Message ID | 1466812312-26093-1-git-send-email-jannh@google.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Jann Horn <jannh@google.com> writes: > diff --git a/fs/orangefs/devorangefs-req.c b/fs/orangefs/devorangefs-req.c > index db170be..a287a66 100644 > --- a/fs/orangefs/devorangefs-req.c > +++ b/fs/orangefs/devorangefs-req.c > @@ -116,6 +116,13 @@ static int orangefs_devreq_open(struct inode *inode, struct file *file) > { > int ret = -EINVAL; > > + /* in order to ensure that the filesystem driver sees correct UIDs */ > + if (file->f_cred->user_ns != &init_user_ns) { > + gossip_err("%s: device cannot be opened outside init_user_ns\n", > + __func__); > + goto out; > + } > + Not necessarily in this patch but the code should also verify that the opener is also in the initial pid namespace as pids are transferred in the upcalls as well. > if (!(file->f_flags & O_NONBLOCK)) { > gossip_err("%s: device cannot be opened in blocking mode\n", > __func__); Eric -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
Your suggestions make sense to me, especially after looking at how other filesystems use init_user_ns... As far as kicking us out of the Kernel, good grief, I hope not, it was hard getting into the kernel! -Mike On Sat, Jun 25, 2016 at 12:29 AM, Eric W. Biederman <ebiederm@xmission.com> wrote: > Jann Horn <jannh@google.com> writes: > >> diff --git a/fs/orangefs/devorangefs-req.c b/fs/orangefs/devorangefs-req.c >> index db170be..a287a66 100644 >> --- a/fs/orangefs/devorangefs-req.c >> +++ b/fs/orangefs/devorangefs-req.c >> @@ -116,6 +116,13 @@ static int orangefs_devreq_open(struct inode *inode, struct file *file) >> { >> int ret = -EINVAL; >> >> + /* in order to ensure that the filesystem driver sees correct UIDs */ >> + if (file->f_cred->user_ns != &init_user_ns) { >> + gossip_err("%s: device cannot be opened outside init_user_ns\n", >> + __func__); >> + goto out; >> + } >> + > > Not necessarily in this patch but the code should also verify that the > opener is also in the initial pid namespace as pids are transferred in > the upcalls as well. > >> if (!(file->f_flags & O_NONBLOCK)) { >> gossip_err("%s: device cannot be opened in blocking mode\n", >> __func__); > > Eric -- To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
diff --git a/fs/orangefs/devorangefs-req.c b/fs/orangefs/devorangefs-req.c index db170be..a287a66 100644 --- a/fs/orangefs/devorangefs-req.c +++ b/fs/orangefs/devorangefs-req.c @@ -116,6 +116,13 @@ static int orangefs_devreq_open(struct inode *inode, struct file *file) { int ret = -EINVAL; + /* in order to ensure that the filesystem driver sees correct UIDs */ + if (file->f_cred->user_ns != &init_user_ns) { + gossip_err("%s: device cannot be opened outside init_user_ns\n", + __func__); + goto out; + } + if (!(file->f_flags & O_NONBLOCK)) { gossip_err("%s: device cannot be opened in blocking mode\n", __func__); diff --git a/fs/orangefs/orangefs-cache.c b/fs/orangefs/orangefs-cache.c index 900a2e3..b6edbe9 100644 --- a/fs/orangefs/orangefs-cache.c +++ b/fs/orangefs/orangefs-cache.c @@ -136,10 +136,10 @@ struct orangefs_kernel_op_s *op_alloc(__s32 type) llu(new_op->tag), get_opname_string(new_op)); - new_op->upcall.uid = from_kuid(current_user_ns(), + new_op->upcall.uid = from_kuid(&init_user_ns, current_fsuid()); - new_op->upcall.gid = from_kgid(current_user_ns(), + new_op->upcall.gid = from_kgid(&init_user_ns, current_fsgid()); } else { gossip_err("op_alloc: kmem_cache_zalloc failed!\n"); diff --git a/fs/orangefs/orangefs-kernel.h b/fs/orangefs/orangefs-kernel.h index 2281882..a6834d4 100644 --- a/fs/orangefs/orangefs-kernel.h +++ b/fs/orangefs/orangefs-kernel.h @@ -600,8 +600,8 @@ int service_operation(struct orangefs_kernel_op_s *op, #define fill_default_sys_attrs(sys_attr, type, mode) \ do { \ - sys_attr.owner = from_kuid(current_user_ns(), current_fsuid()); \ - sys_attr.group = from_kgid(current_user_ns(), current_fsgid()); \ + sys_attr.owner = from_kuid(&init_user_ns, current_fsuid()); \ + sys_attr.group = from_kgid(&init_user_ns, current_fsgid()); \ sys_attr.perms = ORANGEFS_util_translate_mode(mode); \ sys_attr.mtime = 0; \ sys_attr.atime = 0; \ diff --git a/fs/orangefs/orangefs-utils.c b/fs/orangefs/orangefs-utils.c index 2d129b5..c5fbc62 100644 --- a/fs/orangefs/orangefs-utils.c +++ b/fs/orangefs/orangefs-utils.c @@ -153,12 +153,12 @@ static inline int copy_attributes_from_inode(struct inode *inode, */ attrs->mask = 0; if (iattr->ia_valid & ATTR_UID) { - attrs->owner = from_kuid(current_user_ns(), iattr->ia_uid); + attrs->owner = from_kuid(&init_user_ns, iattr->ia_uid); attrs->mask |= ORANGEFS_ATTR_SYS_UID; gossip_debug(GOSSIP_UTILS_DEBUG, "(UID) %d\n", attrs->owner); } if (iattr->ia_valid & ATTR_GID) { - attrs->group = from_kgid(current_user_ns(), iattr->ia_gid); + attrs->group = from_kgid(&init_user_ns, iattr->ia_gid); attrs->mask |= ORANGEFS_ATTR_SYS_GID; gossip_debug(GOSSIP_UTILS_DEBUG, "(GID) %d\n", attrs->group); } diff --git a/fs/orangefs/xattr.c b/fs/orangefs/xattr.c index 5893ddd..e4a070f 100644 --- a/fs/orangefs/xattr.c +++ b/fs/orangefs/xattr.c @@ -79,8 +79,8 @@ ssize_t orangefs_inode_getxattr(struct inode *inode, const char *prefix, return -EINVAL; } - fsuid = from_kuid(current_user_ns(), current_fsuid()); - fsgid = from_kgid(current_user_ns(), current_fsgid()); + fsuid = from_kuid(&init_user_ns, current_fsuid()); + fsgid = from_kgid(&init_user_ns, current_fsgid()); gossip_debug(GOSSIP_XATTR_DEBUG, "getxattr on inode %pU, name %s "
In orangefs_inode_getxattr(), an fsuid is written to dmesg. The kuid is converted to a userspace uid via from_kuid(current_user_ns(), [...]), but since dmesg is global, init_user_ns should be used here instead. In copy_attributes_from_inode(), op_alloc() and fill_default_sys_attrs(), upcall structures are populated with uids/gids that have been mapped into the caller's namespace. However, those upcall structures are read by another process (the userspace filesystem driver), and that process might be running in another namespace. This effectively lets any user spoof its uid and gid as seen by the userspace filesystem driver. To fix the second issue, I just construct the opcall structures with init_user_ns uids/gids and require the filesystem server to run in the init namespace. Since orangefs is full of global state anyway (as the error message in DUMP_DEVICE_ERROR explains, there can only be one userspace orangefs filesystem driver at once), that shouldn't be a problem. [ Why does orangefs even exist in the kernel if everything does upcalls into userspace? What does orangefs do that couldn't be done with the FUSE interface? If there is no good answer to those questions, I'd prefer to see orangefs kicked out of the kernel. Can that be done for something that shipped in a release? According to commit f7ab093f74bf ("Orangefs: kernel client part 1"), they even already have a FUSE daemon, and the only rational reason (apart from "but most of our users report preferring to use our kernel module instead") given for not wanting to use FUSE is one "in-the-works" feature that could probably be integated into FUSE instead. ] This patch has been compile-tested. Signed-off-by: Jann Horn <jannh@google.com> --- fs/orangefs/devorangefs-req.c | 7 +++++++ fs/orangefs/orangefs-cache.c | 4 ++-- fs/orangefs/orangefs-kernel.h | 4 ++-- fs/orangefs/orangefs-utils.c | 4 ++-- fs/orangefs/xattr.c | 4 ++-- 5 files changed, 15 insertions(+), 8 deletions(-)