In addition to worrying about performance and scaling, with accounting enabled or disabled, one should also try to minimize code clutter in key kernel files, such as fork.c
For example, one might, instead of adding 40 lines os fork_connector() code to kernel/fork.c, instead add something like just the #include <linux/connector.h> and the "fork_connector(current->pid, p->pid)" call to kernel/fork.c, where include/linux/connector.h had something like: #ifdef CONFIG_FORK_CONNECTOR static inline void fork_connector(pid_t parent, pid_t child) { if (cn_fork_enable) __fork_connector(parent, child); } #else static inline void fork_connector(pid_t parent, pid_t child) {} #endif Then bury the interesting code in the implementation of __fork_connector(), in drivers/connector/cn_fork.c or some such place. This adds a real function call in the case that cn_fork_enable is set. That code path requires more than that anyway (and it makes kernel stack backtraces more transparent). But it removes 40 lines of fork_connector detail from fork.c. And it avoids marking a 40 line routine as inline ... -- I won't rest till it's the best ... Programmer, Linux Scalability Paul Jackson <[EMAIL PROTECTED]> 1.650.933.1373, 1.925.600.0401 - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/