Add some notes on using these shims to port Linux kernel code.

This commit is contained in:
riastradh 2014-07-03 20:48:19 +00:00
parent b01a0aea47
commit 29a70da5ff
3 changed files with 38 additions and 3 deletions

View File

@ -1,4 +1,4 @@
/* $NetBSD: completion.h,v 1.3 2014/05/05 15:59:11 skrll Exp $ */
/* $NetBSD: completion.h,v 1.4 2014/07/03 20:48:19 riastradh Exp $ */
/*-
* Copyright (c) 2013 The NetBSD Foundation, Inc.
@ -29,6 +29,20 @@
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Notes on porting:
*
* - Linux does not have destroy_completion. You must add it yourself
* in the appropriate place.
*
* - Some Linux code does `completion->done++' or similar. Convert
* that to complete(completion) and suggest the same change upstream,
* unless it turns out there actually is a good reason to do that, in
* which case the Linux completion API should be extended with a
* sensible name for this that doesn't expose the guts of `struct
* completion'.
*/
#ifndef _LINUX_COMPLETION_H_
#define _LINUX_COMPLETION_H_

View File

@ -1,4 +1,4 @@
/* $NetBSD: errno.h,v 1.1 2014/04/07 11:55:29 riastradh Exp $ */
/* $NetBSD: errno.h,v 1.2 2014/07/03 20:48:19 riastradh Exp $ */
/*-
* Copyright (c) 2013 The NetBSD Foundation, Inc.
@ -29,6 +29,15 @@
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Notes on porting:
*
* - Linux consistently passes around negative errno values. NetBSD
* consistently passes around positive ones, except the special magic
* in-kernel ones (EJUSTRETURN, ERESTART, &c.) which should not be
* exposed to userland. Be careful!
*/
#ifndef _LINUX_ERRNO_H_
#define _LINUX_ERRNO_H_

View File

@ -1,4 +1,4 @@
/* $NetBSD: list.h,v 1.1 2013/09/05 15:28:07 skrll Exp $ */
/* $NetBSD: list.h,v 1.2 2014/07/03 20:48:19 riastradh Exp $ */
/*-
* Copyright (c) 2013 The NetBSD Foundation, Inc.
@ -29,6 +29,18 @@
* POSSIBILITY OF SUCH DAMAGE.
*/
/*
* Notes on porting:
*
* - LIST_HEAD(x) means a declaration `struct list_head x =
* LIST_HEAD_INIT(x)' in Linux, but something else in NetBSD.
* Replace by the expansion.
*
* - The `_rcu' routines here are not actually pserialize(9)-safe.
* They need dependent read memory barriers added. Please fix this
* if you need to use them with pserialize(9).
*/
#ifndef _LINUX_LIST_H_
#define _LINUX_LIST_H_