FAQ Search Today's Posts Mark Forums Read
» Video Reviews

» Linux Archive

Linux-archive is a website aiming to archive linux email lists and to make them easily accessible for linux users/developers.


» Sponsor

» Partners

» Sponsor

Go Back   Linux Archive > Debian > Debian Kernel

 
 
LinkBack Thread Tools
 
Old 02-12-2012, 03:51 PM
Ben Hutchings
 
Default Bug#659629: Changes from longterm 2.6.32.57

Package: src:linux-2.6
Version: 2.6.32-41
Severity: important

This is actually based on 2.6.32.57-rc1.

- IB/mlx4: pass SMP vendor-specific attribute MADs to firmware

Fixes handling of some Infiniband protocol extensions that need to be
passed to the device firmware rather than the kernel.

- mm/filemap_xip.c: fix race condition in xip_file_fault()

Not relevant to Debian configurations.

- NFSv4: Fix up the callers of nfs4_state_end_reclaim_reboot

The NFSv4 client was not properly updating some of its state after a
server reboot, presumably resulting in I/O failures and/or memory leaks.
This should fix that.

- NFSv4: The state manager shouldn't exit on errors that were handled
- NFSv4: Ensure the state manager handles NFS4ERR_NO_GRACE correctly

Fix the NFSv4 client to continue running after certain recoverable
errors, which I think would otherwise result in losing access to the
filesystem.

- NFSv4: Handle NFS4ERR_GRACE when recovering an expired lease.

Fixes handling of a server reboot during revalidation of cached state on
the NFSv4 client, which would otherwise result in an I/O failure.

- NFSv4: Fix open recovery

Fixes the NFSv4 client to send new OPEN requests for open files after
the server reboots. Without this, I think the server might allow other
clients to cache files that we have open, resulting in data loss.

- rpc client can not deal with ENOSOCK, so translate it into ENOCONN

Fixes error reporting for RPC requests made during reconnection of the
RPC client socket. The old behaviour apparently could result in NFS I/O
failures during recovery from a server reboot.

- udf: Mark LVID buffer as uptodate before marking it dirty

Fixes a spurious warning when rewriting metadata for this filesystem
after a prior I/O error.

- drm/i915: Fix TV Out refresh rate.

Fixes the display refresh rates for SD TV outputs connected to Intel
GPUs. This doesn't seem to affect the actual signal rate, because the
previous rates would never have worked, but it does apparently fix
flickering.

We won't get this from 2.6.32.57 since we took DRM from 2.6.33, but we
should probably cherry-pick it anyway.

- eCryptfs: Infinite loop due to overflow in ecryptfs_write()

Fixes loop condition for a write more than 4GB beyond the previous end
of a file on this filesystem, on 32-bit systems. This doesn't seem to
be a security issue in itself, but can be combined with the fact that
truncation/extension was not killable (before 2.6.32.56) to produce a
more effective DoS.

- atmel_lcdfb: fix usage of CONTRAST_CTR in suspend/resume
- Staging: asus_oled: fix image processing
- Staging: android: binder: Don't call dump_stack in binder_vma_open
- Staging: android: binder: Fix crashes when sharing a binder file between processes
- usb: gadget: zero: fix bug in loopback autoresume handling

Not relevant to Debian configurations.

- usb: Skip PCI USB quirk handling for Netlogic XLP

Adds a quirk for PCI USB controllers that should have no effect on
Debian-supported systems.

- USB: usbserial: add new PID number (0xa951) to the ftdi driver

New hardware support.

- mmc: cb710 core: Add missing spin_lock_init for irq_lock of struct cb710_chip

Fixes initialisation of a lock structure in this flash card interface
driver. Since we don't enable lock debugging and the structure would be
zero-initialised anyway, I don't believe the bug had any effect on
Debian configurations. But this is obviously correct and harmless.

- net: fix sk_forward_alloc corruptions
- net: sock_queue_err_skb() dont mess with sk_forward_alloc

Removes some accounting of memory used by error/exception packets (ICMP,
local errors and TX timestamps), which was subject to data races. A
remote attacker could provoke a WARNING by sending ICMP packets, but I
don't think any worse effect was possible. Note that memory use by
these packets is still limited by the socket receive buffer size.

Ben.

--
Ben Hutchings
Horngren's Observation:
Among economists, the real world is often a special case.
 

Thread Tools




All times are GMT. The time now is 01:47 PM.

VBulletin, Copyright ©2000 - 2014, Jelsoft Enterprises Ltd.
Content Relevant URLs by vBSEO ©2007, Crawlability, Inc.
Copyright 2007 - 2008, www.linux-archive.org