Fedora Core Doesn't Like to Dual Boot?
schwatoo writes "It seems Fedora Core doesn't like to boot alongside Windows 2K or XP. According to a bug first reported in February on Fedora's bugzilla site it has a tendency to chew up partition maps making it impossible to dual boot into Windows. No one seems to know quite what is causing the problem and a lot of people are ending up with unbootable machines."
if this was a Microsoft problem the amount of bitching and conspiracy theories would never end. Lets see how it plays out.
Only the State obtains its revenue by coercion. - Murray Rothbard
Well who would?
Works fine here with Windows Server 2003.
This guy is way out there
I am nearly 100% sure that the Redhat people are going to straighten this out, if it was a windows problem you know who'd straighten it out? The people on the GRUB or LILO team.
1. Partition in an older, safe system. E.g. knoppix.
2. Install grub in an older, safe system. You should have grub installed already, if you have been using Linux on the machine previously. I never install bootloader anymore, I've been using the same one forever. Just edit the grub config to point to new kernel & root system.
3. Grub should be on the beginning of small boot partition. Never on MBR, if you can avoid it. Always create a 80MB or so partition on the start of every disk, even if you don't plan on using it (yet). This also applies to secondary disks. Kernels should always go to these partitions.
Save your wrists today - switch to Dvorak
Copy and paste from here
It turns out that the bug (#115980) is a result of a few subtle but key changes within the 2.6 kernel. A certain functionality with regards to hard disk geometry has been pulled out, as the kernel developers thought it would be better if userspace utilities took care of this instead. The Bugzilla bug is related to CHS geometry problems, which most likely stems from an error within the parted utility, addressing the BIOS incorrectly. It turns out that BIOS updates tend to fix problems for many users that have been bitten by this "bug". On newer machines, this is basically non-reproducible.
All your partition are belong to us!
Mandrake 10.0 Official also suffers from this problem. This is leading many to believe that it is an issue with the 2.6 kernel, rather than a specific distro.
Fedora Core 2 doesn't like to boot at all, never mind just dual boot systems. So far I have tried installing FC2 on two systems (PII laptop and VIA C3 machine) and both of them fail. The laptop insists that there is not enough disk space and then borks out and the C3 machine just reboots in an endless cycle.
For those who follow Bugzilla the numbers you need are 121819 if you have an ASUS motherboard and 120685 if you have a VIA C3 system. The second link for the C3 is much more involved and a number of the posters are deep into the kernel architecture at the moment.
This is not good, I thought that the test releases were supposed to pick things like this up ?
Ed Almos
Budapest, Hungary
The more corrupt the state, the more numerous the laws. - Tacitus, 56-120 A.D.
http://mailman.linux-thinkpad.org/pipermail/linux
Here's a quick executive summary for those who don't want to read the thread:
Linux 2.6 kernels started to report bogus disk geometries thus some unadjusted partitioning tools create bad partition table resulting unbootable Windows.
Knowledge is power. Knowledge shared is power multiplied.
We wondered if this bug would affect us - and went with rolling out FC1 instead - the kernel 2.6.x + Nvidia driver issue (which I gather will be fixed soon), as well as this seemed too scary.
No sig for you.
You got to know someone at M$ is pissed they never thought of it before.
I had a similar booting problem when installing Gentoo. As Gentoo is very hands-on, and has quite a community, it was easy to find the fix.
First, the fault is Microsoft's. (Seriously, did you expect anything else?). The point is that Windows XP is a hog which believes that it is the one and only system on the computer. Therefore, if it is not on hda, it will put its hands on its ears and start singing aloud "La-la-la-la I can't hear you!". I have Linux on my hda, and WXP getting dustier and dustier on hdb. It would not start until I added the following lines in grub.conf:
title=Windows Xp
rootnoverify (hd1,0)
map (hd1,0) (hd0,0)
map (hd0,0) (hd1,0)
chainloader (hd1,0)+1
I'm not aware of how much Fedora lets the user write their grub.conf, but if they have a GUI tool, it might just not be programmed for this. After all, on my office machine, where Windows has been left on hda1, things works well out-of-the-box. Maybe they assumed everybody would use this configuration.
Victims of 9/11: <3000. Traffic in the US: >30,000/y
the fedora people should atleast put up some serious warnings in their install guides.
:)
I think a headline on the biggest news channel on the internet should serve as a good enough warning
As you'll have seen from various other comments; Mandrake 10 has the same problem therefore if you're using M10 without problems then FC2 will work just as well.
----- Documentation is worth it just to be able to answer all your mail with 'RTFM' - Alan Cox.
FC2 is running fine on my IBM T40, but I had to tell the BIOS to show the hidden partition. With it hidden Anaconda wanted to format the disk. Unfortunately, once I un-hid the recovery partition, installed FC2, both OS's ran fine (XP + FC2), but now the BIOS claims the recover partition is trashed. I'm not 100% convince that I can't do a recover since the recover GUI comes up fine, but I'm not running anything from it. The machine is running fine and the only FC2 problem for me is I'll need a custom kernel to get my suspend on cover close back.
I hate to rain on people's parades, but if you're making a system dual-boot with Windows, the conservative/safe thing to do is NOT install a bootloader.
Just use the one that comes with NT/XP. Of course it is limited in features (esp. compared to GRUB) but it works.
It's not a ton of work either:
Write a LILO bootloader to a partition, use 'dd' to copy that to a file (floppy helps), copy the "file" to Windows, and edit boot.ini to point to it.
Sure, it's not automated, but we're talking just a few steps, and then your're 100% confident that the next upgrade of Windows will not choke.
It would be nice if the PC industry could get "all OS vendors" to agree to universal bootloader, and maybe even get it in the BIOS, but the situation is what it is. You've got to be very careful when dual booting, especially with BETA software.
Sounds like the GRUB and kernel people need to work closer together. I don't know about GRUB, but the kernel has some pretty good testsuites so I am surprised this was not caught by the Linux Test Project (LTP). I'm hearing it's actually a 2.6 kernel problem, and since not a lot of people have upgraded to 2.6 we're hearing about it now.
I recently installed Fedora Core 2 on a computer alongside Windows 2000 and had no trouble dual booting. That particular bug has been seen more often with the Test releases of Fedora Core, as should be expected. If it does happen to you, the problem can be easily fixed by running fixmbr in Recovery Console for 2000/XP.
Some people were guessing the problem is due to the 2.6 kernel reporting a different geometry from 2.4 and the tools not being updated to correct for this. I know that even though I didn't repartition, after installing FC2 over my FC1 installation the start and end cylinders reported by fdisk -l are completely different.
Right, I was hit by this. I'm a linux newbie. But I solved it.
/dev/hda | sfdisk --no-reread -H255 /dev/hda"
./sfdisk to make it work. In my case, i had to add the -force flag to the right hand side of the pipe.
:)
To fix it:
-If you don't already have it, get and install sfdisk (there are RPMs out there, no deps)
Run (presuming your hdd is hda) as root "sfdisk -d
You may have to cd to sbin and replace sfdisk with
That command ran, and then i could run WinXP from Grub just fine
However, FC2 has many other major bugs that I and others have found:
- Nvidia drivers don't work (i know it's nvidia's fault, but it's a stumbling block)
- As Xorg is in use, ATI drivers are a bitch to install (although if you use google there is a very good howto out there).
- The kludge i had to use to get software mixing working (dmix under alsa) was inexcusable. With alsa in 2.6, you'd think by default you'd have software mixing. An OS where I can't listen to XMMS and hear GAIM alerts at the same time is just ludicrous. Even sillier is the fact that GAIM alerts are queued, so when i close XMMS i get a minute solid of notification noises playing. Simultaneous sounds SHOULD work out of the box. Esound and arts are not in the equation any more, as alsa mixing is a much better solution - so why isn't it implemented?
- Totem just won't work. G-Streamer broke totally shortly afterwards.
- There's no easy way to edit your applications menu, without either SUing, or logging in as root. This seems daft for a multiuser OS like linux.
I know these bugs aren't Fedora only, but they need addressing if Fedora wants to remain OS of choice for many.
Gamers Europe - Gaming News. Reviews.
It is interesting that BIOS updates fix that problem for most people. I recently installed FC2 and had exactly the same problem booting between it and XP. I was finally able to fix it by changing the drive geometry setting in my BIOS from Auto (which was using CHS) to LBA. As soon as I did that it started to boot XP again.
From what I remember, not only was Firewire unstable in time for release, but was causing instability even for people without Firewire. I'd rather they held off on including it until it is stable rather than risk data loss by including it prematurely.
Ita erat quando hic adveni.
Actually, that's precisely what it is. 2.6 reports a different layout (typically 16 cylinders instead of, say 255 in the logical one) and grub sets itself up this way. Next, it appears that even when the BIOS was explicitly instructed to use LBA access, this is somehow overriden by GRUB, with the result that trying to launch the NT bootloader fails.
The sfdisk solution on Fedora's bugzilla fails when sfdisk figures a partition does not start at the right cylinder boundary. Apparently, one can try to change the head count only for the windows boot partition, with the hope that it fixes the boot-through-grub problem (I am yet to try this). I guess the biggest problem is for people who don't have an LBA option in BIOS.
As a proof that it's not Fedora-related, I have the same problem with mdk10.
Finally, there seems to be no problem if one sets up the installer's kernel to use LBA access for the hd (no switching to CHS occurs).
Hmm... So you're one of those anti-zealot zealots.
I'm trying to figure out what zealot-free system you'll be able to run...
Linux? right out.
Windows? Nope: OSS is "a cancer", DEVELOPERS! DEVELOPERS!
Sun? No, they were head ABM cheerleaders until a couple of weeks ago.
Apple? Obviously not.
Amiga? Nope: the OS was perfect in every way, only a conspiracy kept it down.
VMS? Ditto
OS/2? Ditto
BeOS? Ditto
*BSD? Maybe, but then again there's more than a few anti-viral license zealots
Netware? Possibly, but now they're in bed with SuSE.
Tandy TRS-80: That's the ticket. Nobody will admit to having used it, much less spew zealotry about it. What's more, you can pick up spare machines on ebay for a couple of bucks.
Win2K introduced "dynamic" disks, which changed how the partition table worked. Partition type 0x42 means the disk is dynamic and the real information is contained at the end of the disk. 0x42 is supposed to be a container partition meant to span the disk and say "don't mess with me". An exception are boot and/or system partitions, as those have to be read early during boot before the dynamic disk stuff is loaded. Thus, boot/system partitions can be type 0x42 but not span the disk.
Anyway, as a wild-assed guess I'd check that out. Perhaps lilo/grub doesn't play well with dynamic disks.
It has worked fine for me in every Redhat since 6.0 and every Mandrake since 7.
PartitionMagic has also worked, although there were some issues with PM 7 that repartitioning with DiskDrake solved.
Show me on the doll where his noodly appendage touched you.
This is Linux kernel 2.6 - Mandrake 10, Suse 9.1 and Fedora Core 2 all suffer from this problem.
Switching to Debian won't help if you want Linux kernel 2.6. Your paritition table will be fubared.
Furthermore, people do know what's causing the problem. The Linux kernel now doesn't show the same disk geometry as the BIOS does. The fix is to use sfdisk to recreate the partition table.
A year or so ago, I tried to install Mandrake 9, Red Hat 8 and Red Hat 7.3 on my old Compaq Deskpro 4000 machine. None of them would install. All of them would either fail to read the partition table - after creating and partitioning and formatting it using the utilities supplied with those distros - or would completely trash the partition table.
Red Hat 7.0, however, would blow on the system with absolutely no problem or complaint whatsoever.
After doing some partition work on my latest system with parted, Partition Magic 5 and Partition Magic 8 cannot in any way read my partition table. Windows (98 first, now 2000 and XP) loads fine, Linux (RH 7.3, Knoppix, other Live CD distros) loads fine, all other partition managers (BootItNG, Ranish) see and handle the partition table. ONLY Partition Magic cannot do anything with the partition table - and it is supposed to be the "premier" partition table manager on the market!
So now we have THIS crap with Fedora Core 2!
Guys, the partition table is NOT rocket science. It's a few bytes on a disk with a few variations in what each byte means. It's been around for decades.
So why in hell can't people who write this stuff GET IT RIGHT? What is the goddamn problem with you programmers?
I realize that hard disk manufacturers are constantly screwing around with their geometry reporting to the BIOS, and of course not writing any Linux drivers, but still a bug of this sort should not exist in any modern OS.
Get it together.
Richard Steven Hack - This sig is TOO GODDAMN SHORT TO DO ANYTHING USEFUL WITH! MORONS!
The fix is here.
The VIA C3 problem didn't get caught because it worked in the betas. The bug involved is in all the 2.6.x kernels but depends on the alignment and size of the kernel. While the beta kernel worked the final kernel didnt get lucky.
Ingo and others are currently working through this one to try and find the cause. At the moment nobody is sure if it is a Linux bug or a CPU errata being tripped.