On 05/16/2017 10:58 PM, Michael Ellerman wrote: > Rob Landley <r...@landley.net> writes: > >> diff --git a/init/main.c b/init/main.c >> index f866510..9ec09ff 100644 >> --- a/init/main.c >> +++ b/init/main.c >> @@ -1055,8 +1049,17 @@ static noinline void __init kernel_init_freeable(void) >> if (sys_access((const char __user *) ramdisk_execute_command, 0) != 0) { >> ramdisk_execute_command = NULL; >> prepare_namespace(); >> + } else if (IS_ENABLED(CONFIG_DEVTMPFS_MOUNT)) { >> + sys_mkdir("/dev", 0755); >> + devtmpfs_mount("/dev"); >> } >> >> + /* Open the /dev/console on the rootfs, this should never fail */ >> + if (sys_open((const char __user *) "/dev/console", O_RDWR, 0) < 0) > > Sorry to pile on,
The correct phrase is "bikeshed". (It's a verb now.) > but while you're moving it do you want _I_ don't, no. I intentionally moved it unmodified. If you want to submit a patch on top of mine, be my guest. > to update this fairly misleading comment. Define "should". (I'll get the popcorn.) > It definitely can fail, eg. if /dev/console doesn't exist, or if no > console driver is registered. /dev/console not existing in an initramfs created by pointing CONFIG_INITRAMFS_SOURCE at a directory created by a normal user was pretty much my initial motivation for poking at this area, yes. That said, /dev/console should always exist. My patch was just finding a different way for it to exist so the condition was satisfied. Meaning the comment isn't exactly _wrong_, just really terse. Feel free to submit a patch rephrasing it. Rob