More syscall messaging commits, and some testing code as well.

Matthew Dillon dillon at apollo.backplane.com
Tue Aug 12 10:33:01 PDT 2003


:	Will the system calls be like atomic transactions? I.e., will it
:	be possible to ^C a program, and the currently executing system
:	call will rollback whatever it was doing?
:
:	I guess what I am asking might be a little superficial...

    Well, I don't expect system calls to operate much differently then
    they do now except that under normal conditions an async system
    call will not have to deal with abort/restart issues due to signals,
    except when a process is trying to exit.

:	IMHO, it will be a flexibility thing; to reinvent the wheel.
:	That is, someone (hint hint :) might want to use the accept()
:	call inside the kernel, without reinventing the wheel.
:
:	Ofcourse, this is (using accept() in kernel) something which not
:	everyone is going to do, but it will be flexbility which can be
:	provided, if someone choose to do so.
:
:-- 
:Hiten M. Pandya
:hmp at xxxxxxxxxxx, hmp at xxxxxxxx

    The kernel has a wonderful API called the 'system call' which it
    cannot use itself.  We definitely want the kernel to be able to use
    this API.  Ultimately access to the API will be accomplished through
    messaging via a layer... kinda like a userland emulation layer except
    for the kernel rather then for userland.

					-Matt
					Matthew Dillon 
					<dillon at xxxxxxxxxxxxx>





More information about the Kernel mailing list