Re: [PATCH 4/7] contrib/elf2dmp: Use lduw_le_p() to read PDB

[Peter Maydell]

On Sun, 3 Mar 2024 at 10:52, Akihiko Odaki <akihiko.odaki@daynix.com> wrote:
>
> This resolved UBSan warnings.
>
> Signed-off-by: Akihiko Odaki <akihiko.odaki@daynix.com>
> ---
>  contrib/elf2dmp/pdb.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/contrib/elf2dmp/pdb.c b/contrib/elf2dmp/pdb.c
> index 40991f5f4c34..2541234205c3 100644
> --- a/contrib/elf2dmp/pdb.c
> +++ b/contrib/elf2dmp/pdb.c
> @@ -19,6 +19,7 @@
>   */
>
>  #include "qemu/osdep.h"
> +#include "qemu/bswap.h"
>
>  #include "pdb.h"
>  #include "err.h"
> @@ -187,7 +188,7 @@ static int pdb_init_symbols(struct pdb_reader *r)
>
>      r->symbols = symbols;
>
> -    r->segments = *(uint16_t *)((const char *)symbols + sizeof(PDB_SYMBOLS) +
> +    r->segments = lduw_le_p((const char *)symbols + sizeof(PDB_SYMBOLS) +
>              symbols->module_size + symbols->offset_size +
>              symbols->hash_size + symbols->srcmodule_size +
>              symbols->pdbimport_size + symbols->unknown2_size +

This is a behaviour change -- previously we did a load in
host-endian order, but now we do it in little-endian order.
Which is correct?

If we need host-endian, then we have lduw_he_p() for that.
If we need little-endian, then maybe other parts of the code
also are loading data in the wrong endianness?

thanks
-- PMM