Re: Towards splitting SCHED_LOCK()

2020-01-15 Thread Martin Pieuchot
On 14/01/20(Tue) 17:30, Martin Pieuchot wrote: > On 14/01/20(Tue) 10:34, Jonathan Gray wrote: > > On Mon, Jan 13, 2020 at 05:02:12PM +0100, Martin Pieuchot wrote: > > > I'm facing a lock ordering issue while profiling the scheduler which > > > cannot be solved without using a different lock for

Re: Towards splitting SCHED_LOCK()

2020-01-14 Thread Theo de Raadt
Mark Kettenis wrote: > > Date: Tue, 14 Jan 2020 10:34:05 +1100 > > From: Jonathan Gray > > > > On Mon, Jan 13, 2020 at 05:02:12PM +0100, Martin Pieuchot wrote: > > > I'm facing a lock ordering issue while profiling the scheduler which > > > cannot be solved without using a different lock for

Re: Towards splitting SCHED_LOCK()

2020-01-14 Thread Mark Kettenis
> Date: Tue, 14 Jan 2020 10:34:05 +1100 > From: Jonathan Gray > > On Mon, Jan 13, 2020 at 05:02:12PM +0100, Martin Pieuchot wrote: > > I'm facing a lock ordering issue while profiling the scheduler which > > cannot be solved without using a different lock for the global sleepqueue > > and the

Re: Towards splitting SCHED_LOCK()

2020-01-14 Thread Martin Pieuchot
On 14/01/20(Tue) 10:34, Jonathan Gray wrote: > On Mon, Jan 13, 2020 at 05:02:12PM +0100, Martin Pieuchot wrote: > > I'm facing a lock ordering issue while profiling the scheduler which > > cannot be solved without using a different lock for the global sleepqueue > > and the runqueues. > > > > The

Re: Towards splitting SCHED_LOCK()

2020-01-14 Thread Jonathan Gray
On Mon, Jan 13, 2020 at 05:02:12PM +0100, Martin Pieuchot wrote: > I'm facing a lock ordering issue while profiling the scheduler which > cannot be solved without using a different lock for the global sleepqueue > and the runqueues. > > The SCHED_LOCK() currently protects both data structures as

Re: Towards splitting SCHED_LOCK()

2020-01-13 Thread Mark Kettenis
> Date: Mon, 13 Jan 2020 17:02:12 +0100 > From: Martin Pieuchot > > I'm facing a lock ordering issue while profiling the scheduler which > cannot be solved without using a different lock for the global sleepqueue > and the runqueues. > > The SCHED_LOCK() currently protects both data structures

Towards splitting SCHED_LOCK()

2020-01-13 Thread Martin Pieuchot
I'm facing a lock ordering issue while profiling the scheduler which cannot be solved without using a different lock for the global sleepqueue and the runqueues. The SCHED_LOCK() currently protects both data structures as well as the related fields in 'struct proc'. One of this fields is