Failure to boot w/CONFIG_PAX_SIZE_OVERFLOW - no output

[x14sg1]

Hello,

I have a machine (32 bit kernel; gcc 5.3.0) that will not boot. I can select a kernel from grub but then the screen goes blank and stays that way. I have tracked it down to CONFIG_PAX_SIZE_OVERFLOW_DISABLE_KILL (until it was removed in 4.2.6) and CONFIG_PAX_SIZE_OVERFLOW (from then on).

With the CONFIG option commented out, I can boot grsecurity kernels from 4.2.6 to 4.3.5.

The "KILL" is happening before netconsole can show me anything (I verified I do get output without the CONFIG option).

As I get nothing on the screen or from netconsole, do you have any suggestions to troubleshoot this?

I have lost access to the site where I would have posted my config/kernel . Any suggestions for me to use now? Or I can email them to you.

Thanks

[PaX Team]

did you try to boot with pax_size_overflow_report_only?

[x14sg1]

Thanks for the reply.

I was not aware of that but will try that next and let you know.

[x14sg1]

Here is the output ...

[ 0.000000] No NUMA configuration found
[ 0.000000] Faking a node at [mem 0x0000000000000000-0x000000012fffffff]
[ 0.000000] PAX: size overflow detected in function x86_numa_init arch/x86/mm/numa.c:691 cicus.311_502 max, count: 45, decl: memblock_clear_hotplug; num: 2; context: fndecl;
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.3.5-grsec-smp #201601311611
[ 0.000000] Hardware name: Dell Inc. Inspiron 530/0FM586, BIOS 1.0.13 03/20/2008
[ 0.000000] 00000000 00000000 c3c01e58 002d59ad c394f30a c3c01e80 0014d7ae c3972370
[ 0.000000] c394f35d c394f30a 000002b3 c3957ec4 c400c0a0 58ad3a28 43e9b7cf c3c01f1c
[ 0.000000] 02e3348c c3957ec4 ea5fa25e 69634953 00000000 8c98b36b 00000000 60e4e858
[ 0.000000] Call Trace:
[ 0.000000] [<002d59ad>] dump_stack+0x41/0x54
[ 0.000000] [<0014d7ae>] report_size_overflow+0x4e/0x60
[ 0.000000] [<02e3348c>] x86_numa_init+0x459/0xdcb
[ 0.000000] [<02e5c435>] ? early_acpi_os_unmap_memory+0x21/0x65
[ 0.000000] [<00003000>] ? nmi_print_seq+0x1a0/0x410
[ 0.000000] [<000464b0>] ? __native_set_fixmap+0x20/0x30
[ 0.000000] [<02e5b61c>] ? acpi_table_parse_entries+0xf8/0x14f
[ 0.000000] [<000464fc>] ? native_set_fixmap+0x3c/0x50
[ 0.000000] [<0003d3e4>] ? default_get_apic_id+0x14/0x40
[ 0.000000] [<02e33edb>] initmem_init+0x1e/0x115
[ 0.000000] [<02e0f50e>] setup_arch+0x12c0/0x151a
[ 0.000000] [<000964f6>] ? vprintk_emit+0x1c6/0x540
[ 0.000000] [<000969e2>] ? vprintk_default+0x12/0x20
[ 0.000000] [<02e08160>] start_kernel+0x84/0x4ef
[ 0.000000] [<000ed5c1>] ? printk+0x12/0x14
[ 0.000000] [<02e08160>] ? start_kernel+0x84/0x4ef
[ 0.000000] [<0009e800>] ? msi_create_irq_domain+0x10/0x180
[ 0.000000] [<02e0732e>] i386_start_kernel+0xf6/0xfd
[ 0.000000] [<00040800>] ? vmalloc_fault+0xe0/0x3a0
[ 0.000000] [<0008f800>] ? proc_sched_show_task+0x10/0x8d0
[ 0.000000] PAX: size overflow detected in function x86_numa_init arch/x86/mm/numa.c:507 cicus.312_503 max, count: 47, decl: memblock_clear_hotplug; num: 2; context: fndecl;
[ 0.000000] CPU: 0 PID: 0 Comm: swapper Not tainted 4.3.5-grsec-smp #201601311611
[ 0.000000] Hardware name: Dell Inc. Inspiron 530/0FM586, BIOS 1.0.13 03/20/2008
[ 0.000000] 00000000 00000000 c3c01e58 002d59ad c394f30a c3c01e80 0014d7ae c3972370
[ 0.000000] c394f35d c394f30a 000001fb c3957f1c c400c0a0 2a7fb137 4436bc12 c3c01f1c
[ 0.000000] 02e3350e c3957f1c ea5fa25e 69634953 00000000 8c98b36b 00000000 60e4e858
[ 0.000000] Call Trace:
[ 0.000000] [<002d59ad>] dump_stack+0x41/0x54
[ 0.000000] [<0014d7ae>] report_size_overflow+0x4e/0x60
[ 0.000000] [<02e3350e>] x86_numa_init+0x4db/0xdcb
[ 0.000000] [<02e5c435>] ? early_acpi_os_unmap_memory+0x21/0x65
[ 0.000000] [<00003000>] ? nmi_print_seq+0x1a0/0x410
[ 0.000000] [<000464b0>] ? __native_set_fixmap+0x20/0x30
[ 0.000000] [<02e5b61c>] ? acpi_table_parse_entries+0xf8/0x14f
[ 0.000000] [<000464fc>] ? native_set_fixmap+0x3c/0x50
[ 0.000000] [<02e33edb>] initmem_init+0x1e/0x115
[ 0.000000] [<02e0f50e>] setup_arch+0x12c0/0x151a
[ 0.000000] [<000964f6>] ? vprintk_emit+0x1c6/0x540
[ 0.000000] [<000969e2>] ? vprintk_default+0x12/0x20
[ 0.000000] [<02e08160>] start_kernel+0x84/0x4ef
[ 0.000000] [<000ed5c1>] ? printk+0x12/0x14
[ 0.000000] [<02e08160>] ? start_kernel+0x84/0x4ef
[ 0.000000] [<0009e800>] ? msi_create_irq_domain+0x10/0x180
[ 0.000000] [<02e0732e>] i386_start_kernel+0xf6/0xfd
[ 0.000000] [<00040800>] ? vmalloc_fault+0xe0/0x3a0
[ 0.000000] [<0008f800>] ? proc_sched_show_task+0x10/0x8d0

[PaX Team]

can you post your config or at least tell me which of CONFIG_ACPI_NUMA and CONFIG_AMD_NUMA is enabled? and i guess Emese will want to get the output files of make arch/x86/mm/numa.o EXTRA_CFLAGS="-fdump-tree-all -fdump-ipa-all" too (email is fine).

[PaX Team]

also try the following patch that should fix an integer truncation (have you got more than 4GB of memory?):

Code: Select all

--- a/arch/x86/mm/numa.c   2015-11-03 01:48:55.603331575 +0100
+++ b/arch/x86/mm/numa.c  2016-02-03 00:53:12.574516515 +0100
@@ -469,7 +469,7 @@
 {
        int i, nid;
        nodemask_t numa_kernel_nodes = NODE_MASK_NONE;
-       unsigned long start, end;
+       u64 start, end;
        struct memblock_region *r;

        /*


[x14sg1]

Your patch solved my problem after changing "u64 long" to "u64".

I assume you no linger need the .config or NUMA files ...

Thanks

[PaX Team]

yeah sorry about the typo, already corrected it in the post and no, don't worry about the earlier request if it all works now.

PS: you could submit this upstream as it's a real bug there.

[x14sg1]

I have reported the bug.

[PaX Team]

can you post a tracking URL please?

[x14sg1]

I have not received a reply yet. Will that have the tracking number in it?

[PaX Team]

where did you report it? if it's lkml or the kernel bugzilla then you have your URL .

[x14sg1]

I sent it via email to the output from " perl scripts/get_maintainer.pl -f arch/x86/mm/numa.c" per the REPORTING-BUGS in the top level kernel directory.