Re: BUG: KCSAN: data-race in mutex_spin_on_owner+0xef/0x1b0

[Peter Zijlstra]

On Fri, Dec 04, 2020 at 11:51:39PM +0530, Naresh Kamboju wrote:
> LKFT started testing KCSAN enabled kernel from the linux next tree.
> Here we have found BUG: KCSAN: data-race in mutex_spin_on_owner
> and several more KCSAN BUGs.
> 
> This report is from an x86_64 machine clang-11 linux next 20201201.
> Since we are running for the first time we do not call this regression.
> 
> [    4.745161] usbcore: registered new interface driver cdc_ether
> [    4.751281] ==================================================================
> [    4.756653] usbcore: registered new interface driver net1080
> [    4.752139] BUG: KCSAN: data-race in mutex_spin_on_owner+0xef/0x1b0

At the very least run your splat through ./scripts/decode_stacktrace.sh

It's impossible to know what the thing is complaining about. I suspect
this is one of those known KCSAN 'bugs' where it can't tell a load+cmp0
is perfectly fine.