netdev - Re: [PATCH net-next v2 0/3] net: bridge: convert multicast to generic rhashtable

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1235eb6f-2048-3fb5-b235-7af5b28bcd2b@cumulusnetworks.com>
Date:   Wed, 5 Dec 2018 01:45:16 +0200
From:   Nikolay Aleksandrov <nikolay@...ulusnetworks.com>
To:     netdev@...r.kernel.org
Cc:     roopa@...ulusnetworks.com, davem@...emloft.net
Subject: Re: [PATCH net-next v2 0/3] net: bridge: convert multicast to generic
 rhashtable

On 12/4/18 5:01 PM, Nikolay Aleksandrov wrote:
> Hi,
> The current bridge multicast code uses a custom rhashtable
> implementation which predates the generic rhashtable API. Patch 01
> converts it to use the generic kernel rhashtable which simplifies the
> code a lot and removes duplicated functionality. The convert also makes
> hash_elasticity obsolete as the generic rhashtable already has such
> checks and has a fixed elasticity of RHT_ELASTICITY (16 currently) so we
> emit a warning whenever elasticity is set and return RHT_ELASTICITY when
> read (patch 02). Since now we have the generic rhashtable which
> autoshrinks we can be more liberal with the default hash maximum so patch
> 03 increases it to 4096 and moves it to a define in br_private.h.
> 
> v2: send the latest version of the set which handles when IGMP snooping
>     is not defined, changes are in patch 01
> 
> Thanks,
>  Nik

Unfortunately I'll have to send a v3, I missed the fact that rhashtable_lookup
requires RCU to be always held and even though br_mdb_ip_get() does the
lookups under multicast_lock, we can still trigger suspicious RCU usage because RCU
is not held in those protected places. Need to add a non-rcu mdb lookup variant.

On a related note I saw Paul's call_rcu patches hit, so I'll wait for those
to go in and will rebase on top of them before sending the v3 as the bridge
change will have a conflict with this set.


Thanks,
 Nik