Al Viro [Sat, 22 Nov 2008 17:33:54 +0000 (17:33 +0000)]
sparc32 cpuinit flase positives
All noise since we don't have CPU hotplug there. However, they
did expose something very odd-looking in there - poke_viking()
does a bunch of identical btfixup each time it's called (i.e.
for each CPU). That one is left alone for now; just the trivial
misannotation fixes.
Signed-off-by: Al Viro <viro@zeniv.linux.org.uk> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Russell King [Sun, 30 Nov 2008 17:38:14 +0000 (17:38 +0000)]
[ARM] amba drivers: don't pass a consumer clock name for devices with unique clocks
Where devices only have one consumer, passing a consumer clock ID
has no real benefit, and it only encourages wrong implementations of
the clk API. Remove it.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King [Sun, 30 Nov 2008 11:45:54 +0000 (11:45 +0000)]
[ARM] Add a common typesafe __io implementation
As Al did for Versatile in 2ad4f86b60b649fd7428265c08d73a3bd360c81b,
add a typesafe __io implementation for platforms to use. Convert
platforms to use this new simple typesafe implementation.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Marcel Holtmann [Sun, 30 Nov 2008 11:17:29 +0000 (12:17 +0100)]
Bluetooth: Fix RFCOMM release oops when device is still in use
It turns out that the following sequence of actions will reproduce the
oops:
1. Create a new RFCOMM device (using RFCOMMCREATEDEV ioctl)
2. (Try to) open the device
3. Release the RFCOMM device (using RFCOMMRELEASEDEV ioctl)
At this point, the "/dev/rfcomm*" device is still in use, but it is gone
from the internal list, so the device id can be reused.
4. Create a new RFCOMM device with the same device id as before
And now kobject will complain that the TTY already exists.
(See http://lkml.org/lkml/2008/7/13/89 for a reproducible test-case.)
This patch attempts to correct this by only removing the device from the
internal list of devices at the final unregister stage, so that the id
won't get reused until the device has been completely destructed.
This should be safe as the RFCOMM_TTY_RELEASED bit will be set for the
device and prevent the device from being reopened after it has been
released.
Based on a report from Vegard Nossum <vegard.nossum@gmail.com>
Marcel Holtmann [Sun, 30 Nov 2008 11:17:29 +0000 (12:17 +0100)]
Bluetooth: Fix format arguments warning
Newer GCC versions are a little bit picky about how to deal with format
arguments:
net/bluetooth/hci_sysfs.c: In function ‘hci_register_sysfs’:
net/bluetooth/hci_sysfs.c:418: warning: format not a string literal and no format arguments
It is simple enough to fix and makes the compiler happy.
With the introduction of CONFIG_DYNAMIC_PRINTK_DEBUG it is possible to
allow debugging without having to recompile the kernel. This patch turns
all BT_DBG() calls into pr_debug() to support dynamic debug messages.
As a side effect all CONFIG_BT_*_DEBUG statements are now removed and
some broken debug entries have been fixed.
Marcel Holtmann [Sun, 30 Nov 2008 11:17:26 +0000 (12:17 +0100)]
Bluetooth: Send HCI Reset command by default on device initialization
The Bluetooth subsystem was not using the HCI Reset command when doing
device initialization. The Bluetooth 1.0b specification was ambiguous
on how the device firmware was suppose to handle it. Almost every device
was triggering a transport reset at the same time. In case of USB this
ended up in disconnects from the bus.
All modern Bluetooth dongles handle this perfectly fine and a lot of
them actually require that HCI Reset is sent. If not then they are
either stuck in their HID Proxy mode or their internal structures for
inquiry and paging are not correctly setup.
To handle old and new devices smoothly the Bluetooth subsystem contains
a quirk to force the HCI Reset on initialization. However maintaining
such a quirk becomes more and more complicated. This patch turns the
logic around and lets the old devices disable the HCI Reset command.
The only device where the HCI_QUIRK_NO_RESET is still needed are the
original Digianswer devices and dongles with an early CSR firmware.
CSR reported that they fixed this for version 12 firmware. The last
official release of version 11 firmware is build ID 115. The first
version 12 candidate was build ID 117.
Marcel Holtmann [Sun, 30 Nov 2008 11:17:21 +0000 (12:17 +0100)]
Bluetooth: Remove deprecated hci_usb driver
The old hci_usb driver has been fully replaced with the new btusb driver
and all major distributions switched to the new driver now. This removes
it since it should not be used at all anymore.
Marcel Holtmann [Sun, 30 Nov 2008 11:17:19 +0000 (12:17 +0100)]
Bluetooth: Fix warnings for bt_key_strings and bt_slock_key_strings
After adding proper lockdep annotations for Bluetooth protocols the case
when lockdep is disabled produced two compiler warnings:
net/bluetooth/af_bluetooth.c:60: warning: ‘bt_key_strings’ defined but not used
net/bluetooth/af_bluetooth.c:71: warning: ‘bt_slock_key_strings’ defined but not used
Fix both of them by adding a CONFIG_DEBUG_LOCK_ALLOC conditional around
them and re-arranging the code a little bit.
sizeof(struct hci_dev_list_req) == 4, so the two bytes immediately
following "dev_num" will never be initialized. When this structure
is copied to userspace, these uninitialized bytes are leaked.
Fix by using kzalloc() instead of kmalloc(). Found using kmemcheck.
Marcel Holtmann [Sun, 30 Nov 2008 11:17:12 +0000 (12:17 +0100)]
Bluetooth: Handle bulk URBs in btusb driver from notify callback
With the addition of usb_unlink_anchored_urbs() it is possible to fully
control the bulk URBs from the notify callback. There is no need to
schedule work and so only do this for the ISOC URBs.
Marcel Holtmann [Sun, 30 Nov 2008 11:17:10 +0000 (12:17 +0100)]
Bluetooth: Add fine grained mem_flags usage to btusb driver
The URB submission routines need more fine grained control for the
mem_flags used by kmalloc(), usb_alloc_urb() and usb_submit_urb() to
better support different caller situations. Add a mem_flags parameter
and give the caller full control.
ALSA: hda - Add a new function to seek for a codec ID
Gateway notebooks have their ID inside codec vendor ID, not at PCI ID. Due to
that, model auto-detection were not possible with the standard seek method.
Ingo Molnar [Sun, 30 Nov 2008 09:20:20 +0000 (10:20 +0100)]
x86: provide readq()/writeq() on 32-bit too, complete
if HAVE_READQ/HAVE_WRITEQ are defined, the full range of readq/writeq
APIs has to be provided to drivers:
drivers/infiniband/hw/amso1100/c2.c: In function 'c2_tx_ring_alloc':
drivers/infiniband/hw/amso1100/c2.c:133: error: implicit declaration of function '__raw_writeq'
So provide them on 32-bit as well. Also, map all the APIs to the
strongest ordering variant. It's way too easy to mess such details
up in drivers and the difference between "memory" and "" constrained
asm() constructs is in the noise range.
Tilman Schmidt [Sun, 30 Nov 2008 05:38:28 +0000 (21:38 -0800)]
gigaset: get rid of info() and warn() macros
Join the move away from the obsolete info() macro, opencoding the
remaining uses. While we're at it, also get rid of the warn() macro
by promoting the three remaining uses to err().
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net>
Tilman Schmidt [Sun, 30 Nov 2008 05:38:04 +0000 (21:38 -0800)]
gigaset: remove unnecessary poll method
The N_GIGASET_M101 line discipline implemented by the ser_gigaset
driver does not transfer any data from/to userspace through the
tty interface. Therefore a poll method is not needed.
Signed-off-by: Tilman Schmidt <tilman@imap.cc> Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/acpi/ec.c: In function 'acpi_ec_ecdt_probe':
drivers/acpi/ec.c:1015: warning: ISO C90 forbids mixed declarations and code
Signed-off-by: Hannes Eder <hannes@hanneseder.net> Acked-by: Alexey Starikovskiy <astarikovskiy@suse.de> Signed-off-by: Len Brown <len.brown@intel.com>
Devices like b44 ethernet can't dma from addresses above 1GB. The driver
handles this cases by falling back to GFP_DMA allocation. But for detecting
the problem it needs to get an indication from dma_mapping_error.
The bug is triggered by using a VMSPLIT option of 2G/2G.
Ingo Molnar [Sat, 29 Nov 2008 19:45:15 +0000 (20:45 +0100)]
sched: prevent divide by zero error in cpu_avg_load_per_task, update
Regarding the bug addressed in:
4cd4262: sched: prevent divide by zero error in cpu_avg_load_per_task
Linus points out that the fix is not complete:
> There's nothing that keeps gcc from deciding not to reload
> rq->nr_running.
>
> Of course, in _practice_, I don't think gcc ever will (if it decides
> that it will spill, gcc is likely going to decide that it will
> literally spill the local variable to the stack rather than decide to
> reload off the pointer), but it's a valid compiler optimization, and
> it even has a name (rematerialization).
>
> So I suspect that your patch does fix the bug, but it still leaves the
> fairly unlikely _potential_ for it to re-appear at some point.
>
> We have ACCESS_ONCE() as a macro to guarantee that the compiler
> doesn't rematerialize a pointer access. That also would clarify
> the fact that we access something unsafe outside a lock.
So make sure our nr_running value is immutable and cannot change
after we check it for nonzero.
Ingo Molnar [Tue, 25 Nov 2008 09:27:49 +0000 (10:27 +0100)]
sched, cpusets: fix warning in kernel/cpuset.c
this warning:
kernel/cpuset.c: In function ‘generate_sched_domains’:
kernel/cpuset.c:588: warning: ‘ndoms’ may be used uninitialized in this function
triggers because GCC does not recognize that ndoms stays uninitialized
only if doms is NULL - but that flow is covered at the end of
generate_sched_domains().
Help out GCC by initializing this variable to 0. (that's prudent anyway)
Also, this function needs a splitup and code flow simplification:
with 160 lines length it's clearly too long.
Russell King [Sat, 29 Nov 2008 19:14:31 +0000 (19:14 +0000)]
[ARM] give RiscPC a NR_IRQS definition and remove default
RiscPC is the only platform using the default setting for NR_IRQS,
so the default NR_IRQS doesn't really make sense; remove it and
make RiscPC provide such a definition.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King [Sat, 29 Nov 2008 11:40:28 +0000 (11:40 +0000)]
[ARM] Hide ISA DMA API when ISA_DMA_API is unset
When ISA_DMA_API is unset, we're not implementing the ISA DMA API,
so there's no point in publishing the prototypes via asm/dma.h, nor
including the machine dependent parts of that API.
This allows us to remove a lot of mach/dma.h files which don't contain
any useful code. Unfortunately though, some platforms put their own
private non-ISA definitions into mach/dma.h, so we leave these behind
and fix the appropriate #include statments.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Russell King [Sat, 29 Nov 2008 10:56:33 +0000 (10:56 +0000)]
[ARM] fix constant values of MAX_DMA_ADDRESS
Since 8d5796d2ec6b5a4e7a52861144e63af438d6f8f7, we have allowed
PAGE_OFFSET to be configurable, so a constant virtual address
for MAX_DMA_ADDRESS is buggy. It should be defined in terms of
PAGE_OFFSET rather than a constant virtual address.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Stefan Richter [Wed, 26 Nov 2008 00:34:25 +0000 (01:34 +0100)]
ieee1394: fix list corruption (reported at module removal)
If there is more than one FireWire controller present, dummy_zero_addr
and dummy_max_addr were added multiple times to different lists, thus
corrupting the lists. Fix this by allocating them dynamically per host
instead of just once globally.
(Perhaps a better address space allocation algorithm could rid us of the
two dummy address spaces.)
This should fix a build error reported on alpha in linux-next:
CC kernel/trace/trace_branch.o
kernel/trace/trace_branch.c: In function 'probe_likely_condition':
kernel/trace/trace_branch.c:44: error: implicit declaration of function 'raw_local_irq_save'
kernel/trace/trace_branch.c:76: error: implicit declaration of function 'raw_local_irq_restore'
Unfortunately, I can't test it since I don't have any Alpha build environment.
Roland Dreier [Sat, 29 Nov 2008 05:55:42 +0000 (21:55 -0800)]
cxgb3: Fix sparse warning and micro-optimize is_pure_response()
The function is_pure_response() does "ntohl(var) & const" and then
essentially just tests whether the result is 0 or not; this can be done
more efficiently by computing "var & htonl(const)" instead and doing the
byte swap at compile time instead of run time.
This change slightly shrinks the compiled code; eg on x86-64 we save a
couple of bswapl instructions:
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-8 (-8)
function old new delta
t3_sge_intr_msix_napi 544 536 -8
and this also has the pleasant side effect of fixing a sparse warning:
drivers/net/cxgb3/sge.c:2313:15: warning: restricted degrades to integer
Signed-off-by: Roland Dreier <rolandd@cisco.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Jack Morgenstein [Sat, 29 Nov 2008 05:29:46 +0000 (21:29 -0800)]
mlx4_core: Save/restore default port IB capability mask
Commit 7ff93f8b ("mlx4_core: Multiple port type support") introduced
support for different port types. As part of that support, SET_PORT
is invoked to set the port type during driver startup. However, as a
side-effect, for IB ports the invocation of this command also sets the
port's capability mask to zero (losing the default value set by FW).
To fix this, get the default ib port capabilities (via a MAD_IFC Port
Info query) during driver startup, and save them for use in the
mlx4_SET_PORT command when setting the port-type to Infiniband.
This patch fixes problems with subnet manager (SM) failover such as
<https://bugs.openfabrics.org/show_bug.cgi?id=1183>, which occurred
because the IsTrapSupported bit in the capability mask was zeroed.
Signed-off-by: Jack Morgenstein <jackm@dev.mellanox.co.il> Signed-off-by: Roland Dreier <rolandd@cisco.com>
This patch adds the power management support into the physical
abstraction layer.
Suspend and resume functions respectively turns on/off the bit 11
into the PHY Basic mode control register.
Generic PHY device starts supporting PM.
In order to support the wake-on LAN and avoid to put in power down
the PHY device, the MDIO is aware of what the Ethernet device wants to do.
Voluntary, no CONFIG_PM defines were added into the sources.
Also generic suspend/resume functions are exported to allow
other drivers use them (such as genphy_config_aneg etc.).
Within the phy_driver_register function, we need to remove the
memset. It overrides the device driver owner and it is not good.
Signed-off-by: Giuseppe Cavallaro <peppe.cavallaro@st.com> Signed-off-by: David S. Miller <davem@davemloft.net>
Ilpo Järvinen [Fri, 28 Nov 2008 23:55:00 +0000 (15:55 -0800)]
ne2: silence static never defined warnings
These warning originate from 50014f1 (ne2: convert to
net_device_ops) which did drop those functions.
drivers/net/ne2.c:140: warning: 'ne_open' declared 'static' but never defined
drivers/net/ne2.c:141: warning: 'ne_close' declared 'static' but never defined
Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@helsinki.fi> Signed-off-by: David S. Miller <davem@davemloft.net>
Russell King [Fri, 28 Nov 2008 16:13:44 +0000 (16:13 +0000)]
[ARM] iop3xx: avoid polluting the kernel's namespace
Avoid unnecessarily pollution of the kernel's namespace by avoiding
mach/hardware.h in mach/io.h, mach/memory.h and mach/timex.h. Include
this header file where necessary.
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
Arjan van de Ven [Fri, 28 Nov 2008 16:19:09 +0000 (08:19 -0800)]
toshiba_acpi: close race in toshiba_acpi driver
the toshiba ACPI driver will, in a failure case, free the rfkill state
before stopping the polling timer that would use this state. More interesting,
in the same failure case handling, it calls the exit function, which also
frees the rfkill state, but after stopping the polling.
If the race happens, a NULL pointer is passed to rfkill_force_state()
which then causes a nice dereference.
Fix the race by just not doing the too-early freeing of the rfkill state.
This appears to be the cause of a hot issue on kerneloops.org; while I
have no solid evidence of that this patch will fix the issue, the race
appears rather real.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Len Brown <len.brown@intel.com>
Alexey Dobriyan [Fri, 28 Nov 2008 19:08:00 +0000 (22:08 +0300)]
sched: move double_unlock_balance() higher
Move double_lock_balance()/double_unlock_balance() higher to fix the following
with gcc-3.4.6:
CC kernel/sched.o
In file included from kernel/sched.c:1605:
kernel/sched_rt.c: In function `find_lock_lowest_rq':
kernel/sched_rt.c:914: sorry, unimplemented: inlining failed in call to 'double_unlock_balance': function body not available
kernel/sched_rt.c:1077: sorry, unimplemented: called from here
make[2]: *** [kernel/sched.o] Error 1
Arjan van de Ven [Fri, 28 Nov 2008 16:36:09 +0000 (08:36 -0800)]
debug warnings: print the DMI board info name in a WARN/WARN_ON
Impact: extend WARN_ON() output with DMI_PRODUCT_NAME
It's very useful for many low level WARN_ON's to find out which
motherboard has the broken BIOS etc... this patch adds a printk
to the WARN_ON code for this.
On architectures without DMI, gcc should optimize the code out.
Signed-off-by: Arjan van de Ven <arjan@linux.intel.com> Signed-off-by: Ingo Molnar <mingo@elte.hu>
projects / linux-2.6-omap-h63xx.git / log