On Fri, Feb 28, 2025 at 10:25:58AM -0800, Paul E. McKenney wrote: > On Fri, Feb 28, 2025 at 06:08:19PM +0100, Uladzislau Rezki wrote: > > On Fri, Feb 28, 2025 at 05:36:47PM +0100, Uladzislau Rezki wrote: > > > On Fri, Feb 28, 2025 at 07:41:40AM -0800, Paul E. McKenney wrote: > > > > On Thu, Feb 27, 2025 at 06:44:15PM +0100, Uladzislau Rezki wrote: > > > > > On Thu, Feb 27, 2025 at 09:26:40AM -0800, Paul E. McKenney wrote: > > > > > > On Thu, Feb 27, 2025 at 09:12:39AM -0800, Boqun Feng wrote: > > > > > > > Hi Ulad, > > > > > > > > > > > > > > I put these three patches into next (and misc.2025.02.27a) for > > > > > > > some > > > > > > > testing, hopefully it all goes well and they can make it v6.15. > > > > > > > > > > > > > > A few tag changed below: > > > > > > > > > > > > > > On Thu, Feb 27, 2025 at 02:16:13PM +0100, Uladzislau Rezki (Sony) > > > > > > > wrote: > > > > > > > > Switch for using of get_state_synchronize_rcu_full() and > > > > > > > > poll_state_synchronize_rcu_full() pair to debug a normal > > > > > > > > synchronize_rcu() call. > > > > > > > > > > > > > > > > Just using "not" full APIs to identify if a grace period is > > > > > > > > passed or not might lead to a false-positive kernel splat. > > > > > > > > > > > > > > > > It can happen, because get_state_synchronize_rcu() compresses > > > > > > > > both normal and expedited states into one single unsigned long > > > > > > > > value, so a poll_state_synchronize_rcu() can miss GP-completion > > > > > > > > when synchronize_rcu()/synchronize_rcu_expedited() concurrently > > > > > > > > run. > > > > > > > > > > > > > > > > To address this, switch to poll_state_synchronize_rcu_full() and > > > > > > > > get_state_synchronize_rcu_full() APIs, which use separate > > > > > > > > variables > > > > > > > > for expedited and normal states. > > > > > > > > > > > > > > > > Link: https://lore.kernel.org/lkml/Z5ikQeVmVdsWQrdD@pc636/T/ > > > > > > > > > > > > > > I switch this into "Closes:" per checkpatch. > > > > > > > > > > > > > > > Fixes: 988f569ae041 ("rcu: Reduce synchronize_rcu() latency") > > > > > > > > Reported-by: cheung wall <zzqq0103....@gmail.com> > > > > > > > > Signed-off-by: Uladzislau Rezki (Sony) <ure...@gmail.com> > > > > > > > > > > > > > > You seem to forget add Paul's Reviewed-by, so I add it in > > > > > > > rcu/next. > > > > > > > Would you or Paul double-check the Reviewed-by should be here? > > > > > > > > > > > > I am good with keeping my Reviewed-by tags. > > > > > > > > > > > Thanks Paul! > > > > > > > > Except that I got this from overnight testing of rcu/dev on the shared > > > > RCU tree: > > > > > > > > WARNING: CPU: 5 PID: 14 at kernel/rcu/tree.c:1636 > > > > rcu_sr_normal_complete+0x5c/0x80 > > > > > > > > I see this only on TREE05. Which should not be too surprising, given > > > > that this is the scenario that tests it. It happened within five > > > > minutes > > > > on all 14 of the TREE05 runs. > > > > > > > Hm.. This is not fun. I tested this on my system and i did not manage to > > > trigger this whereas you do. Something is wrong. > > > > > We have below code to start a new GP, if we detect that processing is > > starved: > > > > <snip> > > /* > > * The "start_new_poll" is set to true, only when this GP is not able > > * to handle anything and there are outstanding users. It happens when > > * the rcu_sr_normal_gp_init() function was not able to insert a dummy > > * separator to the llist, because there were no left any dummy-nodes. > > * > > * Number of dummy-nodes is fixed, it could be that we are run out of > > * them, if so we start a new pool request to repeat a try. It is rare > > * and it means that a system is doing a slow processing of callbacks. > > */ > > if (start_new_poll) > > (void) start_poll_synchronize_rcu(); > > <snip> > > > > we do not use a _full() version, since we need to inform rcu-gp-kthread > > to initiate a new GP. > > > > Any thoughts? > > My kneejerk not-to-be-trusted take is that it does not matter which type > of grace period gets started so long as a grace period does get started. > Presumably you have done the get_state_synchronize_rcu_full() before > this point? > Yes, of course. That code is located in the gp-kthread. get_state_synchronize_rcu_full() is done before.
-- Uladzislau Rezki
