0
votes
==3905== ERROR SUMMARY: 14 errors from 2 contexts (suppressed: 2 from 2)
==3905== 
==3905== 6 errors in context 1 of 2:
==3905== Invalid write of size 4
==3905==    at 0x401BFE: EliminateXr (in /home/suraj/Desktop/project/fm)
==3905==    by 0x402040: fm_elim (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401395: name_fm (in /home/suraj/Desktop/project/fm)
==3905==    by 0x400C38: main (in /home/suraj/Desktop/project/fm)
==3905==  Address 0x51fc724 is 36 bytes inside a block of size 39 alloc'd
==3905==    at 0x4C2A2DB: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==3905==    by 0x401064: alloc_matrix (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401A59: EliminateXr (in /home/suraj/Desktop/project/fm)
==3905==    by 0x402040: fm_elim (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401395: name_fm (in /home/suraj/Desktop/project/fm)
==3905==    by 0x400C38: main (in /home/suraj/Desktop/project/fm)
==3905== 
==3905== 
==3905== 8 errors in context 2 of 2:
==3905== Invalid write of size 4
==3905==    at 0x401B17: EliminateXr (in /home/suraj/Desktop/project/fm)
==3905==    by 0x402040: fm_elim (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401395: name_fm (in /home/suraj/Desktop/project/fm)
==3905==    by 0x400C38: main (in /home/suraj/Desktop/project/fm)
==3905==  Address 0x51fce4c is 12 bytes inside a block of size 15 alloc'd
==3905==    at 0x4C2A2DB: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==3905==    by 0x401064: alloc_matrix (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401A59: EliminateXr (in /home/suraj/Desktop/project/fm)
==3905==    by 0x402040: fm_elim (in /home/suraj/Desktop/project/fm)
==3905==    by 0x401395: name_fm (in /home/suraj/Desktop/project/fm)
==3905==    by 0x400C38: main (in /home/suraj/Desktop/project/fm)
==3905== 
--3905-- 
--3905-- used_suppression:      2 dl-hack3-cond-1
==3905== 
==3905== ERROR SUMMARY: 14 errors from 2 contexts (suppressed: 2 from 2)

I am getting errors from valgrind.

It says that error is located in EliminateXr but i can't really see the error. EliminateXr:

void EliminateXr(float** t,float* q,float*** tnew,float** qnew,int n1,int n2,int* r,int* s,int sprime){

    float** matrix = (float**)alloc_matrix(sprime,(*r)-1, sizeof(float));
    float* vec= (float*)malloc(sprime*sizeof(float));
    int matrixIndex=0;
    int i,k,l;
    for(k = 0; k < n1; ++k){
        for(l = n1; l < n2; ++l){
            for(i=0; i < *r; ++i){
                matrix[matrixIndex][i]=t[k][i]-t[l][i];
            }
            vec[matrixIndex]=q[k]-q[l];
            matrixIndex++;
        }
    }
    for(k = n2; k < *s; ++k){
        for(i=0; i < *r; ++i){
            matrix[matrixIndex][i]=t[k][i];
        }
        vec[matrixIndex]=q[k];
        matrixIndex++;
    }
    *tnew=matrix;
    *qnew=vec;  
    *r=(*r)-1;
    *s=sprime;

}

I am allocating memory for float so i shouldn't get invalid write of size 4. Can anyone explain how i should use this information: Address 0x51fc724 is 36 bytes inside a block of size 39 alloc'd ==3905== at 0x4C2A2DB: malloc (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)

If we look at my alloc_matrix i have a malloc

s = m * n * block + m * sizeof(void*) + block - 1;
p = malloc(s);

block= size of float.

Where are the error according to valgrind?

1
What's the - 1 for when you calculate s?T.C.
For alignment extra (block -1) byte are allocated.user2975699
You need to look at the code in alloc_matrix(), the calculation of the size of the malloc'd memory is wrong.cmaster - reinstate monica
If the addition of block - 1 is for alignment, then you are actually accessing one float after the end of the intended array. Btw, the addition of block - 1 is entirely pointless, malloc() will never return an address that is misaligned for any hardware datatype. But I can't say too much without the code of alloc_matrix().cmaster - reinstate monica

1 Answers

5
votes

My psychic powers say that at this line

float** matrix = (float**)alloc_matrix(sprime,(*r)-1, sizeof(float));

you are allocating a matrix capable of holding sprime rows of floats, each row containing *r - 1 columns.

Yet in your inner loop like this

for(i=0; i < *r; ++i){
    matrix[matrixIndex][i]=t[k][i]-t[l][i];
}

You are accessing the matrix as though it contains *r columns. (i is going from 0 to *r - 1) Note that the value of *r is not changed until later at the end of the function.

Hence, you are trying to write past the end of the buffer, and Valgrind rightfully complains.