Linux 2.6.22.atmel.2, containing a small handful of improvements upon the 2.6.22.atmel.1 release, has been released. For more information and downloads, please see the LinuxPatches page.

...shortly followed by 2.6.22.atmel.1, containing many additional drivers and a few other improvements. Grab it from the LinuxPatches page and give it a spin. Please see the announcement for details.

This also means that the 2.6.23 merge window has opened. Currently queued for inclusion, possibly subject to review, are:

• The Atmel USBA UDC driver
• RTC driver for AP7000
• Watchdog driver for AP7000
• LTV350QV LCD panel driver
• gpio mouse driver (not avr32-specific)
• gpio LEDs driver (not avr32-specific and not developed by anyone here AFAIK, but very useful)

In addition there will be a few general improvements, like configurable STK1000 board setup code based on jumper settings.

Two new mailing lists have been created: "kernel" for kernel-related discussions and "u-boot" for bootloader-related discussions. Read more at the MailingLists page.

New AVR32 Linux kernel patchsets based on the 2.6.20 kernel have been released. Preliminary patches adding support for CompactFlash have also been added.

Two Linux kernel patchsets for the 2.6.20 kernel have been added. The second one, 2.6.20-avr2, fixes a bug in the atmel_serial driver which only triggers when the CompactFlash patchset is applied.

A separate page for the CompactFlash driver has been added, which includes the necessary patches that need to be applied to the kernel along with a description of a few modifications that need to be done to the ATSTK1000 board for this to work properly. The driver should not be considered complete, but it does work fine with at least some cards.

Linux 2.6.20 was released last sunday, with support for the MACB driver out of the box and a handful of other AVR32-related fixes and improvements.

In addition to the macb driver, the following shortlog summarizes all the changes that affected AVR32 to some degree:

Adrian Bunk (1):
      cleanup asm/setup.h userspace visibility

Al Viro (1):
      [NET]: AVR32 checksum annotations and cleanups.

Alan Cox (1):
      tty: preparatory structures for termios revamp

Andrew Victor (1):
      [ARM] 3954/1: AT91: Update drivers for new headers

Benjamin Herrenschmidt (1):
      Driver core: add dev_archdata to struct device

David Howells (1):
      LOG2: Implement a general integer log2 facility in the kernel

Haavard Skinnemoen (13):
      [AVR32] Portmux API update
      [AVR32] Add macb1 platform_device
      [AVR32] Move ethernet tag parsing to board-specific code
      [AVR32] Remove mii_phy_addr and eth_addr from eth_platform_data
      [AVR32] Remove unused file
      [AVR32] Set flow handler for external interrupts
      [AVR32] Put the chip in "stop" mode when halting the system
      [AVR32] Don't include <asm/delay.h>
      [AVR32] Implement intc_get_pending()
      [AVR32] Pass dev parameter to dma_cache_sync()
      [AVR32] Add missing #include <linux/param.h> to delay.c
      [AVR32] Export clear_page symbol
      [AVR32] Update ATSTK1000 defconfig: Enable macb by default

Matthew Wilcox (1):
      Centralise definitions of sector_t and blkcnt_t

Nigel Cunningham (1):
      Add include/linux/freezer.h and move definitions from sched.h

Paul Mundt (1):
      avr32: fixup kprobes preemption handling

Ralf Baechle (3):
      Add struct dev pointer to dma_is_consistent()
      Pass struct dev pointer to dma_cache_sync()
      Optimize D-cache alias handling on fork

Robert P. J. Day (1):
      getting rid of all casts of k[cmz]alloc() calls

I have to admit I was a bit surprised to see how many changes actually affected AVR32 in some way, especially the checksum annotations and cleanups change since I've never even seen it until now. But Al Viro usually knows what he's doing and the changes look good, so I'm not complaining.

Looking forward, here are the changes that have been merged for 2.6.21 so far:

Ahmed S. Darwish (1):
     [AVR32] Use ARRAY_SIZE macro when appropriate

David Brownell (4):
     [AVR32] /proc/interrupts display
     [AVR32] fix serial port setup on ATSTK1000
     [AVR32] ext int fixes
     [AVR32] Fix incorrect invalidation of shared cachelines

Haavard Skinnemoen (9):
     [AVR32] Implement dma_mapping_error()
     [AVR32] GPIO API implementation
     [AVR32] Don't reset PIO state at bootup
     [AVR32] Introduce at32_reserve_pin()
     [AVR32] Add PIOE device and reserve SDRAM pins
     [AVR32] SPI platform code update
     [AVR32] Remove last remains of libgcc
     [AVR32] ssize_t should be long, not int
     [AVR32] Add missing #include <linux/module.h>

Robert P. J. Day (1):
     Remove a couple final references to obsolete verify_area().

In addition, the AtmelSpiDriver has been submitted and will hopefully be included in the 2.6.21 release as well. If so, the 2.6.21 kernel will support Dataflash out of the box on AVR32.

Updated patches for Linux (both 2.6.18 and 2.6.19), U-Boot, Binutils, GCC and uClibc are available. Please test if you feel adventurous.

Also, Linux 2.6.19 has been released with support for AVR32 out of the box.

As usual, there are no guarantees that the patches won't blow up your development board or any other piece of equipment you happen to have on your desk, so use at your own risk. But quite a number of bugs have been fixed, so it might be worth a try.

Unfortunately, even though u-boot 1.1.6 comes with AVR32 support out of the box, no new patchsets against 1.1.6 are available at avr32linux.org yet. Stay tuned for ethernet- and mmc-support in the hopefully not-too-distant future.

A new Application Repository has been added. We encourage everyone who try out new applications on AVR32 Linux to post their efforts there, and to try out the applications already listed.

This application repository will hopefully grow into a comprehensive resource on how to build existing Linux software for AVR32. As a quick scan of the AVR32 forum on avrfreaks.net shows, there's already a significant amount of knowledge around, and collecting this in a single, easily accessible place would be a very good thing.

Each entry in the repository should contain a link to the source code, a patch (if necessary) and instructions on how to build it (in a separate page.) Consolidating instructions that are common to several applications is probably a good idea too -- there is already a page with generic instructions for building applications that utilize autoconf.

Some spammers figured that this site could be used to host their crap. Not anymore.

Three user accounts have been locked down because of HTML Attachment spam. We have also tightened security slightly by, among other things, installing the BlackListPlugin. Please notify the site administrator if this is hindering any legitimate use of the site (which includes pretty much everything even remotely related to AVR32 and Linux.)

If you own one of the user accounts that were locked down and you think that was a mistake, please let us know. You will have quite a lot of explaining to do, though.

Note that this incident was not a security breach of the site -- registered users uploaded attachments to their own homepage, which is a perfectly valid thing to do. However, the attachments were full of links to sleazy sites, whose purpose was to increase the Google rank of those sites.

Sorry for not noticing this immediately. Hopefully, this incident didn't cause too much damage to the site's reputation, and the BlackListPlugin as well as better awareness of the problem will hopefully prevent this from happening again.

It's been a long time since the last news post here, although a lot of things have been happening. The mainstream linux and u-boot trees now feature AVR32 support, drivers have been submitted for review, and avr32linux.org has seen a minor reorganization, including a new "application repository" where everyone can submit information about applications they have working on avr32-linux.

The Linux 2.6.19 kernel will come with support for AVR32 out of the box. Not many drivers have been merged yet, but it supports file systems in parallel flash and console on serial port, so it will boot and function as a somewhat limited Linux system.

Linux drivers for the AP7000 MACB ethernet controller and SPI controller have been submitted for review through the appropriate channels. They will hopefully be included with the 2.6.20 release, but it might take longer. In particular, the SPI driver currently depends on a common GPIO API for AVR32 and AT91, and possibly for the whole kernel.

u-boot 1.1.6 has been released with support for AVR32. There are still a few drivers missing here as well, but the serial port and parallel flash both work.

On the site, the GettingStarted link is gone from the left navigation bar. It's been replaced with a link to a new HowTo section where you, among other things, will find a link to the same GettingStarted page as before.

The LinuxApplications page has been restructured. The information about common porting problems have been moved to a separate page, and a new AVR32 Application Repository has been added. The idea is that whenever someone tries out a new program on AVR32 Linux, he or she will add an entry to this table with a link to a page with more information about things you need to look out for, a patch and some additional information.

I and a few others will add a few entries to this table within the next few days to get the ball rolling.

Another round of updates found its way to the GettingStarted page. Special thanks to squidgit from the AVR32 forum at avrfreaks.net for a nice summary of what he had to do to compile the AVR32 toolchain. Some of the status pages for the LinuxKernel have been updated as well.

In other news, Andrew Morton has released a new -mm patchset. The newest set of LinuxPatches should still apply cleanly, although three of them have been picked up by Andrew, so you should just skip them. These are:

• avr32-set-kbuild_defconfig.patch
• avr32-kprobes-compile-fix.patch
• avr32-include-asm-byteorder-h-from-asm-io-h.patch

The 2.6.18-rc4-mm3 patchset also includes a pretty major bugfix to the atomic bit operations, which isn't included on the LinuxPatches page.

A new experimental patchset is available from the LinuxPatches page. Among the new features are MTD support (for flash-based filsystems) and a driver for the Static Memory Controller.

This patchset is based on the 2.6.18-rc4-mm2 kernel by Andrew Morton, but a rolled-up AVR32 architecture patch is also included so that it can be applied on top of the mainstream 2.6.18-rc4 kernel as well. The following patches from 2.6.18-rc4-mm2 are included in the avr32-arch-6.patch:

avr32-arch.patch
avr32-config_debug_bugverbose-and-config_frame_pointer.patch
avr32-fix-invalid-constraints-for-stcond.patch
avr32-add-support-for-irq-flags-state-tracing.patch
avr32-turn-off-support-for-discontigmem-and-sparsemem.patch
avr32-always-enable-config_embedded.patch
avr32-export-the-find__bit-functions.patch
avr32-add-defconfig-for-at32stk1002.patch
avr32-use-autoconf-instead-of-marker.patch
avr32-dont-assume-anything-about-max_nr_zones.patch
avr32-add-i-o-port-access-primitives.patch
avr32-use-linux-pfnh.patch
avr32-kill-config_discontigmem-support-completely.patch
avr32-fix-bug-in-__avr32_asr64.patch
avr32-switch-to-generic-timekeeping-framework.patch

These patches will most likely be included in the next -mm release:

avr32-set-kbuild_defconfig.patch
avr32-kprobes-compile-fix.patch
avr32-include-asm-byteorder-h-from-asm-io-h.patch

The rest of the patches are mostly new drivers, which will be cleaned up and submitted at some point later:

atmel-macb-driver-3.patch
avr32-rename-usart-devices-to-at91_usart.patch
at91_serial-support-avr32.patch
at91_serial-fix-roundoff-error-in-at91_console_get_options.patch
at91_serial-fix-break-handling.patch
avr32-smc-support.patch
mtd-add-mapping-driver-for-the-at32stk1000-board.patch
avr32-add-platform-device-for-the-at49bv6416-on-at32stk1000.patch
avr32-rename-spi-platform-device-to-spi_atmel.patch
atmel-spi-driver-6.patch
atmel-lcdc-driver.patch

Please post any problems you encounter as a comment below, or use the AVR32 forum at avrfreaks.net.

A new patchset is available from the LinuxPatches page. This includes fixes for a lot of issues found by people on the linux-kernel mailing list, new interrupt handling code based on genirq and experimental preempt support.

Here's the shortlog. Thanks to Andrew Morton, Russell King, Arjan van de Ven, Thomas Gleixner, David Woodhouse, H. Peter Anvin and Nick Piggin for valuable feedback.

• Add MAINTAINERS entries for AVR32 and AT32AP
• Remove EARLY_PRINTK support
• Remove CONFIG_DW_DMAC symbol
• Remove DMA controller framework
• Add Kbuild file for 'make headers_install' on AVR32
• Remove #ifdef __KERNEL__ from asm/atomic.h
• Remove #ifdef __KERNEL__ from asm/bitops.h
• Remove #ifdef __KERNEL__ from asm/dma-mapping.h
• Remove #ifdef __KERNEL__ from asm/mmu_context.h
• Remove #ifdef __KERNEL__ from asm/semaphore.h
• Remove #ifdef __KERNEL__ from asm/thread_info.h
• Move PAGE_SIZE and friends inside #ifdef __KERNEL__
• Wrap __FD_SET and friends inside __KERNEL__
• Kill _syscall[0123456]() and hand-code execve() instead
• Remove obsolete #include <linux/config.h>
• Replace MB -> MiB and KB -> KiB
• Wire up 39 new syscalls
• Use do_div() instead of open-coded div64 in time_init()
• Use for_each_online_cpu in show_interrupts()
• Add asm/futex.h
• Ensure kprobes compiles again
• Switch to genirq framework
• Fix cpu_idle preempt bug
• Process softirqs the usual way
• Handle preempt in the interrupt- and exception handlers

The GettingStarted pages has been extended to the point that it should actually be possible to build the kernel and C library by following the instructions there. All necessary patches have been uploaded as well.

This includes versions of the BinutilsPatches, GccPatches, MicroClibcPatches and UbootPatches which are basically the same as the ones distributed with STK1000 BSP version 1.0. The LinuxPatches have been added earlier.

This is basically the same as the previous patchset, but it applies cleanly to v2.6.18-rc1. Get it from the LinuxPatches page.

The latest patchset available from the LinuxPatches page features two new drivers and some cleanups.

Here's the shortlog:

• Add framebuffer driver for the Atmel AT32/AT91 LCD Controller
• LTV350QV LCD panel driver
• Implement more clock operations and add lcdc device
• Add lcdc and spi devices and reintroduce board_setup_fbmem
• Convert sidsafb driver to use platform_device and struct clk
• Fix at91_serial break handling
• Atmel SPI Driver
• spi_atmel: convert to platform_device framework
• spi_atmel: fix broken parameter validation in setup() and transfer()
• spi_atmel: Revert core workaround for max_speed_hz=0
• spi_atmel: Divide len by two if bits_per_word > 8
• spi_atmel: Enable ENDRX interrupt when rx_buf is set
• spi_atmel: Fix incorrect locking in interrupt routine
• spi_atmel: Line up variable declarations
• spi_atmel: Add some FIXMEs from David Brownell's patch
• spi_atmel: Header file cleanup
• spi_atmel: Introduce new_1 flag and use in sck speed setting
• spi_atmel: Move SPI_ATMEL above SPI_BITBANG
• Define SO_PASSSEC
• Remove obsolete interrupt-specific SA_ flags
• Use named initializers for the mem_res elements
• Move ltv350qv device to SPI bus 0
• Make SPI_ATMEL available to AT91
• Delete asm/platform.h
• Fix sparse warnings
• Fix more sparse warnings
• macb: remove dependency on asm/platform.h
• at91_serial: remove dependency on asm/platform.h

A new experimental patchset against a 2.6.17 git snapshot is available at LinuxPatches, featuring a rewritten clock management framework.

The patchset was also announced at the linux-kernel mailinglist.

The new clock management framework is one step on the way to enable the AVR32 port to use some of the drivers written for the AT91 series of ARM microcontrollers. In fact, the at91_serial driver is already working and can be used as a serial console when booting the AVR32 Linux kernel on STK1000.

Now, the next steps would probably be to either convert some more drivers or start improving the at91_serial driver. I'm not sure yet.

Let's try to use the Blog plugin for site news.

This is one of the first blog postings I've ever made, so please bear with me.

I've been thinking a lot about the best ways to implement a "News page" on a site, and I've come to the conclusion that a Blog is probably the way to go. This way, you can easily provide a title, a "teaser", which will probably end up as the actual news item, and some more in-depth information that will be only visible to people that actually do want to know more (the real blog entry.)

So I'm going to give it a spin and see how it works out. The Blog plugin depends on a huge amount of other wiki plugins, which makes me a bit sceptical about how robust this whole thing is but hey, there's only one way to find out.

If you want to be able to post blog entries on your own, please drop me a message either as a comment here or as mail to _wremoId0. I want to restrict this ability to people who actually do any development or testing on the AVR32 Linux projects, though.

Oh by the way, if you're using Internet Explorer and think the AVR32 penguin logo at the top of the page looks ugly, it's because Microsoft still hasn't figured out how to render PNGs properly. There are workarounds, but I don't really want to "fix" things that aren't broken in the first place. Please consider using a real browser.