------------------------------------------------------------------------
Greg Ungerer -- Chief Software Wizard EMAIL: [email protected]
Snapgear Pty Ltd PHONE: +61 7 3279 1822
825 Stanley St, FAX: +61 7 3279 1820
Woolloongabba, QLD, 4102, Australia WEB: www.SnapGear.com

Christoph Hellwig

2002-10-15 17:12 UTC #1 lore raw

Quoted 9 lines · Greg Ungerer

Hi All,
An updated uClinux patch is available at:
http://www.uclinux.org/pub/uClinux/uClinux-2.5.x/linux-2.5.42uc1.patch.gz
Changelog:
1. v850 update

There are a bunch of CVS .#* files left from this one. v850_defs.h wants updating to the generic generate-asm-offsets.h mechanism (check the toplevel Makefile)
Please stop the ugly symlink hell with the linker scripts - we have vmlinux.lds.S for that.
it should read something like:

#include <linux/config.h>
#ifdef BOARD1
#include "board1.lds"
#else
...
#endif

for v850

Could you please explain the rootfs hacks in v850? I don't think we want those in mainline but rather generic initrd/initramfs.

Also please remove arch/v850/sim/* - that stuff doesn't belong into the kernel tree.

Quoted 1 line

2. cleaned up mm/page_alloc.c

Why do you put set_page_refs into a header? Separating it out looks good to me, but IMHO it should stay in page_alloc.c. BTW, are you sure that you don't need to set the refs in the other caller of prep_new_page? To me it looks like you should and then you could merge it into prep_new_page.

Also, what is CONFIG_CONTIGUOUS_PAGE_ALLOC doing? It seems not fully implemented but adds lots of uglieness :) CONFIG_NO_MMU_LARGE_ALLOCS might want a saner name, btw (CONFIG_LARGE_ALLOCS?).

General commets:
- Config.in files have three-space, not two-space indentation
- I don't think you want to keep around the old binfmt_flat format when merghin into mainline
- MAX_SHARED_LIBS is never used

Greg Ungerer

author

2002-10-16 14:36 UTC #2 lore raw

Hi Christoph,

Quoted 1 line · Christoph Hellwig

There are a bunch of CVS .#* files left from this one.

Cleaned out now.

Quoted 2 lines

v850_defs.h wants updating to the generic generate-asm-offsets.h mechanism (check the toplevel Makefile)

Miles cleaned this up.

Quoted 2 lines

Please stop the ugly symlink hell with the linker scripts - we have vmlinux.lds.S for that.

Hmph. Hardly seems any better...

Quoted 6 lines

Could you please explain the rootfs hacks in v850? I don't think we want those in mainline but rather generic initrd/initramfs.
Also please remove arch/v850/sim/* - that stuff doesn't belong into the kernel tree.

Done.

Quoted 3 lines

Quoted 1 line
2. cleaned up mm/page_alloc.c
Why do you put set_page_refs into a header?

It looked might it might be usefull elsewhere.
(but currently isn't used for anything else).

Quoted 5 lines

Separating it out
looks good to me, but IMHO it should stay in page_alloc.c. BTW, are you sure that you don't need to set the refs in the other caller of prep_new_page? To me it looks like you should and then you could merge it into prep_new_page.

No, you don't want that.

Quoted 2 lines

Also, what is CONFIG_CONTIGUOUS_PAGE_ALLOC doing? It seems not fully implemented but adds lots of uglieness :)

It was a second allocator that more cleverly allocated large memory chunks - not using the power of 2 allocator. I have not carried this into the 2.5 patches. Very nice to have, since you suffer much more from memory fragmentation without VM. But not strictly neccessary, and to minimize and simplify the current patch sets I left it out.

Cleanup out all references to it now.

Quoted 2 lines

CONFIG_NO_MMU_LARGE_ALLOCS might want a saner name, btw (CONFIG_LARGE_ALLOCS?).

Yes, no doubt the NO_MMU name is universally disliked :-) Changed to COFNIG_LARGE_ALLOCS now.

Quoted 3 lines

General commets:
- Config.in files have three-space, not two-space indentation

Fixed.

Quoted 1 line

- I don't think you want to keep around the old binfmt_flat

This isn't old. It is the primary format used on uClinux. ELF and a.out are not practical, since you would need to do the final link/locate on them at exec load time (you won't know what address in memory they will get loaded to until them). You don have the VM luxary of just locating it at a fixed address at compile time.

FLAT format is a light weight, mostly architecture independant way to carry around relocs, and to keep the program binaries small.

Quoted 2 lines

format when merghin into mainline
- MAX_SHARED_LIBS is never used

Another "feature" I have not carried into 2.5. Removed.

Thanks
Greg

Signature · 5 lines

------------------------------------------------------------------------
Greg Ungerer  --  Chief Software Wizard        EMAIL:  [email protected]
SnapGear Pty Ltd                               PHONE:    +61 7 3435 2888
825 Stanley St,                                  FAX:    +61 7 3891 3630
Woolloongabba, QLD, 4102, Australia              WEB:   www.SnapGear.com

Christoph Hellwig

2002-10-16 14:42 UTC #3 lore raw

Quoted 9 lines

This isn't old. It is the primary format used on uClinux. ELF and a.out are not practical, since you would need to do the final link/locate on them at exec load time (you won't know what address in memory they will get loaded to until them). You don have the VM luxary of just locating it at a fixed address at compile time.
FLAT format is a light weight, mostly architecture independant way to carry around relocs, and to keep the program binaries small.

I don't meant binfmt_flat itself but the support for the old-style relocs that is still in the code.

BTW, does binfmt_flat for for any non-uClinux port?

Greg Ungerer

author

2002-10-16 15:30 UTC #4 lore raw

Quoted 6 lines · Christoph Hellwig

Quoted 6 lines
This isn't old. It is the primary format used on uClinux. ELF and a.out are not practical, since you would need to do the final link/locate on them at exec load time (you won't know what address in memory they will get loaded to until them). You don have the VM luxary of just locating it at a fixed address at compile time.

Quoted 7 lines

Quoted 4 lines
FLAT format is a light weight, mostly architecture independant way to carry around relocs, and to keep the program binaries small.
I don't meant binfmt_flat itself but the support for the old-style relocs that is still in the code.

OK, I'll have a look at that.

Quoted 1 line

BTW, does binfmt_flat for for any non-uClinux port?

Don't beleive so.

Regards
Greg

------------------------------------------------------------------------
Greg Ungerer -- Chief Software Wizard EMAIL: [email protected]
SnapGear Pty Ltd PHONE: +61 7 3435 2888
825 Stanley St, FAX: +61 7 3891 3630
Woolloongabba, QLD, 4102, Australia WEB: www.SnapGear.com

Miles Bader

2002-11-26 07:39 UTC #5 lore raw

PATCH 1/1 Make some EXPORT_SYMBOLs dependent on CONFIG_MMU

In reply to Christoph Hellwig

[I'm not sure who to send this to, so I'm guessing. Pointers appreciated!]

Hi,

A few symbols are only defined when CONFIG_MMU=y, but are exported (by kernel/ksyms.c) unconditionally. This patch makes them conditional.

Patch:

ksyms.c renamed from orig/linux-2.5.49-uc0/kernel/ksyms.c +4 -0

diff -ruN -X../cludes ../orig/linux-2.5.49-uc0/kernel/ksyms.c kernel/ksyms.c
@@ -324,7 +324,9 @@
 /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */
 EXPORT_SYMBOL(default_llseek);
 EXPORT_SYMBOL(dentry_open);
+#ifdef CONFIG_MMU
 EXPORT_SYMBOL(filemap_nopage);
+#endif
 EXPORT_SYMBOL(filemap_fdatawrite);
 EXPORT_SYMBOL(filemap_fdatawait);
 EXPORT_SYMBOL(lock_page);

@@ -525,7 +527,9 @@
 EXPORT_SYMBOL(single_release);
 
 /* Program loader interfaces */
+#ifdef CONFIG_MMU
 EXPORT_SYMBOL(setup_arg_pages);
+#endif
 EXPORT_SYMBOL(copy_strings_kernel);
 EXPORT_SYMBOL(do_execve);
 EXPORT_SYMBOL(flush_old_exec);

Thanks,

-Miles

Signature · 2 lines

Love is a snowmobile racing across the tundra.  Suddenly it flips over,
pinning you underneath.  At night the ice weasels come.  --Nietzsche

Miles Bader

2002-11-26 07:42 UTC #6 lore raw

PATCH 1/1 v850 additions to include/linux/elf.h

In reply to Christoph Hellwig

Hi,

This patch adds more stuff to include/linux/elf.h for the v850 (used by the new module loader).

Patch:

linux/elf.h renamed from orig/linux-2.5.49-uc0/include/linux/elf.h +28 -0

diff -ruN -X../cludes ../orig/linux-2.5.49-uc0/include/linux/elf.h include/linux/elf.h
@@ -92,6 +92,9 @@
  */
 #define EM_ALPHA	0x9026
 
+/* Bogus old v850 magic number, used by old tools.  */
+#define EM_CYGNUS_V850	0x9080
+
 /*
  * This is the old interim value for S/390 architecture
  */

@@ -450,6 +453,31 @@
 /* Keep this the last entry.  */
 #define R_390_NUM	27
 
+
+/* v850 relocations.  */
+#define R_V850_NONE		0
+#define R_V850_9_PCREL		1
+#define R_V850_22_PCREL		2
+#define R_V850_HI16_S		3
+#define R_V850_HI16		4
+#define R_V850_LO16		5
+#define R_V850_32		6
+#define R_V850_16		7
+#define R_V850_8		8
+#define R_V850_SDA_16_16_OFFSET	9	/* For ld.b, st.b, set1, clr1,
+					   not1, tst1, movea, movhi */
+#define R_V850_SDA_15_16_OFFSET	10	/* For ld.w, ld.h, ld.hu, st.w, st.h */
+#define R_V850_ZDA_16_16_OFFSET	11	/* For ld.b, st.b, set1, clr1,
+					   not1, tst1, movea, movhi */
+#define R_V850_ZDA_15_16_OFFSET	12	/* For ld.w, ld.h, ld.hu, st.w, st.h */
+#define R_V850_TDA_6_8_OFFSET	13	/* For sst.w, sld.w */
+#define R_V850_TDA_7_8_OFFSET	14	/* For sst.h, sld.h */
+#define R_V850_TDA_7_7_OFFSET	15	/* For sst.b, sld.b */
+#define R_V850_TDA_16_16_OFFSET	16	/* For set1, clr1, not1, tst1,
+					   movea, movhi */
+#define R_V850_NUM		17
+
+
 /* Legal values for e_flags field of Elf64_Ehdr.  */
 
 #define EF_ALPHA_32BIT		1	/* All addresses are below 2GB */

Thanks,

-Miles

Signature · 3 lines

[|nurgle|]  ddt- demonic? so quake will have an evil kinda setting? one that 
            will  make every christian in the world foamm at the mouth? 
[iddt]      nurg, that's the goal

Alan Cox

2002-11-26 15:03 UTC #7 lore raw

Re: [PATCH] v850 additions to include/linux/elf.h

Quoted 4 lines · Miles Bader

Hi,
This patch adds more stuff to include/linux/elf.h for the v850 (used by the new module loader).

To save cluttering a linux/*.h file couldnt the module loader for v850 include an asm/*.h file holding the extra info ?

Miles Bader

2002-11-27 01:13 UTC #8 lore raw

Re: [PATCH] v850 additions to include/linux/elf.h

Quoted 5 lines · Alan Cox

Quoted 2 lines
This patch adds more stuff to include/linux/elf.h for the v850 (used by the new module loader).
To save cluttering a linux/*.h file couldnt the module loader for v850 include an asm/*.h file holding the extra info ?

I suppose so, but the v850 is not exceptional in this regard. The various relocation type constants (R_...) in elf.h are completely unused in the kernel on _every_ architecture, except by the new module loader.

If tidying is needed, then it seems like the best thing would be to move all the arch-specific stuff into the corresponding <asm/elf.h> file for each architecture. I presume the reason this hasn't been done is simply convention -- userland elf.h files are also giant conglomerations of defines for every supported architecture ...

-Miles

Signature · 2 lines

Next to fried food, the South has suffered most from oratory.
  			-- Walter Hines Page

Alan Cox

2002-11-27 12:22 UTC #9 lore raw

Re: [PATCH] v850 additions to include/linux/elf.h

Quoted 5 lines · Miles Bader

If tidying is needed, then it seems like the best thing would be to move all the arch-specific stuff into the corresponding <asm/elf.h> file for each architecture. I presume the reason this hasn't been done is simply convention -- userland elf.h files are also giant conglomerations of defines for every supported architecture ...

I agree entirely. So lets start with v850 8)

Miles Bader

2002-11-28 01:14 UTC #10 lore raw

Re: [PATCH] v850 additions to include/linux/elf.h

Quoted 5 lines · Alan Cox

Quoted 3 lines
If tidying is needed, then it seems like the best thing would be to move all the arch-specific stuff into the corresponding <asm/elf.h> file for each architecture.
I agree entirely. So lets start with v850 8)

Okay. I sent the previous patch to Linus and he doesn't seem to have taken it; perhaps your suggestion will fare better.

I'd like to remain consistent with other archs though -- does anyone want to make a patch (I guess just move all the R_... constants) for them?

-Miles

Signature · 1 line

Yo mama's so fat when she gets on an elevator it HAS to go down.

Miles Bader

2002-10-16 05:36 UTC #11 lore raw

In reply to Greg Ungerer

Hi,

Here's a v850 update for 2.5.42uc1.

I addressed two of Christoph Hellwig's concerns, (1) vmlinux.lds.S [barf] and (2) the asm-constant generation mechanism.

He also complained about using the MD driver &c instead of initrd, but I'm not sure what do about that -- it'd be nice to use a `standard' solution, but when I originally looked at the initrd stuff, it seemed very convoluted and confusing; since earlier lkml discussion had pointed to using MD as the nearest thing to the old blkmem device, that seemed like the way to go.

What's your opinion on this?

Patch:

v850/Makefile renamed from orig/linux-2.5.42uc1/arch/v850/Makefile +15 ~~-39~~

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/arch/v850/Makefile arch/v850/Makefile
@@ -23,32 +23,6 @@
 CFLAGS += -D__linux__ -DUTS_SYSNAME=\"uClinux\"
 
 
-# linker scripts
-ifdef CONFIG_V850E_SIM
-LINKER_SCRIPT = $(arch_dir)/sim.ld
-endif
-ifdef CONFIG_V850E2_SIM85E2C
-LINKER_SCRIPT = $(arch_dir)/sim85e2c.ld
-endif
-ifdef CONFIG_V850E2_FPGA85E2C
-LINKER_SCRIPT = $(arch_dir)/fpga85e2c.ld
-endif
-ifdef CONFIG_V850E2_ANNA
-LINKER_SCRIPT = $(arch_dir)/anna.ld
-endif
-ifdef CONFIG_RTE_CB_MA1
-ifdef CONFIG_ROM_KERNEL
-LINKER_SCRIPT = $(arch_dir)/rte_ma1_cb-rom.ld
-else
-ifdef CONFIG_RTE_CB_MA1_KSRAM
-LINKER_SCRIPT = $(arch_dir)/rte_ma1_cb-ksram.ld
-else
-LINKER_SCRIPT = $(arch_dir)/rte_ma1_cb.ld
-endif
-endif
-endif
-
-
 HEAD := $(arch_dir)/kernel/head.o $(arch_dir)/kernel/init_task.o
 core-y += $(arch_dir)/kernel/
 libs-y += $(arch_dir)/lib/

@@ -73,17 +47,19 @@
 endif
 
 
-# Barf.
-$(arch_dir)/vmlinux.lds.S: $(LINKER_SCRIPT)
-	cp $< $@
-
-
-archmrproper:
-
-archdep:
+prepare: include/asm-$(ARCH)/asm-consts.h
 
-archclean:
-	rm -f vmlinux
-	rm -f root_fs_image.o
-	rm -f $(arch_dir)/kernel/v850_defs.h $(arch_dir)/kernel/v850_defs.d
-	rm -f $(arch_dir)/vmlinux.lds.S $(arch_dir)/vmlinux.lds.s
+# Generate constants from C code for use by asm files
+arch/$(ARCH)/kernel/asm-consts.s: include/asm include/linux/version.h \
+				   include/config/MARKER
+include/asm-$(ARCH)/asm-consts.h.tmp: arch/$(ARCH)/kernel/asm-consts.s
+	@$(generate-asm-offsets.h) < $< > $@
+include/asm-$(ARCH)/asm-consts.h: include/asm-$(ARCH)/asm-consts.h.tmp
+	@echo -n '  Generating $@'
+	@$(update-if-changed)
+
+
+CLEAN_FILES += include/asm-$(ARCH)/asm-consts.h.tmp \
+	       include/asm-$(ARCH)/asm-consts.h \
+	       arch/$(ARCH)/kernel/asm-consts.s \
+	       root_fs_image.o

v850/kernel/Makefile renamed from orig/linux-2.5.42uc1/arch/v850/kernel/Makefile +0 ~~-14~~

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/arch/v850/kernel/Makefile arch/v850/kernel/Makefile
@@ -36,17 +36,3 @@
 
 
 include $(TOPDIR)/Rules.make
-
-
-$(obj)/head.o: $(src)/head.S $(obj)/v850_defs.h
-$(obj)/entry.o: $(src)/entry.S $(obj)/v850_defs.h
-
--include $(obj)/v850_defs.d
-
-$(obj)/v850_defs.h: $(src)/v850_defs.c $(src)/v850_defs.head
-	rm -f $(obj)/v850_defs.d
-	SUNPRO_DEPENDENCIES="$(obj)/v850_defs.d $(obj)/v850_defs.h" \
-	$(CC) $(filter-out -MD,$(CFLAGS)) -S -o $(obj)/v850_defs.s $(src)/v850_defs.c
-	cp $(src)/v850_defs.head $(src)/v850_defs.h
-	grep '^#define' $(obj)/v850_defs.s >> $(obj)/v850_defs.h
-	rm $(obj)/v850_defs.s

v850/kernel/asm-consts.c renamed from orig/linux-2.5.42uc1/arch/v850/kernel/asm-consts.c +60 -0

60 lines not on this page · raw message

v850/kernel/entry.S renamed from orig/linux-2.5.42uc1/arch/v850/kernel/entry.S +1 -1

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/arch/v850/kernel/entry.S arch/v850/kernel/entry.S
@@ -22,7 +22,7 @@
 #include <asm/irq.h>
 #include <asm/errno.h>
 
-#include "v850_defs.h"
+#include <asm/asm-consts.h>
 
 
 /* Make a slightly more convenient alias for C_SYMBOL_NAME.  */

v850/kernel/v850_defs.c renamed from orig/linux-2.5.42uc1/arch/v850/kernel/v850_defs.c +0 ~~-58~~

58 lines not on this page · raw message

v850/kernel/v850_defs.head renamed from orig/linux-2.5.42uc1/arch/v850/kernel/v850_defs.head +0 -5

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/arch/v850/kernel/v850_defs.head arch/v850/kernel/v850_defs.head
@@ -1,5 +0,0 @@
-/*
- * WARNING! This file is automatically generated - DO NOT EDIT!
- */
-
-#define TS_MAGICKEY	0x5a5a5a5a

v850/vmlinux.lds.S renamed from orig/linux-2.5.42uc1/arch/v850/vmlinux.lds.S +27 -0

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/arch/v850/vmlinux.lds.S arch/v850/vmlinux.lds.S
@@ -0,0 +1,27 @@
+#include <linux/config.h>
+
+#ifdef CONFIG_V850E_SIM
+#include "sim.ld"
+#endif
+
+#ifdef CONFIG_V850E2_SIM85E2C
+#include "sim85e2c.ld"
+#endif
+
+#ifdef CONFIG_V850E2_FPGA85E2C
+#include "fpga85e2c.ld"
+#endif
+
+#ifdef CONFIG_V850E2_ANNA
+#include "anna.ld"
+#endif
+
+#ifdef CONFIG_RTE_CB_MA1
+# ifdef CONFIG_ROM_KERNEL
+#  include "rte_ma1_cb-rom.ld"
+# elif CONFIG_RTE_CB_MA1_KSRAM
+#  include "rte_ma1_cb-ksram.ld"
+# else /* !CONFIG_ROM_KERNEL && !CONFIG_RTE_CB_MA1_KSRAM */
+#  include "rte_ma1_cb.ld"
+# endif /* CONFIG_ROM_KERNEL */
+#endif /* CONFIG_RTE_CB_MA1 */

asm-v850/page.h renamed from orig/linux-2.5.42uc1/include/asm-v850/page.h +19 -8

diff -ruN -X../cludes ../orig/linux-2.5.42uc1/include/asm-v850/page.h include/asm-v850/page.h
@@ -21,6 +21,7 @@
 #define PAGE_SIZE       (1UL << PAGE_SHIFT)
 #define PAGE_MASK       (~(PAGE_SIZE-1))
 
+
 /*
  * PAGE_OFFSET -- the first address of the first page of memory. For archs with
  * no MMU this corresponds to the first free page in physical memory (aligned

@@ -30,11 +31,6 @@
 #define PAGE_OFFSET  0x0000000
 #endif
 
-/*
- * MAP_NR -- given an address, calculate the index of the page struct which
- * points to the address's page.
- */
-#define MAP_NR(addr) (((unsigned long)(addr) - PAGE_OFFSET) >> PAGE_SHIFT)
 
 #ifdef __KERNEL__
 #ifndef __ASSEMBLY__

@@ -91,9 +87,11 @@
 
 #endif /* !__ASSEMBLY__ */
 
+
 /* to align the pointer to the (next) page boundary */
 #define PAGE_ALIGN(addr)	(((addr) + PAGE_SIZE - 1) & PAGE_MASK)
 
+
 #ifndef __ASSEMBLY__
 
 extern void __bug (void) __attribute__ ((noreturn));

@@ -116,18 +114,31 @@
 
 #endif /* !__ASSEMBLY__ */
 
-#define __pa(x)			__virt_to_phys ((unsigned long)(x))
-#define __va(x)			((void *)__phys_to_virt ((unsigned long)(x)))
 
 /* No current v850 processor has virtual memory.  */
 #define __virt_to_phys(addr)	(addr)
 #define __phys_to_virt(addr)	(addr)
 
-#define virt_to_page(kaddr) 	(mem_map + MAP_NR (kaddr))
+#define virt_to_pfn(kaddr)	(__virt_to_phys (kaddr) >> PAGE_SHIFT)
+#define pfn_to_virt(pfn)	__phys_to_virt ((pfn) << PAGE_SHIFT)
+
+#define MAP_NR(kaddr) \
+  (((unsigned long)(kaddr) - PAGE_OFFSET) >> PAGE_SHIFT)
+#define virt_to_page(kaddr)	(mem_map + MAP_NR (kaddr))
+#define page_to_virt(page) \
+  ((((page) - mem_map) << PAGE_SHIFT) + PAGE_OFFSET)
+
+#define pfn_to_page(pfn)	virt_to_page (pfn_to_virt (pfn))
+#define page_to_pfn(page)	virt_to_pfn (page_to_virt (page))
+
 #define	virt_addr_valid(kaddr)						\
   (((void *)(kaddr) >= (void *)PAGE_OFFSET) && MAP_NR (kaddr) < max_mapnr)
 
 
+#define __pa(x)		     __virt_to_phys ((unsigned long)(x))
+#define __va(x)		     ((void *)__phys_to_virt ((unsigned long)(x)))
+
+
 #endif /* KERNEL */
 
 #ifdef CONFIG_CONTIGUOUS_PAGE_ALLOC

Thanks,

-Miles

Signature · 1 line

Occam's razor split hairs so well, I bought the whole argument!

Greg Ungerer

author

2002-10-16 12:45 UTC #12 lore raw

Hi Miles,

Quoted 1 line · Miles Bader

Here's a v850 update for 2.5.42uc1.

I have rolled this into, linux-2.5.43uc0.
I have also removed the gdb patches under arch/v850.

Quoted 9 lines

I addressed two of Christoph Hellwig's concerns, (1) vmlinux.lds.S [barf] and (2) the asm-constant generation mechanism.
He also complained about using the MD driver &c instead of initrd, but I'm not sure what do about that -- it'd be nice to use a `standard' solution, but when I originally looked at the initrd stuff, it seemed very convoluted and confusing; since earlier lkml discussion had pointed to using MD as the nearest thing to the old blkmem device, that seemed like the way to go.

I very much like using a specific MTD map driver. I haven't carried my initrd patches through to the 2.5 code yet. But it was pretty simple to do for 2.4...

Regards
Greg