[PATCH] test: efi_selftest: Do not force serial# setting

[Tom Rini]

On Fri, Aug 07, 2020 at 09:56:58PM +0200, Heinrich Schuchardt wrote:
> On 8/7/20 9:50 PM, Tom Rini wrote:
> > On Fri, Aug 07, 2020 at 09:34:06PM +0200, Heinrich Schuchardt wrote:
> >
> >> On 7/31/20 11:42 PM, Tom Rini wrote:
> >>> On Fri, Jul 31, 2020 at 10:12:32AM -0400, Tom Rini wrote:
> >>>
> >>>> As part of the EFI self test we set and check the serial# variable.
> >>>> However, we should not be forcing this setting.  In the case where we
> >>>> are allowed to change the variable it will change, and we will pass the
> >>>> test.  In the case where we cannot change it, force may or may not be
> >>>> allowed, depending on further environment restrictions.  Drop the -f
> >>>> flag here as we do not need it.
> >>>>
> >>>> Cc: Heinrich Schuchardt <xypron.glpk at gmx.de>
> >>>> Signed-off-by: Tom Rini <trini at konsulko.com>
> >>>
> >>> Applied to u-boot/master, thanks!
> >>>
> >>
> >>
> >> With this patch on the Pine64 LTS:
> >>
> >> => => setenv efi_selftest device tree
> >> => => setenv serial# Testing DT
> >> ## Error: Can't overwrite "serial#"
> >> ## Error inserting "serial#" variable, errno=1
> >> =>
> >>
> >> This worked without your patch.
> >>
> >> This patch does not solve any problem.
> >
> > This patch resolved the problems trying to use "env set -f" on platforms
> > where CONFIG_ENV_ACCESS_IGNORE_FORCE is unset and the variable is not
> > protected.  With:
> > commit 0f036bf4b87e6416f5c4d23865a62a62d9073c20
> > Author: Marek Vasut <marex at denx.de>
> > Date:   Tue Jul 7 20:51:33 2020 +0200
> >
> >     env: Warn on force access if ENV_ACCESS_IGNORE_FORCE set
> >
> > applied those platforms fail efi_selftest.  The answer I believe is that
> > on your platform serial# is protected so you do have to force
> > overwriting it in order to change it.  This is not the case of all
> > platforms with a serial# variable.
> 
> So no problem is solved except Marek's fancy warning.

It's not a "fancy warning".  It's an important bit of user feedback.  In
fact, the next question I keep coming back to is, how was, or is, this
working before?  Without Marek's patch, we silently return 0 and don't
change the variable.  With Marek's patch we tell the user we aren't
doing anything and return 0.  So are we actually changing serial# when
it's supposed to be unchangable?

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20200807/16824adb/attachment.sig>