Red Hat Bugzilla – Bug 155638
rescue cd kernel seems to ignore acpi cutoff date
Last modified: 2015-01-04 17:19:06 EST
Description of problem:
The kernel included in the rescue cd seems to ignore the so-called "cutoff date"
for acpi bioses (2001, I believe). Having such a machine (an older, crappy HP) I
can verify that this can cause problems when using anaconda from that cd to do a
network install (it freezes when trying to access the IDE HD certain ways).
Futhermore it works fine without the "noacpi" option passed.
Also, it'd be nice if some more kernel options were doccumented in the f1 menus,
but, that's more of an enhancement.
Steps to Reproduce:
1. Run "linux" from rescue cd and start anaconda without "noacpi" passed to
kernel at boot.
2. Watch anaconda die.
3. Do it with "noacpi" and watch it work.
Use ACPI "Cutoff date" and disable ACPI.
PS: I appologize in advance if I have submit this to the wrong component.
The cutoff date is for machines older than 2001. If yours is made in 2001 and is
broken, it needs specific blacklisting. Please send the output of dmidecode.
Hmm, no, this isn't the issue. It's bios date is 2000.
In fact, the kernel that's installed (which is from the devel branch) works
perfectly, I think I may have run into this exact problem with FC3's rescue disc
The only problem I'm having is on the FC4-test2 rescue cd. Normal rescue mode
works for some odd reason, but I see nothing in dmesg about the cutoff date, and
the ACPI stuff seems to load just fine. When using the graphical installer
(which loads X of course) after specifying only "linux" at the boot prompt, I
notice no "ACPI Cutoff date" warning and it will reliably freeze when trying to
either "Find installations" or format.
When I specify "linux noacpi" it works just fine.
I think it's ignoring the cut off date (old kernel version?) and the problem
remains unnoticable until X is running.
theres no way other than random memory being bitflipped that the cut off date
can be ignored.
The noacpi thing is odd, as disabling acpi is done with 'acpi=off'
Jeremy, is 'noacpi' an anaconda thing to stop the modules being loaded ?
D'oh! Nothing to see here, move along.
I can't seem to reproduce it anymore. Must have been a hardware problem combined
with user stupidity and unscrupulous timing. And yes, upon closer inspection it
does seem to obey the cutoff.
I have seen noacpi floating around on it's own before, but that might just be
other people making the same mistake.
Feel free to flog, grapeshot, or keelhaul accordingly.