I fixed that yesterday in the buildroot.

There's one other thing which the latest kernel changes (or possible the latest mtd) -- the format of JFFS2 sector headers seems to have changed in a way that mtd-utils' mkfs.jffs2 hasn't adopted.  It spews a bunch of warnings at boot time; but otherwise, 2.6.15 is looking pretty good now.


On Jan 20, 2006, at 8:52 AM, Simon de Bakker wrote:

Hi all,

I just tried to run the gumstix with a fresh build of the 2.6.14 buildroot from svn (so 2.6.15 actually :).
At first i had a kernel panic on partitioning the gumstix Flash ROM. The problem was that the external mtd drivers are used in the 2.6.15 kernel while the device class interface has changed. It looks like most of the mtd stuff is now already in the kernel, but just removing the external mtd from the linux.mk does not work. Does anybody know the current status of this?

For now i just patched the external mtd driver to use the new device class interface and everything seems to work ok again.

Simon de Bakker
diff -uNr mtd.orig/drivers/mtd/mtdchar.c mtd/drivers/mtd/mtdchar.c
--- mtd.orig/drivers/mtd/mtdchar.c 2005-08-04 03:05:48.000000000 +0200
+++ mtd/drivers/mtd/mtdchar.c 2006-01-20 16:07:58.000000000 +0100
@@ -24,10 +24,10 @@
  if (!mtd)


- class_device_create(mtd_class, MKDEV(MTD_CHAR_MAJOR, mtd->index*2),
+ class_device_create(mtd_class, NULL, MKDEV(MTD_CHAR_MAJOR, mtd->index*2),
      NULL, "mtd%d", mtd->index);


- class_device_create(mtd_class, 
+ class_device_create(mtd_class, NULL,  
      MKDEV(MTD_CHAR_MAJOR, mtd->index*2+1),
      NULL, "mtd%dro", mtd->index);