From: Stefano Brivio <sbrivio@redhat.com>
To: David Gibson <david@gibson.dropbear.id.au>
Cc: passt-dev@passt.top
Subject: Re: [PATCH v2 05/10] flow: Introduce struct flowside, space for uniform tracking of addresses
Date: Thu, 7 Sep 2023 09:55:26 +0200 [thread overview]
Message-ID: <20230907095526.0345329c@elisabeth> (raw)
In-Reply-To: <ZPlMESqS4YyAoPkH@zatzit>
On Thu, 7 Sep 2023 14:05:37 +1000
David Gibson <david@gibson.dropbear.id.au> wrote:
> On Thu, Sep 07, 2023 at 03:01:40AM +0200, Stefano Brivio wrote:
> > On Mon, 28 Aug 2023 15:41:41 +1000
> > David Gibson <david@gibson.dropbear.id.au> wrote:
> >
> > > Handling of each protocol needs some degree of tracking of the addresses
> > > and ports at the end of each connection or flow. Sometimes that's explicit
> > > (as in the guest visible addresses for TCP connections), sometimes implicit
> > > (the bound and connected addresses of sockets).
> > >
> > > To allow more general abd robust handling, and more consistency across
> > > protocols we want to uniformly track the address and port at each end of
> > > the connection. Furthermore, because we allow port remapping, and we
> > > sometimes need to apply NAT, the addresses and ports can be different as
> > > seen by the guest/namespace and as by the host.
> > >
> > > Introduce 'struct flowside' to keep track of the address and ports of a
> > > flow from a single "side" (guest or host). Store two of these in the
> > > common fields of a flow to track that information for both sides.
> > >
> > > For now we just introduce the structure and fields themselves, along with
> > > some simple helpers. Later patches will actually use these to store useful
> > > information.
> > >
> > > Signed-off-by: David Gibson <david@gibson.dropbear.id.au>
> > > ---
> > > flow.c | 22 ++++++++++++++++++++++
> > > flow.h | 48 ++++++++++++++++++++++++++++++++++++++++++++++++
> > > 2 files changed, 70 insertions(+)
> > >
> > > diff --git a/flow.c b/flow.c
> > > index 12ca8db..a93cf8c 100644
> > > --- a/flow.c
> > > +++ b/flow.c
> > > @@ -7,6 +7,7 @@
> > >
> > > #include <unistd.h>
> > > #include <string.h>
> > > +#include <arpa/inet.h>
> > >
> > > #include "util.h"
> > > #include "passt.h"
> > > @@ -24,6 +25,27 @@ const char *flow_type_str[] = {
> > > /* Global Flow Table */
> > > union flow flowtab[FLOW_MAX];
> > >
> > > +/** flowside_fmt - Format a flowside as a string
> > > + * @fs: flowside to format
> > > + * @buf: Buffer into which to store the formatted version
> > > + * @size: Size of @buf
> > > + *
> > > + * Return: pointer to formatted string describing @fs, or NULL on error
> > > + */
> > > +/* cppcheck-suppress unusedFunction */
> > > +const char *flowside_fmt(const struct flowside *fs, char *buf, size_t size)
> > > +{
> > > + char ebuf[INET6_ADDRSTRLEN], fbuf[INET6_ADDRSTRLEN];
> > > +
> > > + if (!inet_ntop(AF_INET6, &fs->eaddr, ebuf, sizeof(ebuf))
> > > + || !inet_ntop(AF_INET6, &fs->faddr, fbuf, sizeof(fbuf)))
> >
> > For consistency (also with flowside_complete() below):
> >
> > if (!inet_ntop(AF_INET6, &fs->eaddr, ebuf, sizeof(ebuf)) ||
> > !inet_ntop(AF_INET6, &fs->faddr, fbuf, sizeof(fbuf)))
>
> Done.
>
> > > + return NULL;
> > > +
> > > + snprintf(buf, size, "[%s]:%hu <-> [%s]:%hu", fbuf, fs->fport,
> > > + ebuf, fs->eport);
> > > + return (const char *)buf;
> > > +}
> > > +
> > > /**
> > > * flow_table_compact() - Perform compaction on flow table
> > > * @c: Execution context
> > > diff --git a/flow.h b/flow.h
> > > index e212796..9891fcb 100644
> > > --- a/flow.h
> > > +++ b/flow.h
> > > @@ -18,11 +18,59 @@ extern const char *flow_type_str[];
> > > #define FLOW_TYPE(f) \
> > > ((f)->type <= FLOW_MAX ? flow_type_str[(f)->type] : "?")
> > >
> > > +/**
> > > + * struct flowside - Describes a logical packet flow as seen from one "side"
> > > + * @eaddr: Endpoint address (remote address from passt's PoV)
> > > + * @faddr: Forwarding address (local address from passt's PoV)
> > > + * @eport: Endpoint port
> > > + * @fport: Forwarding port
> > > + */
> > > +struct flowside {
> > > + union inany_addr faddr;
> > > + union inany_addr eaddr;
> > > + in_port_t fport, eport;
> >
> > I guess always valid, but uncommon (compared to in_port_t x; in_port_t
> > y;)?
>
> Huh.. I didn't even think about that (and the fact that I did this for
> the ports, but not for the addresses). Changed it to the more
> conventional style.
>
> > > +};
> > > +
> > > +/** flowside_from_af - Initialize a flowside from addresses
> > > + * @fs: flowside to initialize
> > > + * @af: Address family (AF_INET or AF_INET6)
> > > + * @faddr: Forwarding address (pointer to in_addr or in6_addr)
> > > + * @fport: Forwarding port
> > > + * @eaddr: Endpoint address (pointer to in_addr or in6_addr)
> > > + * @eport: Endpoint port
> > > + */
> > > +static inline void flowside_from_af(struct flowside *fs, int af,
> > > + const void *faddr, in_port_t fport,
> > > + const void *eaddr, in_port_t eport)
> > > +{
> > > + inany_from_af(&fs->faddr, af, faddr);
> > > + inany_from_af(&fs->eaddr, af, eaddr);
> > > + fs->fport = fport;
> > > + fs->eport = eport;
> > > +}
> > > +
> > > +/** flowside_complete - Check if flowside is fully initialized
> > > + * @fs: flowside to check
> > > + */
> > > +static inline bool flowside_complete(const struct flowside *fs)
> > > +{
> > > + return !IN6_IS_ADDR_UNSPECIFIED(&fs->faddr) &&
> > > + !IN6_IS_ADDR_UNSPECIFIED(&fs->eaddr) &&
> > > + fs->fport != 0 && fs->eport != 0;
> >
> > Zero is reserved for TCP (which could be problematic anyway if we try
> > to match things),
>
> The more practical consideration here is that a 0 port is used in the
> sockaddr passed to bind() to represent "pick a port for me". The
> point of this is to check that we have a "fully specified" flowside,
> that provides sufficient information to both bind() and connect() a
> socket with no ambiguity on the endpoints.
>
> > but for UDP a "zero" port can actually be used (in
> > the probably desuete sense of "no port").
>
> [Aside: I've never encountered the word desuete before]
>
> By "no port" here are you meaning for UDP traffic that expects no
> response? If that's so we probably neither need or want to create a
> flow for it anyway.
The fact that no response is expected is probably a practical
consequence of this... but RFC 768, "Fields", really says:
Source Port is an optional field, when meaningful, it indicates the port
of the sending process, and may be assumed to be the port to which a
reply should be addressed in the absence of any other information. If
not used, a value of zero is inserted.
> In any case, even if port 0 can be used at the protocol level, I don't
> think it can be used at the socket level: I'm pretty sure the bind()
> behaviour of treating 0 as "pick for me" is true for UDP as well as
> TCP - it's functionality that's basically necessary, and I can't see
> any other way to specify it.
Right, yes.
> > Maybe we should use -1
> > instead?
>
> That doesn't really help - port 65535 is itself valid, so unless we
> widen these fields - which I don't really want to do - it's still
> ambiguous (in fact, worse, because AFAICT port 65535 could actually be
> used in practice).
Oops, sorry, I didn't consider that. Of course.
> Even if we did widen, it's still a problem, because if we got the port
> from, for example, getsockname() on a not-yet-connected socket, that
> will give us 0 and the point of this function is to tell us that's not
> a fully specified endpoint.
Right... forget about this, I didn't consider that.
> > > +}
> > > +
> > > +#define FLOWSIDE_STRLEN (2*(INET6_ADDRSTRLEN+8) + 6)
> >
> > For consistency: "(2 * INET6_ADDRSTRLEN + 8) + 6)".
>
> Done.
>
> > Limited to the usage I've seen in 6/10 (maybe I'm ignoring something):
> > is it worth it to have flowside_fmt() as a function forming a string,
> > rather than something calling debug() with what we want...? At the
> > moment we have tap_packet_debug(), admittedly not very elegant but
> > perhaps more terse than this.
>
> There's at least one more use coming in the remainder of the series,
> and I'd expect to see more as other protocols are added to the flow
> mechanics. I also think it could be a very useful helper when adding
> ad-hoc debugging. So, yes, I think it's worth it.
Ah, okay.
--
Stefano
next prev parent reply other threads:[~2023-09-07 7:55 UTC|newest]
Thread overview: 19+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-08-28 5:41 [PATCH v2 00/10] RFC: Convert TCP connection table to generalisable flow table David Gibson
2023-08-28 5:41 ` [PATCH v2 01/10] flow, tcp: Generalise connection types David Gibson
2023-08-28 5:41 ` [PATCH v2 02/10] flow, tcp: Move TCP connection table to unified flow table David Gibson
2023-08-28 5:41 ` [PATCH v2 03/10] flow, tcp: Consolidate flow pointer<->index helpers David Gibson
2023-09-07 1:01 ` Stefano Brivio
2023-09-07 3:48 ` David Gibson
2023-08-28 5:41 ` [PATCH v2 04/10] flow: Make unified version of flow table compaction David Gibson
2023-08-28 5:41 ` [PATCH v2 05/10] flow: Introduce struct flowside, space for uniform tracking of addresses David Gibson
2023-09-07 1:01 ` Stefano Brivio
2023-09-07 4:05 ` David Gibson
2023-09-07 7:55 ` Stefano Brivio [this message]
2023-08-28 5:41 ` [PATCH v2 06/10] tcp: Move guest side address tracking to flow/flowside David Gibson
2023-08-28 5:41 ` [PATCH v2 07/10] tcp, flow: Perform TCP hash calculations based on flowside David Gibson
2023-08-28 5:41 ` [PATCH v2 08/10] tcp: Re-use flowside_hash for initial sequence number generation David Gibson
2023-08-28 5:41 ` [PATCH v2 09/10] tcp: Maintain host flowside for connections David Gibson
2023-08-28 5:41 ` [PATCH v2 10/10] tcp_splice: Fill out flowside information for spliced connections David Gibson
2023-09-07 1:02 ` Stefano Brivio
2023-09-07 4:14 ` David Gibson
2023-09-07 7:55 ` Stefano Brivio
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20230907095526.0345329c@elisabeth \
--to=sbrivio@redhat.com \
--cc=david@gibson.dropbear.id.au \
--cc=passt-dev@passt.top \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
Code repositories for project(s) associated with this public inbox
https://passt.top/passt
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for IMAP folder(s).