FS

F=S

---

Assume specified user -id and group -id

V8.9 and above

There are three major ways in which sendmail can be run: ^[33] as a set-user-id root process (that is, with the permissions of root regardless of who runs it), as a root process because it was run by root , or as an ordinary process run by an ordinary (nonprivileged) user. When sendmail is running with root privilege and when the F=S delivery agent flag is specified for a delivery agent, sendmail always invokes that delivery agent as the effective user and effective group specified by the U= delivery agent equate. ^[34] If the U= delivery agent equate is unspecified or is specified as zero, it runs as the effective user root . In both instances, the real user and real group IDs remain those of the recipient.

^[33] Beginning with V8.12, the default is to run sendmail as root only when it is executed by root (as it would be if executed from an rc boot-time script). Installing sendmail as a set-user-id root process is discouraged.

^[34] Prior to V8.7 there was no U= delivery agent equate, so F=S always ran as root .

If the F=S flag is omitted from the delivery agent, the following scenarios occur:

• If delivery is to a file, and if the set-user-id bit is set in the file's permission bits, and if the execute-bit is not set, sendmail sets its user and group identities to those of the owner and group of the file.

• Otherwise, if the set-user-id bit is not set, or if deliver is not to a file, and if there is a controlling user (C line) for the address, sendmail sets its identity to that of the controlling user for delivery.

• Otherwise, if the user or group part of the U= delivery agent equate was missing or 0, sendmail assumes the identity of the DefaultUser option (DefaultUser).

• Otherwise, sendmail assumes the identity of the U= delivery agent equate.

If it fails to set its identity, it prints and logs the following error:

 insufficient privileges to change gid, RealGid=  rgid  , RunAsUid=  ruid  , gid=  gid  , egid=  egid  

Note that this F=S flag was revised once for V8.7. Then it was revised again for V8.9, and has remained stable since.