da4458bda2
Change RomFS so that it can use MTD devices directly - without the intercession
of the block layer - as well as using block devices.
This permits RomFS:
(1) to use the MTD direct mapping facility available under NOMMU conditions if
the underlying device is directly accessible by the CPU (including XIP);
(2) and thus to be used when the block layer is disabled.
RomFS can be configured with support just for MTD devices, just for Block
devices or for both. If RomFS is configured for both, then it will treat
mtdblock device files as MTD backing stores, not block layer backing stores.
I tested this using a CONFIG_MMU=n CONFIG_BLOCK=n kernel running on my FRV
board with a RomFS image installed on the mtdram test device. I see my test
program being run XIP:
# cat /proc/maps
...
c0c000b0-c0c01f8c r-xs 00000000 1f:00 144 /mnt/doshm
...
GDB on the kernel can be used to show that these addresses are within the
set-aside RAM space.
Signed-off-by: David Howells <dhowells@redhat.com>
Tested-by: Bernd Schmidt <bernd.schmidt@analog.com>
Signed-off-by: David Woodhouse <David.Woodhouse@intel.com>
13 lines
198 B
Makefile
13 lines
198 B
Makefile
#
|
|
# Makefile for the linux RomFS filesystem routines.
|
|
#
|
|
|
|
obj-$(CONFIG_ROMFS_FS) += romfs.o
|
|
|
|
romfs-y := storage.o super.o
|
|
|
|
ifneq ($(CONFIG_MMU),y)
|
|
romfs-$(CONFIG_ROMFS_ON_MTD) += mmap-nommu.o
|
|
endif
|
|
|