Re: [B.A.T.M.A.N.] B.A.T.M.A.N.-Advanced Multicast Optimizations, v2

By the way, you can also pull these commits directly from http://git.open-mesh.org/?p=t_x/batman-adv.git;a=summary and easily monitor/review new changes here in the "multicast" branch.

This patchset is tagged as "multicast-v2", obviously.

Cheers, Linus

PS: The NDP patchsets are available there as well.

On Sat, Jan 22, 2011 at 02:21:23AM +0100, Linus Lüssing wrote:

Hi everyone,

Here's the next series of patches which should address the comments I got for the first one. Thanks for all the feedback!

Changelog:

• rebasing to commit [65e0869478bce153a799c0e774a117ba5fc78025], using new orig_hash methods
• putting seqno before ttl, 4 byte aligning mcast_packet [01/20]
• adapted compat.h to not use custom lock macros, instead only one macro for netif_addr_lock_bh() in case of older kernel versions
• merged spinlock-irqsave-to-bh commit into previous commits [20/20]
• moved mcast_may_optimize() to soft-interface.c [18/20], removed inlining (won't optimize much anyway...)
• purge_mcast_forw_table, splitted list operations into separate functions [12/20]
• use batman_if refcounting to reduce the time of rcu-locking [13/20]
• do not create nexthop entry if according batman_if is NULL [13/20]
• route_mcast_packet, split into separat functions [13/20]
• fix typo "seperate" [13/20]
• fix typo "i.g." [08/20]
• COMPAT_VERSION to 14 [01/20]
• use rcu-locking+refcounting for orig_node, remove orig_hash_lock [07/20], [17/20]
• made checkpatch-clean
• use __packed instead of __attribute_((packed)) [01/20]
• change tracker_packet_for_each_dest macro [07/20]: make a "break" in this macro to behave like usual, export parts from macro into own functions

TODO:

• directly prepare mcast-tracker-packet in sk_buff

• only create methods / variables in patches that need them

• update mcast-doc

• upload updated mcast-doc to wiki

maybe TODO?

• use hlist instead of list for mcast-table?
• use rcu-locking / refcounting for mcast_forw_table?

Cheers, Linus