[RFC PATCH v4 28/28] board: keymile: common: Use environment to store IVM_* variables.

[Francis Laniel]

Hi.

Le lundi 20 juin 2022, 19:33:24 CEST Tom Rini a écrit :
> On Mon, Jun 20, 2022 at 04:08:32PM +0000, Holger Brunck wrote:
> > > > > On Fri, Jun 17, 2022 at 12:31:58AM +0200, Francis Laniel wrote:
> > > > > > These boards used set_local_var() to store some variables as local
> > > > > > shell.
> > > > > > They then used get_local_var() to retrieve the variables values.
> > > > > > 
> > > > > > Instead of using local shell variables, they should use
> > > > > > environment ones (like a majority of board).
> > > > > > So, this patch converts using local variables to environment ones.
> > > > 
> > > > why do we need to change that? It is intended that we use this hush
> > > > variable infrastructure from u-boot (common/hush.c) for our IVM data
> > > > and not the standard env. We read the IVM at boot time and store these
> > > > values in RAM. It is not intended to store them permanently in the
> > > > flash or wherever the environment is saved. Especially in our case we
> > > > have some boards where the environment is in a i2c EEPROM and we don't
> > > > want to write down to the EEPROM each time when the board is starting.
> > > 
> > > So, the whole series is about updating hush to bring in a new baseline
> > > version of the shell, from busybox.
> > 
> > Ok.
> > 
> > > > > > Signed-off-by: Francis Laniel
> > > > > > <francis.laniel at amarulasolutions.com>
> > > > > > ---
> > > > > > 
> > > > > >  board/keymile/common/common.c | 8 ++++----
> > > > > >  board/keymile/common/ivm.c    | 9 +--------
> > > > > >  2 files changed, 5 insertions(+), 12 deletions(-)
> > > > > > 
> > > > > > diff --git a/board/keymile/common/common.c
> > > > > > b/board/keymile/common/common.c index 3999f48719..72939af36e
> > > > > > 100644
> > > > > > --- a/board/keymile/common/common.c
> > > > > > +++ b/board/keymile/common/common.c
> > > > > > @@ -219,7 +219,7 @@ static int do_setboardid(struct cmd_tbl
> > > > > > *cmdtp, int
> > > > > 
> > > > > flag, int argc,
> > > > > 
> > > > > >  	unsigned char buf[32];
> > > > > >  	char *p;
> > > > > > 
> > > > > > -	p = get_local_var("IVM_BoardId");
> > > > > > +	p = env_get("IVM_BoardId");
> > > > > > 
> > > > > >  	if (!p) {
> > > > > >  	
> > > > > >  		printf("can't get the IVM_Boardid\n");
> > > > > >  		return 1;
> > > > > > 
> > > > > > @@ -228,7 +228,7 @@ static int do_setboardid(struct cmd_tbl
> > > > > > *cmdtp, int
> > > > > 
> > > > > flag, int argc,
> > > > > 
> > > > > >  	env_set("boardid", (char *)buf);
> > > > > >  	printf("set boardid=%s\n", buf);
> > > > > > 
> > > > > > -	p = get_local_var("IVM_HWKey");
> > > > > > +	p = env_get("IVM_HWKey");
> > > > > > 
> > > > > >  	if (!p) {
> > > > > >  	
> > > > > >  		printf("can't get the IVM_HWKey\n");
> > > > > >  		return 1;
> > > > > > 
> > > > > > @@ -272,14 +272,14 @@ static int do_checkboardidhwk(struct cmd_tbl
> > > > > 
> > > > > *cmdtp, int flag, int argc,
> > > > > 
> > > > > >  	 * first read out the real inventory values, these values are
> > > > > >  	 * already stored in the local hush variables
> > > > > >  	 */
> > > > > > 
> > > > > > -	p = get_local_var("IVM_BoardId");
> > > > > > +	p = env_get("IVM_BoardId");
> > > > > > 
> > > > > >  	if (!p) {
> > > > > >  	
> > > > > >  		printf("can't get the IVM_Boardid\n");
> > > > > >  		return 1;
> > > > > >  	
> > > > > >  	}
> > > > > >  	rc = strict_strtoul(p, 16, &ivmbid);
> > > > > > 
> > > > > > -	p = get_local_var("IVM_HWKey");
> > > > > > +	p = env_get("IVM_HWKey");
> > > > > > 
> > > > > >  	if (!p) {
> > > > > >  	
> > > > > >  		printf("can't get the IVM_HWKey\n");
> > > > > >  		return 1;
> > > > > > 
> > > > > > diff --git a/board/keymile/common/ivm.c
> > > > > > b/board/keymile/common/ivm.c index 67db0c50f4..e266d7ce81 100644
> > > > > > --- a/board/keymile/common/ivm.c
> > > > > > +++ b/board/keymile/common/ivm.c
> > > > > > @@ -44,14 +44,7 @@ static int ivm_calc_crc(unsigned char *buf, int
> > > > > > len)
> > > > > > 
> > > > > >  static int ivm_set_value(char *name, char *value)  {
> > > > > > 
> > > > > > -	char tempbuf[256];
> > > > > > -
> > > > > > -	if (value) {
> > > > > > -		sprintf(tempbuf, "%s=%s", name, value);
> > > > > > -		return set_local_var(tempbuf, 0);
> > > > > > -	}
> > > > > > -	unset_local_var(name);
> > > > > > -	return 0;
> > > > > > +	return env_set(name, value);
> > > > 
> > > > this means we are now writing always down to the permanent environment
> > > > or? And this I would really like to avoid in our case.
> > > 
> > > Note that "env_set" does not force a save of the running environment.
> > 
> > Ah yes you are right. But still for the first boot of our board we will
> > call saveenv to store an initial environment and with this change it
> > would also write down the IVM data which is currently only stored
> > temporary in RAM.
> > 
> > > These variables will be exposed to the CLI run-time, which I am not sure
> > > if they are today, so if the user then does "saveenv" they will be
> > > written.  That I think would be a functional difference.
> > 
> > yes exactly and I wonder if this functionality will be also possible after
> > the rework. I mean our use case (even it seems we are the only ones using
> > it) is quite useful I think. We read out inventory data from an EEPROM
> > and they are parsed and temporary stored in RAM and then we are able to
> > use them as any other environment variables without the need to store
> > them permanently. I also would like to avoid this as the data should be
> > permanently in the IVM only and not stored a second time permanently  in
> > flash.
> 
> So, I'll start by noting that the environment variable flag support
> isn't as well documented as I would like.  But my super quick glance
> makes me wonder if ENV_FLAGS_VARACCESS_WRITEABLE does what you're
> thinking about, and if not, if it would be useful to extend the flag
> support to include a "hidden" flag, or otherwise way to indicate that
> variables A/B/C should never be saved.  This could be useful for other
> cases as well.

Again, sorry for the late reply.

To be honest, this board is the only one which presents this behavior.
I will think about a mechanism to handle your use case.
Nonetheless, as you are the exception and this series begins to be a bit big I 
think in a first time I will not address it.
I will just add these patches to make the CI happy, but I will mark them as 
"do not merge".
So, we can merge this series and you can still use the old hush shell, as I do 
not think we plan about making the new version the default right now.

What do you think?
Does it seem acceptable for you?