On 9 Jun 2002, ulrich mensfeld wrote:
The udf filesystem seems to be inefficient at handling many small files.
The pktcdvd module is bypassing the the I/O elevator when creating write requests for the CDRW drive. This can make performance really suffer when
i copied the linux source-tree to the cdrw with ext2-fs now. My experiences on 'cp -a /usr/src/linux/* /mnt/cdwriter':
After a short time my system 'freezes' nearly completely for about 15 minutes. That means, i couldn't use keyboard nor mouse, even the LED's numlock on the keyboard wouldn't react of the num-key. But the cdrw was blinking and was written obviously; and the system was allways working as an internetrouter for my son's windows-pc. Then at first i get back the mouse (i did that all under kde3, in some xterms); after some additional minutes the keyboard.
This probably happens because all of your memory is filled up with dirty buffers waiting to be written to the CDRW. This means that all processes that tries to allocate memory may have to wait a very long time before memory becomes available. There was some talk on the linux kernel mailing list some time ago about limiting the amount of dirty data for slow devices, but I don't think any code has been written yet.
This freeze never happened when copying to udf-fs. The whole copying-time was about 30 minutes.
Maybe this is because the udf filesystem is slower for this case so there won't be so many dirty buffers laying around.
Another problem (or is that normal using ext2?): mounting the cdrw (after writing to it) in my cdrom instead of the burner, i couldn't see any files.
I would have thought it would work, but I don't have a CDROM drive to test this. Does the same thing work if you use the udf filesystem? Did the mount command succeed for the ext2 case or did you get some error message?
there is a mixed read/write load. The 2.5 version of pktcdvd has fixed this problem, but a backport is not easy because it relies heavily on the new bio infrastructure in 2.5.
Is the 2.5 kernel worth a try? I'm running no system with special hardware, only a standard-pc with duron 800, matrox 100 and symbios-logic scsi. Would it be stable on this platform?
I don't know if 2.5 will work with this setup. I guess the only way to find out is to try it. I wouldn't run 2.5 on a system containing important data though.
What about the corrupted files? Is that managed in this version?
I don't know what's causing the corruption, so I don't know if it is fixed. What did the corruption look like in your case? In my case the corruption was always a 2Kb block at file offset 0x5800 or 0x6000. -- Peter Osterlund - petero2@telia.com http://w1.894.telia.com/~u89404340