Don't typecast cg_meta to uint8_t * and access it like an array. This is
not endian safe and breaks on big endian systems. Found the hard way by
denis@. Instead use the same shit and mask operations as used in the
other functions operating on cg_meta.
OK denis@
allow vlan tags (and therefore vlan interfaces) on top of vports.
bluhm@ really wants to test vlan interfaces on vports even though
veb as a vlan aware switch allows vports with the right "untagged"
pvid configured to do the same thing with less layering.
Implementation of a fast non-cryptographic hash function for small inputs
Based on HashLen0to16() from CityHash64. Useful for hashing a few ids
or pointer values that are local data into a hash. For such cases siphash
is overkill and this seems to be fast and good enough.
OK tb@
Modify the behavior of tmpfs limitations. Keep the global memory limit
of all tmpfs instances, however each instance mounted with '-s' option
decreases available limit by given value. This value can't be greater
than available space. The instance mounted with with '-s' option
consumes it's own limit and does no impact to other tmpfs instances. The
instances mounted without '-s' option share all available limit and do
not affect on instances mounted with '-s'. Now it is impossible to mount
instance with '-s' option if passed value is greater than available
limit.
Tested with dbp(1) by kirill, ok mpi
For example, this machine has 2G of RAM, so tmpfs is limited with 1G.
Please note, the +- 1Mb depends of how 'dd', 'df' and 'mount_tmpfs'
expose the sizes.
# mount_tmpfs tmpfs /mnt/t0
# mount_tmpfs tmpfs /mnt/t1
# mount_tmpfs -s 720M tmpfs /mnt/t2
[104 lines not shown]
openssl ts: simplify create_nonce()
Just your average dumb TS code. Instead of handrolling a random ASN.1
integer generator, we can use BN_rand() and convert the resulting bn to
an ASN1_INTEGER. All this then also works without reaching into ASN1_STRING.
ok kenjiro
add Private VLAN support to veb(4) as per RFC 5517
this extends the core etherbridge functionality so it stores the
extra information needed to be usable by a pvlan aware bridge. in
practice this means address entries in etherbridges store an extra
vlan id on top of the one i just added to make veb vlan aware. a
mac address remains scoped by what is now referred to as a "primary"
vlan id, but has an additional "secondary" vlan id recorded against
it. ie, the "key" to a address lookup in etherbridge is made up of
the mac address and the primary vlan id, and the value returned by
the lookup is the port the mac address should go to and the secondary
vlan id.
the non veb users of the etherbridge code just hard code these extra
vlan ids to 0.
because normal vlans function identically to a primary vlan in a
pvlan topology (or primary vlans in pvlan function like normal
vlans), the veb forwarding code pretends normal vlans are pvlans,
[20 lines not shown]
Update to the current version of the bnxt structure definitions from FreeBSD.
This is necessary for adding support for newer hardware generations and
should also help with work on LRO. This shouldn't result in any functional
changes yet however.
add a sshbuf_get_nulterminated_string() function to pull a \0-
terminated string from a sshbuf. Intended to be used to improve
parsing of SOCKS headers for dynamic forwarding.
ok deraadt; feedback Tim van der Molen
Determine how many queue pairs we have by looking at the I40E_PFLAN_QALLOC
register, rather than assuming we have the full capacity of the whole
chip, which is likely to be split among 2 or 4 functions.
ok jan@ dlg@
In the Adj-RIB-Out dmetric has no meaning (by default there is only one
path). Also nhflags are not needed any more since the nexthop selection
happens right before addition to the Adj-RIB-Out in up_prep_adjout().
OK tb@
Split up the PREFIX flags, split the name space and use PREFIX_ADJOUT
for those that only affect that struct.
In the process retire PREFIX_FLAG_ADJOUT.
OK tb@
time.h: const correct tm_zone in struct tm
The Base Specifications, Issue 8 incorporated Austin Group Defect 1533
adding long tm_gmtoff (which we already have) and const char *tm_zone,
which we have but without const. Adapt our struct. This was tested in
an amd64 bulk without fallout and fixes a bunch of compiler warnings
from strptime.c r1.32 where gmt[] and utc[] were moved to rodata.
https://austingroupbugs.net/view.php?id=1533
ok jca millert
Duplicate struct prefix into struct prefix_adjout and adjust code
to work with that.
Make sure that anything in the adj-rib-out uses the new struct. In
some cases code has to be duplicated since it is no longer possible to
use one code path to cover both Local-RIB and Adj-Out-RIB. This is
mostly in the code handling the 'bgpctl show rib' family of commands.
OK tb@
