linux-kernel - Re: Converting DEVICE_ATTR to DEVICE_ATTR_{RO,RW,WO} and changing function names at the same time

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: <20161222173341.GA17317@kroah.com>
Date:   Thu, 22 Dec 2016 18:33:41 +0100
From:   Greg KH <gregkh@...uxfoundation.org>
To:     Julia Lawall <julia.lawall@...6.fr>
Cc:     Guenter Roeck <linux@...ck-us.net>, cocci@...teme.lip6.fr,
        linux-hwmon@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: Converting DEVICE_ATTR to DEVICE_ATTR_{RO,RW,WO} and changing
 function names at the same time

On Thu, Dec 22, 2016 at 04:45:45PM +0100, Julia Lawall wrote:
> 
> 
> On Thu, 22 Dec 2016, Guenter Roeck wrote:
> 
> > On 12/22/2016 04:29 AM, Julia Lawall wrote:
> > >
> > >
> > > On Wed, 21 Dec 2016, Guenter Roeck wrote:
> > >
> > > > Hi Julia,
> > > >
> > > > On Wed, Dec 21, 2016 at 08:39:38PM +0100, Julia Lawall wrote:
> > > > >
> > > > >
> > > > > On Wed, 21 Dec 2016, Guenter Roeck wrote:
> > > > >
> > > > > > Hi Julia,
> > > > > >
> > > > > > On Wed, Dec 21, 2016 at 03:05:37PM +0100, Julia Lawall wrote:
> > > > > > > A solution is below: the semantic patch, an explanation of the
> > > > > > > semantic
> > > > > > > patch, and the results.  I have only tried to compile the results
> > > > > > > (make
> > > > > > > drivers/hwmon/).  Two affected files were not considered for
> > > > > > > compilation:
> > > > > > >
> > > > > > > drivers/hwmon/vexpress-hwmon.o
> > > > > > > drivers/hwmon/jz4740-hwmon.o
> > > >
> > > > I compile tested those two patches. If possible please drop
> > > > vexpress-hwmon.c
> > > > from the patch series; the changes in that file don't add any value.
> > > >
> > > > I compile tested all files, and reviewed the patch. It all looks good.
> > > > Please submit the series.
> > > >
> > > > Again, thanks a lot for your help!
> > >
> > > I have sent the patches.  I adjusted the semantic patch so that the
> > > indentation of function parameters/arguments would only change if the
> > > length of the function name changes.
> > >
> > > Do you think this could be of more general interest in the Linux kernel?
> > > Since the semantic patch works pretty well, I could add it to the
> > > scripts/coccinelle directory?  Previously, however, I got some negative
> > > feedback about this change, because people felt that the new names hid the
> > > actual behavior, so I didn't pursue it.
> > >
> >
> > I do think it would add a lot of value, if for nothing else as an excellent
> > example
> > of what can be done with coccinelle.
> >
> > I actually liked the name changes. I think it is a good idea if the function
> > name
> > reflects the sysfs attribute it serves (isn't that exactly what it does, ie
> > its
> > behavior ?). But, as you have experienced, some people inadvertently did not
> > like
> > it. Given that, I am not sure if it is worth adding it to the kernel source
> > tree.
> > Maybe you could submit it as RFC so it is at least on record.
> >
> > Anyway, for SENSOR_DEVICE_ATTR(), I'll have to be a bit more flexible since
> > the function _will_ be reused. I'll need something like
> > 	SENSOR_DEVICE_ATTR_{RO,RW,WO}(attr, func, param)
> 
> Chosen at random,
> 
> static SENSOR_DEVICE_ATTR_2(sf2_point4_fan1, S_IRUGO | S_IWUSR,
>                         show_sf2_point, store_sf2_point, 4, 1);
> 
> should become
> 
> static SENSOR_DEVICE_ATTR_2_RW(sf2_point4_fan1, sf2_point, 4, 1); ?
> 
> And the functions should be renamed with show and store at the end?
> 
> > Maybe Greg would be open to something like
> > DEVICE_ATTR_FUNC_{RO,RW,WO}(attr,func)
> > to accommodate the "I want my own function name" crowd ? That would also solve
> > the case where the function is reused for multiple attributes.
> 
> Actually, it was the DEVICE_ATTR_{RO,RW,WO} that wasn't liked.  It doesn't
> show the exact permission numbers.  The fact that not all DEVICE_ATTR uses
> can be changed due to function reuse is awkward, though.  Greg, do you
> have any thoughts about that?

I think the large majority can, and should, be switched to use the
RO,RW,WO variants.  Developers should never be having to specify sysfs
file permissions if at all possible, as they get them wrong too many
times, as we have found out in the past...

> Currently, there are around 1100 calls to DEVICE_ATTR_{RO,RW,WO}.

Those are fine, why do people object to them?

Yes, reusing the function names is not possible, but note, that is a
usually a rare case, the "normal" case for a sysfs attribute can use
these macros just fine.

thanks,

greg k-h