Enabling swap in embedded systems

If your embedded system running Linux does not have enough memory, you can enable swap to get more memory. However if your platform does not have MMU (Memory Management Unit) as is the case for Sigma Designs EM8620 series, it won’t support swap, so forget it. If your platform does have MMU, as is the case for many newer platforms such as Sigma Designs SMP8630, SMP8640 and SMP8650 series, you can enable swap support. First you’ll have to make sure swap support is enabled in your kernel: and swapon/swapoff is enabled in busybox. So for example if you have an IDE harddisk with the second partition configured as swap. (Use fdisk to create a partition and mkswap /dev/hda2 to initialize the partition), you can enable the swap as follows: If you get out-of-memory killer kernel error, you can change the “swapiness” to avoid oom-killer to kick in.

Advertisements

How to Generate CVS Changelog Automatically

There is a perl script called cvs2cl.pl provided by Red-Bean You can get the latest version thru cvs: cvs -d :pserver:[email protected]:/usr/local/cvs co cvs2cl/cvs2cl.pl Copy the script in a directory that belong to the path e.g.: cp cvs2cl.pl /usr/local/bin Then just checkout your project and run cvs2cl.pl to generate a ChangeLog that will look like: 2010-01-18 developer1 * app/cnxapp/Makefile: Added support for SMP8652 target 2010-01-17 developer2 * app/cnxapp/cnx_download.c: Added HTTP resume support [Bug 1250]: Fixed potential buffer overflow … where developer1 and developer2 are the linux usernames of the developers, app/cnxapp/Makefile, app/cnxapp/cnx_download.c the files that have been modified, and the comments correspond to the messages input during cvs commit. cvs2cl.pl also have other options such as XML output, date selection etc… Just type cvs2cl.pl –help to get the full options.

Block CVS Commits without message

If you are in my case, and you found it difficult to make some people in your team make proper comments/messages or any comments at all (cvs commit -m “Description of the changes”) while committing their changes to CVS, here’s what can be done to block commit without comments and with short comments. Checkout the CVSROOT directory cvs co CVSROOT Edit verifymsg in CVSROOT as follows #Set the verifymsg file to run a script DEFAULT /usr/local/bin/validate-cvs-log.sh Create /usr/local/bin/validate-cvs-log.sh script (make it executable for all cvs users): #!/bin/sh # # validate-cvs-log.sh logfile # test that log message has some characters in it if [ cat $1 | wc -c -lt 12 ] ; then echo “log message too short; please enter a description for the changes” exit 1 else exit 0 fi Commit verifymsg cvs commit -m “Added script to set minimum message size to 12 characters” verifymsg That’s it ! Next time a developer tries to commit changes without messages/comments …

How to do a framebuffer screenshot

I’ll explain how to do framebuffer screenshots on 16-bit and 32-bit framebuffer. For 16-bit this is fully based on http://docs.blackfin.uclinux.org/doku.php?id=framebuffer Capturing screenshots Whatever the bit-depth of your framebuffer, the first step is to capture the frambuffer raw data on the board: Now the you need to take the raw image, and convert it to a standard image format. This step depends on what type of display is there Converting 16-bit Framebuffer screenshot (RGB565) into png To convert the raw rgb data extracted from /dev/fb0, use iraw2png perl script To do the conversion, type the following command in the host: where 640 and 480 are respectively the width and height of your framebuffer. This has been tried on a 16-bit framebuffer on EM8620 series. Converting 32-bit Framebuffer screenshot (ARGB, RGBA, BGRA…) into png The solution proposed here is not as neat as the blackfin’s solution for 16-bit framebuffer, however this still works. First you’ll need to install Gimp 2 in your …

Advertisements

uClinux kernel panic: Stack overflow

If you’re using ucLinux, you may get kernel panic errors coming out of nowhere. There may be several reasons (buffer overflow, out of memory..), but the most common is stack overflow for the process or one of the threads. To increase the stack size of a flat binary you’ll need to adjust the LDFLAGS as follows: This will set the stack size to 64KB. To change the stack size of a thread (e.g. 32KB below), you’ll need to set the stack size attribute: How to detect which thread suffers from stack overflow? First, you can check your code for recursive function calls and local variables (especially arrays) both of which will be added at runtime to the stack to estimate what should be the stack size. So if you have large arrays you may use a pointer + a call to malloc instead. If this can not fix the issue, you can use the functions below to return the used …

Digital Signage Standard by POPAI

Currently digital signage implementations are mainly proprietary, and that means once you choose a vendor you are stuck with it, unless you completely change your network and the way you manage your content or handles separate systems. If the digital signage industry really wants to take off some standards are needed. One of the first standard for digital signage has been released by POPAI (Point of Purchase Advertising International). This first standard is to promote interoperability between different providers by defining Screen Media Formats (Click to open the standard) basically telling which video and audio codecs and containers (“Wrappers”)¬† and which and picture formats¬† should/could be supported by digital signage players and providing different levels of standard profiles and extended profiles a bit like it is done in MPEG-4 specifications. POPAI also used to provide some media samples to test your system previously available at www.popai.com/DS/ContentSamples but the link is now dead. You may have to contact them to get …

Advertisements