linux-kernel - Re: [PATCH 0/4] regulator: add and use a helper for setting supply names

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:   Fri, 30 Aug 2019 13:06:00 +0200
From:   Bartosz Golaszewski <brgl@...ev.pl>
To:     Thierry Reding <thierry.reding@...il.com>
Cc:     Jens Axboe <axboe@...nel.dk>,
        Jonathan Hunter <jonathanh@...dia.com>,
        JC Kuo <jckuo@...dia.com>,
        Kishon Vijay Abraham I <kishon@...com>,
        Liam Girdwood <lgirdwood@...il.com>,
        Mark Brown <broonie@...nel.org>,
        Mathias Nyman <mathias.nyman@...el.com>,
        Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
        linux-ide@...r.kernel.org, linux-tegra@...r.kernel.org,
        Linux Kernel Mailing List <linux-kernel@...r.kernel.org>,
        linux-usb@...r.kernel.org,
        Bartosz Golaszewski <bgolaszewski@...libre.com>
Subject: Re: [PATCH 0/4] regulator: add and use a helper for setting supply names

pt., 30 sie 2019 o 10:15 Thierry Reding <thierry.reding@...il.com> napisał(a):
>
> On Fri, Aug 30, 2019 at 09:17:36AM +0200, Bartosz Golaszewski wrote:
> > From: Bartosz Golaszewski <bgolaszewski@...libre.com>
> >
> > There are ~80 users of regulator bulk APIs that set the supply names
> > in a for loop before using the bulk helpers.
> >
> > This series proposes to add a helper function for setting supply names
> > and uses it in a couple tegra drivers. If accepted: a coccinelle script
> > should be easy to develop that would address this in other drivers.
> >
> > Bartosz Golaszewski (4):
> >   regulator: provide regulator_bulk_set_supply_names()
> >   ahci: tegra: use regulator_bulk_set_supply_names()
> >   phy: tegra: use regulator_bulk_set_supply_names()
> >   usb: host: xhci-tegra: use regulator_bulk_set_supply_names()
> >
> >  drivers/ata/ahci_tegra.c           |  6 +++---
> >  drivers/phy/tegra/xusb.c           |  6 +++---
> >  drivers/regulator/helpers.c        | 21 +++++++++++++++++++++
> >  drivers/usb/host/xhci-tegra.c      |  5 +++--
> >  include/linux/regulator/consumer.h | 12 ++++++++++++
> >  5 files changed, 42 insertions(+), 8 deletions(-)
>
> The diffstat here suggests that this isn't really helpful. Even if you
> subtract the regulator code that adds the new helper, you actually have
> a zero diffstat (and a positive one in one case).
>

You are right when it comes to LoC stats, but bloat-o-meter says it's
quite useful when it comes to actual code size.

This is what we're adding:

$ ./scripts/bloat-o-meter drivers/regulator/helpers.o.old
drivers/regulator/helpers.o
add/remove: 3/0 grow/shrink: 0/0 up/down: 72/0 (72)
Function                                     old     new   delta
regulator_bulk_set_supply_names                -      32     +32
__kstrtab_regulator_bulk_set_supply_names       -      32     +32
__ksymtab_regulator_bulk_set_supply_names       -       8      +8
Total: Before=4438, After=4510, chg +1.62%

And this is what we're removing just from one driver:

$ ./scripts/bloat-o-meter ./drivers/ata/ahci_tegra.o.old
./drivers/ata/ahci_tegra.o
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-52 (-52)
Function                                     old     new   delta
tegra_ahci_probe                            1688    1636     -52
Total: Before=3474, After=3422, chg -1.50%

> Instead, why don't we take this one step further and roll a bit more of
> the boilerplate into the new helper, something along these lines:
>
>         struct regulator_bulk_data *
>         devm_regulator_bulk_get_from_names(struct device *dev,
>                                            const char * const *supply_names,
>                                            unsigned int num_supplies)
>         {
>                 struct regulator_bulk_data *data;
>                 unsigned int i;
>                 int err;
>
>                 data = devm_kcalloc(dev, num_supplies, sizeof(*data));
>                 if (!data)
>                         return ERR_PTR(-ENOMEM);
>
>                 for (i = 0; i < num_supplies; i++)
>                         data[i].supply = supply_names[i];
>
>                 err = devm_regulator_bulk_get(dev, num_supplies, data);
>                 if (err < 0)
>                         return ERR_PTR(err);
>
>                 return data;
>         }
>
> That seems to be still a very common pattern and gets rid of a lot more
> boilerplate, so your diffstat will actually start to be negative at some
> point.
>

This looks like a good candidate for the next step, but I'd say this
change is still worth applying.

Bart

> I suppose one could add a non-managed variant for this as well, but I'm
> counting 6 uses of regulator_bulk_get() vs. ~140 uses of the managed
> variant, so I'm not sure it's worth it.
>
> Thierry