Re: i915 potential deadlock

[Rodrigo Vivi <rodrigo.vivi@intel.com>]

On Sat, Nov 02, 2024 at 01:46:15AM +0000, Alexander Coffin wrote:
>    Hi Rodrigo,                                                                  
>                                                                                 
>    > I'm sorry for that. Our list is moderated for non-subscribers because of   
>    the amount of spams we receive daily. But I eventually got to your msg and   
>    allowed it to go through.                                                    
>                                                                                 
>    I am sorry about sending the email twice in that case. In hindsight, I       
>    probably should have subscribed to the list and then sent the email to       
>    avoid this issue, but I don't interact with email lists that often so I      
>    didn't think of doing that.                                                  
>                                                                                 
>    > Could you please file a bug in our gitlab?                                 
>                                                                                 
>    Of course.                                                                   
>    [1]https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12677             

Very interesting race...

Bringing some folks here that could have some ideas. I don't believe that we
should be accepting that plane pin call before we finish the probe's initial
plane object.

Folks, any quick idea?


>                                                                                 
>    > Well, so many things changed from 6.1... Is it possible to test with a     
>    drm-tip kernel and see if it is still reproduced?                            
>                                                                                 
>    Sorry, I mispoke in my last email that 6.1 is the latest LTS release. 6.6    
>    is the latest LTS release. 6.1 is the default for Debian though and isn't    
>    considered old there, and that is the kernel that I use on my personal       
>    desktop, personal laptop, server, and work laptop.                           

Nevermind. This part of the code hasn't changed since 6.1, so we likely still have
this bug.

We usually ask to test a recent kernel so it is easier to backport some know fix.
But this is not the case.

>                                                                                 
>    I was hoping that this bug could be debugged (or at least attempted) with    
>    a single deadlock report (that has worked well most other times I have       
>    reported similar issues to kernel maintainers for other drivers).            
>                                                                                 
>    If I knew how to recreate this issue I would gladly try out any kernel       
>    version for a quick test. Or if you have any things that you specifically    
>    want me to test then I would be happy to try a newer kernel version.         
>    However, I am not really comfortable daily driving a bleeding edge kernel    
>    version on my work laptop as of right now. I may be able to read through     
>    the source code to try to understand what could be triggering this issue     
>    and then figure out how to use that information to recreate it on a newer    
>    version, but as I don't think I have seen the issue since I reported it      
>    I'm not sure how likely it is that I stumble upon the issue by chance        
>    again. I would feel comfortable daily driving random kernels on my           
>    personal laptop, but I use an XPS 9370 for my personal laptop and I          
>    believe an XPS 15 9530 at work (if this is helpful information I can         
>    double check this and provide more system details) so they aren't the same   
>    hardware (I have no idea if the hardware is relevant).                       
>                                                                                 
>    Do you know where I could find information on the versions supported by      
>    Intel of the i915 driver? I can't seem to find anything from doing a few     
>    minutes of Google searching.                                                 
>                                                                                 
>    Best wishes,                                                                 
>                                                                                 
>    Alex Coffin                                                                  
> 
> References
> 
>    Visible links
>    1. https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/12677