Suspend/resume on i915 stop working between r278348 and r278741.

Gleb Smirnoff glebius at FreeBSD.org
Sun Feb 15 18:53:19 UTC 2015


On Sun, Feb 15, 2015 at 07:45:44PM +0100, Ranjan1018 . wrote:
R> > On Sun, Feb 15, 2015 at 03:37:17PM +0100, Ranjan1018 . wrote:
R> > > Hi,
R> > > the bug was introduced in r278473. Suspend/resume works again if I
R> > disable
R> > > the x2APIC support with hw.x2apic_enable=0 in /boot/loader.conf, tested
R> > in
R> > > r278473 and r278741.
R> >
R> > So how it is related to i915 ?
R> >
R> > Sorry, just my guess.
R> 
R> 
R> > Try the patch.
R> >
R> > diff --git a/sys/x86/x86/local_apic.c b/sys/x86/x86/local_apic.c
R> > index 1809fa6..c8ea35e 100644
R> > --- a/sys/x86/x86/local_apic.c
R> > +++ b/sys/x86/x86/local_apic.c
R> > @@ -541,6 +541,9 @@ native_lapic_setup(int boot)
R> >
R> >         saveintr = intr_disable();
R> >
R> > +       if (!boot && x2apic_mode)
R> > +               native_lapic_enable_x2apic();
R> > +
R> >         la = &lapics[lapic_id()];
R> >         KASSERT(la->la_present, ("missing APIC structure"));
R> >         maxlvt = (lapic_read32(LAPIC_VERSION) & APIC_VER_MAXLVT) >>
R> > MAXLVTSHIFT;
R> >
R> 
R> Unfortunately does not work for me. Tested with r278803.

Me neither.

Also, I noticed that in my case the suspend sequence is also behaving
differently from pre-x2apic kernel.

With new kernel, suspend sequence doesn't switch from X to vt(4), screen
switches to black directly from X. So, probably the resume isn't broken
at all, instead suspend is broken: it puts machine in a state that can't
be resumed.

-- 
Totus tuus, Glebius.


More information about the freebsd-current mailing list