Minimal CPU for i586 Tumbleweed?
Hi all, what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)? I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs). So before I'm filing bug reports, the question is if this is still supposed to work. Best regards, seife -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman
Hi Am 02.05.22 um 09:53 schrieb Stefan Seyfried:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
You'll definitely need a CPU with SSE2 support. [1] That rules out most of what can be considered 586. Pentium-M looks like it's compatible, so the problem might be elsewhere. Best regards Thomas [1] https://en.wikipedia.org/wiki/SSE2
Best regards,
seife
-- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
On Monday 2022-05-02 10:56, Thomas Zimmermann wrote:
Am 02.05.22 um 09:53 schrieb Stefan Seyfried:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
You'll definitely need a CPU with SSE2 support.
That's just bullshit. I have here a Tumbleweed 20220308 on an Athlon XP (SSE1 only). Granted, Mesa does not run because of SSE2 shenanigans, but it can do basic Xorg with nouveau (how else could I investigate that Mesa didn't work...).
Am 02.05.22 um 11:05 schrieb Jan Engelhardt:
That's just bullshit.
That was highly unnecessary.
I have here a Tumbleweed 20220308 on an Athlon XP (SSE1 only).
Granted, Mesa does not run because of SSE2 shenanigans, but it can do basic Xorg with nouveau (how else could I investigate that Mesa didn't work...). Thomas tried to help. And that your Tumbleweed runs on Athlon XP does *not* help Stefan. So you just add an insult, but nothing else.
Greetings, Stephan
On Monday 2022-05-02 11:17, Stephan Kulow wrote:
what is the minimal suitable CPU type for openSUSE Tumbleweed
I have here a Tumbleweed 20220308 on an Athlon XP (SSE1 only).
Granted, Mesa does not run because of SSE2 shenanigans, but it can do basic Xorg with nouveau (how else could I investigate that Mesa didn't work...).
Thomas tried to help. And that your Tumbleweed runs on Athlon XP does *not* help Stefan. So you just add an insult, but nothing else.
I absolutely disagree with your views. 1. The original question asked for "minimal". 2. We can surely agree that the result of doing a Tumbleweed text installation with the "Server-/Text console" pattern (however it is called) is a reasonable interpretation for "minimal". 3. Thomas made a blunt statement that is factually untrue with respect to (1) and (2). That's like going into a car dealership, posing as an associate and telling interested parties "yeah it doesn't have that". That's either a great way of upselling them something else, but just as easily can make them leave because they did not find what they were hoping for. 4. I specified information past (2), namely about Xorg and Mesa which extends the textonly install. Naming Athlon also established temporal capabilities; Dothan M is somewhat newer than AXP, and if AXP can do it, the user expectation is/should be that Dothan can do it, too. But that you did not want to see those three data points from (4) and instead focused on the one expletive in the mail is really telling..
after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
Extrapolating "it works for me" should lead to the conclusion that requiring rodata=off is either a relatively new issue, or one that does not manifest on AXP. I did not comment on hangs yet, but here goes: Hangs were not observed yb me; rather, things crashed as they should (SIGILL due to SSE2, naturally).
Am 02.05.22 um 12:18 schrieb Jan Engelhardt:
Thomas tried to help. And that your Tumbleweed runs on Athlon XP does *not* help Stefan. So you just add an insult, but nothing else. I absolutely disagree with your views. And I encourage you to express your disagreement in the most direct way you feel. I strongly disagree with your use of words though. And if a mail of yours start with calling names, I won't give the rest of it a fair chance - and I'm sure I'm not alone with that reading practise.
Greetings, Stephan
Hi Am 02.05.22 um 11:05 schrieb Jan Engelhardt:
On Monday 2022-05-02 10:56, Thomas Zimmermann wrote:
Am 02.05.22 um 09:53 schrieb Stefan Seyfried:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
You'll definitely need a CPU with SSE2 support.
That's just bullshit.
Thank you for this nice and welcoming response.
I have here a Tumbleweed 20220308 on an Athlon XP (SSE1 only).
Granted, Mesa does not run because of SSE2 shenanigans, but it can do basic Xorg with nouveau (how else could I investigate that Mesa didn't work...).
Yes, exactly. You can run a subset of the packages. Some of the graphics libraries need SSE2 and as a result, the modern desktops tend to not work or they are unreliable. I had success with the more basic ones, but Gnome or KDE is flaky. Best regards Thomas -- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
Hi Am 02.05.22 um 09:53 schrieb Stefan Seyfried:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
Forgot to mention: you'll need every MiB of RAM you can get. I had an Athlon XP with 2 GiB of memory, which ran sluggish. Upgrading to 3 GiB at least made made it OK-ish. It was a notable improvement especially with graphics environments. Best regards Thomas
Best regards,
seife
-- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there). But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot. Do you experience problems only when entering graphics mode and thus desktop features or is it already the console mode not working? Richard.
Best regards,
seife
-- Richard Biener <rguenther@suse.de> SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany; GF: Ivo Totev; HRB 36809 (AG Nuernberg)
On Mon, May 02, 2022 at 11:24:05AM +0200, Richard Biener wrote:
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there).
But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot.
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-) https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868 [...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) [...] I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ... Thanks, Stefan Public Key available ------------------------------------------------------ Stefan Dirsch (Res. & Dev.) SUSE Software Solutions Germany GmbH Tel: 0911-740 53 0 Maxfeldstraße 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ---------------------------------------------------------------- (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer ----------------------------------------------------------------
Hi Stefan Am 02.05.22 um 11:54 schrieb Stefan Dirsch:
On Mon, May 02, 2022 at 11:24:05AM +0200, Richard Biener wrote:
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there).
But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot.
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-)
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868
[...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs
Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) [...]
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ...
Maybe you're refering to librsvg? [1] Since it has been rewritten, it depends on SSE2 (by some compiler flag?). IIRC it's been the lib that killed quite a few desktop environments on my i686. My comment in [1] indicates that there are a few more components with SSE2. Best regards Thomas [1] https://bugzilla.opensuse.org/show_bug.cgi?id=1077870
Thanks, Stefan
Public Key available ------------------------------------------------------ Stefan Dirsch (Res. & Dev.) SUSE Software Solutions Germany GmbH Tel: 0911-740 53 0 Maxfeldstraße 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ---------------------------------------------------------------- (HRB 36809, AG Nürnberg) Geschäftsführer: Felix Imendörffer ----------------------------------------------------------------
-- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
On Mon, May 02, 2022 at 12:13:57PM +0200, Thomas Zimmermann wrote:
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ...
Maybe you're refering to librsvg? [1] Since it has been rewritten, it depends on SSE2 (by some compiler flag?). IIRC it's been the lib that killed quite a few desktop environments on my i686. My comment in [1] indicates that there are a few more components with SSE2.
No, but seems other components are affected as well. BTW, I patched Mesa build so systems without SSE2 are still working, since only Intel driver was affected by this change. And Intel GPUs pre-dated to year 2000 are no longer supported anyway (DRI1 drivers have been dropped long time ago). ;-) CU, Stefan Public Key available ------------------------------------------------------ Stefan Dirsch (Res. & Dev.) SUSE Software Solutions Germany GmbH Tel: 0911-740 53 0 Maxfeldstraße 5 FAX: 0911-740 53 479 D-90409 Nürnberg http://www.suse.de Germany ---------------------------------------------------------------- (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev ----------------------------------------------------------------
On Mon, 2 May 2022, Stefan Dirsch wrote:
On Mon, May 02, 2022 at 11:24:05AM +0200, Richard Biener wrote:
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there).
But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot.
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-)
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868
[...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs
Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) [...]
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ...
OK, I suppose on https://get.opensuse.org/tumbleweed/ the System Requirements for the 32bit version could be clarified. IIRC some of the codecs in Firefox even assume AVX and 32bit support is spotty so eventually advertising the 32bit version less by targeting "vintage" systems only would help. Richard. -- Richard Biener <rguenther@suse.de> SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany; GF: Ivo Totev; HRB 36809 (AG Nuernberg)
Hi Am 02.05.22 um 12:16 schrieb Richard Biener:
On Mon, 2 May 2022, Stefan Dirsch wrote:
On Mon, May 02, 2022 at 11:24:05AM +0200, Richard Biener wrote:
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there).
But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot.
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-)
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868
[...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs
Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) [...]
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ...
OK, I suppose on https://get.opensuse.org/tumbleweed/ the System Requirements for the 32bit version could be clarified. IIRC some of the codecs in Firefox even assume AVX and 32bit support is spotty so eventually advertising the 32bit version less by targeting "vintage" systems only would help.
Having seen this discussion about 32-bit support several times, I'm wondering if we should remove all packages with SSE2 from the official repos. Most of upstream seems to have given up on it anyway. For the affected components, the installer could still offer a link to optional repositories on OBS; with a clear message about the additional HW requirements. Best regards Thomas
Richard.
-- Thomas Zimmermann Graphics Driver Developer SUSE Software Solutions Germany GmbH Maxfeldstr. 5, 90409 Nürnberg, Germany (HRB 36809, AG Nürnberg) Geschäftsführer: Ivo Totev
On Mon, 2 May 2022, Thomas Zimmermann wrote:
Hi
Am 02.05.22 um 12:16 schrieb Richard Biener:
On Mon, 2 May 2022, Stefan Dirsch wrote:
On Mon, May 02, 2022 at 11:24:05AM +0200, Richard Biener wrote:
On Mon, 2 May 2022, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
I'm not aware of any concious limit set apart from some CPU features supported only by some Pentiums and up for atomics. The concious limits probably come from the kernel and kernel folks could clarify there (but you should be fine there).
But of course nowadays everybody assumes 64bit and SSE2 as Thomas said so things tend to bitrot.
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-)
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868
[...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs
Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) [...]
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ...
OK, I suppose on https://get.opensuse.org/tumbleweed/ the System Requirements for the 32bit version could be clarified. IIRC some of the codecs in Firefox even assume AVX and 32bit support is spotty so eventually advertising the 32bit version less by targeting "vintage" systems only would help.
Having seen this discussion about 32-bit support several times, I'm wondering if we should remove all packages with SSE2 from the official repos. Most of upstream seems to have given up on it anyway.
For the affected components, the installer could still offer a link to optional repositories on OBS; with a clear message about the additional HW requirements.
The problem is of course identifying packages that do not support systems without CPU feature X (in this case X == SSE2). Presence of SSE2 instruction is not a good enough indication (the usage could be cpuid gated). Richard. -- Richard Biener <rguenther@suse.de> SUSE Software Solutions Germany GmbH, Maxfeldstrasse 5, 90409 Nuernberg, Germany; GF: Ivo Totev; HRB 36809 (AG Nuernberg)
On Monday 2022-05-02 11:54, Stefan Dirsch wrote:
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-) [...] I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember.
The rust compiler itself can produce sse2-avoiding code if you invoke it with the right options (-Ctarget-cpu=...). It's just that we need to make it the *default*, and it's unclear where in the package the default value for -Ctarget-cpu is located so that it can be modified appropriately. https://bugzilla.opensuse.org/show_bug.cgi?id=1162283
At least with regards to Mesa the decision to require SSE2 has been an decision by intention. ;-)
https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/9868
[...] What does this mean for users? On Linux raises the default minimum processor spec to SSE2 supporting CPUs
Intel requirements raise from P5 (1993) to Netburst (2000) AMD requirements raise from Athlon(1999/2000) to Athlon 64 (2003) Via requirements raise from C3(2001) to C7 (2005) I think the choice for SSE2 is not a very great one. From performance
On Mon, 2 May 2022 11:54:23 +0200 Stefan Dirsch wrote: point of view Athlon XP CPUs are still sufficient for many tasks. Especially for compiled code I think it makes not much sense to require SSE2. Hand crafted assembler code would be a different story.
[...]
I think there was another component/tool, which requires SSE2 now? Was it rust? I can't remember. I think it was something, which has been invented long after SSE2 was avaialble ... a very early problem was the openssl FIPS check - which can be avoided. https://bugzilla.opensuse.org/show_bug.cgi?id=1032165
other pieces are Chromium and Qt Webengine: https://bugzilla.opensuse.org/show_bug.cgi?id=1183493#c1 Apart from these limitations, current Tumbleweed still works sufficiently on an Athlon XP with a Matrox video card using the kernel parameter: iomem=relaxed Kind regards, Dieter
Hi Again, (Replying directly to myself, to not single one of the helpful comments out :-) Executive summary: already fixed in Kernel:HEAD After I was finally able to boot with "rodata=off acpi=off" (otherwise the system would hang up relatively hard, during transition from initrd to root from disk) I did investigate the "acpi=off" issue but did not find the culprit, then just for testing installed the latest 5.18-rc5 from Kernel:HEAD, which magically fixed everything (no rodata=off or acpi=off needed anymore). I now also found that the Pentium M "Dothan" (Model 13 Stepping 6) is actually capable of running the PAE kernel. It also features sse2 already. Yes, I'm not expecting 4k youtube playback in Firefox (this will even fail because of the slow 54mbit WiFi module :-) The machine is intended to be used in the garage to do some programming and debugging on motor control units and maybe it will even ride a few kilometers strapped to one of my motorbikes. The toughbook is especially suited for this job because 1) it is tough ;-) 2) it has a real serial port 3) it has a display which is readable in plain sunlight (then via its passive-reflective mode) 4) it is really tough. I just killed a thinkpad display, just by dropping it from maybe 80cm (motorcycle seat) onto the concrete floor. They are not building them anymore like they used to... ;-) Additionally, I would totally understand if we make a conscious decision some time in the future to drop off support for > 15 years old hardware (even if I have quite some of that lying around ;-)), I was just not sure if this was already the case here and if we had forgotten to communicate this more clearly. I'm happy that this is not the case, though. Apparently, I just happened to hit a bad time window when updating when my machine was temporarily broken :-) Thanks for all the hints, have a lot of fun... seife On 02.05.22 09:53, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)? -- Stefan Seyfried
"For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman
Hey, acpi=off works here as well, even without rodata=off. Even the YAST-Gui installer works. Tumbleweed is amazing! Bernd Am 02.05.22 um 12:48 schrieb Stefan Seyfried:
Hi Again,
(Replying directly to myself, to not single one of the helpful comments out :-)
Executive summary: already fixed in Kernel:HEAD
After I was finally able to boot with "rodata=off acpi=off" (otherwise the system would hang up relatively hard, during transition from initrd to root from disk) I did investigate the "acpi=off" issue but did not find the culprit, then just for testing installed the latest 5.18-rc5 from Kernel:HEAD, which magically fixed everything (no rodata=off or acpi=off needed anymore).
I now also found that the Pentium M "Dothan" (Model 13 Stepping 6) is actually capable of running the PAE kernel. It also features sse2 already.
Yes, I'm not expecting 4k youtube playback in Firefox (this will even fail because of the slow 54mbit WiFi module :-)
The machine is intended to be used in the garage to do some programming and debugging on motor control units and maybe it will even ride a few kilometers strapped to one of my motorbikes. The toughbook is especially suited for this job because 1) it is tough ;-) 2) it has a real serial port 3) it has a display which is readable in plain sunlight (then via its passive-reflective mode) 4) it is really tough. I just killed a thinkpad display, just by dropping it from maybe 80cm (motorcycle seat) onto the concrete floor. They are not building them anymore like they used to... ;-)
Additionally, I would totally understand if we make a conscious decision some time in the future to drop off support for > 15 years old hardware (even if I have quite some of that lying around ;-)), I was just not sure if this was already the case here and if we had forgotten to communicate this more clearly. I'm happy that this is not the case, though.
Apparently, I just happened to hit a bad time window when updating when my machine was temporarily broken :-)
Thanks for all the hints, have a lot of fun...
seife
On 02.05.22 09:53, Stefan Seyfried wrote:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
Hi Bernd, On 02.05.22 13:24, Bernd Ritter wrote:
Hey,
acpi=off works here as well, even without rodata=off. Even the YAST-Gui installer works. Tumbleweed is amazing!
After finishing the installation, try adding the Kernel:HEAD/standard repository and installing kernel-default from there. With the new kernel, everything seems to work fine without rodata= and acpi=off. Have fun :-) seife -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman
Hey Stefan, I have the same ambition here. I want to run openSUSE Tumbleweed on a IBM Thinkpad T42 with a Pentium-M and 1 GB RAM (all 32bit of course, hence Tumbleweed). The boot seems as well to be stuck at "A" if I boot without boot options. With "rodata=off" it gets to point "B". In the end I want to run console only, but with modern tools. I would be happy to try stuff, if anybody has an idea. https://paste.opensuse.org/0ccd01ab Cheers, Bernd Am 02.05.22 um 09:53 schrieb Stefan Seyfried:
Hi all,
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
So before I'm filing bug reports, the question is if this is still supposed to work.
Best regards,
seife
On Mon, May 02, 2022 at 11:14:57AM +0000, Bernd Ritter wrote:
Hey Stefan,
I have the same ambition here. I want to run openSUSE Tumbleweed on a IBM Thinkpad T42 with a Pentium-M and 1 GB RAM (all 32bit of course, hence Tumbleweed).
The boot seems as well to be stuck at "A" if I boot without boot options. With "rodata=off" it gets to point "B".
In the end I want to run console only, but with modern tools. I would be happy to try stuff, if anybody has an idea.
Since zram was enabled you are already running the installer. You can try adding some debug options like linuxrc.log=/dev/console linemode=1 linuxrc.debug=1 See also https://en.opensuse.org/SDB:Linuxrc HTH Michal
On Monday 2022-05-02 09:53, Stefan Seyfried wrote:
what is the minimal suitable CPU type for openSUSE Tumbleweed (I know that i586 is deprecated / less supported, but I guess there is still a "minimal viable target system" for it)?
I updated a Pentium-M (Dothan, probably) based Toughbook CF-51 and after finding that I need "rodata=off" kernel parameter for the kernel to even boot, it fails miserably later in userspace apparently (it just hangs).
I have since updated the Athlon machine to TW du jour (so from 20220308 -> 20220428), no rodata=off was needed for 20220428 on the Athlon. You mentioned it's fixed in Kernel:HEAD, but I do not immediately find in the suse-kernel git log anything about rodata in 2022. I have a hang on poweroff now, oh well.
Hi Jan, On 02.05.22 14:14, Jan Engelhardt wrote:
I have since updated the Athlon machine to TW du jour (so from 20220308 -> 20220428), no rodata=off was needed for 20220428 on the Athlon. You mentioned it's fixed in Kernel:HEAD, but I do not immediately find in the suse-kernel git log anything about rodata in 2022.
Maybe it's specific to the machine ¯\_(ツ)_/¯ I just did find the parameter by going through the kernel source and checking the boot sequence and finding that the next thing after the "Freeing initmem..." was the rodata stuff :-) Adding in need for "acpi=off" this might really be something relatively cpu specific. I'd guess if 5.17 would not boot on any 32bit x86 machine, someone would have noticed earlier ;-)
I have a hang on poweroff now, oh well.
Next thing to debug ;-) -- Stefan Seyfried "For a successful technology, reality must take precedence over public relations, for nature cannot be fooled." -- Richard Feynman
participants (9)
-
Bernd Ritter -
dieter -
Jan Engelhardt -
Michal Suchánek -
Richard Biener -
Stefan Dirsch -
Stefan Seyfried -
Stephan Kulow -
Thomas Zimmermann