Re: [PATCHv2 2/2] userns: control capabilities of some user namespaces

[Mahesh Bandewar (àààà ààààààà)]

On Sat, Nov 25, 2017 at 10:40 PM, Serge E. Hallyn <serge@xxxxxxxxxx> wrote:
> Quoting Mahesh Bandewar (mahesh@xxxxxxxxxxxx):
>> From: Mahesh Bandewar <maheshb@xxxxxxxxxx>
>>
>> With this new notion of "controlled" user-namespaces, the controlled
>> user-namespaces are marked at the time of their creation while the
>> capabilities of processes that belong to them are controlled using the
>> global mask.
>>
>> Init-user-ns is always uncontrolled and a process that has SYS_ADMIN
>> that belongs to uncontrolled user-ns can create another (child) user-
>> namespace that is uncontrolled. Any other process (that either does
>> not have SYS_ADMIN or belongs to a controlled user-ns) can only
>> create a user-ns that is controlled.
>>
>> global-capability-whitelist (controlled_userns_caps_whitelist) is used
>> at the capability check-time and keeps the semantics for the processes
>> that belong to uncontrolled user-ns as it is. Processes that belong to
>> controlled user-ns however are subjected to different checks-
>>
>>    (a) if the capability in question is controlled and process belongs
>>        to controlled user-ns, then it's always denied.
>>    (b) if the capability in question is NOT controlled then fall back
>>        to the traditional check.
>>
>> Signed-off-by: Mahesh Bandewar <maheshb@xxxxxxxxxx>
>
> Acked-by: Serge Hallyn <serge@xxxxxxxxxx>
>
> Although a few comment addition requests below:
>
>> ---
>> v2:
>>   Don't recalculate user-ns flags for every setns() call.
>> v1:
>>   Initial submission.
>>
>>  include/linux/capability.h     |  1 +
>>  include/linux/user_namespace.h | 20 ++++++++++++++++++++
>>  kernel/capability.c            |  5 +++++
>>  kernel/user_namespace.c        |  4 ++++
>>  security/commoncap.c           |  8 ++++++++
>>  5 files changed, 38 insertions(+)
>>
>> diff --git a/include/linux/capability.h b/include/linux/capability.h
>> index 7d79a4689625..a1fd9e460379 100644
>> --- a/include/linux/capability.h
>> +++ b/include/linux/capability.h
>> @@ -251,6 +251,7 @@ extern bool ptracer_capable(struct task_struct *tsk, struct user_namespace *ns);
>>  extern int get_vfs_caps_from_disk(const struct dentry *dentry, struct cpu_vfs_cap_data *cpu_caps);
>>  int proc_douserns_caps_whitelist(struct ctl_table *table, int write,
>>                                void __user *buff, size_t *lenp, loff_t *ppos);
>
> Here and at the definition below, please add a comment explaining
> that a controlled cap is defined as not being in the sysctl.
>
will do in v3.

>> +bool is_capability_controlled(int cap);
>>
>>  extern int cap_convert_nscap(struct dentry *dentry, void **ivalue, size_t size);
>>
>> diff --git a/include/linux/user_namespace.h b/include/linux/user_namespace.h
>> index 3fe714da7f5a..647f825c7b5f 100644
>> --- a/include/linux/user_namespace.h
>> +++ b/include/linux/user_namespace.h
>> @@ -23,6 +23,7 @@ struct uid_gid_map {        /* 64 bytes -- 1 cache line */
>>  };
>>
>>  #define USERNS_SETGROUPS_ALLOWED 1UL
>> +#define USERNS_CONTROLLED     2UL
>>
>>  #define USERNS_INIT_FLAGS USERNS_SETGROUPS_ALLOWED
>>
>> @@ -103,6 +104,16 @@ static inline void put_user_ns(struct user_namespace *ns)
>>               __put_user_ns(ns);
>>  }
>>
>
> Please add a comment explaining that a controlled ns
> is one created by a user which did not have CAP_SYS_ADMIN
> (or descended from such an ns).
>
will do in v3.

>> +static inline bool is_user_ns_controlled(const struct user_namespace *ns)
>> +{
>> +     return ns->flags & USERNS_CONTROLLED;
>> +}
>> +
>> +static inline void mark_user_ns_controlled(struct user_namespace *ns)
>> +{
>> +     ns->flags |= USERNS_CONTROLLED;
>> +}
>> +
>>  struct seq_operations;
>>  extern const struct seq_operations proc_uid_seq_operations;
>>  extern const struct seq_operations proc_gid_seq_operations;
>> @@ -161,6 +172,15 @@ static inline struct ns_common *ns_get_owner(struct ns_common *ns)
>>  {
>>       return ERR_PTR(-EPERM);
>>  }
>> +
>> +static inline bool is_user_ns_controlled(const struct user_namespace *ns)
>> +{
>> +     return false;
>> +}
>> +
>> +static inline void mark_user_ns_controlled(struct user_namespace *ns)
>> +{
>> +}
>>  #endif
>>
>>  #endif /* _LINUX_USER_H */
>> diff --git a/kernel/capability.c b/kernel/capability.c
>> index 4a859b7d4902..bffe249922de 100644
>> --- a/kernel/capability.c
>> +++ b/kernel/capability.c
>> @@ -511,6 +511,11 @@ bool ptracer_capable(struct task_struct *tsk, struct user_namespace *ns)
>>  }
>>
>>  /* Controlled-userns capabilities routines */
>> +bool is_capability_controlled(int cap)
>> +{
>> +     return !cap_raised(controlled_userns_caps_whitelist, cap);
>> +}
>> +
>>  #ifdef CONFIG_SYSCTL
>>  int proc_douserns_caps_whitelist(struct ctl_table *table, int write,
>>                                void __user *buff, size_t *lenp, loff_t *ppos)
>> diff --git a/kernel/user_namespace.c b/kernel/user_namespace.c
>> index c490f1e4313b..600c7dcb9ff7 100644
>> --- a/kernel/user_namespace.c
>> +++ b/kernel/user_namespace.c
>> @@ -139,6 +139,10 @@ int create_user_ns(struct cred *new)
>>               goto fail_keyring;
>>
>>       set_cred_user_ns(new, ns);
>> +     if (!ns_capable(parent_ns, CAP_SYS_ADMIN) ||
>> +         is_user_ns_controlled(parent_ns))
>> +             mark_user_ns_controlled(ns);
>> +
>>       return 0;
>>  fail_keyring:
>>  #ifdef CONFIG_PERSISTENT_KEYRINGS
>> diff --git a/security/commoncap.c b/security/commoncap.c
>> index fc46f5b85251..89103f16ac37 100644
>> --- a/security/commoncap.c
>> +++ b/security/commoncap.c
>> @@ -73,6 +73,14 @@ int cap_capable(const struct cred *cred, struct user_namespace *targ_ns,
>>  {
>>       struct user_namespace *ns = targ_ns;
>>
>> +     /* If the capability is controlled and user-ns that process
>> +      * belongs-to is 'controlled' then return EPERM and no need
>> +      * to check the user-ns hierarchy.
>> +      */
>> +     if (is_user_ns_controlled(cred->user_ns) &&
>> +         is_capability_controlled(cap))
>> +             return -EPERM;
>
> I'd be curious to see the performance impact on this on a regular
> workload (kernel build?) in a controlled ns.
>
Should it affect? If at all, it should be +ve since, the recursive
user-ns hierarchy lookup is avoided with the above check if the
capability is controlled. The additional cost otherwise is this check
per cap_capable() call.

>> +
>>       /* See if cred has the capability in the target user namespace
>>        * by examining the target user namespace and all of the target
>>        * user namespace's parents.
>> --
>> 2.15.0.448.gf294e3d99a-goog