PDA

View Full Version : Disk access issue



pinecloud
14th November 2008, 04:19 AM
I recently bought and attempted to install Fedora 9 to one of my system. I could not install it. Please allow me to suggest the following 3 item to improve linux.:(:(:(:(:(:(:(

1. Linux should give the user more flexibility about disk layout. If the user gives mounting point for each partition then why your install program stutter and keep rebooting?

2. Linux should not force to reboot when any disk layout is written. Check out Solaris! It never reboot after fdisk or format session.

3. Linux must update block device access table at kernel level to be par with the current disk size trends - the age of terabyte disk like Barracuda ES why not sda31 or sda23 instead of sda15 limitation?

In the recent past. I had given up Fedora 6, 7 and 8 which hang on SCSI CDROM boot displaying “cipio error”. Therefore I recently bought a pre-built DELL machine with SATA disks and DVD.

I was so happy to see the Fedora DVD booting into a professional looking 1280 x 1024 installation screen by auto detecting video card and monitor but even then I was not able to install at all.

I have been an ardent supporter of “util linux fdisk” program since it let you work with exact CHS numbers and more ever it can be used as partition recovery tool for fat16 and fat32 file system shared in a system before the days of advanced R-Studio style recovery editors.

But at this date most “util linux fdisk” causes kernel panic when it sees disks over 127GB. In turn, Fedora did not panic, but it was such dissapointing experience to find out the way Fedora 9 installer reacted to it. It kept rebooting both in graphic and text mode when I set custom disk layout to my liking. Perhaps underlying partitionning program did not know how to deal with maximum allowed 15 partitions. It acted like “disk druid” which I am most feared of. It did not seem to like predefined clean cut boundaries nor did large disks.

In the mid 1990, Linux did have more flexibility (Caldera to be on top because it relied on fdisk and user just needed to specify mounting point) in regard to custom disk layout over Windows, Solaris, BSD and like. Is it a current linux trend to force the default layout upon us? If I wanted my way it just keep rebooting time after time until I just give up?

I am not interested anything like LVM. I began to like Linux and Unix for being able to customise disk layout. Years ago when disks were of 9 to 18 GB, I used to split /usr/lib, /usr/X11, /export and swp between several disks up to 7.

Now the disks have got considerably faster, but at least I want to split between two or three disks to increase performance. I also manually partition the disk by cylinder and sector number for exact allocation. Of course, I followed “15 partitions max per disk rule”. Lot of times I have exactly 15 partitions due to large disk size nearing 300GB. Fedora installer accused me of more than 15 partitions but it was not true. I had 11 partitions on sdb and 15 partitions on sdc. sda is solely used as storage for compressed images of all the partitions on the sdb and sdc.

I never liked the way most of your Linux installer keep rebooting more than douzen times – It is very frustrating and waste of time. I wasted entire day yesterday. Installer should be able to write a script onto disk like IBM Aix and Sun Solaris do once the user gives installer a few partitions so that user does not have to go through the same routine. Solaris X86 has it’s own glitch but it got better and better except their inconsistency of partition naming convention between GRUB and Sun OS mtab entry like “c0d0t0p0 and so on” to confuse me when setting up multiboot chaining.

I hope to see the Fedora 10 adheres to proposed ability handling disk over 127GB and real 15 partition per disk as its kernel designed to do. If not then I would have to use a HBA from ICP Vortex and two or three large disks to appear to OS as if seven small disks (through virtual int 13 addresses from 80h to 86h on system side of on board pci bridge on hba). I remember Redhat 9 refused to mount anything SATA or SAS as late as 2004 whereas defunct Caldera 3.1 handled them well and reliably installed onto Serial ATA and Serial SCSI thinking them as plain IDE or SCSI in 2003. But adding just one large disk over 127GB on the same HBA crushed the any Caldera system.

Terminology of 15 partitions maximum also is very ambiguous. It seems to me that it is pre-defined in current kernel, sda sda1 sda2 . . . . . sda15 as device files. Therefore, if I create one extended partition taking entire disk space with 12 logical partitions in that extended partition, Kernel still sees as 16 partitions even though actuality is 13 since device space between sda2 and sda4 are not used. This scheme needs to be urgently updated to meet today’s disk space. Giving one more bit gives you 32 device space per disk that is more realistic for today’s disk size. I know this sda sda1. . .sda15 (sda without number being MBR/partition table and being one device) scheme go back to 1990 and no one made any effort to update it.

nick.stumpos
14th November 2008, 04:48 AM
sorry to take away from your rant, but this is a users board, and the devs don't tend to frequent it, if you really want your voice to be heard you should probably hit up the devel mailing lists

marcrblevins
14th November 2008, 07:04 AM
But at this date most “util linux fdisk” causes kernel panic when it sees disks over 127GB. In turn, Fedora did not panic, but it was such dissapointing experience to find out the way Fedora 9 installer reacted to it. It kept rebooting both in graphic and text mode when I set custom disk layout to my liking. Perhaps underlying partitionning program did not know how to deal with maximum allowed 15 partitions. It acted like “disk druid” which I am most feared of. It did not seem to like predefined clean cut boundaries nor did large disks.

fdisk can handle over 127 gig, you are probably using an old ATA33 controller to hit that 127 gig limit.

pinecloud
14th November 2008, 08:30 PM
Thank you for the comment. No, I am using Intel 945 Chipset and 82801GB (circa 2006) PCI-Express SATA port embedded in chipset. It is a recent Dell Dimension Dual Core machine. If I had space, I would much rather use most familiar PCIX64 GDT8623RZ made by ICP Vortex with faithful Seagate Cheetahs 15.5. But these Cheetahs need a 4" cooling fan per every two drives to keep temprature under 35C degrees. On other machines that I have build. I have not had any problem like this because everything is SCSI. Except I always had to use two floppies for installation. Kernel panic issue was Caldera 3.1 (2.4 Kernel Circa 2003) and Slackware recovery set with earliest 2.6 (sirca 2006) kernel. Do you think "Util Linux Fdisk" got improved since then? That would be great. Util Linux is the cleanest fdisk you can find in this world that I trust. In fact every distribution should include it as an option.
It is my opinion that all the user should know at least cylinder count build a disk layout with multification factor of 7.84429 devided by 1.05 to 1.12 for inode overheads into spread sheet file before using fdisk. This table know where to go if something happens to the parttions because it has starting and ending location of each partition. I save them into 512byte files. Pinecloud.