exec: Remove __FMODE_EXEC from uselib()
Subsystems (3)
PATCH 1/1 exec: Remove __FMODE_EXEC from uselib()
Path-based LSMs will bypass uselib() "open" checks since commit 4759ff71f23e ("exec: Check __FMODE_EXEC instead of in_execve for LSMs"), so don't set __FMODE_EXEC during uselib(). The LSM "open" and eventual "mmap" hooks will be restored. (uselib() never set current->in_execve.)
Other things that checked __FMODE_EXEC:
- fs/fcntl.c is just doing a bitfield sanity check.
- nfs_open_permission_mask() is only checking for the "unreadable exec" case, which is not an issue for uselib(), which sets MAY_READ, unlike execve().
- fsnotify would no longer see uselib() as FS_OPEN_EXEC_PERM, but rather as FS_OPEN_PERM, but this is likely a bug fix, as uselib() isn't an exec: it's more like mmap(), which fsnotify doesn't intercept.
--- fs/exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
fs/exec.c +1 -1
@@ -128,7 +128,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library)struct filename *tmp =getname (library);int error =PTR_ERR (tmp);static const struct open_flags uselib_flags = { - .open_flag =O_LARGEFILE |O_RDONLY | __FMODE_EXEC, + .open_flag =O_LARGEFILE |O_RDONLY , .acc_mode =MAY_READ |MAY_EXEC , .intent =LOOKUP_OPEN , .lookup_flags =LOOKUP_FOLLOW ,
Signature · 1 line
2.34.1
Quoted 4 lines · Kees Cook
Path-based LSMs will bypass
uselib()"open" checks since commit4759ff71f23e("exec: Check__FMODE_EXECinstead ofin_execvefor LSMs"), so don't set__FMODE_EXECduringuselib(). The LSM "open" and eventual "mmap" hooks will be restored. (uselib()never setcurrent->in_execve.)
Ah, nevermind, I see Linux's commit has taken care of this already:
https://git.kernel.org/linus/3eab830189d94f0f80f34cbff609b5bb54002679
Signature · 1 line
Kees Cook
Quoted 9 lines · Kees Cook
Path-based LSMs will bypass
uselib()"open" checks since commit4759ff71f23e("exec: Check__FMODE_EXECinstead ofin_execvefor LSMs"), so don't set__FMODE_EXECduringuselib(). The LSM "open" and eventual "mmap" hooks will be restored. (uselib()never setcurrent->in_execve.)Other things that checked
__FMODE_EXEC:-
fs/fcntl.cis just doing a bitfield sanity check.
Quoted 7 lines
-
nfs_open_permission_mask()is only checking for the "unreadable exec" case, which is not an issue foruselib(), which setsMAY_READ, unlikeexecve().- fsnotify would no longer see
uselib()asFS_OPEN_EXEC_PERM, but rather asFS_OPEN_PERM, but this is likely a bug fix, asuselib()isn't an exec: it's more likemmap(), which fsnotify doesn't intercept.
OK, I went back to the original discussion with Steve Grubb and Matthew Bobrowski who asked for FS_OPEN_EXEC_PERM and AFAICT this change in uselib() should be fine wrt usescases we discussed. That doesn't mean there cannot be some userspace which will get broken by this (in which case we'd have to revert or find some other solution) but I'm willing to try. I'm also CCing Steve & Matthew for input but from my side feel free to add:
Honza
Quoted 33 lines
Reported-by Jann Horn <[email protected]>Closes https://lore.kernel.org/lkml/CAG48ez017tTwxXbxdZ4joVDv5i8FLWEjk=K_z1Vf=pf0v1=cTg@mail.gmail.com/ open hereFixes 4759ff71f23e ("exec: Check __FMODE_EXEC instead of in_execve for LSMs")Suggested-by Linus Torvalds <[email protected]>Cc Kevin Locke <[email protected]>Cc Eric Biederman <[email protected]>Cc Alexander Viro <[email protected]>Cc Christian Brauner <[email protected]>Cc Jan Kara <[email protected]>Signed-off-by Kees Cook <[email protected]>--- fs/exec.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)fs/exec.c +1
-1@@ -128,7 +128,7 @@ SYSCALL_DEFINE1(uselib, const char __user *, library)struct filename *tmp =getname (library);int error =PTR_ERR (tmp);static const struct open_flags uselib_flags = { - .open_flag =O_LARGEFILE |O_RDONLY | __FMODE_EXEC, + .open_flag =O_LARGEFILE |O_RDONLY , .acc_mode =MAY_READ |MAY_EXEC , .intent =LOOKUP_OPEN , .lookup_flags =LOOKUP_FOLLOW ,Signature · 1 line
2.34.1
Signature · 2 lines
Jan Kara <[email protected]> SUSE Labs, CR