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>] [day] [month] [year] [list]
Message-ID: <CAEZbON6Lcj2a4hVeQ-7BwmiCg-B-=ujeQ1aG+SKmWgjbT=aA9w@mail.gmail.com>
Date:   Tue, 5 Mar 2019 09:59:29 -0800
From:   Ravi Chandra Sadineni <ravisadineni@...omium.org>
To:     Daisuke Nojiri <dnojiri@...gle.com>
Cc:     Guenter Roeck <groeck@...gle.com>,
        Benson Leung <bleung@...omium.org>,
        Enric Balletbo i Serra <enric.balletbo@...labora.com>,
        Guenter Roeck <groeck@...omium.org>,
        linux-kernel <linux-kernel@...r.kernel.org>,
        Todd Broch <tbroch@...gle.com>
Subject: Re: [PATCH V2] cros_ec: Expose sysfile to force battery cut-off on shutdown.

Hi Daisuke,

On Tue, Mar 5, 2019 at 9:29 AM Daisuke Nojiri <dnojiri@...gle.com> wrote:
>>
>> +       if (len == 11 && !strncmp(buf, "at-shutdown", len)) {
>> +               param->flags = EC_BATTERY_CUTOFF_FLAG_AT_SHUTDOWN;
>> +       } else {
>> +               kfree(msg);
>> +               return -EINVAL;
>> +       }
>
>
> Why don't we just let the EC decide what flag is legal or illegal? If you hard code 'at-shutdown', other calls to the sysfs will be blocked even if the EC supports other flags.
That might not be a wise option. Allowing random values can have
unintended effects. For example consider what would happen if write
random values to EC today.  Any flag value other than
EC_BATTERY_CUTOFF_FLAG_AT_SHUTDOWN will cause the battery to cutoff
immediately before the system shutdowns.
In future, if EC exposes more attributes, it will not be that
difficult to add support here.
>
> On Mon, Mar 4, 2019 at 5:00 PM Ravi Chandra Sadineni <ravisadineni@...omium.org> wrote:
>>
>> Hi Guenter,
>>
>> On Sun, Mar 3, 2019 at 5:13 PM Guenter Roeck <groeck@...gle.com> wrote:
>> >
>> > On Sun, Mar 3, 2019 at 4:54 PM RaviChandra Sadineni <ravisadineni@...omium.org> wrote:
>> >>
>> >> On chromebooks, power_manager daemon normally shutsdown(S5) the device
>> >> when the battery charge falls below 4% threshold. Chromeos EC then
>> >> normally spends an hour in S5 before hibernating. If the battery charge
>> >> falls below critical threshold in the mean time, EC does a battery cutoff
>> >> instead of hibernating. On some chromebooks, S5 is optimal enough
>> >> resulting in EC hibernating without battery cut-off. This results in
>> >> battery deep-discharging. This is a bad user experience as battery
>> >> has to trickle charge before booting when the A.C is plugged in the next
>> >> time.
>> >>
>> >> This patch exposes a sysfs file for an userland daemon to suggest EC if it
>> >> has to do a battery cut-off instead of hibernating when the system enters
>> >> S5 next time.
>> >>
>> >> Signed-off-by: RaviChandra Sadineni <ravisadineni@...omium.org>
>> >> ---
>> >> V2: Use kstrtobool() instead of kstrtou8() and add documentation.
>> >>
>> >>  .../ABI/testing/sysfs-class-chromeos          |  8 ++++
>> >>  drivers/platform/chrome/cros_ec_sysfs.c       | 37 +++++++++++++++++++
>> >>  2 files changed, 45 insertions(+)
>> >>  create mode 100644 Documentation/ABI/testing/sysfs-class-chromeos
>> >>
>> >> diff --git a/Documentation/ABI/testing/sysfs-class-chromeos b/Documentation/ABI/testing/sysfs-class-chromeos
>> >> new file mode 100644
>> >> index 000000000000..44d3cee6e7ae
>> >> --- /dev/null
>> >> +++ b/Documentation/ABI/testing/sysfs-class-chromeos
>> >> @@ -0,0 +1,8 @@
>> >> +What:           /sys/class/chromeos/cros_ec/cutoff_at_shutdown
>> >> +Date:           February 2019
>> >> +Contact:        Ravi Chandra Sadineni <ravisadineni@...omium.org>
>> >> +Description:
>> >> +                On writing '[Yy1]' to cutoff_at_shutdown property,
>> >> +                kernel sends a host command to EC requesting battery
>> >> +                cutoff on next shutdown. If AC is plugged in before
>> >> +                next shutdown, EC resets this flag.
>> >> diff --git a/drivers/platform/chrome/cros_ec_sysfs.c b/drivers/platform/chrome/cros_ec_sysfs.c
>> >> index f34a50121064..f5168ce8bfc7 100644
>> >> --- a/drivers/platform/chrome/cros_ec_sysfs.c
>> >> +++ b/drivers/platform/chrome/cros_ec_sysfs.c
>> >> @@ -322,14 +322,51 @@ static ssize_t kb_wake_angle_store(struct device *dev,
>> >>         return count;
>> >>  }
>> >>
>> >> +/* Battery cut-off control */
>> >> +static ssize_t cutoff_at_shutdown_store(struct device *dev,
>> >> +                                  struct device_attribute *attr,
>> >> +                                  const char *buf, size_t count)
>> >> +{
>> >> +       struct ec_params_battery_cutoff *param;
>> >> +       struct cros_ec_command *msg;
>> >> +       int ret;
>> >> +       struct cros_ec_dev *ec = container_of(
>> >> +                       dev, struct cros_ec_dev, class_dev);
>> >> +       bool cutoff_battery;
>> >> +
>> >> +       if (kstrtobool(buf, &cutoff_battery))
>> >> +               return -EINVAL;
>> >> +
>> >> +       if (!cutoff_battery)
>> >> +               return count;
>> >> +
>> >
>> >
>> > It is quite unusual to only be able to set this option, but not to be able to reset it. I think that warrants an explanation and reasoning. Also, if clearing the flag is not supported, I would expect an attempt to do so to return an error. There should also be an explanation in the code explaining why the attribute is write-only.
>> Added documentation for the reason behind making this flag write only.
>> Instead of taking a boolean value, made this attribute more generic
>> and took 'at-shutdown' as an option. This way it will be easy to
>> extend this sysfs api in the future (if EC exposes more attributes for
>> this host command).
>> >
>> > Guenter
>> >
>> >>
>> >> +       msg = kmalloc(sizeof(*msg) + EC_HOST_PARAM_SIZE, GFP_KERNEL);
>> >> +       if (!msg)
>> >> +               return -ENOMEM;
>> >> +
>> >> +       param = (struct ec_params_battery_cutoff *)msg->data;
>> >> +       msg->command = EC_CMD_BATTERY_CUT_OFF + ec->cmd_offset;
>> >> +       msg->version = 1;
>> >> +       param->flags = EC_BATTERY_CUTOFF_FLAG_AT_SHUTDOWN;
>> >> +       msg->outsize = sizeof(*param);
>> >> +       msg->insize = 0;
>> >> +       ret = cros_ec_cmd_xfer_status(ec->ec_dev, msg);
>> >> +       kfree(msg);
>> >> +       if (ret < 0)
>> >> +               return ret;
>> >> +       return count;
>> >> +}
>> >> +
>> >>  /* Module initialization */
>> >>
>> >>  static DEVICE_ATTR_RW(reboot);
>> >>  static DEVICE_ATTR_RO(version);
>> >>  static DEVICE_ATTR_RO(flashinfo);
>> >>  static DEVICE_ATTR_RW(kb_wake_angle);
>> >> +static DEVICE_ATTR_WO(cutoff_at_shutdown);
>> >>
>> >>  static struct attribute *__ec_attrs[] = {
>> >> +       &dev_attr_cutoff_battery_at_shutdown.attr,
>> >>         &dev_attr_kb_wake_angle.attr,
>> >>         &dev_attr_reboot.attr,
>> >>         &dev_attr_version.attr,
>> >> --
>> >> 2.20.1
>> >>
>> Thanks,
>
>
>>
>> Ravi

Thanks,
Ravi

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ