[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <87o762m31v.fsf@nvidia.com>
Date: Fri, 9 Aug 2024 11:48:09 +0200
From: Petr Machata <petrm@...dia.com>
To: Jakub Kicinski <kuba@...nel.org>
CC: Petr Machata <petrm@...dia.com>, "David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>, Paolo Abeni <pabeni@...hat.com>,
<netdev@...r.kernel.org>, Ido Schimmel <idosch@...dia.com>, David Ahern
<dsahern@...nel.org>, Donald Sharp <sharpd@...dia.com>, Simon Horman
<horms@...nel.org>, Przemek Kitszel <przemyslaw.kitszel@...el.com>,
<mlxsw@...dia.com>
Subject: Re: [PATCH net-next v2 0/6] net: nexthop: Increase weight to u16
Jakub Kicinski <kuba@...nel.org> writes:
> On Wed, 7 Aug 2024 16:13:45 +0200 Petr Machata wrote:
>> In CLOS networks, as link failures occur at various points in the network,
>> ECMP weights of the involved nodes are adjusted to compensate. With high
>> fan-out of the involved nodes, and overall high number of nodes,
>> a (non-)ECMP weight ratio that we would like to configure does not fit into
>> 8 bits. Instead of, say, 255:254, we might like to configure something like
>> 1000:999. For these deployments, the 8-bit weight may not be enough.
>>
>> To that end, in this patchset increase the next hop weight from u8 to u16.
>>
>> Patch #1 adds a flag that indicates whether the reserved fields are zeroed.
>> This is a follow-up to a new fix merged in commit 6d745cd0e972 ("net:
>> nexthop: Initialize all fields in dumped nexthops"). The theory behind this
>> patch is that there is a strict ordering between the fields actually being
>> zeroed, the kernel declaring that they are, and the kernel repurposing the
>> fields. Thus clients can use the flag to tell if it is safe to interpret
>> the reserved fields in any way.
>>
>> Patch #2 contains the substantial code and the commit message covers the
>> details of the changes.
>>
>> Patches #3 to #6 add selftests.
>
> I did update iproute2 to the branch you sent me last time, but tests
> are not happy:
>
> # IPv6 groups functional
> # ----------------------
> # TEST: Create nexthop group with single nexthop [ OK ]
> # TEST: Get nexthop group by id [ OK ]
> # TEST: Delete nexthop group by id [ OK ]
> # TEST: Nexthop group with multiple nexthops [ OK ]
> # TEST: Nexthop group updated when entry is deleted [ OK ]
> # TEST: Nexthop group with weighted nexthops [ OK ]
> # TEST: Weighted nexthop group updated when entry is deleted [ OK ]
> # TEST: Nexthops in groups removed on admin down [ OK ]
> # TEST: Multiple groups with same nexthop [ OK ]
> # TEST: Nexthops in group removed on admin down - mixed group [ OK ]
> # TEST: Nexthop group can not have a group as an entry [ OK ]
> # TEST: Nexthop group with a blackhole entry [ OK ]
> # TEST: Nexthop group can not have a blackhole and another nexthop [ OK ]
> # TEST: Nexthop group replace refcounts [ OK ]
> # WARNING: Unexpected route entry
> # TEST: 16-bit weights [FAIL]
> #
> # IPv6 resilient groups functional
> # --------------------------------
> # TEST: Nexthop group updated when entry is deleted [ OK ]
> # TEST: Nexthop buckets updated when entry is deleted [ OK ]
> # TEST: Nexthop group updated after replace [ OK ]
> # TEST: Nexthop buckets updated after replace [ OK ]
> # TEST: Nexthop group updated when entry is deleted - nECMP [ OK ]
> # TEST: Nexthop buckets updated when entry is deleted - nECMP [ OK ]
> # TEST: Nexthop group updated after replace - nECMP [ OK ]
> # TEST: Nexthop buckets updated after replace - nECMP [ OK ]
> # WARNING: Unexpected route entry
> # TEST: 16-bit weights [FAIL]
>
> https://netdev-3.bots.linux.dev/vmksft-net/results/718641/2-fib-nexthops-sh/stdout
This failure mode is consistent with non-updated iproute2. I only pushed
to the iproute2 repository after having sent the kernel patches, so I
think you or your automation have picked up the old version. Can you try
again, please? I retested on my end and it still works.
Powered by blists - more mailing lists