Re: PC Engines APU panics on boot since 4.2.4

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



On Sun, 1 Nov 2015, Greg KH wrote:
> On Sun, Nov 01, 2015 at 07:26:28PM +0100, Marc Haber wrote:
> > Hi,
> > 
> > On Sat, Oct 31, 2015 at 03:23:45PM -0700, Greg KH wrote:
> > > On Sat, Oct 31, 2015 at 10:15:51PM +0100, Marc Haber wrote:
> > > > I run a PC Engines APU in my basement for local infrastructure with
> > > > KVM. Kernels 4.2.4 and 4.2.5 panic on bootup pretty fast after the
> > > > usual "x86: Booted up 1 node, 2 CPUs", "smpboot: Total of 2 processors
> > > > activated (3999.89 BogoMIPS)" message. The message I would expect
> > > > instead of the kernel panic would be "devtmpfs: initialized", as one
> > > > can see on a successful 4.2.3 boot with identical kernel configuration.
> > > 
> > > If you use 'git bisect' between 4.2.3 and 4.2.5, can you find the
> > > problem commit for us to help track this down?
> > 
> > The winner is ...
> > [21f751a634f3b5eaf916f940e47f327fbc6e9727] x86/ioapic: Force affinity setting in setup_ioapic_dest()
> > 		   
> > After rolling this back in 4.2.5, the APU now runs fine.
> 
> Ok, this matches up with commit 21f751a634f3b5eaf916f940e47f327fbc6e9727
> in Linus's tree.
> 
> Thomas and Mika, any thoughts about this?
> 
> Marc, do you have this same problem with Linus's tree at the moment as
> well?

I just sent a pull request to Linus with a fix for this. Patch below.

Thanks,

	tglx

---
commit ababae44108b0e94b58eef6cb5bd830bd040a47f
Author: Werner Pawlitschko <werner.pawlitschko@xxxxxxxx>
Date:   Tue Oct 27 09:08:04 2015 +0900

    x86/ioapic: Prevent NULL pointer dereference in setup_ioapic_dest()
    
    Commit 4857c91f0d19 changed the way how irq affinity is setup in
    setup_ioapic_dest() from using the core helper function to
    unconditionally calling the irq_set_affinity() callback of the
    underlying irq chip.
    
    That results in a NULL pointer dereference for the rare case where the
    underlying irq chip is lapic_chip which has no irq_set_affinity()
    callback. lapic_chip is occasionally used for the timer interrupt (irq
    0).
    
    The fix is simple: Check the availability of the callback instead of
    calling it unconditionally.
    
    Fixes: 4857c91f0d19 "x86/ioapic: Force affinity setting in setup_ioapic_dest()"
    Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx>
    Cc: stable@xxxxxxxxxxxxxxx

diff --git a/arch/x86/kernel/apic/io_apic.c b/arch/x86/kernel/apic/io_apic.c
index bb6bfc01cb82..4f2821527014 100644
--- a/arch/x86/kernel/apic/io_apic.c
+++ b/arch/x86/kernel/apic/io_apic.c
@@ -2547,7 +2547,9 @@ void __init setup_ioapic_dest(void)
 			mask = apic->target_cpus();
 
 		chip = irq_data_get_irq_chip(idata);
-		chip->irq_set_affinity(idata, mask, false);
+		/* Might be lapic_chip for irq 0 */
+		if (chip->irq_set_affinity)
+			chip->irq_set_affinity(idata, mask, false);
 	}
 }
 #endif
--
To unsubscribe from this list: send the line "unsubscribe stable" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Kernel]     [Kernel Development Newbies]     [Linux USB Devel]     [Video for Linux]     [Linux Audio Users]     [Yosemite Hiking]     [Linux Kernel]     [Linux SCSI]