linux-kernel - Re: [PATCH v1] driver core: Reject pointless SYNC_STATE

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] [day] [month] [year] [list]

Date:   Wed, 29 Sep 2021 12:01:07 -0700
From:   Saravana Kannan <saravanak@...gle.com>
To:     "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ulf Hansson <ulf.hansson@...aro.org>,
        "Cc: Android Kernel" <kernel-team@...roid.com>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>
Subject: Re: [PATCH v1] driver core: Reject pointless SYNC_STATE_ONLY device links

On Wed, Sep 29, 2021 at 5:12 AM Rafael J. Wysocki <rafael@...nel.org> wrote:
>
> On Wed, Sep 29, 2021 at 2:51 AM Saravana Kannan <saravanak@...gle.com> wrote:
> >
> > SYNC_STATE_ONLY device links intentionally allow cycles because cyclic
> > sync_state() dependencies are valid and necessary.
> >
> > However a SYNC_STATE_ONLY device link where the consumer and the supplier
> > are the same device is pointless because the device link would be deleted
> > as soon as the device probes (because it's also the consumer) and won't
> > affect when the sync_state() callback is called. It's a waste of CPU cycles
> > and memory to create this device link. So reject any attempts to create
> > such a device link.
> >
> > Fixes: 05ef983e0d65 ("driver core: Add device link support for SYNC_STATE_ONLY flag")
> > Reported-by: Ulf Hansson <ulf.hansson@...aro.org>
> > Signed-off-by: Saravana Kannan <saravanak@...gle.com>
> > ---
> >  drivers/base/core.c | 2 +-
> >  1 file changed, 1 insertion(+), 1 deletion(-)
> >
> > diff --git a/drivers/base/core.c b/drivers/base/core.c
> > index 15986cc2fe5e..eed27933ac4d 100644
> > --- a/drivers/base/core.c
> > +++ b/drivers/base/core.c
> > @@ -716,7 +716,7 @@ struct device_link *device_link_add(struct device *consumer,
> >          * SYNC_STATE_ONLY link, we don't check for reverse dependencies
> >          * because it only affects sync_state() callbacks.
> >          */
> > -       if (!device_pm_initialized(supplier)
> > +       if (!device_pm_initialized(supplier) || consumer == supplier
>
> Why do we need to get all the way down to here in order to return NULL
> in the consumer == supplier case?

This is where it used to be checked before the "Fixes commit" so I
added it back here. But sure, I can move it up there.

-Saravana

>
> IMO this should be checked at the beginning along with !consumer and !supplier.
>
> >             || (!(flags & DL_FLAG_SYNC_STATE_ONLY) &&
> >                   device_is_dependent(consumer, supplier))) {
> >                 link = NULL;
> > --
> > 2.33.0.685.g46640cef36-goog
> >