Re: ramdisk corruption problems - was: RE: pivot_root and initrd kern el panic woes

[Tachino Nobuhiro]

At Thu, 20 Dec 2001 11:46:23 -0800,
Linus Torvalds wrote:
> 
> The problem is that having buffers doesn't necessarily always mean that
> they are valid, nor that _all_ of them are valid.

  If following sequence occurs, ramdisk_readpage() may clear the valid buffer
data. I'm not sure whether this really occurs, but if it does,
I think lock_buffer()/unlock_buffer() may be required.

	CPU1					CPU2
---------------------------------------------------------------------------
	call ext2_alloc_branch()		call ramdisk_readpage()

	bh = getblk();
	lock_buffer(bh);

	write data to bh.
						ramdisk_readpage()
						if (!buffer_uptodate(bh))
							do memset()

	mark_buffer_uptodate()
	unlock_buffer(bh);