BadMEM patches/utils for Linux Kernel / News: Recent posts

The BadMEM project, aiming at being able to use partly-fault memory modules in a linux environment, now supports all current kernel based on the 2.6.x hive of the Linux Kernel's tree.

The kernel patches can now be downloaded at the usual project's place at SourceForge (see http://sourceforge.net/project/showfiles.php?group_id=24797 ).

The BadMEM project, aiming at being able to use partly-fault memory modules in a linux environment, now supports almost every kernel based on the 2.4.x hive of the Linux Kernel's tree. However, it is most important that with the today's updates, the latest version of the 2.4.x series, 2.4.31 is fully supported.

The kernel patches can now be downloaded at the usual project's place at SourceForge.

After the patch for Kernel Version 2.4.22 has settle, the patch for 2.4.23 was released yesterday. Currently there are no new features in the new patch.

Great success on attempts to custom writes to any SDRAM SPD-EEPROM!

Based on the latest efforts of this project, it is now possible to write to each eeprom on every SDRAM module with a so-called SPD EEPROM on it (almost every today's SDRAM module has one). This makes it possible to "burn" any BadMEM pattern into the module. During boot time of the Linux kernel it will be possible to lock bad memory spots soon. Only a special parallel-port driven SPD-adapter made by Madex in Germany is neccessary even to fix bugs in the manufactor's SPD-programming.

During my last efforts this night, I found out that the current version of Memtest86 (Version 3.0) surely has a bug in the compution of badram/badmem patterns. The bug can (but need not) lead to lost addresses especially on so called bit borders. A patch will be available very soon.

Hi Folks,

I just finished my work on the dangling problem, a hard problem on 0-bits in the mask that can cause wrong address counts. Due to this I was forced to use a sub-optimal hash algorithm for computing the combine costs of two patterns in the early 1.x's. Now, this has been resolved and the code is O(1) again. Tests with all your patterns that you have sent me are through and have been successful. This change makes it possible to use the utilities' library in the memtest86 environment. ... read more