Bizzare filesystem bug

Carnildo · Guru Joined: 17 Jun 2004 Posts: 595

I've got a 4GB CompactFlash card I'm trying to set up as the boot disk for one of my computers. I've got it partitioned and formatted as ext2, and I'm copying the filesystem over one top-level directory at a time using "cp -x -a". However, some of the files aren't being copied correctly.

For example, after copying, the new /bin/tempfile contains the contents of "/usr/portage/x11-plugins/wmjsql/wmjsql-0.5.ebuild", enough nulls to pad to the 4096-byte mark, the contents of "/usr/portage/x11-plugins/wmjsql/ChangeLog", enough nulls to pad it out to the 8192-byte mark, the contents of "/usr/portage/x11-plugins/wmjsql/Manifest", and enough nulls to pad it out to the original file's size. The new /bin/cpio starts with what looks like the directory entry for "/usr/portage/x11-plugins/wmifs/wmifs-1.3_beta1-r2.ebuild", followed by various files from that general area of the Portage tree.

When I was partitioning the disk, I had to use fdisk, as cfdisk quit with the message "FATAL ERROR: Bad primary partition 0: Partition ends in the final partial cylinder". The card is slightly smaller than another 4GB card that I'm successfully using as a hard disk, at 3997163520 bytes versus 4034838528 bytes. It also has a very different logical geometry: 245 heads, 63 sectors/track, 505 cylinders versus the working card's 128 heads, 63 sectors/track, 977 cylinders.

Any ideas?

hoacker · Posted: Sun Oct 23, 2011 9:01 pm Post subject:

I'd think of three possibilities:
1) The original files are already broken
2) "cp" is doing something really weird
3) The CF card is broken

I'd say that 3) is the most likely. Do you have a spare CF card to test?

Carnildo · Guru Joined: 17 Jun 2004 Posts: 595

1) isn't the case: the source files I've checked look fine, and if some of the corrupted files (eg. libc.so) were corrupted at the source, the system would be totally non-functional. I think I can rule out 2), first because only a few files are corrupted, and second because copying the files that were corrupted one at a time works fine.

There are two other options I can think of.
4) The computer's disk controller isn't correctly handing a CompactFlash card connected to a PATA socket (this shouldn't be an issue, though, since the only difference between CF and PATA is the size of the socket)
5) Something's wrong with the partition information or logical geometry that's causing data to not be read/written correctly.

s_bernstein · Posted: Mon Oct 24, 2011 6:10 am Post subject:

I would try another CF. I had some weird problems with partially broken CFs in the past, like important files magically vanishing from the file system. I didn't realized early enogh, that it was the card that was broken, so a few month later the card suddenly stopped working.

MacGyver031 · Posted: Mon Oct 24, 2011 6:39 am Post subject:

Could you do a badblocks scanning to verify that the CF is fully functional?

Ant P. · Watchman Joined: 18 Apr 2009 Posts: 6920

Make sure you don't have a cheap ripoff card that reports a fake size but implements it as a circular buffer over tiny storage. They exist.

Carnildo · Guru Joined: 17 Jun 2004 Posts: 595