| Message ID | 20221011071128.3470622-3-houtao@huaweicloud.com (mailing list archive) |
|---|---|
| State | Superseded |
| Headers | show |
| Series | Remove unnecessary RCU grace period chaining | expand |
On Tue, Oct 11, 2022 at 03:11:27PM +0800, Hou Tao wrote: > From: Hou Tao <houtao1@huawei.com> > > Local storage map is accessible for both sleepable and normal bpf > program, so its memory is freed by using both call_rcu_tasks_trace() and > kfree_rcu() to wait for both RCU-tasks-trace grace period and RCU grace > period to expire. > > However According to the current implementation of RCU-tasks-trace, one > RCU-tasks-trace grace period waits for one RCU grace period, so there is > no need to call kfree_rcu(), it is safe to call kfree() directly. Again, this is true, but this is an implementation detail that is not guaranteed in future versions of the kernel. But if this additional call_rcu() is causing trouble, I could add some API member that returned true in kernels where it does happen to be the case that call_rcu_tasks_trace() implies a call_rcu()-style grace period. The BPF local storage code could then complain or adapt, as appropriate. Again, thoughts? Thanx, Paul > Signed-off-by: Hou Tao <houtao1@huawei.com> > --- > kernel/bpf/bpf_local_storage.c | 10 ++++++++-- > 1 file changed, 8 insertions(+), 2 deletions(-) > > diff --git a/kernel/bpf/bpf_local_storage.c b/kernel/bpf/bpf_local_storage.c > index 802fc15b0d73..18a2dd611635 100644 > --- a/kernel/bpf/bpf_local_storage.c > +++ b/kernel/bpf/bpf_local_storage.c > @@ -84,20 +84,26 @@ bpf_selem_alloc(struct bpf_local_storage_map *smap, void *owner, > return NULL; > } > > +/* Now RCU Tasks grace period implies RCU grace period, so no need to call > + * kfree_rcu(), just call kfree() directly. > + */ > void bpf_local_storage_free_rcu(struct rcu_head *rcu) > { > struct bpf_local_storage *local_storage; > > local_storage = container_of(rcu, struct bpf_local_storage, rcu); > - kfree_rcu(local_storage, rcu); > + kfree(local_storage); > } > > +/* Now RCU Tasks grace period implies RCU grace period, so no need to call > + * kfree_rcu(), just call kfree() directly. > + */ > static void bpf_selem_free_rcu(struct rcu_head *rcu) > { > struct bpf_local_storage_elem *selem; > > selem = container_of(rcu, struct bpf_local_storage_elem, rcu); > - kfree_rcu(selem, rcu); > + kfree(selem); > } > > /* local_storage->lock must be held and selem->local_storage == local_storage. > -- > 2.29.2 >
diff --git a/kernel/bpf/bpf_local_storage.c b/kernel/bpf/bpf_local_storage.c index 802fc15b0d73..18a2dd611635 100644 --- a/kernel/bpf/bpf_local_storage.c +++ b/kernel/bpf/bpf_local_storage.c @@ -84,20 +84,26 @@ bpf_selem_alloc(struct bpf_local_storage_map *smap, void *owner, return NULL; } +/* Now RCU Tasks grace period implies RCU grace period, so no need to call + * kfree_rcu(), just call kfree() directly. + */ void bpf_local_storage_free_rcu(struct rcu_head *rcu) { struct bpf_local_storage *local_storage; local_storage = container_of(rcu, struct bpf_local_storage, rcu); - kfree_rcu(local_storage, rcu); + kfree(local_storage); } +/* Now RCU Tasks grace period implies RCU grace period, so no need to call + * kfree_rcu(), just call kfree() directly. + */ static void bpf_selem_free_rcu(struct rcu_head *rcu) { struct bpf_local_storage_elem *selem; selem = container_of(rcu, struct bpf_local_storage_elem, rcu); - kfree_rcu(selem, rcu); + kfree(selem); } /* local_storage->lock must be held and selem->local_storage == local_storage.