As a printer is not a necessary-for-boot service, I don't intend to add any code
to initscripts for it (ideally, there shouldn't be *any* code in initscripts for
USB, it should all be hotplug).

Did this work in 7.1?

> As a printer is not a necessary-for-boot service, I don't intend to add any code
> to initscripts for it

May be I didn't state my problem clear enough - if the printer is plugged in and
turned on when the machine is rebooted, the printer will not be recognized at
any point during or *after* the boot process. This means that I have to do
things *manually* if I want to be able to use my printer.

> Did this work in 7.1?

As it turned out, between the time I got the printer and the time I started
upgrading to RC1, I never rebooted, so I have no idea...

P.S. I do not think this is a hotplug issue. Hotplug does everything correctly
if I plug printer after the machine is booted (see the 3rd log from my initial
report).

The question I would like to ask - at which point should I expect the printer
driver to be loaded if I am booting the system with printer plugged in and
turned on?

One possible answer - it should happen when USB core modules are first loaded
(as shown in my 2nd log in initial report). But that would mean that hotplug
should be ready to go (all sysctls, etc should be done) before usb modules are
loaded, which is not the case for rc.sysinit in current initscripts.
Are there any better answers to this question?

hotplug is supposed to handle the cold-plug cases too; that's why it's a hotplug
problem. The code is in initscripts *only* for the things you'd need to use
the console if something bad happens (fsck failure, etc.)

Please try hotplug-2001_04_24-11 from http://people.redhat.com/teg/

Reported fixed in that version.

Unfortunatelly, this problem still exists in Enigma (which has
hotplug-2001_04_24-11).

Boot messages (grep -i usb /var/log/messages):

Oct  2 19:40:04 home rc.sysinit: Mounting USB filesystem:  succeeded
Oct  2 19:40:04 home rc.sysinit: Initializing USB controller (usb-uhci):  succeeded
Oct  2 19:44:20 home kernel: usb.c: registered new driver usbdevfs
Oct  2 19:44:20 home kernel: usb.c: registered new driver hub
Oct  2 19:44:20 home kernel: usb-uhci.c: $Revision: 1.259 $ time 17:18:11 Sep  6
2001
Oct  2 19:44:20 home kernel: usb-uhci.c: High bandwidth mode enabled
Oct  2 19:44:20 home kernel: usb-uhci.c: USB UHCI at I/O 0xff80, IRQ 11
Oct  2 19:44:21 home kernel: usb-uhci.c: Detected 2 ports
Oct  2 19:44:21 home kernel: usb.c: new USB bus registered, assigned bus number 1
Oct  2 19:44:21 home kernel: hub.c: USB hub found
Oct  2 19:44:21 home kernel: usb-uhci.c: v1.251:USB Universal Host Controller
Interface driver
Oct  2 19:44:21 home kernel: hub.c: USB new device connect on bus1/2, assigned
device number 2
Oct  2 19:44:21 home kernel: usb.c: USB device 2 (vend/prod 0x3f0/0x317) is not
claimed by any active driver.

After manual "modprobe printer":
Oct  3 12:08:30 home kernel: usb.c: registered new driver usblp
Oct  3 12:08:30 home kernel: printer.c: usblp0: USB Bidirectional printer dev 2
if 0 alt 1
Oct  3 12:08:30 home kernel: printer.c: v0.8:USB Printer Device Class driver

Where and how is cold-plugging supposed to be initiated? According to the
hotplug home page, it is supposed to be done by /etc/rc.d/init.d/hotplug;
hotplug's man page also talks about /etc/init.d/hotplug, but the hotplug RPM
does not include any init.d sripts!

Initialization of the USB-controller is done in the rc.sysinit scripts.

You are missing the point!
Yes, initialization of USB *controller* is done in rc.sysinit. However, when the
USB *controller* is initialized it does not yet know about the USB *devices*, so
if a device is plugged in before the controller is initialized, the device
driver is not loaded.

In a normal hotplug operation the cold plugging (e.g. recognition of the devices
that are already plugged in) of USB devices is performed by
/etc/hotplug/usb.rc start
which, in turn, is supposed to be called from
/etc/init.d/hotplug startUnfortunatelly, /etc/init.d/hotplug is *missing* from
the hotplug package!

The first paragraph of the "How do these Hotplug Tools work?" section of the
HotPlug Overview - http://linux-hotplug.sourceforge.net/?selected=overview states:

 There is a "hotplug" system service (/etc/rc.d/init.d/hotplug) that's invoked
during system booting, which handles initialization including "coldplugging" by
invoking /etc/hotplug/type.rc files for each type the system supports. These
types include usb, pci for PCI and Cardbus, net for networking, and so on; these
types are defined by the Linux kernel.

I do believe the documentation is wrong. /etc/init.d/hotplug should not be
necessary. As your messages note, it recognized the printer fine; why it
didn't find the right driver is a separate problem.

USB devices are detected/hotplugged when the USB controller is initialized -
which is what the hotplug init script does. We need USB keyboards active before
this, and thus it's initialized in rc.sysinit. Why your device isn't detected
properly I'm not sure, but as you can see from the logs it's trying.

I disagree. I do not think it's really trying. If you think the documentation is
wrong, please tell me *exactly* at what point during the boot process and *using
which mechanism* is the cold-plugging supposed to work. 

What you've said so far seems to imply the answer "when the USB controller is
initialized in rc.sysinit" and "using the hotplug" which can not be true since
the hotplug sysctl goes *after* USB initialization in rc.sysinit, so hoplug
subsystem is not active when the USB controller is initialized.

teg> USB devices are detected/hotplugged when the USB controller is initialized -
teg> which is what the hotplug init script does.

But how can hotplug init script do anything when *it's missing from the RPM*?

See also bug 55332 for another instance of this problem.

Because the kernel calls /sbin/hotplug (or whatever you tell it in
/proc/sys/kernel/hotplug )

teg> Because the kernel calls /sbin/hotplug (or whatever you tell it in
teg> /proc/sys/kernel/hotplug )

But does not this mean that we are coming back to this being an rc.sysinit
problem? Currently rc.sysinit loads usb core modules *before* it does "sysctl -w
kernel.hotplug", so how could this work?

Again, this comes to the same question I haven't seen any answer to - at what
exact point in the bootup process do you want coldplugging to happen? Before
that question is answered, it's impossible to say what script is responsible for
making sure this really happens and so far you seem to be just bouncing this
back and forth, essentially saying "it's not my script that is responsible for
that, it somebody else's"...

Cold plugging should work as hotplug events are generated by the kernel when the
USB controller is initialized; hence, it goes through the normal hotplug
mechanisms, hence, no init script is needed. The bug is elsewhere.

> Cold plugging should work as hotplug events are generated by the kernel when the
> USB controller is initialized; hence, it goes through the normal hotplug
> mechanisms, 

How would it work if hotplug sysctl goes *after* USB initialization is sysinit?
(We are going in circles! I am essentially repeating what I've already said in
my comments from 2001-10-28 10:41:00).

Also, I remember hearing about some problems with the hotplug scripts not doing
the right thing when /usr is not mounted - could that be a problem?

Because before the sysctl happens, the value of kernel.hotplug is the default,
'/sbin/hotplug'. (Yes, the sysctl is somewhat redundant.)

It works on the cases I've tested... including usb-storage, webcams etc.