Christoph Steiger
0 siblings, 1 reply; 4+ messages in thread
From: Dominique Martinet @ 2026-11-20 12:56 UTC (permalink / raw)
To: christoph.steiger; +Cc: cip-dev, Akihito Irie, Dominique Martinet, Jan Kiszka
Christoph Steiger via cip-dev wrote on Fri, Nov 20, 2026 at 01:01:09PM +0100:
> > and since the linux upstream CVEs are fed to NVD/OSV databases then
> > scanners will pick up the many CVEs still affecting the releases, so one
> > can triage them and export the result as VEX as seems to be the current
> > best practices; but if we don't provide a VEX there doesn't seem to be
> > any way to describe the kernel version as e.g. v5.10.272-cip79 with all
> > the fixes covered by CIP automatically covered.
>
> It is true there currently is no globally understood way to reference CIP
> kernels. It would be simple enough to use something similar to the kernel
> e.g. pkg:generic/linux-cip-kernel@<version>. The second step would then be
> to start integration with all the scanners. I dont think there are currently
> any efforts in that direction though.
Right, just deciding on a pkg format won't do much unless we start using
it for data feeds scanners pick up, the question is how useful that
would be given the state most scanners are in...
Just adding pkg:generic/linux-kernel@<version> will make scanners report
hundreds of problems with no way to automatically filter out what isn't
built in or what isn't used, check by category etc; that needs specific
tooling.
But to be fair that's not a problem with the CIP kernel - the tools just
haven't caught up yet, and I'm not sure how much will be done by whom...
> > I've found about cip-kernel-sec[1] and kernel-cve-triage[2] and we can
> > use it to automatically check boxes in a VEX, but would it make sense to
> > publish this data as well somewhere so that "standard" scanners
> > (dependency-track/grype/osv-scanner/etc) automatically do it in the
> > first place?
>
> Is there any information you are missing from the data published in
> cip-kernel-sec? It is not integrated with all the scanners, but I think
> everything that we need should already be public.
I honestly haven't spent enough time with it yet to say, sorry.
As far as data goes, it's even properly tracking CVEs introduced by
cip-specific backports (e.g. CVE-2025-38656), so it's really quite
good - thank you!
I'm curious about how we could track a vuln introduced by the backport
process itself (e.g. bringing in patch X without patch Y caused a bug
not present in any upstream kernel, thus without assigned CVE by the
kernel CNA), but I don't think there's much we can do about it, and it's
a bit off topic... :)
Since kernel-cve-triage makes use of "affected files" I assume the data
is not meant to be standalone, but to complete kernel's vulns.git that
has this info, is that correct?
But anyway, my problem isn't triaging locally than deciding what we want
to tell our users.
The main problem with local triage is that it only knows about the
vunlerabilities public at the time of the triage, our users need to have
enough information in the SBOM we publish to be able to reproduce the
triage and get useful information regarding new vulnerabilities that
came out after we published our data.
That means that either we need to document in our product manual how to
use CIP's kernel-cve-triage, providing the necessary information to run
it, and have our users run these scripts manually, or we need to work on
some standardized SBOM that will be enough for "any tooling" to provide
useful information (but as said above we're not there yet)
I think the kernel SBOMs you pointed at is a great way in that
direction - since triage needs to know which source files were used for
build and things like that it should be a good input file for it, and
eventually other tools will pick it up (eventually with a bit of help :-))
> > Just to be clear I'm not asking for free lunch here:
> > - My main question is, if people here publish a SBOM, what do they use
> > to describe their kernel to others, is there any recommendation
> > somewhere I didn't see?
>
> I actually posted a lengthy idea about including kernel SBOMs as part of
> standard CIP kernel builds on this mailing list yesterday, but that is stuck
> in limbo somewhere.
Great timing, I'll wait for this to get unstuck!
> Since v7.2 the mainline kernel provides scripts to generate kernel SBOMs
> [1]. That is as close as we can get to a standard way of describing your
> kernel, and it is conveniently already in the correct format for all of the
> compliance stuff. My suggestion is to backport the SBOM generation to all
> CIP SLTS kernels and generate them as part of a standard build.
>
> The kernel SBOM can also be used for CVE triaging, very similar to what is
> done in kernel-cve-triage. Yocto for example is using their kernel SBOMs for
> exactly that purpose.
Thanks for bringing this up, I saw it when it was added, noted to have a
closer look and promptly forgot.
I didn't test it yet, but from a quick look at the documentation it's
definitely a great step forward from what we're shipping (which is
basically useless, scan of the built binary kernel/modules only)
We'll definitely want to use this as well, the standalone version could
probably be used without any backport but it makes sense to pick it up.
> > - Are there any plans going forward for kernel? If so is there anything
> > we can help with?
>
> I hoped to start a discussion about this exact topic with my post yesterday.
> @Jan can we do anything to get it unstuck? I can repost too if it is not
> recoverable for whatever reason.
I'm on the list, I'll check it out when it goes through, thanks again!
--
Dominique Martinet | Asmadeus
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: [cip-dev] Best practices to write CIP kernel details in a SBOM?
2026-11-20 12:56 ` [cip-dev] " Dominique Martinet
@ 2026-11-20 13:35 ` Christoph Steiger
0 siblings, 0 replies; 4+ messages in thread
From: Christoph Steiger @ 2026-11-20 13:35 UTC (permalink / raw)
To: Dominique Martinet; +Cc: cip-dev, Akihito Irie, Dominique Martinet, Jan Kiszka
On 11/20/26 1:56 PM, Dominique Martinet wrote:
> Christoph Steiger via cip-dev wrote on Fri, Nov 20, 2026 at 01:01:09PM +0100:
>>> and since the linux upstream CVEs are fed to NVD/OSV databases then
>>> scanners will pick up the many CVEs still affecting the releases, so one
>>> can triage them and export the result as VEX as seems to be the current
>>> best practices; but if we don't provide a VEX there doesn't seem to be
>>> any way to describe the kernel version as e.g. v5.10.272-cip79 with all
>>> the fixes covered by CIP automatically covered.
>>
>> It is true there currently is no globally understood way to reference CIP
>> kernels. It would be simple enough to use something similar to the kernel
>> e.g. pkg:generic/linux-cip-kernel@<version>. The second step would then be
>> to start integration with all the scanners. I dont think there are currently
>> any efforts in that direction though.
>
> Right, just deciding on a pkg format won't do much unless we start using
> it for data feeds scanners pick up, the question is how useful that
> would be given the state most scanners are in...
> Just adding pkg:generic/linux-kernel@<version> will make scanners report
> hundreds of problems with no way to automatically filter out what isn't
> built in or what isn't used, check by category etc; that needs specific
> tooling.
>
> But to be fair that's not a problem with the CIP kernel - the tools just
> haven't caught up yet, and I'm not sure how much will be done by whom...
>
>>> I've found about cip-kernel-sec[1] and kernel-cve-triage[2] and we can
>>> use it to automatically check boxes in a VEX, but would it make sense to
>>> publish this data as well somewhere so that "standard" scanners
>>> (dependency-track/grype/osv-scanner/etc) automatically do it in the
>>> first place?
>>
>> Is there any information you are missing from the data published in
>> cip-kernel-sec? It is not integrated with all the scanners, but I think
>> everything that we need should already be public.
>
> I honestly haven't spent enough time with it yet to say, sorry.
>
> As far as data goes, it's even properly tracking CVEs introduced by
> cip-specific backports (e.g. CVE-2025-38656), so it's really quite
> good - thank you!
> I'm curious about how we could track a vuln introduced by the backport
> process itself (e.g. bringing in patch X without patch Y caused a bug
> not present in any upstream kernel, thus without assigned CVE by the
> kernel CNA), but I don't think there's much we can do about it, and it's
> a bit off topic... :)
>
> Since kernel-cve-triage makes use of "affected files" I assume the data
> is not meant to be standalone, but to complete kernel's vulns.git that
> has this info, is that correct?
>
Yes it is supplementary, but it should mostly match. kernel-cve-triage
for example, among other things, tries to figure out which .c files are
affected by vulnerable headers as this is needed for some of the analysis.
>
> But anyway, my problem isn't triaging locally than deciding what we want
> to tell our users.
> The main problem with local triage is that it only knows about the
> vunlerabilities public at the time of the triage, our users need to have
> enough information in the SBOM we publish to be able to reproduce the
> triage and get useful information regarding new vulnerabilities that
> came out after we published our data.
>
> That means that either we need to document in our product manual how to
> use CIP's kernel-cve-triage, providing the necessary information to run
> it, and have our users run these scripts manually, or we need to work on
> some standardized SBOM that will be enough for "any tooling" to provide
> useful information (but as said above we're not there yet)
>
> I think the kernel SBOMs you pointed at is a great way in that
> direction - since triage needs to know which source files were used for
> build and things like that it should be a good input file for it, and
> eventually other tools will pick it up (eventually with a bit of help :-))
>
I think doing the triage with SBOMs as part of the product lifecycle is
the way forward. Once you have the SBOM and the vulnerability from
vulns.git it is very easy to check if the kernel is affected (<20 lines
of python with no dependencies). No checkout of the kernel tree, no
build artifacts, just the SBOM which you want to (or have to) store anyways.
CIP could provide the kernel SBOMs as in mainline, and also the
vulnerabilities with affected file information in cip-kernel-sec. That
would allow an easy workflow for users. And we would be somewhat
compatible with mainline, which should allow us to reuse any tooling
created for it with minimal or no further effort.
>>> Just to be clear I'm not asking for free lunch here:
>>> - My main question is, if people here publish a SBOM, what do they use
>>> to describe their kernel to others, is there any recommendation
>>> somewhere I didn't see?
>>
>> I actually posted a lengthy idea about including kernel SBOMs as part of
>> standard CIP kernel builds on this mailing list yesterday, but that is stuck
>> in limbo somewhere.
>
> Great timing, I'll wait for this to get unstuck!
>
>> Since v7.2 the mainline kernel provides scripts to generate kernel SBOMs
>> [1]. That is as close as we can get to a standard way of describing your
>> kernel, and it is conveniently already in the correct format for all of the
>> compliance stuff. My suggestion is to backport the SBOM generation to all
>> CIP SLTS kernels and generate them as part of a standard build.
>>
>> The kernel SBOM can also be used for CVE triaging, very similar to what is
>> done in kernel-cve-triage. Yocto for example is using their kernel SBOMs for
>> exactly that purpose.
>
> Thanks for bringing this up, I saw it when it was added, noted to have a
> closer look and promptly forgot.
> I didn't test it yet, but from a quick look at the documentation it's
> definitely a great step forward from what we're shipping (which is
> basically useless, scan of the built binary kernel/modules only)
> > We'll definitely want to use this as well, the standalone version could
> probably be used without any backport but it makes sense to pick it up.
>
>>> - Are there any plans going forward for kernel? If so is there anything
>>> we can help with?
>>
>> I hoped to start a discussion about this exact topic with my post yesterday.
>> @Jan can we do anything to get it unstuck? I can repost too if it is not
>> recoverable for whatever reason.
>
> I'm on the list, I'll check it out when it goes through, thanks again!
>
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~2026-11-20 13:35 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-20 1:29 Best practices to write CIP kernel details in a SBOM? Dominique Martinet
2026-11-20 12:01 ` Christoph Steiger
2026-11-20 12:56 ` [cip-dev] " Dominique Martinet
2026-11-20 13:35 ` Christoph Steiger
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).