This is a sixth draft of an implementation of more general
"connection" tracking, as described at:
https://pad.passt.top/p/NewForwardingModel
This series changes the TCP connection table and hash table into a
more general flow table that can track other protocols as well. Each
flow uniformly keeps track of all the relevant addresses and ports,
which will allow for more robust control of NAT and port forwarding.
ICMP and UDP are converted to use the new flow table.
This is based on the most recent series of flowtable preliminaries.
NOTE: this fails the pasta UDP perf tests. This is one aspect of a
curly issue I'm not sure how to deail with yet. So, this isn't ready
for merge yet, but I'm posting because there have been a *lot* of
changes since v5, and it could do with another round of review,
Other caveats:
* We roughly double the size of a connection/flow entry
* UDP still has a number of forwarding bugs because we don't consider
the local address when looking for a socket to use. There are at
least some FIXMEs noted for this.
Changes since v5:
* flowside_from_af() is now static
* Small fixes to state verification
* Pass protocol specific types into deferred/timer callbacks
* No longer require complete forwarding address info for the hash
table (we won't have it for UDP)
* Fix bugs with logging of flow addresses
* Make sure to initialise sin_zero field sockaddr_from_inany
* Added patch better typing parameters to flow type specific callbacks
* Terminology change "forwarded side" to "target side"
* Assorted wording and style tweaks based on Stefano's review
* Fold introduction of struct flowside and populating the initiating
side together
* Manage outbound addresses via the flow table as well
* Support for UDP
* Correct type of 'b' in flowside_lookup() (was a signed int)
Changes since v4:
* flowside_from_af() no longer fills in unspecified addresses when
passed NULL
* Split and rename flow hash lookup function
* Clarified flow state transitions, and enforced where practical
* Made side 0 always the initiating side of a flow, rather than
letting the protocol specific code decide
* Separated pifs from flowside addresses to allow better structure
packing
Changes since v3:
* Complex rebase on top of the many things that have happened
upstream since v2.
* Assorted other changes.
* Replace TAPFSIDE() and SOCKFSIDE() macros with local variables.
Changes since v2:
* Cosmetic fixes based on review
* Extra doc comments for enum flow_type
* Rename flowside to flowaddrs which turns out to make more sense in
light of future changes
* Fix bug where the socket flowaddrs for tap initiated connections
wasn't initialised to match the socket address we were using in the
case of map-gw NAT
* New flowaddrs_from_sock() helper used in most cases which is cleaner
and should avoid bugs like the above
* Using newer centralised workarounds for clang-tidy issue 58992
* Remove duplicate definition of FLOW_MAX as maximum flow type and
maximum number of tracked flows
* Rebased on newer versions of preliminary work (ICMP, flow based
dispatch and allocation, bind/address cleanups)
* Unified hash table as well as base flow table
* Integrated ICMP
Changes since v1:
* Terminology changes
- "Endpoint" address/port instead of "correspondent" address/port
- "flowside" instead of "demiflow"
* Actually move the connection table to a new flow table structure in
new files
* Significant rearrangement of earlier patchs on top of that new
table, to reduce churn
David Gibson (26):
flow: Common address information for initiating side
flow: Common address information for target side
tcp, flow: Remove redundant information, repack connection structures
tcp: Obtain guest address from flowside
tcp: Manage outbound address via flow table
tcp: Simplify endpoint validation using flowside information
tcp_splice: Eliminate SPLICE_V6 flag
tcp, flow: Replace TCP specific hash function with general flow hash
flow, tcp: Generalise TCP hash table to general flow hash table
tcp: Re-use flow hash for initial sequence number generation
icmp: Remove redundant id field from flow table entry
icmp: Obtain destination addresses from the flowsides
icmp: Look up ping flows using flow hash
icmp: Eliminate icmp_id_map
icmp: Manage outbound socket address via flow table
flow, tcp: Flow based NAT and port forwarding for TCP
flow, icmp: Use general flow forwarding rules for ICMP
fwd: Update flow forwarding logic for UDP
udp: Create flow table entries for UDP
udp: Direct traffic from tap according to flow table
udp: Direct traffic from host to guest tap according to flow table
udp: Direct spliced traffic according to flow table
udp: Remove 'splicesrc' tracking
udp: Remove tap port flags field
udp: Remove rdelta port forwarding maps
udp: Eliminate 'splice' flag from epoll reference
Makefile | 2 +-
conf.c | 14 +-
flow.c | 402 +++++++++++++++++++++++++-
flow.h | 45 +++
flow_table.h | 45 ++-
fwd.c | 177 +++++++++++-
fwd.h | 9 +
icmp.c | 105 +++----
icmp_flow.h | 2 -
inany.h | 31 +-
passt.h | 3 +
tap.c | 11 -
tap.h | 1 -
tcp.c | 524 +++++++++-------------------------
tcp_buf.c | 6 +-
tcp_conn.h | 51 ++--
tcp_internal.h | 10 +-
tcp_splice.c | 98 +------
tcp_splice.h | 5 +-
udp.c | 757 +++++++++++++++++++++++--------------------------
udp.h | 22 +-
udp_flow.h | 25 ++
util.c | 6 +-
util.h | 3 +
24 files changed, 1328 insertions(+), 1026 deletions(-)
create mode 100644 udp_flow.h
--
2.45.2