Return-Path: linux-nfs-owner@vger.kernel.org Received: from fieldses.org ([174.143.236.118]:54933 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751921AbaGJP7H (ORCPT ); Thu, 10 Jul 2014 11:59:07 -0400 Date: Thu, 10 Jul 2014 11:59:05 -0400 From: "J. Bruce Fields" To: Jeff Layton Cc: linux-nfs@vger.kernel.org Subject: Re: [PATCH v4 001/100] nfsd: close potential race between delegation break and laundromat Message-ID: <20140710155905.GF26851@fieldses.org> References: <1404842668-22521-1-git-send-email-jlayton@primarydata.com> <1404842668-22521-2-git-send-email-jlayton@primarydata.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii In-Reply-To: <1404842668-22521-2-git-send-email-jlayton@primarydata.com> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Tue, Jul 08, 2014 at 02:02:49PM -0400, Jeff Layton wrote: > Bruce says: > > There's also a preexisting expire_client/laundromat vs break race: > > - expire_client/laundromat adds a delegation to its local > reaplist using the same dl_recall_lru field that a delegation > uses to track its position on the recall lru and drops the > state lock. > > - a concurrent break_lease adds the delegation to the lru. > > - expire/client/laundromat then walks it reaplist and sees the > lru head as just another delegation on the list.... > > Fix this race by checking the dl_time under the state_lock. If we find > that it's not 0, then we know that it has already been queued to the LRU > list and that we shouldn't queue it again. > > In the case of destroy_client, we must also ensure that we don't hit > similar races by ensuring that we don't move any delegations to the > reaplist with a dl_time of 0. Just bump the dl_time by one before we > drop the state_lock. We're destroying the delegations anyway, so a 1s > difference there won't matter. > > The fault injection code also requires a bit of surgery here: > > First, in the case of nfsd_forget_client_delegations, we must prevent > the same sort of race vs. the delegation break callback. For that, we > just increment the dl_time to ensure that a delegation callback can't > race in while we're working on it. > > We can't do that for nfsd_recall_client_delegations, as we need to have > it actually queue the delegation, and that won't happen if we increment > the dl_time. The state lock is held over that function, so we don't need > to worry about these sorts of races there. > > There is one other potential bug nfsd_recall_client_delegations though. > Entries on the victims list are not dequeued before calling > nfsd_break_one_deleg. That's a potential list corruptor, so ensure that > we do that there. > > Reported-by: "J. Bruce Fields" > Signed-off-by: Jeff Layton > --- > fs/nfsd/nfs4state.c | 40 +++++++++++++++++++++++++++++++++------- > 1 file changed, 33 insertions(+), 7 deletions(-) > > diff --git a/fs/nfsd/nfs4state.c b/fs/nfsd/nfs4state.c > index 763aeeb67ccf..633b34fd6c92 100644 > --- a/fs/nfsd/nfs4state.c > +++ b/fs/nfsd/nfs4state.c > @@ -1287,6 +1287,8 @@ destroy_client(struct nfs4_client *clp) > while (!list_empty(&clp->cl_delegations)) { > dp = list_entry(clp->cl_delegations.next, struct nfs4_delegation, dl_perclnt); > list_del_init(&dp->dl_perclnt); > + /* Ensure that deleg break won't try to requeue it */ > + ++dp->dl_time; > list_move(&dp->dl_recall_lru, &reaplist); > } > spin_unlock(&state_lock); > @@ -2933,10 +2935,14 @@ static void nfsd_break_one_deleg(struct nfs4_delegation *dp) > * it's safe to take a reference: */ > atomic_inc(&dp->dl_count); > > - list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru); > - > - /* Only place dl_time is set; protected by i_lock: */ > - dp->dl_time = get_seconds(); > + /* > + * If the dl_time != 0, then we know that it has already been > + * queued for a lease break. Don't queue it again. > + */ > + if (dp->dl_time == 0) { Any reason not to just make that if (dp->dl_time) return; ? I don't know if it matters right now, but it might also be useful to know that no more work will get queued once dl_time is set. --b. > + list_add_tail(&dp->dl_recall_lru, &nn->del_recall_lru); > + dp->dl_time = get_seconds(); > + } > > block_delegations(&dp->dl_fh); > > @@ -5081,8 +5087,23 @@ static u64 nfsd_find_all_delegations(struct nfs4_client *clp, u64 max, > > lockdep_assert_held(&state_lock); > list_for_each_entry_safe(dp, next, &clp->cl_delegations, dl_perclnt) { > - if (victims) > + if (victims) { > + /* > + * It's not safe to mess with delegations that have a > + * non-zero dl_time. They might have already been broken > + * and could be processed by the laundromat outside of > + * the state_lock. Just leave them be. > + */ > + if (dp->dl_time != 0) > + continue; > + > + /* > + * Increment dl_time to ensure that delegation breaks > + * don't monkey with it now that we are. > + */ > + ++dp->dl_time; > list_move(&dp->dl_recall_lru, victims); > + } > if (++count == max) > break; > } > @@ -5107,14 +5128,19 @@ u64 nfsd_forget_client_delegations(struct nfs4_client *clp, u64 max) > > u64 nfsd_recall_client_delegations(struct nfs4_client *clp, u64 max) > { > - struct nfs4_delegation *dp, *next; > + struct nfs4_delegation *dp; > LIST_HEAD(victims); > u64 count; > > spin_lock(&state_lock); > count = nfsd_find_all_delegations(clp, max, &victims); > - list_for_each_entry_safe(dp, next, &victims, dl_recall_lru) > + while (!list_empty(&victims)) { > + dp = list_first_entry(&victims, struct nfs4_delegation, > + dl_recall_lru); > + list_del_init(&dp->dl_recall_lru); > + dp->dl_time = 0; > nfsd_break_one_deleg(dp); > + } > spin_unlock(&state_lock); > > return count; > -- > 1.9.3 >