several bugfixes/workarounds in mtd snapshot of 20041226 && 2.4.28

Dear MTD maintainers,

I suggest several bugfixes and workarounds against the snapshot of 20041226 (inserted into a vanilla 2.4.28) - see the attached patch.

I'm using DiskOnChip devices in various sorts of Advantech hardware. I'm compiling my MTD drivers for the DiskOnChip alone - I don't want to include all the other MTD stuff (useless to me).

Bugs fixed / worked around:

1) if you flush the partition table and issue a BLKRRPART
(i.e., you re-partition the diskonchip), the reported device
size shrinks to one half of the original. If you do this again
before a reboot, the size shrinks again (now one quarter).
Seems like the driver would re-init device size to a size denominated in 1k blocks, rather than sectors (512B).

2) in my DiskOnChip-only setup, CONFIG_MTD_MAP_BANK_WIDTH_1
is missing in the .config. I've added that straight to $LINUX/include/linux/map.h

3) two #includes seem to be missing in $MTD/nand/diskonchip.c

4) the Reed-Solomon library hooks are messed up in $LINUX/lib/Config.in

I hope this helps

Frank Rysanek

@@ -315,7 +315,7 @@
 	}
 
 	grok_partitions(gd, minor, 1 << tr->part_bits, 
-			tr->blkcore_priv->sizes[minor]);
+                        dev->size);
 	up(&mtd_table_mutex);
 
 	return 0;

@@ -14,6 +14,10 @@
 #include <asm/io.h>
 #include <asm/bug.h>
 
+#ifndef CONFIG_MTD_CFI_GEOMETRY
+#define CONFIG_MTD_MAP_BANK_WIDTH_1
+#endif
+
 #ifdef CONFIG_MTD_MAP_BANK_WIDTH_1
 #define map_bankwidth(map) 1
 #define map_bankwidth_is_1(map) (map_bankwidth(map) == 1)

@@ -24,6 +24,8 @@
 #include <linux/sched.h>
 #include <linux/delay.h>
 #include <linux/rslib.h>
+#include <linux/moduleparam.h>
+#include <linux/rslib.h>
 #include <asm/io.h>
 
 #include <linux/mtd/mtd.h>

@@ -43,13 +43,15 @@
 
 if [ "$CONFIG_MTD_DOCPROBE" = "y" -o \
      "$CONFIG_MTD_NAND_RTC_FROM4" = "y" -o \
-     "$MTD_NAND_DISKONCHIP" = "y" ]; then
+     "$CONFIG_MTD_NAND_DISKONCHIP" = "y" ]; then
    define_tristate CONFIG_REED_SOLOMON y
+   define_tristate CONFIG_REED_SOLOMON_DEC16 y
 else
   if [ "$CONFIG_MTD_DOCPROBE" = "m" -o \
      "$CONFIG_MTD_NAND_RTC_FROM4" = "m" -o \
-     "$MTD_NAND_DISKONCHIP" = "m" ]; then
+     "$CONFIG_MTD_NAND_DISKONCHIP" = "m" ]; then
      define_tristate CONFIG_REED_SOLOMON m
+     define_tristate CONFIG_REED_SOLOMON_DEC16 y
   else
      define_tristate CONFIG_REED_SOLOMON n
   fi

Committed. Thanks.

I think that's a Config.in bug -- it shouldn't happen on 2.6.

rslib.h was already there; I've added moduleparam.h which was not.

Committed; thanks.

dwmw2