Re: [PATCH v3 1/4] cmd: bind: Add unbind command with driver filter

[Miquel Raynal <miquel.raynal@bootlin.com>]

Hi Marek,

marex@denx.de wrote on Wed, 2 Aug 2023 16:38:47 +0200:

> On 8/2/23 09:48, Miquel Raynal wrote:
> > Hi Marek,
> > 
> > marex@denx.de wrote on Wed, 2 Aug 2023 01:07:46 +0200:
> >   
> >> On 8/1/23 20:53, Miquel Raynal wrote:  
> >>> Hi Marek,
> >>>
> >>> marex@denx.de wrote on Mon, 31 Jul 2023 16:40:04 +0200:  
> >>>    >>>> On 7/31/23 16:25, Miquel Raynal wrote:  
> >>>>> Hi Marek,
> >>>>>
> >>>>> marex@denx.de wrote on Mon, 31 Jul 2023 16:08:19 +0200:  
> >>>>>     >>>> On 7/31/23 15:58, Miquel Raynal wrote:  
> >>>>>>> Hi Marek,
> >>>>>>>
> >>>>>>> marex@denx.de wrote on Mon, 31 Jul 2023 15:50:58 +0200:  
> >>>>>>>      >>>> On 7/31/23 15:36, Miquel Raynal wrote:  
> >>>>>>>>> Hi Marek,
> >>>>>>>>>
> >>>>>>>>> marex@denx.de wrote on Mon, 31 Jul 2023 13:44:25 +0200:  
> >>>>>>>>>       >>>> On 7/31/23 11:31, Miquel Raynal wrote:  
> >>>>>>>>>>> Hi Marek,
> >>>>>>>>>>>
> >>>>>>>>>>> marex@denx.de wrote on Sat, 29 Jul 2023 16:57:09 +0200:  
> >>>>>>>>>>>        >>>> Extend the driver core to perform lookup by both OF node and driver  
> >>>>>>>>>>>> bound to the node. Use this to look up specific device instances to
> >>>>>>>>>>>> unbind from nodes in the unbind command. One example where this is
> >>>>>>>>>>>> needed is USB peripheral controller, which may have multiple gadget
> >>>>>>>>>>>> drivers bound to it. The unbind command has to select that specific
> >>>>>>>>>>>> gadget driver instance to unbind from the controller, not unbind the
> >>>>>>>>>>>> controller driver itself from the controller.
> >>>>>>>>>>>>
> >>>>>>>>>>>> USB ethernet gadget usage looks as follows with this change. Notice
> >>>>>>>>>>>> the extra 'usb_ether' addition in the 'unbind' command at the end.
> >>>>>>>>>>>> "
> >>>>>>>>>>>> bind /soc/usb-otg@49000000 usb_ether
> >>>>>>>>>>>> setenv ethact usb_ether
> >>>>>>>>>>>> setenv loadaddr 0xc2000000
> >>>>>>>>>>>> setenv ipaddr 10.0.0.2
> >>>>>>>>>>>> setenv serverip 10.0.0.1
> >>>>>>>>>>>> setenv netmask 255.255.255.0
> >>>>>>>>>>>> tftpboot 0xc2000000 10.0.0.1:test.file
> >>>>>>>>>>>> unbind /soc/usb-otg@49000000 usb_ether
> >>>>>>>>>>>> "
> >>>>>>>>>>>>
> >>>>>>>>>>>> Signed-off-by: Marek Vasut <marex@denx.de>
> >>>>>>>>>>>> ---  
> >>>>>>>>>>>
> >>>>>>>>>>> I am no longer getting wrong pointer dereferences, the SPL is working in
> >>>>>>>>>>> recovery mode, TFTP "File not found" errors are no longer a problem and
> >>>>>>>>>>> I did not experience any reset while tftp'ing regular files.
> >>>>>>>>>>>
> >>>>>>>>>>> One last remaining request on my side is the need for using fastboot as
> >>>>>>>>>>> well which does no longer work as-is:  
> >>>>>>>>>>>        >>> => fastboot usb 0  
> >>>>>>>>>>> couldn't find an available UDC
> >>>>>>>>>>> g_dnl_register: failed!, error: -19
> >>>>>>>>>>> exit not allowed from main input shell.
> >>>>>>>>>>>
> >>>>>>>>>>> Can you advise what bind/unbind command would be necessary here?  
> >>>>>>>>>>
> >>>>>>>>>> Either 'unbind usb_ether' or run 'dm tree' -> look up the path to usb_ether in the tree (it will be hanging under usb_peripheral or some such), and then use 'unbind <that path>'.  
> >>>>>>>>>
> >>>>>>>>> Nice `dm tree` command, never used it before.
> >>>>>>>>>
> >>>>>>>>> Even when I unbind usb_ether I still get the same error:  
> >>>>>>>>>       >>> => unbind /ocp/usb@47400000/usb@47401000  
> >>>>>>>>> => fastboot usb 0  
> >>>>>>>>> couldn't find an available UDC
> >>>>>>>>> g_dnl_register: failed!, error: -19
> >>>>>>>>> exit not allowed from main input shell.
> >>>>>>>>>
> >>>>>>>>> Is there a specific gadget driver which I should bind again manually?  
> >>>>>>>>
> >>>>>>>> Can you share the output of dm tree before/after unbind ?
> >>>>>>>>
> >>>>>>>> fastboot should auto-bind to the right thing.  
> >>>>>>>
> >>>>>>> Ok. Apparently it does not, but I don't have any clue why. If you want
> >>>>>>> me to check something else I will. Here is the output:
> >>>>>>>
> >>>>>>> U-Boot 2023.07-00806-g979e7443428 (Jul 31 2023 - 11:17:06 +0200)  
> >>>>>>
> >>>>>> [...]  
> >>>>>>    >>>>>     watchdog      0  [ + ]   omap3_wdt             |   |-- wdt@44e35000  
> >>>>>>>      misc          0  [ + ]   ti-musb-wrapper       |   |-- usb@47400000
> >>>>>>>      usb           0  [ + ]   ti-musb-peripheral    |   |   |-- usb@47401000
> >>>>>>>      ethernet      1  [ + ]   usb_ether             |   |   |   `-- usb_ether
> >>>>>>>      bootdev       3  [   ]   eth_bootdev           |   |   |       `-- usb_ether.bootdev
> >>>>>>>      usb           0  [   ]   ti-musb-host          |   |   `-- usb@47401800
> >>>>>>>      ethernet      0  [ + ]   eth_cpsw              |   |-- ethernet@4a100000
> >>>>>>>      bootdev       2  [   ]   eth_bootdev           |   |   `-- ethernet@4a100000.bootdev
> >>>>>>>      simple_bus   67  [   ]   ti_sysc               |   |-- target-module@53100000
> >>>>>>>      simple_bus   68  [   ]   ti_sysc               |   |-- target-module@53500000
> >>>>>>>      simple_bus   69  [   ]   ti_sysc               |   `-- target-module@56000000
> >>>>>>>      clk          62  [   ]   fixed_clock           |-- clk_mcasp0_fixed
> >>>>>>>      bootstd       0  [   ]   bootstd_drv           |-- bootstd
> >>>>>>>      bootmeth      0  [   ]   bootmeth_efi          |   |-- efi
> >>>>>>>      bootmeth      1  [   ]   bootmeth_extlinux     |   |-- extlinux
> >>>>>>>      bootmeth      2  [   ]   bootmeth_pxe          |   |-- pxe
> >>>>>>>      bootmeth      3  [   ]   vbe_simple            |   `-- vbe_simple
> >>>>>>>      timer         0  [ + ]   omap_timer            `-- timer@0  
> >>>>>>> => unbind /ocp/usb@47400000/usb@47401000  
> >>>>>>
> >>>>>> The commit message of this patch contains the following example:
> >>>>>> unbind /soc/usb-otg@49000000 usb_ether
> >>>>>> so in your case, try
> >>>>>> unbind /ocp/usb@47400000/usb@47401000 usb_ether  
> >>>>>
> >>>>> Does not work, unfortunately:  
> >>>>> => unbind /ocp/usb@47400000/usb@47401000 usb_ether  
> >>>>> Cannot find a device with path /ocp/usb@47400000/usb@47401000  
> >>>>> => unbind /ocp/usb@47400000/usb@47401000  
> >>>>
> >>>> Did this even work before , i.e. without this series, if you were to use USB ethernet first and then fastboot second (that sequence is important) ?  
> >>>
> >>> Yes it did.  
> >>>    >>>> Can you debug this ? Basically the problem that happens is that if you run:  
> >>>> unbind /ocp/usb@47400000/usb@47401000
> >>>> this unbind ti-musb-peripheral usb@47401000 instead of just unbinding
> >>>> usb_ether usb_ether  
> >>>
> >>> True.  
> >>>    >>>> Maybe just try  
> >>>> unbind usb_ether  
> >>>    >>> => unbind usb_ether  
> >>> unbind - Unbind a device from a driver
> >>>
> >>> Usage:
> >>> unbind <node path> [<driver>]
> >>> unbind <class> <index>
> >>> unbind <class> <index> <driver>  
> >>>    >>>> or  
> >>>> unbind usb_ether usb_ether  
> >>>    >>> => unbind usb_ether usb_ether  
> >>> usb_ether is not a valid uclass  
> >>>    >>>> ti-musb-peripheral usb@47401000 has to stay in the dm tree output after the unbind. Or maybe try  
> >>>>
> >>>> bind /ocp/usb@47400000/usb@47401000 ti-musb-peripheral  
> >>>
> >>> Indeed when we perform the unbind, ti-musb-peripheral is removed. Yes,
> >>> binding it again like you advise makes it come back.  
> >>>    >>>> and then the fastboot thing  
> >>>
> >>> Unfortunately after binding it again, so with the following dm tree
> >>> snippet:
> >>>
> >>>    misc          0  [ + ]   ti-musb-wrapper       |   |-- usb@47400000
> >>>    usb           0  [   ]   ti-musb-host          |   |   |-- usb@47401800
> >>>    usb           0  [   ]   ti-musb-peripheral    |   |   `-- usb@47401000
> >>>
> >>> It still fails:  
> >>>    >>> => fastboot usb 0  
> >>> couldn't find an available UDC
> >>> g_dnl_register: failed!, error: -19
> >>> exit not allowed from main input shell.
> >>>
> >>> Without this series, this works:  
> >>> => unbind /ocp/usb@47400000/usb@47401000
> >>> => bind /ocp/usb@47400000/usb@47401000 ti-musb-peripheral
> >>> => fastboot usb 0  
> >>
> >> OK, so fastboot did not work even without this series without obscure workarounds ?  
> > 
> > I'm happy you call the unbind/bind sequence an "obscure workaround" ;-)
> > 
> > The command "fastboot usb 0" works without your series, before tftp
> > or after tftp, it does not matter, it used to work. Applying this
> > series makes it always fail.
> > 
> > I was just reporting that, before your series, calling unbind would
> > also remove ti-musb-peripheral, but calling bind again with the above
> > line makes the usb_ether gadget and fastboot work again. Whereas with
> > your series the unbind behaves identically but the bind command does not
> > "fix" fastboot (tftp works fine in v3).
> >   
> >> Can you try and debug this ?  
> > 
> > The breakage is not part of the existing code base, it was introduced
> > by the series. Maybe you can try with any other board with USB
> > peripheral and see if the behavior is identical?
> >   
> >> Does 'fastboot usb 1' work with this series by any chance ?  
> > 
> > When fastboot works, starting it "blocks" the CLI. When I unbind the
> > peripheral driver on USB 0, the fastboot command does not block any
> > more and yet does not print anything, it just returns "immediately"
> > without any log.
> > 
> > When I run 'fastboot usb 1' it acts like that as well, nothing happens,
> > without error. It acts identically before and after this series.  
> 
> Please test v4

I will.

Thanks,
Miquèl