HP OpenVMS Systems Documentation

All logical or physical I/O to a spooled device requires privilege.

The LOG_IO privilege allows the user's process to execute the Queue I/O Request ($QIO) system service to perform logical-level I/O operations. LOG_IO privilege is also required for certain device-control functions, such as setting permanent terminal elements.

The PHY_IO privilege allows the user's process to execute the Queue I/O Request ($QIO) system service to perform physical-level I/O operations. The PHY_IO privilege also grants LOG_IO privilege.

To create a permanent mailbox or mark it for deletion requires PRMMBX privilege.

5.3.7 Kinds of Auditing Performed

The following types of events can be audited, provided the security administrator enables auditing for the appropriate event class:

The profile of clusterwide disks and tapes is stored in the object database VMS$OBJECTS.DAT, but other object profiles have to be reset each time the system starts up.

5.4 Files

A file is a named array of fixed-size (512-byte) data blocks with an associated set of attributes. In OpenVMS systems, the file class includes both data files and directory files. The operating system provides full security protection for individual disk files stored on Files-11 On-Disk Structure Level 2 (ODS-2) volumes. Tape files are collectively protected by the protection code on the volume but are not protected on an individual basis.

The file object differs from other protected objects in one important way: because files provide more flexibility than any other object class, files do not acquire their profiles from a template. Section 5.4.5 describes the rules the operating system applies in assigning a profile.

5.4.1 Naming Rules

A file specification is a string of 1 to 255 characters. See the OpenVMS User's Manual for a full description.

5.4.2 Types of Access

The file class supports the following types of access:

The following conditions apply to file access:

• General rules
  To access a file, you must have permission to access the file and the volume on which it resides. When attempting to access a file by name, you must have read or execute access to the directory containing the file. An access check of the volume is required before either a directory or a file access is considered. The protection of a directory file can restrict access to files in the directory, so even though a group of users has access to a file, they can be prevented from accessing it by name if they lack proper access to the directory in which the file is located.

  Note You can access a file by its file identifier. When users do so, they bypass the directory file protection. Therefore, you must not rely entirely on directory file protection to control access to a file.

• For write access
  To write to a file, the operating system requires that you have both read and write access.
• File ownership changes
  To change the ownership of a file, you must have control access and hold both the old and new identifiers with the Resource attribute. (Your own UIC identifier always carries the Resource attribute.)

Before you can create a file, the operating system checks to see that you have satisfied the following conditions:

• You must have adequate disk space. This includes both available disk blocks and sufficient disk quota (assuming quotas are enabled).
• You have read and write access to the previous file version. You must also have delete access to the oldest file version if the file has a nonzero version limit and the new version would exceed this number.
• You have write access to the directory where the file is being created.
• You have read, write, and create access to the volume on which the file is to be stored.

The new file obtains its owner, protection code, and ACL from a number of sources. The ownership assignment of a new file is done independently of protection and ACL.

5.4.5.1 Rules for Assigning Ownership

If any of the following conditions are true, then you can assign an identifier as the owner of a file:

• The identifier matches your process UIC.
• You hold the identifier with the Resource attribute.
• You hold GRPPRV privilege and the identifier's group number matches your UIC group.
• You hold SYSPRV privilege.

A file receives its owner identifier from the first applicable source that you are allowed to assign:

• The explicit assignment of an owner at creation with the /OWNER_UIC qualifier to the CREATE or COPY command
• The previous version
• The parent directory
• The process UIC

See Section 8.8.1.2 for a description of how resource identifiers can own files and directories.

5.4.5.2 Rules for Assigning a Protection Code and ACL

The sources of a new file's protection code and ACL are similar to those of ownership and are considered in the same order. The system assigns a file's protection code and ACL from one of the following sources:

1. The explicit assignment of elements at creation
   You can create a file with the CREATE command or the COPY command. You use the CREATE/DIRECTORY command in the case of a directory.
   To assign a protection code when creating a file, add the /PROTECTION qualifier to the COPY or CREATE command. After creating the file, you can use the SECURITY/ACL command to add an ACL.
   For example, the following command copies a file from the device USE1 to the default disk directory. The protection code defines the protection for the newly created file PAYSORT.DAT so that users with system UICs can read and write to the file. The owner has all types of access, and other users in the owner's group can read and write to the file. All other users have no access through the protection code.

   $ COPY USE1:[PAYDATA]PAYROLL.DAT PAYSORT.DAT - _$ /PROTECTION=(SYSTEM:RW,OWNER:RWED,GROUP:RW,WORLD)

2. The profile of the previous version of the file, if one exists
   Whenever you create a new version of the file, the new version is created with the protection code and ACL of the earlier version (unless, of course, you make an explicit assignment).
3. A Default Protection ACE and Default ACL on the parent directory
   Without either an explicit assignment or a previous version of a file, the operating system looks at the directory where the file is being created.
   With data files, the system looks for a Default Protection ACE and assigns the protection code specified by that ACE. (See Section 4.5.6 for an example.) If any ACE in the directory's ACL has the Default attribute, then the file inherits that ACE as well. (Refer to Section 4.4.7 for an example.)
   With directory files, the system assigns the protection code of the parent directory, less any delete access. If the directory happens to be a top-level directory, the protection is taken from the master file directory (MFD). Newly created subdirectories inherit the ACL of the parent directory, even ACEs with the Default attribute. Only ACEs with the Nopropagate attribute are omitted.
4. The UIC and protection defaults of the process issuing the command
   If the directory ACL does not have a Default Protection ACE, the default process protection is used. The system parameter RMS_FILEPROT establishes this value, and the operating system assigns it to your process during login. However, the value derived at login may be changed with the DCL command SET PROTECTION/DEFAULT. (For example, you can put this command in your login command procedure to set default protection.) Use the DCL command SHOW PROTECTION to display the default process protection.
5. One of the above with provision for the user creating the file
   When you create a file in a directory owned by a resource identifier and you hold the identifier with the Resource attribute, the new file inherits its protection code and ACL in the same way as any other file.
   The operating system modifies the file's ACL in some cases to provide the creator with access to the new file. If the directory ACL has a Creator ACE, that ACE defines the access the creator has to the file. If the Creator ACE specifies no access, no additional ACE is created. Without such an ACE, the operating system adds an ACE to the file's ACL that gives the creator control access plus the access specified in the owner field of the file's protection code.

The output file of a COPY command is treated as a newly created file and so is assigned a new security profile. The security profiles of the input files are immaterial.

However, a renamed file by default retains its existing security profile. To assign a new security profile, as if the file were newly created, use the DCL command RENAME/INHERIT_SECURITY. This causes the file to be assigned a security profile.

Section 5.4.5.1 and Section 5.4.5.2 explain how a security profile is assigned.

5.4.6 Kinds of Auditing Performed

The following types of events can be audited, provided the security administrator enables auditing for the appropriate event class:

Ordinary file protection mechanisms control who can access a file, but they do not address the problem of protecting old data that remains on disk after a file is deleted.

When a file is deleted, its header is removed from the directory, but its contents remain intact on disk until it is overwritten. Because data exists on a disk, it is necessary to protect deleted or purged file information from disk scavenging.

The OpenVMS operating system solves the problem of disk scavenging with the combination of the two following techniques:

• Overwriting disk blocks before they are allocated
• Setting a high-water mark on allocated blocks

A security administrator or user can apply an erasure pattern to individual files on a volume or to a complete volume. An erasure pattern is a repeated sequence of bits written over a file when the file is deleted or purged.

The security administrator can ensure that every block on a volume starts off with the erasure pattern by specifying the /ERASE qualifier when the volume is initialized, as follows:

If the volume is mounted, the security administrator can automatically apply the erasure pattern to the space occupied by a file when it is deleted by specifying the /ERASE_ON_DELETE qualifier, as follows:

Note that this technique has no effect on existing files.

Alternatively, the security administrator may ask users to specify the erasure pattern on a file-by-file basis by using the /ERASE qualifier when entering the DCL commands SET FILE, DELETE, and PURGE.

Security administrators can also write an erase routine by using the $ERAPAT system service. The routine specifies to the system the erasure pattern and number of passes to be used to erase disk blocks.

5.4.7.2 Setting a High-water Mark

When the operating system allocates disk blocks for a file, it automatically sets a high-water mark. The high-water mark indicates how far the file has been written in its allotted space on the disk. All blocks in the file up to the high-water mark are guaranteed to have been written since they were allocated to the file. Users are not permitted to read beyond the high-water mark and thus cannot read stale data that they did not actually write.

A more conservative but costly technique is to erase all disk blocks before allocation. The erase-on-allocate technique is used when the file is open allowing any form of shared access or nonsequential access. When blocks are erased on allocation, the file's high-water mark is set to point to the end of the newly allocated and erased space.

By default, high-water marking is enabled when the volume is initialized. The security administrator can disable high-water marking for a specific volume by using the DCL command SET VOLUME/NOHIGHWATER_MARKING.

5.4.7.3 Accessibility of Data in a File

Once the file system allocates disk blocks for a file, users can read or write to them at any time. The high-water mark identifies the physical end of file, beyond which the user cannot read. However, an application can reposition the logical end-of-file mark and leave data in the area between the logical and the physical end of the file. Any block of file data can later be read, regardless of the logical end-of-file mark.

An application largely determines how allocated disk blocks are managed. For example, OpenVMS RMS services shorten a sequential file by resetting the logical end-of-file position to the beginning of the current record. It does not deallocate space between the end-of-file position and the physical end of the file, nor does it overwrite the records between the end-of-file position and the physical end of the file with an erase pattern.

Thus, blocks written to a file can remain available regardless of the end-of-file mark. If you want to erase the data between the logical end of the file and the physical end of the file, your application program must overwrite the data you want deleted. On OpenVMS systems, a common way to accomplish this is to create a new version of the file using the DCL command COPY.

5.4.8 Suggestions for Optimizing File Security

Use the following precautions to protect your files and directories:

• Purge your files regularly. Delete unnecessary files. This keeps your directories to a minimum and simplifies the task of regularly checking the protection and ownership on your files.
• Use the DCL command DIRECTORY/SECURITY regularly to monitor the ownership, protection code, and ACLs on your files. A user who succeeds in obtaining sufficient privilege may change the protection or ownership on your files, allowing access immediately and in the future. If you perform these checks frequently, you can detect and report unexplained changes in file protection or ownership.
• Pay special attention to the protection on your mail files; normally, they should be accessible only to you and the system (for mail delivery and backups).
• When you place ACLs on your files, be sure you know exactly which users hold the identifiers you have specified. (This generally requires consultation with your site security administrator.)
• Follow your site security administrator's recommendations to prevent disk scavenging. You may be requested to use the /ERASE qualifier on the SET FILE, DELETE, and PURGE commands for some or all of your files.
• Always protect files and directories that contain command procedures and executable programs. Carefully control the granting of write access to these directories and files. This is particularly important if you have any of the more powerful privileges or access to sensitive files.

OpenVMS memory management services allow processes to communicate through shared memory pages called global sections. Using global sections, two or more processes can map the same page into their individual virtual address spaces, thereby sharing the same page of code or data.

A global section can provide access to a disk file (called a file-backed global section), provide access to dynamically created storage (called a page file-backed global section), or provide access to specific physical memory (called a page frame number [PFN] global section). A global section object may be either temporary or permanent.

The operating system supports two types of global section objects:

• Group global sections are shareable memory sections potentially available to all processes in the same group.
• System global sections are shareable memory sections potentially available to all processes in the system.