Re: [PATCH 1/3] uprobes/core: Remove redundant lock_page/unlock_page

[Peter Zijlstra <peterz@infradead.org>]

On Thu, 2012-05-31 at 20:37 +0530, Srikar Dronamraju wrote:
> * Peter Zijlstra <peterz@infradead.org> [2012-05-31 13:58:38]:
> 
> > On Thu, 2012-05-31 at 17:16 +0530, Srikar Dronamraju wrote:
> > > From: Srikar Dronamraju <srikar@linux.vnet.ibm.com>
> > > 
> > > No need to lock the page when copying the opcode in read_opcode().
> > 
> > It would be good if the changelog said _why_ this is so :-)
> 
> In read_opcode(), we have the reference for the page and we only are reading
> from the the page. i.e we are neither modifying the page contents, not
> the page attributes. 

Fair enough, so put that in the changelog. The changelog should explain
things, not raise questions.

> Existing kernel code has enough examples where we read the contents
> of the page without taking the page lock.

Yes, but that doesn't tell us this site is ok, doing it because others
do isn't an argument.

> Further this was discussed here too https://lkml.org/lkml/2012/4/17/361.

That wasn't a discussion, that was two people saying they don't know.