Re: [PATCH] cowlinks v2

[Davide Libenzi]

On 22 Mar 2004, Eric W. Biederman wrote:

> Davide Libenzi <davidel@xxxxxxxxxxxxxxx> writes:
> 
> > There has been a misunderstanding. I thought you were talking about a 
> > userspace solution ala fl-cow. Of course if you are inside the kernel you 
> > can catch both explicit writes and page syncs.
> 
> Right.  Although there is nothing prevent the copy to be in user space
> even with the trigger hooks down in the write path.

How do you insert yourself before the first page fault to do the COW, from 
userspace (open+mmap)? You can obviously hook mmap(2) and if a PROT_WRITE 
is requested, you COW from there. But then you have an whole bunch of new 
problems (again, when done from userspace) because, just to begin with, 
you need a stateful interception layer (while fl-cow for example is stateless).
In my modest usage scenario for my fl-cow shell (emacs+patch+diff+gcc) 
I've found that when something opens in RDWR, it really writes the file at 
some point during the opened session. So moving the COW down in the write 
path helps little or nothing. There may be as well other use cases where 
applications do frequently open in RDWR even w/out ever touching the file.



- Davide


-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/