Re: Regression on ARMs in next-20170531

[Andrew Morton]

On Tue, 6 Jun 2017 10:36:48 -0400 Johannes Weiner <hannes@xxxxxxxxxxx> wrote:

> On Tue, Jun 06, 2017 at 05:30:10AM -0700, Tony Lindgren wrote:
> > PC is at __mod_node_page_state+0x2c/0xc8
> > LR is at __per_cpu_offset+0x0/0x8
> > pc : [<c0280078>]    lr : [<c0d07d6c>]    psr: 200001d3
> > sp : c0d01eec  ip : 00000000  fp : 00000001
> > r10: c0c7cf68  r9 : 00008000  r8 : 00000000
> > r7 : 00000001  r6 : 00000006  r5 : 2ea2d000  r4 : 00000007
> > r3 : 00000007  r2 : 00000001  r1 : 00000006  r0 : c0dc1fc0
> > Flags: nzCv  IRQs off  FIQs off  Mode SVC_32  ISA ARM  Segment none
> > Control: 10c5387d  Table: 8000404a  DAC: 00000051
> > Process swapper (pid: 0, stack limit = 0xc0d00218)
> > Stack: (0xc0d01eec to 0xc0d02000)
> > 1ee0:                            400001d3 c0dc1fc0 c028018c 00000001 c1599440
> > 1f00: c0d58834 efd83000 00000000 c02af214 01000000 c157a890 00002000 00008000
> > 1f20: 00000001 00000001 00008000 c02aeb4c 00000000 00008000 c0d58834 00008000
> > 1f40: 01008000 c0c23a88 c0d58834 c1580034 400001d3 c02afa9c 00000000 c086b230
> > 1f60: c0d58834 000000c0 01000000 c157a78c c0abe0fc 00000080 00002000 c0dd4000
> > 1f80: efffec40 c0c55a48 00000000 c0c23a88 c157a78c c0c5be48 c0c5bde8 c157a890
> > 1fa0: c0dd4000 c0c25a9c 00000000 ffffffff c0dd4000 c0d07940 c0dd4000 c0c00abc
> > 1fc0: ffffffff ffffffff 00000000 c0c006a0 00000000 c0c55a48 c0dd4214 c0d07958
> > 1fe0: c0c55a44 c0d0cae4 8000406a 411fc093 00000000 8000807c 00000000 00000000
> > [<c0280078>] (__mod_node_page_state) from [<c028018c>] (mod_node_page_state+0x2c/0x4c)
> > [<c028018c>] (mod_node_page_state) from [<c02af214>] (cache_alloc_refill+0x654/0x898)
> > [<c02af214>] (cache_alloc_refill) from [<c02afa9c>] (kmem_cache_alloc+0x2d4/0x364)
> > [<c02afa9c>] (kmem_cache_alloc) from [<c0c23a88>] (create_kmalloc_cache+0x20/0x8c)
> > [<c0c23a88>] (create_kmalloc_cache) from [<c0c25a9c>] (kmem_cache_init+0xac/0x11c)
> > [<c0c25a9c>] (kmem_cache_init) from [<c0c00abc>] (start_kernel+0x1b8/0x3d8)
> 
> That's the one Russell analyzed and I misinterpreted. We put a fix
> into -next to initialize pgdat->per_cpu_nodestats in time for slab
> initialization during boot.
> 
> Is today's -next working again?

I'll be even less than usually functional for the next week, so please
cc Stephen on any -next hotfixes.