Re: linux-next: boot failure after merge of the akpm tree

[Nicholas Piggin]

Stephen Rothwell's on May 30, 2019 4:17 pm:
> Hi all,
> 
> My qemu boot (PowerPC le guest on PowerPC le host, with and without kvm,
> using a kernel built with powerpc_pseries_le_defconfig) oopses during boot
> like this:
> 
> -----------------------------------------------------------------------------
> numa: Node 0 CPUs: 0
> Using standard scheduler topology
> devtmpfs: initialized
> clocksource: jiffies: mask: 0xffffffff max_cycles: 0xffffffff, max_idle_ns: 19112604462750000 ns
> futex hash table entries: 256 (order: -1, 32768 bytes)
> ------------[ cut here ]------------
> kernel BUG at mm/vmalloc.c:472!
> Oops: Exception in kernel mode, sig: 5 [#1]
> LE PAGE_SIZE=64K MMU=Hash SMP NR_CPUS=2048 NUMA pSeries
> Modules linked in:
> CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.2.0-rc2 #2
> NIP:  c000000000369b18 LR: c000000000369c74 CTR: c000000000176e30
> REGS: c00000007e6636e0 TRAP: 0700   Not tainted  (5.2.0-rc2)
> MSR:  8000000002029033 <SF,VEC,EE,ME,IR,DR,RI,LE>  CR: 24024882  XER: 20000000
> CFAR: c000000000369c78 IRQMASK: 0 
> GPR00: c000000000369c74 c00000007e663970 c00000000119c100 0000000000000001 
> GPR04: 000000007ec20000 00000001f4fe19cb 00000001f5398c84 c000000001380000 
> GPR08: 0000000000000000 0000000000000001 0000000000000001 00000000000002b2 
> GPR12: 0000000000004000 c000000001380000 c000000000010fc0 0000000000000001 
> GPR16: 0000000000010000 800000000000018e c000000000df9988 0000000000000000 
> GPR20: 0000000000010000 0000000000002dc2 0000000000000dc0 0000000000000022 
> GPR24: c00000007e2204c0 0000000000000dc2 0000000000010000 c00a000000000000 
> GPR28: c008000000000000 0000000000010000 ffffffffffffffff 0000000000000dc0 
> NIP [c000000000369b18] __vmalloc_node_range+0x1f8/0x410
> LR [c000000000369c74] __vmalloc_node_range+0x354/0x410
> Call Trace:
> [c00000007e663970] [c000000000369c74] __vmalloc_node_range+0x354/0x410 (unreliable)
> [c00000007e663a70] [c000000000369d80] __vmalloc+0x50/0x60
> [c00000007e663ae0] [c000000000299a98] bpf_prog_alloc_no_stats+0x58/0x120
> [c00000007e663b20] [c000000000299b90] bpf_prog_alloc+0x30/0xe0
> [c00000007e663b60] [c000000000a49dd8] bpf_prog_create+0x68/0x100
> [c00000007e663ba0] [c000000000f4f2a8] ptp_classifier_init+0x4c/0x80
> [c00000007e663be0] [c000000000f4b9e8] sock_init+0xe0/0x100
> [c00000007e663c10] [c000000000010b60] do_one_initcall+0x60/0x2c0
> [c00000007e663ce0] [c000000000ee45b0] kernel_init_freeable+0x37c/0x478
> [c00000007e663db0] [c000000000010fe4] kernel_init+0x2c/0x148
> [c00000007e663e20] [c00000000000c0cc] ret_from_kernel_thread+0x5c/0x70
> Instruction dump:
> 60000000 2c230000 418200dc e9580020 79e91f24 7c6a492a 40920170 8138002c 
> 394f0001 794f0020 7f895040 419dffbc <0fe00000> 60000000 3f400001 4bfffedc 
> ---[ end trace 49ed8f97d467e164 ]---
> 
> Kernel panic - not syncing: Attempted to kill init! exitcode=0x00000005
> -----------------------------------------------------------------------------
> 
> The BUG is:
> 
> 	BUG_ON(page_shift != PAGE_SIZE);
> 
> in the !CONFIG_HAVE_ARCH_HUGE_VMAP version of vmap_hpages_range().
> 
> I am guessing this is something to do with the vmalloc changes in Andrew's
> patches (or it could be the fixup I did to Nick's patch).
> 
> I have reverted
> 
>   c353e2997976 ("mm/vmalloc: hugepage vmalloc mappings")
>   a826492f28d9 ("mm: move ioremap page table mapping function to mm/")
> 
> (and my fix up) for today and things seem to work (if only because the
> BUG() has been removed :-)).

Good to know, maybe I didn't test powerpc without later enabling 
patches...

The series also has a compile bug on ARM I have to work out, so
yeah drop those for now, I'll post a v2. The large system map patches
that I posted in that series can stay I think.

Thanks,
Nick