The original patch for long division on x86 didn't take into account
the use of short circuit logic for checking if peer is NULL before
testing it. Here is a revised patch to v3.16.
get_seconds is removed in 5.11; its replacement ktime_get_real_seconds
is available since 3.19. The timestamps should not be affected by clock
resets, so will be switched to ktime_get_seconds.
The DEBUG definition in xt_pknock.h causes a compiler warning if one
adds a DEBUG define to xt_pknock.c to enable pr_debug. Since it only
controls some debugging output in libxt_pknock.c, it would make sense to
move the definition there, but let's just disable the debugging instead.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
Replace some instances of `printk(KERN_ERR PKNOCK ...)`. We define
`pr_fmt`, so `pr_err` is equivalent.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
The group ID used by xt_pknock is configurable, but pknlusr hard-codes
it. Modify pknlusr to accept an optional ID from the command line.
Group IDs range from 1 to 32 and each ID appears in the group bitmask
at position `group_id - 1`.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
A return-value of zero is not an error, so there is no point calling
perror, but since we have not requested and do not expect a zero-length
datagram, we treat it as EOF and exit.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
Use struct initialization and drop memset. We do not need to set the port
ID, since the kernel will do it for us.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
We only need to specify the address at our end, and given that we are
receiving messages, not sending them, calling it `src_addr` is
misleading.
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
shash_desc::flags was removed from the kernel in 5.1.
That assignment was actually superfluous anyway, because crypto.desc
is zero-initialized when crypto is initialized (xt_pknock.c, ll.
110ff.).
Signed-off-by: Jeremy Sowden <jeremy@azazel.net>
When the PK_CRYPTO pre-processor flag got removed in
v1.47.1-2-g66f213e, one of the removal cases was misapplied; the body
of an "#ifndef PK_CRYPTO" was left in rather than the whole section
being removed.
