* Re: [linux-next:master 952/3956] crypto/blake2b_generic.c:73:13: warning: stack frame size of 9776 bytes in function 'blake2b_compress_one_generic'
[not found] <202101160841.jUXjdS7j-lkp@intel.com>
@ 2021-01-16 1:59 ` Eric Biggers
2021-01-16 15:36 ` Arnd Bergmann
0 siblings, 1 reply; 2+ messages in thread
From: Eric Biggers @ 2021-01-16 1:59 UTC (permalink / raw)
To: kernel test robot
Cc: kbuild-all, clang-built-linux, Linux Memory Management List,
Herbert Xu, linux-crypto, Arnd Bergmann
On Sat, Jan 16, 2021 at 08:59:50AM +0800, kernel test robot wrote:
> tree: https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git master
> head: b3a3cbdec55b090d22a09f75efb7c7d34cb97f25
> commit: 28dcca4cc0c01e2467549a36b1b0eacfdb01236c [952/3956] crypto: blake2b - sync with blake2s implementation
> config: powerpc64-randconfig-r021-20210115 (attached as .config)
> compiler: clang version 12.0.0 (https://github.com/llvm/llvm-project 5b42fd8dd4e7e29125a09a41a33af7c9cb57d144)
> reproduce (this is a W=1 build):
> wget https://raw.githubusercontent.com/intel/lkp-tests/master/sbin/make.cross -O ~/bin/make.cross
> chmod +x ~/bin/make.cross
> # install powerpc64 cross compiling tool for clang build
> # apt-get install binutils-powerpc64-linux-gnu
> # https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/commit/?id=28dcca4cc0c01e2467549a36b1b0eacfdb01236c
> git remote add linux-next https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git
> git fetch --no-tags linux-next master
> git checkout 28dcca4cc0c01e2467549a36b1b0eacfdb01236c
> # save the attached .config to linux build tree
> COMPILER_INSTALL_PATH=$HOME/0day COMPILER=clang make.cross ARCH=powerpc64
>
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp@intel.com>
>
> All warnings (new ones prefixed by >>):
>
> >> crypto/blake2b_generic.c:73:13: warning: stack frame size of 9776 bytes in function 'blake2b_compress_one_generic' [-Wframe-larger-than=]
> static void blake2b_compress_one_generic(struct blake2b_state *S,
> ^
> 1 warning generated.
>
>
> vim +/blake2b_compress_one_generic +73 crypto/blake2b_generic.c
>
> 48
> 49 #define G(r,i,a,b,c,d) \
> 50 do { \
> 51 a = a + b + m[blake2b_sigma[r][2*i+0]]; \
> 52 d = ror64(d ^ a, 32); \
> 53 c = c + d; \
> 54 b = ror64(b ^ c, 24); \
> 55 a = a + b + m[blake2b_sigma[r][2*i+1]]; \
> 56 d = ror64(d ^ a, 16); \
> 57 c = c + d; \
> 58 b = ror64(b ^ c, 63); \
> 59 } while (0)
> 60
> 61 #define ROUND(r) \
> 62 do { \
> 63 G(r,0,v[ 0],v[ 4],v[ 8],v[12]); \
> 64 G(r,1,v[ 1],v[ 5],v[ 9],v[13]); \
> 65 G(r,2,v[ 2],v[ 6],v[10],v[14]); \
> 66 G(r,3,v[ 3],v[ 7],v[11],v[15]); \
> 67 G(r,4,v[ 0],v[ 5],v[10],v[15]); \
> 68 G(r,5,v[ 1],v[ 6],v[11],v[12]); \
> 69 G(r,6,v[ 2],v[ 7],v[ 8],v[13]); \
> 70 G(r,7,v[ 3],v[ 4],v[ 9],v[14]); \
> 71 } while (0)
> 72
> > 73 static void blake2b_compress_one_generic(struct blake2b_state *S,
> 74 const u8 block[BLAKE2B_BLOCK_SIZE])
> 75 {
> 76 u64 m[16];
> 77 u64 v[16];
> 78 size_t i;
> 79
> 80 for (i = 0; i < 16; ++i)
> 81 m[i] = get_unaligned_le64(block + i * sizeof(m[i]));
> 82
> 83 for (i = 0; i < 8; ++i)
> 84 v[i] = S->h[i];
> 85
> 86 v[ 8] = BLAKE2B_IV0;
> 87 v[ 9] = BLAKE2B_IV1;
> 88 v[10] = BLAKE2B_IV2;
> 89 v[11] = BLAKE2B_IV3;
> 90 v[12] = BLAKE2B_IV4 ^ S->t[0];
> 91 v[13] = BLAKE2B_IV5 ^ S->t[1];
> 92 v[14] = BLAKE2B_IV6 ^ S->f[0];
> 93 v[15] = BLAKE2B_IV7 ^ S->f[1];
> 94
> 95 ROUND(0);
> 96 ROUND(1);
> 97 ROUND(2);
> 98 ROUND(3);
> 99 ROUND(4);
> 100 ROUND(5);
> 101 ROUND(6);
> 102 ROUND(7);
> 103 ROUND(8);
> 104 ROUND(9);
> 105 ROUND(10);
> 106 ROUND(11);
> 107 #ifdef CONFIG_CC_IS_CLANG
> 108 #pragma nounroll /* https://bugs.llvm.org/show_bug.cgi?id=45803 */
> 109 #endif
> 110 for (i = 0; i < 8; ++i)
> 111 S->h[i] = S->h[i] ^ v[i] ^ v[i + 8];
> 112 }
> 113
>
> ---
> 0-DAY CI Kernel Test Service, Intel Corporation
> https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org
Looks like the clang bug that causes large stack usage in this function
(https://bugs.llvm.org/show_bug.cgi?id=45803 which is still unfixed) got
triggered again. Note that the function only has 264 bytes of local variables,
so there's no reason why it should use anywhere near 9776 bytes of stack space.
I'm not sure what we can do about this. Last time the solution was commit
0c0408e86dbe which randomly added a 'pragma nounroll' to the loop at the end.
Anyone have any better idea than randomly trying adding optimization pragmas and
seeing what makes the report go away?
Also this was reported with clang 12.0.0 which is a prerelease version, so I'm
not sure how much I'm supposed to care about this report.
- Eric
^ permalink raw reply [flat|nested] 2+ messages in thread