Hi,
Basile STARYNKEVITCH wrote:
>
> Yes. Of course, i meant a server able to give memory back (not wanting
> more memory -this is just malloc- but capable of free-ing some
> unimportant memory like buffer caches). I believe a FIFO discipline is
> enough -the kernel would send some FS_MEMORY messages until it has
> enough memory (or perhaps it could be a memory server?? which also
> implements malloc?? -or an mmap or sbrk primitive for allocating
> virtual|real memory- i'm not clear on that point). FS_MEMORY would be
> sent in turn to all servers until there is enough memory. I don't know
> how this FS_MEMORY (that i called MEMORY_LOW before) would be
> implemented.
The message implementation is pretty simple, the biggest problem is how the
kernel will decide that any particular process is a server and thus capable
of responding to a FS_MEMORY message. I guess the easiest thing would be to
check to see which processes have created a port and then go and try them.
I have a feeling that this isn't quite so straightforward because the kernel
isn't a normal client, but I guess this could be worked around.
Regards,
Dave
Received on Wed Nov 23 07:36:32 1994
This archive was generated by hypermail 2.1.8 : Thu Sep 22 2005 - 15:12:10 PDT