[PATCH] distro_boot: Fix bootfile env after calling boot_extlinux

[Simon Glass]

Hi Art,

On Tue, 12 Oct 2021 at 21:45, Art Nikpal <email2tema at gmail.com> wrote:
>
> Yes changes inside include/config_distro_bootcmd.h not the best solution for this issue.
> I think it is better to change sysboot cmd and i have prepared another solution already!
> https://patchwork.ozlabs.org/project/uboot/patch/20211013033912.3297227-1-art@khadas.com/
> how about this ?

Seem reasonable to me, but could you check out my syslinux refactor
series? It could use some review.

http://patchwork.ozlabs.org/project/uboot/list/?series=264265


- Simon

>
> On Wed, Oct 13, 2021 at 5:07 AM Tom Rini <trini at konsulko.com> wrote:
>>
>> On Tue, Oct 12, 2021 at 02:31:18PM -0600, Simon Glass wrote:
>> > Hi Tom,
>> >
>> > On Tue, 12 Oct 2021 at 13:44, Tom Rini <trini at konsulko.com> wrote:
>> > >
>> > > On Tue, Oct 12, 2021 at 04:55:44PM +0800, Artem Lapkin wrote:
>> > >
>> > > > Problem
>> > > >
>> > > > PXE cannot boot normally after Sysboot changed the bootfile env (called
>> > > > from boot_extlinux) from the default "boot.scr.uimg" to
>> > > > "/boot/extlinux/extlinux.conf".
>> > > >
>> > > > In addition, an unbootable extlinux configuration will also make the PXE
>> > > > boot unbootable, because it will use the incorrect path "/boot/extlinux/"
>> > > > from the bootfile env.
>> > > >
>> > > > Solution
>> > > >
>> > > > Save and restore default bootfile env value when boot_extlinux is used.
>> > > >
>> > > > Example
>> > > > ================================================================
>> > > > Hit SPACE in 2 seconds to stop autoboot
>> > > > ... is now current device
>> > > > Found /boot/extlinux/extlinux.conf
>> > > > Retrieving file: /boot/extlinux/extlinux.conf
>> > > > 413 bytes read in 2 ms (201.2 KiB/s)
>> > > > Skipping Krescue for failure retrieving kernel
>> > > > SCRIPT FAILED: continuing...
>> > > > ...
>> > > > Speed: 1000, full duplex
>> > > > BOOTP broadcast 1
>> > > > DHCP client bound to address 192.168.11.151 (8 ms)
>> > > > Using ethernet at ff3f0000 device
>> > > > TFTP from server 192.168.11.1; our IP address is 192.168.11.151
>> > > > Filename '/boot/extlinux/pxelinux.cfg/default'.
>> > > > Not retrying...
>> > > > ================================================================
>> > > >
>> > > > Signed-off-by: Artem Lapkin <art at khadas.com>
>> > > > ---
>> > > >  include/config_distro_bootcmd.h | 2 ++
>> > > >  1 file changed, 2 insertions(+)
>> > > >
>> > > > diff --git a/include/config_distro_bootcmd.h b/include/config_distro_bootcmd.h
>> > > > index 3f724aa10f..db3d1b2362 100644
>> > > > --- a/include/config_distro_bootcmd.h
>> > > > +++ b/include/config_distro_bootcmd.h
>> > > > @@ -445,7 +445,9 @@
>> > > >                               "${devnum}:${distro_bootpart} "           \
>> > > >                               "${prefix}${boot_syslinux_conf}; then "   \
>> > > >                       "echo Found ${prefix}${boot_syslinux_conf}; "     \
>> > > > +                     "bootfile_bak=${bootfile}; "                      \
>> > > >                       "run boot_extlinux; "                             \
>> > > > +                     "setenv bootfile ${bootfile_bak}; "               \
>> > > >                       "echo SCRIPT FAILED: continuing...; "             \
>> > > >               "fi\0"                                                    \
>> > > >       \
>> > >
>> > > We've had this kind of problem before, and the answer is that variables
>> > > should be local, not global in scope.  In this case, I see that the way
>> > > the pxe/sysboot code works is that we have to env_set("..") in one place
>> > > to env_get("..") in another, so I don't see a way around this.
>> > >
>> > > Reviewed-by: Tom Rini <trini at konsulko.com>
>> >
>> > IMO a better approach will be the bootflow implementation. I hope to
>> > get v2 out early next week.
>>
>> I'm not sure if the bootflow way of going here would or would not have
>> the same problem, or perhaps a slightly different problem.  At heart
>> here, "sysboot" calls env_set(...) and then calls in to the pxe code
>> which does env_get(...).  So now I wonder how this would be fixed in
>> bootflow, since we aren't dealing with the environment directly.
>>
>> --
>> Tom