Petr Štetiar <yn...@true.cz> writes: > Indeed, it looks like a regression caused by upstream commit e5f31552674e > ("ethernet: fix PTP_1588_CLOCK dependencies") in conjuction with images > produced by our buildbots, which use `CONFIG_ALL_KMODS=y` config setting, so > it likely makes `PTP_1588_CLOCK=m` and thus `CONFIG_NET_DSA_MV88E6XXX=m`. > > This simple revert/workaround seems to fix it: > > diff --git a/drivers/net/dsa/mv88e6xxx/Kconfig > b/drivers/net/dsa/mv88e6xxx/Kconfig > index 7a2445a34eb7..634a48e6616b 100644 > --- a/drivers/net/dsa/mv88e6xxx/Kconfig > +++ b/drivers/net/dsa/mv88e6xxx/Kconfig > @@ -2,7 +2,6 @@ > config NET_DSA_MV88E6XXX > tristate "Marvell 88E6xxx Ethernet switch fabric support" > depends on NET_DSA > - depends on PTP_1588_CLOCK_OPTIONAL > select IRQ_DOMAIN > select NET_DSA_TAG_EDSA > select NET_DSA_TAG_DSA
That look like it, yes. But the problem is much more generic, isn't it? I don't see how CONFIG_ALL_KMODS can ever be safe. AFAIK, there are lots of tristate symbols depending on a pattern like config FOO tristate "Build foo" depends on BAR || BAR=n where turning BAR from n to m might change FOO from y to m. Won't it? Bjørn _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel