On 05.11.2018 [11:54:39 +0700], Dennis Zhou wrote: > Hi Nish, > > On Thu, Nov 01, 2018 at 12:06:44PM -0700, Tejun Heo wrote: > > ---------- Forwarded message --------- > > From: Nishanth Aravamudan <naravamudan@xxxxxxxxxxxxxxxx> > > Date: Thu, Nov 1, 2018 at 3:03 PM > > Subject: Kernel panic when enabling cgroup2 io controller at runtime > > To: Tejun Heo <tj@xxxxxxxxxx>, Li Zefan <lizefan@xxxxxxxxxx>, Johannes > > Weiner <hannes@xxxxxxxxxxx> > > Cc: <cgroups@xxxxxxxxxxxxxxx> > > > > > > Hi, > > > > tl;dr: I see a kernel NULL pointer dereference with Linus' master > > (7c6c54b5) when enabling the IO cgroup2 controller at runtime. Is this > > PEBKAC and if so what config option am I missing? > > I don't think you're missing something. I ran a patch series here that > changed blkcg to do more accurate accounting. However, it seems that I > didn't correctly handle all the cases. My guess is this is what caused > the oops. It has been reverted in b5f2954d30c7. > > The original patch series is [1]. > > > > > [ 1015.243027] BUG: unable to handle kernel NULL pointer dereference at > > 0000000000000000 <snip> > Thanks for providing the oops and the details! Do you mind testing rc1 > to make sure this issue is resolved? I can confirm that rebased onto Linus' master (4.20-rc2), the oopsen are gone. > Second, can you tell me a little more about your disk setup so I can > more easily reproduce it? The oops above has the raid10 driver in the > call stack. This test matchine has 6 partitions in a software raid1 (/boot) and another 6 partitions in a software raid10 which then has LVM on top. I guess it must have been the latter stack that was throwing the oops? What specific details do you need to help reproduce the issue? Sorry for the delay in replying on my end, I was on vacation. Thanks! -Nish
