Related link: http://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=115980
In response to my recent article on problems booting a Wndows XP partition after installing Fedora Core 2 a couple of the Fedora developers wrote me. They defended the choice of not treating the bug as a show-stopper and also scolded me a bit for possibly giving others ammunition to use FUD against the Fedora project. Their thoughtful response deserves the same forum as my warning and complaint, so with their permission I’ve posted relevant excerpts.
Jack Aboutboul responded:
I was alerted of this blog entry just a short time ago, and after reading it, I was pretty shocked at what was posted. I just wanted to send you some information regarding the bug you talk about and hope that you can somehow update this blog entry with the correct information, so as not to cause confusion.
The bug you speak about in the article stems [from] a few subtle,yet key feature changes that have happened within the 2.6 kernel. For 2.6 the kernel developers pulled out certain functionality from within the kernel related to Hard Disk geometry. In 2.6 the developers though that it may be wiser to have user space take control of things such as HD geometry. There is still discussion as to whether or not this was such a bright idea.
In addition to that, it is noted in the bugzilla report that the cause of this bug seems to be a CHS geometry problem. This problem most likely stems from an error within the parted utility, addressing bios incorrectly. In fact, many users with this bug reported that it disappeared after updating their bios to newer versions. Other than that, exhaustive testing has been done on this bug and newer machines did not seem to have this problem. It was almost totally non-reproducible.
I hope that I have been able to shed some light on this issue. You do very great work, and we salute you for trying to bring to the surface what seems like a stop-ship bug. However, I hope that the information I have provided you with will help you further investigate this problem so that you can update your blog, and spare the Fedora Project some great amounts of nonsensical FUD, that will undoubtedly be spread around by many unknowing readers of your blog.
Mr. Aboutboul in a later message said:
It is an issue that people should know about. The truth of the matter is, that almost none of the developers dual boot, and hence,no testing. Even the few that did try it, did so on IBM laptops and desktops, and didnt have any problems.
How do you think distributions should handle elusive bugs that may not be their fault, and could be wrongly used as fuel for FUD?
I stand by my original blog
Thanks to those who responded to my blog, and I want to reiterate that Fedora Core is my distro of choice, and I'd be as sad as anyone else if people misused this matter as a reason for spreading FUD about Fedora. I hope that one of the strengths of Fedora is a community that prides itself in making sure that others who give the OS a try get the most gain from the experience and see the least harm.
I spent several hours looking into this entire matter before posting the original blog. I read the bug entry from top to bottom. I read all the relevant mailing list postings I could find. I even did some searches to see if I could see some patterns to the problems. I'm no stranger to Linux, nor to basic hardware architecture, including basics of disk geometry. I did not seem to me that upgrading the BIOS was a reliable fix, nor was manually setting the BIOS to LBA mode either before or after the FC2 install. It fixed things for a few people, but not for many others.
The lack of a reliable and well-understood fix combined with the fact that I don't think people have been well enough warned about this potential problem was my reason for raising the alarm. I really am worried about people rushing in and inadvertently breaking their XP partitions. Best case they lose valuable time reinstalling and restoring. Worst case they lose priceless data.
I also want to point out that I'm also almost Windows free, and have been for years. I have the WinXP that came with my Dell laptop and once in a blue moon I need to boot into it. I run a consulting business and sometimes I have no choice but to deal with a Windows-only app. (Side note: I bought VMWare in hopes of eliminating this need). Naturally I'm sympathetic to others in my shoes and I think we should be careful of marginalizing people who still for any number of reasons need to access Windows in addition to Linux.
--Uche
http://uche.ogbuji.net
OT: Fused participle
The owl Fowler bugbear.
"I really am worried about people rushing in..."
I meant the more proper
"I really am worried about people's rushing in..."
Then again I may be one of the last few people on earth who take pains to avoid fused participles.
--Uche
Potential solution from NTFS Resize FAQ
http://mlf.linux.rulez.org/mlf/ezaz/ntfsresize.html
http://mlf.linux.rulez.org/mlf/ezaz/ntfsresize.html#troubleshoot
Definitely a Problem!
Some have argued that the bug is not a problem since a dual boot configuration is a transitory configuration. They couldnt be more wrong. I am a student and every CS student I know has a dual boot configuration so that they can develop, use and learn on both Windows and Linux simulataneously. 4 hours and one Windows recovery CD use later I finally got XP back but still can't boot into Fedora. I was happy with Redhat 9 and I am disappointed that Fedora developers should be so irresponsible as to dump such a buggy system as a final release on users.
XP/FC2 multiboot problems
I seem to have determined what the problem is; at least for my hardware anyway. Using two different ASUS P3B-F boards (processor is irrelevant if this is a BIOS disk geometry issue, which I doubt, as you will see later) and 3 different hard drives and setups I reproduced the problem and managed to fix it in both instances.
It appears that the GRUB commands in the
/etc/grub.conffile are incorrect with respect to booting other operating systems. FC2 for some reason writes the commands for the other operating systems as such:rootnoverify (hd0,0)chainloader +1
Astute readers will note that this is correct for booting Windows operating systems (which works when Windows is installed on the same disk, but not on a different disk) but will not boot any Linux distro. In the case of Windows being installed on a different disk, the
mapcommand must be added beforerootnoverify(twomapcommands, incidentally) so that GRUB will think disk 1 is disk 0 and disk 0 is disk 1. The makeactive command needs to be added before thechainloadercommand, to boot, pun intended.To fix the Linux distros, the commands should be:
root (hd0,0)kernel [name and location of kernel]
initrd [name of initial ram disk image]
Hopefully, this information will help others running into this issue and prevent them from having to re-install their other operating systems.
Definitely a Problem!
I have used linux (RadHat and SuSE) for several yers. And on almost every laptop I have had I have been running booth Linux and Windows (I need booth environments in my work), and I have never had a problem with either installations or dualboot, and didn't expect to have when I was going to install Fedora 3 on my IBM Thinkpad T41 last week.
Of cause I ran into the mentioned dualboot bug, I was unable to boot Windows XP, and none of the mentioned workarounds worked - tried to rewrite partition table with other Header values (240 and 255), rewrited the MBR several times, run fixboot and fixmbr from the windows xp recovery CD but nothing worked until I wiped all partitions and reinstalled Windows.
However now I am not sure I dare to install Fedora again, and this is in my opinion a huge threat towards booth Fedora and Linux towards the desktop in common...