diff options
Diffstat (limited to 'src/H5FDmpio.c')
| -rw-r--r-- | src/H5FDmpio.c | 45 |
1 files changed, 33 insertions, 12 deletions
diff --git a/src/H5FDmpio.c b/src/H5FDmpio.c index c0b14ce..fe737c6 100644 --- a/src/H5FDmpio.c +++ b/src/H5FDmpio.c @@ -860,14 +860,19 @@ H5FD__mpio_open(const char *name, unsigned flags, hid_t fapl_id, haddr_t H5_ATTR file->mpi_size = mpi_size; /* Only processor p0 will get the filesize and broadcast it. */ - if (mpi_rank == 0) + if (mpi_rank == 0) { + /* If MPI_File_get_size fails, broadcast file size as -1 to signal error */ if (MPI_SUCCESS != (mpi_code = MPI_File_get_size(fh, &file_size))) - HMPI_GOTO_ERROR(NULL, "MPI_File_get_size failed", mpi_code) + file_size = (MPI_Offset)-1; + } /* Broadcast file size */ if (MPI_SUCCESS != (mpi_code = MPI_Bcast(&file_size, (int)sizeof(MPI_Offset), MPI_BYTE, 0, comm))) HMPI_GOTO_ERROR(NULL, "MPI_Bcast failed", mpi_code) + if (file_size < 0) + HMPI_GOTO_ERROR(NULL, "MPI_File_get_size failed", mpi_code) + /* Determine if the file should be truncated */ if (file_size && (flags & H5F_ACC_TRUNC)) { /* Truncate the file */ @@ -1264,10 +1269,14 @@ H5FD__mpio_read(H5FD_t *_file, H5FD_mem_t H5_ATTR_UNUSED type, hid_t H5_ATTR_UNU rank0_bcast = TRUE; /* Read on rank 0 Bcast to other ranks */ - if (file->mpi_rank == 0) + if (file->mpi_rank == 0) { + /* If MPI_File_read_at fails, push an error, but continue + * to participate in following MPI_Bcast */ if (MPI_SUCCESS != (mpi_code = MPI_File_read_at(file->f, mpi_off, buf, size_i, buf_type, &mpi_stat))) - HMPI_GOTO_ERROR(FAIL, "MPI_File_read_at failed", mpi_code) + HMPI_DONE_ERROR(FAIL, "MPI_File_read_at failed", mpi_code) + } + if (MPI_SUCCESS != (mpi_code = MPI_Bcast(buf, size_i, buf_type, 0, file->comm))) HMPI_GOTO_ERROR(FAIL, "MPI_Bcast failed", mpi_code) } /* end if */ @@ -1311,11 +1320,21 @@ H5FD__mpio_read(H5FD_t *_file, H5FD_mem_t H5_ATTR_UNUSED type, hid_t H5_ATTR_UNU if (!rank0_bcast || (rank0_bcast && file->mpi_rank == 0)) { /* How many bytes were actually read? */ #if MPI_VERSION >= 3 - if (MPI_SUCCESS != (mpi_code = MPI_Get_elements_x(&mpi_stat, buf_type, &bytes_read))) + if (MPI_SUCCESS != (mpi_code = MPI_Get_elements_x(&mpi_stat, buf_type, &bytes_read))) { #else - if (MPI_SUCCESS != (mpi_code = MPI_Get_elements(&mpi_stat, MPI_BYTE, &bytes_read))) + if (MPI_SUCCESS != (mpi_code = MPI_Get_elements(&mpi_stat, MPI_BYTE, &bytes_read))) { #endif - HMPI_GOTO_ERROR(FAIL, "MPI_Get_elements failed", mpi_code) + if (rank0_bcast && file->mpi_rank == 0) { + /* If MPI_Get_elements(_x) fails for a rank 0 bcast strategy, + * push an error, but continue to participate in the following + * MPI_Bcast. + */ + bytes_read = -1; + HMPI_DONE_ERROR(FAIL, "MPI_Get_elements failed", mpi_code) + } + else + HMPI_GOTO_ERROR(FAIL, "MPI_Get_elements failed", mpi_code) + } } /* end if */ /* If the rank0-bcast feature was used, broadcast the # of bytes read to @@ -1695,17 +1714,19 @@ H5FD__mpio_truncate(H5FD_t *_file, hid_t H5_ATTR_UNUSED dxpl_id, hbool_t H5_ATTR HMPI_GOTO_ERROR(FAIL, "MPI_Barrier failed", mpi_code) /* Only processor p0 will get the filesize and broadcast it. */ - /* (Note that throwing an error here will cause non-rank 0 processes - * to hang in following Bcast. -QAK, 3/17/2018) - */ - if (0 == file->mpi_rank) + if (0 == file->mpi_rank) { + /* If MPI_File_get_size fails, broadcast file size as -1 to signal error */ if (MPI_SUCCESS != (mpi_code = MPI_File_get_size(file->f, &size))) - HMPI_GOTO_ERROR(FAIL, "MPI_File_get_size failed", mpi_code) + size = (MPI_Offset)-1; + } /* Broadcast file size */ if (MPI_SUCCESS != (mpi_code = MPI_Bcast(&size, (int)sizeof(MPI_Offset), MPI_BYTE, 0, file->comm))) HMPI_GOTO_ERROR(FAIL, "MPI_Bcast failed", mpi_code) + if (size < 0) + HMPI_GOTO_ERROR(FAIL, "MPI_File_get_size failed", mpi_code) + if (H5FD_mpi_haddr_to_MPIOff(file->eoa, &needed_eof) < 0) HGOTO_ERROR(H5E_INTERNAL, H5E_BADRANGE, FAIL, "cannot convert from haddr_t to MPI_Offset") |