[rfc] use ext3 instead of ext2

Bryan Kadzban bryan at kadzban.is-a-geek.net
Sat Jul 17 05:06:22 PDT 2004


Ian Molton wrote:
> On Fri, 16 Jul 2004 17:35:04 -0400
>> Well, no, only with a preemptible kernel
> 
> You could run a non-preemptable SMP kernel...

That's right, I keep forgetting that preempt is built on top of the SMP
code.  OK, so "only with a preemptible or SMP kernel".

>> (assuming the driver lets the lock go when the syscall is finished)
> 
> Locks arent (necessarily) held for the duration of a syscall.

Right (and locking that's finer grained than syscall level is *always*
good, when possible).  But what I meant was, any locks that the syscall
takes need to be released by the time it's finished running.  There
aren't any syscalls AFAIK that that require some other syscall to be
called later to clean stuff up, because then if the userspace program
exits unexpectedly, that lock (or memory area, or whatever) will never
be released (or freed, or whatever).

I should have said "some time before the syscall is finished", rather
than "when the syscall is finished", basically.  "When" is too strong of
a restriction.



More information about the lfs-dev mailing list