Messages in this thread Patch in this message | | | Date | Mon, 24 Apr 2017 15:58:05 +0900 | From | Satoru Takeuchi <> | Subject | [BUG] x86: failed to boot a kernel on a Ryzen machine |
| |
Recently I bought a new Ryzen machine. When I tried to test v4.11-rc8 on it, it failed to boot with the following panic log.
``` ... [ 0.227720] raid6: sse2x1 gen() 7985 MB/s [ 0.295709] raid6: sse2x1 xor() 8181 MB/s [ 0.363706] raid6: sse2x2 gen() 17531 MB/s [ 0.431699] raid6: sse2x2 xor() 11098 MB/s [ 0.499693] raid6: sse2x4 gen() 18509 MB/s [ 0.567688] raid6: sse2x4 xor() 10177 MB/s [ 0.571692] invalid opcode: 0000 [#1] SMP [ 0.572312] Modules linked in: [ 0.572822] CPU: 0 PID: 1 Comm: swapper/0 Not tainted 4.11.0-rc8-ktest #1 [ 0.573734] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Ubuntu-1.8.2-1ubuntu1 04/01/2014 [ 0.575040] task: ffff8f791e1c0000 task.stack: ffff9c72c00d0000 [ 0.575865] RIP: 0010:raid6_avx21_gen_syndrome+0x3d/0x120 [ 0.576634] RSP: 0018:ffff9c72c00d3d70 EFLAGS: 00010246 [ 0.577376] RAX: 0000000000000000 RBX: ffff9c72c00d3dc0 RCX: 00000000fffedb97 [ 0.578327] RDX: 0000000000000000 RSI: 0000000000001000 RDI: 0000000000000012 [ 0.579283] RBP: ffff9c72c00d3da0 R08: 0000000000000000 R09: 00000000000000cd [ 0.580243] R10: 00000000fffedb86 R11: ffffffffa617008d R12: 0000000000001000 [ 0.581211] R13: ffff8f791e39e000 R14: ffff8f791e39f000 R15: 0000000000000012 [ 0.582163] FS: 0000000000000000(0000) GS:ffff8f791fc00000(0000) knlGS:0000000000000000 [ 0.583324] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 0.584128] CR2: 0000000000000000 CR3: 000000001be09000 CR4: 00000000003006f0 [ 0.585078] Call Trace: [ 0.594952] raid6_select_algo+0x116/0x30b [ 0.595592] ? libcrc32c_mod_init+0x2b/0x2b [ 0.596240] do_one_initcall+0x53/0x1a0 [ 0.596843] ? parse_args+0x2cf/0x490 [ 0.597421] kernel_init_freeable+0x182/0x21c [ 0.598077] ? rest_init+0x80/0x80 [ 0.598626] kernel_init+0xe/0x100 [ 0.599175] ret_from_fork+0x2c/0x40 [ 0.599741] Code: 55 41 54 53 48 89 d3 48 8d 14 c5 00 00 00 00 41 89 ff 49 89 f4 48 83 ec 08 4c 8b 2c c3 4c 8b 74 13 08 48 89 55 d0 e8 53 ed a9 ff <c5> fd 6f 05 2b 2d 4e 00 c5 e5 ef db 4d 85 e4 48 8b 55 d0 0f 84 [ 0.602215] RIP: raid6_avx21_gen_syndrome+0x3d/0x120 RSP: ffff9c72c00d3d70 [ 0.603154] ---[ end trace 17ee01f86b8fc548 ]--- [ 0.603850] Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b [ 0.603850] [ 0.605276] ---[ end Kernel panic - not syncing: Attempted to kill init! exitcode=0x0000000b ... ```
This panic occured at lib/raid6/avx2.c#raid6_avx21_gen_syndrome() and this function consists of many AVX instructions.
lib/raid6/avx2.c: ``` /* * Plain AVX2 implementation */ static void raid6_avx21_gen_syndrome(int disks, size_t bytes, void **ptrs) { u8 **dptr = (u8 **)ptrs; u8 *p, *q; int d, z, z0;
z0 = disks - 3; /* Highest data disk */ p = dptr[z0+1]; /* XOR parity */ q = dptr[z0+2]; /* RS syndrome */
kernel_fpu_begin();
asm volatile("vmovdqa %0,%%ymm0" : : "m" (raid6_avx2_constants.x1d[0])); asm volatile("vpxor %ymm3,%ymm3,%ymm3"); /* Zero temp */
for (d = 0; d < bytes; d += 32) { asm volatile("prefetchnta %0" : : "m" (dptr[z0][d])); asm volatile("vmovdqa %0,%%ymm2" : : "m" (dptr[z0][d]));/* P[0] */ asm volatile("prefetchnta %0" : : "m" (dptr[z0-1][d])); asm volatile("vmovdqa %ymm2,%ymm4");/* Q[0] */ asm volatile("vmovdqa %0,%%ymm6" : : "m" (dptr[z0-1][d])); for (z = z0-2; z >= 0; z--) { asm volatile("prefetchnta %0" : : "m" (dptr[z][d])); asm volatile("vpcmpgtb %ymm4,%ymm3,%ymm5"); asm volatile("vpaddb %ymm4,%ymm4,%ymm4"); asm volatile("vpand %ymm0,%ymm5,%ymm5"); asm volatile("vpxor %ymm5,%ymm4,%ymm4"); asm volatile("vpxor %ymm6,%ymm2,%ymm2"); asm volatile("vpxor %ymm6,%ymm4,%ymm4"); asm volatile("vmovdqa %0,%%ymm6" : : "m" (dptr[z][d])); } asm volatile("vpcmpgtb %ymm4,%ymm3,%ymm5"); asm volatile("vpaddb %ymm4,%ymm4,%ymm4"); asm volatile("vpand %ymm0,%ymm5,%ymm5"); asm volatile("vpxor %ymm5,%ymm4,%ymm4"); asm volatile("vpxor %ymm6,%ymm2,%ymm2"); asm volatile("vpxor %ymm6,%ymm4,%ymm4");
asm volatile("vmovntdq %%ymm2,%0" : "=m" (p[d])); asm volatile("vpxor %ymm2,%ymm2,%ymm2"); asm volatile("vmovntdq %%ymm4,%0" : "=m" (q[d])); asm volatile("vpxor %ymm4,%ymm4,%ymm4"); }
asm volatile("sfence" : : : "memory"); kernel_fpu_end(); }
```
This problem can be bypassed by the following patch.
``` diff --git a/arch/x86/Makefile b/arch/x86/Makefile index 2d44933..b589097 100644 --- a/arch/x86/Makefile +++ b/arch/x86/Makefile @@ -162,7 +162,7 @@ asinstr := $(call as-instr,fxsaveq (%rax),-DCONFIG_AS_FXSAVEQ=1) asinstr += $(call as-instr,pshufb %xmm0$(comma)%xmm0,-DCONFIG_AS_SSSE3=1) asinstr += $(call as-instr,crc32l %eax$(comma)%eax,-DCONFIG_AS_CRC32=1) avx_instr := $(call as-instr,vxorps %ymm0$(comma)%ymm1$(comma)%ymm2,-DCONFIG_AS_AVX=1) -avx2_instr :=$(call as-instr,vpbroadcastb %xmm0$(comma)%ymm1,-DCONFIG_AS_AVX2=1) +#avx2_instr :=$(call as-instr,vpbroadcastb %xmm0$(comma)%ymm1,-DCONFIG_AS_AVX2=1) avx512_instr :=$(call as-instr,vpmovm2b %k1$(comma)%zmm5,-DCONFIG_AS_AVX512=1) sha1_ni_instr :=$(call as-instr,sha1msg1 %xmm0$(comma)%xmm1,-DCONFIG_AS_SHA1_NI=1) sha256_ni_instr :=$(call as-instr,sha256msg1 %xmm0$(comma)%xmm1,-DCONFIG_AS_SHA256_NI=1) ```
I suspect that there is a problem about handling AVX2 in the linux kernel or in Ryzen.
# Build environent
## Hardware
CPU: Ryzen 1800x
## Software
- distro: Ubuntu 16.04 x86_64 - gcc: gcc (Ubuntu 5.4.0-6ubuntu1~16.04.4) 5.4.0 20160609
## Build target kernel version
linux v4.11-r8
## Test environment
## Hardware
A VM running on the above mentioned hardware.
## Software
- distro: Ubuntu 16.04
# Additional information
- .config is attached in this mail - This problam also happens on v4.10 - When I tested v4.10 in my previous Core i5 machine, it succeeded to boot.
Thanks, Satoru
[unhandled content-type:application/octet-stream] | |