linux-kernel - Re: [PATCH] PM / wakeup: Avoid dev

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: <CANA+-vDNkSG39SLd0_B_L8gkDkfQ2uJLcuvCaRgHkU8DiBzMTw@mail.gmail.com>
Date:   Sat, 27 Jul 2019 10:37:07 -0700
From:   Tri Vo <trong@...roid.com>
To:     "Rafael J. Wysocki" <rafael@...nel.org>
Cc:     Stephen Boyd <swboyd@...omium.org>,
        "Rafael J. Wysocki" <rjw@...ysocki.net>,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        Linux PM <linux-pm@...r.kernel.org>,
        Kalesh Singh <kaleshsingh@...gle.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        Ravi Chandra Sadineni <ravisadineni@...omium.org>
Subject: Re: [PATCH] PM / wakeup: Avoid dev_name collisions in wakeup class

On Sat, Jul 27, 2019 at 6:10 AM Rafael J. Wysocki <rafael@...nel.org> wrote:
>
> On Sat, Jul 27, 2019 at 3:11 AM Stephen Boyd <swboyd@...omium.org> wrote:
> >
> > If a device is wakeup capable and the driver calls device_wakeup_init()
> > on it during probe and then userspace writes 'enabled' to that device's
> > power/wakeup file in sysfs we'll try to create the same named wakeup
> > device in sysfs. The kernel will complain about duplicate file names.

Thanks for reporting the issue, Stephen!
> >
> > sysfs: cannot create duplicate filename '/devices/virtual/wakeup/1-1.1'
> > kobject_add_internal failed for 1-1.1 with -EEXIST, don't try to register things with the same name in the same directory.
> >
> > It may be advantageous to not write 'enabled' to the wakeup file (see
> > wakeup_store()) from userspace for these devices because we allocate
> > devices and register them and then throw them all away later on if the
> > device driver has already initialized the wakeup attribute. The
> > implementation currently tries to avoid taking locks here so it seems
> > best to optimize that path in a separate patch.
> >
> > Let's rename the wakeup class devices as 'wakeupN' with an IDA that's
> > simple enough to just return some sort of number. In addition, let's
> > make the device registering the wakeup the parent and include a 'name'
> > attribute in case userspace wants to figure out the type of wakeup it is
> > (in the case of virtual wakeups) or the device associated with the
> > wakeup. This makes it easier for userspace to go from /sys/class/wakeup
> > to a place in the device hierarchy where the wakeup is generated from
> > like an input device.
> >
> > Cc: Tri Vo <trong@...roid.com>
> > Cc: Kalesh Singh <kaleshsingh@...gle.com>
> > Cc: Greg Kroah-Hartman <gregkh@...uxfoundation.org>
> > Cc: Ravi Chandra Sadineni <ravisadineni@...omium.org>
> > Signed-off-by: Stephen Boyd <swboyd@...omium.org>
>
> I'd rather change the commit that introduced this issue which is only
> in linux-next for now.

Raphael, could you roll back my patch? I'll work with Stephen to fix it.