hi, Paul, On Wed, Sep 25, 2024 at 01:57:45AM -0700, Paul E. McKenney wrote: > > > > If you fix the issue in a separate patch/commit (i.e. not just a new > > version of > > the same patch/commit), kindly add following tags > > | Reported-by: kernel test robot <[email protected]> > > | Closes: https://lore.kernel.org/oe-lkp/[email protected] > > > > > > [ 169.560380][ T76] BUG: using __this_cpu_read() in preemptible > > [00000000] code: kworker/1:2/76 > > [ 169.561229][ T76] caller is try_check_zero (kernel/rcu/srcutree.c:485 > > kernel/rcu/srcutree.c:1149) > > Does this diff fix things for you? Here the per-CPU access is a > heuristic that reduces overhead once the bulk of the CPUs have done > srcu_read_lock_lite(). So it is OK to get migrated just after the check.
yes, this patch fixed the random failures in our tests. Tested-by: kernel test robot <[email protected]> we applied patch upon 26cb612c43 and run the same tests up to 200 times, didn't observe the issue. thanks! ========================================================================================= compiler/kconfig/rootfs/runtime/tbox_group/test/testcase/torture_type: gcc-12/x86_64-randconfig-006-20240921/debian-12-x86_64-20240206.cgz/300s/vm-snb/cpuhotplug/rcutorture/srcud commit: daf812f27d6bb ("srcu: Standardize srcu_data pointers to "sdp" and similar") 26cb612c431de ("srcu: Add srcu_read_lock_lite() and srcu_read_unlock_lite()") f2f1565029268 <--- your patch daf812f27d6bbe4c 26cb612c431de5093ae6a3a67ec f2f1565029268e27760307bc82b ---------------- --------------------------- --------------------------- fail:runs %reproduction fail:runs %reproduction fail:runs | | | | | :50 66% 33:50 0% :200 dmesg.BUG:using__this_cpu_read()in_preemptible > > Thanx, Paul > > ------------------------------------------------------------------------ > > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c > index f3db4d42e70ab..07147efcb64d3 100644 > --- a/kernel/rcu/srcutree.c > +++ b/kernel/rcu/srcutree.c > @@ -482,7 +482,7 @@ static unsigned long srcu_readers_unlock_idx(struct > srcu_struct *ssp, int idx) > */ > static bool srcu_readers_active_idx_check(struct srcu_struct *ssp, int idx) > { > - bool did_gp = !!(__this_cpu_read(ssp->sda->srcu_reader_flavor) & > SRCU_READ_FLAVOR_LITE); > + bool did_gp = !!(raw_cpu_read(ssp->sda->srcu_reader_flavor) & > SRCU_READ_FLAVOR_LITE); > unsigned long unlocks; > > unlocks = srcu_readers_unlock_idx(ssp, idx); > > ------------------------------------------------------------------------
