[opensuse-kernel] openSUSE 10.3 beta 1: problem with libata and hard drive geometry
I tried running the openSUSE 10.3b1 installer (using the GNOME x86 CD), and ran across a problem with detected disk geometry on a Thinkpad T41p, with no other OS installed. I have had problems in the past with openSUSE 10.2: fdisk detects a C/H/S geometry of 9729/255/63, while the kernel itself would detect a much higher C value, with H=16. On my first install things worked fine for a while, until the system locked up during use a few days afterwards (running the latest kernel), which is probably due to the kernel accessing the disk (and writing to it) using the wrong geometry. On 10.2 I can just pass 'hda=9729,255,63' to force the kernel to use the same geometry used to partition the disk. On 10.3, I tried passing 'sda=9729,255,63' but to no avail; YaST would warn that parted could not access the disk and thus I could not touch the partition table, and dropping to a console and checking with fdisk shows that only 9300-something cylinders are detected. Has something changed in the kernel between 2.6.18 and 2.6.22 that causes even fdisk/parted to see a different geometry? How can I override this at boot time? Out of curiosity, I've seen claims that keeping a Windows partition on disk would force parted etc. to use the disk geometry that Windows sees, so a consistent geometry is always used. Any truth to that? Many thanks, -- Michel -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
On 18/08/07, Michel Salim
I tried running the openSUSE 10.3b1 installer (using the GNOME x86 CD), and ran across a problem with detected disk geometry on a Thinkpad T41p, with no other OS installed.
I have had problems in the past with openSUSE 10.2: fdisk detects a C/H/S geometry of 9729/255/63, while the kernel itself would detect a much higher C value, with H=16. On my first install things worked fine for a while, until the system locked up during use a few days afterwards (running the latest kernel), which is probably due to the kernel accessing the disk (and writing to it) using the wrong geometry.
On 10.2 I can just pass 'hda=9729,255,63' to force the kernel to use the same geometry used to partition the disk.
On 10.3, I tried passing 'sda=9729,255,63' but to no avail; YaST would warn that parted could not access the disk and thus I could not touch the partition table, and dropping to a console and checking with fdisk shows that only 9300-something cylinders are detected.
9364 cylinders, the rest presumably gobbled up by the Hidden Protected Area. Funnily, setting it to "disabled" in the BIOS does not affect things. I tried using "hwprobe=-modules.pata" as specified in the release notes, but now encounter another problem: YaST would read the hard drive OK, and would verify the installation CD OK as well, but after that, could not find the installation medium, and would continue asking for "openSUSE 10.3 CD 1" to be inserted. This looks like a YaST bug -- is there a way to circumvent this? e.g. manually mounting the CD in a specific directory. Failing that, can I instruct hwprobe to just skip pata for ide0 (hard drive is attached as hda) but use pata for ide1 (optical drive was hdc), and thus YaST would find the optical drive where it expected it? Thanks, -- Michel -- To unsubscribe, e-mail: opensuse-kernel+unsubscribe@opensuse.org For additional commands, e-mail: opensuse-kernel+help@opensuse.org
participants (1)
-
Michel Salim