Brian, I not only wish you luck, I'm praying for you.  I mean that
literally.  The issues with getting a STREAMS into a Linux kernel
are not technical issues at all; they're full-fledged religious
issues.  I know who the priests of the other religion are; been there,
done that (an important one was mentioned on this list recently; I'm
not going to honor the person by repeating their name).

You need prayer, not just wishes, but you have them from me. And
Godspeed to you, my friend (on this issue, if nothing else)... 8^)

-John

Brian F. G. Bidulock wrote:
John,

Yes, LiS correctly uses u.generic_ip only within its own filesystem.

My intention is to ultimately contribute Linux Fast-STREAMS patch into
the 2.4 and 2.6 kernels, and, as part of the kernel, should then be far
easier to maintain than LiS.  That's the plan anyway.  Wish me luck.

--brian

On Wed, 08 Oct 2003, John A. Boyd Jr. wrote:


One other thing, quickly.  Lis should never attempt to modify, or barely
otherwise even use, an inode belonging to any other filesystem.  Unless,
of course, someone who doesn't understand this to be the intent changes
the code to do otherwise.

Otherwise, the u.generic_ip field is used appropriately.  We've known
about this since LiS 2.8, at least, and I changed the code way back then
to avoid using non-LiS inodes directly.

I have to tell you, Brian, I shudder to think what you're facing in
terms of maintenance effort, if you're assuming that you'll have to
patch the kernel just for things to work.  But, that's not news to
you that I feel that way, I'm sure.  8^)

-John

Brian F. G. Bidulock wrote:

John,

LiS attaches stream heads to the u.generic_ip pointer, which
unfortunately is part of the file-specific union in the inode structure.
If LiS attempts to attach a stream head to a real filesystem inode the
way that Linux Fast-STREAMS is doing it, it will risk corrupting the
real filesystem's file system specific data in the inode.



_______________________________________________
Linux-streams mailing list
[EMAIL PROTECTED]
http://gsyc.escet.urjc.es/mailman/listinfo/linux-streams





_______________________________________________ Linux-streams mailing list [EMAIL PROTECTED] http://gsyc.escet.urjc.es/mailman/listinfo/linux-streams

Reply via email to