On Fri, Dec 09, 2022 at 04:38:38PM -0800, Paul E. McKenney wrote: > On Sat, Dec 10, 2022 at 01:06:16AM +0100, Jiri Olsa wrote: > > On Fri, Dec 09, 2022 at 03:34:45PM -0800, Jakub Kicinski wrote: > > > On Sat, 10 Dec 2022 00:32:07 +0100 Daniel Borkmann wrote: > > > > fwiw, these should not be necessary, Documentation/RCU/checklist.rst : > > > > > > > > [...] One example of non-obvious pairing is the XDP feature in networking, > > > > which calls BPF programs from network-driver NAPI (softirq) context. BPF > > > > relies heavily on RCU protection for its data structures, but because the > > > > BPF program invocation happens entirely within a single local_bh_disable() > > > > section in a NAPI poll cycle, this usage is safe. The reason that this usage > > > > is safe is that readers can use anything that disables BH when updaters use > > > > call_rcu() or synchronize_rcu(). [...] > > > > > > FWIW I sent a link to the thread to Paul and he confirmed > > > the RCU will wait for just the BH. > > > > so IIUC we can omit the rcu_read_lock/unlock on bpf_prog_run_xdp side > > > > Paul, > > any thoughts on what we can use in here to synchronize bpf_dispatcher_change_prog > > with bpf_prog_run_xdp callers? > > > > with synchronize_rcu_tasks I'm getting splats like: > > https://lore.kernel.org/bpf/20221209153445.22182ca5@xxxxxxxxxx/T/#m0a869f93404a2744884d922bc96d497ffe8f579f > > > > synchronize_rcu_tasks_rude seems to work (patch below), but it also sounds special ;-) > > It sounds like we are all talking past each other, leaving me no > choice but to supply a wall of text: > > It is quite true that synchronize_rcu_tasks_rude() will wait > for bh-disabled regions of code, just like synchronize_rcu() > and synchronize_rcu_tasks() will. However, please note that > synchronize_rcu_tasks() never waits on any of the idle tasks. So the > usual approach in tracing is to do both a synchronize_rcu_tasks() and > synchronize_rcu_tasks_rude(). One way of overlapping the resulting > pair of grace periods is to use synchronize_rcu_mult(). > > But none of these permit readers to sleep. That is what > synchronize_rcu_tasks_trace() is for, but unlike both > synchronize_rcu_tasks() and synchronize_rcu_tasks_rude(), > you must explicitly mark the readers with rcu_read_lock_trace() > and rcu_read_unlock_trace(). This is used to protect sleepable > BPF programs. > > Now, synchronize_rcu() will also wait on bh-disabled lines of code, with > the exception of such code in the exception path, way deep in the idle > loop, early in the CPU-online process, or late in the CPU-offline process. > You can recognize the first two categories of code by the noinstr tags > on the functions. > > And yes, synchronize_rcu_rude() is quite special. ;-) > > Does this help, or am I simply adding to the confusion? I see, so as Alexei said to synchronize bpf_prog_run_xdp callers, we should be able to use just synchronize_rcu, because it's allways called just in bh-disabled code thanks, jirka
