From: "Pali Rohár" <pali.rohar@gmail.com>
To: Hans de Goede <hdegoede@redhat.com>
Cc: Darren Hart <dvhart@infradead.org>,
Andy Shevchenko <andy@infradead.org>,
Henrique de Moraes Holschuh <ibm-acpi@hmh.eng.br>,
Jacek Anaszewski <jacek.anaszewski@gmail.com>,
Pavel Machek <pavel@ucw.cz>,
platform-driver-x86@vger.kernel.org, linux-leds@vger.kernel.org
Subject: Re: [PATCH v8 6/7] platform/x86/dell-laptop: Protect kbd_state against races
Date: Tue, 21 Feb 2017 15:53:00 +0100 [thread overview]
Message-ID: <20170221145300.GL9795@pali> (raw)
In-Reply-To: <ec4bf07b-366b-ae3a-9c08-f0ce85c824dd@redhat.com>
On Tuesday 21 February 2017 15:42:02 Hans de Goede wrote:
> Hi,
>
> On 21-02-17 15:25, Pali Rohár wrote:
> >On Tuesday 21 February 2017 15:18:14 Hans de Goede wrote:
> >>Hi,
> >>
> >>On 21-02-17 15:06, Pali Rohár wrote:
> >>>On Thursday 09 February 2017 16:44:16 Hans de Goede wrote:
> >>>>The kbd led code has multiple entry points each of which modifies the
> >>>>kbd_state by reading it, modifying a copy, writing the copy and on
> >>>>error setting the modified copy writing back the original state.
> >>>>
> >>>>This is racy, so add a mutex protection the read-modify-write cycle
> >>>>on each of the entry points.
> >>>
> >>>Is this mutex really needed? kbd_get_state and kbd_set_state are already
> >>>locked by mutex. Which situation is trying this patch fix?
> >>
> >>Yes this is really necessary, between getting the state
> >>and storing it writing to another sysfs attribute
> >>may cause a change to the state which will then get
> >>overwritten by the write of the earlier gotten state.
> >>
> >>This is a classic read-modify-write race and as such
> >>needs protection.
> >
> >Right, to preserve all changes (by all concurrent modifications) it is
> >needed to do that kbd_get_state() + modify + kbd_set_state() atomically.
> >
> >So this patch fix all concurrent modifications by kernel.
> >
> >But does not fix race condition when both userspace and kernel want to
> >change keyboard brightness settings.
>
> Yes using libsmbios is ALWAYS racy, as said in my previous mails,
> people really should use the proper kernel interfaces rather then
> directly poking hw from userspace, but just because one path is
> racy is not a good reason to not fix the races in another path,
> esp, when that other path is the preferred way to do things
> and is actually the path which all modern desktop environments use.
Yes, I'm not against this kernel fix. I just pointed that it does not
fix race condition between userspace and kernel...
Anyway, patch seems good so add my Reviewed-by.
> Regards,
>
> Hans
>
>
>
> >
> >>Regards,
> >>
> >>Hans
> >>
> >>
> >>>
> >>>>Signed-off-by: Hans de Goede <hdegoede@redhat.com>
> >>>>---
> >>>>Changes in v8:
> >>>>-New patch in v8 of this patch-set
> >>>>---
> >>>>drivers/platform/x86/dell-laptop.c | 112 +++++++++++++++++++++++++------------
> >>>>1 file changed, 76 insertions(+), 36 deletions(-)
> >>>>
> >>>>diff --git a/drivers/platform/x86/dell-laptop.c b/drivers/platform/x86/dell-laptop.c
> >>>>index a2913a5..70951f3 100644
> >>>>--- a/drivers/platform/x86/dell-laptop.c
> >>>>+++ b/drivers/platform/x86/dell-laptop.c
> >>>>@@ -1133,6 +1133,7 @@ static u8 kbd_previous_level;
> >>>>static u8 kbd_previous_mode_bit;
> >>>>
> >>>>static bool kbd_led_present;
> >>>>+static DEFINE_MUTEX(kbd_led_mutex);
> >>>>
> >>>>/*
> >>>> * NOTE: there are three ways to set the keyboard backlight level.
> >>>>@@ -1562,9 +1563,11 @@ static ssize_t kbd_led_timeout_store(struct device *dev,
> >>>> }
> >>>> }
> >>>>
> >>>>+ mutex_lock(&kbd_led_mutex);
> >>>>+
> >>>> ret = kbd_get_state(&state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>> new_state = state;
> >>>> new_state.timeout_value = value;
> >>>>@@ -1572,9 +1575,12 @@ static ssize_t kbd_led_timeout_store(struct device *dev,
> >>>>
> >>>> ret = kbd_set_state_safe(&new_state, &state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>>- return count;
> >>>>+ ret = count;
> >>>>+out:
> >>>>+ mutex_unlock(&kbd_led_mutex);
> >>>>+ return ret;
> >>>>}
> >>>>
> >>>>static ssize_t kbd_led_timeout_show(struct device *dev,
> >>>>@@ -1634,9 +1640,11 @@ static ssize_t kbd_led_triggers_store(struct device *dev,
> >>>> if (trigger[0] != '+' && trigger[0] != '-')
> >>>> return -EINVAL;
> >>>>
> >>>>+ mutex_lock(&kbd_led_mutex);
> >>>>+
> >>>> ret = kbd_get_state(&state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>> if (kbd_triggers_supported)
> >>>> triggers_enabled = kbd_is_trigger_mode_bit(state.mode_bit);
> >>>>@@ -1650,18 +1658,24 @@ static ssize_t kbd_led_triggers_store(struct device *dev,
> >>>> if (strcmp(trigger+1, kbd_led_triggers[i]) != 0)
> >>>> continue;
> >>>> if (trigger[0] == '+' &&
> >>>>- triggers_enabled && (state.triggers & BIT(i)))
> >>>>- return count;
> >>>>+ triggers_enabled && (state.triggers & BIT(i))) {
> >>>>+ ret = count;
> >>>>+ goto out;
> >>>>+ }
> >>>> if (trigger[0] == '-' &&
> >>>>- (!triggers_enabled || !(state.triggers & BIT(i))))
> >>>>- return count;
> >>>>+ (!triggers_enabled || !(state.triggers & BIT(i)))) {
> >>>>+ ret = count;
> >>>>+ goto out;
> >>>>+ }
> >>>> trigger_bit = i;
> >>>> break;
> >>>> }
> >>>> }
> >>>>
> >>>>- if (trigger_bit == -1)
> >>>>- return -EINVAL;
> >>>>+ if (trigger_bit == -1) {
> >>>>+ ret = -EINVAL;
> >>>>+ goto out;
> >>>>+ }
> >>>>
> >>>> new_state = state;
> >>>> if (trigger[0] == '+')
> >>>>@@ -1677,22 +1691,29 @@ static ssize_t kbd_led_triggers_store(struct device *dev,
> >>>> new_state.triggers &= ~BIT(2);
> >>>> }
> >>>> if ((kbd_info.triggers & new_state.triggers) !=
> >>>>- new_state.triggers)
> >>>>- return -EINVAL;
> >>>>+ new_state.triggers) {
> >>>>+ ret = -EINVAL;
> >>>>+ goto out;
> >>>>+ }
> >>>> if (new_state.triggers && !triggers_enabled) {
> >>>> new_state.mode_bit = KBD_MODE_BIT_TRIGGER;
> >>>> kbd_set_level(&new_state, kbd_previous_level);
> >>>> } else if (new_state.triggers == 0) {
> >>>> kbd_set_level(&new_state, 0);
> >>>> }
> >>>>- if (!(kbd_info.modes & BIT(new_state.mode_bit)))
> >>>>- return -EINVAL;
> >>>>+ if (!(kbd_info.modes & BIT(new_state.mode_bit))) {
> >>>>+ ret = -EINVAL;
> >>>>+ goto out;
> >>>>+ }
> >>>> ret = kbd_set_state_safe(&new_state, &state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>> if (new_state.mode_bit != KBD_MODE_BIT_OFF)
> >>>> kbd_previous_mode_bit = new_state.mode_bit;
> >>>>- return count;
> >>>>+ ret = count;
> >>>>+out:
> >>>>+ mutex_unlock(&kbd_led_mutex);
> >>>>+ return ret;
> >>>>}
> >>>>
> >>>>static ssize_t kbd_led_triggers_show(struct device *dev,
> >>>>@@ -1749,12 +1770,16 @@ static ssize_t kbd_led_als_enabled_store(struct device *dev,
> >>>> if (ret)
> >>>> return ret;
> >>>>
> >>>>+ mutex_lock(&kbd_led_mutex);
> >>>>+
> >>>> ret = kbd_get_state(&state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>>- if (enable == kbd_is_als_mode_bit(state.mode_bit))
> >>>>- return count;
> >>>>+ if (enable == kbd_is_als_mode_bit(state.mode_bit)) {
> >>>>+ ret = count;
> >>>>+ goto out;
> >>>>+ }
> >>>>
> >>>> new_state = state;
> >>>>
> >>>>@@ -1774,15 +1799,20 @@ static ssize_t kbd_led_als_enabled_store(struct device *dev,
> >>>> new_state.mode_bit = KBD_MODE_BIT_ON;
> >>>> }
> >>>> }
> >>>>- if (!(kbd_info.modes & BIT(new_state.mode_bit)))
> >>>>- return -EINVAL;
> >>>>+ if (!(kbd_info.modes & BIT(new_state.mode_bit))) {
> >>>>+ ret = -EINVAL;
> >>>>+ goto out;
> >>>>+ }
> >>>>
> >>>> ret = kbd_set_state_safe(&new_state, &state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>> kbd_previous_mode_bit = new_state.mode_bit;
> >>>>
> >>>>- return count;
> >>>>+ ret = count;
> >>>>+out:
> >>>>+ mutex_unlock(&kbd_led_mutex);
> >>>>+ return ret;
> >>>>}
> >>>>
> >>>>static ssize_t kbd_led_als_enabled_show(struct device *dev,
> >>>>@@ -1817,18 +1847,23 @@ static ssize_t kbd_led_als_setting_store(struct device *dev,
> >>>> if (ret)
> >>>> return ret;
> >>>>
> >>>>+ mutex_lock(&kbd_led_mutex);
> >>>>+
> >>>> ret = kbd_get_state(&state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>> new_state = state;
> >>>> new_state.als_setting = setting;
> >>>>
> >>>> ret = kbd_set_state_safe(&new_state, &state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>>
> >>>>- return count;
> >>>>+ ret = count;
> >>>>+out:
> >>>>+ mutex_unlock(&kbd_led_mutex);
> >>>>+ return ret;
> >>>>}
> >>>>
> >>>>static ssize_t kbd_led_als_setting_show(struct device *dev,
> >>>>@@ -1913,27 +1948,32 @@ static int kbd_led_level_set(struct led_classdev *led_cdev,
> >>>> u16 num;
> >>>> int ret;
> >>>>
> >>>>+ mutex_lock(&kbd_led_mutex);
> >>>>+
> >>>> if (kbd_get_max_level()) {
> >>>> ret = kbd_get_state(&state);
> >>>> if (ret)
> >>>>- return ret;
> >>>>+ goto out;
> >>>> new_state = state;
> >>>> ret = kbd_set_level(&new_state, value);
> >>>> if (ret)
> >>>>- return ret;
> >>>>- return kbd_set_state_safe(&new_state, &state);
> >>>>- }
> >>>>-
> >>>>- if (kbd_get_valid_token_counts()) {
> >>>>+ goto out;
> >>>>+ ret = kbd_set_state_safe(&new_state, &state);
> >>>>+ } else if (kbd_get_valid_token_counts()) {
> >>>> for (num = kbd_token_bits; num != 0 && value > 0; --value)
> >>>> num &= num - 1; /* clear the first bit set */
> >>>> if (num == 0)
> >>>>- return 0;
> >>>>- return kbd_set_token_bit(ffs(num) - 1);
> >>>>+ ret = 0;
> >>>>+ else
> >>>>+ ret = kbd_set_token_bit(ffs(num) - 1);
> >>>>+ } else {
> >>>>+ pr_warn("Keyboard brightness level control not supported\n");
> >>>>+ ret = -ENXIO;
> >>>> }
> >>>>
> >>>>- pr_warn("Keyboard brightness level control not supported\n");
> >>>>- return -ENXIO;
> >>>>+out:
> >>>>+ mutex_unlock(&kbd_led_mutex);
> >>>>+ return ret;
> >>>>}
> >>>>
> >>>>static struct led_classdev kbd_led = {
> >>>
> >
--
Pali Rohár
pali.rohar@gmail.com
next prev parent reply other threads:[~2017-02-21 14:53 UTC|newest]
Thread overview: 56+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-02-09 15:44 [PATCH v8 0/7] platform/x86: Notify userspace about hotkeys changing kbd-backlight brightness Hans de Goede
2017-02-09 15:44 ` [PATCH v8 1/7] platform/x86/thinkpad_acpi: Stop setting led_classdev brightness directly Hans de Goede
2017-02-09 18:09 ` Henrique de Moraes Holschuh
2017-03-01 11:27 ` Pali Rohár
2017-03-01 11:57 ` Hans de Goede
2017-03-01 12:00 ` Pali Rohár
2017-03-01 12:04 ` Hans de Goede
2017-03-01 14:24 ` Marco Trevisan (Treviño)
2017-02-09 15:44 ` [PATCH v8 2/7] platform/x86/thinkpad_acpi: Use brightness_set_blocking callback for LEDs Hans de Goede
2017-02-09 18:00 ` Henrique de Moraes Holschuh
2017-02-09 15:44 ` [PATCH v8 3/7] platform/x86/thinkpad: Call led_classdev_notify_brightness_hw_changed on kbd brightness change Hans de Goede
2017-02-09 18:08 ` Henrique de Moraes Holschuh
2017-02-13 22:52 ` Andy Shevchenko
2017-02-14 9:25 ` Hans de Goede
2017-02-14 9:33 ` Andy Shevchenko
2017-02-17 3:45 ` Darren Hart
2017-02-14 9:36 ` Pali Rohár
2017-02-09 15:44 ` [PATCH v8 4/7] platform/x86/dell-*: Add a generic dell-laptop notifier chain Hans de Goede
2017-02-21 14:18 ` Pali Rohár
2017-02-09 15:44 ` [PATCH v8 5/7] platform/x86/dell-laptop: Refactor kbd_led_triggers_store() Hans de Goede
2017-02-21 14:02 ` Pali Rohár
2017-02-09 15:44 ` [PATCH v8 6/7] platform/x86/dell-laptop: Protect kbd_state against races Hans de Goede
2017-02-21 14:06 ` Pali Rohár
2017-02-21 14:18 ` Hans de Goede
2017-02-21 14:25 ` Pali Rohár
2017-02-21 14:42 ` Hans de Goede
2017-02-21 14:53 ` Pali Rohár [this message]
2017-02-09 15:44 ` [PATCH v8 7/7] platform/x86/dell-*: Call led_classdev_notify_brightness_hw_changed on kbd brightness change Hans de Goede
2017-02-21 14:11 ` Pali Rohár
2017-02-21 14:40 ` Hans de Goede
2017-02-21 14:50 ` Pali Rohár
2017-02-21 14:56 ` Hans de Goede
2017-02-21 15:13 ` Pali Rohár
2017-02-21 16:14 ` Hans de Goede
2017-02-21 17:08 ` Pali Rohár
2017-02-22 8:36 ` Hans de Goede
2017-02-22 8:49 ` Pali Rohár
2017-02-22 10:24 ` Hans de Goede
2017-02-22 12:01 ` Pali Rohár
2017-02-22 12:20 ` Hans de Goede
2017-03-01 11:15 ` Pali Rohár
2017-03-01 12:02 ` Hans de Goede
2017-03-01 12:55 ` Pali Rohár
2017-03-01 13:58 ` Hans de Goede
2017-03-03 12:00 ` Pali Rohár
2017-03-06 13:39 ` Hans de Goede
2017-03-16 10:11 ` Hans de Goede
2017-02-21 20:47 ` Jacek Anaszewski
2017-02-09 20:21 ` [PATCH v8 0/7] platform/x86: Notify userspace about hotkeys changing kbd-backlight brightness Jacek Anaszewski
2017-02-11 20:08 ` Pavel Machek
2017-03-01 23:10 ` Andy Shevchenko
2017-03-02 14:12 ` Hans de Goede
2017-03-02 14:22 ` Pali Rohár
2017-03-02 14:30 ` Hans de Goede
2017-03-02 14:34 ` Pali Rohár
2017-03-02 15:27 ` Andy Shevchenko
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20170221145300.GL9795@pali \
--to=pali.rohar@gmail.com \
--cc=andy@infradead.org \
--cc=dvhart@infradead.org \
--cc=hdegoede@redhat.com \
--cc=ibm-acpi@hmh.eng.br \
--cc=jacek.anaszewski@gmail.com \
--cc=linux-leds@vger.kernel.org \
--cc=pavel@ucw.cz \
--cc=platform-driver-x86@vger.kernel.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).