Mike Isely [Wed, 30 Aug 2006 03:07:11 +0000 (22:07 -0500)]
cypress_m8: implement graceful failure handling
When receiving a fatal error from the USB core, e.g. EILSEQ (which can
happen if the polling interval is too short), fail gracefully.
Previously the driver would fill the log with useless error messages
or (more alarmingly) silently spin forever trying to write updated
control information to the device. This change implements a new flag
which if cleared indicates that the driver has failed. The flag will
be set on initialization, cleared on fatal errors, and anything else
that touches the USB port in the driver will abort if the flag is
clear. When the flag is cleared, a message will be logged indicating
that the driver has failed.
Signed-off-by: Mike Isely <isely@pobox.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Mike Isely [Wed, 30 Aug 2006 03:07:07 +0000 (22:07 -0500)]
cypress_m8: improve control endpoint error handling
Fix usb core function error return checks to look for negative errno
values, not positive errno values. This bug had rendered those checks
useless. Also remove attempted error recovery on control endpoints
for EPIPE - with control endpoints EPIPE does not indicate a halted
endpoint so trying to recover with usb_clear_halt() is not the correct
action.
Signed-off-by: Mike Isely <isely@pobox.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Mike Isely [Wed, 30 Aug 2006 03:07:04 +0000 (22:07 -0500)]
cypress_m8: use usb_fill_int_urb where appropriate
Rather than directly filling in URB fields, it's safer to use
usb_fill_int_urb(). This improves robustness of the driver; URB
changes in the future will not go uninitialized here. That point not
withstanding, this driver should at least be self-consistent. Either
use usb_fill_int_urb() everywhere or don't bother with it all.
Signed-off-by: Mike Isely <isely@pobox.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Mike Isely [Wed, 30 Aug 2006 03:06:59 +0000 (22:06 -0500)]
cypress_m8: use appropriate URB polling interval
The polling interval for the device can't always be 1msec. If it is
too quick, the device can fail causing a fatal (to the driver) EILSEQ
error from the USB core. The actual correct value is reported by the
device as part of its configuration data, so use that value as the
default. On a DeLorme Earthmate for example, the device reports that
it wants a 6msec interval. As part of this fix, the "interval" module
option has been fixed as well; the device's default can be overridden
by specifying interval=<value> as a module option.
Signed-off-by: Mike Isely <isely@pobox.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Tue, 19 Sep 2006 14:08:43 +0000 (10:08 -0400)]
USB: fix root-hub resume when CONFIG_USB_SUSPEND is not set
This patch (as786) removes a redundant test and fixes a problem
involving repeated system sleeps when CONFIG_USB_SUSPEND is not set.
During the first wakeup, the root hub's dev.power.power_state.event
field doesn't get updated, causing it not to be suspended during the
second sleep transition.
This takes care of the issue raised by Rafael J. Wysocki and Mattia
Dongili.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 19:47:11 +0000 (15:47 -0400)]
usbcore: remove usb_suspend_root_hub
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 <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch (as739) adds the basic infrastructure for USB autosuspend
and autoresume. The main features are:
PM usage counters added to struct usb_device and struct
usb_interface, indicating whether it's okay to autosuspend
them or they are currently in use.
Flag added to usb_device indicating whether the current
suspend/resume operation originated from outside or as an
autosuspend/autoresume.
Flag added to usb_driver indicating whether the driver
supports autosuspend. If not, no device bound to the driver
will be autosuspended.
Mutex added to usb_device for protecting PM operations.
Unlike the device semaphore, the locking rule for the pm_mutex
is that you must acquire the locks going _up_ the device tree.
New routines handling autosuspend/autoresume requests for
interfaces and devices.
Suspend and resume requests are propagated up the device tree
(but not outside the USB subsystem).
work_struct added to usb_device, for carrying out delayed
autosuspend requests.
Autoresume added (and autosuspend prevented) during probe and
disconnect.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 19:46:48 +0000 (15:46 -0400)]
usbcore: store each usb_device's level in the tree
This patch (as778) adds a field to struct usb_device to store the
device's level in the USB tree. In itself this number isn't really
important. But the overhead is very low, and in a later patch it will
be used for preventing bogus warnings from the lockdep checker.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Thu, 7 Sep 2006 23:54:22 +0000 (16:54 -0700)]
USB: UB: Let cdrecord to see a device with media absent
The command "cdrecord dev=/dev/uba x.iso" prints nasty garbage if a blank
is not in the drive. This happens because drivers have to set req->errors
separately from just returning zero uptodate with end_that_request_first,
end_that_request_last. These functions only set error in BIO, but sg_io()
ignores it.
Since we're on it, let cdrecord access a device when ->changed is set.
It's useful if someone wants to know device capabilities without burning
anything.
Signed-Off-By: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
David Brownell [Sat, 2 Sep 2006 10:13:45 +0000 (03:13 -0700)]
USB: net2280: update dma buffer allocation
This updates the code handling dma-coherent buffer allocations, basically
reusing code from the musb_hdrc driver. Instead of trying to work around two
significant limitations of the dma framework (memory wastage for buffers
smaller than a page, and inconsistency between calling context requirements
for allocation and free) this just works around one of them (the latter).
So count this as two steps forward (bugfixes: the latter issue could cause
errors on some platforms, and some MIPS changes broke code for the former),
and one step back (increasing cross-platform memory wastage).
Plus linelength and whitespace fixes; and minor data segment shrinkage.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
David Brownell [Wed, 30 Aug 2006 20:24:56 +0000 (13:24 -0700)]
USB: ethernet gadget avoids zlps for musb_hdrc
For systems using the Mentor HDRC controllers we get better TX DMA throughput
if we can avoid falling back to PIO to write zero length packets ... so tell
the driver to avoid ZLPs.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 15:35:21 +0000 (11:35 -0400)]
usbmon: don't call mon_dmapeek if DMA isn't being used
This patch (as755b) fixes a bug in usbmon. Rather than assuming all
USB host controllers use DMA, the code will check the usb_bus data
structure. If DMA isn't used, we don't want to try peeking into a
non-existent DMA buffer!
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 15:32:52 +0000 (11:32 -0400)]
usbcore: trim down usb_bus structure
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 <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 15:29:56 +0000 (11:29 -0400)]
usbcore: Add flag for whether a host controller uses DMA
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 <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 30 Aug 2006 15:27:36 +0000 (11:27 -0400)]
USB: remove struct usb_operations
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 <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 21 Aug 2006 16:08:19 +0000 (12:08 -0400)]
usbcore: help drivers to change device configs
It's generally a bad idea for USB interface drivers to try to change a
device's configuration, and usbcore doesn't provide any way for them
to do it. However in a few exceptional circumstances it can make
sense. This patch (as767) adds a roundabout mechanism to help drivers
that may need it.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 21 Aug 2006 16:00:53 +0000 (12:00 -0400)]
usb-storage: fix for UFI LUN detection
The UFI specification doesn't permit devices to indicate non-existent
LUNs in the manner prescribed by the SCSI spec. This patch (as773)
sets a special flag so that the SCSI scanner will recognize these
devices and treat them specially.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Matthew Dharm <mdharm-usb@one-eyed-alien.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Jules Villard [Tue, 22 Aug 2006 20:40:15 +0000 (22:40 +0200)]
USB: fix typo in drivers/usb/gadget/Kconfig
This tiny patch fixes a typo in drivers/usb/gadget/Kconfig. The typo
is present in 2.6.18-rc4 and in the corresponding -mm tree (and AFAIK,
FYI and FWIW was present in previous kernel versions as well).
Eric Sesterhenn [Mon, 21 Aug 2006 22:31:05 +0000 (15:31 -0700)]
USB: fix signedness issue in drivers/usb/gadget/ether.c
another gcc 4.1 signdness warning:
drivers/usb/gadget/ether.c:2028: warning: comparison of unsigned expression < 0 is always false
length is assigned the value of usb_ep_queue() which returns an int.
Directly after this it is checked for < 0, which can never be true. Making
length an int makes the error check work again.
Signed-off-by: Eric Sesterhenn <snakebyte@gmx.de> Acked-by: David Brownell <david-b@pacbell.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch marks some USB core's functions parameters as const. This
improves the design (we're saying to the caller that its parameter is
not going to be modified) and may help in compiler's optimisation work.
Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
This patch teaches the USB stack handling of WUSB devices (those whose
speed is USB_SPEED_VARIABLE). For these devices, we need to set ep0's
maxpacketsize to 512 (even though the device descriptor reports it as
0xff).
New code being pushed to linuxuwb.org requires this patch to connect WUSB
devices.
Change usb_get_configuration() so that it is more tolerant to devices
with bad configuration descriptors (it'll make it ignore
configurations that fail to load).
Matthew Dharm [Mon, 14 Aug 2006 00:30:14 +0000 (17:30 -0700)]
USB Storage: add rio karma eject support
This changeset from Keith Bennett (via Bob Copeland) moves the Karma
initializer to its own file and adds trapping of the START_STOP command to
enable eject of the device.
Signed-off-by: Keith Bennett <keith@mcs.st-and.ac.uk> Signed-off-by: Bob Copeland <me@bobcopeland.com> Signed-off-by: Matthew Dharm <mdharm-usb@one-eyed-alien.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
o CodingStyle fixes
o Removes trailing spaces
o Do not make not needed initialiation of automatic variables
o Use usb_endpoint_* functions
o If we get an error in the write URB callback print an error message instead
of a debug one
(Pretty unrelated changes, but spliting this up doesn't pay off as our main
changes are just CodingStyle fixes).
Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Fri, 11 Aug 2006 20:52:39 +0000 (16:52 -0400)]
usbcore: khubd and busy-port handling
We don't want khubd to start interfering in the device-resume process
merely because the PORT_STATUS_C_SUSPEND feature happens to be set.
Ports need to be marked as busy while a resume is taking place.
In addition, so long as ports are marked as busy, khubd won't be able to
clear their various status-change features. On an interrupt-driven root
hub this could lead to an interrupt storm. Root hub IRQs should not be
re-enabled until the busy_bits value is equal to 0.
This patch (as765) fixes these two potential problems.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Fri, 11 Aug 2006 20:01:45 +0000 (16:01 -0400)]
usbcore: make hcd_endpoint_disable wait for queue to drain
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 <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Fri, 11 Aug 2006 15:33:58 +0000 (11:33 -0400)]
UHCI: increase Resume-Detect-off delay
The UHCI controller in my laptop takes longer to turn off the
Resume-Detect bit than the 4 us allowed by uhci-hcd. Presumably other
computers will have the same problem.
This patch (as752) increases the maximum delay to 10 us, which should be
plenty, and uses polling to avoid penalizing systems which can turn the
bit off more quickly.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Aleksey Gorelov [Wed, 9 Aug 2006 00:24:08 +0000 (17:24 -0700)]
USB: Properly unregister reboot notifier in case of failure in ehci hcd
If some problem occurs during ehci startup, for instance, request_irq fails,
echi hcd driver tries it best to cleanup, but fails to unregister reboot
notifier, which in turn leads to crash on reboot/poweroff.
The following patch resolves this problem by not using reboot notifiers
anymore, but instead making ehci/ohci driver get its own shutdown method. For
PCI, it is done through pci glue, for everything else through platform driver
glue.
One downside: sa1111 does not use platform driver stuff, and does not have its
own shutdown hook, so no 'shutdown' is called for it now. I'm not sure if it
is really necessary on that platform, though.
Signed-off-by: Aleks Gorelov <dared1st@yahoo.com> Cc: Alan Stern <stern@rowland.harvard.edu> Cc: David Brownell <david-b@pacbell.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
These functions makes USB driver's code simpler when dealing with endpoints
by avoiding them from accessing the endpoint's descriptor structure directly
when they only need to know the endpoint's transfer type and/or
direction.
Please, read each functions' documentation in order to know how to use
them.
Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
USB: making the kernel -Wshadow clean - USB & completion
include/linux/usb.h causes a lot of -Wshadow warnings - fix them.
include/linux/usb.h:901: warning: declaration of 'complete' shadows a global declaration
include/linux/completion.h:52: warning: shadowed declaration is here
include/linux/usb.h:932: warning: declaration of 'complete' shadows a global declaration
include/linux/completion.h:52: warning: shadowed declaration is here
include/linux/usb.h:967: warning: declaration of 'complete' shadows a global declaration
include/linux/completion.h:52: warning: shadowed declaration is here
Ben Williamson [Tue, 1 Aug 2006 01:28:16 +0000 (11:28 +1000)]
USB: gmidi: New USB MIDI Gadget class driver.
This driver is glue between the USB gadget interface
and the ALSA MIDI interface. It allows us to appear
as a MIDI Streaming device to a host system on the
other end of a USB cable.
This includes linux/usb/audio.h and linux/usb/midi.h
containing definitions from the relevant USB specifications
for USB audio and USB MIDI devices.
The following changes have been made since the first RFC
posting:
* Bug fixes to endpoint handling.
* Workaround for USB_REQ_SET_CONFIGURATION handling,
not understood yet.
* Added SND and SND_RAWMIDI dependencies in Kconfig.
* Moved usb_audio.h and usb_midi.h to usb/*.h
* Added module parameters for ALSA card index and id.
* Added module parameters for USB descriptor IDs and strings.
* Removed some unneeded stuff inherited from zero.c, more to go.
* Provide DECLARE_* macros for the variable-length structs.
* Use kmalloc instead of usb_ep_alloc_buffer.
* Limit source to 80 columns.
* Return actual error code instead of -ENOMEM in a few places.
Signed-off-by: Ben Williamson <ben.williamson@greyinnovation.com> Cc: David Brownell <david-b@pacbell.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Milan Svoboda [Mon, 26 Jun 2006 14:19:00 +0000 (07:19 -0700)]
USB: fix ep_config to return correct value
This patch fixes ep_config to return correct value. Without patch
ep_config returns submitted lenght minus 4 on succes. With this
patch applied, whole submitted lenght is returned.
ep_config parses submitted data and if buffer starts with (int) 1
it is parsed, otherwise error is reported. Problem is that ep_config
returns size of buffer minus 4 on success. I think that size of buffer
should be returned instead, because there were no problems and
all data were processed.
Signed-off-by: Milan Svoboda <msvoboda@ra.rockwell.com> Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
David Brownell [Mon, 31 Jul 2006 14:29:39 +0000 (07:29 -0700)]
USB: build fixes: ohci-omap
The ohci-omap code has diverged from the working version in the linux-omap
tree; this syncs up the versions:
- Another clock is needed in various cases
- The omap-1510 iommu code needs to be #ifdeffed out on newer parts
- Saner use of the HCD framework
- Various other changes, e.g. a Nokia 770 quirk
And some minor dead-whitespace removal.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
- split wacom.c into 4 files: wacom.h, wacom_wac.h, wacom_sys.c, and wacom_wac.c
- where wacom_sys.c deals with system specific code,
- and wacom_wac.c deals with Wacom specific code
USB: usbcore get rid of the timer in usb_start_wait_urb()
This patch uses completion timeout instead of a timer to implement
a timeout when submitting an URB in usb_start_wait_urb().
It also fixes a small issue. With the previous code, if no timeout
happened and the URB's status was set to ECONNRESET value, the code
assumed wrongly that a timeout had occured.
Commit b512504e5671f83638be0ddr085c4b1832f623d3 made ipaq_open() a bit
messy by moving the read urb submission far from its usb_fill_bulk_urb()
call and the comment explaining what it does.
This patch put they together again. Although only compiled tested, should
not break the fix introduced by b512504e5671f83638be0ddr085c4b1832f623d3,
of course.
Signed-off-by: Luiz Fernando N. Capitulino <lcapitulino@mandriva.com.br> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Randy Dunlap [Mon, 7 Aug 2006 22:56:40 +0000 (15:56 -0700)]
usbnet: printk format warning
Fix printk format warning(s):
drivers/usb/net/usbnet.c:654: warning: int format, different type arg (arg 3)
The fact that rx_urb_size happens to be a size_t has propagated all the way
back to this printk. It's fragile to be using %z in this case - let's just
typecast the args instead.
Signed-off-by: Randy Dunlap <rdunlap@xenotime.net> Signed-off-by: Andrew Morton <akpm@osdl.org> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
USB: pl2303: cosmetic changes to pl2303_buf_{clear, data_avail}
Changes the functions pl2303_buf_clear and pl2303_buf_data_avail for
the purpose of keeping them under the 80 column limit, making them
more similar to similar functions and making then simpler.
Daniel Ritz [Mon, 31 Jul 2006 19:43:24 +0000 (21:43 +0200)]
usbtouchscreen: version 0.4
changes over 0.3:
- some more eGalax device IDs (from eGalax driver/spec)
- return the error code in probe()
- 3M/MTouch init fixes, tested by Don Alexander
- eGalax fixes for bugs in multi-packet handling, spottet by Pieter Grimmerink
- support for some eTurboTouch devices, mostly by Pieter Grimmerink
- support for Gunze AHL61 controller (untested, but simple enough)
Signed-off-by: Daniel Ritz <daniel.ritz@gmx.ch> Cc: Pieter Grimmerink <p.grimmerink@inepro.com> Cc: Don Alexander <debug@roosoft.ltd.uk> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:13:50 +0000 (22:13 -0400)]
usbcore: suspending devices with no driver
Since usb_generic can be unbound from a USB device, we need to be able
to handle the possibility that a suspend or resume request arrives for a
device with no driver. This patch (as735) arranges things so that
resume requests will fail and suspend requests will use the standard USB
port-suspend code. Attempts to suspend or resume an unbound interface
are handled similarly (although the error caused by trying to resume an
unbound interface is dropped by the calling routine).
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:13:04 +0000 (22:13 -0400)]
usbcore: fix up device and power state tests
This patch (as734) rationalizes the various tests of device state and
power states. There are duplications and mistaken tests in several
places.
Perhaps the most interesting challenge is where the hub driver tests to
see that all the child devices are suspended before allowing itself to
be suspended. When CONFIG_USB_SUSPEND is set the test is
straightforward, since we expect that the children _will_ be suspended.
But when CONFIG_USB_SUSPEND isn't set, it's not so clear what should be
done. The code compromises by checking the child's
power.power_state.event field.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:12:19 +0000 (22:12 -0400)]
usbcore: set device and power states properly
This patch (as733) fixes up the places where device states and power
states are set in usbcore. Right now things are duplicated or missing;
this should straighten things out.
The idea is that udev->state is USB_STATE_SUSPENDED exactly when the
device's upstream port has been suspended, whereas
udev->dev.power.power_state.event reflects the result of the last call
to the suspend/resume routines (which might not actually change the
device state, especially if CONFIG_USB_SUSPEND isn't set).
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:11:44 +0000 (22:11 -0400)]
usbcore: track whether interfaces are suspended
Currently we rely on intf->dev.power.power_state.event for tracking
whether intf is suspended. This is not a reliable technique because
that value is owned by the PM core, not by usbcore. This patch (as718b)
adds a new flag so that we can accurately tell which interfaces are
suspended and which aren't.
At first one might think these flags aren't needed, since interfaces
will be suspended along with their devices. It turns out there are a
couple of intermediate situations where that's not quite true, such as
while processing a remote-wakeup request.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:11:02 +0000 (22:11 -0400)]
usbcore: resume device resume recursion
This patch (as717b) removes the existing recursion in hub resume code:
Resuming a hub will no longer automatically resume the devices attached
to the hub.
At the same time, it adds one level of recursion: Suspending a USB
device will automatically suspend all the device's interfaces. Failure
at an intermediate stage will cause all the already-suspended interfaces
to be resumed. Attempts to suspend or resume an interface by itself will
do nothing, although they won't return an error. Thus the regular
system-suspend and system-resume procedures should continue to work as
before; only runtime PM will be affected.
The patch also removes the code that tests state of the interfaces
before suspending a device. It's no longer needed, since everything
gets suspended together.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:10:15 +0000 (22:10 -0400)]
usbcore: split suspend/resume for device and interfaces
This patch (as716b) splits up the core suspend and resume routines into
two parts each: one for handling devices and one for handling
interfaces. The behavior of the parts should be the same as in the old
unified code.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:09:35 +0000 (22:09 -0400)]
usbcore: make usb_generic a usb_device_driver
This patch (as714b) makes usb_generic into a usb_device_driver capable
of being probed and unbound, just like other drivers. A fair amount of
the work that used to get done during discovery or removal of a USB
device have been moved to the probe and disconnect methods of
usb_generic: creating the sysfs attributes and selecting an initial
configuration. However the normal behavior should continue to be the
same as before.
We will now have the possibility of creating other USB device drivers,
They will assist with exporting devices to remote systems
(USB-over-TCPIP) or to paravirtual guest operating systems.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:08:49 +0000 (22:08 -0400)]
usbcore: add usb_device_driver definition
This patch (as732) adds a usb_device_driver structure, for representing
drivers that manage an entire USB device as opposed to just an
interface. Support routines like usb_register_device_driver,
usb_deregister_device_driver, usb_probe_device, and usb_unbind_device
are also added.
Unlike an earlier version of this patch, the new code is type-safe. To
accomplish this, the existing struct driver embedded in struct
usb_driver had to be wrapped in an intermediate wrapper. This enables
the core to tell at runtime whether a particular struct driver belongs
to a device driver or to an interface driver.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:08:06 +0000 (22:08 -0400)]
usbcore: move code among source files
This revised patch (as713b) moves a few routines among source files in
usbcore. Some driver-related code in usb.c (claiming interfaces and
matching IDs) is moved to driver.c, where it belongs. Also the
usb_generic stuff in driver.c is moved to a new source file: generic.c.
(That's the reason for revising the patch.) Although not very big now,
it will get bigger in a later patch.
None of the code has been changed; it has only been re-arranged.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:07:21 +0000 (22:07 -0400)]
usbcore: rename usb_suspend_device to usb_port_suspend
This revised patch (as715b) renames usb_suspend_device to
usb_port_suspend, usb_resume_device to usb_port_resume, and
finish_device_resume to finish_port_resume. There was no objection to
the original version of the patch so this should be okay to apply.
The revision was needed only because I have re-arranged the order of the
earlier patches.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:06:36 +0000 (22:06 -0400)]
usb-skeleton: don't submit URBs after disconnection
This patch (as712b) is a slight revision of one submitted earlier. It
fixes the usb-skeleton example driver so that it won't try to submit
URBs after skel_disconnect() has returned. This could cause errors, if
the driver was unbound and then a different driver was bound to the
device. It also fixes a couple of small bugs in the skel_write()
routine.
The revised patch uses a slightly different test, suggested by Dave
Brownell, for determining whether to free a transfer buffer. It's a
little clearer than the earlier version.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:05:56 +0000 (22:05 -0400)]
usbfs: detect device unregistration
This patch (as711b) is a revised version of an earlier submission. It
modifies the usbfs code to detect when a device has been unregistered from
usbfs, even if the device is still connected. Although this can't happen
now, it will be able to happen after the upcoming changes to usb_generic.
Nobody objected to this patch when it was submitted before, so it should
be okay to apply this version. The revision is merely to take into
account the changes introduced by as723, which touches the same driver.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Sun, 2 Jul 2006 02:05:01 +0000 (22:05 -0400)]
usbfs: private mutex for open, release, and remove
The usbfs code doesn't provide sufficient mutual exclusion among open,
release, and remove. Release vs. remove is okay because they both
acquire the device lock, but open is not exclusive with either one. All
three routines modify the udev->filelist linked list, so they must not
run concurrently.
Apparently someone gave this a minimum amount of thought in the past by
explicitly acquiring the BKL at the start of the usbdev_open routine.
Oddly enough, there's a comment pointing out that locking is unnecessary
because chrdev_open already has acquired the BKL.
But this ignores the point that the files in /proc/bus/usb/* are not
char device files; they are regular files and so they don't get any
special locking. Furthermore it's necessary to acquire the same lock in
the release and remove routines, which the code does not do.
Yet another problem arises because the same file_operations structure is
accessible through both the /proc/bus/usb/* and /dev/usb/usbdev* file
nodes. Even when one of them has been removed, it's still possible for
userspace to open the other. So simple locking around the individual
remove routines is insufficient; we need to lock the entire
usb_notify_remove_device notifier chain.
Rather than rely on the BKL, this patch (as723) introduces a new private
mutex for the purpose. Holding the BKL while invoking a notifier chain
doesn't seem like a good idea.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Thu, 6 Jul 2006 19:37:42 +0000 (15:37 -0400)]
usbcore: add configuration_string to attribute group
This patch (as737b) does a very small cleanup of core/sysfs.c by adding
the configuration_string attribute file to the existing attribute group
instead of treating it separately. It doesn't need this separate
treatment because unlike the other device string attributes, it changes
along with the active configuration.
The patch also fixes a simple typo (which, oddly enough, doesn't seem to
bother the compiler).
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
projects / linux-2.6 / log