Re: mmap()ed AF_NETLINK: lockdep and sleep-in-atomic warnings

Thomas Graf [off-list ref] wrote:

On 07/13/15 at 10:11pm, Cong Wang wrote:

quoted

Caused by:

commit 21e4902aea80ef35afc00ee8d2abdea4f519b7f7
Author: Thomas Graf [off-list ref]
Date:   Fri Jan 2 23:00:22 2015 +0100

    netlink: Lockless lookup with RCU grace period in socket release

    Defers the release of the socket reference using call_rcu() to
    allow using an RCU read-side protected call to rhashtable_lookup()

    This restores behaviour and performance gains as previously
    introduced by e341694 ("netlink: Convert netlink_lookup() to use
    RCU protected hash table") without the side effect of severely
    delayed socket destruction.

    Signed-off-by: Thomas Graf [off-list ref]
    Signed-off-by: David S. Miller [off-list ref]


We can't hold mutex lock in a rcu callback, perhaps we could
defer the mmap ring cleanup to a workqueue.

The socket should be dead at this point. It might be simpler to
add a netlink_release_ring() function which doesn't require
locking at all. It would also get rid of all the special handling
for closing vs. non-closing.

Do you know of any users of mmap'd netlink?
Support for this was never added to libmnl; given the fact that netlink
mmap doesn't have any advantages vs. socket interface i'd just revert it
(or mark CONFIG_NETLINK_MMAP broken -- although I don't see
 how the unconditional copy on tx could ever be avoided).