On 2015-10-30 13:54, Toke Høiland-Jørgensen wrote: > Felix Fietkau <[email protected]> writes: > >> How is this more useful than simply checking a hash of the kernel >> module in /lib/modules and comparing that with the corresponding file >> on the host? > > Because the value in /sys/module/ is from the *loaded* module. So it > catches the case where the module file is updated but not reloaded (was > bitten by this a couple of times). > > Also, you can put the module source git hash as the build id, so it can > be referred back to the source, rather than the compiled file. I'm still not convinced that this is very useful - if you have issues that you sometimes reinstall modules, but don't reload them and have to check the id, why not just fix your workflow instead? You could just make a simple wrapper script that reinstalls and reloads your module and complains if reloading fails.
When I do wifi driver development, I have to reload a chain of modules frequently. Since this is tedious and error prone, I have a script that simply pulls the latest package from my laptop via HTTP, unloads all relevant modules, then reloads them. Either way, if this is really necessary for you, I'm okay with adding support for keeping the build-id, but setting this in the package makefile (especially in the Build/Exports section) seems rather quirky to me. Since this is a rather exotic debugging-only feature, why not just add a global config option for it? - Felix _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
