使用 Open MPI 运行并行程序时出现分段错误

在我之前的文章中，我需要在 10 台计算机之间分发 pgm 文件的数据。在 Jonathan Dursi 和 Shawn Chin 的帮助下，我集成了代码。 我可以编译我的程序，但出现分段错误。我跑了，但什么也没发生

mpirun -np 10 ./exmpi_2 balloons.pgm output.pgm

结果是

[ubuntu:04803] *** Process received signal ***
[ubuntu:04803] Signal: Segmentation fault (11)
[ubuntu:04803] Signal code: Address not mapped (1)
[ubuntu:04803] Failing at address: 0x7548d0c
[ubuntu:04803] [ 0] [0x86b410]
[ubuntu:04803] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x186b00]
[ubuntu:04803] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04803] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x141bd6]
[ubuntu:04803] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04803] *** End of error message ***
--------------------------------------------------------------------------
mpirun noticed that process rank 1 with PID 4803 on node ubuntu exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------

然后我尝试使用 valgrind 运行来调试程序并生成 output.pgm

valgrind mpirun -np 10 ./exmpi_2 balloons.pgm output.pgm

结果是

==4632== Memcheck, a memory error detector
==4632== Copyright (C) 2002-2009, and GNU GPL'd, by Julian Seward et al.
==4632== Using Valgrind-3.6.0.SVN-Debian and LibVEX; rerun with -h for copyright info
==4632== Command: mpirun -np 10 ./exmpi_2 2.pgm 10.pgm
==4632==
==4632== Syscall param sched_setaffinity(mask) points to unaddressable byte(s)
==4632==    at 0x4215D37: syscall (syscall.S:31)
==4632==    by 0x402B335: opal_paffinity_linux_plpa_api_probe_init (plpa_api_probe.c:56)
==4632==    by 0x402B7CC: opal_paffinity_linux_plpa_init (plpa_runtime.c:37)
==4632==    by 0x402B93C: opal_paffinity_linux_plpa_have_topology_information (plpa_map.c:494)
==4632==    by 0x402B180: linux_module_init (paffinity_linux_module.c:119)
==4632==    by 0x40BE2C3: opal_paffinity_base_select (paffinity_base_select.c:64)
==4632==    by 0x40927AC: opal_init (opal_init.c:295)
==4632==    by 0x4046767: orte_init (orte_init.c:76)
==4632==    by 0x804A82E: orterun (orterun.c:540)
==4632==    by 0x804A3EE: main (main.c:13)
==4632==  Address 0x0 is not stack'd, malloc'd or (recently) free'd
==4632==
[ubuntu:04638] *** Process received signal ***
[ubuntu:04639] *** Process received signal ***
[ubuntu:04639] Signal: Segmentation fault (11)
[ubuntu:04639] Signal code: Address not mapped (1)
[ubuntu:04639] Failing at address: 0x7548d0c
[ubuntu:04639] [ 0] [0xc50410]  
[ubuntu:04639] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0xde4b00]
[ubuntu:04639] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04639] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0xd9fbd6]
[ubuntu:04639] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04639] *** End of error message ***
[ubuntu:04640] *** Process received signal ***
[ubuntu:04640] Signal: Segmentation fault (11)
[ubuntu:04640] Signal code: Address not mapped (1)
[ubuntu:04640] Failing at address: 0x7548d0c
[ubuntu:04640] [ 0] [0xdad410]
[ubuntu:04640] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0xe76b00]
[ubuntu:04640] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04640] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0xe31bd6]
[ubuntu:04640] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04640] *** End of error message ***
[ubuntu:04641] *** Process received signal ***
[ubuntu:04641] Signal: Segmentation fault (11)
[ubuntu:04641] Signal code: Address not mapped (1)
[ubuntu:04641] Failing at address: 0x7548d0c
[ubuntu:04641] [ 0] [0xe97410]
[ubuntu:04641] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x1e8b00]
[ubuntu:04641] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04641] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x1a3bd6]
[ubuntu:04641] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04641] *** End of error message ***
[ubuntu:04642] *** Process received signal ***
[ubuntu:04642] Signal: Segmentation fault (11)
[ubuntu:04642] Signal code: Address not mapped (1)
[ubuntu:04642] Failing at address: 0x7548d0c
[ubuntu:04642] [ 0] [0x92d410]
[ubuntu:04642] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x216b00]
[ubuntu:04642] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04642] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x1d1bd6]
[ubuntu:04642] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04642] *** End of error message ***
[ubuntu:04643] *** Process received signal ***
[ubuntu:04643] Signal: Segmentation fault (11)
[ubuntu:04643] Signal code: Address not mapped (1)
[ubuntu:04643] Failing at address: 0x7548d0c
[ubuntu:04643] [ 0] [0x8f4410]
[ubuntu:04643] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x16bb00]
[ubuntu:04643] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04643] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x126bd6]
[ubuntu:04643] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04643] *** End of error message ***
[ubuntu:04638] Signal: Segmentation fault (11)
[ubuntu:04638] Signal code: Address not mapped (1)
[ubuntu:04638] Failing at address: 0x7548d0c
[ubuntu:04638] [ 0] [0x4f6410]
[ubuntu:04638] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x222b00]
[ubuntu:04638] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04638] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x1ddbd6]
[ubuntu:04638] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04638] *** End of error message ***
[ubuntu:04644] *** Process received signal ***
[ubuntu:04644] Signal: Segmentation fault (11)
[ubuntu:04644] Signal code: Address not mapped (1)
[ubuntu:04644] Failing at address: 0x7548d0c
[ubuntu:04644] [ 0] [0x61f410]
[ubuntu:04644] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x1a3b00]
[ubuntu:04644] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04644] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x15ebd6]
[ubuntu:04644] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04644] *** End of error message ***
[ubuntu:04645] *** Process received signal ***
[ubuntu:04645] Signal: Segmentation fault (11)
[ubuntu:04645] Signal code: Address not mapped (1)
[ubuntu:04645] Failing at address: 0x7548d0c
[ubuntu:04645] [ 0] [0x7a3410]
[ubuntu:04645] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x1d5b00]
[ubuntu:04645] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04645] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x190bd6]
[ubuntu:04645] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04645] *** End of error message ***
[ubuntu:04647] *** Process received signal ***
[ubuntu:04647] Signal: Segmentation fault (11)
[ubuntu:04647] Signal code: Address not mapped (1)
[ubuntu:04647] Failing at address: 0x7548d0c
[ubuntu:04647] [ 0] [0xf54410]
[ubuntu:04647] [ 1] /lib/tls/i686/cmov/libc.so.6(fclose+0x1a0) [0x2bab00]
[ubuntu:04647] [ 2] ./exmpi_2(main+0x78e) [0x80492c2]
[ubuntu:04647] [ 3] /lib/tls/i686/cmov/libc.so.6(__libc_start_main+0xe6) [0x275bd6]
[ubuntu:04647] [ 4] ./exmpi_2() [0x8048aa1]
[ubuntu:04647] *** End of error message ***
--------------------------------------------------------------------------
mpirun noticed that process rank 2 with PID 4639 on node ubuntu exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------
6 total processes killed (some possibly by mpirun during cleanup)
==4632==
==4632== HEAP SUMMARY:
==4632==     in use at exit: 158,751 bytes in 1,635 blocks
==4632==   total heap usage: 10,443 allocs, 8,808 frees, 15,854,537 bytes allocated
==4632==
==4632== LEAK SUMMARY:
==4632==    definitely lost: 81,655 bytes in 112 blocks
==4632==    indirectly lost: 5,108 bytes in 91 blocks
==4632==      possibly lost: 1,043 bytes in 17 blocks
==4632==    still reachable: 70,945 bytes in 1,415 blocks 
==4632==         suppressed: 0 bytes in 0 blocks
==4632== Rerun with --leak-check=full to see details of leaked memory
==4632==
==4632== For counts of detected and suppressed errors, rerun with: -v
==4632== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 96 from 9)

有人可以帮我解决这个问题吗？这是我的源代码

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include "mpi.h"
#include <syscall.h>

#define SIZE_X 640
#define SIZE_Y 480




 int main(int argc, char **argv)
{
FILE *FR,*FW;
int ierr;
int rank, size;
int ncells;
int greys[SIZE_X][SIZE_Y];
int rows,cols, maxval;

int mystart, myend, myncells;
const int IONODE=0;
int *disps, *counts, *mydata;
int *data;
int i,j,temp1;
char dummy[50]="";





ierr = MPI_Init(&argc, &argv);
if (argc != 3) {
    fprintf(stderr,"Usage: %s infile outfile\n",argv[0]);
    fprintf(stderr,"outputs the negative of the input file.\n");
    return -1;
}            

ierr  = MPI_Comm_rank(MPI_COMM_WORLD, &rank);
ierr = MPI_Comm_size(MPI_COMM_WORLD, &size);
if (ierr) {
    fprintf(stderr,"Catastrophic MPI problem; exiting\n");
    MPI_Abort(MPI_COMM_WORLD,1);
}

    if (rank == IONODE) {
            //if (read_pgm(argv[1], &greys, &rows, &cols, &maxval)) {
            //   fprintf(stderr,"Could not read file; exiting\n");
              //   MPI_Abort(MPI_COMM_WORLD,2);

         rows=SIZE_X;
         cols=SIZE_Y;
         maxval=255;
         FR=fopen(argv[1], "r+");

         fgets(dummy,50,FR);
         do{  fgets(dummy,50,FR); } while(dummy[0]=='#');
         fgets(dummy,50,FR);

     for (j = 0; j <cols; j++)
     {
       for (i = 0; i <rows; i++)
       {
           fscanf(FR,"%d",&temp1);
         greys[i][j] = temp1;
       }
     }
}

    ncells = rows*cols;
    disps = (int *)malloc(size * sizeof(int));
    counts= (int *)malloc(size * sizeof(int));
    data = &(greys[0][0]); /* we know all the data is contiguous */

/* everyone calculate their number of cells */
ierr = MPI_Bcast(&ncells, 1, MPI_INT, IONODE, MPI_COMM_WORLD);
myncells = ncells/size;
mystart = rank*myncells;
myend   = mystart + myncells - 1;
if (rank == size-1) myend = ncells-1;
myncells = (myend-mystart)+1;
mydata = (int *)malloc(myncells * sizeof(int));

/* assemble the list of counts.  Might not be equal if don't divide evenly. */
ierr = MPI_Gather(&myncells, 1, MPI_INT, counts, 1, MPI_INT, IONODE, MPI_COMM_WORLD);
if (rank == IONODE) {
    disps[0] = 0;
    for (i=1; i<size; i++) {
        disps[i] = disps[i-1] + counts[i-1];
    }
}

/* scatter the data */
ierr = MPI_Scatterv(data, counts, disps, MPI_INT, mydata, myncells, MPI_INT, IONODE, MPI_COMM_WORLD);

/* everyone has to know maxval */
ierr = MPI_Bcast(&maxval, 1, MPI_INT, IONODE, MPI_COMM_WORLD);

for (i=0; i<myncells; i++)
    mydata[i] = maxval-mydata[i];

/* Gather the data */
ierr = MPI_Gatherv(mydata, myncells, MPI_INT, data, counts, disps, MPI_INT, IONODE, MPI_COMM_WORLD);

if (rank == IONODE)
{
//      write_pgm(argv[2], greys, rows, cols, maxval);
  FW=fopen(argv[2], "w");
  fprintf(FW,"P2\n%d %d\n255\n",rows,cols);    
  for(j=0;j<cols;j++)
    for(i=0;i<rows;i++)
   fprintf(FW,"%d ", greys[i][j]);
}

free(mydata);
if (rank == IONODE) {
    free(counts);
    free(disps);
    //free(&(greys[0][0]));
    //free(greys);

}
fclose(FR);
fclose(FW);
MPI_Finalize();
return 0;
}

这是输入图像http://orion.math.iastate.edu/burkardt/data/pgm/balloons.pgm http://orion.math.iastate.edu/burkardt/data/pgm/balloons.pgm

恭喜；代码almost运行完全完美，几乎在最后几行代码处就死掉了。

使用 valgrind 会更清楚这个问题，但是使用 MPI 或任何涉及程序启动器的东西运行 valgrind 时必须更加棘手。代替：

valgrind mpirun -np 10 ./exmpi_2 balloons.pgm output.pgm

它做了 mpirun 的 valgrind，你并不真正关心，你想做

mpirun -np 10 valgrind ./exmpi_2 balloons.pgm output.pgm

-- 也就是说，您想要启动 10 个 valgrind，每个运行一个进程的 exmpi_2。如果你这样做（并且你已经使用 -g 进行了编译），你会在最后发现 valgrind 输出如下：

==6303==  Access not within mapped region at address 0x1
==6303==    at 0x387FA60C17: fclose@@GLIBC_2.2.5 (in /lib64/libc-2.5.so)
==6303==    by 0x401222: main (pgm.c:124)

.. 这就是全部；你有all进程执行fclose()s，当只有一个进程有一个句柄时fopen()首先是 ed 文件。只需更换

fclose(FR);
fclose(FW);

with

if (rank == IONODE) {
    fclose(FR);
    fclose(FW);
}

似乎对我有用。