x86_64 hang on hpet

[oneoverzero]

I am new to Fedora but have decided to take the leap for a project I am working on. I am experiencing a system hang during initial boot. After installing boot media and beginning boot via "Install or Upgrade Existing System" I receive about three pages of preliminary boot/install text and a hard hang at

Time: hpet clocksource has been installed

I waited up to 20 minutes for progress but with no DVD or HDD action the install/boot is definitely hung.

Details:
OS: Fedora 7 x86_64 (both Live and DVD ISOs attempted and fail at same place, both SHA1 hash verified OK via SHA1SUM)

System: Intel S5000PSLSASR MB, 2 quad core x64 Xeon E5345 cpus, 16GB 667 ECC RAM, 300GB 15000rpm SAS, 2 500GB SATA, LITEON DVD, onboard video and NICs (chipsets unknown right now)

Things I have tried: disable legacy USB in BIOS, disable SAS in BIOS, nousb command line boot option, searched the Fedora 7 install guides and of course checked Google.

Google turns up a few complaints on hangs at hpet during initial boot/install but there are no solutions posted. I came across this issue which seems to bear directly on the problem (summary of link: hpet conflicts on x86_64 platforms causing hangs discovered August 2007), but I am unsure if this defect has been fixed in Fedora 7.

I am very willing to peel back the hood and try things, but could use a bit of direction from more experienced users. Any help on this issue would be greatly appreciated.


Thanks,

one

[buddha]

Have you tried disabling HPET support in the BIOS?

My motherboard's BIOS has a setting on its Power Management page to enable or disable HPET.

[buddha]

another thing you could try is adding hpet=disable to the kernel boot line

hit TAB at the first screen to edit the boot line

[oneoverzero]

Many thanks for the help, Buddha.

I checked my BIOS (AMI 1.20.1093) but I do not have any mention of HPET. I also tried the suggested hpet=disable but this had no effect (produced exactly the same output) on boot. I checked Bugzilla and came across this bug (summary of link: hpet-related hangs on non-tickless kernels). The workaround there suggests using nohpet as the boot command line. I used this approach and received a hang at

Time: acpi_pm clocksource has been installed

Per Bugzilla, this defect was patched in rawhide in July but I am unfamiliar with how bugs fixed in the daily builds make ther way into the more official product. I am still looking for a path forward on this. Do I have some sort of core hardware incompatibility? Do I need to go to rawhide builds to address this issue (something I would prefer not to do given my need for stability on this system) or are there other things I can try to workaround this issue?

Thanks

one

[oneoverzero]

After further research on hangs associated with acpi_pm, this post surfaced. I tried setting nohpet and acpi=off at command line and am now installing. I have no real idea what impact these changes will have but seeing an install screen is a welcome sight.

If I have further issues, I will post here, but otherwise this is a fix.

[buddha]

Glad to see you found a way past the hangs. I read over the bugzilla bug you mentioned and it seems that the x86_64 tickless kernel still has some issues with HPET.

If you need a very stable system, Fedora might not be the right choice for you. It changes very fast and since it tracks the front end of development, you do run into issues sometimes.

Your trouble with HPET seems a bit odd though since I thought HPET was an Intel thing. I'm not familiar with what looks like a server motherboard you have. I wonder if a BIOS firmware update might have better support for HPET and/or ACPI.

[oneoverzero]

Originally Posted by buddha

If you need a very stable system, Fedora might not be the right choice for you. It changes very fast and since it tracks the front end of development, you do run into issues sometimes.

Stability is important for me, but so are features. Of course, I want the best of both worlds, which is one reason why I went with Fedora. I can live with the odd defect here or there and treat it as a learning experience as long as there is a path forward that leads to success.

Originally Posted by buddha

Your trouble with HPET seems a bit odd though since I thought HPET was an Intel thing. I'm not familiar with what looks like a server motherboard you have. I wonder if a BIOS firmware update might have better support for HPET and/or ACPI.

AFAIK, the HPET issue is Intel-related and the problems associated with it are in some way associated to IA64 (Itanium?) and x86_64 HPET conflicts. Since the S5000PSL board is an Intel board, this issue is not as unexpected as it might first seem. Regarding BIOS updates, I searched for new BIOS and even came across a post associated with newer versions of the kernel having issues with the S5000 server board. FC6 is apparently fine on this board, but F7 is problematic, but Intel's website is so crappy I couldn't find the older versions of BIOS they recommended when trying F7 on that particular mainboard. Oddly, I have another project/server I am working on that has a Intel Core 2 Duo (x64 architecture) and AMI BIOS that does have HPET control in the BIOS. *shrug*

It's just my luck to make my first Linux-based project as rough as possible wrt installation. As if these issues were not enough, I missed the checkbox during install that allowed me to supply additional boot options. This resulted in me not including nohpet and acpi=off during initial boot config and repeated "blank screen" hangs after I completed what I thought was a successful installation. I am sure there is an easier way to supply these settings post-mortum (as it were), but even booting from the Live CD I wasn't able to get to the boot config file to make these changes. A clean re-install and a careful reading of the screens fixed the problem. I am now up and running and thus far have no known issues.

Thanks again for your help.