[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: proxy-defpager
From: |
Roland McGrath |
Subject: |
Re: proxy-defpager |
Date: |
Tue, 6 Aug 2002 15:52:33 -0400 (EDT) |
I used S_ as the prefix, to match our norm. (It probably wouldn't hurt to
make hurd/default_pager.defs include hurd_types.defs like all the others do
so that it would obey the normal -DSERVERPREFIX=S_ method. But I didn't
change it, and had avoided that originally, to minimize the modification of
the original Mach .defs file. Not that I think that really matters.) The
story with the user stubs is the same in swapon, and I made proxy-defpager
make its own stubs as swapon does.
I am ambivalent about including the new interface in libc. I guess the old
one is there is might as well just switch it. But I think it would
probably be better for this interface not to be in libhurduser. Aside from
default_pager_object_create, it's only used by swapon. One idea is not to
have default_pager_object_create be the thing that any Hurd user normally
uses, but rather say what they should do is use shm_open and mmap it or
getdport+io_map to get the memobj. What shm_open really means is creat of
a file in /dev/shm/, a tmpfs directory--in fact it would be better just to
use dir_mkfile in /dev/shm for an anonymous object. Then tmpfs is the only
thing that ever uses default_pager_object_create, and it seems ok for the
interface to be sort of "private" to the core Hurd programs.