[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: HDB style locks (was "kermit")



Daniel Quinlan writes:
> 
> >>> FSSTND decided against the SCO and SVR4 implementations for various
> >>> reasons.
> 
> (Didn't I explain this before?)
> 
> a) There are proposed plans to change the device numbering scheme in
>    Linux eventually (that might conceivably affect this).

No, it wouldn't.  Linux will always use major/minor numbers to specify
devices; the only changes might come in which devices are associated with
which numbers and how many bits can be stored for each number.

> b) Some devices share the same device, but have different major
>    numbers (i.e., /dev/cua0 and /dev/ttyS0).

Well, HDB doesn't fix this either, but it is a valid point: as the
benefit of the SVR4 scheme is to help prevent lost locks due to device
"aliasing", cases where it doesn't do its job show a strong argument
against it.  Also, since Debian and FSSTND have declared that symlinks
do not belong in /dev, there really isn't much the SVR4 scheme can help
with.

Oh, well.  A perfect solution would be a set of C library functions
lock_device() and unlock_device(), the first returning a status code
and optionally waiting until the device comes free.  That way, it
wouldn't matter if /var/lock/LCK..whatever files or SYSV semaphores
were used.  But it's too late to go back...



Reply to: