Re: [lkp] [mm] 122708b1b9: PANIC: early exception

From: Catalin Marinas <catalin.marinas@arm.com>
To: Andrew Morton <akpm@linux-foundation.org>
Cc: kernel test robot <xiaolong.ye@intel.com>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	Vignesh R <vigneshr@ti.com>,
	lkp@01.org
Subject: Re: [lkp] [mm]  122708b1b9: PANIC: early exception
Date: Wed, 17 Aug 2016 21:36:19 +0100	[thread overview]
Message-ID: <20160817203618.GA16393@MBP.local> (raw)
In-Reply-To: <20160817121808.bf31e27382554bf532368c38@linux-foundation.org>

On Wed, Aug 17, 2016 at 12:18:08PM -0700, Andrew Morton wrote:
> On Wed, 17 Aug 2016 17:10:28 +0100 Catalin Marinas <catalin.marinas@arm.com> wrote:
> > On Wed, Aug 17, 2016 at 11:51:41PM +0800, kernel test robot wrote:
> > > FYI, we noticed the following commit:
> > > 
> > > https://github.com/0day-ci/linux Catalin-Marinas/mm-kmemleak-Avoid-using-__va-on-addresses-that-don-t-have-a-lowmem-mapping/20160816-232733
> > > commit 122708b1b91eb3d253baf86a263ead0f1f5cac78 ("mm: kmemleak: Avoid using __va() on addresses that don't have a lowmem mapping")
> > > 
> > > in testcase: boot
> > > 
> > > on test machine: 1 threads qemu-system-i386 -enable-kvm with 320M memory
> > > 
> > > caused below changes:
> > > 
> > > +--------------------------------+------------+------------+
> > > |                                | 304bec1b1d | 122708b1b9 |
> > > +--------------------------------+------------+------------+
> > > | boot_successes                 | 3          | 0          |
> > > | boot_failures                  | 5          | 8          |
> > > | invoked_oom-killer:gfp_mask=0x | 1          |            |
> > > | Mem-Info                       | 1          |            |
> > > | BUG:kernel_test_crashed        | 4          |            |
> > > | PANIC:early_exception          | 0          | 8          |
> > > | EIP_is_at__phys_addr           | 0          | 8          |
> > > | BUG:kernel_hang_in_boot_stage  | 0          | 2          |
> > > | BUG:kernel_boot_hang           | 0          | 6          |
> > > +--------------------------------+------------+------------+
> > 
> > Please disregard this patch. I posted v2 here:
> > 
> > http://lkml.kernel.org/g/1471426130-21330-1-git-send-email-catalin.marinas@arm.com
> > 
> > (and I'm eager to see the kbuild/kernel test robot results ;))
> 
> I don't see how the v1->v2 changes could fix a panic?

This particular panic is avoided (rather than fixed) in v2 because the
config used above has kmemleak disabled, hence there is no
__pa(high_memory) call.

But you are right, it is likely to trigger once kmemleak is enabled, I
think because __pa(high_memory) use isn't valid. I need to reproduce it
tomorrow (UK time) but a workaround is to test against __pa(high_memory
- 1) or (max_low_pfn << PAGE_SHIFT).

-- 
Catalin