Skip to content

Commit

Permalink
misc: lis3lv02d: Fix false-positive WARN on various HP models
Browse files Browse the repository at this point in the history
commit 3641762 upstream.

Before this commit lis3lv02d_get_pwron_wait() had a WARN_ONCE() to catch
a potential divide by 0. WARN macros should only be used to catch internal
kernel bugs and that is not the case here. We have been receiving a lot of
bug reports about kernel backtraces caused by this WARN.

The div value being checked comes from the lis3->odrs[] array. Which
is sized to be a power-of-2 matching the number of bits in lis3->odr_mask.

The only lis3 model where this array is not entirely filled with non zero
values. IOW the only model where we can hit the div == 0 check is the
3dc ("8 bits 3DC sensor") model:

int lis3_3dc_rates[16] = {0, 1, 10, 25, 50, 100, 200, 400, 1600, 5000};

Note the 0 value at index 0, according to the datasheet an odr index of 0
means "Power-down mode". HP typically uses a lis3 accelerometer for HDD
fall protection. What I believe is happening here is that on newer
HP devices, which only contain a SDD, the BIOS is leaving the lis3 device
powered-down since it is not used for HDD fall protection.

Note that the lis3_3dc_rates array initializer only specifies 10 values,
which matches the datasheet. So it also contains 6 zero values at the end.

Replace the WARN with a normal check, which treats an odr index of 0
as power-down and uses a normal dev_err() to report the error in case
odr index point past the initialized part of the array.

Fixes: 1510dd5 ("lis3lv02d: avoid divide by zero due to unchecked")
Cc: stable@vger.kernel.org
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=785814
BugLink: https://bugzilla.redhat.com/show_bug.cgi?id=1817027
BugLink: https://bugs.centos.org/view.php?id=10720
Link: https://lore.kernel.org/r/20210217102501.31758-1-hdegoede@redhat.com
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
  • Loading branch information
jwrdegoede authored and gregkh committed May 14, 2021
1 parent cb506a1 commit 709e692
Showing 1 changed file with 16 additions and 5 deletions.
21 changes: 16 additions & 5 deletions drivers/misc/lis3lv02d/lis3lv02d.c
Original file line number Diff line number Diff line change
Expand Up @@ -208,23 +208,31 @@ static int lis3_3dc_rates[16] = {0, 1, 10, 25, 50, 100, 200, 400, 1600, 5000};
static int lis3_3dlh_rates[4] = {50, 100, 400, 1000};

/* ODR is Output Data Rate */
static int lis3lv02d_get_odr(struct lis3lv02d *lis3)
static int lis3lv02d_get_odr_index(struct lis3lv02d *lis3)
{
u8 ctrl;
int shift;

lis3->read(lis3, CTRL_REG1, &ctrl);
ctrl &= lis3->odr_mask;
shift = ffs(lis3->odr_mask) - 1;
return lis3->odrs[(ctrl >> shift)];
return (ctrl >> shift);
}

static int lis3lv02d_get_pwron_wait(struct lis3lv02d *lis3)
{
int div = lis3lv02d_get_odr(lis3);
int odr_idx = lis3lv02d_get_odr_index(lis3);
int div = lis3->odrs[odr_idx];

if (WARN_ONCE(div == 0, "device returned spurious data"))
if (div == 0) {
if (odr_idx == 0) {
/* Power-down mode, not sampling no need to sleep */
return 0;
}

dev_err(&lis3->pdev->dev, "Error unknown odrs-index: %d\n", odr_idx);
return -ENXIO;
}

/* LIS3 power on delay is quite long */
msleep(lis3->pwron_delay / div);
Expand Down Expand Up @@ -816,9 +824,12 @@ static ssize_t lis3lv02d_rate_show(struct device *dev,
struct device_attribute *attr, char *buf)
{
struct lis3lv02d *lis3 = dev_get_drvdata(dev);
int odr_idx;

lis3lv02d_sysfs_poweron(lis3);
return sprintf(buf, "%d\n", lis3lv02d_get_odr(lis3));

odr_idx = lis3lv02d_get_odr_index(lis3);
return sprintf(buf, "%d\n", lis3->odrs[odr_idx]);
}

static ssize_t lis3lv02d_rate_set(struct device *dev,
Expand Down

0 comments on commit 709e692

Please sign in to comment.