Unclear about unwriteprotecting

Unclear about unwriteprotecting

[elf]

I've built the mtd driver into a 2.2.17 kernel using the mtd-20000704
version.  I am accessing a DOC 2000 (24MB) that was created using the
M-Systems's doc driver.  It has two partitions, one formatted for
ext2.

The NFTL driver sees the partitions find and I can mount them.
However, I cannot write.  Nothing in the code or on the list has
enlightened me.  What am I missing?

Building from CVS for 2.2.x

[elf]

It appears that at least one of the include references in the CVS code
(mtd-snapshot-20020504) is not compatible with the 2.2.x kernel
series.

I'm building the kernel from a directory in my /home.  The kernel
running on the development system is 2.4.x.  I unpacked a stock 2.2.17
kernel, patched it with mtd-2.2.17.patch, and then ran patchin.sh.  I
I checked the links and found them proper and intact.

The first problem was that the spinlock types were defined twice
incompatibly.  In compactmac.h, the inclusion of linux/spinlock.h at
line 362 pulls the file from /usr/include instead of the linux source
tree version.  Changing this include to asm/spinlock.h eliminated the
problem, though I'm not confident that this is the right way to fix
it.

Then, the pci.c driver in mtd/maps failed to build, but for a reason I
couldn't deduce.  The compiler complains that the ioremap_noncache
symbol is missing.  

What might I be missing?  Is there another kernel patch?

Re: Unclear about unwriteprotecting

[David Woodhouse]

elf@buici.com said:
> I've built the mtd driver into a 2.2.17 kernel using the mtd-20000704
> version.  I am accessing a DOC 2000 (24MB) that was created using the
> M-Systems's doc driver.  It has two partitions, one formatted for
> ext2.

> The NFTL driver sees the partitions find and I can mount them.
> However, I cannot write.  Nothing in the code or on the list has
> enlightened me.  What am I missing? 

You're using code that's almost two years old. Don't do that.

--
dwmw2

Latest CVS and 2.2.x

[elf]

On Tue, May 07, 2002 at 04:29:47PM +0100, David Woodhouse wrote:
> When building from the CVS tree itself, the makefiles should add the CVS 
> tree's include directory first on the include path, so it should work OK. 
> When building by copying the CVS code into your kernel tree, you should 
> copy the headers over so that should work too.
> 
> What did you do, and precisely what were the errors?

I sent this over the weekend to the list:

  It appears that at least one of the include references in the CVS code
  (mtd-snapshot-20020504) is not compatible with the 2.2.x kernel
  series.

  I'm building the kernel from a directory in my /home.  The kernel
  running on the development system is 2.4.x.  I unpacked a stock 2.2.17
  kernel, patched it with mtd-2.2.17.patch, and then ran patchin.sh.  I
  I checked the links and found them proper and intact.

  The first problem was that the spinlock types were defined twice
  incompatibly.  In compactmac.h, the inclusion of linux/spinlock.h at
  line 362 pulls the file from /usr/include instead of the linux source
  tree version.  Changing this include to asm/spinlock.h eliminated the
  problem, though I'm not confident that this is the right way to fix
  it.

  Then, the pci.c driver in mtd/maps failed to build, but for a reason I
  couldn't deduce.  The compiler complains that the ioremap_noncache
  symbol is missing.

  What might I be missing?  Is there another kernel patch?

What I'm doing is unpacking the CVS tree, using the patch and patchin
as I described and then building from the kernel source tree as I
normally would.  This worked with the two year old version.

Re: Latest CVS and 2.2.x

[David Woodhouse]

elf@buici.com said:
> 
>   The first problem was that the spinlock types were defined twice
>   incompatibly.  In compactmac.h, the inclusion of linux/spinlock.h at
>   line 362 pulls the file from /usr/include instead of the linux source
>   tree version.  Changing this include to asm/spinlock.h eliminated the
>   problem, though I'm not confident that this is the right way to fix
>   it.

That's very broken. Are you sure that include/linux/spinlock.h exists in 
your 2.2 kernel tree?

>   Then, the pci.c driver in mtd/maps failed to build, but for a reason I
>   couldn't deduce.  The compiler complains that the ioremap_noncache
>   symbol is missing. 

That's probably been introduced since 2.2. Add something like...

	#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0)
	#define ioremap_noncache ioremap
	#endif

... to compatmac.h and then if it works, send me the diff.



--
dwmw2