Subject: re: kern/34010: undefined reference to `uvm_mremap' in LKM compat_linux.o.
To: None <kern-bug-people@netbsd.org, gnats-admin@netbsd.org,>
From: matthew green <mrg@eterna.com.au>
List: netbsd-bugs
Date: 07/16/2006 20:05:03
The following reply was made to PR kern/34010; it has been noted by GNATS.

From: matthew green <mrg@eterna.com.au>
To: YAMAMOTO Takashi <yamt@mwd.biglobe.ne.jp>
Cc: kern-bug-people@netbsd.org, gnats-admin@netbsd.org,
	netbsd-bugs@netbsd.org, gnats-bugs@NetBSD.org
Subject: re: kern/34010: undefined reference to `uvm_mremap' in LKM compat_linux.o. 
Date: Mon, 17 Jul 2006 06:04:11 +1000

    > -file	uvm/uvm_mremap.c		compat_linux
    > +file	uvm/uvm_mremap.c
    
    i'd like to explain the intent of the current code
    so that an lkm/configure guru can decide what would be a good fix.
    
    - its primary purpose is a backend of linux mremap system call.
    
    - however, what uvm_mremap.c implements is more than linux mremap.
      eg. it's intended to support inter-vm_map move.
    
    - i made it compat_linux dependent because no one else uses it currently.
    
    - otoh, i put it in uvm/ because its functionality can be useful for others.
      (and i was thinking about introducing netbsd-native version of
      the system call.  it makes significant performance difference for
      some applications, especially for ones written for linux. :-)
 
 
 given this, i would make 2 changes.  until we introduce our own
 mremap(2), mark the above as "compat_linux | lkm" so that it is
 available for LKM's.  then, i think we should have mremap(2) and
 then we can just remove the conditional in uvm/files.uvm entirely.