> I didn't quite understand what you are hoping to achieve with this patch > set, but if the command would serve the purpose, then that would be
Okay, I will come back with something soon. Thank's for support Le lun. 3 nov. 2025 à 09:22, Ahmad Fatoum <[email protected]> a écrit : > > Hi Anis, > > On 11/3/25 3:09 PM, anis chali wrote: > >>> strlen can't return negative numbers. > > ack, right it is size_t (unsigned) > > > >>> The original function had a number of safety checks that are missing here: > >>> > >>> - It only ran the code when in EFI payload mode, as the same barebox > >>> binary could be chainloaded without EFI > >>> > >>> - It checked that /boot is mounted and gave a useful error message to hint > >>> at e.g., EFIFS missing, which we are missing out on here. > >>> > >>> I am not yet convinced this is strictly an improvement. > > > > Yes you are right but I didn't reintegrate the code in generic driver > > since it will just use Kconfigs to reference a state dtb file. > > In case this code was accepted there is no context to do the check, > > but I ensured that barebox gives a behaves in way that the user > > will quickly see that there is no /boot directory with the message > > telling No such file or directory, anyway you suggestion to add -f to > > state command > > is more elegant, I will do some experimentations. the aim of this > > patch is to support a state.dtb given in runtime which permits > > userspace interaction with > > native board because I didn't figure out how barebox state command > > could communicate with a barebox intree state. > > I didn't quite understand what you are hoping to achieve with this patch > set, but if the command would serve the purpose, then that would be > best, I think. > > Cheers, > Ahmad > > > > >>> If you have the need to customize where your state comes from, how about > >>> adding > >>> a state -f /path/to/state.dtb command option and calling it from your > >>> init script? > > right, more elegant suggestion, I will try it. > > > > Thank's for feedback > > > > Anis. > > > > Le lun. 3 nov. 2025 à 01:46, Ahmad Fatoum <[email protected]> a écrit > > : > >> > >> Hi, > >> > >> On 03.11.25 05:38, [email protected] wrote: > >>> From: Chali Anis <[email protected]> > >>> > >>> Add a driver to use an external state dtb, gives the ability to define > >>> an external state at compile time. useful for yocto or buildroot defining > >>> a state.dtb that will be passed to barebox at compile time vi a defconfig > >>> fragment. > >> > >> We already had code that did this unconditionally for EFI. > >> Can you explain why this needs to be customizable? > >> > >>> +config EXTERNAL_STATE > >>> + tristate "Use external barebox state dtb" > >>> + depends on OFDEVICE > >>> + depends on STATE > >>> + help > >>> + This permits the use of an extranl dtb state blob > >> > >> external > >> > >>> + which permits to dynamicly at compile time specify > >> > >> dynamically > >> > >>> + an external blob vi EXTERNAL_STATE_DTB_PATH > >>> + > >>> +config EXTERNAL_STATE_DTB_PATH > >>> + string "the external barebox state dtb path" > >>> + depends on EXTERNAL_STATE > >> > >> Really needs help text as EXTERNAL makes me think of a file on the > >> build host. > >> > >>> +static int state_external_init(void) > >>> +{ > >>> + const char *dt_path = CONFIG_EXTERNAL_STATE_DTB_PATH; > >>> + struct device_node *state_root = NULL; > >>> + size_t size; > >>> + void *fdt; > >>> + int ret; > >>> + > >>> + if (strlen(dt_path) <= 0) > >>> + return -EINVAL; > >> > >> strlen can't return negative numbers. > >> > >> The original function had a number of safety checks that are missing here: > >> > >> - It only ran the code when in EFI payload mode, as the same barebox > >> binary could be chainloaded without EFI > >> > >> - It checked that /boot is mounted and gave a useful error message to hint > >> at e.g., EFIFS missing, which we are missing out on here. > >> > >> > >> I am not yet convinced this is strictly an improvement. > >> > >> If you have the need to customize where your state comes from, how about > >> adding > >> a state -f /path/to/state.dtb command option and calling it from your init > >> script? > >> > >> Cheers, > >> Ahmad > >> > >>> + > >>> + fdt = read_file(dt_path, &size); > >>> + if (!fdt) { > >>> + pr_info("unable to read %s: %m\n", dt_path); > >>> + return 0; > >>> + } > >>> + > >>> + state_root = of_unflatten_dtb(fdt, size); > >>> + if (!IS_ERR(state_root)) { > >>> + struct device_node *np = NULL; > >>> + struct state *state; > >>> + > >>> + ret = barebox_register_of(state_root); > >>> + if (ret) > >>> + pr_warn("Failed to register device-tree: %pe\n", > >>> ERR_PTR(ret)); > >>> + > >>> + np = of_find_node_by_alias(state_root, "state"); > >>> + > >>> + state = state_new_from_node(np, false); > >>> + if (IS_ERR(state)) > >>> + return PTR_ERR(state); > >>> + > >>> + ret = state_load(state); > >>> + if (ret != -ENOMEDIUM) > >>> + pr_warn("Failed to load persistent state, > >>> continuing with defaults, %d\n", > >>> + ret); > >>> + > >>> + return 0; > >>> + } > >>> + > >>> + return -EINVAL; > >>> +} > >>> + > >>> +late_initcall(state_external_init); > >> > >> > >> -- > >> Pengutronix e.K. | | > >> Steuerwalder Str. 21 | http://www.pengutronix.de/ | > >> 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | > >> Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | > > > > -- > Pengutronix e.K. | | > Steuerwalder Str. 21 | http://www.pengutronix.de/ | > 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 | > Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 | >
