Re: [x86/Kconfig] fac54e2bfb: kernel_BUG_at_arch/x86/mm/physaddr.c

From: Song Liu
Date: Wed Feb 16 2022 - 13:16:25 EST

Next message: Sui Jingfeng: "[PATCH v8 0/3] drm/lsdc: add drm driver for loongson display controller"
Previous message: nick . hawkins: "[PATCH] dt-bindings: soc: Add HPE GXP SOC binding"
In reply to: kernel test robot: "[x86/Kconfig] fac54e2bfb: kernel_BUG_at_arch/x86/mm/physaddr.c"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]

+ Nicholas Piggin

> On Feb 16, 2022, at 5:00 AM, kernel test robot <oliver.sang@xxxxxxxxx> wrote:
>
>
>
> Greeting,
>
> FYI, we noticed the following commit (built with gcc-9):
>
> commit: fac54e2bfb5be2b0bbf115fe80d45f59fd773048 ("x86/Kconfig: Select HAVE_ARCH_HUGE_VMALLOC with HAVE_ARCH_HUGE_VMAP")
> https://git.kernel.org/cgit/linux/kernel/git/next/linux-next.git master
>
> in testcase: boot
>
> on test machine: qemu-system-i386 -enable-kvm -cpu SandyBridge -smp 2 -m 4G
>
> caused below changes (please refer to attached dmesg/kmsg for entire log/backtrace):
>
>
>
> If you fix the issue, kindly add following tag
> Reported-by: kernel test robot <oliver.sang@xxxxxxxxx>
>
>
> [ 44.587744][ T1] kernel BUG at arch/x86/mm/physaddr.c:76!
> [ 44.589159][ T1] invalid opcode: 0000 [#1] SMP PTI
> [ 44.590151][ T1] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 5.16.0-11620-gfac54e2bfb5b #1
> [ 44.590151][ T1] EIP: __phys_addr (arch/x86/mm/physaddr.c:76 (discriminator 1))
> [ 44.590151][ T1] Code: 00 8d 76 00 83 05 20 92 8a c5 01 83 15 24 92 8a c5 00 89 f0 5b 5e 5d c3 8d 74 26 00 83 05 e0 91 8a c5 01 83 15 e4 91 8a c5 00 <0f> 0b 83 05 e8 91 8a c5 01 83 15 ec 91 8a c5 00 83 05 f0 91 8a c5
> All code
> ========
> 0: 00 8d 76 00 83 05 add %cl,0x5830076(%rbp)
> 6: 20 92 8a c5 01 83 and %dl,-0x7cfe3a76(%rdx)
> c: 15 24 92 8a c5 adc $0xc58a9224,%eax
> 11: 00 89 f0 5b 5e 5d add %cl,0x5d5e5bf0(%rcx)
> 17: c3 retq
> 18: 8d 74 26 00 lea 0x0(%rsi,%riz,1),%esi
> 1c: 83 05 e0 91 8a c5 01 addl $0x1,-0x3a756e20(%rip) # 0xffffffffc58a9203
> 23: 83 15 e4 91 8a c5 00 adcl $0x0,-0x3a756e1c(%rip) # 0xffffffffc58a920e
> 2a:* 0f 0b ud2 <-- trapping instruction
> 2c: 83 05 e8 91 8a c5 01 addl $0x1,-0x3a756e18(%rip) # 0xffffffffc58a921b
> 33: 83 15 ec 91 8a c5 00 adcl $0x0,-0x3a756e14(%rip) # 0xffffffffc58a9226
> 3a: 83 .byte 0x83
> 3b: 05 f0 91 8a c5 add $0xc58a91f0,%eax
>
> Code starting with the faulting instruction
> ===========================================
> 0: 0f 0b ud2
> 2: 83 05 e8 91 8a c5 01 addl $0x1,-0x3a756e18(%rip) # 0xffffffffc58a91f1
> 9: 83 15 ec 91 8a c5 00 adcl $0x0,-0x3a756e14(%rip) # 0xffffffffc58a91fc
> 10: 83 .byte 0x83
> 11: 05 f0 91 8a c5 add $0xc58a91f0,%eax
> [ 44.590151][ T1] EAX: 00000000 EBX: 00000000 ECX: 00000000 EDX: 00000000
> [ 44.590151][ T1] ESI: f7000000 EDI: f7000000 EBP: c6d85dd8 ESP: c6d85db4
> [ 44.590151][ T1] DS: 007b ES: 007b FS: 00d8 GS: 0000 SS: 0068 EFLAGS: 00010246
> [ 44.590151][ T1] CR0: 80050033 CR2: ff7ff000 CR3: 05854000 CR4: 000406b0
> [ 44.590151][ T1] DR0: 00000000 DR1: 00000000 DR2: 00000000 DR3: 00000000
> [ 44.590151][ T1] DR6: fffe0ff0 DR7: 00000400
> [ 44.590151][ T1] Call Trace:
> [ 44.590151][ T1] ? vmap_pages_range_noflush (mm/vmalloc.c:594)
> [ 44.590151][ T1] __vmalloc_area_node (mm/vmalloc.c:622 mm/vmalloc.c:2995)
> [ 44.590151][ T1] ? __get_vm_area_node+0xf5/0x200
> [ 44.590151][ T1] __vmalloc_node_range (mm/vmalloc.c:3108)
> [ 44.590151][ T1] __vmalloc_node (mm/vmalloc.c:3157)
> [ 44.590151][ T1] ? txInit.cold (fs/jfs/jfs_txnmgr.c:296)
> [ 44.590151][ T1] vmalloc (mm/vmalloc.c:3190)
> [ 44.590151][ T1] ? txInit.cold (fs/jfs/jfs_txnmgr.c:296)
> [ 44.590151][ T1] txInit.cold (fs/jfs/jfs_txnmgr.c:296)
> [ 44.590151][ T1] ? mempool_free (mm/mempool.c:509)
> [ 44.590151][ T1] ? mempool_create_node (mm/mempool.c:270)
> [ 44.590151][ T1] ? mempool_alloc_slab (mm/mempool.c:517)
> [ 44.590151][ T1] ? init_omfs_fs (fs/jfs/super.c:934)
> [ 44.590151][ T1] init_jfs_fs (fs/jfs/super.c:959)
> [ 44.590151][ T1] ? init_omfs_fs (fs/jfs/super.c:934)
> [ 44.590151][ T1] do_one_initcall (init/main.c:1297)
> [ 44.590151][ T1] ? rdinit_setup (init/main.c:1354)
> [ 44.590151][ T1] ? rcu_read_lock_sched_held (include/linux/lockdep.h:283 kernel/rcu/update.c:125)
> [ 44.590151][ T1] do_initcalls (init/main.c:1370 init/main.c:1386)
> [ 44.590151][ T1] kernel_init_freeable (init/main.c:1405 init/main.c:1610)
> [ 44.590151][ T1] ? rest_init (init/main.c:1491)
> [ 44.590151][ T1] kernel_init (init/main.c:1499)
> [ 44.590151][ T1] ret_from_fork (arch/x86/entry/entry_32.S:772)
> [ 44.590151][ T1] Modules linked in:
> [ 44.630667][ T1] ---[ end trace 0000000000000000 ]---
> [ 44.631743][ T1] EIP: __phys_addr (arch/x86/mm/physaddr.c:76 (discriminator 1))
> [ 44.632726][ T1] Code: 00 8d 76 00 83 05 20 92 8a c5 01 83 15 24 92 8a c5 00 89 f0 5b 5e 5d c3 8d 74 26 00 83 05 e0 91 8a c5 01 83 15 e4 91 8a c5 00 <0f> 0b 83 05 e8 91 8a c5 01 83 15 ec 91 8a c5 00 83 05 f0 91 8a c5
> All code
> ========
> 0: 00 8d 76 00 83 05 add %cl,0x5830076(%rbp)
> 6: 20 92 8a c5 01 83 and %dl,-0x7cfe3a76(%rdx)
> c: 15 24 92 8a c5 adc $0xc58a9224,%eax
> 11: 00 89 f0 5b 5e 5d add %cl,0x5d5e5bf0(%rcx)
> 17: c3 retq
> 18: 8d 74 26 00 lea 0x0(%rsi,%riz,1),%esi
> 1c: 83 05 e0 91 8a c5 01 addl $0x1,-0x3a756e20(%rip) # 0xffffffffc58a9203
> 23: 83 15 e4 91 8a c5 00 adcl $0x0,-0x3a756e1c(%rip) # 0xffffffffc58a920e
> 2a:* 0f 0b ud2 <-- trapping instruction
> 2c: 83 05 e8 91 8a c5 01 addl $0x1,-0x3a756e18(%rip) # 0xffffffffc58a921b
> 33: 83 15 ec 91 8a c5 00 adcl $0x0,-0x3a756e14(%rip) # 0xffffffffc58a9226
> 3a: 83 .byte 0x83
> 3b: 05 f0 91 8a c5 add $0xc58a91f0,%eax
>
> Code starting with the faulting instruction
> ===========================================
> 0: 0f 0b ud2
> 2: 83 05 e8 91 8a c5 01 addl $0x1,-0x3a756e18(%rip) # 0xffffffffc58a91f1
> 9: 83 15 ec 91 8a c5 00 adcl $0x0,-0x3a756e14(%rip) # 0xffffffffc58a91fc
> 10: 83 .byte 0x83
> 11: 05 f0 91 8a c5 add $0xc58a91f0,%eax

Hi Nicholas,

I guess you know the HAVE_ARCH_HUGE_VMALLOC best.

In the commit

fac54e2bfb5b ("x86/Kconfig: Select HAVE_ARCH_HUGE_VMALLOC with HAVE_ARCH_HUGE_VMAP")

I was trying to enable huge vmalloc for x86. This report shows that
it doesn't really work for 32-bit x86.

I also confirmed the following change fix it by 32-bit x86 (by
disabling huge vmalloc).

Do you think this is something we can easily fix for 32-bit x86?
If not, I guess we should just go ahead disable it for 32-bit x86.

Thanks,
Song

diff --git i/arch/x86/Kconfig w/arch/x86/Kconfig
index 995f2dc28631..0d08c36dfff1 100644
--- i/arch/x86/Kconfig
+++ w/arch/x86/Kconfig
@@ -158,7 +158,7 @@ config X86
select HAVE_ALIGNED_STRUCT_PAGE if SLUB
select HAVE_ARCH_AUDITSYSCALL
select HAVE_ARCH_HUGE_VMAP if X86_64 || X86_PAE
- select HAVE_ARCH_HUGE_VMALLOC if HAVE_ARCH_HUGE_VMAP
+ select HAVE_ARCH_HUGE_VMALLOC if X86_64
select HAVE_ARCH_JUMP_LABEL
select HAVE_ARCH_JUMP_LABEL_RELATIVE
select HAVE_ARCH_KASAN if X86_64

>
>
> To reproduce:
>
> # build kernel
> cd linux
> cp config-5.16.0-11620-gfac54e2bfb5b .config
> make HOSTCC=gcc-9 CC=gcc-9 ARCH=i386 olddefconfig prepare modules_prepare bzImage modules
> make HOSTCC=gcc-9 CC=gcc-9 ARCH=i386 INSTALL_MOD_PATH=<mod-install-dir> modules_install
> cd <mod-install-dir>
> find lib/ | cpio -o -H newc --quiet | gzip > modules.cgz
>
>
> git clone https://github.com/intel/lkp-tests.git
> cd lkp-tests
> bin/lkp qemu -k <bzImage> -m modules.cgz job-script # job-script is attached in this email
>
> # if come across any failure that blocks the test,
> # please remove ~/.lkp and /lkp dir to run from a clean state.
>
>
>
> ---
> 0DAY/LKP+ Test Infrastructure Open Source Technology Center
> https://lists.01.org/hyperkitty/list/lkp@xxxxxxxxxxxx Intel Corporation
>
> Thanks,
> Oliver Sang
>
> <config-5.16.0-11620-gfac54e2bfb5b><job-script.txt><dmesg.xz>

Next message: Sui Jingfeng: "[PATCH v8 0/3] drm/lsdc: add drm driver for loongson display controller"
Previous message: nick . hawkins: "[PATCH] dt-bindings: soc: Add HPE GXP SOC binding"
In reply to: kernel test robot: "[x86/Kconfig] fac54e2bfb: kernel_BUG_at_arch/x86/mm/physaddr.c"
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]