The IFNET_LOCK was added to avoid data races on if_flags for IFF_ALLMULTI.
Unfortunatetly it caused a deadlock instead. A known scenario causing a
deadlock is to occur the following two operations concurrently: (a) a removal of
an IP adddres assigned to an interface and (b) a manipulation of multicast
groups to the interface. The resource dependency graph is like this:
softnet_lock => IFNET_LOCK => psref_target_destroy => softint => softnet_lock
Thanks to the previous commit that avoids data races on if_flags for
IFF_ALLMULTI by another approach, we can remove IFNET_LOCK and defuse the
deadlock.
PR kern/54189
Now, we have M_CSUM_TSOv[46] bit in ifp->if_csum_flags_tx when
TSO[46] is enabled for the interface. So we can simply check
whether TSO[46] is required in a packet but missing in the
interface by (sw_csum & M_CSUM_TSOv[46]).
Note that this is a very rare case where TSO[46] is suddenly
turned off during a packet passing b/w TCP and IP.
part of PR kern/53562
OK msaitoh
ip_undefer_csum -> in_undefer_cksum
in_delayed_cksum -> in_undefer_cksum_tcpudp
The two previous names were inconsistent and misleading.
Put the two functions into in_offload.c. Add comments to explain what
we're doing.
The same could be done for IPv6.
The check enforces a 4-byte-aligned size for the option mbuf. If the size
is not multiple of 4, the computation of ip_hl gets truncated in the
output path. There is no reason for this check not to be present on VAX.
While here add a KASSERT in ip_insertoptions to enforce the assumption.
Discussed briefly on tech-net@
ipsec4_get_policy and ipsec6_get_policy
ipsec4_delete_pcbpolicy and ipsec6_delete_pcbpolicy
The already-existing ipsec_get_policy() function is inlined in the new
one.
* Add a KASSERT in ip_output(), we expect (at least) the IP header to be
here.
* In ip_fragment(), declare two variables instead of recomputing the
values each time. Add an XXX for ipoff, it seems to me we should also
remove IP_RF.
* Rename the arguments of ip_optcopy().
* Style: use NULL for pointers, remove ()s for return statements, and
add whitespaces for clarity.
No real functional change.
2) Change the IP_RECVPKTINFO option to control the generation of
IP_PKTINFO control messages, the way it's done in Solaris.
3) Remove the superfluous IP_RECVPKTINFO control message.
4) Change the IP_PKTINFO option to do different things depending on
the parameter it's supplied with:
- If it's sizeof(int), assume it's being used as in Linux:
- If it's non-zero, turn on the IP_RECVPKTINFO option.
- If it's zero, turn off the IP_RECVPKTINFO option.
- If it's sizeof(struct in_pktinfo), assume it's being used as in
Solaris, to set a default for the source interface and/or
source address for outgoing packets on the socket.
5) Return what Linux or Solaris compatible code expects, depending
on data size, and just added a fallback to a Linux (and current NetBSD)
compatible value if the size is unknown (as it is now), or,
in the future, if the calling application specifies a receiving
buffer that doesn't match either data item.
From: Tom Ivar Helbekkmo
curlwp_bindx must be called in LIFO order, i.e., we can't call curlwp_bind
and curlwp_bindx like this:
bound1 = curlwp_bind();
bound2 = curlwp_bind();
curlwp_bindx(bound1);
curlwp_bindx(bound2);
ip_outout did so if NET_MPSAFE. Fix it.
It reduces C&P codes such as "#ifndef NET_MPSAFE KERNEL_LOCK(1, NULL); ..."
scattered all over the source code and makes it easy to identify remaining
KERNEL_LOCK and/or softnet_lock that are held even if NET_MPSAFE.
No functional change
The source address or output interface can be specified by adding IP_PKTINFO
to the control part of the message on a SOCK_DGRAM or SOCK_RAW socket.
Reviewed by ozaki-r@ and christos@. thanks.
rather than blindly picking the first matcing address from any interface
when testing source address validity.
This allows another interface to have the same address, but be detached.
christos