PARTUUID isn't the same as UUID, as it's the GPT identifier for the partition rather than the identifier for the filesystem. Therefore, it's unfair to declare these two as synonymous. But, since libblkid can't resolve a PARTUUID back to a device, we remove this support for the time being. This reverts commit e02ceaa.
The kernel supports this, but blkid doesn't really. A little bit of refactoring, and everybody's happy.
derp, the kernel does this for us.
this is, in theory, faster for some noodle noggin who spends a lot of time loading modules in early userspace because they're hashed versions of the human readable files. they also don't compress as well and they're larger.
* 'master' of github.com:falconindy/geninit: init: remove support for disablemodules on the cmdline api: ignore EEXIST return in add_module geninit: show config file in use with ((verbose)) allow exec on /run init: use /run/modprobe.d instead /etc/modprobe.d init: limit /dev to 1024k error on module not found in API builders/usb: trim down module adding lsinitramfs: sort and columnize module output api: add basic method for adding mount hooks init: draw mount handlers from /mount. avoid statvfs. initial support for custom mount handlers api: fix src::dest file adding when not a binary
module-init-tools 3.13 supports modprobe.blacklist=$mod, so this should be used instead.
this is part of a slightly larger issue. EEXIST isn't necessarily bad -- it just means we're overlapping.
Since /tmp and /dev/shm are world writeable, it doesn't make sense to artifically inhibit exec from /run. Also, explicitly list mount flags for clarity.
we also switch to using 'blacklist' instead of hard blacklisting via install /bin/false since we're now expecting that this file be read after leaving early userspace.
trims off a small amount of execution time, woo.
Taken from module-init-tools, a '.' won't be found in a module name, so we can simplify how we string trim file/module names.