Bug#663534: Changes from longterm 2.6.32.59
Package: src:linux-2.6
Version: 2.6.32-41
Severity: important
This is based on 2.6.32.59-rc1.
- compat: Re-add missing asm/compat.h include to fix compile breakage on s390
- [IA64] Remove COMPAT_IA32 support
Fix build breakages in 2.6.32.58 due to is_compat_task() not being
defined in all cases. The first is for s390 without CONFIG_COMPAT,
which is not a configuration we use. The second is for ia64 with
CONFIG_COMPAT, which we do use. It disables CONFIG_IA32_SUPPORT and
CONFIG_COMPAT, which are known to be broken. Unfortunately this
breaks ABI compatibility, so we will have to do things differently
(already committed).
- writeback: fixups for !dirty_writeback_centisecs
Fixes some cases where if vm.dirty_writeback_centisecs=0 the kernel
will spin in (effectively) an infinite loop.
- bsg: fix sysfs link remove warning
Fixes WARNING on removal of some SCSI-generic devices. No real harm
is done, so far as I know.
- eCryptfs: Handle failed metadata read in lookup
In some odd cases, ecryptfs may return the lower filesystem's file
size (including padding) rather than the plaintext file size. When
programs read the entire file they will see some padding. The file
is likely to be corrupted if it is then rewritten.
- [S390] KEYS: Enable the compat keyctl wrapper on s390x
Add 32-bit compatibility for the keyctl system call on s390x. Since
we currently only support 32-bit s390 userland on 64-bit kernel, this
has evidently been completely broken for our users!
- cifs: fix dentry refcount leak when opening a FIFO on lookup
This network filesystem will leak memory and then crash (oops) when
unmounted if requested to open a file that is a special file on the
server. This is obviously exploitable by a file server for denial
of service.
- mac80211: zero initialize count field in ieee80211_tx_rate
Belt-and-braces fix for #663198. I think it's bogus for 2.6.32.y.
- net/usbnet: avoid recursive locking in usbnet_stop()
Fixes potential deadlock when setting a USB networking interface down.
- regset: Prevent null pointer reference on readonly regsets
The regset interface extends ptrace() and core dumps to more registers
and pseudo-registers. All such registers should be read-write, but
the 'IOPERM32' and 'IOPERM64' pseudo-registers for x86 are read-only.
Attempting to write them through ptrace() results in a crash (oops).
This is CVE-2012-1097.
- regset: Return -EFAULT, not -EIO, on host-side memory fault
Cleanup for regset; not sure why it's here but it's harmless.
- watchdog: hpwdt: clean up set_memory_x call for 32 bit
This driver was changed in 2.6.32.53 to ensure that the BIOS code it
depends on is executable. However, it did this wrongly, resulting in
incorrect changes to kernel memory mappings.
Ben.
-- System Information:
Debian Release: wheezy/sid
APT prefers stable-updates
APT policy: (500, 'stable-updates'), (500, 'proposed-updates'), (500, 'unstable'), (500, 'stable'), (1, 'experimental')
Architecture: i386 (x86_64)
Kernel: Linux 3.2.0-2-amd64 (SMP w/2 CPU cores)
Locale: LANG=en_GB.utf8, LC_CTYPE=en_GB.utf8 (charmap=UTF-8)
Shell: /bin/sh linked to /bin/dash
Reply to: