[Karmic] SRU: Update to stable kernel

Andy Whitcroft apw at canonical.com
Fri Nov 6 11:39:37 UTC 2009

On Tue, Nov 03, 2009 at 08:24:09PM -0800, Leann Ogasawara wrote:

> > > The Upstream Stable Kernel brings in the following 47 patches:

Overall these looks sane.  There are two which are not obviously required,
the maintaners update and the pax update as neither contributes to any
of our builds; that said we lose nothing by taking them and staying in
sync with stable at this stage is likely worthwhile.

The most worrying appear to be the almost complete rewrite of USB ftdi_sio,
but this is only for one device manufacturer and seems to fix a number
of regressions which would render the device unsusable already.

So overall, details below:

Acked-by: Andy Whitcroft <apw at canonical.com>


UBUNTU: [Config] Move position of CONFIG_LIRC_ENE0100 for ports configs

  Benign change, not sure how the config system ended up with this out
  of order in the first place.  It would also trigger build issues if
  broken so is very low risk.


  Benign.  Sensible for tracking.

tty: Make flush_to_ldisc() locking more robust

  A large looking patch but in essence relativly simple moving to atomic
  test_and_set to ensure only one thread can set and unset TTY_FLUSHING.

mac80211: fix vlan and optimise RX

  Patch sounds like it does a sane thing, fixinf VLAN handling which
  would affect server installs.

iwlwifi: incorrect method used for finding valid OTP blocks

  This one looks to do what it suggests.  Seems reasonable applying to
  iwl which is a common driver for working h/w.

usb-storage: Workaround devices with bogus sense size
Revert "UBUNTU: [Upstream] (drop after 2.6.31) usb-storage: Workaround d

  These simply replace out existing commit with the official version.
  They appear identicle.

Staging: rt2860sta: prevent a panic when disabling when associated

  Looks obviously correct.

sysfs: Allow sysfs_notify_dirent to be called from interrupt context.

  Fixes locking, looks sane.

bsdacct: switch credentials for writing to the accounting file

  Looks reasonable.

ALSA: Don't assume i2c device probing always succeeds

  Simple check for no device found, propogating the error.  Looks

i2c: Hide probe errors caused by ACPI resource conflicts

  Reduces pointless errors to users while retaining the main error.
  Seems reasonable.

macintosh: Don't assume i2c device probing always succeeds

  This patch moves the linkage between clients to a new head, which will
  avoid the panic.  Confusingly it is using a predeclaration of the
  head pointers, but it is not at all obvious in C that it is a
  predeclaration.  Looks ok when applied.

MAINTAINERS: Fix Riku Voipio's address

  Trivial maintainers change.

e1000e: swap max hw supported frame size between 82574 and 82583
Revert "e1000e: swap max hw supported frame size between 82574 and 82583

  Again just an update to the mainline version of an early fix.

uvesafb/connector: Disallow unpliviged users to send netlink packets
pohmelfs/connector: Disallow unpliviged users to configure pohmelfs
dst/connector: Disallow unpliviged users to configure dst
dm/connector: Only process connector packages from privileged processes

  All tighten up permissions on connectors.  Seems like a sensible
  security fix.

connector: Removed the destruct_data callback since it is always kfree_s
connector: Provide the sender's credentials to the callback
connector: Keep the skb in cn_callback_data

  Look to be pre-requisites for the security fixes.

Bluetooth: Let HIDP grab the device reference for connections
Bluetooth: Add extra device reference counting for connections

  Cleans up device discovery events to userspace.  Seems sensible.

Bluetooth: Disconnect HIDRAW devices on disconnect

  Ensures HIDRAW devices go away.  Without this connect/disconnect
  cycles cause these to build up and overflow X's input list.  Seems

TPM: fix pcrread

  Fixes TPM size handling.  Looks reasonable.

futex: Fix wakeup race by setting TASK_INTERRUPTIBLE before queue_me()

  Fixes userspace hangs with futexes.  Looks sane.

futex: Detect mismatched requeue targets

  Fixes suprious wakeup queue triggers.  Looks sane.

intel-iommu: Cope with broken HP DC7900 BIOS

  Handles specific broken BIOS, also fixes error handling.  Looks sane.

arm, cris, mips, sparc, powerpc, um, xtensa: fix build with bash 4.0

  All build system fixes.  Should be easy to test and fixes things on
  arm.  Seems sensible.

USB: storage: When a device returns no sense data, call it a Hardware Er

  Fixes flat spin retrying sense on bad devices, reporting a sensible
  error into the bargain.  Looks good.

USB: Fix throttling in generic usbserial driver

  Fixes lost packets to slow devices.  Seems sane.

USB: cp210x: Add support for the DW700 UART

  Simple device quirk.

USB: ipaq: fix oops when device is plugged in

  Prevents oops on insert for ipaq.  Looks sane.

USB: option: Toshiba G450 device id

  Simple device quirk.

USB: serial: don't call release without attach

  Prevents OOPS on insertion of some sick devices.  Looks sane.

USB: digi_acceleport: Fix broken unthrottle.

  Fixes regression.  Looks sane.

USB: pl2303: fix error characters not being reported to ldisc

  Fixes regression.  Looks sane.

USB: ftdi_sio: re-implement read processing
USB: ftdi_sio: clean up read completion handler
USB: ftdi_sio: remove unused rx_byte counter
USB: ftdi_sio: remove tty->low_latency

  A pretty large update to fix up a number of regressions.  This is a
  single SIO provider so low impact.  Likely worth the risk.

x86/paravirt: Use normal calling sequences for irq enable/disable

  Prevents boot crash with stack-protector, which we have enabled.
  Looks sensible.

tracing/filters: Fix memory leak when setting a filter

  Memory leak using tracing.  Looks sane.

ARM: pxa: workaround errata #37 by not using half turbo switching

  pxa specific.  Simple quirk, no risk.

SCSI: sg: Free data buffers after calling blk_rq_unmap_user

  Corrects free sequence.  Looks sane.

SCSI: Fix protection scsi_data_buffer leak

  Memory leak.  Looks sane.

SCSI: Retry ADD_TO_MLQUEUE return value for EH commands

  Ensure we retry EH commands correctly.  Looks sane.

More information about the kernel-team mailing list