Pete Zaitcev [Tue, 14 Aug 2007 20:19:16 +0000 (13:19 -0700)]
usblp: Fix a double kfree
If submit fails, slab hits a BUG() because of a double kfree.
The today's lesson is, you cannot just slap USB_FREE_BUFFER on code
without adjusting the error paths.
The patch is made bigger by opportunistic refactoring.
Signed-Off-By: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Tue, 14 Aug 2007 07:37:51 +0000 (00:37 -0700)]
usbmon: Drop DMA mapping for setup packet
Setup packet must be visible in virtual space. There's absolutely no
good reason to implement any kind of zero-copy transfer of 8 bytes, and
the documentation in usb.h is explicit about it. So, drop DMA remapping.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Tue, 14 Aug 2007 07:33:40 +0000 (00:33 -0700)]
usbmon: Update pipe removal to suit my taste
This is a set of small updates to Alan's work to make the code more to
my liking. Mostly premature optimizations, but also direction of control
transfers in the binary interface was always out.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Cc: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Tue, 14 Aug 2007 05:54:29 +0000 (22:54 -0700)]
usblp: Cosmetics
This is a small bunch of cosmetic fixes:
- Timeout is not a write timeout anymore, rename
- Condition in poll was confusingly backwards, invert and simplify
- The comment log gave a wrong impression of version 0.13, terminate it.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Tue, 14 Aug 2007 05:50:13 +0000 (22:50 -0700)]
usblp: mutex in usblp_check_status
Add a mutex to protect the ->statusbuf. Not really an issue, because CUPS
is single-threaded when it talks to the printer, but I feel safer this way.
This should be deadlock-free, but I kept this as a separate patch in case
someone ends running a git bisect.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 8 Aug 2007 15:59:18 +0000 (11:59 -0400)]
USB: don't touch sysfs stuff when altsetting is unchanged
This patch (as955) prevents the interface-related sysfs files and
endpoint pseudo-devices from being deleted and recreated when a call
to usb_set_interface() specifies the current altsetting. Since the
altsetting doesn't get changed, there's no need to do anything.
Furthermore, avoiding changes to the endpoint devices will be
necessary in the future. This code is called from usb_reset_device(),
which gets invoked for reset-resume processing, but upcoming changes
to the PM and driver cores will make it impossible to register devices
while a suspend/resume transition is in progress. Since we don't need
to re-register those endpoint devices anyhow, it's best to skip the
whole thing.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Wed, 8 Aug 2007 15:48:02 +0000 (11:48 -0400)]
USB: make HCDs responsible for managing endpoint queues
This patch (as954) implements a suggestion of David Brownell's. Now
the host controller drivers are responsible for linking and unlinking
URBs to/from their endpoint queues. This eliminates the possiblity of
strange situations where usbcore thinks an URB is linked but the HCD
thinks it isn't. It also means HCDs no longer have to check for URBs
being dequeued before they were fully enqueued.
In addition to the core changes, this requires changing every host
controller driver and the root-hub URB handler. For the most part the
required changes are fairly small; drivers have to call
usb_hcd_link_urb_to_ep() in their urb_enqueue method,
usb_hcd_check_unlink_urb() in their urb_dequeue method, and
usb_hcd_unlink_urb_from_ep() before giving URBs back. A few HCDs make
matters more complicated by the way they split up the flow of control.
In addition some method interfaces get changed. The endpoint argument
for urb_enqueue is now redundant so it is removed. The unlink status
is required by usb_hcd_check_unlink_urb(), so it has been added to
urb_dequeue.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> CC: David Brownell <david-b@pacbell.net> CC: Olav Kongas <ok@artecdesign.ee> CC: Tony Olech <tony.olech@elandigitalsystems.com> CC: Yoshihiro Shimoda <shimoda.yoshihiro@renesas.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
These USB API functions will do the full authorization/deauthorization
to be used for a device. When authorized we effectively allow a
configuration to be set. Reverse that when deauthorized.
Effectively this means that we have to clean all the configuration
descriptors on deauthorize and reload them when we authorized. We could
do without throwing them out for wired devices, but for wireless, we can
read them only after authenticating, and thus, when authorizing an
authenticated device we would need to read them. So to simplify, always
release them on deauthorize(), re-read them on authorize().
Also fix leak reported by Ragner Magalhaes; in usb_deauthorize_device(),
bNumConfigurations was being set to zero before the for loop, and thus
the different raw descriptors where never being freed.
usb: split usb_new_device for clarity and refactoring
This patch takes hub.c:usb_new_device() and splits it in three parts:
- The actual actions of adding a new device (quirk detection,
announcement and autoresume tracking)
- Actual discovery and probing of the configuration and interfaces
(split into __usb_configure_device())
- Configuration of the On-the-go parameters (split into
__usb_configure_device_otg()).
The fundamental reasons for doing this split are clarity (smaller
functions are easier to maintain) and to allow part of the code to be
reused when authorizing devices to connect.
When a device is authorized connection, we need to run through the
hoops we didn't run when it was connected but not authorized, which is
basically parsing the configurations and probing
them. usb_configure_device() will do that for us.
If called and the device is not authorized to be used, then we won't
choose a configuration (as they are not a concept that exists for an
unauthorized device). However, the device is added to the system.
usb: add the concept of default authorization to USB hosts
This introduces /sys/bus/devices/usb*/authorized_default; it dictates
what is going to be the default authorization state for devices
connected to the host. User space can set that using the sysfs file.
We hook to the root hub instead of to the device controller as it is
quite easy to get to it in sysfs from the device structure (device
5-4.3 is usb5) vs. backtracking to the controller device.
By default it is set to be 'authorized' (!0) for normal, wired USB
devices and 'unauthorized' (0) for Wireless USB devices.
As suggested by Adrian Bunk, make authorized_default static
This just modifies 'struct usb_device' to contain the 'authorized'
bit. It also adds a 'wusb' bit. This is needed because nonauthorized
(and thus non-authenticated) wusb devices will fail certain kind of
simple requests (such as string descriptors). By knowing the device is
WUSB, we just avoid them.
David Brownell [Thu, 2 Aug 2007 07:00:38 +0000 (00:00 -0700)]
USB: gadget: ethernet gadget cleanups, shrinkage
Clean up the ethernet gadget, using newer APIs and conventions:
- gadget_is_dualspeed() and gadget_is_otg() ... #ifdef removal
- Remove many now-needless #includes
- Use the VERBOSE_DEBUG convention
- Minor whitespace fixes.
- Fix a warning from "sparse".
Surprisingly, this saved about 2K of code (16%) on a fullspeed-only
ARMv4 platform. I'm bit puzzled by that (it's so much!), but approve
of the result.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
David Brownell [Thu, 2 Aug 2007 06:58:22 +0000 (23:58 -0700)]
USB: gadget: gadget_is_{dualspeed,otg} predicates and cleanup
This adds two small inlines to the gadget stack, which will
often evaluate to compile-time constants. That can help
shrink object code and remove #ifdeffery.
- gadget_is_dualspeed(), currently always a compile-time
constant (depending on which controller is selected).
- gadget_is_otg(), usually a compile time "false", but this
is a runtime test if the platform enables OTG (since it's
reasonable to populate boards with different USB sockets).
It also updates two peripheral controller drivers to use these:
- fsl_usb2_udc, mostly OTG-related bugfixes: non-OTG devices
must follow the rules about drawing VBUS power, and OTG ones
need to reject invalid SET_FEATURE requests.
- omap_udc, just scrubbing a bit of #ifdeffery.
And also gadgetfs, which lost some #ifdefs and moved to a more
standard handling of DEBUG and VERBOSE_DEBUG.
The main benefits come from patches which will follow.
Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Nico Erfurth [Thu, 2 Aug 2007 20:25:20 +0000 (22:25 +0200)]
USB: Add drivers/usb/misc/iowarrior.c to the Makefile
This patch adds the new iowarrior module to the Makefile in drivers/usb.
Currently the module isn't build unless another driver from usb/misc is
selected.
Alan Stern [Thu, 2 Aug 2007 19:06:54 +0000 (15:06 -0400)]
USB: separate out endpoint queue management and DMA mapping routines
This patch (as953) separates out three key portions from
usb_hcd_submit_urb(), usb_hcd_unlink_urb(), and usb_hcd_giveback_urb()
and puts them in separate functions of their own. In the next patch,
these functions will be called directly by host controller drivers
while holding their private spinlocks, which will remove the
possibility of some unpleasant races.
The code responsible for mapping and unmapping DMA buffers is also
placed into a couple of separate subroutines, for the sake of
cleanliness and consistency.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Thu, 2 Aug 2007 19:05:45 +0000 (15:05 -0400)]
USB: update spinlock usage for root-hub URBs
This patch (as952) adjusts the spinlock usage in the root-hub
emulation part of usbcore, to make it match more closely the pattern
used by regular host controller drivers. To wit: The private lock
(usb_hcd_root_hub_lock) is held throughout the important parts, and it
is dropped temporarily without re-enabling interrupts around the call
to usb_hcd_giveback_urb().
A nice side effect is that the code now avoids calling
local_irq_save(), thereby becoming more RT-friendly.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Thu, 2 Aug 2007 19:04:52 +0000 (15:04 -0400)]
USB: cleanup for previous patches
This patch (as951) cleans up a few loose ends from earlier patches.
Redundant checks for non-NULL urb->dev are removed, as are checks of
urb->dev->bus (which can never be NULL). Conversely, a check for
non-NULL urb->ep is added to the unlink paths.
A homegrown round-down-to-power-of-2 loop is simplified by using the
ilog2 routine. The comparison in usb_urb_dir_in() is made more
transparent.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:09:28 +0000 (17:09 -0400)]
USB: avoid urb->pipe in usbfs
This patch (as948) removes most of the references to urb->pipe from
the usbfs routines in devio.c. The one tricky aspect is in
snoop_urb(), which can be called before the URB is submitted and which
uses usb_urb_dir_in(). For this to work properly, the URB's direction
flag must be set manually in proc_do_submiturb().
The patch also fixes a minor bug; the wValue, wIndex, and wLength
fields were snooped in proc_do_submiturb() without conversion from
le16 to CPU-byte-ordering.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:08:43 +0000 (17:08 -0400)]
USB: address-0 handling during device initialization
This patch (as947) changes the device initialization and enumeration
code in hub.c; now udev->devnum will be set to 0 while the device is
being accessed at address 0. Until now this wasn't needed because the
address value was passed as part of urb->pipe; without that field the
device address must be stored elsewhere.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:07:21 +0000 (17:07 -0400)]
USB: avoid using urb->pipe in usbcore
This patch (as946) eliminates many of the uses of urb->pipe in
usbcore. Unfortunately there will have to be a significant API
change, affecting all USB drivers, before we can remove it entirely.
This patch contents itself with changing only the interface to
usb_buffer_map_sg() and friends: The pipe argument is replaced with a
direction flag. That can be done easily because those routines get
used in only one place.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:06:16 +0000 (17:06 -0400)]
USB: add direction bit to urb->transfer_flags
This patch (as945) adds a bit to urb->transfer_flags for recording the
direction of the URB. The bit is set/cleared automatically in
usb_submit_urb() so drivers don't have to worry about it (although as
a result, it isn't valid until the URB has been submitted). Inline
routines are added for easily checking an URB's direction. They
replace calls to usb_pipein in the DMA-mapping parts of hcd.c.
For non-control endpoints, the direction is determined directly from
the endpoint descriptor. However control endpoints are
bi-directional; for them the direction is determined from the
bRequestType byte and the wLength value in the setup packet.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:05:22 +0000 (17:05 -0400)]
USB: add ep->enable
This patch (as944) adds an explicit "enabled" field to the
usb_host_endpoint structure and uses it in place of the current
mechanism. This is merely a time-space tradeoff; it makes checking
whether URBs may be submitted to an endpoint simpler. The existing
mechanism is efficient when converting urb->pipe to an endpoint
pointer, but it's not so efficient when urb->ep is used instead.
As a side effect, the procedure for enabling an endpoint is now a
little more complicated. The ad-hoc inline code in usb.c and hub.c
for enabling ep0 is now replaced with calls to usb_enable_endpoint,
which is no longer static.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Alan Stern [Mon, 30 Jul 2007 21:04:37 +0000 (17:04 -0400)]
USB: add urb->ep
This patch (as943) prepares the way for eliminating urb->pipe by
introducing an endpoint pointer into struct urb. For now urb->ep
is set by usb_submit_urb() from the pipe value; eventually drivers
will set it themselves and we will remove urb->pipe completely.
The patch also adds new inline routines to retrieve an endpoint
descriptor's number and transfer type, essentially as replacements for
usb_pipeendpoint and usb_pipetype.
usb_submit_urb(), usb_hcd_submit_urb(), and usb_hcd_unlink_urb() are
converted to use the new field and new routines. Other parts of
usbcore will be converted in later patches.
Signed-off-by: Alan Stern <stern@rowland.harvard.edu> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Mike Nuss [Wed, 1 Aug 2007 20:24:30 +0000 (13:24 -0700)]
USB: OHCI handles more ZFMicro quirks
The ZF Micro OHCI controller exhibits unexpected behavior that seems to be
related to high load. Under certain conditions, the controller will
complete a TD, remove it from the endpoint's queue, and fail to add it to
the donelist. This causes the endpoint to appear to stop responding. Worse,
if the device is removed while in that state, OHCI will hang while waiting
for the orphaned TD to complete. The situation is not recoverable without
rebooting.
This fix enhances the scope of the existing OHCI_QUIRK_ZFMICRO flag:
1. A watchdog routine periodically scans the OHCI structures to check
for orphaned TDs. In these cases the TD is taken back from the
controller and completed normally.
2. If a device is removed while the endpoint is hung but before the
watchdog catches the situation, any outstanding TDs are taken back
from the controller in the 'sanitize' phase.
The ohci-hcd driver used to print "INTR_SF lossage" in this situation;
this changes it to the universally accurate "ED unlink timeout". Other
instances of this message presumably have different root causes.
Both this Compaq quirk and a NEC quirk are now properly compiled out for
non-PCI builds of this driver.
Signed-off-by: Mike Nuss <mike@terascala.com> Signed-off-by: David Brownell <dbrownell@users.sourceforge.net> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Mon, 23 Jul 2007 09:17:49 +0000 (02:17 -0700)]
usblp: Make use of URB_FREE_BUFFER
Employ the new API URB_FREE_BUFFER that we've got. There was talk of a combined
constructor for this case, but apparently it's not happening, so just set the
flag explicitly for now.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
Pete Zaitcev [Mon, 23 Jul 2007 08:58:15 +0000 (01:58 -0700)]
usblp: Implement the ENOSPC convention
This patch implements a mode when a printer returns ENOSPC when it runs
out of paper. The default remains the same as before. An application which
wishes to use this function has to enable it explicitly with an ioctl
LPABORT.
This is done on a request by our (Fedora) CUPS guy, Tim Waugh. The API is
similar enough to the lp0's one that CUPS works with both (but see below),
but it's has some differences.
Most importantly, the abort mode is persistent in case of lp0: once tunelp
was run your cat fill blow up until you reboot or run tunelp again. For
usblp, I made it so the abort mode is only in effect as long as device
is open. This way you can mix and match CUPS and cat(1) freely and nothing
bad happens even if you run out of paper. It is also safer in the face
of any unexpected crashes.
It has to be noted that mixing LPABORT and O_NONBLOCK is not advised.
It probably does not do what you want: instead of returning -ENOSPC
it will always return -EAGAIN (because it would otherwise block while
waiting for the paper). Applications which use O_NONBLOCK should continue
to use LPGETSTATUS like before.
Finally, CUPS actually requires patching to take full advantage of this.
It has several components; those which invoke LPABORT work, but some of
them need the ioctl added. This is completely compatible, you can mix
old CUPS and new kernels or vice versa.
Signed-off-by: Pete Zaitcev <zaitcev@redhat.com> Signed-off-by: Greg Kroah-Hartman <gregkh@suse.de>
USB: Remove unneeded pointer intf from speedtch_upload_firmware()
This trivial patch removes the unneeded pointer intf returned from
usb_ifnum_to_if(), which is never used. The check for NULL can be simply done
by if (!usb_ifnum_to_if(usb_dev, 2)).
David S. Miller [Fri, 12 Oct 2007 05:15:08 +0000 (22:15 -0700)]
[ZLIB]: Fix external builds of zlib_inflate code.
Move zlib_inflate_blob() out into it's own source file,
infutil.c, so that things like the powerpc zImage builder
in arch/powerpc/boot/Makefile don't end up trying to
compile it.
Signed-off-by: David S. Miller <davem@davemloft.net>
Make sure and not dump reserved areas of device space.
Touching some of these causes machine check exceptions on boards
like D-Link DGE-550SX.
Coding note, used a complex switch statement rather than bitmap
because it is easier to relate the block values to the documentation
rather than looking at a encoded bitmask.
Signed-off-by: Stephen Hemminger <shemminger@linux-foundation.org> Signed-off-by: David S. Miller <davem@davemloft.net>
Linus Torvalds [Fri, 12 Oct 2007 04:55:47 +0000 (21:55 -0700)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc
* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/paulus/powerpc: (408 commits)
[POWERPC] Add memchr() to the bootwrapper
[POWERPC] Implement logging of unhandled signals
[POWERPC] Add legacy serial support for OPB with flattened device tree
[POWERPC] Use 1TB segments
[POWERPC] XilinxFB: Allow fixed framebuffer base address
[POWERPC] XilinxFB: Add support for custom screen resolution
[POWERPC] XilinxFB: Use pdata to pass around framebuffer parameters
[POWERPC] PCI: Add 64-bit physical address support to setup_indirect_pci
[POWERPC] 4xx: Kilauea defconfig file
[POWERPC] 4xx: Kilauea DTS
[POWERPC] 4xx: Add AMCC Kilauea eval board support to platforms/40x
[POWERPC] 4xx: Add AMCC 405EX support to cputable.c
[POWERPC] Adjust TASK_SIZE on ppc32 systems to 3GB that are capable
[POWERPC] Use PAGE_OFFSET to tell if an address is user/kernel in SW TLB handlers
[POWERPC] 85xx: Enable FP emulation in MPC8560 ADS defconfig
[POWERPC] 85xx: Killed <asm/mpc85xx.h>
[POWERPC] 85xx: Add cpm nodes for 8541/8555 CDS
[POWERPC] 85xx: Convert mpc8560ads to the new CPM binding.
[POWERPC] mpc8272ads: Remove muram from the CPM reg property.
[POWERPC] Make clockevents work on PPC601 processors
...
Fixed up conflict in Documentation/powerpc/booting-without-of.txt manually.
[POWERPC] Add legacy serial support for OPB with flattened device tree
Currently find_legacy_serial_ports() can find no serial ports on the
OPB with flattened device tree. Thus no legacy boot console can be
initialized. Just the early udbg console works, which is initialized
with udbg_init_44x_as1 on the UART's physical address specified in
kernel config. This happens because we look for ns16750 serial
devices only and expect opb node to have a device type property. This
patch makes it look for ns16550-compatible devices and use
of_device_is_compatible() for opb in case device type is not
specified.
Signed-off-by: Valentine Barshak <vbarshak@ru.mvista.com> Signed-off-by: Paul Mackerras <paulus@samba.org>
Paul Mackerras [Thu, 11 Oct 2007 10:37:10 +0000 (20:37 +1000)]
[POWERPC] Use 1TB segments
This makes the kernel use 1TB segments for all kernel mappings and for
user addresses of 1TB and above, on machines which support them
(currently POWER5+, POWER6 and PA6T).
We detect that the machine supports 1TB segments by looking at the
ibm,processor-segment-sizes property in the device tree.
We don't currently use 1TB segments for user addresses < 1T, since
that would effectively prevent 32-bit processes from using huge pages
unless we also had a way to revert to using 256MB segments. That
would be possible but would involve extra complications (such as
keeping track of which segment size was used when HPTEs were inserted)
and is not addressed here.
Parts of this patch were originally written by Ben Herrenschmidt.
Grant Likely [Wed, 10 Oct 2007 18:31:51 +0000 (04:31 +1000)]
[POWERPC] XilinxFB: Add support for custom screen resolution
Some custom implementations of the xilinx fb can use resolutions other
than 640x480. This patch allows the resolution to be specified in the
device tree or the xilinx_platform_data structure.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca> Signed-off-by: Paul Mackerras <paulus@samba.org>
Grant Likely [Wed, 10 Oct 2007 18:31:46 +0000 (04:31 +1000)]
[POWERPC] XilinxFB: Use pdata to pass around framebuffer parameters
The call to xilinxfb_assign is getting unwieldy when adding features
to the Xilinx framebuffer driver. Change xilinxfb_assign() to accept
a pointer to a xilinxfb_platform_data structure to prepare for adding
additition configuration options.
Signed-off-by: Grant Likely <grant.likely@secretlab.ca> Signed-off-by: Paul Mackerras <paulus@samba.org>
Linus Torvalds [Fri, 12 Oct 2007 02:43:13 +0000 (19:43 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/roland/infiniband: (87 commits)
mlx4_core: Fix section mismatches
IPoIB: Allow setting policy to ignore multicast groups
IB/mthca: Mark error paths as unlikely() in post_srq_recv functions
IB/ipath: Minor fix to ordering of freeing and zeroing of tid pages.
IB/ipath: Remove redundant link state checks
IB/ipath: Fix IB_EVENT_PORT_ERR event
IB/ipath: Better handling of unexpected GPIO interrupts
IB/ipath: Maintain active time on all chips
IB/ipath: Fix QHT7040 serial number check
IB/ipath: Indicate a couple of chip bugs to userspace
IB/ipath: iba6110 rev4 no longer needs recv header overrun workaround
IB/ipath: Use counters in ipath_poll and cleanup interrupts in ipath_close
IB/ipath: Remove duplicate copy of LMC
IB/ipath: Add ability to set the LMC via the sysfs debugging interface
IB/ipath: Optimize completion queue entry insertion and polling
IB/ipath: Implement IB_EVENT_QP_LAST_WQE_REACHED
IB/ipath: Generate flush CQE when QP is in error state
IB/ipath: Remove redundant code
IB/ipath: Future proof eeprom checksum code (contents reading)
IB/ipath: UC RDMA WRITE with IMMEDIATE doesn't send the immediate
...
Linus Torvalds [Fri, 12 Oct 2007 02:40:14 +0000 (19:40 -0700)]
Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6
* 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/davem/net-2.6: (867 commits)
[SKY2]: status polling loop (post merge)
[NET]: Fix NAPI completion handling in some drivers.
[TCP]: Limit processing lost_retrans loop to work-to-do cases
[TCP]: Fix lost_retrans loop vs fastpath problems
[TCP]: No need to re-count fackets_out/sacked_out at RTO
[TCP]: Extract tcp_match_queue_to_sack from sacktag code
[TCP]: Kill almost unused variable pcount from sacktag
[TCP]: Fix mark_head_lost to ignore R-bit when trying to mark L
[TCP]: Add bytes_acked (ABC) clearing to FRTO too
[IPv6]: Update setsockopt(IPV6_MULTICAST_IF) to support RFC 3493, try2
[NETFILTER]: x_tables: add missing ip6t_modulename aliases
[NETFILTER]: nf_conntrack_tcp: fix connection reopening
[QETH]: fix qeth_main.c
[NETLINK]: fib_frontend build fixes
[IPv6]: Export userland ND options through netlink (RDNSS support)
[9P]: build fix with !CONFIG_SYSCTL
[NET]: Fix dev_put() and dev_hold() comments
[NET]: make netlink user -> kernel interface synchronious
[NET]: unify netlink kernel socket recognition
[NET]: cleanup 3rd argument in netlink_sendskb
...
Fix up conflicts manually in Documentation/feature-removal-schedule.txt
and my new least favourite crap, the "mod_devicetable" support in the
files include/linux/mod_devicetable.h and scripts/mod/file2alias.c.
(The latter files seem to be explicitly _designed_ to get conflicts when
different subsystems work with them - that have an absolutely horrid
lack of subsystem separation!)
Linus Torvalds [Fri, 12 Oct 2007 02:21:23 +0000 (19:21 -0700)]
Merge branch 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus
* 'upstream' of git://ftp.linux-mips.org/pub/scm/upstream-linus: (80 commits)
[MIPS] tlbex.c: Cleanup __init usage.
[MIPS] WRPPMC serial support move to platform device
[MIPS] R1: Fix hazard barriers to make kernels work on R2 also.
[MIPS] VPE: reimplement ELF loader.
[MIPS] cleanup WRPPMC include files
[MIPS] Add BUG_ON assertion for attempt to run kernel on the wrong CPU type.
[MIPS] SMP: Use ISO C struct initializer for local structs.
[MIPS] SMP: Kill useless casts.
[MIPS] Kill num_online_cpus() loops.
[MIPS] SMP: Implement smp_call_function_mask().
[MIPS] Make facility to convert CPU types to strings generally available.
[MIPS] Convert list of CPU types from #define to enum.
[MIPS] Optimize get_unaligned / put_unaligned implementations.
[MIPS] checkfiles: Fix "need space after that ','" errors.
[MIPS] Fix "no space between function name and open parenthesis" warnings.
[MIPS] Allow hardwiring of the CPU type to a single type for optimization.
[MIPS] tlbex: Size optimize code by declaring a few functions inline.
[MIPS] pg-r4k.c: Dump the generated code
[MIPS] Cobalt: Remove cobalt_machine_power_off()
[MIPS] Cobalt: Move reset port definition to arch/mips/cobalt/reset.c
...
* git://git.kernel.org/pub/scm/linux/kernel/git/bart/ide-2.6: (24 commits)
ide: use only ->set_pio_mode method for programming PIO modes (take 2)
sis5513: don't change UDMA settings when programming PIO
it8213/piix/slc90e66: don't change DMA settings when programming PIO
alim15x3: PIO mode setup fixes
siimage: fix ->set_pio_mode method to select PIO data transfer
cs5520: don't enable VDMA in ->speedproc
sc1200: remove redundant warning message from sc1200_tune_chipset()
ide-pmac: PIO mode setup fixes (take 3)
icside: fix ->speedproc to return on unsupported modes (take 5)
sgiioc4: use ide_tune_dma()
amd74xx/via82cxxx: use ide_tune_dma()
ide: add ide_set{_max}_pio() (take 4)
ide: Kconfig face-lift
ide: move ide_rate_filter() calls to the upper layer (take 2)
sis5513: add ->udma_filter method for chipset_family >= ATA_133
ide: mode limiting fixes for user requested speed changes
ide: add missing ide_rate_filter() calls to ->speedproc()-s
ide: call udma_filter() before resorting to the UltraDMA mask
ide: make jmicron match vendor and device class
pdc202xx_new: switch to using pci_get_slot() (take 2)
...
Linus Torvalds [Fri, 12 Oct 2007 02:20:16 +0000 (19:20 -0700)]
Merge branch 'for-linus' of git://git.o-hand.com/linux-rpurdie-leds
* 'for-linus' of git://git.o-hand.com/linux-rpurdie-leds:
leds: Update Cobalt Qube series front LED support
leds: Add Cobalt Raq series LEDs support
leds: Rename leds-cobalt driver
Linus Torvalds [Fri, 12 Oct 2007 02:19:50 +0000 (19:19 -0700)]
Merge branch 'for-linus' of git://git.o-hand.com/linux-rpurdie-backlight
* 'for-linus' of git://git.o-hand.com/linux-rpurdie-backlight:
backlight: Convert corgi backlight driver into a more generic driver
backlight: Add Samsung LTV350QV LCD driver
backlight: Fix cr_bllcd allocations and error paths
backlight/leds: Make two structs static
Linus Torvalds [Fri, 12 Oct 2007 02:14:22 +0000 (19:14 -0700)]
Merge branch 'block-2.6.24' of git://git.kernel.dk/data/git/linux-2.6-block
* 'block-2.6.24' of git://git.kernel.dk/data/git/linux-2.6-block: (37 commits)
[BLOCK] Fix failing compile with BLK_DEV_IO_TRACE=n
compat_ioctl: move floppy handlers to block/compat_ioctl.c
compat_ioctl: move cdrom handlers to block/compat_ioctl.c
compat_ioctl: move BLKPG handling to block/compat_ioctl.c
compat_ioctl: move hdio calls to block/compat_ioctl.c
compat_ioctl: handle blk_trace ioctls
compat_ioctl: add compat_blkdev_driver_ioctl()
compat_ioctl: move common block ioctls to compat_blkdev_ioctl
Sysace: Don't enable IRQ until after interrupt handler is registered
Sysace: sparse fixes
Sysace: Minor coding convention fixup
drivers/block/umem: use DRIVER_NAME where appropriate
drivers/block/umem: trim trailing whitespace
drivers/block/umem: minor cleanups
drivers/block/umem: use dev_printk()
drivers/block/umem: move private include away from include/linux
Sysace: Labels in C code should not be indented.
Sysace: Add of_platform_bus binding
Sysace: Move IRQ handler registration to occur after FSM is initialized
Sysace: minor rework and cleanup changes
...
Linus Torvalds [Fri, 12 Oct 2007 02:13:44 +0000 (19:13 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/hskinnemoen/avr32-2.6:
[AVR32] Fix random segfault with preemption
[AVR32] Don't use __builtin_xchg()
[AVR32] ngw100 i2c-gpio tweaks
[AVR32] Ignore a few irrelevant syscalls
[AVR32] SMC configuration in clock cycles
[AVR32] Drop support for redundant "keepinitrd" boot-time parm.
[AVR32] Make dma_sync_*_for_cpu no-ops
[AVR32] Remove unneeded 8K alignment of .text section
[AVR32] Kill a few hardcoded constants in vmlinux.lds
[AVR32] rename vmlinux.lds
[AVR32] fix command line parsing in early_parse_fbmem
[AVR32] checkstack support
[AVR32] Wire up USBA device
[AVR32] add multidrive support for pio driver
[AVR32] /sys/kernel/debug/at32ap_clk
[AVR32] Move AT32_PM_BASE definition into pm.h
Linus Torvalds [Fri, 12 Oct 2007 02:11:51 +0000 (19:11 -0700)]
Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/blackfin-2.6
* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/cooloney/blackfin-2.6: (74 commits)
Blackfin serial driver: pending a unique anomaly id, tie the break flood issue to ANOMALY_05000230
blackfin enable arbitary speed serial setting
Blackfin arch: Remove cruft - CONFIG_DEBUG_SERIAL_EARLY_INIT and DEBUG_KERNEL_START
Blackfin arch: fix typo in register name
Blackfin arch: trim the Blackfin arch MAINTAINERS list
Blackfin arch: fix bug libstdc++ calling writev with an iovec containing { NULL, 0 } fails on Blackfin
Blackfin arch: Export strcpy - occasionally get module link failures otherwise
Blackfin arch: the load address is not safe to point to as a workaround for ANOMALY 05000281
Blackfin arch: show_mem can not be marked as init, since it is called during OOM condition
Blackfin arch: flush/inv the correct range when using write back cache and fix bugs find by dmacopy
Blackfin arch: update kgdb patch
Blackfin arch: Comply with revised Anomaly Workarounds for BF533 05000311 and BF561 05000323
Blackfin arch: Print out debug info, as early as possible
Blackfin arch: Enable earlyprintk earlier - so any error after our interrupt tables are set up will print out
Blackfin arch: fix endless loop bug when a double fault happens
Blackfin arch: Initial patch to add earlyprintk support
Blackfin arch: add TWIx_REGBASE and SPIx_REGBASE to specific CPU header files, use the new REGBASE for board platform resources
Blackfin arch: modify the insX/outsX and dma_insX/dma_outsX to be compatible with other archs
Blackfin arch: add more common defines for output sections
Blackfin arch: cleanup IO and DMA_IO API function definitions according to other arches
...
David S. Miller [Fri, 12 Oct 2007 01:08:29 +0000 (18:08 -0700)]
[NET]: Fix NAPI completion handling in some drivers.
In order for the list handling in net_rx_action() to be
correct, drivers must follow certain rules as stated by
this comment in net_rx_action():
/* Drivers must not modify the NAPI state if they
* consume the entire weight. In such cases this code
* still "owns" the NAPI instance and therefore can
* move the instance around on the list at-will.
*/
A few drivers do not do this because they mix the budget checks
with reading hardware state, resulting in crashes like the one
reported by takano@axe-inc.co.jp.
BNX2 and TG3 are taken care of here, SKY2 fix is from Stephen
Hemminger.
Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Fri, 12 Oct 2007 00:36:13 +0000 (17:36 -0700)]
[TCP]: Limit processing lost_retrans loop to work-to-do cases
This addition of lost_retrans_low to tcp_sock might be
unnecessary, it's not clear how often lost_retrans worker is
executed when there wasn't work to do.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Fri, 12 Oct 2007 00:35:41 +0000 (17:35 -0700)]
[TCP]: Fix lost_retrans loop vs fastpath problems
Detection implemented with lost_retrans must work also when
fastpath is taken, yet most of the queue is skipped including
(very likely) those retransmitted skb's we're interested in.
This problem appeared when the hints got added, which removed
a need to always walk over the whole write queue head.
Therefore decicion for the lost_retrans worker loop entry must
be separated from the sacktag processing more than it was
necessary before.
It turns out to be problematic to optimize the worker loop
very heavily because ack_seqs of skb may have a number of
discontinuity points. Maybe similar approach as currently is
implemented could be attempted but that's becoming more and
more complex because the trend is towards less skb walking
in sacktag marker. Trying a simple work until all rexmitted
skbs heve been processed approach.
Maybe after(highest_sack_end_seq, tp->high_seq) checking is not
sufficiently accurate and causes entry too often in no-work-to-do
cases. Since that's not known, I've separated solution to that
from this patch.
Noticed because of report against a related problem from TAKANO
Ryousei <takano@axe-inc.co.jp>. He also provided a patch to
that part of the problem. This patch includes solution to it
(though this patch has to use somewhat different placement).
TAKANO's description and patch is available here:
...In short, TAKANO's problem is that end_seq the loop is using
not necessarily the largest SACK block's end_seq because the
current ACK may still have higher SACK blocks which are later
by the loop.
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Fri, 12 Oct 2007 00:34:57 +0000 (17:34 -0700)]
[TCP]: No need to re-count fackets_out/sacked_out at RTO
Both sacked_out and fackets_out are directly known from how
parameter. Since fackets_out is accurate, there's no need for
recounting (sacked_out was previously unnecessarily counted
in the loop anyway).
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
projects / linux-2.6 / log