Maxime, York,

I've just sent a patch that removes 'get_char' callback from the env drivers. This should restore the old behaviour we had before supporting multiple environment drivers (for all but eeprom, of course).


On 08.02.2018 23:10, Maxime Ripard wrote:
On Thu, Feb 08, 2018 at 10:52:20AM +0100, Simon Goldschmidt wrote:
On 08.02.2018 09:47, Maxime Ripard wrote:
On Wed, Feb 07, 2018 at 02:17:12PM -0800, York Sun wrote:
Commit 8a3a7e2270b3 ("env: Pass additional parameters to the env
lookup function") dropped the default action if driver doesn't have
get_char() defined. This causes failure to get environmental
variables from NOR flash. Add back this default action for now.

Signed-off-by: York Sun <>
CC: Maxime Ripard <>
Limited test on LS1043ARDB.

   env/env.c | 2 +-
   1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/env/env.c b/env/env.c
index edfb575..210bae2 100644
--- a/env/env.c
+++ b/env/env.c
@@ -159,7 +159,7 @@ int env_get_char(int index)
                int ret;
                if (!drv->get_char)
-                       continue;
+                       return *(uchar *)(gd->env_addr + index);
Thinking more about this, I think this would break the case where the
first environment in your list has no get_char method, but the second

How can we decide which way is wanted? With your patch below, we might end
up loading chars from a low-prio environment (which is not CRC checked) in
the early boot stage. Later, we load the environment from another env driver
with higher priority.
Ah, right.

That's why I suggested removing the 'get_char' callback from the env drivers
:-) Early boot stage environment lookups would still work the old way
reading from 'gd->env_addr + index'.
If that works on York's board, I'm all in.


U-Boot mailing list

Reply via email to