Use stdio. Don't lock to read -- we only ever rename new ccaches into
place, or overwrite endtimes to delete entries, or overwrite part of the
realm name of cc config entries.
Dropping locks around ccache iterator stepping strongly implied that we
don't expect truncation, that we only expect appends (and the
overwriting done to delete entries).
Don't unlock -- let close(2) do it, thus making fewer system calls.
The ASN.1 functions copy_Realm(), copy_PrincipalName() and
copy_EncryptionKey() can fail. Check the return and perform error
handling as appropriate.
Change-Id: I2b3629d19db96eb41d1cd554cef1dca99745e753
When comparing the first component of a multi-component
string that uses a field separator (e.g., '/' or ':'), be sure
to include the separator in the comparison to ensure that the
comparison does not succeed due to a prefix match.
Change-Id: Ieec3e73cb015a83bb652ec11cd7d83e57755c3d2
Fix the numeric error value comparison performed by error_cmp().
Identified by AppChecker and previously fixed by Darwin.
Change-Id: I0a72fd381bef0b34b7e7d155bfff03ab4196d38e
if defined(HAVE_GETAUXVAL) && !defined(GETAUXVAL_SETS_ERRNO) the
stack variable declaration for 'a' is missing:
const auxv_t *a
Declare 'a' so that Heimdal can build again.
Change-Id: If64dd497f240688b76f3dd6139ac68c2ff149a0e
In 13c8a2c212 permits fclose() to
be issued on 'stdin' and forgets to use the N_() macro for text
strings.
Change-Id: I5e9ac1354da8dcff5277c39e4784a768ad76afdd
Though a race was never observed, the re-sync, slave DB deletion,
master changes and restart did not take into account late changes
in the master before slave down.
Creation of a random key HDB entry takes 3 updates, not one. Instead of
hard-coding this magic count, wait for the appearance of the expected
entry with no "invalid" flag among its attributes.
When obtaining a remote TGT krbtgt/REALM2@REALM2, an intermediate
cross-realm TGT obtained for krbtgt/REALM2@REALM1 is not equivalent
to the TGT we seek, and must not be stored under its name.
When the cred passed krb5_build_ap_req() has a different name for the actual
ticket (e.g., because the entry came from a ccache with an alias name as the
entry name) then we were putting a Ticket on the wire with the name from the
cred rather than from the Ticket in the cred. We don't think this is intended
or desirable. The server should see the Ticket _exactly_ as minted by the KDC.
Perhaps AP-REQ should have used an OCTET STRING to contain the Ticket given that
Ticket is a PDU, which would make a byte-for-byte copy trivial, but as it is it
uses Ticket instead. Therefore this commit has it decode the Ticket from the
cred and then C struct assign that value to the AP-REP's ticket field -- this
then copies the Ticket as exactly as vended.
Commit f469fc6 (2010-10-02) inadvertently caused the previous hop realm
to not be added to the transit path of issued tickets. This may, in
some cases, enable bypass of capath policy in Heimdal versions 1.5
through 7.2.
Note, this may break sites that rely on the bug. With the bug some
incomplete [capaths] worked, that should not have. These may now break
authentication in some cross-realm configurations.
If the file is not a regular file, the 'fname' buffer will be freed
before it is passed to the krb5_set_error_message() routine. That
this is true is not obvious from code inspection but 'newfname'
and 'fname' refer to the same memory.
Change-Id: I7780eae9ceaf01b245df488a2f7d7f406770864f
The output of `wc -l` includes leading white-space, and at least
in FreeBSD 11, `expr 1 + "$foo"` fails when "$foo" (captured as
`wc -l` output) has leading whitespace. Instead, just emit one
more line for "wc" to count.
