Re: question about nfs_execute_read: why do we need to do lock_kernel?

["Xin Zhao" <uszhaoxin@gmail.com>]

Thanks for your reply. So the only reason is for rpc auditing? If so,
why not just lock the code that updating the audit information? Now
the code is:

lock_kernel()
rpc_execute()
unlock_kernel().

That means the kernel will be blocked when rpc is executed, which
could take long time. Even if rpc_execute() won't take very long, this
implementation still looks inefficient. That's why I am a little
confused on this point.

Any further thought?

Xin


On 4/25/06, Trond Myklebust <trond.myklebust@fys.uio.no> wrote:
> On Tue, 2006-04-25 at 00:57 -0400, Xin Zhao wrote:
> > This question may be dumb. But I am curious why in nfs_execute_read()
> > function, rpc_execute  is bracketed with lock_kernel() and
> > unlock_kernel()?
>
> We're keeping the BKL for the moment simply because we are not done
> auditing all the RPC code for potential races. When that is done, we
> will be able to remove it.
>
> Cheers,
>   Trond
>
>