linux.git/drivers/usb/core/hcd.c, branch v2.6.20

usb hub: fix root hub code so it takes more than 15 devices per root hub

2006-12-01T22:23:26+00:00

Wireless USB Host Controllers accept a large number of devices per
host, which shows up as a large number of ports in its root hub.

When the number of ports in a hub device goes over 16, the activation
of the hub fails with the cryptic message in klogd.

hub 2-0:1.0: activate --> -22

Following this further, it was seen that:

hub_probe()
  hub_configure()
    generates pipe number

    pseudo allocates buffer 'maxp' bytes in size using usb_maxpacket()

      The endpoint descriptor for a root hub interrupt endpoint is
      declared in
      drivers/usb/core/hcd.c:hs_rh_config_descriptor and declares it
      to be size two (supporting 15 devices max).

    hub_activate()
      usb_hcd_submit_urb()
        rh_urb_enqueue()
          urb->pipe is neither int nor ctl, so it errors out
            rh_queue_status()
              Returns -EINVAL because the buffer length is smaller
              than the minimum needed to report all the hub port
              bits as in accordance with USB2.0[11.12.3]. There has
              to be trunc((PORTS + 1 + 7) / 8) bytes of space at
              least.

Alan Stern confirmed that the reason for reading maxpktsize and not
the right amount is because some hubs are known to return more data
and thus cause overflow. 

So this patch simply changes the code to make the interrupt endpoint's
max packet size be at least the minimum required by USB_MAXCHILDREN
(instead of a fixed magic number) and add documentation for that. This
way we are always ahead of the limit.

Signed-off-by: Inaky Perez-Gonzalez 
Signed-off-by: Greg Kroah-Hartman

IRQ: Maintain regs pointer globally rather than passing to IRQ handlers

2006-10-05T14:10:12+00:00

Maintain a per-CPU global "struct pt_regs *" variable which can be used instead
of passing regs around manually through all ~1800 interrupt handlers in the
Linux kernel.

The regs pointer is used in few places, but it potentially costs both stack
space and code to pass it around.  On the FRV arch, removing the regs parameter
from all the genirq function results in a 20% speed up of the IRQ exit path
(ie: from leaving timer_interrupt() to leaving do_IRQ()).

Where appropriate, an arch may override the generic storage facility and do
something different with the variable.  On FRV, for instance, the address is
maintained in GR28 at all times inside the kernel as part of general exception
handling.

Having looked over the code, it appears that the parameter may be handed down
through up to twenty or so layers of functions.  Consider a USB character
device attached to a USB hub, attached to a USB controller that posts its
interrupts through a cascaded auxiliary interrupt controller.  A character
device driver may want to pass regs to the sysrq handler through the input
layer which adds another few layers of parameter passing.

I've build this code with allyesconfig for x86_64 and i386.  I've runtested the
main part of the code on FRV and i386, though I can't test most of the drivers.
I've also done partial conversion for powerpc and MIPS - these at least compile
with minimal configurations.

This will affect all archs.  Mostly the changes should be relatively easy.
Take do_IRQ(), store the regs pointer at the beginning, saving the old one:

	struct pt_regs *old_regs = set_irq_regs(regs);

And put the old one back at the end:

	set_irq_regs(old_regs);

Don't pass regs through to generic_handle_irq() or __do_IRQ().

In timer_interrupt(), this sort of change will be necessary:

	-	update_process_times(user_mode(regs));
	-	profile_tick(CPU_PROFILING, regs);
	+	update_process_times(user_mode(get_irq_regs()));
	+	profile_tick(CPU_PROFILING);

I'd like to move update_process_times()'s use of get_irq_regs() into itself,
except that i386, alone of the archs, uses something other than user_mode().

Some notes on the interrupt handling in the drivers:

 (*) input_dev() is now gone entirely.  The regs pointer is no longer stored in
     the input_dev struct.

 (*) finish_unlinks() in drivers/usb/host/ohci-q.c needs checking.  It does
     something different depending on whether it's been supplied with a regs
     pointer or not.

 (*) Various IRQ handler function pointers have been moved to type
     irq_handler_t.

Signed-Off-By: David Howells 
(cherry picked from 1b16e7ac850969f38b375e511e3fa2f474a33867 commit)

[PATCH] namespaces: utsname: use init_utsname when appropriate

2006-10-02T14:57:21+00:00

In some places, particularly drivers and __init code, the init utsns is the
appropriate one to use.  This patch replaces those with a the init_utsname
helper.

Changes: Removed several uses of init_utsname().  Hope I picked all the
	right ones in net/ipv4/ipconfig.c.  These are now changed to
	utsname() (the per-process namespace utsname) in the previous
	patch (2/7)

[akpm@osdl.org: CIFS fix]
Signed-off-by: Serge E. Hallyn 
Cc: Kirill Korotaev 
Cc: "Eric W. Biederman" 
Cc: Herbert Poetzl 
Cc: Andrey Savochkin 
Cc: Serge Hallyn 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

USB: allow both root-hub interrupts and polling

2006-09-28T22:36:45+00:00

Originally I didn't think any host controller driver would ever use
interrupts and polling at the same time, but it turns out ohci-hcd wants
to do exactly that.  This patch (as788) makes it possible.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

USB: Fix alignment of buffer passed down to ->hub_control()

2006-09-28T22:36:43+00:00

Implementations assume the buffer is at least 4 byte aligned.

Signed-off-by: David S. Miller 
Cc: David Brownell 
Signed-off-by: Greg Kroah-Hartman

usbcore: remove usb_suspend_root_hub

2006-09-27T18:58:57+00:00

This patch (as740) removes the existing support for autosuspend of
root hubs.  That support fit in rather awkwardly with the rest of
usbcore and it was used only by ohci-hcd.  It won't be needed any more
since the hub driver will take care of autosuspending all hubs, root
or external.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

usbcore: trim down usb_bus structure

2006-09-27T18:58:56+00:00

As part of the ongoing program to flatten out the HCD bus-glue layer,
this patch (as771b) eliminates the hcpriv, release, and kref fields
from struct usb_bus.  hcpriv and release were not being used for
anything worthwhile, and kref has been moved into the enclosing
usb_hcd structure.

Along with those changes, the patch gets rid of usb_bus_get and
usb_bus_put, replacing them with usb_get_hcd and usb_put_hcd.

The one interesting aspect is that the dev_set_drvdata call was
removed from usb_put_hcd, where it clearly doesn't belong.  This means
the driver private data won't get reset to NULL.  It shouldn't cause
any problems, since the private data is undefined when no driver is
bound.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

usbcore: Add flag for whether a host controller uses DMA

2006-09-27T18:58:56+00:00

This patch (as770b) introduces a new field to usb_bus: a flag
indicating whether or not the host controller uses DMA.  This serves
to encapsulate the computation.  It also means we will have only one
spot to update if the DMA API changes.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

USB: remove struct usb_operations

2006-09-27T18:58:56+00:00

All of the currently-supported USB host controller drivers use the HCD
bus-glue framework.  As part of the program for flattening out the glue
layer, this patch (as769) removes the usb_operations structure.  All
function calls now go directly to the HCD routines (slightly renamed
to remain within the "usb_" namespace).

The patch also removes usb_alloc_bus(), because it's not useful in the
HCD framework and it wasn't referenced anywhere.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman

usbcore: make hcd_endpoint_disable wait for queue to drain

2006-09-27T18:58:54+00:00

The inconsistent lock state problem in usbcore (the one that shows up
when an HCD is unloaded) comes down to two inter-related problems:

	usb_rh_urb_dequeue() isn't set up to be called with interrupts
	disabled.

	hcd_endpoint_disable() doesn't wait for all URBs on the
	endpoint's queue to complete.

The two problems are related because the one type of URB that isn't
likely to be complete when hcd_endpoint_disable() returns is a root-hub
URB.  Right now usb_rh_urb_dequeue() waits for them to complete, and it
assumes interrupts are enabled so it can wait.  But
hcd_endpoint_disable() calls it with interrupts disabled.

Now, it should be legal to unlink root-hub URBs with interrupts
disabled.  The solution is to move the waiting into
hcd_endpoint_disable(), where it belongs.  This patch (as754) does that.

It turns out to be completely safe to replace the del_timer_sync() with
a simple del_timer().  It doesn't matter if the timer routine is
running; hcd_root_hub_lock will synchronize the two threads and the
status URB will complete with an unlink error, as it should.

Signed-off-by: Alan Stern 
Signed-off-by: Greg Kroah-Hartman