Incorrect f.s. info

Zigbert has described this problem here:

He burns to a CD in multisession mode, then burns another track, but after mounting the CD the 'ls' command shows the old contents. The CD has to be physically ejected then reinserted then mounted to force the system to update the f.s. info.

I think this problem is only with the new libata PATA/SATA drivers, not with the old IDE drivers, so you won't have it on the 4.1retro Puppy.

I have just been sitting here for a few hours, trying all sorts of things to try to force a refresh without having to eject the CD, but no-go. This is a fundamental problem with the kernel and libata, and yet another thing they need to address.

Posted on 18 Sep 2008, 19:23


Posted on 19 Sep 2008, 4:48 by Rarsa
Maybe autoeject
Maybe that's why all burning software I've ever used ejects the CD after burning even if it has to be reinserted for CD verification.

That sounds like a posible "woraround".

Posted on 19 Sep 2008, 8:29 by Jesse
Cdrom experience
Cdroms break some unix notions, they are not precisely just a block device, and not precisely just a character device. They are both. The usual notion for a data cdrom disk is that it is a block device with a filesystem, burning a cdrom changes the number of blocks (via ioctls / scsi commands), and the kernel block cache doesn't pay attention to those commands to notice the changes.
To get the kernel to notice the changes, the burning program must close its filedescriptor, ask the cdrom drive to clear its buffer cache, make sure no other program has the filedescriptor open, clear the kernels block cache for the cdrom drive, wait for the kernel to free its cdrom filedescriptor structures, reopen the filedescriptor, ask the cdrom to be rescanned. And then this may not work with all cdrom drives, but I suppose you could then eject it.
Phew! Sometimes its just easier to eject! which does all those things above.

Posted on 20 Sep 2008, 7:24 by BarryK
Not for IDE
NOTE: My comment above that the problem only occurs with PATA/SATA and not with the old IDE drivers is only supposition, as zigbert has only recently raised this as a bug for Pburn, and only in relation to a /dev/sr0 device.

Posted on 20 Sep 2008, 14:43 by Zigbert
Pburn 2.0.4 and Pbackup 3.1.4 is uploaded.

Posted on 20 Sep 2008, 14:47 by Zigbert
jake_take reported that Pschedule doesn't keep crontab for next boot on his usb installation, because /var/ is not saved. This works in frugal install.

Is it supposed to differ, or ...

Posted on 20 Sep 2008, 19:45 by BarryK
/var not saved
yes, that is inconsistent. I'll look into it.