Hello CentOS folks - this is a first email to propose a RISC-V SIG within CentOS, specifically to support a CentOS Stream port for RISC-V. The idea is to show that CentOS Stream can run effectively on many existing RISC-V platforms, as these have already been demonstrated through Fedora. I'm very interested in feedback on this, as I'd like to promote to the CentOS board as soon as possible.
thanks
Jeffrey "Jefro" Osier-Mixon | josiermi@redhat.com Distinguished Community Architect, Red Hat Office of the CTO Automotive, RISC-V, Edge & IoT Communities
On Thu, Feb 6, 2025 at 2:10 PM Jeffrey Osier-Mixon jefro@redhat.com wrote:
Hello CentOS folks - this is a first email to propose a RISC-V SIG within CentOS, specifically to support a CentOS Stream port for RISC-V. The idea is to show that CentOS Stream can run effectively on many existing RISC-V platforms, as these have already been demonstrated through Fedora. I'm very interested in feedback on this, as I'd like to promote to the CentOS board as soon as possible.
This sounds like a fantastic SIG to create, particularly if it can build and integrate with the work done in Fedora directly. A couple of questions.
1. The subject says "Datacenter SIG" but there's a notable lack of datacenter RISC-V hardware... anywhere. Is it safe to assume the SIG would build to that eventually but start with some of the more developer oriented hardware available today?
2. Are there expectations for hosting and integrating new hardware into CBS, or is there some other plan in mind?
3. Would it make sense to use an ELN build of Fedora RISC-V as the seed buildroot, or perhaps do the work against the existing Fedora secondary arch hardware entirely?
josh
This sounds like a fantastic SIG to create, particularly if it can build and integrate with the work done in Fedora directly. A couple of questions.
- The subject says "Datacenter SIG" but there's a notable lack of
datacenter RISC-V hardware... anywhere. Is it safe to assume the SIG would build to that eventually but start with some of the more developer oriented hardware available today?
Thanks, Josh. One of the purposes is to ensure that CentOS is ready when hardware does become available, as we expect the need from the community to escalate quickly. Right now there are a few performant boards on the market with many more announced and expected to be GA this year.
- Are there expectations for hosting and integrating new hardware
into CBS, or is there some other plan in mind?
- Would it make sense to use an ELN build of Fedora RISC-V as the
seed buildroot, or perhaps do the work against the existing Fedora secondary arch hardware entirely?
These are excellent questions that I would propose as first topics for the SIG to determine.
On Thu, Feb 6, 2025 at 2:09 PM Jeffrey Osier-Mixon jefro@redhat.com wrote:
Hello CentOS folks - this is a first email to propose a RISC-V SIG within CentOS, specifically to support a CentOS Stream port for RISC-V. The idea is to show that CentOS Stream can run effectively on many existing RISC-V platforms, as these have already been demonstrated through Fedora. I'm very interested in feedback on this, as I'd like to promote to the CentOS board as soon as possible.
I don't think this SIG makes sense just yet, since we should be focusing our work at the Fedora level first. Until the integration and initial release of Fedora RISC-V occurs, I'm not sure we want to be in the business of developing a port in CentOS given the lack of infrastructure and contributors to support it.
I don't think this SIG makes sense just yet, since we should be focusing our work at the Fedora level first. Until the integration and initial release of Fedora RISC-V occurs, I'm not sure we want to be in the business of developing a port in CentOS given the lack of infrastructure and contributors to support it.
Thanks, Neal - appreciate your opinion as always. I believe RISC-V will be relevant sooner rather than later. For what it's worth, there has been a ton of work in Fedora thanks to teams both inside and outside Red Hat, and many images available for dozens of platforms - it is only a matter of time before RISC-V is an officially supported architecture, and I feel it makes sense to have a CentOS destination as well. Community is always a chicken/egg situation, but it costs very little to support a SIG and the benefits could be substantial.
best,
On Thu, Feb 6, 2025 at 4:31 PM Jeffrey Osier-Mixon jefro@redhat.com wrote:
I don't think this SIG makes sense just yet, since we should be focusing our work at the Fedora level first. Until the integration and initial release of Fedora RISC-V occurs, I'm not sure we want to be in the business of developing a port in CentOS given the lack of infrastructure and contributors to support it.
Thanks, Neal - appreciate your opinion as always. I believe RISC-V will be relevant sooner rather than later. For what it's worth, there has been a ton of work in Fedora thanks to teams both inside and outside Red Hat, and many images available for dozens of platforms - it is only a matter of time before RISC-V is an officially supported architecture, and I feel it makes sense to have a CentOS destination as well. Community is always a chicken/egg situation, but it costs very little to support a SIG and the benefits could be substantial.
In this case, it costs quite a lot to support this kind of SIG. Most SIGs aren't talking about essentially maintaining a fork of a distribution for a new platform, and the CentOS community is not set up well to handle that. We have plenty of historical evidence to show we *can't* do that (see the failed AltArch and CentOS for Raspberry Pi efforts).
Not to mention, Fedora doesn't even yet have RISC-V as a supported architecture due to lack of hardware, developer support, and commitments from sponsors. Not even Red Hat is seriously doing work in Fedora RISC-V officially as far as I can see. I've met some of the RISC-V vendors who mention to me Red Hat making and maintaining a Fedora RISC-V build, but the people they mention working on it are not participating in Fedora at all.
There are serious misalignments right now that have to be resolved long before we talk about a CentOS RISC-V effort of any kind.
Just circling back - I want to thank Neil and Josh for comments. The most expedient thing is to surface this RISC-V support within the existing ISA SIG, rather than starting a new SIG. I met with the ISA SIG in their last meeting and we discussed it at length.
I also wanted to add - there is indeed serious work on RISC-V happening within Red Hat, which will be more obvious as we go forward. There is finally some hardware available with a BMC that we can use within a lights-out data center, and there is already work ongoing in the Fedora community to host boards for that purpose, with the goal to set up RISC-V as a fully supported architecture in Fedora within the next 9-10 months. Fu We (wefu) has been doing this porting work for several years now, as has davidlt in the Fedora RISC-V SIG.
So the work is happening, it is just slow and not obvious yet. Hardware has always been the blocking factor. That is clearly changing for the better this year.
Jeffrey "Jefro" Osier-Mixon | josiermi@redhat.com Distinguished Community Architect, Red Hat Office of the CTO Automotive, RISC-V, Edge & IoT Communities
On Thu, Feb 6, 2025 at 4:38 PM Neal Gompa ngompa13@gmail.com wrote:
On Thu, Feb 6, 2025 at 4:31 PM Jeffrey Osier-Mixon jefro@redhat.com wrote:
I don't think this SIG makes sense just yet, since we should be focusing our work at the Fedora level first. Until the integration and initial release of Fedora RISC-V occurs, I'm not sure we want to be in the business of developing a port in CentOS given the lack of infrastructure and contributors to support it.
Thanks, Neal - appreciate your opinion as always. I believe RISC-V will
be relevant sooner rather than later. For what it's worth, there has been a ton of work in Fedora thanks to teams both inside and outside Red Hat, and many images available for dozens of platforms - it is only a matter of time before RISC-V is an officially supported architecture, and I feel it makes sense to have a CentOS destination as well. Community is always a chicken/egg situation, but it costs very little to support a SIG and the benefits could be substantial.
In this case, it costs quite a lot to support this kind of SIG. Most SIGs aren't talking about essentially maintaining a fork of a distribution for a new platform, and the CentOS community is not set up well to handle that. We have plenty of historical evidence to show we *can't* do that (see the failed AltArch and CentOS for Raspberry Pi efforts).
Not to mention, Fedora doesn't even yet have RISC-V as a supported architecture due to lack of hardware, developer support, and commitments from sponsors. Not even Red Hat is seriously doing work in Fedora RISC-V officially as far as I can see. I've met some of the RISC-V vendors who mention to me Red Hat making and maintaining a Fedora RISC-V build, but the people they mention working on it are not participating in Fedora at all.
There are serious misalignments right now that have to be resolved long before we talk about a CentOS RISC-V effort of any kind.
-- 真実はいつも一つ！/ Always, there's only one truth! _______________________________________________ devel mailing list -- devel@lists.centos.org To unsubscribe send an email to devel-leave@lists.centos.org
On 26/02/2025 01:27, Jeffrey Osier-Mixon wrote:
Just circling back - I want to thank Neil and Josh for comments. The most expedient thing is to surface this RISC-V support within the existing ISA SIG, rather than starting a new SIG. I met with the ISA SIG in their last meeting and we discussed it at length.
I also wanted to add - there is indeed serious work on RISC-V happening within Red Hat, which will be more obvious as we go forward. There is finally some hardware available with a BMC that we can use within a lights-out data center, and there is already work ongoing in the Fedora community to host boards for that purpose, with the goal to set up RISC- V as a fully supported architecture in Fedora within the next 9-10 months. Fu We (wefu) has been doing this porting work for several years now, as has davidlt in the Fedora RISC-V SIG.
So the work is happening, it is just slow and not obvious yet. Hardware has always been the blocking factor. That is clearly changing for the better this year.
Hi Jeffrey,
Can you elaborate on this ? Myself currently not aware of any hardware coming to either CentOS Stream build infra (internal) nor for SIGs (so behind cbs.centos.org) WRT RISC-V .. :)
Or is your idea/proposal just to start a SIG about discussing what would be eventually needed (either for Stream and in parallel for SIGs) and come with a plan ?
Thanks Fabian. I will get the story about incoming hardware and be in touch, but I'm certain there is hardware inbound for CBS soon. Work is being done right now with individual boards on people's desks.
Jeffrey "Jefro" Osier-Mixon | josiermi@redhat.com Distinguished Community Architect, Red Hat Office of the CTO Automotive, RISC-V, Edge & IoT Communities
On Thu, Feb 27, 2025 at 2:46 AM Fabian Arrotin arrfab@centos.org wrote:
On 26/02/2025 01:27, Jeffrey Osier-Mixon wrote:
Just circling back - I want to thank Neil and Josh for comments. The most expedient thing is to surface this RISC-V support within the existing ISA SIG, rather than starting a new SIG. I met with the ISA SIG in their last meeting and we discussed it at length.
I also wanted to add - there is indeed serious work on RISC-V happening within Red Hat, which will be more obvious as we go forward. There is finally some hardware available with a BMC that we can use within a lights-out data center, and there is already work ongoing in the Fedora community to host boards for that purpose, with the goal to set up RISC- V as a fully supported architecture in Fedora within the next 9-10 months. Fu We (wefu) has been doing this porting work for several years now, as has davidlt in the Fedora RISC-V SIG.
So the work is happening, it is just slow and not obvious yet. Hardware has always been the blocking factor. That is clearly changing for the better this year.
Hi Jeffrey,
Can you elaborate on this ? Myself currently not aware of any hardware coming to either CentOS Stream build infra (internal) nor for SIGs (so behind cbs.centos.org) WRT RISC-V .. :)
Or is your idea/proposal just to start a SIG about discussing what would be eventually needed (either for Stream and in parallel for SIGs) and come with a plan ?
-- Fabian Arrotin The CentOS Project | https://www.centos.org gpg key: 17F3B7A1 | @arrfab[@fosstodon.org]
devel mailing list -- devel@lists.centos.org To unsubscribe send an email to devel-leave@lists.centos.org
-
Fabian Arrotin -
Jeffrey Osier-Mixon -
Josh Boyer -
Neal Gompa