[U-Boot-Users] [PATCH] new CONFIG_RETURN_FROM_BOOTM

[Jerry Van Baren <gerald.vanbaren@smiths-aerospace.com>]

Wilson Callan wrote:
> Hi,
> 
> I've added an option similar to the  netretry=no environment option 
> which allows a script to control the retry operation for networking.  
> This option controls a retry operation for bootm.  Here is the README 
> entry:
> 
> +- CONFIG_RETURN_FROM_BOOTM
> +
> +        Return from the bootm command instead of rebooting.  This
> +        is useful for setting up multiple boot image options in a
> +        script.  It allows the script to control the retry
> +        operation.  For example, these scripts setup a system that
> +        will attempt a primary flash image, then a backup image
> +        location, then resort to TFTP:
> +          "boot_tftp=dhcp 3c0;bootm 3c0;reset\0" /* if fails, reset */ \
> +          "boot_flash=bootm ff050000 3c0;bootm ffc00000 3c0;run 
> boot_tftp;reset\0"
> 
> 
> CHANGELOG:
> 
> Added CONFIG_RETURN_FROM_BOOTM to control retry operation of bootm command.
> 
> Signed-off-by: Wilson Callan <wcallan@savantav.com>

NAK

This is a *BAD* idea as is.  See the comment in cmd_bootm.c line 298:
<http://www.denx.de/cgi-bin/gitweb.cgi?p=u-boot.git;a=blob;f=common/cmd_bootm.c;h=ba286f1e636c1f6d36ea41b0366b6e192ef03d0f;hb=HEAD#l298>
  298         /*
  299          * We have reached the point of no return: we are going to
  300          * overwrite all exception vector code, so we cannot easily
  301          * recover from any failures any more...
  302          */

If you return, one of four things will happen:
a) U-boot will continue to work.
b) It may work by pure luck.
c) It may work for a while and then go goofy and crash.
d) It may simply crash.
Unless you can guarantee (a), this is not an acceptable patch.  Note 
that the current code does *NOT* guarantee (a).  My experience is that 
it is (c) or (b), in that order (DAMHIKT).

Trivia: I have some pending reorganization and other improvements in my 
u-boot-fdt repo that fix some erroneous returns.  One Of These Days they 
will be pulled into the mainline repo.
<http://www.denx.de/cgi-bin/gitweb.cgi?p=u-boot/u-boot-fdt.git;a=blob;f=common/cmd_bootm.c;h=ba286f1e636c1f6d36ea41b0366b6e192ef03d0f;hb=HEAD>

One thought is to copy the interrupt vectors to a set-aside area so 
that, if something fails after line 298, you can restore the interrupt 
vectors and _then_ return.  Note that you have to guarantee your u-boot 
(static) variables are not overwritten by the aborted bootm command as 
well.  This /shouldn't/ be an issue, but it could be since stupid^W 
ignorant users can load in poorly chosen memory locations.

Good luck,
gvb