netdev - Re: [PATCH bpf-next v2 1/2] net: xdp: refactor XDP_QUERY_PROG{,

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]

Date:   Mon, 3 Jun 2019 11:04:36 +0200
From:   Björn Töpel <bjorn.topel@...il.com>
To:     Jakub Kicinski <jakub.kicinski@...ronome.com>
Cc:     Saeed Mahameed <saeedm@...lanox.com>,
        "toke@...hat.com" <toke@...hat.com>,
        "daniel@...earbox.net" <daniel@...earbox.net>,
        "ast@...nel.org" <ast@...nel.org>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        "magnus.karlsson@...el.com" <magnus.karlsson@...el.com>,
        "bjorn.topel@...el.com" <bjorn.topel@...el.com>,
        "brouer@...hat.com" <brouer@...hat.com>,
        "bpf@...r.kernel.org" <bpf@...r.kernel.org>
Subject: Re: [PATCH bpf-next v2 1/2] net: xdp: refactor XDP_QUERY_PROG{,_HW}
 to netdev

On Sat, 1 Jun 2019 at 21:57, Jakub Kicinski
<jakub.kicinski@...ronome.com> wrote:
>
> On Fri, 31 May 2019 19:18:17 +0000, Saeed Mahameed wrote:
> > > +   if (!bpf_op || flags & XDP_FLAGS_SKB_MODE)
> > > +           mode = XDP_FLAGS_SKB_MODE;
> > > +
> > > +   curr_mode = dev_xdp_current_mode(dev);
> > > +
> > > +   if (!offload && curr_mode && (mode ^ curr_mode) &
> > > +       (XDP_FLAGS_DRV_MODE | XDP_FLAGS_SKB_MODE)) {
> >
> > if i am reading this correctly this is equivalent to :
> >
> > if (!offload && (curre_mode != mode))
> > offlad is false then curr_mode and mode must be DRV or GENERIC ..
>
> Naw, if curr_mode is not set, i.e. nothing installed now, we don't care
> about the diff.
>
> > better if you keep bitwise operations for actual bitmasks, mode and
> > curr_mode are not bitmask, they can hold one value each .. according to
> > your logic..
>
> Well, they hold one bit each, whether one bit is a bitmap perhaps is
> disputable? :)
>
> I think the logic is fine.
>

Hmm, but changing to:

       if (!offload && curr_mode && mode != curr_mode)

is equal, and to Saeed's point, clearer. I'll go that route in a v3.

> What happened to my request to move the change in behaviour for
> disabling to a separate patch, tho, Bjorn? :)

Actually, I left that out completely. This patch doesn't change the
behavior. After I realized how the flags *should* be used, I don't
think my v1 change makes sense anymore. My v1 patch was to give an
error if you tried to disable, say generic if drv was enabled via
"auto detect/no flags". But this is catched by looking at the flags.

What I did, however, was moving the flags check into change_fd so that
the driver doesn't have to do the check. E.g. the Intel drivers didn't
do correct checking of flags.