Re: [fixup! EXP srcu] 7bf8d644f8: WARNING:at_kernel/rcu/srcutree.c:#process_srcu

[Oliver Sang]

hi, Paul,

On Wed, Mar 16, 2022 at 09:53:42AM -0700, Paul E. McKenney wrote:
> On Mon, Mar 14, 2022 at 01:38:14PM -0700, Paul E. McKenney wrote:
> > On Sat, Mar 12, 2022 at 11:59:19PM +0800, kernel test robot wrote:
> > > 
> > > 
> > > Greeting,
> > > 
> > > FYI, we noticed the following commit (built with gcc-9):
> > > 
> > > commit: 7bf8d644f869260f63eaaedebabdd2054f59ecc4 ("fixup! EXP srcu: Check for too many zero-delay srcu_reschedule() calls")
> > > https://github.com/ammarfaizi2/linux-block paulmck/linux-rcu/dev.2022.03.01a
> > > 
> > > in testcase: boot
> > > 
> > > on test machine: qemu-system-x86_64 -enable-kvm -cpu SandyBridge -smp 2 -m 16G
> > > 
> > > caused below changes (please refer to attached dmesg/kmsg for entire log/backtrace):
> > > 
> > > 
> > > 
> > > If you fix the issue, kindly add following tag
> > > Reported-by: kernel test robot <oliver.sang@xxxxxxxxx>
> > > 
> > > 
> > > [ 27.088275][ T42] WARNING: CPU: 1 PID: 42 at kernel/rcu/srcutree.c:1555 process_srcu (kernel/rcu/srcutree.c:1555 (discriminator 3)) 
> > > [   27.089252][   T42] Modules linked in: libata(+) serio_raw i2c_piix4 parport_pc(+) parport ip_tables
> > > [   27.090253][   T42] CPU: 1 PID: 42 Comm: kworker/1:1 Not tainted 5.17.0-rc1-00191-g7bf8d644f869 #1
> > > [   27.091212][   T42] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.12.0-1 04/01/2014
> > > [   27.092176][   T42] Workqueue: rcu_gp process_srcu
> > > [ 27.092718][ T42] RIP: 0010:process_srcu (kernel/rcu/srcutree.c:1555 (discriminator 3)) 
> > > [ 27.093653][ T42] Code: 6b f8 48 b8 00 00 00 00 00 fc ff df 48 c1 ea 03 80 3c 02 00 0f 85 91 02 00 00 48 8b 43 f8 31 f6 48 83 f8 32 0f 86 02 fe ff ff <0f> 0b e9 fb fd ff ff 4c 8d b3 58 ff ff ff 48 b8 00 00 00 00 00 fc
> > 
> > I had a feeling that the simple fix wasn't going to cut it, so thank you
> > for finding a scenario that proves that to be the case.  Fix in progress,
> > to be folded into the original with attribution.
> 
> OK, the attribution isn't quite there yet, but the squished-together
> commit is here:
> 
> ea0098f19115 ("srcu: Prevent expedited GPs and blocking readers from consuming CPU")
> 
> Does this version behave better for you?

we tested this commit, found the issue gone.
please be noted the issue on 7bf8d644f8 is random and ~25% rate.
so I queued to run 100 times for this new commit, which is finally clean
according to tests.

=========================================================================================
compiler/kconfig/rootfs/sleep/tbox_group/testcase:
  gcc-9/x86_64-rhel-8.3-func/debian-10.4-x86_64-20200603.cgz/1/vm-snb/boot

commit:
  feebd043ec6b ("Revert "srcu: Avoid expedited GPs and blocking readers from consuming CPU"")
  7bf8d644f869 ("fixup! EXP srcu: Check for too many zero-delay srcu_reschedule() calls")
  ea0098f19115 ("srcu: Prevent expedited GPs and blocking readers from consuming CPU")

feebd043ec6b9eab 7bf8d644f869260f63eaaedebab ea0098f191152c7fe3db5181a2e
---------------- --------------------------- ---------------------------
       fail:runs  %reproduction    fail:runs  %reproduction    fail:runs
           |             |             |             |             |
           :16          25%           4:16           0%            :100   dmesg.RIP:process_srcu
           :16          25%           4:16           0%            :100   dmesg.WARNING:at_kernel/rcu/srcutree.c:#process_srcu


> 
> 							Thanx, Paul
>