still wandering through the bowels of u-boot environment manipulation code, and i see this in common/env_common.c:

/* * Look up the variable from the default environment */ char *getenv_default(const char *name) { char *ret_val; unsigned long really_valid = gd->env_valid; unsigned long real_gd_flags = gd->flags;

/* Pretend that the image is bad. */ gd->flags &= ~GD_FLG_ENV_READY; gd->env_valid = 0; <--- ????? ret_val = getenv(name); gd->env_valid = really_valid; gd->flags = real_gd_flags; return ret_val; }

fair enough ... in order to read a variable explicitly from the (default) environment, temporarily fake that there is nothing in the hash table, so this statement makes sense:

gd->flags &= ~GD_FLG_ENV_READY;

but is this one immediately afterwards also necessary?

gd->env_valid = 0;

because once you get to getenv() as defined in cmd/nvedit.c:

char *getenv(const char *name) { if (gd->flags & GD_FLG_ENV_READY) { /* after import into hashtable */ ENTRY e, *ep;

WATCHDOG_RESET();

e.key = name; e.data = NULL; hsearch_r(e, FIND, &ep, &env_htab, 0);

return ep ? ep->data : NULL; }

/* restricted capabilities before import */ if (getenv_f(name, (char *)(gd->env_buf), sizeof(gd->env_buf)) > 0) return (char *)(gd->env_buf);

return NULL; }

it seems that the only test is for GD_ENV_FLG_READY, so setting gd->env_valid=0 seems unnecessary. but perhaps i'm not reading far enough.

rday

p.s. just to be clear, the meanings of those two gd members is:

gd->flags & GD_FLG_ENV_READY means the environment has been imported into a hash table, and

gd->env_valid means the CRC for the environment in persistent storage is correct.

do i understand that correctly?