kernel weekly news – 14.05.2011

Posted: May 14, 2011 in kernel

Hello everyone !

-We start with Ingo Molnar’s perf fixes, Paul E. McKenney has
rcu fixes for 2.6.40, Dave Airlie announces drm fixes , mainly
mm, radeon and nouveau, Greg Kroah Hartman announces
2.6.38.6 (http://git.kernel.org/?p=linux/kernel/git/stable/linux-2.6.38.y.git;a=summary)
and also 2.6.32.40 and 2.6.33.13 and … and….

-Linus announces 2.6.39-rc7 as follows :

 So things have been pretty quiet, and unless something major comes up
I believe that this will be the last -rc.

There's only about a hundred commits here, and the bulk of the diffs
are to fs/hpfs which has been marked as BROKEN since early in the
merge window because of the final BKL removal. So it is a regression
fix. That said, I admittedly considered just leaving hpfs broken for
the 39 release and doing this fix series in the next merge window.
Whatever. Might as well take it now, since it has no impact outside of
hpfs, and certainly thus couldn't make things -worse-.

Outside of the hpfs patches, there really isn't all that much
interesting going on. Some hw breakpoint ptrace races, some cifs
updates, and then just random small fixes, mostly to drivers. Nothing
noticeable really stands out, it really has been pretty quiet.

But please do test, just to make sure that 39-final is good.

                      Linus 

-Joerg Roedel of AMD fame has some IOMMU updates
(of course, for AMD CPUs :-) targetted at 2.6.40,
Konrad Rzeszutek has xen stable/blkfront updates/fixes
also aimed at 2.6.40, Alex Elder has a few xfs updates
for .39 and David Miller has the following report on
his work on the networking tree:

 There's several OOPS'ers and reverts in here.

I think we now have all of the worst ones fixed from the regression
list, and I would recommend doing just one more -rc to get this all
sorted out and tested properly.  But of course that is completely up
to you.

1) ipheth regressed because it had a hard dependency upon NET_IP_ALIGN
   being defined always as 2, get rid of that assumption.  From Ben
   Hutchings.

2) IPV6 REJECT module puts random values in TOS field, fix from
   Fernando Luis Vazquez Cao.

3) When an ipv4 fragmentation entry expires via a timer, we have to
   revalidate the route otherwise we can crash.  Fix from Eric
   Dumazet.

4) In usbnet, usbnet_bh can be scheduled too early during resume
   resulting in flood of RX frames but no reclaim, and this leads to
   running out of atomic memory.  Don't allow usbnet_bh to schedule
   until the device is brought completely up.  Fix from Ming Lei.

5) TCP cubic can divide by zero in some extreme cases, fix from
   Stephen Hemminger.

6) SLIP and SLCAN devices return incorrect values from their ldisc
   open method, from Matvejchikov Ilya and Oliver Hartkopp.

7) VLAN GVRP state is undone at the wrong moment, causing crashes during
   batched device delete.  From Eric Dumazet.

8) dev_close() mistakenly had it's IFF_UP check removed, this has to be
   put back otherwise we can crash during batched device teardown, in
   particular with bonding.  Fix from Eric DUmazet.

9) PCH_GBE "checksum correct" logic on RX is reversed (hardware sets
   the status bit on checksum failure, clears it on success), from
   Toshiharu Okada.

10) vmxnet3 does not take ->cmd_lock consistently with interrupts disabled,
    as is warned by lockdep.  Fix from Roland Dreier.

11) DCCP feature options length needs to be validated properly,
    otherwise we can end up working with negative lengths, fix from
    Dan Rosenberg.

13) Fix regressions in ebtables compat support, from Eric Dumazet and
    Florian Westphal.

14) IPVS namespace support can leave objects referenced indefinitely until
    reboot.  Fixes from Hans Schillstrom.

15) DSCP netfilter code forgets to invert mask, from Fernando Luis
    Vazquez Cao.

16) Revert a buggy xt_conntrack change that broke handling of locally
    generated packets.  From Florian Westphal and Jan Engelhardt.

17) Inter-family packet output was busted in IPSEC, we need to split
    the operation into two parts, ->output() and ->output_finish(), to
    make sure we work in the context of the correct protocol (ipv4 vs
    ipv6) at each step.  Fix from Steffen Klassert.

18) We cannot allow ESN handling when anti-reply detection is disabled,
    also from Steffen Klassert.

Please pull, thanks a lot! 

-Dave Airlie has drm fixes (nouveau and radeon) :

 two fixes for issues myself and Alex discovered trying to bring up accel 
on the cayman hardware. The other change is a bit gross but is all I'm 
comfortable with doing this close to a release, it fixes an build failure 
on alpha due to the first attempt at supporting Xen + drm drivers, we 
actually backed out the change that needs these paths to work so I've just 
commented out the two paths for now until Konrad and Thomas work out how 
to do this cleanly or just go back to having no Xen support, I seem to 
remember someone worrying about Xen making crap out of drivers once merged 
and I'm hoping we aren't the first victim, but I'll give them some more 
time to find an acceptable solution. 

-Martin Schwidefsky announces s390 patches for -rc7 , Samuel Ortiz has
MFD fixes for 2.6.39, John W. Linville comes with wireless fixes – for 2.6.39
also , dated 11.05-, Sage Weil updates the ceph tree, for .39 again and
David Miller has updates for the sparc tree (32-bit sparc, that is).

-Konrad Rzeszutek Wilk announces xen stable bugfixes for -rc7,
Takashi Iwai updates sound, random-tracing(stacktrace) updates
are announced by Frederic Weisbecker and…

-…I would like to wish you a nice weekend and see you!

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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