[PATCH v2 2/2] NFSv4 account for selinux security context when deciding to share superblock

[Olga Kornievskaia]

Trond/Anna,

I'd like your opinion here. Some static checking flags a "ctx"
assignment in nfs_fill_super() in the new patch. In an existing code
there is a check for it is NULL before dereferencing. However, "ctx"
can never be null. nfs_get_tree_common() which calls nfs_fill_super()
and passes in "ctx" gets it from the passed in "fs_context". If the
passed in arg can be null then we are dereferencing in var assignment
so things would blow up there. So "ctx" can never be null.

Should I create another clean up patch to remove the check for null
ctx in nfs_fill_super() to make static analyzers happy?

On Fri, Feb 19, 2021 at 3:19 AM Dan Carpenter <dan.carpenter at oracle.com> wrote:
>
> Hi Olga,
>
> url:    https://github.com/0day-ci/linux/commits/Olga-Kornievskaia/Add-new-hook-to-compare-new-mount-to-an-existing-mount/20210219-035957
> base:   git://git.linux-nfs.org/projects/trondmy/linux-nfs.git linux-next
> config: i386-randconfig-m021-20210215 (attached as .config)
> compiler: gcc-9 (Debian 9.3.0-15) 9.3.0
>
> If you fix the issue, kindly add following tag as appropriate
> Reported-by: kernel test robot <lkp at intel.com>
> Reported-by: Dan Carpenter <dan.carpenter at oracle.com>
>
> smatch warnings:
> fs/nfs/super.c:1061 nfs_fill_super() error: we previously assumed 'ctx' could be null (see line 1029)
>
> vim +/ctx +1061 fs/nfs/super.c
>
> 62a55d088cd87d Scott Mayhew      2019-12-10  1021  static void nfs_fill_super(struct super_block *sb, struct nfs_fs_context *ctx)
> f7b422b17ee5ee David Howells     2006-06-09  1022  {
> 54ceac45159860 David Howells     2006-08-22  1023       struct nfs_server *server = NFS_SB(sb);
> f7b422b17ee5ee David Howells     2006-06-09  1024
> f7b422b17ee5ee David Howells     2006-06-09  1025       sb->s_blocksize_bits = 0;
> f7b422b17ee5ee David Howells     2006-06-09  1026       sb->s_blocksize = 0;
> 6a74490dca8974 Bryan Schumaker   2012-07-30  1027       sb->s_xattr = server->nfs_client->cl_nfs_mod->xattr;
> 6a74490dca8974 Bryan Schumaker   2012-07-30  1028       sb->s_op = server->nfs_client->cl_nfs_mod->sops;
> 5eb005caf5383d David Howells     2019-12-10 @1029       if (ctx && ctx->bsize)
>                                                             ^^^
> Check for NULL
>
> 5eb005caf5383d David Howells     2019-12-10  1030               sb->s_blocksize = nfs_block_size(ctx->bsize, &sb->s_blocksize_bits);
> f7b422b17ee5ee David Howells     2006-06-09  1031
> 6a74490dca8974 Bryan Schumaker   2012-07-30  1032       if (server->nfs_client->rpc_ops->version != 2) {
> 54ceac45159860 David Howells     2006-08-22  1033               /* The VFS shouldn't apply the umask to mode bits. We will do
> 54ceac45159860 David Howells     2006-08-22  1034                * so ourselves when necessary.
> 54ceac45159860 David Howells     2006-08-22  1035                */
> 1751e8a6cb935e Linus Torvalds    2017-11-27  1036               sb->s_flags |= SB_POSIXACL;
> 54ceac45159860 David Howells     2006-08-22  1037               sb->s_time_gran = 1;
> 20fa1902728698 Peng Tao          2017-06-29  1038               sb->s_export_op = &nfs_export_ops;
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1039       } else
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1040               sb->s_time_gran = 1000;
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1041
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1042       if (server->nfs_client->rpc_ops->version != 4) {
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1043               sb->s_time_min = 0;
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1044               sb->s_time_max = U32_MAX;
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1045       } else {
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1046               sb->s_time_min = S64_MIN;
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1047               sb->s_time_max = S64_MAX;
> 54ceac45159860 David Howells     2006-08-22  1048       }
> f7b422b17ee5ee David Howells     2006-06-09  1049
> ab88dca311a372 Al Viro           2019-12-10  1050       sb->s_magic = NFS_SUPER_MAGIC;
> 54ceac45159860 David Howells     2006-08-22  1051
> ab88dca311a372 Al Viro           2019-12-10  1052       /* We probably want something more informative here */
> ab88dca311a372 Al Viro           2019-12-10  1053       snprintf(sb->s_id, sizeof(sb->s_id),
> ab88dca311a372 Al Viro           2019-12-10  1054                "%u:%u", MAJOR(sb->s_dev), MINOR(sb->s_dev));
> 1fcb79c1b21801 Deepa Dinamani    2019-03-26  1055
> ab88dca311a372 Al Viro           2019-12-10  1056       if (sb->s_blocksize == 0)
> ab88dca311a372 Al Viro           2019-12-10  1057               sb->s_blocksize = nfs_block_bits(server->wsize,
> ab88dca311a372 Al Viro           2019-12-10  1058                                                &sb->s_blocksize_bits);
> f7b422b17ee5ee David Howells     2006-06-09  1059
> ab88dca311a372 Al Viro           2019-12-10  1060       nfs_super_set_maxbytes(sb, server->maxfilesize);
> 52a2a3a4af9af7 Olga Kornievskaia 2021-02-18 @1061       server->has_sec_mnt_opts = ctx->has_sec_mnt_opts;
>                                                                                    ^^^^^^^^^^^^^^^^^^^^^
> Unchecked dereference.  Is the earlier NULL check necessary?  (Actually
> on my system with a built cross function DB, I see that the earlier
> NULL check can be removed.  If the cross function DB were built then
> Smatch would not have printed this warning about inconsistent NULL
> checks).
>
> f7b422b17ee5ee David Howells     2006-06-09  1062  }
> f7b422b17ee5ee David Howells     2006-06-09  1063
>
> ---
> 0-DAY CI Kernel Test Service, Intel Corporation
> https://lists.01.org/hyperkitty/list/kbuild-all@lists.01.org