Re: Linux-rt fails to boot [any release past 3.12.5-rt7] on all of my machines

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

* jordan | 2014-02-05 20:54:59 [-0500]:

>Hey list,
Hi jordan,

>I've been meaning to post, but alas, I am working three jobs right now
>(just took on more work) and haven't had a whole lot of free time to
>investigate my boot problems on linux-rt (-rt9, -rt10, -rt11, -rt13).
>
>On 3.12.8-rt11 - i was able to get the kernel to boot via reverting two
>patches, from the split queue;
>
>'timer-Raise-softirq-if-there-s-irq_work.patch'
>'timers-do-not-raise-softirq-unconditionally.patch'

It was reported to the list that
timers-do-not-raise-softirq-unconditionally.patch sometimes stalled the
machine. The timer-Raise-softirq-if-there-s-irq_work.patch should have
fixes one common case and -rt13 should have coverted them all.

>(i haven't tried reverting on -rt13, as i haven't had the time to do that).

It should work on -rt13 without reverting them.

>however, while this was enough to boot - my system was unstable. thus i
>downgraded my kernel back to trusty 3.12.5-rt7 - which works flawlessly for
>me. Next, I saw the -rt13 release, thinking maybe the boot issue was
>finally solved [according to the change log], but no - it fails just as
>before... I tried to add debug to grub's commandline but in the end, it
>wasn't really helpful. [the last thing shown is something about PnP and
>then ACPI];
>
>[    0.255329] pnp: PnP ACPI: found 10 devices
>[    0.255330] ACPI: bus type PNP unregistered
>
>(which i would normally see anyway.)

this looks like a missing event (like timer interrupt or RCU wakeup and
"timers-do-not-raise-softi…" might be a source of this but then it
should be fixed).

>I might have some time over the weekend (hopefully), to take a closer look
>- but i  thought i would at least notify the list ~ as i saw another user
>have the same booting problems and he has the same CPU as i do; AMD
>Phenom 965 X4 [not exactly exotic H/W and every RT release has worked
>great on my setup, until 3.12.x-rt8. Now, nothing works properly.
>[note: vanilla 3.12.x kernels boot just fine].
>
>anyway, if anyone has some suggestions - that i could follow up on over the
>weekend that would be fantastic.

At [0] is my quilt queue containing tags for releases between the
working and non-working release. As mentioned before,
timers-do-not-raise-softirq-unconditionally.patch will cause stalls
which is finally fixed by
timer-rt-Always-raise-the-softirq-if-there-s-irq_wor.patch.

Looking at my queue between -rt8 and now there is nothing big except for
the "raise softirq unconditional" patch.
If you decide to bisect the queue to find the offending post -rt8 patch
that causes trouble, please note that take these patches
| timers-do-not-raise-softirq-unconditionally.patch
| timer-Raise-softirq-if-there-s-irq_work.patch
| timer-rt-Always-raise-the-softirq-if-there-s-irq_wor.patch
should be applied together since the later two are fixes for the former.

[0] git://git.kernel.org/pub/scm/linux/kernel/git/bigeasy/rt-devel.git

>cheers
>
>Jordan

Sebastian
--
To unsubscribe from this list: send the line "unsubscribe linux-rt-users" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html