Re: [PATCH v3] tracing: Have format file honor EVENT_FILE_FL_FREED

[Mathias Krause]

On 30.07.24 17:15, Steven Rostedt wrote:
> On Tue, 30 Jul 2024 11:06:57 -0400
> Steven Rostedt <rostedt@xxxxxxxxxxx> wrote:
> 
>> Link: https://lore.kernel.org/all/20240719204701.1605950-1-minipli@xxxxxxxxxxxxxx/
>>
>> Cc: stable@xxxxxxxxxxxxxxx
>> Fixes: b63db58e2fa5d ("eventfs/tracing: Add callback for release of an eventfs_inode")
>> Reported-by: Mathias Krause <minipli@xxxxxxxxxxxxxx>
>> Tested-by: Mathias Krause <minipli@xxxxxxxxxxxxxx>
> 
> Hi Mathias,
> 
> I kept your "Tested-by" tag because you did test the part that fixes the
> bug you reported. I just added on top of that one to handle other possibly
> cases. Let me know if you want me to drop that tag, or if you tested this
> version as well.

Hi Steven,

I retested this version and it's still fixing the bug, so you can keep
the "Tested-by" tag.

Also nice catch with covering the write part of "filter" as well!
Testing this specifically, I saw a few ENODEV's, so it's definitely
possible to hit this one as well.

Thanks,
Mathias

> 
> Thanks,
> 
> -- Steve
> 
> 
>> Signed-off-by: Steven Rostedt (Google) <rostedt@xxxxxxxxxxx>
>> ---
>> Changes since v3: https://lore.kernel.org/20240726132811.306a449e@xxxxxxxxxxxxxxxxxxxx
>>
>> - After inspecting the code, there's several users of event_file_data()
>>   that return the trace_event_file descriptor when it could be set to be
>>   freed. Add a new helper function: event_file_file() that will make sure
>>   the event_mutex is held and check the flag and return NULL if the file is
>>   freed.