Hi there, I went on a rather "foreign" excursion with an owrt-based build here (I don't need to elaborate on specifics right? ;), and then kept having issues with udev hotplug auto loading of modules failure on this box [did I manage to hit all search keywords?]. Simple example: MCS7830 usbnet card, which should result in having the mcs7830.ko kernel module *auto-loaded* by udev as soon as plugged into the not-quite-OpenWrt router. Not so. lsusb succeeded, yet module functional only via *manual* loading.
Finally it dawned upon me that a modinfo mcs7830.ko failed to list the entire content: # modinfo mcs7830.ko filename: /lib/modules/3.10.24/mcs7830.ko license: GPL description: USB to network adapter MCS7830) alias: usb:v0DF6p0021d*dc*dsc*dp*ic*isc*ip*in* alias: usb:v9710p7730d*dc*dsc*dp*ic*isc*ip*in* alias: usb:v9710p7830d*dc*dsc*dp*ic*isc*ip*in* alias: usb:v9710p7832d*dc*dsc*dp*ic*isc*ip*in* depends: usbnet intree: Y vermagic: 3.10.24 mod_unload MIPS32_R2 32BIT (note the alias lines which are available here now yet which were getting removed with CONFIG_MODULE_STRIPPED active) As a result kmod's depmod -avw (or some such) executed on the target device seemed to be unable to compile its ID-based lookup table --> auto-load of modules not possible, due to missing ID mapping, as long as CONFIG_MODULE_STRIPPED remained active. Thus, while OpenWrt itself perhaps does not have much use for udev-based auto loading of modules (I guess modules usually are installed per-package, and then probably also registered for mechanical loading each), I'd deem the CONFIG_MODULE_STRIPPED setting to be (while not actively DANGEROUS) at least rather harmful. In light of this *default-enabled* functionality-breaking result the Kconfig descriptions of CONFIG_MODULE_STRIPPED: "Reduce module size" "Remove module parameter descriptions, author info, version, aliases, device tables, etc." as provided by e.g. ./target/linux/generic/patches-3.12/204-module_strip.patch seem to be rather very toned down. It would thus be nice to have this situation improved, perhaps by improving descriptions and possibly altering the default setting. Put differently, this default-enabled "hack" (quote original patch ;) results in a gross departure from functionality expected to be standard by upstream, which in itself is worrisome already. Also, what kind of space savings are we talking about here anyway? I cannot imagine these measly strings to constitute a sufficiently high percentage of module size... (except for drastic candidates such as VID/PID of pl2303 or ftdi_sio driver) Thanks, Andreas Mohr _______________________________________________ openwrt-devel mailing list [email protected] https://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
