Re: ffsv2 extattr support

["Thomas Schmitt" <scdbackup%gmx.net@localhost>]

Hi,

Emmanuel Dreyfus:

> IIRC, we must store a trailing 0 because of an oddity in the FreeBSD API.

It causes a slight but possibly painful incompatibility to other
systems. An appended 0 might have a meaning in some special case.
So one should not flood the "user" namespace with them.

What API/ABI might be affected ?


I believe to see that the current way in getextattr.c is wrong
if option -n sets variable "flag_null" to 1.

Pity there is no valgrind to prove that this is bad memory usage:

                mkbuf(&buf, &buflen, val_len);
                ...
                        error = extattr_set_file(argv[arg_counter],
                            attrnamespace, attrname, buf,
                            val_len + flag_null);

In any case, the result proves that flag_null was indeed 1:

  $ echo c >/mnt/ffs1/c
  $ setextattr -n user bla c_with_n /mnt/ffs1/c
  $ getextattr -s user bla /mnt/ffs1/c
  /mnt/ffs1/c     "c_with_n\000\000"


> How is it done in FreeBSD sources?

Without unconditionally adding 1 to strlen().

  
http://svnweb.freebsd.org/base/head/usr.sbin/extattr/rmextattr.c?revision=248995&view=markup#l186

        if (what == EASET) {
                mkbuf(&buf, &buflen, strlen(argv[0]) + 1);
                strcpy(buf, argv[0]);
                argc--; argv++;
        }

Line 205:
                        len = strlen(buf) + flag_null;

Line 210:
                ret = extattr_set_file(argv[arg_counter],
                    attrnamespace, attrname, buf, len);


Have a nice day :)

Thomas