🚀 go-pugleaf

jrm2k@uvacs.cs.Virginia.EDU (Jeff Michel) writes:
"Is this all a fluke?  Is there support (eg. a manual or commented header
"file) for using threads in the kernel?  What differences exist between the
"threads at the kernel level and at the user level?  Is it safe to use kernel
"threads in a device driver?

i was told [apropos my q. is lwp in the sunos5 compat lib] that *no*
special kernel resources were added to support lwp; it's all in the
standard c lib.

-- 
		+--------------------------------------------------------------+
Andrew Hay	|	one step beyond the outer limits of the twilight       |
a_d_hay@att.com	|	zone in the darkened room at the night gallery	       |
		+--------------------------------------------------------------+

In article <1993Jan25.155627.15411@cbnews.cb.att.com>, adh@cbnews.cb.att.com (andrew.d.hay) writes:
> jrm2k@uvacs.cs.Virginia.EDU (Jeff Michel) writes:
> "Is this all a fluke?  Is there support (eg. a manual or commented header
> "file) for using threads in the kernel?  What differences exist between the
> "threads at the kernel level and at the user level?  Is it safe to use kernel
> "threads in a device driver?
> 
> i was told [apropos my q. is lwp in the sunos5 compat lib] that *no*
> special kernel resources were added to support lwp; it's all in the
> standard c lib.

But there is a compile flag to configure a SunOS kernel which supports lwp's.
In fact, when you look at some of the OS sources, you will notice that the
kernel differs in some places if lwp's are supported.

There might even be threads at the kernel level. Consider this: Asynchronous I/O
via aread() and awrite() is only supported when lwp's are supported by the
kernel as well. This leads me to believe that lwp's are used to implement
asynchr. I/O. This doesn't surprise, in fact, that's what threads are for besides
other things. Now, maybe one could write kernel threads under SunOS 4.1 similar
to asynchr. I/O...

Of course, you can try to migrate to SunOS 5.x where threads are in the kernel.
If you can get the OS installed, that is.

Frank
mueller@cs.fsu.edu

jrm2k@uvacs.cs.Virginia.EDU (Jeff Michel) writes:
>"Is this all a fluke?  Is there support (eg. a manual or commented header
>"file) for using threads in the kernel?

 From my SunOS 4.1.2 kernel config file:
#
# The following two options are needed for asynchronous I/O.
#
options LWP             # kernel threads
options ASYNCHIO        # asynch I/O (requires LWP)

 I've never found documentation on what those options are really good for.
I would appreciate any info others might have.

--
|                                                                   Rob Quinn |
|                                                            rjq@phys.ksu.edu |
|                                                       QuinnBob@KSUVM.BITNET |

>That is, in addition to the lightweight processes library (-llwp) provided for
>users, there *appears* to be support for *kernel* threads.  Perhaps this is in
>order to support multiple asynchronous I/O operations and the like.

That's precisely why it's there.

They basically stuck code derived from "-llwp" into the kernel.

>Is this all a fluke?

No.

>Is there support (eg.  a manual or commented header file) for using
>threads in the kernel?

No.  It was stuck in (4.1 was the first release with it) only to support
the async I/O calls; proceed at your own risk.

>options LWP             # kernel threads
>options ASYNCHIO        # asynch I/O (requires LWP)
>
> I've never found documentation on what those options are really good for.

The first option is good for enabling the second option; SunOS 4.x only uses
kernel LWPs to support asynch I/O.

The second option is good for Sybase, which uses async I/O on raw
partitions, and may be useful for other applications such as databases
or tape dump/restore that use async I/O on raw disks or tapes.