Download/kernel/2.6.16/026test012.1/changes

From OpenVZ Virtuozzo Containers Wiki
< Download‎ | kernel‎ | 2.6.16‎ | 026test012.1
Revision as of 10:13, 21 March 2008 by Kir (talk | contribs) (created)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Changes

  • CPU Fair scheduler.
  • Mainstream security fixes (up to 2.6.16.15).
  • Compilation fixes.
  • UBC accounting fixes.

Config changes

Same as 026test009.1, plus:
Added:

  • +CONFIG_FAIRSCHED=y
  • +CONFIG_SCHED_VCPU=y
  • +CONFIG_INPUT_EVDEV=y

Removed:

  • -CONFIG_DETECT_SOFTLOCKUP (was y)

For the complete list of changes in this release, see git changelog for kernel 026test012.1.

Patches

diff-cpt-comp-errno-fix-20060424

Patch from Dmitry Mishin <dim@openvz.org>:
Fixed cpt compilation on ia32 with gcc 4.0.2

diff-cpt-comp-fix-20060424

Patch from Dmitry Mishin <dim@openvz.org>:
Fixed compilation for CONFIG_VZ_CHECKPOINT=y

  • static eventpoll_ops
  • strut for __put_namespace
  • common files for cpt and rst modules
  • emt64 kernel thread creation

diff-cpt-comp-htlb-ipv6-fix-20060426

Patch from Dmitry Mishin <dim@openvz.org>:
[CPT] Fixed compilation with CONFIG_HUGETLB and CONFIG_IPV6 enabled.

diff-fairsched-20060502

Patch from Kirill Korotaev <dev@openvz.org>:
Fair cpu scheduler

diff-fairsched-init-boot-20060510

Patch from Pavel Emelianov <xemul@openvz.org>:
Initialize vcpus for idle_vsched and default_vsched in migration_call

Normally this should be done after CPU_UP_PREPARE event, but some arches (i386) call init_idle() before it, so init vcpus there as well.

This patch fixes boot process for x86_64 smp kernel.

diff-fairsched-ppc-task-cpu-20060502

Patch from Kirill Korotaev <dev@openvz.org>:
Fix usage of task_cpu on powerpc.

diff-fairsched-up-comp-20060502

Patch from Pavel Emelianov <xemul@openvz.org>:
Fixed fairsched compilation for UP

diff-fairsched-vcpu-del-fix-20060503

Patch from Pavel Emelianov <xemul@openvz.org>:
Don't migrate migration thread from dying vcpu on VPS stop.

Otherwise on destroying the last cpu destination will be NR_CPUS cpu and it will cause an OOPs in __migrate_task() (#61877).

diff-fairsched-yield-20060502

Patch from Kirill Korotaev <dev@openvz.org>:
Fix vsched_del_vcpu to work under lock_kernel()

vzioctls are called under lock_kernel() as all other ioctl's. This lead to the fact that scheduler code can do cpu_relax() with lock_kernel() being held, which isn't good anyhow. I replaced one such non-fatal cpu_relax() with yield(). It is not fatal since vcpu has no more tasks and should be switched from at this moment.

diff-faudit-statfs-fix-20060421

Patch from Pavel Emelianov <xemul@openvz.org>:
Compat layer in 64bit systems didn't call faudit on statfs calls.

Now faudit uses struct kstatfs for patching statfs data and compat layer uses it.

OpenVZ Bug #142.

diff-lockkernel-20060502

Patch from Kirill Korotaev <dev@openvz.org>:
Remove lock_kernel() from vzioctls, since nothing in our code requires it...

diff-merge-2.6.16.13-20050504

Patch from OpenVZ team <devel@openvz.org>:
Merged 2.6.16.13 from /linux/kernel/git/stable/linux-2.6.16.y

diff-merge-2.6.16.15-20060510

Patch from OpenVZ team <devel@openvz.org>:
Merged 2.6.16.15 from /linux/kernel/git/stable/linux-2.6.16.y

diff-merge-2.6.16.9-20060424

Patch from OpenVZ team <devel@openvz.org>:
Merge 2.6.16.9 from /linux/kernel/git/stable/linux-2.6.16.y

diff-oops-decode-fix

Patch from Pavel Emelianov <xemul@openvz.org>:
Some fixes for disabling of decoding calltraces

  • some places were missed in emt64
  • make it "on" by default

diff-sched-stats-20060502

Patch from Pavel Emelianov <xemul@openvz.org>:
Scheduler tasks statistics needed with fairsched.

Adds percpu counters for number of running, sleeping, stopped etc tasks and updates them in proper places. nr_xxx() functions are patched to use them.

Patch is needed for fairsched to calculate these numbers correctly.

diff-ubc-cache-counter-free-20060503

Patch from Pavel Emelianov <xemul@openvz.org>:
Free ubc cache counters on cache destroy.

diff-ubc-mremap-fix-20060503

Patch from Pavel Emelianov <xemul@openvz.org>:
Charge correct size in sys_mremap.

Since only (new_len - old_len) region is added to address space this very size should be charged, not the whole new_len.

OpenVZ Bug #141.

diff-ubc-pbc-comp-fix-20060422

Patch from Pavel Emelianov <xemul@openvz.org>:
Compilation without USER_RSS_ACCOUNTING fix.

ub_init_pbc() is in kernel/ub/ub_page_bc.c which is compiled under CONFIG_USER_RSS_ACCOUNTING, not CONFIG_USER_RESOURCE

diff-ubc-vmrss-fix-20060421

Patch from Pavel Emelianov <xemul@openvz.org>:

Handle of mapping (with vma merging) of /dev/zero

  • lost ub_unused_privvm_dec in zeromap_pte_range caused "ub helds xxx in unused_privvmpages on put" message
  • do_mmap_pgoff can perform such a thing:
# allocate a new VMA
# populate it with pages via file->f_ops->mmap
# try to merge this VMA with vma_merge
# free VMA without inserting it in to list
# as the result the expanded with vma_merge vm_area doesn't account populated rss. So vma_rss counter must be tuned in do_mmap_pgoff.

diff-ubc-xip-include-20060428

Patch from Pavel Emelianov <xemul@openvz.org>:
Add missed include in mm/filemap_xip.c

Gentoo bug #131250

diff-ve-higetlb-fix-20060506

Patch from Pavel Emelianov <xemul@openvz.org>:
Fix hugetlb compilation for CONFIG_VE=n case

diff-ve-kstat-fix-20060424

Patch from Pavel Emelianov <xemul@openvz.org>:
Fix mm/vmscan.c compilation

 CC      mm/vmscan.o
 mm/vmscan.c: In function `refill_inactive_zone':
 mm/vmscan.c:1246: warning: ISO C90 forbids mixed declarations and code

OpenVZ Bug #146.

diff-ve-netdev-comp-fix-20060424

Patch from Dmitry Mishin <dim@openvz.org>:
Fixed compilation for CONFIG_VE_NETDEV=y

diff-venet-resetsaddr-20060420

Patch from Denis Lunev <den@openvz.org>:

This patch sets skb destination address as a source in ip_send_reply (TCP reset)

diff-ve-nf-comp-fix-20060424

Patch from Dmitry Mishin <dim@openvz.org>:
Fix compilation for CONFIG_VE_IPTABLES=y

diff-ve-sched-stat-iowait-20060417

Patch from Dmitry Mishin <dim@openvz.org>:

Per VPS i/o wait time was accounted wrongly as presense of uninterruptible tasks, not those in IO wait state.

http://forum.openvz.org/index.php?t=tree&goto=2686&#msg_2686

diff-ve-vzmon-comp-fix-20060424

Patch from Dmitry Mishin <dim@openvz.org>:
Fixed compilation for CONFIG_VZ_WDOG=y