Goals for first release (June/USENIX)

Matthew Dillon dillon at apollo.backplane.com
Thu Mar 11 12:40:06 PST 2004


:Hi !
:
:>     * Syscall interface layer and messaging (true async syscalls), and
:>       related upcall and other interfaces, such as placing the syscall
:>       compatibility layer into userland.
:Does this imply async IO as well ? Or is this still way too hard ?

    Yup, it does imply asynch I/O.

:>     * Userland VFS and DEV interfaces.  Not yet started.  I'll consider this
:>       goal as 'done' when I have sucessfully created a userland 'vn' like
:>       device and migrated, say, ext2fs, to userland.  As part of this I
:>       am going to try to 'fix' the buffer cache code to reduce the number
:>       of kvm mappings it has to do.
:Heeeey, are vfs environments included in that map ?
:(Thinking of the other thread on packages...)

    Maybe, maybe not.  As per another recent email I posted today we might
    not actually need vfs environments to bootstrap our packaging system.

:I know, I've been laying low on the userland scheduler.
:But I still want that thing done. My personal goal is to
:have basic encapsulation until May. And an alternative scheduler at least
:as a possible option in the kernel config for 1.0.
:Kernel modules for that I don't see before autumn...
:
:The toughest issue for me - concerning encapsulation alone - is where to get
:run- and sleeptimes  from the system in an appropriate and uniform manner.
:lwkt makes life easier, but not in that respect. Any hints ?
:(Maybe I'm just being blind...)
:
:Cheers
:Peter

    You'd have to be a bit more specific.  Run and sleep times are generally
    accounted for.  The only scheduler-specific case I guess is how it handles
    wakeup after sleeping for a long period of time.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Kernel mailing list