On Wed, Jan 21, 2009 at 11:58:25AM +0530, Ananth N Mavinakayanahalli wrote: > On Mon, Jan 19, 2009 at 03:20:31PM -0800, Roland McGrath wrote: > > Thanks for working on this, Ananth. (Btw, it's "embed.") > > > > I think it would be less disruptive (and materially no different) > > to leave utrace_flags as it is. That field is the one (and only) > > that is used in hot paths (or used anywhere outside utrace.c). > > It might in future get moved around to stay in a cache-hot part > > of task_struct, for example. > > > > The long comment above struct utrace is really all about implementation > > details inside utrace.c and I don't think you should move that commentary > > to the header file. Instead, put a comment saying that the contents of > > struct utrace and their use is entirely private to kernel/utrace.c and it > > is only defined in the header to make its size known for struct task_struct > > layout (and init_task.h). > > > > I committed some cosmetic changes that will make for a little less flutter > > in your patch. > > Here is V2 of the patch. Tested and works fine. Same two tests on > powerpc fail, all tests pass on x86, while there are some occurances of > the ptrace.c WARN_ON. > > Roland, > I've tried to tweak the comments appropriately. Please feel free to > modify them as you consider fit.
Roland, Any updates on this and the utrace upstream integration front? Ananth