kernel weekly news – 07.01.2012

Posted: January 7, 2012 in kernel

Hello everyone!

-David Miller announces networking fixes, as below:

1) mlx4 driver stops working after a ring size change due to forgotten
clear of the queue vector value, fix from Yevgeny Petrilin.

2) RX multicast filter in skge is not restored after resume, breaking
ipv6 amongst other things, fix from Florian Zumbiehl.

3) Fix sync message handling in IPVS, from Julian Anastasov.

4) ctnetlink timeout sanity checks never work due to signedness,
fix from Xi Wang.

-Greg KH announces the release of kernels, 3.0.15 and 3.1.7.
Arnaldo Carvalho de Melo has perf/core fixes and improvements,
John W. Linville has wireless fixes and David Miller updates networking

Probably the bluetooth 10 second resume delay fix is the most
interesting fix here, especially if you’re actually hitting it.

The rest are small fixes isolated to specific drivers or packet

1) Revert bluetooth HCI reset timeout increase, it adds a 10 second delay
during resume in Thinkpad laptops and potentially other machines. From
Gustavo F. Padovan.

2) New packet QFQ scheduler using wrong types to compute queueing limits
leading to overflows, fix from Eric Dumazet.

3) Fix regression in b43 driver conversion to new RX descriptor header
format, it broke PIO mode where the header length needs to be computed
differently. Fix from Guennadi Liakhovetski.

4) Fix build dependencies of FEC ethernet driver, from Wolfram Sang.

5) mwifiex gets OOPS when scan and connect occur simultaneously, fix from
Amitkumar Karwar.

6) ath9k crashes in access point mode because it does aggregation
handling during power save even if TX aggregation is not enabled.
Fix from Mohammed Shafi Shajakhan.

7) Update CAN MAINTAINERS entry, from Oliver Hartkopp.

-Al Viro has vfs updates, Keith Packard updates drm-intel,
Neil Brown has important md changes for 3.3 and Grant Likely
has gpio and spi changes, also for 3.3.

-Linus Torvalds finally announces the release of Linux 3.2:

So 3.2 is out, and the merge window for 3.3 is thus officially open.

I delayed 3.2 first a few days to wait for the final linux-next
(“final” in the sense that that’s what I’ll fetch to decide whether
something has been in linux-next for 3.3 or not), and then some more
as people were coming back from holidays and sorting out some
regressions. So we do have a few last-minute reverts and small fixes.

Still, there’s not a whole lot of changes since -rc7 (shortlog
appended), and almost all of them are *tiny*. So despite the few
annoying last-minute reverts, I’m feeling pretty happy about it.

-Ingo Molnar has core/{locking,memblock,printk}, RCU, perf and scheduler
changes, and also timer, x86/{apic,asm,cleanup,cpu,debug,efi,mce,microcode},
Geert Uytterhoeven has m68k updates for 3.3, Ingo steps up again with
x86/{mm,platform,rdrand} changes, Joerg Roedel has IOMMU updates for
3.3-rc1 and Tony Luck has pstore changes.

-Guenter Roeck has hwmon updates for 3.3:

No new drivers this time, only cleanups, minor fixes, and added support for
new chips to existing drivers.

Most of the patches have been in -next for a while, but there are some cleanups
and minor fixes which I only added in the last couple of days (the patches
submitted by Frans). Some of those may be missing in the -next snapshot you use
as base to determine what to accept and what not (all are in next-20120105,
though). I have no problem taking those out and resubmit later into an -rc
if you prefer. Please let me know.

-Ben Myers has XFS updates for -rc1, Jiri Kosina updates HID, APM
and trivial, James Bottomley has SCSI patches and David Miller
has networking changes:

The most prominent thing here are the TX byte queue limit changes
from Tom Herbert and others at Google.

The idea is that you can limit the amount of packet data, in bytes,
that can get queued up to a TX queue of a networking device. And as
a result one can tune things such that high priority packets get
serviced with a reasonable amount of latency whilst not subjecting
the hardware queue to emptying when data is available to send.

Driver’s need to make some minor changes to support this feature, and
drivers for most of the most commonly used ethernet devices have been
transformed. In particular niu, bnx2, igb, e1000e, forcedeth, tg3,
bnx2x, sfc, and sky2 have support.

Eric Dumazet added a new interface for drivers called build_skb() which
allows the driver to DMA RX packets into kmalloc()’d buffers, and
then allocate the packet metadata on the local cpu right when we push
the packet up into the stack for better cache locality. Currently
tg3, bnx2, and bnx2x have been converted to use this new interface.

Eric also extended our RED packet scheduler to support Adaptive RED
wherein the random-drop/tail-drop thresholds are adjusted dynamically
in response to traffic.

Jiri Pirko has introduced a new “teaming” network device which is
intended to be a more cleanly designed, scalable, and simpler
replacement for the bonding driver. Currently round-robin and
active-backup modes are implemented.

Netlink socket dumping is now supported for UDP and AF_UNIX sockets
thanks to Pavel Emelyanov. For AF_UNIX sockets we report the peers as
well as the pending connection IDs. The reporting of socket memory
usage is also now more complete.

Socket memory pressure can now be enforced on a per-cgroup level,
from Glauber Costa.

As is usually the case the majority of this pull is a boatload of
driver updates.

-Greg Ungerer updates the m68knommu tree, Benjamin Herrenschmidt
updates powerpc.git, Ingo Molnar is yet again back with x86/syscall
changes for 3.3, Steven Whitehouse updates GFS2 and that’s it for
this week. See ya!


Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s