Re: [PATCH 2/4] i3c: master: Check if devices have i3c_dev_boardinfo on i3c_master_add_i3c_dev_locked()

[Boris Brezillon]

On Thu, 29 Aug 2019 15:57:32 +0000
Vitor Soares <Vitor.Soares@xxxxxxxxxxxx> wrote:

> -----Original Message-----
> From: Boris Brezillon 
> <boris.brezillon@xxxxxxxxxxxxx> 
> Sent: Thursday, August 29, 2019 4:25 
> PM
> To: Vitor Soares <Vitor.Soares@xxxxxxxxxxxx>
> Cc: 
> linux-kernel@xxxxxxxxxxxxxxx; devicetree@xxxxxxxxxxxxxxx; 
> linux-i3c@xxxxxxxxxxxxxxxxxxx; bbrezillon@xxxxxxxxxx; robh+dt@xxxxxxxxxx; 
> mark.rutland@xxxxxxx; Joao.Pinto@xxxxxxxxxxxx
> Subject: Re: [PATCH 2/4] 
> i3c: master: Check if devices have i3c_dev_boardinfo on 
> i3c_master_add_i3c_dev_locked()
> 
> On Thu, 29 Aug 2019 15:07:08 +0000
> Vitor Soares <Vitor.Soares@xxxxxxxxxxxx> wrote:
> 
> > From: Boris Brezillon   
> <boris.brezillon@xxxxxxxxxxxxx>
> > Date: Thu, Aug 29, 2019 at 15:39:41
> >   
> 
> > > On Thu, 29 Aug 2019 16:39:18 +0200
> > > Boris Brezillon <boris.brezillon@xxxxxxxxxxxxx> wrote:
> > >     
> > > > On Thu, 29 Aug 2019 14:00:44 +0000
> > > > Vitor Soares <Vitor.Soares@xxxxxxxxxxxx> wrote:
> > > >     
> > > > > Hi Boris,
> > > > > 
> > > > > From: Boris Brezillon <boris.brezillon@xxxxxxxxxxxxx>
> > > > > Date: Thu, Aug 29, 2019 at 11:44:57
> > > > >       
> > > > > > On Thu, 29 Aug 2019 12:19:33 +0200
> > > > > > Vitor Soares <Vitor.Soares@xxxxxxxxxxxx> wrote:
> > > > > >         
> > > > > > > The I3C devices described in DT might not be attached to the master which
> > > > > > > doesn't allow to assign a specific dynamic address.        
> > > > > > 
> > > > > > I remember testing this when developing the framework, so, unless
> > > > > > another patch regressed it, it should already work. I suspect patch 1
> > > > > > is actually the regressing this use case.        
> > > > > 
> > > > > For today it doesn't address the case where the device is described with 
> > > > > static address = 0, which isn't attached to the controller.      
> > > > 
> > > > Hm, I'm pretty sure I had designed the code to support that case (see
> > > > [1]). It might be buggy, but nothing we can't fix I guess.
> > > >     
> > > 
> > > [1]https://urldefense.proofpoint.com/v2/url?u=https-3A__elixir.bootlin.com_linux_v5.3-2Drc6_source_drivers_i3c_master.c-23L1898&d=DwICAg&c=DPL6_X_6JkXFx7AXWqB0tg&r=qVuU64u9x77Y0Kd0PhDK_lpxFgg6PK9PateHwjb_DY0&m=IXS1ygIgEo5vwajk0iwd5aBDVBzRnVTjO3cg4iBmGNc&s=HC-AcYm-AZPrUBoALioej_BDnqOtJHltr39Z2yPkuU4&e=     
> > 
> > That is only valid if you have olddev which will only exist if static 
> > address != 0.  
> 
> Hm, if you revert patch 1 (and assuming the device is properly defined
> in the DT), you should have olddev != NULL when reaching that point. If
> that's not the case there's a bug somewhere that should be fixed.
> 
> No, because the device is not attached.

Oh, my bad, I see what you mean now. This is definitely a bug and should
have the Fixes tags. I mean, even if we don't care about dynamic
address assignment, I3C drivers might care about the ->of_node that's
attached to the device.