Re: [cfs_trace_lock_tcd] BUG: unable to handle kernel NULL pointer dereference at 00000050

[Fengguang Wu]

Hi James,

On Wed, Apr 18, 2018 at 02:59:15PM +0100, James Simmons wrote:
> > Hello,
> >
> > FYI this happens in mainline kernel 4.17.0-rc1.
> > It looks like a new regression.
> >
> > [    7.587002]  lnet_selftest_init+0x2c4/0x5d9:
> > 						lnet_selftest_init at drivers/staging/lustre/lnet/selftest/module.c:134
> > [    7.587002]  ? lnet_selftest_exit+0x8d/0x8d:
> > 						lnet_selftest_init at drivers/staging/lustre/lnet/selftest/module.c:90
>
> Are you running lnet selftest ?

Perhaps yes -- it's randconfig boot test and the .config does include
CONFIG_LNET_SELFTEST:

       CONFIG_LNET=y
       CONFIG_LNET_MAX_PAYLOAD=1048576
==>     CONFIG_LNET_SELFTEST=y
       CONFIG_LNET_XPRT_IB=y

> Is this a UMP setup?

Yes, .config has:

       # CONFIG_SMP is not set

> The reason I ask is that their is a SMP handling bug in lnet
> selftest. If you look at the mailing list I pushed a SMP patch
> series. Can you try that series and tell me if it works for you.

So it looks your fixup patch is not for this case? Anyway the
reproduce-* script attached in the previous email should be fairly
straightforward to try out for reproducing the bug.

Thanks,
Fengguang