linux-kernel - Re: [PATCH v2 01/13] seqnum_ops: Introduce Sequence Number Ops

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20201117173839.GO29991@casper.infradead.org>
Date:   Tue, 17 Nov 2020 17:38:39 +0000
From:   Matthew Wilcox <willy@...radead.org>
To:     Shuah Khan <skhan@...uxfoundation.org>
Cc:     corbet@....net, keescook@...omium.org, gregkh@...uxfoundation.org,
        peterz@...radead.org, linux-doc@...r.kernel.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH v2 01/13] seqnum_ops: Introduce Sequence Number Ops

On Tue, Nov 17, 2020 at 09:34:24AM -0700, Shuah Khan wrote:
> > seqnum_inc() should just return the new value -- seqnum_inc_return is
> > too verbose.  And do we not need a seqnum_add()?
> 
> I had the patch series with seqnum_inc() all ready to go and then
> revisited the choice. My thinking is that matching the current atomic
> api that has _inc() and inc_return() might be less confusing. That

No, it's more confusing.  I know you're converting things from using
atomic_t, but you really need to think about this in terms of "What
makes sense for this API".  Unless you really want to have inc that
returns void and inc_return that returns the new value, having only
inc_return makes no sense.

> being said, I have no problems with making just _inc(). The reason
> for 32 and 64 appended is based on comments that it including size
> in the api makes it very clear.

By putting 32 and 64 in the name of the API, I would contend you're making
people think about something that they should not need to think about.

> No need for atomic_add() - inc_return() is sufficient for this use-case.

I haven't looked at the various potential users of this API, but there
are often cases where we account, eg, number of bytes transmitted.

There are also cases where read-and-zero would be a useful operation
to have.  I'm thinking about sampling statistics.