tech-kern archive

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

Re: revivesa status 2008/07/09



On Sun, Jul 13, 2008 at 02:58:55PM +0100, Mindaugas Rasiukevicius wrote:
> Greg Troxel <gdt%ir.bbn.com@localhost> wrote:
> > To me, the binary compatibility issue is the biggest one - while running
> > 0.8 binaries is perhaps no longer important, 5.0 kernel on 4.0 userland
> > is still relevant.
> 
> AFAIK, all you need is just to upgrade your libpthread, and your 4.0 userland
> would work fine. Your old applications would work fine too.

Actually, no. As others have noted, the 4.0 libc didn't have the right 
syscall stubs needed by the 1:1 libpthread, so this doesn't just work.

Also, see:

http://www.NetBSD.org/cgi-bin/query-pr-single.pl?number=38804

It seems there are incompatabilities between the size of a mutex in the 
old and new libraries. I gather it is now bigger? There is a clean fix, 
which is to bump the libpthread major number. However the 5.0 libpthread 
then isn't a drop-in replacement for the 4.0 one due to the major version 
bump.

> But instead of that, you want to add ~3000 lines of code, which is quite
> complex, and not very resistant to the DoS attacks. I am not convinced :)

What DoS attack_s_? I'm aware of one, which is that SA really needs upcall
stacks for things work & things go bonkers when there aren't enough.  You
can kill the errant app, but the SA system should probably handle this
better. Are you aware of others?

I ask because DoS is a _quite_ reasonable concern about adding the code. 
It however has a simple solution, which is to just kill the application if 
it gets into the upcall-generating code and we don't have a stack 
available.

Take care,

Bill

Attachment: pgpMNUpQDMeJT.pgp
Description: PGP signature



Home | Main Index | Thread Index | Old Index