Re: [PATCH rcu 12/12] srcu: Block less aggressively for expedited grace periods

[Paul E. McKenney]

On Tue, Jun 21, 2022 at 10:00:07AM +0800, Zhangfei Gao wrote:
> 
> 
> On 2022/6/21 上午6:20, Paul E. McKenney wrote:
> > Commit 282d8998e997 ("srcu: Prevent expedited GPs and blocking readers
> > from consuming CPU") fixed a problem where a long-running expedited SRCU
> > grace period could block kernel live patching.  It did so by giving up
> > on expediting once a given SRCU expedited grace period grew too old.
> > 
> > Unfortunately, this added excessive delays to boots of embedded systems
> > running on qemu that use the ARM IORT RMR feature.  This commit therefore
> > makes the transition away from expediting less aggressive, increasing
> > the per-grace-period phase number of non-sleeping polls of readers from
> > one to three and increasing the required grace-period age from one jiffy
> > (actually from zero to one jiffies) to two jiffies (actually from one
> > to two jiffies).
> > 
> > Fixes: 282d8998e997 ("srcu: Prevent expedited GPs and blocking readers from consuming CPU")
> > Signed-off-by: Paul E. McKenney <paulmck@xxxxxxxxxx>
> > Reported-by: Zhangfei Gao <zhangfei.gao@xxxxxxxxxx>
> > Reported-by: chenxiang (M)" <chenxiang66@xxxxxxxxxxxxx>
> > Cc: Shameerali Kolothum Thodi  <shameerali.kolothum.thodi@xxxxxxxxxx>
> > Cc: Paolo Bonzini <pbonzini@xxxxxxxxxx>
> > Link: https://lore.kernel.org/all/20615615-0013-5adc-584f-2b1d5c03ebfc@xxxxxxxxxx/
> 
> Test on 5.19-rc1 with this patch with qemu boot with -bios QEMU_EFI-2022.fd,
> seems not work, same as rc1.
> 
> real    2m42.948s
> user    0m2.843s
> sys     0m1.170s
> 
> qemu: stable-6.1
> 
> build/aarch64-softmmu/qemu-system-aarch64 -machine
> virt,gic-version=3,iommu=smmuv3 \
> -enable-kvm -cpu host -m 1024 \
> -kernel /home/linaro/Image -initrd /home/linaro/tmp/ramdisk-new.img
> -nographic -append \
> "rdinit=init console=ttyAMA0 earlycon=pl011,0x9000000 kpti=off acpi=force" \
> -bios QEMU_EFI-2022.fd

Understood.  This patch fixes some cases, but not your case.  Which is
why you guys are experimenting with additional changes.  In the meantime,
this patch helps at least some people.  I look forward to you guys have
an appropriate solution that I can pull in on top of this one.

Or, if the solution shows up quickly enough, I can replace this patch
with your guys' solution.

							Thanx, Paul

> Thanks
> > ---
> >   kernel/rcu/srcutree.c | 20 +++++++++++++-------
> >   1 file changed, 13 insertions(+), 7 deletions(-)
> > 
> > diff --git a/kernel/rcu/srcutree.c b/kernel/rcu/srcutree.c
> > index 50ba70f019dea..0db7873f4e95b 100644
> > --- a/kernel/rcu/srcutree.c
> > +++ b/kernel/rcu/srcutree.c
> > @@ -513,7 +513,7 @@ static bool srcu_readers_active(struct srcu_struct *ssp)
> >   #define SRCU_INTERVAL		1	// Base delay if no expedited GPs pending.
> >   #define SRCU_MAX_INTERVAL	10	// Maximum incremental delay from slow readers.
> > -#define SRCU_MAX_NODELAY_PHASE	1	// Maximum per-GP-phase consecutive no-delay instances.
> > +#define SRCU_MAX_NODELAY_PHASE	3	// Maximum per-GP-phase consecutive no-delay instances.
> >   #define SRCU_MAX_NODELAY	100	// Maximum consecutive no-delay instances.
> >   /*
> > @@ -522,16 +522,22 @@ static bool srcu_readers_active(struct srcu_struct *ssp)
> >    */
> >   static unsigned long srcu_get_delay(struct srcu_struct *ssp)
> >   {
> > +	unsigned long gpstart;
> > +	unsigned long j;
> >   	unsigned long jbase = SRCU_INTERVAL;
> >   	if (ULONG_CMP_LT(READ_ONCE(ssp->srcu_gp_seq), READ_ONCE(ssp->srcu_gp_seq_needed_exp)))
> >   		jbase = 0;
> > -	if (rcu_seq_state(READ_ONCE(ssp->srcu_gp_seq)))
> > -		jbase += jiffies - READ_ONCE(ssp->srcu_gp_start);
> > -	if (!jbase) {
> > -		WRITE_ONCE(ssp->srcu_n_exp_nodelay, READ_ONCE(ssp->srcu_n_exp_nodelay) + 1);
> > -		if (READ_ONCE(ssp->srcu_n_exp_nodelay) > SRCU_MAX_NODELAY_PHASE)
> > -			jbase = 1;
> > +	if (rcu_seq_state(READ_ONCE(ssp->srcu_gp_seq))) {
> > +		j = jiffies - 1;
> > +		gpstart = READ_ONCE(ssp->srcu_gp_start);
> > +		if (time_after(j, gpstart))
> > +			jbase += j - gpstart;
> > +		if (!jbase) {
> > +			WRITE_ONCE(ssp->srcu_n_exp_nodelay, READ_ONCE(ssp->srcu_n_exp_nodelay) + 1);
> > +			if (READ_ONCE(ssp->srcu_n_exp_nodelay) > SRCU_MAX_NODELAY_PHASE)
> > +				jbase = 1;
> > +		}
> >   	}
> >   	return jbase > SRCU_MAX_INTERVAL ? SRCU_MAX_INTERVAL : jbase;
> >   }
>