[RFC] dfu: remove UPDATE_TFTP

[AKASHI Takahiro]

On Wed, Jul 15, 2020 at 04:42:20PM +0200, Heinrich Schuchardt wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA256
> 
> On 15.07.20 16:06, Michal Simek wrote:
> >
> >
> > On 15. 07. 20 15:23, Lukasz Majewski wrote:
> >> Hi Heinrich,
> >>
> >> On 15.07.20 13:24, Lukasz Majewski wrote:
> >>>>> Hi Heinrich,
> >>>>>
> >>>>>> Using UPDATE_TFTP the firmware can be updated from tFTP
> >>>>>> by writing to NOR flash. The same is possible by defining
> >>>>>> a dfu command in CONFIG_PREBOOT.
> >>>>>>
> >>>>>> The dfu command cannot only write to NOR but also to
> >>>>>> other devices. In README.dfutfp UPDATE_TFTP has been
> >>>>>> marked as deprecated.
> >>>>>
> >>>>> Could you also write in the proper README the steps
> >>>>> necessary to have the same functionality as with
> >>>>> UPDATE_TFTP with dfu and preboot? I think that it would be
> >>>>> good to have it written down in ./doc.
> >>>>>
> >>
> >> doc/README.dfutftp already describes the necessary steps as:
> >>
> >> * enable CONFIG_DFU_TFTP * enable CONFIG_PREBOOT * put a string
> >> like "dfu tftp 0 mmc 0" into the "preboot" env variable
> >>
> >> Do you want me to explicitly write that "dfu tftp 0 mtd 0"
> >> replaces the former auto-update functionality?
> >
> > I am curious if that command is really dfu tftp 0 mtd 0.
> 
> Lukasz was asking about the replacement of the update that was
> executed when running with CONFIG_UPDATE_TFTP=y. This function that I
> am removing could only write to NOR flash. The update defined via
> CONFIG_PREBOOT is much more flexible and can server other storage media.

One of my concerns regarding CONFIG_PREBOOT is that it depends
on the "preboot" variable.
What this means is that, if a user unsets this variable (accidentally
or not), 'automatic firmware update' won't be executed.
This can be a critical issue if the system administrator intends to
*enforce* the update.
(I know there are more issues regarding enforcement though.)

That is why I added UEFI capsule update hook in main_loop() as
a compile-time option.

Thanks,
-Takahiro Akashi


> The different accessible devices are described in doc/README.dfu.
> "ram" is one of them.
> 
> Best regards
> 
> Heinrich
> 
> >
> > I have just tested it on zynqmp with copying images to ram and
> > command is just dfu tftp ram 0. It means I think it would be worth
> > to list options which are really used by different back ends
> > because it is not fully clear from help how to exactly define it.
> >
> > dfu tftp [<interface> <dev>] [<addr>] - device firmware upgrade via
> > TFTP on device <dev>, attached to interface <interface> [<addr>] -
> > address where FIT image has been stored
> >
> > Thanks, Michal
> >
> 
> -----BEGIN PGP SIGNATURE-----
> 
> iQIzBAEBCAAdFiEEbcT5xx8ppvoGt20zxIHbvCwFGsQFAl8PFcUACgkQxIHbvCwF
> GsSwLQ//bU8AISZ5q4cgLcnAVPYnBPbVtoPXlyRROnX6XW3eDjq506Ah0Rk8UwYH
> lhLSM1m+bfJZzVtWUjymlRGjuaB1HLzcyd1PbfK0+50mMaVnFBV5PMU4ue6Ju+zT
> BukpUvtQ4KLpooo2QhjCz4/oXmur782RUJGq6hb7mDbk6R0+L8coqoiglLid5374
> GdOQhgTAa8jWu0QxIBbdeOOrTSZz1kOBaShYffKXfBCaUfMOf/2T6sxDrSuTB8/N
> WTeKdbTAsx5BVEY8s9EaAy6a7kWf8mb/7LdooFUMQ2noNbdw/ucfiNCA+H2JNTK1
> GjvzDgViWAF0A0sXu6c1fqFT+4xo2lpRvG0wXsz3zRwpggmlAVX5PPi7o4Xj2HVF
> 7HPBVxz4AjJ3YpOpyLm0rEVEXnwBdke/jy/2r0VRyn5QMFaqAMEkrA39/wHrjj6t
> 0s4EUeuy5IjMf51zUrTAE/J9xHowLwjldODRgwmWoEYShJRgZp+J0yr2nhGTwpyS
> 91ESrHzkok/5BPkYFYGulBeeJFDOBegA7N7xec5e6YsXrWEPDiOpS5I1AfbycceV
> lERhJpHIgPPjpaMgNcATKeti85o7veNe31IMoW758i5kbjQFOB8NlvtSSu97/2/K
> eB/dIrOdnzC+URaun+MEt0T4HYPYgEuGAsBPQJKYrTw0sFhJRkg=
> =JXjF
> -----END PGP SIGNATURE-----