Re: possible trace_printk() bug in v5.19-rc1

[Chuck Lever III <chuck.lever@oracle.com>]

> On Jun 16, 2022, at 11:34 AM, Steven Rostedt <rostedt@goodmis.org> wrote:
> 
> On Thu, 16 Jun 2022 15:22:26 +0000
> Chuck Lever III <chuck.lever@oracle.com> wrote:
> 
>> Hi Steve-
>> 
>> Since v5.19-rc1, trace_printk() no longer displays a function
>> name, but rather a kernel address. I don't have time at the
>> moment to track this down to a specific commit, just parking
>> it here in case someone can get to it sooner than me.
>> 
>> <idle>-0 [003] 269.397568: bprint: 0xffffffffc0cccab7: nf=0xffff888172dbfba0 still on the LRU or a dispose list
>> <idle>-0 [003] 269.397576: bprint: 0xffffffffc0cccafa: nf=0xffff888172dbfba0
>> <idle>-0 [003] 269.397583: bprint: 0xffffffffc0cccab7: nf=0xffff8881726cd4d8 still on the LRU or a dispose list
>> <idle>-0 [003] 269.397584: bprint: 0xffffffffc0cccafa: nf=0xffff8881726cd4d8
>> 
> 
> It uses the normal vsprintf() in the kernel (that is, it prints the
> function address just like printk() does with %pS). So if trace_printk() is
> broken, then so is printk().
> 
> But one reason for this is if you have a trace_printk() in a module, do the
> record, and then remove the module. The function name will also be removed
> from kallsyms, leaving you with only the function address.
> 
> Did you remove any modules before displaying the trace?

I haven't explicitly removed nfsd.ko, and lsmod says it's still there.
And, trace_printk was working as expected on v5.18.


--
Chuck Lever