On Sep 28 2015 or thereabouts, Mika Westerberg wrote: > Some newer Intel Skylake based Dell laptops with Win8 precision touchpad > fail when initial feature reports are fetched from it. Below is an example > output with some additional debug included: > > i2c_hid i2c-DLL0704:01: Fetching the HID descriptor > i2c_hid i2c-DLL0704:01: __i2c_hid_command: cmd=20 00 > i2c_hid i2c-DLL0704:01: HID Descriptor: 1e 00 00 01 99 02 21 00 24 ... > ... > i2c_hid i2c-DLL0704:01: i2c_hid_get_report > i2c_hid i2c-DLL0704:01: __i2c_hid_command: cmd=22 00 38 02 23 00 > i2c_hid i2c-DLL0704:01: report (len=4): 04 00 08 05 > i2c_hid i2c-DLL0704:01: report id 13 > i2c_hid i2c-DLL0704:01: i2c_hid_get_report > i2c_hid i2c-DLL0704:01: __i2c_hid_command: cmd=22 00 3d 02 23 00 > i2c_hid i2c-DLL0704:01: failed to retrieve report from device. > i2c_hid i2c-DLL0704:01: report id 7 > i2c_hid i2c-DLL0704:01: i2c_hid_get_report > i2c_hid i2c-DLL0704:01: __i2c_hid_command: cmd=22 00 37 02 23 00 > i2c_hid i2c-DLL0704:01: report (len=259): 03 01 07 fc 28 fe 84 40 ... > i2c_hid i2c-DLL0704:01: report id 4 > i2c_hid i2c-DLL0704:01: i2c_hid_get_report > i2c_hid i2c-DLL0704:01: __i2c_hid_command: cmd=22 00 34 02 23 00 > > We manage to fetch few reports but then the touchpad dies: > > i2c_designware i2c_designware.1: i2c_dw_handle_tx_abort: lost arbitration > i2c_hid i2c-DLL0704:01: failed to retrieve report from device. > > it eventually pulls the whole I2C bus low: > > i2c_designware i2c_designware.1: controller timed out > i2c_hid i2c-DLL0704:01: failed to set a report to device. > > Fix this by preventing initial feature report retrieval for Win8 devices. > Instead we fetch reports as needed in mt_feature_mapping(). This prevents > fetching reports which might cause problems with the device in question. > > Suggested-by: Benjamin Tissoires <benjamin.tissoires@xxxxxxxxxx> > Signed-off-by: Mika Westerberg <mika.westerberg@xxxxxxxxxxxxxxx> > --- > This is a second version. The first one which just prevented initial report > reading can be found here: > > https://lkml.org/lkml/2015/9/24/191 > > This version tries to do what Benjamin suggested by reading feature values > as needed in mt_feature_mapping(). This fixes the issue in the Dell laptop > I have here. Thanks for the re-spin Mika. I just have one comment that needs discussion: > > drivers/hid/hid-multitouch.c | 42 +++++++++++++++++++++++++++++++++++++++++- > 1 file changed, 41 insertions(+), 1 deletion(-) > > diff --git a/drivers/hid/hid-multitouch.c b/drivers/hid/hid-multitouch.c > index 661b4fce1a5d..d9ca2b96ab68 100644 > --- a/drivers/hid/hid-multitouch.c > +++ b/drivers/hid/hid-multitouch.c > @@ -309,6 +309,38 @@ static struct attribute_group mt_attribute_group = { > .attrs = sysfs_attrs > }; > > +static void mt_get_feature(struct hid_device *hdev, struct hid_report *report) > +{ > + int ret, size = hid_report_len(report); > + u8 *buf; > + > + /* > + * Only fetch the feature report if initial reports are not already > + * been retrieved. Currently this is only done for Windows 8 touch > + * devices. > + */ > + if (!(hdev->quirks & HID_QUIRK_NO_INIT_REPORTS)) > + return; Some Win 7 panels have this quirk set too because they do not support either querying random feature or input reports. I think this test is safe given that this is, I think, what the Windows driver does (I remember that it queries for the features it needs). However I am not 100% sure if this will not break one low quality touchscreen. I would be in favor of applying the patch in its current shape, but if Jiri thinks we need a little bit more caution here, we would need to add a test regarding the Win8 capability here. Jiri? If we apply it, this patch is: Reviewed-by: Benjamin Tissoires <benjamin.tissoires@xxxxxxxxxx> Cheers, Benjamin > + > + buf = hid_alloc_report_buf(report, GFP_KERNEL); > + if (!buf) > + return; > + > + ret = hid_hw_raw_request(hdev, report->id, buf, size, > + HID_FEATURE_REPORT, HID_REQ_GET_REPORT); > + if (ret < 0) { > + dev_warn(&hdev->dev, "failed to fetch feature %d\n", > + report->id); > + } else { > + ret = hid_report_raw_event(hdev, HID_FEATURE_REPORT, buf, > + size, 0); > + if (ret) > + dev_warn(&hdev->dev, "failed to report feature\n"); > + } > + > + kfree(buf); > +} > + > static void mt_feature_mapping(struct hid_device *hdev, > struct hid_field *field, struct hid_usage *usage) > { > @@ -327,6 +359,8 @@ static void mt_feature_mapping(struct hid_device *hdev, > > break; > case HID_DG_CONTACTMAX: > + mt_get_feature(hdev, field->report); > + > td->maxcontact_report_id = field->report->id; > td->maxcontacts = field->value[0]; > if (!td->maxcontacts && > @@ -343,6 +377,7 @@ static void mt_feature_mapping(struct hid_device *hdev, > break; > } > > + mt_get_feature(hdev, field->report); > if (field->value[usage->usage_index] == MT_BUTTONTYPE_CLICKPAD) > td->is_buttonpad = true; > > @@ -1029,8 +1064,13 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id) > * reports. Fortunately, the Win8 spec says that all touches > * should be sent during each report, making the initialization > * of input reports unnecessary. > + * > + * In addition some touchpads do not behave well if we read > + * all feature reports from them. Instead we prevent > + * initial report fetching and then selectively fetch each > + * report we are interested in. > */ > - hdev->quirks |= HID_QUIRK_NO_INIT_INPUT_REPORTS; > + hdev->quirks |= HID_QUIRK_NO_INIT_REPORTS; > > td = devm_kzalloc(&hdev->dev, sizeof(struct mt_device), GFP_KERNEL); > if (!td) { > -- > 2.5.1 > -- To unsubscribe from this list: send the line "unsubscribe linux-input" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
