Hello,

I am installing Mandrake 9.2 on a compact flash drive. Long ago I used a RedHat 6.2 system installed on a compact flash drive and it was configured to be a read-only files system. If my memory is correct that was because the compact flash only has a finite number of writes and the linux logging can quickly comsume those writes. Again I recall having to pass the 'ro' kernel parameter. Other than those recollections I have no idea of the specifics for creating the system.

QUESTION/S: Is my recollection antiquated? Does anyone have advice for creating developer linux installs on compact flash drives? If so I would really appreciate your input, otherwise I am proceding with the normal install and hoping for the best.

Thanks,

I don't know that "ro" is neccessary, but noatime (man fstab, man mount) sure is. Otherwise, every time a file is accessed, a write will be generated. What is your goal with this installation?

Matir,


Thanks for the quick reply. I am building a system that is going to be placed on an outdoor research robot. In being such I need the hardened aspects of the compact flash as well as the ability to develop code for robot control.

I guess the question then is this: will there be any need for data collection or storage? I don't think mounting the system ro is really neccessary, though I would NOT put any sort of swapping on compact flash

Matir,

No Data collection or storage on the robot. All information gets packaged and sent to the user interface via wireless serial radios or a cell phone.

I currently have a bear bones install on the compact flash drive.

My fstab set noatime for the / and /home partitions. It also has a swap partition. Should I change the swap partition to be on a ramdisk?

Yes, having swap on the flash drive would certainly cause a LOT of write activity. However, swap on a ram disk would rather defeat the point. I would just try to build a system with sufficient ram, or use a secondary drive for swap. (If a flash drive, it will probably need to be tested and/or replaced often).

DuckSew,

I am about to do basically the same thing you are working on now.. I am using a small, old desktop PC (basically just the motherboard/proc/ram) as a kind of "fake embedded" system.

My approach is to build a working system on a real hard drive partition until I get it down to the correct size (128MB) and with the correct read-only/ramdisk setup.

I am about to start the ramdisk part of the process... do you have any documentation or references on creating and using a ramdisk in this type of application?


Thanks,
RG

Don't you just love it when you ask a question, then 2 minutes later find the answer yourself?

http://silent.gumph.org/content/4/1/...nux-on-cf.html

That howto explains it very well...