2.6.36 merge window part 1

By Jonathan Corbet
August 4, 2010

The 2.6.36 merge window got off to a rather slow start; Linus, perhaps, has been spending too much time with his coffee maker and not enough at the keyboard. Things got rolling, though, on the afternoon of August 4; as of this writing, about 2600 patches have been merged into the mainline. Here is a summary of what has been seen so far.

User-visible changes include:

The 9p filesystem has gained support for extended attributes and a new, Linux-specific variant of the 9p2000 protocol called 9p2000.L.
The CIFS filesystem can now make use of FS-Cache to keep local copies of files for performance.
The TOMOYO Linux security module has a new "interactive enforcing mode," allowing an administrator to permit policy violations at run time. It is intended to help when installing application updates (which require policy changes) on running production systems.
At long last, the AppArmor security module has been merged.
Rafael Wysocki's wakeup_count mechanism has been merged. This feature is intended to make it possible to suspend the system without having to worry about races with wakeup events; it thus hopes to solve part of the problem addressed by suspend blockers.
Support for the LIRC infrared controller API has been merged, along with a long list of LIRC drivers. LIRC is one of the larger pieces of out-of-tree code which is still shipped by many distributors, so this merge should help bring distributor kernels that much closer to the mainline.
New drivers:
- Boards and systems: Bluewater Systems Snapper 9260/9G20 modules, HP t5325 thin client systems, NXP Semiconductor LPC32xx-based systems, and Eukrea CPUIMX51 and CPUIMX35 modules.
- Input: Atmel QT602240 I2C touchscreens, Analog Devices ADXL34x three-axis digital accelerometers, and Cypress cy8ctmg110 touchscreens.
- Miscellaneous: ARM Ltd. character LCD displays, HTC "Dream" (G1 handset) GPIO lines, and Intel "intelligent power sharing" controllers.
- Networking: Freescale Flexcan CAN controllers, ESD USB/2 CAN/USB interfaces, Chelsio T4-based gigabit and 10Gb Ethernet adapters with PCI-E SR-IOV virtual functions, and CAIF protocol drivers on slave SPI interfaces
- Video4Linux: i.MX27/i.MX25 camera sensor interfaces, SunPlus SPCA1528-based USB cameras, SQ Technologies SQ930X-based USB cameras, Windows Media Center Edition eHome infrared transceivers, and Freescale VIU video engines.

Changes visible to kernel developers include:

The ARM architecture has lost support for the "discontigmem" memory model; it is expected that everybody is using sparsemem at this point. ARM has also switched from the old bootmem allocator to memblock (formerly LMB) and added support for the -fstack-protector GCC feature.
The DMAPI hooks have been dropped from the XFS filesystem, indicating that the XFS developers do not ever expect to get hierarchical storage management at this level merged.
The PM_QOS API has changed again; quality-of-service requests are now added with:
```
    void pm_qos_add_request(struct pm_qos_request_list *request,
			    int pm_qos_class, s32 value);
```
The biggest change is that the request structure must now be allocated by the caller; this shifts a bit of work but, importantly, allows this function to be called in atomic context.

The merge window can be expected to remain open until around August 15, unless Linus decides to surprise developers by making it shorter.

Index entries for this article
Kernel	Releases/2.6.36

2.6.36 merge window part 1

Posted Aug 5, 2010 8:00 UTC (Thu) by YannisDas (guest, #60144) [Link] (2 responses)

Jonathan, how about an article on the out-of-tree pieces of code out there? There are big pieces that are actively developed (others not so much anymore) and are waiting for inclusion.

I think it's very interesting to learn what is really out there and what can be expected in the (near?) future.

Out of tree modules

Posted Aug 5, 2010 10:16 UTC (Thu) by TRS-80 (guest, #1804) [Link] (1 responses)

A quick way to see some of what's out there is what's supported by module-assistant, although some are just for easy installation of development versions on old kernels, and DKMS is starting to take over. m-a list|awk -F- '/source/ {ORS=" ";print $1}' on Ubuntu 10.04 gives

acerhk acx100 affix alsa arla at76c503a bcm4400 bcm5700 cdfs cipe cloop comedi cpad cryptoapi cryptoloop dazuko ddrmat device3dfx drbd0.7 drbd8 dvb e100 eagle em8300 exmap fglrx freeswan ftape ftpfs fuse fwatch gpib hostap hubcot i2c ieee80211 ipw2100 ipw2200 ivtv kqemu linux linux lirc lm loop loop lufs madwifi mga misdn ndiswrapper nozomi nvidia nvidia nvidia openafs openswan ov511 pcmcia plex86 ppscsi qc qla2x00 realtime rt2400 rt2500 rt2570 rtai shfs sl spca5xx squashfs sysprof thinkpad tidev translucency tun unicorn unionfs userlink vaiostat video4linux virtualbox wacom xdslusb xlibmesa zaptel

Personally, I found myself having to compile DAHDI (nee zaptel) recently for asterisk.

Out of tree modules

Posted Aug 12, 2010 4:23 UTC (Thu) by mchehab (subscriber, #41156) [Link]

This list is incomplete and wrong. By looking on it, I've got the impression that Ubuntu has a weird concept of "out-of-tree": they have DKMS packages for things that are ever upstream or that are at upstream for a long time, like alsa, dvb, ipw*, ov511, pcmcia, v4l, ivtv, etc. It looks like that there are separate packages for drivers that also provide testing trees to allow compilation as separate drivers to easy driver development (like v4l/dvb) instead of properly porting the upstream patches to their kernels.

There's actually a list of the out of tree drivers at:
http://www.linuxdriverproject.org/foswiki/bin/view/Main/O...

The list is a little outdated, as it still shows some devices that were already merged, but it is a good start. Unfortunately, it doesn't show what drivers are actually shipped with the major distros.

2.6.36 merge window part 1

Posted Aug 5, 2010 16:42 UTC (Thu) by cesarb (subscriber, #6266) [Link]

For future reference, a permanent link to the blog post mentioned as the first link on this article: http://torvalds-family.blogspot.com/2010/08/13744-supplie...

2.6.36 merge window part 1

Posted Aug 5, 2010 22:06 UTC (Thu) by BenHutchings (subscriber, #37955) [Link]

I added support for 64-bit networking stats on 32-bit systems (dependent on driver support). Eric Dumazet then implemented this for many drivers.

These drivers should now provide 64-bit stats on all architectures: 8021q, bnx2, bonding, cxgb4, ixgbe, loopback, macvlan, sfc, tg3.

2.6.36 merge window part 1

Posted Aug 7, 2010 8:29 UTC (Sat) by lbt (subscriber, #29672) [Link]

LIRC --- \o/

2.6.36 merge window part 1

Posted Aug 9, 2010 14:40 UTC (Mon) by nye (guest, #51576) [Link] (1 responses)

>The 2.6.36 merge window got off to a rather slow start; Linus, perhaps, has been spending too much time with his coffee maker

Might be worth linking directly to the post, for the benefit of readers from the future?

http://torvalds-family.blogspot.com/2010/08/13744-supplie...

2.6.36 merge window part 1

Posted Aug 9, 2010 14:55 UTC (Mon) by nye (guest, #51576) [Link]

Aaand now I notice somebody already has :P.