... updating the cpus_allowed mask in the task struct. The dependency of cpusets on ... . It lets us avoid checking for a changed cpuset 'cpus' setting in ... directly following the patch: task-containersv11-add-tasks-file-interface.patch kernel/cgroup ...

linux-kernel - Paul Jackson - Oct 3 2007 - 04:42

... tasks file is there, by design, and has been there for ... it is not an urgent fix, in my experience. But there ... given some recent cpuset hotplug patches being worked by Cliff Wickman ... this bug, such that reattaching tasks to their current cpuset becomes ...

linux-kernel - Paul Jackson - Oct 3 2007 - 16:16

... the set_cpus_allowed() call, which checks for intersection between your new cpumask and ... and dropping the sched_hotcpu_mutex for each task, just around > the call to ... > consistency of cpu_online_map with the tasks cpus_allowed masks. > It's needed ...

linux-kernel - David Rientjes - Oct 12 2007 - 11:13

... skips the > > attachment of a task to the group it is ... . Cpusets depends on reattaching a task to its current > > cpuset, in ... the cpus_allowed mask in the > > task struct. > > Can you explain a ... to rewrite each pid in its 'tasks' file back to that ...

linux-kernel - Paul Jackson - Oct 3 2007 - 13:58

... to rewrite each pid in its 'tasks' file back to that > ... scheduler. What's the rationale for this? Given that later in ... done when changing the 'cpus' file manually. What's wrong with, ... be forked between reading the tasks file and doing the writes ...

linux-kernel - Paul Menage - Oct 3 2007 - 14:10

... made the tradeoff to make this task list walk an explicitly invoked user ... ) cgroups has a per-cgroup task list, and the above loop has cost linear in the number of tasks actually in the cgroup, plus ( ... ll ask Andrew to kill this patch of mine, and I will ...

linux-kernel - Paul Jackson - Oct 3 2007 - 16:52

... /sched.c: > > void sched_migrate_task(struct task_struct *task, > cpumask_t cpus_allowed) > { > mutex_lock(&sched_hotcpu_mutex); > set_cpus_allowed(task, ... Taking and dropping the sched_hotcpu_mutex for each task, just around the call to ...

linux-kernel - Paul Jackson - Oct 11 2007 - 19:15

... to just save references to the tasks. > > > > struct cgroup_iter it; > > struct task_struct *p, ... page is full) { need_repeat = true; break; } } } for each saved task p { set_cpus_allowed(p, new_cpumask ...

linux-kernel - Paul Menage - Oct 6 2007 - 17:09

... cs->css.cgroup, &it))) { > get_task_struct(p); > tasks[i++] = p; > } > cgroup_iter_end(cs->css.cgroup ... kernel/cgroup.c:cgroup_tasks_open(). Look for how pidarray[] is setup. And ... supporting the reading of the 'tasks' file by user code, this ...

linux-kernel - Paul Jackson - Oct 6 2007 - 15:59

... set_cpus_allowed(p, new_mask); if (!retval) { /* Check for races with cpuset updates */ cpus_allowed ... now outside the * permitted cpumask for the cpuset. Since a * change ... cpus resets the * cpumask for each task, do the same thing ...

linux-kernel - Paul Menage - Oct 10 2007 - 16:46

... be the best in terms of a clean interface. Then you could protect the whole thing by ... kernel/sched.c: void sched_migrate_task(struct task_struct *task, cpumask_t cpus_allowed) { mutex_lock(&sched_hotcpu_mutex); set_cpus_allowed(task, ...

linux-kernel - David Rientjes - Oct 10 2007 - 16:59

... () which skips the > attachment of a task to the group it is already in has ... > removed. Cpusets depends on reattaching a task to its current > cpuset, in order to trigger updating the cpus_allowed mask in the > task struct. Can you explain a bit more ...

linux-kernel - Paul Menage - Oct 3 2007 - 11:51

... > > The getting and putting of the tasks will prevent them from exiting or > > ... > as the end result is that the task's cpus_allowed are within the > cpuset's ... retval = set_cpus_allowed(p, new_mask); and allow a task to have a cpu outside of the cpuset ...

linux-kernel - David Rientjes - Oct 7 2007 - 02:15

... @google.com> wrote: > The getting and putting of the tasks will prevent them from exiting or > being deallocated prematurely. But ... achieved, as long as the end result is that the task's cpus_allowed are within the cpuset's cpus_allowed. Paul - To ...

linux-kernel - Paul Menage - Oct 6 2007 - 17:11

... > as if we were the actual tasks themselves. Which should simplify > locking. The Big ... some finer grain locking is essential for performance on higher processor count systems ( ... that depend on, or modify, data subject to such invariants must be done ...

linux-kernel - Paul Jackson - Oct 11 2007 - 21:23

... full) { > need_repeat = true; > break; > } > } > } > for each saved task p { > set_cpus_allowed(p, ... store the ones not yet fixed. Uunfortunately, I need to put ... Given that cgroups is targeted for 2.6.24, and that ...

linux-kernel - Paul Jackson - Oct 6 2007 - 17:41

... > > struct cgroup_iter it; > > struct task_struct *p, **tasks; > > int i = 0; > > > > cgroup_iter_start(cs->css.cgroup, & ... (cs->css.cgroup, &it))) { > > get_task_struct(p); > > tasks[i++] = p; > > } > > cgroup_iter_end(cs->css.cgroup, &it) ...

linux-kernel - David Rientjes - Oct 7 2007 - 02:13

... > > The key kernel routine for updating a tasks cpus_allowed > cannot be called while ... that callback. > > That, or my original patch, which calls the attach routine > even ... 's solution of grabbing references to tasks during the iteration and then doing ...

linux-kernel - Paul Menage - Oct 6 2007 - 16:53

... > after I've kmalloc'd the tasks[] array, filling in the tasks[] array. > > > > On a big cgroup on a big system, this could ... are quite a bit higher than for a line of user code. ... and end up harboring latent bugs for months or years, by which time ...

linux-kernel - Paul Jackson - Oct 11 2007 - 18:03

... css.cgroup, &it); This isn't working for me. The key kernel routine for updating a tasks cpus_allowed cannot be called while holding ... in plugging that callback. That, or my original patch, which calls the attach routine even if ...

linux-kernel - Paul Jackson - Oct 6 2007 - 04:24