Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 18 Jan 2017 22:29:10 -0500
From: David Windsor <>
To: Kees Cook <>
Cc: Elena Reshetova <>, 
	"" <>, Arnd Bergmann <>, 
	Thomas Gleixner <>, Ingo Molnar <>, 
	"H. Peter Anvin" <>, Peter Zijlstra <>, 
	Will Deacon <>, Greg KH <>, 
	Hans Liljestrand <>
Subject: Re: [RFCv2 PATCH 09/18] net: convert from atomic_t to refcount_t

On Wed, Jan 18, 2017 at 8:11 PM, Kees Cook <> wrote:
> On Wed, Jan 18, 2017 at 10:39 AM, David Windsor <> wrote:
>> The relevant part of the patch has been cut off, but it appears that
>> you've fixed some of the issues we identified earlier as corner cases
>> as to reference counting in net/.  In particular, inetpeer.c has its
>> own garbage collection system in place which frees shared objects when
>> their reference count became -1 (rather than 0).  The proposed
>> solution was to do a global +1 on this refcounting scheme, then
>> replace unsupported atomic_*() functions with appropriate refcount_*()
>> calls.
>> When submitting this to netdev, it may make sense to separate out
>> these changes: first, do a global +1 (while still using atomic_t),
>> then convert to refcount_t.  I'm already working on this now, but I
>> didn't know if you wanted to follow this approach or not.
> David, if you've got the global +1 patches ready, let's start feeding
> those to netdev ASAP. We can convert them to refcount_t more easily
> after that.

Alright, let me get them in order and I'll submit them to netdev.  I
just wanted to make sure that an alternate approach hadn't been
decided upon.


> -Kees
> --
> Kees Cook
> Nexus Security

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.