DPDK patches and discussions
 help / color / mirror / Atom feed
From: "Robin Jarry" <rjarry@redhat.com>
To: "Vladimir Medvedkin" <vladimir.medvedkin@intel.com>, <dev@dpdk.org>
Subject: rte_fib network order bug
Date: Tue, 12 Nov 2024 10:31:26 +0100	[thread overview]
Message-ID: <D5K3GCR2XSBG.287JQOMCHLET6@redhat.com> (raw)

Hi Vladimir,

I started playing with the new RTE_FIB_F_NETWORK_ORDER flag and I found 
that it does not work at all.

rte_fib is based on rte_rib to perform all slow path lookups and 
modifications. Unfortunately, rte_rib does not handle network order 
addresses. This causes the added routes to be incorrectly masked and 
thus inserted at the wrong place in the dir24 structure.

Here is the config I am using:

static struct rte_fib_conf fib_conf = {
	.type = RTE_FIB_DIR24_8,
	.default_nh = 0,
	.max_routes = 65536,
	.rib_ext_sz = 0,
	.dir24_8 = {
		.nh_sz = RTE_FIB_DIR24_8_8B,
		.num_tbl8 = 1 << 15,
	},
	.flags = RTE_FIB_F_NETWORK_ORDER,
};

And here is a short gdb session demonstrating the issue:

rte_fib_add (fib=0x16ed79740, ip=67305985, depth=24 '\030', next_hop=6438576192) at ../subprojects/dpdk/lib/fib/rte_fib.c:126
126             if ((fib == NULL) || (fib->modify == NULL) ||
(gdb) 
129             return fib->modify(fib, ip, depth, next_hop, RTE_FIB_ADD);
(gdb) p (char*)inet_ntoa(ip)
$9 = 0x7ffff72a471c "1.2.3.4"
(gdb) p depth
$10 = 24 '\030'
(gdb) p fib->flags 
$11 = 1 (
(gdb) s
dir24_8_modify (fib=0x16ed79740, ip=67305985, depth=24 '\030', next_hop=6438576192, op=0)
    at ../subprojects/dpdk/lib/fib/dir24_8.c:452
452             struct rte_rib_node *tmp = NULL;
(gdb) n
455             int ret = 0;
(gdb) 
458             if ((fib == NULL) || (depth > RTE_FIB_MAXDEPTH))
(gdb) 
461             dp = rte_fib_get_dp(fib);
(gdb) 
462             rib = rte_fib_get_rib(fib);
(gdb) 
465             if (next_hop > get_max_nh(dp->nh_sz))
(gdb) 
468             ip &= rte_rib_depth_to_mask(depth);
(gdb) p (char*)inet_ntoa(ip)
$12 = 0x7ffff72a471c "1.2.3.4"
(gdb) p depth
$13 = 24 '\030'
(gdb) n
470             node = rte_rib_lookup_exact(rib, ip, depth);
(gdb) p (char*)inet_ntoa(ip)
$14 = 0x7ffff72a471c "0.2.3.4"
(gdb) p (char*)inet_ntoa(rte_rib_depth_to_mask(depth))
$15 = 0x7ffff72a471c "0.255.255.255"

As you can see, the generated mask is invalid. It should have been 
255.255.255.0.

rte_rib would need to have similar network order support and should 
inherit it from rte_fib.

Let me know if you need more information.


             reply	other threads:[~2024-11-12  9:31 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-11-12  9:31 Robin Jarry [this message]
2024-11-13 10:42 ` Medvedkin, Vladimir
2024-11-13 13:27   ` Robin Jarry
2024-11-13 19:39     ` Medvedkin, Vladimir
2024-11-14  7:43       ` Morten Brørup
2024-11-14 10:18         ` Robin Jarry
2024-11-14 14:35           ` Morten Brørup

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=D5K3GCR2XSBG.287JQOMCHLET6@redhat.com \
    --to=rjarry@redhat.com \
    --cc=dev@dpdk.org \
    --cc=vladimir.medvedkin@intel.com \
    /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.
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 NNTP newsgroup(s).