12:46 ` Peter Zijlstra 0 siblings, 2 replies 15+ messages in threadįrom: Paul E. > + schedule_work_on(smp_processor_id(), &sched_clock_work) > smp_mb() /* matches sched_clock_init_late() */ > diff -git a/kernel/sched/clock.c b/kernel/sched/clock.c > 1 file changed, 1 insertion(+), 1 deletion(-) Thank you! I will give it a shot after the other tests complete. On Thu, at 04:24:20PM +0100, Peter Zijlstra wrote: To: Peter Zijlstra +Cc: linux-kernel, mingo, fweisbec McKenney 0 siblings, 1 reply 15+ messages in threadįrom: Paul E. + schedule_work_on(smp_processor_id(), &sched_clock_work) Smp_mb() /* matches sched_clock_init_late() */ Lightly testedġ file changed, 1 insertion(+), 1 deletion(-) > and to the uses of the pointer that it returns. > This commit therefore disables preemption across this call to this_scd() > by default and the new this_scd() function accesses per-CPU variables. > This happens because workqueue handlers run with preemption enabled > ? call_usermodehelper_exec_async+0x12a/0x130 > Workqueue: events _clear_sched_clock_stable > Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs > caller is debug_smp_processor_id+0x17/0x20 > BUG: using smp_processor_id() in preemptible code: kworker/3:1/49 > The v4.11-rc1 kernel emits the following splat in some configurations: McKenney +Cc: linux-kernel, mingo, fweisbec McKenney 0 siblings, 1 reply 15+ messages in threadįrom: Peter Zijlstra 15:24 UTC ( / raw) 21:53 clock: Fix smp_processor_id() in preemptible bug Paul E. * Re: clock: Fix smp_processor_id() in preemptible bug Printk(KERN_INFO "sched_clock: Marking unstable (%lld, %lld)tick_gtod, gtod_offset, Gtod_offset = (scd->tick_raw + raw_offset) - (scd->tick_gtod) * Attempt to make the stable->unstable transition -154,7 +154,10 static void _clear_sched_clock_stable(struct work_struct *work) struct sched_clock_data *scd = this_scd() Static void _clear_sched_clock_stable(struct work_struct *work) ? call_usermodehelper_exec_async+0x12a/0x130 Workqueue: events _clear_sched_clock_stable Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs caller is debug_smp_processor_id+0x17/0x20 BUG: using smp_processor_id() in preemptible code: kworker/3:1/49 The v4.11-rc1 kernel emits the following splat in some configurations: To: linux-kernel +Cc: mingo, peterz, fweisbec 15:24 ` Peter Zijlstra 0 siblings, 1 reply 15+ messages in threadįrom: Paul E. I think Stanislav even roams these forums once in a while.Clock: Fix smp_processor_id() in preemptible bug LKML Archive on help / color / mirror / Atom feed * clock: Fix smp_processor_id() in preemptible bug 21:53 Paul E. I have contributed to a lot of the code, including some BIOS modifications, but Stanislav and Volker are the ones to ask. However, I am not an active member of the Bochs team, so don't take my word for it. It is and will be a massive re-write unfortunately. This is a challenge that I don't think many wish the start, though I have looked into it myself and gave up after hours (and hours) of work. To get it to work with UEFI, a number of changes would have to be done to both the main core of Bochs as well as the Legacy BIOS code to allow both types. Bochs was written for the Legacy type and had no idea there would be a new type of BIOS in later years.Ĭurrently, the Bochs code initializes parts of the virtual machine before the BIOS is called and also relies upon certain aspects of a Legacy BIOS. There is quite a bit of work that would need to be done to Bochs to get it to work with both a Legacy BIOS and an UEFI firmware. Would be glad if somebody would join and explain. We have no clue what even required to be done to support OVMF in Bochs It is actually even worse than you think.īochs has only two active developers which do not and absolute lack of interest from the community - looks like everybody prefer to contribute QEMU.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |