Messages in this thread | | | Date | Mon, 11 Apr 2011 16:31:10 +0800 | From | Lai Jiangshan <> | Subject | Re: [RFC PATCH 4/5] RCU: Add TASK_RCU_OFFSET |
| |
On 04/11/2011 01:12 PM, Paul E. McKenney wrote:
> -static inline struct task_struct *next_thread(const struct task_struct *p) > -{ > - return list_entry_rcu(p->thread_group.next, > - struct task_struct, thread_group); > -} > +/* Avoid #include hell for inlining rcu_read_lock(). */ > +#define next_thread(p) \ > + list_entry_rcu((p)->thread_group.next, struct task_struct, thread_group) >
It is strange for me. The user of the API "next_thread(p)" must has two headers included, sched.h and rculist.h
I know this is a very popular pattern of user space code, is it OK for kernel? I think a file(even a header file) uses API(Marco), it should includes the the corresponding headers, it reduces surprises(example, the name of "next_thread()" has no "rcu", it is not expected that rcuxxxx.h is required).
I admit the work will become very much simpler if this pattern is allowed.
man fcntl: #include <unistd.h> #include <fcntl.h>
int fcntl(int fd, int cmd, ... /* arg */ );
| |