Re: Linux problems

["Thomas Schmitt" <scdbackup@gmx.net>]

Hi,

> There is a potential problem with your idea about doing something in open().
> Open() is a C library call, not a kernel API call.

Oh, not for general open(). But for the code of
the hd , sg , and sr drivers which they perform
on open(). The same code that is able to spoil
our burns. To my experience an 
  open("/dev/sr0", O_RDONLY);
already sends SCSI commands which can do harm.

Some instance in Linux must trigger those specific
commands on the generic gesture open(). For now i
am reluctant to dive into this architecture.
Exactly one year ago i got dragged into the lower
levels of burning and i still understand only half
of what a CD actually is. (E.g. video CD and mixed
mode and the colored books and and and)
I don't want to open a new can of worms before i
am done with the old one. :))


> Also, I've not found a way to guarantee that my patch never causes the type
> of deadlock that forces at least on program to fail.  I don't consider this
> fatal, but others might disagree.

Tell me more about your deadlock problem.
Give me a sketch of how your idea would work.
Just so i can join you with pondering.


Have a nice day :)

Thomas