Both the events and flags fields in tcp_splice_conn have several bits
which are per-side, e.g. OUT_WAIT_0 for side 0 and OUT_WAIT_1 for side 1.
This necessitates some rather awkward ternary expressions when we need
to get the relevant bit for a particular side.
Simplify this by using a parameterised macro for the bit values. This
needs a ternary expression inside the macros, but makes the places we use
it substantially clearer.
That simplification in turn allows us to use a loop across each side to
implement several things which are currently open coded to do equivalent
things for each side in turn.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We have a handful of places where we use a loop to step through each side
of a flow or flows, and we're probably going to have mroe in future.
Introduce a macro to implement this loop for convenience.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
In various places we have variables named 'side' or similar which always
have the value 0 or 1 (INISIDE or TGTSIDE). Given a flow, this refers to
a specific side of it. Upcoming flow table work will make it more useful
for "side" to refer to a specific side of a specific flow. To make things
less confusing then, prefer the name term "side index" and name 'sidei' for
variables with just the 0 or 1 value.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: Fixed minor detail in comment to struct flow_common]
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
TCP (both regular and spliced) and ICMP both have macros to retrieve the
relevant protcol specific flow structure from a flow index. In most cases
what we actually want is to get the specific flow from a sidx. Replace
those simple macros with a more precise inline, which also asserts that
the flow is of the type we expect.
While we're they're also add a pif_at_sidx() helper to get the interface of
a specific flow & side, which is useful in some places.
Finally, fix some minor style issues in the comments on some of the
existing sidx related helpers.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Currently we ignore all events other than EPOLLIN on UDP sockets. This
means that if we ever receive an EPOLLERR event, we'll enter an infinite
loop on epoll, because we'll never do anything to clear the error.
Luckily that doesn't seem to have happened in practice, but it's certainly
fragile. Furthermore changes in how we handle UDP sockets with the flow
table mean we will start receiving error events.
Add handling of EPOLLERR events. For now we just read the error from the
error queue (thereby clearing the error state) and print a debug message.
We can add more substantial handling of specific events in future if we
want to.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Allow sockaddr_ntop() to format AF_UNSPEC socket addresses. There do exist
a few cases where we might legitimately have either an AF_UNSPEC or a real
address, such as the origin address from MSG_ERRQUEUE. Even in cases where
we shouldn't get an AF_UNSPEC address, formatting it is likely to make
things easier to debug if we ever somehow do.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We abort the UDP socket handler if the no_udp flag is set. But if UDP
was disabled we should never have had a UDP socket to trigger the handler
in the first place. If we somehow did, ignoring it here isn't really going
to help because aborting without doing anything is likely to lead to an
epoll loop. The same is the case for the TCP socket and timer handlers and
the no_tcp flag.
Change these checks on the flag to ASSERT()s. Similarly add ASSERT()s to
several other entry points to the protocol specific code which should never
be called if the protocol is disabled.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Through an oversight this was previously declared as a public function
although it's only used in udp.c and there is no prototype in any header.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
UDP and/or TCP can be disabled with the --no-udp and --no-tcp options.
However, when this is specified, it's still possible to configure forwarded
ports for the disabled protocol. In some cases this will open sockets and
perform other actions, which might not be safe since the entire protocol
won't be initialised.
Check for this case, and explicitly forbid it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
A bug in kernel TCP may lead to a deadlock where a zero window is sent
from the guest peer, while it is unable to send out window updates even
after socket reads have freed up enough buffer space to permit a larger
window. In this situation, new window advertisements from the peer can
only be triggered by data packets arriving from this side.
However, currently such packets are never sent, because the zero-window
condition prevents this side from sending out any packets whatsoever
to the peer.
We notice that the above bug is triggered *only* after the peer has
dropped one or more arriving packets because of severe memory squeeze,
and that we hence always enter a retransmission situation when this
occurs. This also means that the implementation goes against the
RFC-9293 recommendation that a previously advertised window never
should shrink.
RFC-9293 seems to permit that we can continue sending up to the right
edge of the last advertised non-zero window in such situations, so that
is what we do to resolve this situation.
It turns out that this solution is extremely simple to implememt in the
code: We just omit to save the advertised zero-window when we see that
it has shrunk, i.e., if the acknowledged sequence number in the
advertisement message is lower than that of the last data byte sent
from our side.
When that is the case, the following happens:
- The 'retr' flag in tcp_data_from_tap() will be 'false', so no
retransmission will occur at this occasion.
- The data stream will soon reach the right edge of the previously
advertised window. In fact, in all observed cases we have seen that
it is already there when the zero-advertisement arrives.
- At that moment, the flags STALLED and ACK_FROM_TAP_DUE will be set,
unless they already have been, meaning that only the next timer
expiration will open for data retransmission or transmission.
- When that happens, the memory squeeze at the guest will normally have
abated, and the data flow can resume.
It should be noted that although this solves the problem we have at
hand, it is a work-around, and not a genuine solution to the described
kernel bug.
Suggested-by: Stefano Brivio <sbrivio@redhat.com>
Signed-off-by: Jon Maloy <jmaloy@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
[sbrivio: Minor fix in commit title and commit reference in comment
to workaround
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
>From linux-6.9.0 the kernel will contain
commit 05ea491641d3 ("tcp: add support for SO_PEEK_OFF socket option").
This new feature makes is possible to call recv_msg(MSG_PEEK) and make
it start reading data from a given offset set by the SO_PEEK_OFF socket
option. This way, we can avoid repeated reading of already read bytes of
a received message, hence saving read cycles when forwarding TCP
messages in the host->name space direction.
In this commit, we add functionality to leverage this feature when
available, while we fall back to the previous behavior when not.
Measurements with iperf3 shows that throughput increases with 15-20
percent in the host->namespace direction when this feature is used.
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Jon Maloy <jmaloy@redhat.com>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
This test program checks for particular behaviour regardless of order of
operations. So, we step through the test with all possible orders for
a number of different of parts. Or at least, we're supposed to, a copy
pasta error led to using the same order for two things which should be
independent.
Fixes: 299c407501 ("doc: Add program to document and test assumptions about SO_REUSEADDR")
Reported-by: David Taylor <davidt@yadt.co.uk>
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Add a test program verifying that we're able to discard datagrams from a
socket without needing a big discard buffer, by using a zero length recv().
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
For the approach we intend to use for handling UDP flows, we have some
pretty specific requirements about how SO_REUSEADDR works with UDP sockets.
Specifically SO_REUSEADDR allows multiple sockets with overlapping bind()s,
and therefore there can be multiple sockets which are eligible to receive
the same datagram. Which one will actually receive it is important to us.
Add a test program which verifies things work the way we expect, which
documents what those expectations are in the process.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
When we receive datagrams on a socket, we need to split them into batches
depending on how they need to be forwarded (either via a specific splice
socket, or via tap). The logic to do this, is somewhat awkwardly split
between udp_buf_sock_handler() itself, udp_splice_send() and
udp_tap_send().
Move all the batching logic into udp_buf_sock_handler(), leaving
udp_splice_send() to just send the prepared batch. udp_tap_send() reduces
to just a call to tap_send_frames() so open-code that call in
udp_buf_sock_handler().
This will allow separating the batching logic from the rest of the datagram
forwarding logic, which we'll need for upcoming flow table support.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
udp_buf_sock_handler(), udp_splice_send() and udp_tap_send loosely, do four
things between them:
1. Receive some datagrams from a socket
2. Split those datagrams into batches depending on how they need to be
sent (via tap or via a specific splice socket)
3. Prepare buffers for each datagram to send it onwards
4. Actually send it onwards
Split (1) and (3) into specific helper functions. This isn't
immediately useful (udp_splice_prepare(), in particular, is trivial),
but it will make further reworks clearer.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Since we split our packet frame buffers into different pieces, we have
a single buffer per IP version for the ethernet header, rather than one
per frame. This makes sense since our ethernet header is alwaus the same.
However we initialise those buffers udp[46]_eth_hdr inside a per frame
loop. Pull that outside the loop so we just initialise them once.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
The only differences between these arrays are that udp4_l2_iov is
pre-initialised to point to the IPv4 ethernet header, and IPv4 per-frame
header and udp6_l2_iov points to the IPv6 versions.
We already have to set up a bunch of headers per-frame, including updating
udp[46]_l2_iov[i][UDP_IOV_PAYLOAD].iov_len. It makes more sense to adjust
the IOV entries to point at the correct headers for the frame than to have
two complete sets of iovecs.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
We have separate mmsghdr arrays for splicing IPv4 and IPv6 packets, where
the only difference is that they point to different sockaddr buffers for
the destination address.
Unify these by having the common array point at a sockaddr_inany as the
address. This does mean slightly more work when we're about to splice,
because we need to write the whole socket address, rather than just the
port. However it removes 32 mmsghdr structures and we're going to need
more flexibility constructing that target address for the flow table.
Because future changes might mean that the address isn't always loopback,
change the name of the common address from *_localname to udp_splicename.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Make the salient points about these various arrays clearer with renames:
* udp_l2_iov_sock and udp[46]_l2_mh_sock don't really have anything to do
with L2. They are, however, specific to receiving not sending. Rename
to udp_iov_recv and udp[46]_mh_recv.
* udp[46]_l2_iov_tap is redundant - "tap" implies L2 and vice versa.
Rename to udp[46]_l2_iov
* udp[46]_localname are (for now) pre-populated with the local address but
the more salient point is that these are the destination address for the
splice arrays. Rename to udp[46]_splice_to
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
udp_buf_sock_handler() takes the epoll reference from the receiving socket,
and passes the UDP relevant part on to several other functions. Future
changes are going to need several different epoll types for UDP, and to
pass that information through to some of those functions. To avoid extra
noise in the patches making the real changes, change those functions now
to take the full epoll reference, rather than just the UDP part.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
To implement the TCP hash table, we need an invalid (NULL-like) value for
flow_sidx_t. We use FLOW_SIDX_NONE for that, but for defensiveness, we
treat (usually) anything with an out of bounds flow index the same way.
That's not always done consistently though. In flow_at_sidx() we open code
a check on the flow index. In tcp_hash_probe() we instead compare against
FLOW_SIDX_NONE, and in some other places we use the fact that
flow_at_sidx() will return NULL in this case, even if we don't otherwise
need the flow it returns.
Clean this up a bit, by adding an explicit flow_sidx_valid() test function.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
sock_l4() creates a socket of the given IP protocol number, and adds it to
the epoll state. Currently it determines the correct tag for the epoll
data based on the protocol. However, we have some future cases where we
might want different semantics, and therefore epoll types, for sockets of
the same protocol. So, change sock_l4() to take the epoll type as an
explicit parameter, and determine the protocol from that.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
UNIX_SOCK_MAX is the maximum number we'll append to the socket path
if we generate it automatically. If it's given on the command line,
it can be up to UNIX_PATH_MAX (including the terminating character)
long.
UNIX_SOCK_MAX happened to kind of fit because it's 100 (instead of
108).
Commit ceddcac74a ("conf, tap: False "Buffer not null terminated"
positives, CWE-170") fixed the wrong problem: the right fix for the
problem at hand was actually commit cc287af173 ("conf: Fix
incorrect bounds checking for sock_path parameter").
Fixes: ceddcac74a ("conf, tap: False "Buffer not null terminated" positives, CWE-170")
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Spotted by Coverity, harmless as we would consider that successful
and check on the socket later from the timer, but printing a debug
message in that case is definitely wise, should it ever happen.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Spotted by Coverity just recently. Not that it really matters as
MAXDNSRCH always appears to be defined as 1025, while a full domain
name can have up to 253 characters: it would be a bit pointless to
have a longer search domain.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
cppcheck 2.14 warns that the scope of the rport variable could be
reduced: do that, as reverted commit c80fa6a6bb ("udp: Make rport
calculation more local") did, but keep the temporary variable of
in_port_t type, otherwise the sum gets promoted to int.
While at it, add a comment explaining why we calculate rport like
this instead of directly using the sum as array index.
Reported-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
This reverts commit c80fa6a6bb, as it
reintroduces the issue fixed by commit 1e6f92b995 ("udp: Fix 16-bit
overflow in udp_invert_portmap()").
Reported-by: Laurent Jacquot <jk@lutty.net>
Link: https://bugs.passt.top/show_bug.cgi?id=80
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
If we daemonised, we can't use standard error. If we didn't, it's
rather annoying to have all those messages on standard error anyway,
and kind of pointless too, as the messages we wanted to print were
printed to standard error anyway.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
If a log file is configured, we would otherwise open a connection to
the system logger (if any), print any message that we might have
before we initialise the log file, and then keep that connection
around for no particular reason.
Call __openlog() as an alternative to the log file setup, instead.
This way, we might skip printing some messages during the
initialisation phase, but they're probably not really valuable to
have in a system log, and we're going to print them to standard
error anyway.
Suggested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Now that we have logging functions embedding perror() functionality,
we can make _some_ calls more terse by using them. In many places,
the strerror() calls are still more convenient because, for example,
they are used in flow debugging functions, or because the return code
variable of interest is not 'errno'.
While at it, convert a few error messages from a scant perror style
to proper failure descriptions.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
perror() prints directly to standard error, but in many cases standard
error might be already closed, or we might want to skip logging, based
on configuration. Our logging functions provide all that.
While at it, make errors more descriptive, replacing some of the
existing basic perror-style messages.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
In many places, we have direct perror() calls, which completely bypass
logging functions and log files.
They are definitely convenient: offer similar convenience with
_perror() logging variants, so that we can drop those direct perror()
calls.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
After commit 15001b39ef ("conf: set the log level much earlier"), we
had a phase during initialisation when messages wouldn't be printed to
standard error anymore.
Commit f67238aa86 ("passt, log: Call __openlog() earlier, log to
stderr until we detach") fixed that, but only for the case where no
log files are given.
If a log file is configured, vlogmsg() will not call passt_vsyslog(),
but during initialisation, LOG_PERROR is set, so to avoid duplicated
prints (which would result from passt_vsyslog() printing to stderr),
we don't call fprintf() from vlogmsg() either.
This is getting a bit too complicated. Instead of abusing LOG_PERROR,
define an internal logging flag that clearly represents that we're not
done with the initialisation phase yet.
If this flag is not set, make sure we always print to stderr, if the
log mask matches.
Reported-by: Yalan Zhang <yalzhang@redhat.com>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
We currently use a LOG_EMERG log mask to represent the fact that we
don't know yet what the mask resulting from configuration should be,
before the command line is parsed.
However, we have the necessity of representing another phase as well,
that is, configuration is parsed but we didn't daemonise yet, or
we're not ready for operation yet. The next patch will add that
notion explicitly.
Mapping these cases to further log levels isn't really practical.
Introduce boolean log flags to represent them, instead of abusing
log priorities.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
The original behaviour of printing messages to standard error by
default when running from a non-interactive terminal was introduced
because the first KubeVirt integration draft used to start passt in
foreground and get messages via standard error.
For development purposes, the system logger was more convenient at
that point, and passt was running from interactive terminals only if
not started by the KubeVirt integration.
This behaviour was introduced by 84a62b79a2 ("passt: Also log to
stderr, don't fork to background if not interactive").
Later, I added command-line options in 1e49d194d0 ("passt, pasta:
Introduce command-line options and port re-mapping") and accidentally
reversed this condition, which wasn't a problem as --stderr could
force printing to standard error anyway (and it was used by KubeVirt).
Nowadays, the KubeVirt integration uses a log file (requested via
libvirt configuration), and the same applies for Podman if one
actually needs to look at runtime logs. There are no use cases left,
as far as I know, where passt runs in foreground in non-interactive
terminals.
Seize the chance to reintroduce some sanity here. If we fork to
background, standard error is closed, so --stderr is useless in that
case.
If we run in foreground, there's no harm in printing messages to
standard error, and that accidentally became the default behaviour
anyway, so --stderr is not needed in that case.
It would be needed for non-interactive terminals, but there are no
use cases, and if there were, let's log to standard error anyway:
the user can always redirect standard error to /dev/null if needed.
Before we're up and running, we need to print to standard error anyway
if something happens, otherwise we can't report failure to start in
any kind of usage, stand-alone or in integrations.
So, make --stderr do nothing, and deprecate it.
While at it, drop a left-over comment about --foreground being the
default only for interactive terminals, because it's not the case
anymore.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
If we don't run in foreground, we close standard error as we
daemonise, so it makes no sense to check if the controlling terminal
is an interactive terminal or if --force-stderr was given, to decide
if we want to log to standard error.
Make --force-stderr depend on --foreground.
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
In multiple occasions, especially when passt(1) and pasta(1) are used
in integrations such as the one with Podman, the ability to override
earlier options on the command line with later one would have been
convenient.
Recently, to debug a number of issues happening with Podman, I would
have liked to ask users to share a debug log by passing --debug as
additional option, but pasta refuses --quiet (always passed by Podman)
and --debug at the same time.
On top of this, Podman lets users specify other pasta options in its
containers.conf(5) file, as well as on the command line.
The options from the configuration files are appended together with
the ones from the command line, which makes it impossible for users to
override options from the configuration file, if duplicated options
are refused, unless Podman takes care of sorting them, which is
clearly not sustainable.
For --debug and --trace, somebody took care of this on Podman side at:
https://github.com/containers/common/pull/2052
but this doesn't fix the issue with other options, and we'll have
anyway older versions of Podman around, too.
I think there's some value in telling users about duplicated or
conflicting options, because that might reveal issues in integrations
or accidental misconfigurations, but by now I'm fairly convinced that
the downsides outweigh this.
Drop checks about duplicate options and mutually exclusive ones. In
some cases, we need to also undo a couple of initialisations caused
by earlier options, but this looks like a simplification, overall.
Notable exception: --stderr still conflicts with --log-file, because
users might have the expectation that they don't actually conflict.
But they do conflict in the existing implementation, so it's safer
to make sure that the users notice that.
Suggested-by: Paul Holzinger <pholzing@redhat.com>
Suggested-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Tested-by: Paul Holzinger <pholzing@redhat.com>
If routing daemons set up host routes, for example FRR via OSPF as in
the reported issue, they might add nexthop identifiers (not objects)
that are generally not valid in the target namespace. Strip them off
as well, otherwise we'll get EINVAL from the kernel.
Link: https://github.com/containers/podman/issues/22960
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
The SPDX identifier states GPL-2.0-or-later but the copyright section
mentions GPL-3.0 or later causing a mismatch.
Also, only correctly refers to GPL instead of AGPL.
Signed-off-by: Danish Prakash <contact@danishpraka.sh>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
To implicitly resolve possible dependencies between routes as we
duplicate them into the target namespace, we go through a set of n
routes n times, and ignore EEXIST responses to netlink messages (we
already inserted the route) and ENETUNREACH (we didn't insert the
route yet, but we need to insert another one first).
Until now, we didn't ignore EHOSTUNREACH responses. However,
NetworkManager users with multiple non-subnet routes for the same
interface report that pasta exits with "no route to host" while
duplicating routes.
This happens because NetworkManager sets the 'noprefixroute' attribute
on addresses, meaning that the kernel won't create subnet routes
automatically depending on the prefix length of the address. We copy
this attribute as we copy the address into the target namespace, and
as a result, the kernel doesn't create subnet routes in the target
namespace either.
This means that the gateway for routes that are inserted later can be
unreachable at some points during the sequence of route duplication.
That is, we don't just have dependencies between regular routes, but
we can also have dependencies between regular routes and subnet
routes, as subnet routes are not automatically inserted in advance.
Link: https://github.com/containers/podman/issues/22824
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
While commit f919dc7a4b ("conf, netlink: Don't require a default
route to start") sounded reasonable in the assumption that, if we
don't find default routes for a given address family, we can still
proceed by selecting an interface with any route *iff it's the only
one for that protocol family*, Jelle reported a further issue in a
similar setup.
There, multiple interfaces are present, and while remote container
connectivity doesn't matter for the container, local connectivity is
desired. There are no default routes, but those multiple interfaces
all have non-default routes, so we should just pick one and start.
Pick the first interface reported by the kernel with any route, if
there are no default routes. There should be no harm in doing so.
Reported-by: Jelle van der Waa <jvanderwaa@redhat.com>
Reported-by: Martin Pitt <mpitt@redhat.com>
Link: https://bugzilla.redhat.com/show_bug.cgi?id=2277954
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Reviewed-by: Paul Holzinger <pholzing@redhat.com>
Commit e1a2e2780c ("tcp: Check if connection is local or low RTT
was seen before using large MSS") added a call to bind() before we
issue a connect() to the target for an outbound connection.
If bind() fails, but neither with EADDRNOTAVAIL, nor with EACCESS, we
can conclude that the target address is a local (host) address, and we
can use an unlimited MSS.
While at it, according to the reasoning of that commit, if bind()
succeeds, we would know right away that nobody is listening at that
(local) address and port, and we don't even need to call connect(): we
can just fail early and reset the connection attempt.
But if non-local binds are enabled via net.ipv4.ip_nonlocal_bind or
net.ipv6.ip_nonlocal_bind sysctl, binding to a non-local address will
actually succeed, so we can't rely on it to fail in general.
The visible issue with the existing behaviour is that we would reset
any outbound connection to non-local addresses, if non-local binds are
enabled.
Keep the significant optimisation for local addresses along with the
bind() call, but if it succeeds, don't draw any conclusion: close the
socket, grab another one, and proceed normally.
This will incur a small latency penalty if non-local binds are
enabled (we'll likely fetch an existing socket from the pool but
additionally call close()), or if the target is local but not bound:
we'll need to call connect() and get a failure before relaying that
failure back.
Link: https://github.com/containers/podman/issues/23003
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
In fc8f0f8c ("siphash: Use incremental rather than all-at-once siphash
functions") we removed the older interface to the SipHash implementation,
which took fixed sized blocks of data. However, we forgot to remove the
prototypes for those functions, so do that now.
Fixes: fc8f0f8c48 ("siphash: Use incremental rather than all-at-once siphash functions")
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
Mostly, udp_sock_handler() is independent of how the datagrams it processes
will be forwarded (tap or splice). However, it also updates the msg_name
fields for spliced sends, which doesn't really make sense here. Move it
into udp_splice_send() which is all about spliced sends. This does
potentially mean we'll update the field to the same value several times,
but we're going to need this in future anyway: with the extensions the
flow table allows, it might not be the same value each time after all.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
udp_sock_handler() takes a number of datagrams from sockets that depending
on their addresses could be forwarded either to the L2 interface ("tap")
or to another socket ("spliced"). In the latter case we can also only
send packets together if they have the same source port, and therefore
are sent via the same socket.
To reduce the total number of system calls we gather contiguous batches of
datagrams with the same destination interface and socket where applicable.
The determination of what the target is is made by udp_mmh_splice_port().
It returns the source port for splice packets and -1 for "tap" packets.
We find batches by looking ahead in our queue until we find a datagram
whose "splicefrom" port doesn't match the first in our current batch.
udp_mmh_splice_port() is moderately expensive, and unfortunately we
can call it twice on the same datagram: once as the (last + 1) entry
in one batch (to check it's not in that batch), then again as the
first entry in the next batch.
Avoid this by keeping track of the "splice port" in the metadata structure,
and filling it in one entry ahead of the one we're currently considering.
This is a bit subtle, but not that hard. It will also generalise better
when we have more complex possibilities based on the flow table.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
udp_mmh_splice_port() is used to determine if a UDP datagram can be
"spliced" (forwarded via a socket instead of tap). We only invoke it if
the origin socket has the 'splice' flag set.
Fold the checking of the flag into the helper itself, which makes the
caller simpler. It does mean we have a loop looking for a batch of
spliceable or non-spliceable packets even in the case where the flag is
clear. This shouldn't be that expensive though, since each call to
udp_mmh_splice_port() will return without accessing memory in that case.
In any case we're going to need a similar loop in more cases with upcoming
flow table work.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
sock_l4() creates, binds and otherwise prepares a new socket. It builds
the socket address to bind from separately provided address and port.
However, we have use cases coming up where it's more natural to construct
the socket address in the caller.
Prepare for this by adding sock_l4_sa() which takes a pre-constructed
socket address, and rewriting sock_l4() in terms of it.
Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
buf_size is set to sizeof(pkt_buf) by default. And it seems more correct
to provide the actual size of the buffer.
Later a buf_size of 0 will allow vhost-user mode to detect
guest memory buffers.
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>
it was needed by a draft version of vhost-user, it is not needed
anymore.
Signed-off-by: Laurent Vivier <lvivier@redhat.com>
Reviewed-by: David Gibson <david@gibson.dropbear.id.au>
Signed-off-by: Stefano Brivio <sbrivio@redhat.com>