[olug] Drive letter assignments
Christopher Cashell
topher-olug at zyp.org
Thu Apr 14 02:23:02 UTC 2011
On Wed, Apr 13, 2011 at 8:13 PM, Kevin D. Snodgrass
<kdsnodgrass at yahoo.com> wrote:
> No, I don't think so. I consider this to >NOT< be a bug, just the way things are with the new udevd plus parallel driver initializaion. I understand the usefulness of udev, sort of. (Why are people worried about /dev being populated with thousands of entries? They don't take up much space, just entries in the /dev directory...)
The cleaning up of /dev is actually more of a bonus item from udev,
rather than the goal. Years ago, hardware was static. Every device
that was going to be connected to and managed by a computer was
connected prior to starting it up. The easy way for Linux to handle
that was to just load up driver modules, have them interrogate the
system to see if matching hardware was connected, configure the
hardware, then move on to the next one. When you finished processing
all the hardware, you were done, and everything was static from that
point until the next reboot.
Then along came Hot Swap had drives. And PCMCIA. And USB. And
Firewire. And so on.
All of a sudden, the hardware in a machine couldn't be relied on to
stay static. Hardware could be added or removed at any time. A
couple of different subsystems were attempted to handle this until
udev came along and offered a level of flexibility, configurability,
and capability that could handle modern systems.
Now, instead of just doing a static scan of hardware on bootup, the
kernel sends an event to udev every time a piece of hardware is added
or removed from the system, or changes its state. udev then takes
actions and makes configurations based on its rules and the udev
database.
Oh yeah, and it cleans up all the crap in /dev, so entries there are
actually meaningful and only exist when a device matches it (reducing
confusion and complexity when users try to deal with hardware from the
OS). ;-)
> Kevin D. Snodgrass
--
Christopher
More information about the OLUG
mailing list