Commmit 58ad436fcf49810aa006016107f494c9ac9013db
Author: Johannes Berg <johannes.berg at intel.com>
genetlink: fix family dump race
added genl_lock() call to ctrl_dumpfamily(), which potentially
can deadlock. Suppose the following case:
genl_rcv_msg(): !family->parallel_ops -> genl_lock()
genl_family_rcv_msg()
netlink_dump_start()
netlink_dump()
ctrl_dumpfamily() --> genl_lock()
Take in account possibility of genl_lock() in genl_rcv_msg() and do
not acquire genl_mutex in ctrl_dumpfamily() for a family that supports
parallel_ops.
Signed-off-by: Sergey Senozhatsky <[email protected]>
---
net/netlink/genetlink.c | 6 ++++++
1 file changed, 6 insertions(+)
diff --git a/net/netlink/genetlink.c b/net/netlink/genetlink.c
index f85f8a2..fca3659 100644
--- a/net/netlink/genetlink.c
+++ b/net/netlink/genetlink.c
@@ -791,6 +791,12 @@ static int ctrl_dumpfamily(struct sk_buff *skb, struct netlink_callback *cb)
int fams_to_skip = cb->args[1];
bool need_locking = chains_to_skip || fams_to_skip;
+ if (need_locking) {
+ /* genl_mutex could be already locked in genl_rcv_msg() */
+ rt = genl_family_find_byid(cb->nlh->nlmsg_type);
+ need_locking = need_locking && rt->parallel_ops;
+ }
+
if (need_locking)
genl_lock();
> + if (need_locking) {
> + /* genl_mutex could be already locked in genl_rcv_msg() */
> + rt = genl_family_find_byid(cb->nlh->nlmsg_type);
> + need_locking = need_locking && rt->parallel_ops;
> + }
This is equivalent to reverting the patch because parallel_ops will be false for the genl family. :-)
johannes
--
Intel GmbH
Dornacher Strasse 1
85622 Feldkirchen/Muenchen, Deutschland
Sitz der Gesellschaft: Feldkirchen bei Muenchen
Geschaeftsfuehrer: Christian Lamprechter, Hannes Schwaderer, Douglas Lusk
Registergericht: Muenchen HRB 47456
Ust.-IdNr./VAT Registration No.: DE129385895
Citibank Frankfurt a.M. (BLZ 502 109 00) 600119052
On (08/20/13 14:03), Berg, Johannes wrote:
> > + if (need_locking) {
> > + /* genl_mutex could be already locked in genl_rcv_msg() */
> > + rt = genl_family_find_byid(cb->nlh->nlmsg_type);
> > + need_locking = need_locking && rt->parallel_ops;
> > + }
>
> This is equivalent to reverting the patch because parallel_ops will be false for the genl family. :-)
>
sorry, didn't notice that there is already a thread on lkml regarding
this issue. will follow that discussion. the reason was that I see an
actual deadlock on my machine during certain network activities, but
lockdep is silent.
thanks,
-ss
> johannes
> --
>
> Intel GmbH
> Dornacher Strasse 1
> 85622 Feldkirchen/Muenchen, Deutschland
> Sitz der Gesellschaft: Feldkirchen bei Muenchen
> Geschaeftsfuehrer: Christian Lamprechter, Hannes Schwaderer, Douglas Lusk
> Registergericht: Muenchen HRB 47456
> Ust.-IdNr./VAT Registration No.: DE129385895
> Citibank Frankfurt a.M. (BLZ 502 109 00) 600119052
>