On Mon, 2018-08-27 at 13:47 -0400, Jeff Layton wrote:
> POSIX mandates that open fds and their associated file locks should be
> preserved across an execve. This works, unless the process is
> multithreaded at the time that execve is called.
>
> In that case, we'll end up unsharing the files_struct but the locks will
> still have their fl_owner set to the address of the old one. Eventually,
> when the other threads die and the last reference to the old
> files_struct is put, any POSIX locks get torn down since it looks like
> a close occurred on them.
>
> The result is that all of your open files will be intact with none of
> the locks you held before execve. The simple answer to this is "use OFD
> locks", but this is a nasty surprise and it violates the spec.
>
> Fix this by doing unshare_files later during exec, after we've already
> killed off the other threads in the process. This helps ensure that we
> only unshare the files_struct during exec when it is truly shared with
> other processes.
>
> Note that because the unshare_files call is now done just after
> de_thread, we need a mechanism to pass the displaced files_struct back
> up to __do_execve_file. This is done via a new displaced_files field
> inside the linux_binprm.
>
> Cc: Eric W. Biederman <ebied...@xmission.com>
> Reported-by: Daniel P. Berrangé <berra...@redhat.com>
> Signed-off-by: Jeff Layton <jlay...@kernel.org>
> ---
> fs/exec.c | 19 +++++++++----------
> include/linux/binfmts.h | 1 +
> 2 files changed, 10 insertions(+), 10 deletions(-)
>
> diff --git a/fs/exec.c b/fs/exec.c
> index ca25f805ebad..a45b0cae5817 100644
> --- a/fs/exec.c
> +++ b/fs/exec.c
> @@ -1262,6 +1262,10 @@ int flush_old_exec(struct linux_binprm * bprm)
> if (retval)
> goto out;
>
> + retval = unshare_files(&bprm->displaced_files);
> + if (retval)
> + goto out;
> +
> /*
> * Must be called _before_ exec_mmap() as bprm->mm is
> * not visibile until then. This also enables the update
> @@ -1712,8 +1716,7 @@ static int __do_execve_file(int fd, struct filename
> *filename,
> int flags, struct file *file)
> {
> char *pathbuf = NULL;
> - struct linux_binprm *bprm;
> - struct files_struct *displaced;
> + struct linux_binprm *bprm = NULL;
> int retval;
>
> if (IS_ERR(filename))
> @@ -1735,10 +1738,6 @@ static int __do_execve_file(int fd, struct filename
> *filename,
> * further execve() calls fail. */
> current->flags &= ~PF_NPROC_EXCEEDED;
>
> - retval = unshare_files(&displaced);
> - if (retval)
> - goto out_ret;
> -
> retval = -ENOMEM;
> bprm = kzalloc(sizeof(*bprm), GFP_KERNEL);
> if (!bprm)
> @@ -1831,8 +1830,8 @@ static int __do_execve_file(int fd, struct filename
> *filename,
> kfree(pathbuf);
> if (filename)
> putname(filename);
> - if (displaced) {
> - put_files_struct(displaced);
> + if (bprm->displaced_files) {
> + put_files_struct(bprm->displaced_files);
Note that this is broken (bprm is freed above this point). It's simple
enough to fix, but I'll hold off on resending until I hear some feedback
on the general approach.
> } else {
> spin_lock(¤t->files->file_lock);
> current->files->in_exec = false;
> @@ -1855,8 +1854,8 @@ static int __do_execve_file(int fd, struct filename
> *filename,
> kfree(pathbuf);
>
> out_files:
> - if (displaced) {
> - reset_files_struct(displaced);
> + if (bprm && bprm->displaced_files) {
> + reset_files_struct(bprm->displaced_files);
> } else {
> spin_lock(¤t->files->file_lock);
> current->files->in_exec = false;
> diff --git a/include/linux/binfmts.h b/include/linux/binfmts.h
> index c05f24fac4f6..d7ec384bb1b0 100644
> --- a/include/linux/binfmts.h
> +++ b/include/linux/binfmts.h
> @@ -49,6 +49,7 @@ struct linux_binprm {
> unsigned int taso:1;
> #endif
> unsigned int recursion_depth; /* only for search_binary_handler() */
> + struct files_struct * displaced_files;
> struct file * file;
> struct cred *cred; /* new credentials */
> int unsafe; /* how unsafe this exec is (mask of
> LSM_UNSAFE_*) */
--
Jeff Layton <jlay...@kernel.org>
