Package: src:linux-2.6
Version: 2.6.32-39
Severity: important

[Actually based on 2.6.32.50-rc1.]

i2c-algo-bit: Generate correct i2c address sequence for 10-bit target

Fixes addressing of some I2C devices on a bit-banged bus, which would
otherwise be inacessible. Hardware support, I suppose.

eCryptfs: Extend array bounds for all filename chars

Fixes trivial information leak (not user data but constant data from the
module).

PCI hotplug: shpchp: don't blindly claim non-AMD 0x7450 device IDs

Fixes device matching for this driver, part of our bug #638863.

ARM: 7161/1: errata: no automatic store buffer drain

Has no effect on Debian configurations.

ALSA: lx6464es - fix device communication via command bus

This is apparently not necessary and I expect this will be dropped.

SUNRPC: Ensure we return EAGAIN in xs_nospace if congestion is cleared

Fixes spurious SunRPC communication failures, which I assume affect NFS
reliability.

timekeeping: add arch_offset hook to ktime_get functions

Has no effect on Debian release architectures, but apparently fixes
reporting of the monotonic clock time on alpha and m68k.

p54spi: Add missing spin_lock_init
p54spi: Fix workqueue deadlock

Fix deadlocks in this wireless network driver.

nl80211: fix MAC address validation

I think this fixes an information leak or (unlikely) local DoS
exploitable with CAP_NET_ADMIN, when using a wireless network device.

gro: reset vlan_tci on reuse

This should fix CVE-2011-1576. We already have a different fix.

staging: usbip: bugfix for deadlock

Fixes a potential deadlock when using this protocol for remote USB.

staging: comedi: fix oops for USB DAQ devices.

Fixes crash if one of these devices is unplugged while in use.

Staging: comedi: fix signal handling in read and write

Fixes a bug that would appear to delay killing a process using COMEDI
(data acquisitoon) devices. The bug may have worse effects; I'm not
sure but the fix is pretty obvious.

USB: whci-hcd: fix endian conversion in qset_clear()

Fixes operation of Wireless USB on big-endian architectures.

usb: ftdi_sio: add PID for Propox ISPcable III
usb: option: add SIMCom SIM5218
USB: usb-storage: unusual_devs entry for Kingston DT 101 G2

More new hardware support.

SCSI: scsi_lib: fix potential NULL dereference

Mitigation for kernel bugs that result in an invalid SCSI request. Such
a bug should result in an assertion failure with a meaningful log
message, rather than an 'oops'.

SCSI: Silencing 'killing requests for dead queue'

Replaces a dubious informational message with more accurate messages.

cifs: fix cifs stable patch cifs-fix-oplock-break-handling-try-2.patch

This adds a missing piece of a previous fix made in 2.6.32.32. These
fixes relate to 'oplock break' messages, by which the CIFS server tells
us to stop caching a file that another client is accessing. It looks
like the change in 2.6.32.32 could disable read caching when the server
only tells us to disable write caching, and this corrects that.

sched, x86: Avoid unnecessary overflow in sched_clock

Fixes crash after the system has been running for 2^54 ns (208.5 days).

x86/mpparse: Account for bus types other than ISA and PCI

Fixes crash at boot on some systems with EISA (our bug #586494).

oprofile, x86: Fix crash when unloading module (nmi timer mode)

The description seems clear enough.

genirq: Fix race condition when stopping the irq thread

Some IRQ handlers are configured to run in threads. This fixes a bug
that can cause such a thread to hang rather than exiting when the
handler is removed.

tick-broadcast: Stop active broadcast device when replacing it
clockevents: Set noop handler in clockevents_exchange_device()

Tick broadcast is a generic facility for waking up processors that have
their own interrupt timers but on which the timers stop when they enter
a low-power state. If there are multiple clock devices suitable for
this purpose then they might both be enabled, when only one should be.
I'm not sure quite what the result of that would be, but it's certainly
not good. These changes fix that bug.

Ben.

--
Ben Hutchings
In a hierarchy, every employee tends to rise to his level of incompetence.