|Main index||Section 9||Options|
The crget() function allocates memory for a new structure, sets its reference count to 1, and initializes its lock.
The crhold() function increases the reference count on the credential.
The crfree() function decreases the reference count on the credential. If the count drops to 0, the storage for the structure is freed.
The crshared() function returns true if the credential is shared. A credential is considered to be shared if its reference count is greater than one.
The crcopy() function copies the contents of the source (template) credential into the destination template. The uidinfo structure within the destination is referenced by calling uihold(9).
The crcopysafe() function copies the current credential associated with the process p into the newly allocated credential cr. The process lock on p must be held and will be dropped and reacquired as needed to allocate group storage space in cr.
The crdup() function allocates memory for a new structure and copies the contents of cr into it. The actual copying is performed by crcopy().
The crsetgroups() function sets the cr_groups and cr_ngroups variables and allocates space as needed. It also truncates the group list to the current maximum number of groups. No other mechanism should be used to modify the cr_groups array except for updating the primary group via assignment to cr_groups.
The cru2x() function converts a ucred structure to an xucred structure. That is, it copies data from cr to xcr; it ignores fields in the former that are not present in the latter (e.g., cr_uidinfo), and appropriately sets fields in the latter that are not present in the former (e.g., cr_version).
crshared() returns 0 if the credential has a reference count greater than 1; otherwise, 1 is returned.
In the common case, credentials required for access control decisions are used in a read-only manner. In these circumstances, the thread credential td_ucred should be used, as it requires no locking to access safely, and remains stable for the duration of the call even in the face of a multi-threaded application changing the process credentials from another thread.
During a process credential update, the process lock must be held across check and update, to prevent race conditions. The process credential, td->td_proc->p_ucred, must be used both for check and update. If a process credential is updated during a system call and checks against the thread credential are to be made later during the same system call, the thread credential must also be refreshed from the process credential so as to prevent use of a stale value. To avoid this scenario, it is recommended that system calls updating the process credential be designed to avoid other authorization functions.
If temporarily elevated privileges are required for a thread, the thread credential can by replaced for the duration of an activity, or for the remainder of the system call. However, as a thread credential is often shared, appropriate care should be taken to make sure modifications are made to a writable credential through the use of crget() and crcopy().
Caution should be exercised when checking authorization for a thread or process perform an operation on another thread or process. As a result of temporary elevation, the target thread credential should never be used as the target credential in an access control decision: the process credential associated with the thread, td->td_proc->p_ucred, should be used instead. For example, p_candebug(9) accepts a target process, not a target thread, for access control purposes.
|UCRED (9)||January 23, 2019|
|Main index||Section 9||Options|
|“||What will happen when the 32-bit Unix date goes negative in mid-January 2038 does not bear thinking about.||”|
|— Henry Spencer|