2022-09-20 17:13:38

by Johnothan King

[permalink] [raw]
Subject: [PATCH v2] HID: nintendo: check analog user calibration for plausibility

Arne Wendt writes:
Cheap clone controllers may (falsely) report as having a user
calibration for the analog sticks in place, but return
wrong/impossible values for the actual calibration data.
In the present case at mine, the controller reports having a
user calibration in place and successfully executes the read
commands. The reported user calibration however is
min = center = max = 0.

This pull request addresses problems of this kind by checking the
provided user calibration-data for plausibility (min < center < max)
and falling back to the default values if implausible.

I'll note that I was experiencing a crash because of this bug when using
the GuliKit KingKong 2 controller. The crash manifests as a divide by
zero error in the kernel logs:
kernel: divide error: 0000 [#1] PREEMPT SMP NOPTI

Changes in v2:
- Move the plausibility check to joycon_read_stick_calibration() and
have that function return -EINVAL if the check fails.
- In the plausibility check, change >= to ==. hid_field_extract() never
returns a negative value, so a scenario involving min > center or
center > max is impossible.
- To reduce code duplication, move the code for setting default
calibration values into a single function called
joycon_use_default_calibration().

Link: https://github.com/nicman23/dkms-hid-nintendo/pull/25
Link: https://github.com/DanielOgorchock/linux/issues/36
Co-authored-by: Arne Wendt <[email protected]>
Signed-off-by: Johnothan King <[email protected]>
---
drivers/hid/hid-nintendo.c | 58 ++++++++++++++++++++++----------------
1 file changed, 33 insertions(+), 25 deletions(-)

diff --git a/drivers/hid/hid-nintendo.c b/drivers/hid/hid-nintendo.c
index 6028af3c3aae..ca58f5440159 100644
--- a/drivers/hid/hid-nintendo.c
+++ b/drivers/hid/hid-nintendo.c
@@ -760,12 +760,34 @@ static int joycon_read_stick_calibration(struct joycon_ctlr *ctlr, u16 cal_addr,
cal_y->max = cal_y->center + y_max_above;
cal_y->min = cal_y->center - y_min_below;

- return 0;
+ /* check if values are plausible */
+ if (cal_x->min == cal_x->center ||
+ cal_x->center == cal_x->max ||
+ cal_y->min == cal_y->center ||
+ cal_y->center == cal_y->max)
+ ret = -EINVAL;
+
+ return ret;
}

static const u16 DFLT_STICK_CAL_CEN = 2000;
static const u16 DFLT_STICK_CAL_MAX = 3500;
static const u16 DFLT_STICK_CAL_MIN = 500;
+static void joycon_use_default_calibration(struct joycon_ctlr *ctlr,
+ struct joycon_stick_cal *cal_x,
+ struct joycon_stick_cal *cal_y,
+ const char *stick, int ret)
+{
+ hid_warn(ctlr->hdev,
+ "Failed to read %s stick cal, "
+ "using defaults; e=%d\n",
+ stick, ret);
+
+ cal_x->center = cal_y->center = DFLT_STICK_CAL_CEN;
+ cal_x->max = cal_y->max = DFLT_STICK_CAL_MAX;
+ cal_x->min = cal_y->min = DFLT_STICK_CAL_MIN;
+}
+
static int joycon_request_calibration(struct joycon_ctlr *ctlr)
{
u16 left_stick_addr = JC_CAL_FCT_DATA_LEFT_ADDR;
@@ -793,38 +815,24 @@ static int joycon_request_calibration(struct joycon_ctlr *ctlr)
&ctlr->left_stick_cal_x,
&ctlr->left_stick_cal_y,
true);
- if (ret) {
- hid_warn(ctlr->hdev,
- "Failed to read left stick cal, using dflts; e=%d\n",
- ret);

- ctlr->left_stick_cal_x.center = DFLT_STICK_CAL_CEN;
- ctlr->left_stick_cal_x.max = DFLT_STICK_CAL_MAX;
- ctlr->left_stick_cal_x.min = DFLT_STICK_CAL_MIN;
-
- ctlr->left_stick_cal_y.center = DFLT_STICK_CAL_CEN;
- ctlr->left_stick_cal_y.max = DFLT_STICK_CAL_MAX;
- ctlr->left_stick_cal_y.min = DFLT_STICK_CAL_MIN;
- }
+ if (ret)
+ joycon_use_default_calibration(ctlr,
+ &ctlr->left_stick_cal_x,
+ &ctlr->left_stick_cal_y,
+ "left", ret);

/* read the right stick calibration data */
ret = joycon_read_stick_calibration(ctlr, right_stick_addr,
&ctlr->right_stick_cal_x,
&ctlr->right_stick_cal_y,
false);
- if (ret) {
- hid_warn(ctlr->hdev,
- "Failed to read right stick cal, using dflts; e=%d\n",
- ret);
-
- ctlr->right_stick_cal_x.center = DFLT_STICK_CAL_CEN;
- ctlr->right_stick_cal_x.max = DFLT_STICK_CAL_MAX;
- ctlr->right_stick_cal_x.min = DFLT_STICK_CAL_MIN;

- ctlr->right_stick_cal_y.center = DFLT_STICK_CAL_CEN;
- ctlr->right_stick_cal_y.max = DFLT_STICK_CAL_MAX;
- ctlr->right_stick_cal_y.min = DFLT_STICK_CAL_MIN;
- }
+ if (ret)
+ joycon_use_default_calibration(ctlr,
+ &ctlr->right_stick_cal_x,
+ &ctlr->right_stick_cal_y,
+ "right", ret);

hid_dbg(ctlr->hdev, "calibration:\n"
"l_x_c=%d l_x_max=%d l_x_min=%d\n"
--
2.37.3



2022-09-20 22:28:57

by Daniel Ogorchock

[permalink] [raw]
Subject: Re: [PATCH v2] HID: nintendo: check analog user calibration for plausibility

On Tue, Sep 20, 2022 at 12:54 PM Johnothan King
<[email protected]> wrote:
>
> Arne Wendt writes:
> Cheap clone controllers may (falsely) report as having a user
> calibration for the analog sticks in place, but return
> wrong/impossible values for the actual calibration data.
> In the present case at mine, the controller reports having a
> user calibration in place and successfully executes the read
> commands. The reported user calibration however is
> min = center = max = 0.
>
> This pull request addresses problems of this kind by checking the
> provided user calibration-data for plausibility (min < center < max)
> and falling back to the default values if implausible.
>
> I'll note that I was experiencing a crash because of this bug when using
> the GuliKit KingKong 2 controller. The crash manifests as a divide by
> zero error in the kernel logs:
> kernel: divide error: 0000 [#1] PREEMPT SMP NOPTI
>
> Changes in v2:
> - Move the plausibility check to joycon_read_stick_calibration() and
> have that function return -EINVAL if the check fails.
> - In the plausibility check, change >= to ==. hid_field_extract() never
> returns a negative value, so a scenario involving min > center or
> center > max is impossible.
> - To reduce code duplication, move the code for setting default
> calibration values into a single function called
> joycon_use_default_calibration().
>
> Link: https://github.com/nicman23/dkms-hid-nintendo/pull/25
> Link: https://github.com/DanielOgorchock/linux/issues/36
> Co-authored-by: Arne Wendt <[email protected]>
> Signed-off-by: Johnothan King <[email protected]>
> ---
> drivers/hid/hid-nintendo.c | 58 ++++++++++++++++++++++----------------
> 1 file changed, 33 insertions(+), 25 deletions(-)
>

Reviewed-by: Daniel J. Ogorchock <[email protected]>

Thanks for the fix. This seems like a good way around invalid calibration data.

-Daniel

2022-09-20 22:31:42

by Daniel Ogorchock

[permalink] [raw]
Subject: Re: [PATCH v2] HID: nintendo: check analog user calibration for plausibility

> static const u16 DFLT_STICK_CAL_CEN = 2000;
> static const u16 DFLT_STICK_CAL_MAX = 3500;
> static const u16 DFLT_STICK_CAL_MIN = 500;
> +static void joycon_use_default_calibration(struct joycon_ctlr *ctlr,
> + struct joycon_stick_cal *cal_x,
> + struct joycon_stick_cal *cal_y,
> + const char *stick, int ret)
> +{
> + hid_warn(ctlr->hdev,
> + "Failed to read %s stick cal, "
> + "using defaults; e=%d\n",
> + stick, ret);
> +

Sorry, missed this on my first readthrough. I think the coding style
mentions not to break up the logged string into multiple lines, since
it'll harm greppability.

-Daniel

2022-09-21 00:59:06

by Johnothan King

[permalink] [raw]
Subject: Re: [PATCH v2] HID: nintendo: check analog user calibration for plausibility

I'll submit a third version that fixes the warning string. I've also
made another minor tweak to joycon_use_default_calibration() for
the v3 patch.

- Johnothan King

------- Original Message -------
On Tuesday, September 20th, 2022 at 3:06 PM, Daniel Ogorchock <[email protected]> wrote:


> > static const u16 DFLT_STICK_CAL_CEN = 2000;
> > static const u16 DFLT_STICK_CAL_MAX = 3500;
> > static const u16 DFLT_STICK_CAL_MIN = 500;
> > +static void joycon_use_default_calibration(struct joycon_ctlr *ctlr,
> > + struct joycon_stick_cal *cal_x,
> > + struct joycon_stick_cal *cal_y,
> > + const char *stick, int ret)
> > +{
> > + hid_warn(ctlr->hdev,
> > + "Failed to read %s stick cal, "
> > + "using defaults; e=%d\n",
> > + stick, ret);
> > +
>
>
> Sorry, missed this on my first readthrough. I think the coding style
> mentions not to break up the logged string into multiple lines, since
> it'll harm greppability.
>
> -Daniel