Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-02-02 Thread Luca Abeni
Hi Peter, On 01/31/2015 10:56 AM, Peter Zijlstra wrote: On Fri, Jan 30, 2015 at 10:35:02AM +, Juri Lelli wrote: So, we do the safe thing only in case of throttling. No, even for the !throttle aka running tasks. We only use dl_{runtime,deadline,period} for replenishment, until that time

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-02-02 Thread Juri Lelli
On 31/01/2015 09:56, Peter Zijlstra wrote: On Fri, Jan 30, 2015 at 10:35:02AM +, Juri Lelli wrote: So, we do the safe thing only in case of throttling. No, even for the !throttle aka running tasks. We only use dl_{runtime,deadline,period} for replenishment, until that time we observe

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-31 Thread Peter Zijlstra
On Fri, Jan 30, 2015 at 10:35:02AM +, Juri Lelli wrote: So, we do the safe thing only in case of throttling. No, even for the !throttle aka running tasks. We only use dl_{runtime,deadline,period} for replenishment, until that time we observe the old runtime/deadline set by the previous

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-30 Thread Juri Lelli
Hi Peter, On 28/01/2015 14:08, Peter Zijlstra wrote: On Thu, Jan 15, 2015 at 02:35:46PM +0100, Luca Abeni wrote: From what I understand we should either modify the tasks run/sleep stats when we change its parameters or we should schedule a delayed release of the bandwidth delta (when it

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-30 Thread Luca Abeni
Hi Peter, On 01/28/2015 03:08 PM, Peter Zijlstra wrote: On Thu, Jan 15, 2015 at 02:35:46PM +0100, Luca Abeni wrote: From what I understand we should either modify the tasks run/sleep stats when we change its parameters or we should schedule a delayed release of the bandwidth delta (when it

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-28 Thread Luca Abeni
Hi Peter, On 01/28/2015 03:08 PM, Peter Zijlstra wrote: On Thu, Jan 15, 2015 at 02:35:46PM +0100, Luca Abeni wrote: From what I understand we should either modify the tasks run/sleep stats when we change its parameters or we should schedule a delayed release of the bandwidth delta (when it

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-28 Thread Peter Zijlstra
On Thu, Jan 15, 2015 at 02:35:46PM +0100, Luca Abeni wrote: From what I understand we should either modify the tasks run/sleep stats when we change its parameters or we should schedule a delayed release of the bandwidth delta (when it reaches its 0-lag point, if thats in the future). I

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-15 Thread Luca Abeni
Hi Kirill, On 01/14/2015 01:43 PM, Kirill Tkhai wrote: [...] Say we have a userspace task that evaluates and changes runtime parameters for other tasks (basically what Luca is doing IIRC), and the changes keep resetting the sleep time, the whole guarantee system comes down, rendering the

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-15 Thread Peter Zijlstra
On Thu, Jan 15, 2015 at 12:23:43PM +0100, Luca Abeni wrote: There are some parts of the patch that I do not understand (for example: if I understand well, if the task is not throttled you set dl_new to 1... And if it is throttled you change its current runtime and scheduling deadline...

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-15 Thread Luca Abeni
Hi Peter, On 01/15/2015 01:23 PM, Peter Zijlstra wrote: On Thu, Jan 15, 2015 at 12:23:43PM +0100, Luca Abeni wrote: There are some parts of the patch that I do not understand (for example: if I understand well, if the task is not throttled you set dl_new to 1... And if it is throttled you

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-14 Thread Kirill Tkhai
13.01.2015, 17:04, Peter Zijlstra pet...@infradead.org: On Tue, Jan 12, 2015 at 12:26:40PM +0300, Kirill Tkhai wrote:  Well, I'm inclined to agree to Luca's viewpoint. We should not change  parameters of a throttled task or we may affect other tasks.  Could you explain your viewpoint more? How

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-13 Thread Juri Lelli
Hi all, really sorry for the huge delay in replying to this! :( On 07/01/2015 12:29, Kirill Tkhai wrote: On Ср, 2015-01-07 at 08:01 +0100, Luca Abeni wrote: Hi Kirill, On Tue, 06 Jan 2015 02:07:21 +0300 Kirill Tkhai tk...@yandex.ru wrote: On Пн, 2015-01-05 at 16:21 +0100, Luca Abeni

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-13 Thread Peter Zijlstra
On Tue, Jan 12, 2015 at 12:26:40PM +0300, Kirill Tkhai wrote: Well, I'm inclined to agree to Luca's viewpoint. We should not change parameters of a throttled task or we may affect other tasks. Could you explain your viewpoint more? How does this affects other tasks? I agree with Juri and

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-13 Thread Kirill Tkhai
Hi, Juri, 13.01.2015, 11:10, Juri Lelli juri.le...@arm.com: Hi all, really sorry for the huge delay in replying to this! :( On 07/01/2015 12:29, Kirill Tkhai wrote:  On Ср, 2015-01-07 at 08:01 +0100, Luca Abeni wrote:  Hi Kirill,  On Tue, 06 Jan 2015 02:07:21 +0300  Kirill Tkhai

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-09 Thread Kirill Tkhai
09.01.2015, 14:15, Luca Abeni luca.ab...@unitn.it: Hi Kirill, On 01/07/2015 02:04 PM, Kirill Tkhai wrote: [...]  If in the future we allow non-privileged users to increase deadline,  we will reflect that in __setparam_dl() too.  Ok.  Does my patch help you? It helps me, but anyway I need

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-09 Thread Luca Abeni
Hi Kirill, On 01/07/2015 02:04 PM, Kirill Tkhai wrote: [...] If in the future we allow non-privileged users to increase deadline, we will reflect that in __setparam_dl() too. Ok. Does my patch help you? It helps me, but anyway I need your confirmation. Sorry about the delay... Anyway, I

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-07 Thread Kirill Tkhai
On Ср, 2015-01-07 at 13:45 +0100, Luca Abeni wrote: On 01/07/2015 01:29 PM, Kirill Tkhai wrote: [...] Based on your comments, I suspect my patch can be further simplified by moving the call to init_dl_task_timer() in __sched_fork(). It seems this way has problems. The first one is that

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-07 Thread Kirill Tkhai
On Ср, 2015-01-07 at 08:01 +0100, Luca Abeni wrote: Hi Kirill, On Tue, 06 Jan 2015 02:07:21 +0300 Kirill Tkhai tk...@yandex.ru wrote: On Пн, 2015-01-05 at 16:21 +0100, Luca Abeni wrote: [...] For reference, I attach the patch I am using locally (based on what I suggested in my

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-07 Thread Luca Abeni
On 01/07/2015 01:29 PM, Kirill Tkhai wrote: [...] Based on your comments, I suspect my patch can be further simplified by moving the call to init_dl_task_timer() in __sched_fork(). It seems this way has problems. The first one is that task may become throttled again, and we will start dl_timer

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-07 Thread Luca Abeni
On 01/07/2015 02:04 PM, Kirill Tkhai wrote: [...] and further enqueue_task() places it on the dl_rq. I was under the impression that no further enqueue_task() will happen (since the task is throttled, it is not on runqueue, so __sched_setscheduler() will not dequeue/enqueue it). But I am

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-06 Thread Luca Abeni
Hi Kirill, On Tue, 06 Jan 2015 02:07:21 +0300 Kirill Tkhai tk...@yandex.ru wrote: On Пн, 2015-01-05 at 16:21 +0100, Luca Abeni wrote: [...] For reference, I attach the patch I am using locally (based on what I suggested in my previous mail) and seems to work fine here. Based on your

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-05 Thread Kirill Tkhai
On Пн, 2015-01-05 at 16:21 +0100, Luca Abeni wrote: Hi Kirill, On 01/04/2015 11:51 PM, Kirill Tkhai wrote: Hi, Luca, I've just notived this. [...] I think we should do something like below. hrtimer_init() is already called in sched_fork(), so we shouldn't do this twice. Also, we

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-05 Thread Luca Abeni
Hi Kirill, On 01/04/2015 11:51 PM, Kirill Tkhai wrote: Hi, Luca, I've just notived this. [...] I think we should do something like below. hrtimer_init() is already called in sched_fork(), so we shouldn't do this twice. Also, we shouldn't reinit dl_throttled etc if timer is pending, and we

Re: Another SCHED_DEADLINE bug (with bisection and possible fix)

2015-01-04 Thread Kirill Tkhai
Hi, Luca, I've just notived this. 30.12.2014, 02:27, luca abeni luca.ab...@unitn.it: Hi all, when running some experiments on current git master, I noticed a regression respect to version 3.18 of the kernel: when invoking sched_setattr() to change the SCHED_DEADLINE parameters of a task

Another SCHED_DEADLINE bug (with bisection and possible fix)

2014-12-29 Thread luca abeni
Hi all, when running some experiments on current git master, I noticed a regression respect to version 3.18 of the kernel: when invoking sched_setattr() to change the SCHED_DEADLINE parameters of a task that is already scheduled by SCHED_DEADLINE, it is possible to crash the system. The bug can