I am working on getting my FC2 build released, but I am worried about the problem with dual-booting Windows that I have seen. Is there any way to get the legacy partitioning info for the installer without any manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem? I've seen solutions that use the legacy EDD information to change the heads to 255/240, although I haven't been able to determine which disks to perform this on. I assume that int13_dev80 is only for device 0x80, thus the discussion about disks other than 0x80. Is Red Hat working on a fix/respin?
Here is what I have so far in the %pre (untested):
------------- modprobe edd # It seems to be only the heads that are incorrect H=$(echo "`cat /sys/firmware/edd/int13_dev80/legacy_max_head` + 1" | bc -l) S=$(cat /sys/firmware/edd/int13_dev80/legacy_sectors_per_track) C=$(echo "scale=0;`cat /sys/firmware/edd/int13_dev80/sectors` / $H / $S" | bc -l) # I'm not sure how to determine which drives are present (maybe # /tmp/hda vs. /tmp/sda??
echo "bios_cyl:$C bios_head:$H bios_sect:$S" > /proc/ide/hda/settings -------------
Does anyone see any problems with this--or any better solutions? I still haven't figured out how to catch the exception where the system disks are SCSI and the data disks are IDE (/ on SCSI).
Thanks,
Forrest
On Tue, 2004-06-29 at 13:10 -0700, Taylor, ForrestX wrote:
I am working on getting my FC2 build released, but I am worried about the problem with dual-booting Windows that I have seen. Is there any way to get the legacy partitioning info for the installer without any manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem?
I still have yet to reproduce the problem. If you have a way to reliably reproduce it and are willing to test, there's some work going on in upstream parted to try to make it a non-issue. As it stands right now, I'm following the discussion and it seems sane, but I can't do much in the way of testing. If I have someone who can reproduce and who will test, it's easy enough for me to make an updates.img with an updated libparted that will get used.
I've seen solutions that use the legacy EDD information to change the heads to 255/240, although I haven't been able to determine which disks to perform this on. I assume that int13_dev80 is only for device 0x80, thus the discussion about disks other than 0x80.
This is a hack and there are comments in various places about it not being good enough. It definitely doesn't at all feel robust to me.
Jeremy
On Tue, 2004-06-29 at 13:25, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:10 -0700, Taylor, ForrestX wrote:
I am working on getting my FC2 build released, but I am worried about the problem with dual-booting Windows that I have seen. Is there any way to get the legacy partitioning info for the installer without any manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem?
I still have yet to reproduce the problem. If you have a way to reliably reproduce it and are willing to test, there's some work going on in upstream parted to try to make it a non-issue. As it stands right now, I'm following the discussion and it seems sane, but I can't do much in the way of testing. If I have someone who can reproduce and who will test, it's easy enough for me to make an updates.img with an updated libparted that will get used.
Really? I thought that it was an issue for everyone. I haven't had time to even test it with Windows yet--I thought I'd get a fix first and test that. I'll see if I can reliably reproduce the problem and then I'll test libparted. Anaconda grabs the libparted from the installation tree during buildinstall, doesn't it? I can also test a newer parted with buildinstall.
I've seen solutions that use the legacy EDD information to change the heads to 255/240, although I haven't been able to determine which disks to perform this on. I assume that int13_dev80 is only for device 0x80, thus the discussion about disks other than 0x80.
This is a hack and there are comments in various places about it not being good enough. It definitely doesn't at all feel robust to me.
It certainly is a hack, so I'll try to reproduce the problem and test a newer parted.
Thanks,
Forrest
On Tue, 2004-06-29 at 13:35 -0700, Taylor, ForrestX wrote:
On Tue, 2004-06-29 at 13:25, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:10 -0700, Taylor, ForrestX wrote:
I am working on getting my FC2 build released, but I am worried about the problem with dual-booting Windows that I have seen. Is there any way to get the legacy partitioning info for the installer without any manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem?
I still have yet to reproduce the problem. If you have a way to reliably reproduce it and are willing to test, there's some work going on in upstream parted to try to make it a non-issue. As it stands right now, I'm following the discussion and it seems sane, but I can't do much in the way of testing. If I have someone who can reproduce and who will test, it's easy enough for me to make an updates.img with an updated libparted that will get used.
Really? I thought that it was an issue for everyone. I haven't had time to even test it with Windows yet--I thought I'd get a fix first and test that. I'll see if I can reliably reproduce the problem and then I'll test libparted. Anaconda grabs the libparted from the installation tree during buildinstall, doesn't it? I can also test a newer parted with buildinstall.
It would be far easier if it were an issue for everyone ;-) My new laptop came with XP pre-installed and I happily shrunk the partition and everything worked without any problems at all. The libparted gets pulled from the package tree, yes, but I haven't built a CVS parted package yet. You can if you'd like (or my offer definitely stands of making an updates.img)
I've seen solutions that use the legacy EDD information to change the heads to 255/240, although I haven't been able to determine which disks to perform this on. I assume that int13_dev80 is only for device 0x80, thus the discussion about disks other than 0x80.
This is a hack and there are comments in various places about it not being good enough. It definitely doesn't at all feel robust to me.
It certainly is a hack, so I'll try to reproduce the problem and test a newer parted.
That would be much appreciated by me at least :-)
Jeremy
On Tue, 2004-06-29 at 13:39, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:35 -0700, Taylor, ForrestX wrote:
On Tue, 2004-06-29 at 13:25, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:10 -0700, Taylor, ForrestX wrote:
I am working on getting my FC2 build released, but I am worried about the problem with dual-booting Windows that I have seen. Is there any way to get the legacy partitioning info for the installer without any manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem?
I still have yet to reproduce the problem. If you have a way to reliably reproduce it and are willing to test, there's some work going on in upstream parted to try to make it a non-issue. As it stands right now, I'm following the discussion and it seems sane, but I can't do much in the way of testing. If I have someone who can reproduce and who will test, it's easy enough for me to make an updates.img with an updated libparted that will get used.
Really? I thought that it was an issue for everyone. I haven't had time to even test it with Windows yet--I thought I'd get a fix first and test that. I'll see if I can reliably reproduce the problem and then I'll test libparted. Anaconda grabs the libparted from the installation tree during buildinstall, doesn't it? I can also test a newer parted with buildinstall.
It would be far easier if it were an issue for everyone ;-) My new laptop came with XP pre-installed and I happily shrunk the partition and everything worked without any problems at all. The libparted gets pulled from the package tree, yes, but I haven't built a CVS parted package yet. You can if you'd like (or my offer definitely stands of making an updates.img)
Yes, I'd like to see an updates.img (for those who don't rebuild from the tree), as well as a CVS parted package.
Speaking of shrinking XP--how did you shrink the NTFS partition? I was thinking of adding ntfsresize-static to the build and allowing users to shrink the NTFS/FAT32 partitions if they existed. Thoughts?
Forrest
On Tue, 2004-06-29 at 13:39, Jeremy Katz wrote:
It certainly is a hack, so I'll try to reproduce the problem and test a newer parted.
That would be much appreciated by me at least :-)
So if I do run into this problem, and cannot boot into Windows, what would be the most beneficial thing to do? Reimage Windows onto the hard drive and reverify that installing FC2 will not allow it to boot (see if I can do it twice in a row)? Try a new libparted/parted?
Forrest
On Tue, 2004-06-29 at 13:39, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:35 -0700, Taylor, ForrestX wrote:
On Tue, 2004-06-29 at 13:25, Jeremy Katz wrote:
On Tue, 2004-06-29 at 13:10 -0700, Taylor, ForrestX wrote:
I am working on getting my FC2 build released, but I am worried
about
the problem with dual-booting Windows that I have seen. Is
there any
way to get the legacy partitioning info for the installer
without any
manual intervention? Are there any updates for the kernel/parted/anaconda that would remedy this problem?
I still have yet to reproduce the problem. If you have a way to reliably reproduce it and are willing to test, there's some work
going
on in upstream parted to try to make it a non-issue. As it stands
right
now, I'm following the discussion and it seems sane, but I can't
do much
in the way of testing. If I have someone who can reproduce and
who will
test, it's easy enough for me to make an updates.img with an
updated
libparted that will get used.
Really? I thought that it was an issue for everyone. I haven't had time to even test it with Windows yet--I thought I'd get a fix first
and
test that. I'll see if I can reliably reproduce the problem and
then
I'll test libparted. Anaconda grabs the libparted from the
installation
tree during buildinstall, doesn't it? I can also test a newer
parted
with buildinstall.
It would be far easier if it were an issue for everyone ;-) My new laptop came with XP pre-installed and I happily shrunk the partition and everything worked without any problems at all. The libparted gets pulled from the package tree, yes, but I haven't built a CVS parted package yet. You can if you'd like (or my offer definitely stands of making an updates.img)
Well, after many days of testing on various machines, I have never been able to reproduce the problem. I've released the discs into the wild here, so I'll see if anyone else can reproduce the problem. One thing to note is that I did use all the latest updates in the build, so perhaps the newer kernel fixed the problem.
Forrest
anaconda-devel@lists.fedoraproject.org
-
Jeremy Katz -
Taylor, ForrestX