> -----Original Message-----
> From: devel-boun...@open-fcoe.org [mailto:devel-boun...@open-fcoe.org]
> On Behalf Of Chris Leech
> Sent: Monday, September 10, 2012 10:09 AM
> To: devel@open-fcoe.org
> Subject: [Open-FCoE] netdev removal and timing of call to dev_mc_del
> 
> There's some odd timing of the fcoe modules use of dev_mc_del when an
> fcoe interface removal is triggered by the notifier for a netdev
> removal.  Basically, the multicast address cleanup in fcoe is
> happening after the netdev code already forcefully flushed all the
> addresses.
> 
> This commit moved the cleanup that was inline with the notifier to the
> deferred workqueue context.
> 
>   [SCSI] fcoe: Rearrange fcoe port and NPIV port cleanup
>   commit b2085a4efc1a00375b77d5cbfe73a549c9d7d65b
>   Author:     Neerav Parikh <neerav.par...@intel.com>
>   AuthorDate: Mon Jun 20 16:59:51 2011 -0700
> 
> But some of that was expected to be completed before returning from
> the notifier by the netdev layer.
> 
> It's not really a big problem, but it is more noticeable in the Red
> Hat kernel where this code was backported and paired with older
> multicast list handling code that prints a warning when it finds
> unfreed multicast addresses after the netdev notifiers have been
> called.
> 
> Is it worth trying to move the address deletion back to the notifier
> context?
> 
> - Chris
One thing that I remember when I moved this code was because of the 
rtnl_lock ordering and dependency on the rtnl_lock in the cleanup path.
That was one of the reasons for delaying the cleanup in the delayed work
context and not doing it inline. So, if you try to move the address 
deletion back to the notifier context then probably you'll have to take 
care of that as well. Also, there might be some other nuances with the 
ordering of the rtnl_lock/fcoe_config_mutex locks that are being used in 
the fcoe create/destroy paths that might need some review to avoid any
deadlocks.


> _______________________________________________
> devel mailing list
> devel@open-fcoe.org
> https://lists.open-fcoe.org/mailman/listinfo/devel
_______________________________________________
devel mailing list
devel@open-fcoe.org
https://lists.open-fcoe.org/mailman/listinfo/devel

Reply via email to