-H. Peter Anvin has x86 and x86/kdump for 3.9:
Three groups of fixes:
1. Make sure we don’t execute the early microcode patching if
family < 6, since it would touch MSRs which don't exist on those
families, causing crashes.2. The Xen partial emulation of HyperV can be dealt with more
gracefully than just disabling the driver.3. More EFI variable space magic. In particular, variables hidden
from runtime code need to be taken into account too.
#########################################
I am sending this particular patchset under its own cover because I
consider it to be a judgement call.The kexec/kdump people have found several problems with the support
for loading over 4 GiB that was introduced in this merge cycle. This
is partly due to a number of design problems inherent in the way the
various pieces of kdump fit together (it is pretty horrifically manual
in many places.)After a *lot* of iterations this is the patchset that was agreed upon,
but of course it is now very late in the cycle. However, because it
changes both the syntax and semantics of the crashkernel option, it
would be desirable to avoid a stable release with the broken
interfaces.
-David Miller has networking and SPARC fixes:
1) ax88796 does 64-bit divides which causes link errors on ARM, fix
from Arnd Bergmann.2) Once an improper offload setting is detected on an SKB we don’t
rate limit the log message so we can very easily live lock. From
Ben Greear.3) Openvswitch cannot report vport configuration changes reliably
because it didn’t preallocate the netlink notification message
before changing state. From Jesse Gross.4) The effective UID/GID SCM credentials fix, from Linus.
5) When a user explicitly asks for wireless authentication, cfg80211
isn’t told about the AP detachment leaving inconsistent state. Fix
from Johannes Berg.6) Fix self-MAC checks in batman-adv on multi-mesh nodes, from Antonio
Quartulli.7) Revert build_skb() change sin IGB driver, can result in memory
corruption. From Alexander Duyck.8) Fix setting VLANs on virtual functions in IXGBE, from Greg Rose.
9) Fix TSO races in qlcnic driver, from Sritej Velaga.
10) In bnx2x the kernel driver and UNDI firmware can try to program
the chip at the same time, resulting in corruption. Add proper
synchronization. From Dmitry Kravkov.11) Fix corruption of status block in firmware ram in bxn2x, from
Ariel Elior.12) Fix load balancing hash regression of bonding driver in forwarding
configurations, from Eric Dumazet.13) Fix TS ECR regression in TCP by calling tcp_replace_ts_recent() in
all the right spots, from Eric Dumazet.14) Fix several bonding bugs having to do with address manintainence,
including not removing address when configuration operations encounter
errors, missed locking on the address lists, missing refcounting on
VLAN objects, etc. All from Nikolay Aleksandrov.15) Add workarounds for firmware bugs in LTE qmi_wwan devices, wherein
the devices fail to add a proper ethernet header while on LTE
networks but otherwise properly do so on 2G and 3G ones. From
Bjørn Mork.
################################################
1) Fix race in sparc64 TLB shootdowns, we have to synchronize with the
sibling cpus completing if we are passing them a reference via
pointer to a data structure.2) Fix cleaning of bitmaps in sparc32, from Akinobu Mita.
3) Fix various sparc header mistakes, some of which resulted in
userland build breakage. From Sam Ravnborg.4) Kill ghost declarations and defines missed when several bits of
code got deleted recently.
-Linus Torvalds announces the release of kernel 3.9-rc8:
Yes, I was really hoping (and originally planning) to release 3.9
final this weekend, but we had enough issues that I just didn’t feel
comfy about it. It was borderline, and none of the issues were huge,
and maybe I could have called this just 3.9 and opened the merge
window, but hey, another week won’t hurt.The bulk of the changes here are networking (both core and drivers),
but there’s arch fixes (sparc, x86, arm, powepc) and random other
stuff. A couple of reverts and some cleanups. The shortlog gives some
flavor of the details. Nothing here on its own would have delayed 3.9
for me, but I was hoping for a calmer week.Please don’t send me pull requests unless it’s something really
critical, and let’s aim for a nice calm 3.9 release next weekend, ok?
