Skip to contentSkip to conversation

atmel_mxt_ts misc

PATCH Superseded Jiada Wang 2 participants 87 messages Mar 31, 2020 – Apr 18, 2020

Subsystems (8)

v10 of 11 Superseded Latest is v11 →

lore
Share & tools
lore

Jiada Wang

author
#0 lore raw

v10 cover atmel_mxt_ts misc

This patch-set forward ports Nick Dyer's work in ndyer/linux github repository as long as some other features and fixes

Balasubramani Vivekanandan (2):
Input: atmel_mxt_ts: Limit the max bytes transferred in an i2c
transaction
Input: atmel_mxt_ts: use gpiod_set_value_cansleep for reset pin

Dean Jenkins (1):
Input: atmel_mxt_ts: return error from
mxt_process_messages_until_invalid()

Deepak Das (6):
Input: Atmel: improve error handling in mxt_start()
Input: Atmel: improve error handling in mxt_initialize()
Input: Atmel: improve error handling in mxt_update_cfg()
Input: Atmel: Improve error handling in mxt_initialize_input_device()
Input: Atmel: handle ReportID "0x00" while processing T5 messages
Input: Atmel: use T44 object to process T5 messages

George G. Davis (1):
input: atmel_mxt_ts: export GPIO reset line via sysfs

Jiada Wang (10):
Input: introduce input_mt_report_slot_inactive
dt-bindings: input: atmel: add suspend mode support
Input: atmel_mxt_ts: Rename mxt_fw_version_show to fw_version_show
Input: atmel_mxt_ts: Rename mxt_hw_version_show to hw_version_show
Input: atmel_mxt_ts: rename mxt_update_fw_store to update_fw_store
dt-bindings: input: atmel: provide name of configuration file
dt-bindings: input: atmel: support to specify input name
Input: atmel_mxt_ts - eliminate data->raw_info_block
dt-bindings: input: atmel: support to set max bytes transferred
Input: atmel_mxt_ts - Fix compilation warning

Karl Tsou (1):
Input: atmel_mxt_ts - add debug for T92 gesture and T93 touch seq msgs

Kautuk Consul (2):
Input: atmel_mxt_ts - Change call-points of mxt_free_* functions
Input: atmel_mxt_ts - rely on calculated_crc rather than file
config_crc

Naveen Chakka (2):
input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen
status
input: atmel_mxt_ts: added sysfs interface to update atmel T38 data

Nick Dyer (26):
Input: atmel_mxt_ts - rework sysfs init/remove
Input: atmel_mxt_ts - only read messages in mxt_acquire_irq() when
necessary
Input: atmel_mxt_ts - split large i2c transfers into blocks
Input: atmel_mxt_ts - output status from T48 Noise Supression
Input: atmel_mxt_ts - output status from T42 Touch Suppression
Input: atmel_mxt_ts - implement T9 vector/orientation support
Input: atmel_mxt_ts - implement T15 Key Array support
Input: atmel_mxt_ts - handle reports from T47 Stylus object
Input: atmel_mxt_ts - implement support for T107 active stylus
Input: atmel_mxt_ts - release touch state during suspend
Input: atmel_mxt_ts - add regulator control support
Input: atmel_mxt_ts - report failures in suspend/resume
Input: atmel_mxt_ts - allow specification of firmware file name
Input: atmel_mxt_ts - handle cfg filename via pdata/sysfs
Input: atmel_mxt_ts - allow input name to be specified in platform
data
Input: atmel_mxt_ts - refactor firmware flash to extract context into
struct
Input: atmel_mxt_ts - refactor code to enter bootloader into separate
func
Input: atmel_mxt_ts - combine bootloader version query with probe
Input: atmel_mxt_ts - improve bootloader state machine handling
Input: atmel_mxt_ts - rename bl_completion to chg_completion
Input: atmel_mxt_ts - make bootloader interrupt driven
Input: atmel_mxt_ts - delay enabling IRQ when not using regulators
Input: atmel_mxt_ts - implement I2C retries
Input: atmel_mxt_ts - orientation is not present in hover
Input: atmel_mxt_ts - implement debug output for messages
Input: atmel_mxt_ts - implement improved debug message interface

Nikhil Ravindran (1):
Input: atmel_mxt_ts: Add support for run self-test routine.

Sanjeev Chugh (1):
Input: atmel_mxt_ts: Implement synchronization during various
operation

karl tsou (1):
Input: atmel_mxt_ts - add config checksum attribute to sysfs

keerthikumarp (1):
input: atmel_mxt_ts: Add Missing Delay for reset handling of Atmel
touch panel controller in detachable displays.
---
v10:
Following commits have been updated
input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status
dt-bindings: input: atmel: add suspend mode support
Input: atmel_mxt_ts: Implement synchronization during various operation

Re-order commits to avoid compilation error

v9:
Following commits have been added
Input: atmel_mxt_ts: rename mxt_update_fw_store to update_fw_store
Input: atmel_mxt_ts: Rename mxt_hw_version_show to hw_version_show
Input: atmel_mxt_ts: Rename mxt_fw_version_show to fw_version_show

Addressed dev_attrs related checkpatch warnings

v8:
Fix checkpatch errors and warnings

Fix issue in commit (" Input: atmel_mxt_ts - only read messages in mxt_acquire_irq() when necessary")
reported by Dmitry Osipenko

Cleanup coding style for commits
Input: atmel_mxt_ts - add regulator control support
Input: atmel_mxt_ts - improve bootloader state machine handling

v7:
Fix regression found when updating firmware
Following commits have been updated to fix regression found when updating firmware
Input: atmel_mxt_ts - improve bootloader state machine handling
Input: atmel_mxt_ts - make bootloader interrupt driven
input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status
Input: atmel_mxt_ts: Implement synchronization during various operation

v6:
Fix issue in commit ("Input: introduce input_mt_report_slot_inactive") reported by kernel test robot

v5:
Following commits have been updated to address warnings & errors reported by kbuild test robot
Input: atmel_mxt_ts - make bootloader interrupt driven
Input: atmel_mxt_ts - add debug for T92 gesture and T93 touch seq msgs

Following commit has been updated
Input: introduce input_mt_report_slot_inactive

v4:
Following commit in v3 patch-set has been removed
Input: switch to use return value of input_mt_report_slot_state

Following commit has been updated to address checkpatch warning
Input: atmel_mxt_ts: Implement synchronization during various operation

v3:
Following commits have been updated compared to v2 patchset
Input: atmel_mxt_ts - implement debug output for messages
- added inline comment
Input: atmel_mxt_ts - add debug for T92 gesture and T93 touch seq msg
- changed dev_info() to dev_dbg()

v2:
Following commit in v1 patchset has been split into two commits
Input: introduce input_mt_report_slot_inactive

Following commits have been updated compared to v1 patchset
Input: atmel_mxt_ts - split large i2c transfers into blocks
Input: atmel_mxt_ts - output status from T42 Touch Suppression

Following commits in v1 patchset have been squashed
Input: touchscreen: Atmel: Add device tree support for T15 key array objects
Input: atmel_mxt_ts - check data->input_dev is not null in mxt_input_sync()
Input: atmel_mxt_ts - check firmware format before entering bootloader
Input: atmel_mxt_ts: update stale use_retrigen_workaround flag
input: atmel_mxt_ts: move bootloader probe from mxt_initialize()
input: Atmel: limit the max bytes transferred while reading T5 messages
Input: atmel_mxt_ts: Use msecs_to_jiffies() instead of HZ
Input: atmel_mxt_ts: Use complete when in_bootloader true
Input: atmel_mxt_ts: Prevent crash due to freeing of input device
input: atmel_mxt_ts: Add NULL check for sysfs attribute debug_msg_attr

Following commits in v1 patchset have been dropped:
Input: atmel_mxt_ts - configure and use gpios as real gpios
Input: touchscreen: Atmel: Enable IRQ_DISABLE_UNLAZY flag for interrupt
Input: atmel_mxt_ts - add memory access interface via sysfs
Input: atmel_mxt_ts: Remove sysfs attributes during driver detach
Input: atmel_mxt_ts: Avoid race condition in freeing of input device

v1: initial version

---
 .../bindings/input/atmel,maxtouch.txt         |   17 +
 MAINTAINERS                                   |    1 +
 drivers/hid/hid-alps.c                        |    3 +-
 drivers/hid/hid-multitouch.c                  |    6 +-
 drivers/input/misc/xen-kbdfront.c             |    2 +-
 drivers/input/mouse/elan_i2c_core.c           |    2 +-
 drivers/input/touchscreen/atmel_mxt_ts.c      | 2270 ++++++++++++++---
 drivers/input/touchscreen/cyttsp4_core.c      |    5 +-
 drivers/input/touchscreen/cyttsp_core.c       |    2 +-
 drivers/input/touchscreen/melfas_mip4.c       |    4 +-
 drivers/input/touchscreen/mms114.c            |    2 +-
 drivers/input/touchscreen/raspberrypi-ts.c    |    2 +-
 drivers/input/touchscreen/stmfts.c            |    2 +-
 include/dt-bindings/input/atmel_mxt_ts.h      |   23 +
 include/linux/input/mt.h                      |    5 +
 15 files changed, 1985 insertions(+), 361 deletions(-)
 create mode 100644 include/dt-bindings/input/atmel_mxt_ts.h
Signature · 1 line
2.17.1
author
#1 lore raw

PATCH v10 1/55 Input: introduce input_mt_report_slot_inactive

input_mt_report_slot_state() ignores the tool when the slot is closed. which has caused a bit of confusion.
This patch introduces input_mt_report_slot_inactive() to report slot inactive state.
replaces all input_mt_report_slot_state() with input_mt_report_slot_inactive() in case of close of slot.

---
 drivers/hid/hid-alps.c                     | 3 +--
 drivers/hid/hid-multitouch.c               | 6 ++----
 drivers/input/misc/xen-kbdfront.c          | 2 +-
 drivers/input/mouse/elan_i2c_core.c        | 2 +-
 drivers/input/touchscreen/atmel_mxt_ts.c   | 7 +++----
 drivers/input/touchscreen/cyttsp4_core.c   | 5 ++---
 drivers/input/touchscreen/cyttsp_core.c    | 2 +-
 drivers/input/touchscreen/melfas_mip4.c    | 4 ++--
 drivers/input/touchscreen/mms114.c         | 2 +-
 drivers/input/touchscreen/raspberrypi-ts.c | 2 +-
 drivers/input/touchscreen/stmfts.c         | 2 +-
 include/linux/input/mt.h                   | 5 +++++
 12 files changed, 21 insertions(+), 21 deletions(-)
drivers/hid/hid-alps.c +1 -2
index ae79a7c66737..36ca1d815d53 100644
@@ -387,8 +387,7 @@ static int u1_raw_event(struct alps_dev *hdata, u8 *data, int size)
 				input_report_abs(hdata->input,
 					ABS_MT_PRESSURE, z);
 			} else {
-				input_mt_report_slot_state(hdata->input,
-					MT_TOOL_FINGER, 0);
+				input_mt_report_slot_inactive(hdata->input);
 			}
 		}
 
drivers/hid/hid-multitouch.c +2 -4
index 362805ddf377..e2ce790ff4a4 100644
@@ -896,7 +896,7 @@ static void mt_release_pending_palms(struct mt_device *td,
 		clear_bit(slotnum, app->pending_palm_slots);
 
 		input_mt_slot(input, slotnum);
-		input_mt_report_slot_state(input, MT_TOOL_PALM, false);
+		input_mt_report_slot_inactive(input);
 
 		need_sync = true;
 	}
@@ -1640,9 +1640,7 @@ static void mt_release_contacts(struct hid_device *hid)
 		if (mt) {
 			for (i = 0; i < mt->num_slots; i++) {
 				input_mt_slot(input_dev, i);
-				input_mt_report_slot_state(input_dev,
-							   MT_TOOL_FINGER,
-							   false);
+				input_mt_report_slot_inactive(input_dev);
 			}
 			input_mt_sync_frame(input_dev);
 			input_sync(input_dev);
drivers/input/misc/xen-kbdfront.c +1 -1
index 24bc5c5d876f..a1bba722b234 100644
@@ -146,7 +146,7 @@ static void xenkbd_handle_mt_event(struct xenkbd_info *info,
 		break;
 
 	case XENKBD_MT_EV_UP:
-		input_mt_report_slot_state(info->mtouch, MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(info->mtouch);
 		break;
 
 	case XENKBD_MT_EV_SYN:
drivers/input/mouse/elan_i2c_core.c +1 -1
index 8719da540383..3f9354baac4b 100644
@@ -938,7 +938,7 @@ static void elan_report_contact(struct elan_tp_data *data,
 		input_report_abs(input, ABS_MT_TOUCH_MINOR, minor);
 	} else {
 		input_mt_slot(input, contact_num);
-		input_mt_report_slot_state(input, MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(input);
 	}
 }
 
drivers/input/touchscreen/atmel_mxt_ts.c +3 -4
index ae60442efda0..a2189739e30f 100644
@@ -822,8 +822,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 		 * have happened.
 		 */
 		if (status & MXT_T9_RELEASE) {
-			input_mt_report_slot_state(input_dev,
-						   MT_TOOL_FINGER, 0);
+			input_mt_report_slot_inactive(input_dev);
 			mxt_input_sync(data);
 		}
 
@@ -839,7 +838,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 		input_report_abs(input_dev, ABS_MT_TOUCH_MAJOR, area);
 	} else {
 		/* Touch no longer active, close out slot */
-		input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, 0);
+		input_mt_report_slot_inactive(input_dev);
 	}
 
 	data->update_input = true;
@@ -947,7 +946,7 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 		dev_dbg(dev, "[%u] release\n", id);
 
 		/* close out slot */
-		input_mt_report_slot_state(input_dev, 0, 0);
+		input_mt_report_slot_inactive(input_dev);
 	}
 
 	data->update_input = true;
drivers/input/touchscreen/cyttsp4_core.c +2 -3
index 6bcffc930384..02a73d9a4def 100644
@@ -744,8 +744,7 @@ static void cyttsp4_report_slot_liftoff(struct cyttsp4_mt_data *md,
 
 	for (t = 0; t < max_slots; t++) {
 		input_mt_slot(md->input, t);
-		input_mt_report_slot_state(md->input,
-			MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(md->input);
 	}
 }
 
@@ -845,7 +844,7 @@ static void cyttsp4_final_sync(struct input_dev *input, int max_slots, int *ids)
 		if (ids[t])
 			continue;
 		input_mt_slot(input, t);
-		input_mt_report_slot_state(input, MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(input);
 	}
 
 	input_sync(input);
drivers/input/touchscreen/cyttsp_core.c +1 -1
index 3f5d463dbeed..697aa2c158f7 100644
@@ -340,7 +340,7 @@ static void cyttsp_report_tchdata(struct cyttsp *ts)
 			continue;
 
 		input_mt_slot(input, i);
-		input_mt_report_slot_state(input, MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(input);
 	}
 
 	input_sync(input);
drivers/input/touchscreen/melfas_mip4.c +2 -2
index 247c3aaba2d8..f67efdd040b2 100644
@@ -391,7 +391,7 @@ static void mip4_clear_input(struct mip4_ts *ts)
 	/* Screen */
 	for (i = 0; i < MIP4_MAX_FINGERS; i++) {
 		input_mt_slot(ts->input, i);
-		input_mt_report_slot_state(ts->input, MT_TOOL_FINGER, 0);
+		input_mt_report_slot_inactive(ts->input);
 	}
 
 	/* Keys */
@@ -534,7 +534,7 @@ static void mip4_report_touch(struct mip4_ts *ts, u8 *packet)
 	} else {
 		/* Release event */
 		input_mt_slot(ts->input, id);
-		input_mt_report_slot_state(ts->input, MT_TOOL_FINGER, 0);
+		input_mt_report_slot_inactive(ts->input);
 	}
 
 	input_mt_sync_frame(ts->input);
drivers/input/touchscreen/mms114.c +1 -1
index 69c6d559eeb0..776ed9e4629d 100644
@@ -549,7 +549,7 @@ static int __maybe_unused mms114_suspend(struct device *dev)
 	/* Release all touch */
 	for (id = 0; id < MMS114_MAX_TOUCH; id++) {
 		input_mt_slot(input_dev, id);
-		input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, false);
+		input_mt_report_slot_inactive(input_dev);
 	}
 
 	input_mt_report_pointer_emulation(input_dev, true);
drivers/input/touchscreen/raspberrypi-ts.c +1 -1
index 0e2e08f3f433..ef6aaed217cf 100644
@@ -100,7 +100,7 @@ static void rpi_ts_poll(struct input_dev *input)
 	released_ids = ts->known_ids & ~modified_ids;
 	for_each_set_bit(i, &released_ids, RPI_TS_MAX_SUPPORTED_POINTS) {
 		input_mt_slot(input, i);
-		input_mt_report_slot_state(input, MT_TOOL_FINGER, 0);
+		input_mt_report_slot_inactive(input);
 		modified_ids &= ~(BIT(i));
 	}
 	ts->known_ids = modified_ids;
drivers/input/touchscreen/stmfts.c +1 -1
index b6f95f20f924..b54cc64e4ea6 100644
@@ -198,7 +198,7 @@ static void stmfts_report_contact_release(struct stmfts_data *sdata,
 	u8 slot_id = (event[0] & STMFTS_MASK_TOUCH_ID) >> 4;
 
 	input_mt_slot(sdata->input, slot_id);
-	input_mt_report_slot_state(sdata->input, MT_TOOL_FINGER, false);
+	input_mt_report_slot_inactive(sdata->input);
 
 	input_sync(sdata->input);
 }
include/linux/input/mt.h +5 -0
index 9e409bb13642..3b8580bd33c1 100644
@@ -100,6 +100,11 @@ static inline bool input_is_mt_axis(int axis)
 bool input_mt_report_slot_state(struct input_dev *dev,
 				unsigned int tool_type, bool active);
 
+static inline void input_mt_report_slot_inactive(struct input_dev *dev)
+{
+	input_mt_report_slot_state(dev, 0, false);
+}
+
 void input_mt_report_finger_count(struct input_dev *dev, int count);
 void input_mt_report_pointer_emulation(struct input_dev *dev, bool use_count);
 void input_mt_drop_unused(struct input_dev *dev);
Signature · 1 line
2.17.1
author
#2 lore raw

PATCH v10 2/55 Input: atmel_mxt_ts - rework sysfs init/remove

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

An error in the sysfs init may otherwise interfere with the async return from the firmware loader

(cherry picked from ndyer/linux/for-upstream commit 3114584ae77c2b03b6dad87174f010d002e9c05d) [gdavis: Forward port and fixup conflicts. Also fixed sysfs leaks in
both the mxt_initialize() and mxt_probe() error return cases.]

[jiada: keep call mxt_initialize() before sysfs creation
replace S_IWUSR with 0200]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 64 +++++++++++++++++++-----
 1 file changed, 52 insertions(+), 12 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +52 -12
index a2189739e30f..49bdf5cf3a0d 100644
@@ -2086,10 +2086,14 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 	return 0;
 
 err_free_mem:
+	data->input_dev = NULL;
 	input_free_device(input_dev);
 	return error;
 }
 
+static int mxt_sysfs_init(struct mxt_data *data);
+static void mxt_sysfs_remove(struct mxt_data *data);
+
 static int mxt_configure_objects(struct mxt_data *data,
 				 const struct firmware *cfg);
 
@@ -2141,16 +2145,24 @@ static int mxt_initialize(struct mxt_data *data)
 	if (error)
 		return error;
 
+	error = mxt_sysfs_init(data);
+	if (error)
+		return error;
+
 	error = request_firmware_nowait(THIS_MODULE, true, MXT_CFG_NAME,
 					&client->dev, GFP_KERNEL, data,
 					mxt_config_cb);
 	if (error) {
 		dev_err(&client->dev, "Failed to invoke firmware loader: %d\n",
 			error);
-		return error;
+		goto err_free_sysfs;
 	}
 
 	return 0;
+
+err_free_sysfs:
+	mxt_sysfs_remove(data);
+	return error;
 }
 
 static int mxt_set_t7_power_cfg(struct mxt_data *data, u8 sleep)
@@ -2803,6 +2815,7 @@ static int mxt_load_fw(struct device *dev, const char *fn)
 		if (ret)
 			goto release_firmware;
 
+		mxt_sysfs_remove(data);
 		mxt_free_input_device(data);
 		mxt_free_object_table(data);
 	} else {
@@ -2909,16 +2922,25 @@ static ssize_t mxt_update_fw_store(struct device *dev,
 	return count;
 }
 
+static DEVICE_ATTR(update_fw, 0200, NULL, mxt_update_fw_store);
+
+static struct attribute *mxt_fw_attrs[] = {
+	&dev_attr_update_fw.attr,
+	NULL
+};
+
+static const struct attribute_group mxt_fw_attr_group = {
+	.attrs = mxt_fw_attrs,
+};
+
 static DEVICE_ATTR(fw_version, S_IRUGO, mxt_fw_version_show, NULL);
 static DEVICE_ATTR(hw_version, S_IRUGO, mxt_hw_version_show, NULL);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
-static DEVICE_ATTR(update_fw, S_IWUSR, NULL, mxt_update_fw_store);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
 	&dev_attr_hw_version.attr,
 	&dev_attr_object.attr,
-	&dev_attr_update_fw.attr,
 	NULL
 };
 
@@ -2926,6 +2948,28 @@ static const struct attribute_group mxt_attr_group = {
 	.attrs = mxt_attrs,
 };
 
+static int mxt_sysfs_init(struct mxt_data *data)
+{
+	struct i2c_client *client = data->client;
+	int error;
+
+	error = sysfs_create_group(&client->dev.kobj, &mxt_attr_group);
+	if (error) {
+		dev_err(&client->dev, "Failure %d creating sysfs group\n",
+			error);
+		return error;
+	}
+
+	return 0;
+}
+
+static void mxt_sysfs_remove(struct mxt_data *data)
+{
+	struct i2c_client *client = data->client;
+
+	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+}
+
 static void mxt_start(struct mxt_data *data)
 {
 	switch (data->suspend_mode) {
@@ -3112,19 +3156,14 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	if (error)
 		return error;
 
-	error = sysfs_create_group(&client->dev.kobj, &mxt_attr_group);
+	error = sysfs_create_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (error) {
-		dev_err(&client->dev, "Failure %d creating sysfs group\n",
+		dev_err(&client->dev, "Failure %d creating fw sysfs group\n",
 			error);
-		goto err_free_object;
+		return error;
 	}
 
 	return 0;
-
-err_free_object:
-	mxt_free_input_device(data);
-	mxt_free_object_table(data);
-	return error;
 }
 
 static int mxt_remove(struct i2c_client *client)
@@ -3132,7 +3171,8 @@ static int mxt_remove(struct i2c_client *client)
 	struct mxt_data *data = i2c_get_clientdata(client);
 
 	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
+	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
Signature · 1 line
2.17.1
author
#3 lore raw

PATCH v10 3/55 Input: atmel_mxt_ts - only read messages in mxt_acquire_irq() when necessary

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

The workaround of reading all messages until an invalid is received is a way of forcing the CHG line high, which means that when using edge-triggered interrupts the interrupt can be acquired.

With level-triggered interrupts the workaround is unnecessary.

Also, most recent maXTouch chips have a feature called RETRIGEN which, when enabled, reasserts the interrupt line every cycle if there are messages waiting. This also makes the workaround unnecessary.

Note: the RETRIGEN feature is only in some firmware versions/chips, it's not valid simply to enable the bit.

(cherry picked from ndyer/linux/for-upstream commit 1ae4e8281e491b22442cd5acdfca1862555f8ecb) [gdavis: Fix conflicts due to v4.6-rc7 commit eb43335c4095 ("Input:
atmel_mxt_ts - use mxt_acquire_irq in mxt_soft_reset").]

[jiada: reset use_retrigen_workaround at beginning of mxt_check_retrigen()
call mxt_check_retrigen() after mxt_acquire_irq() in mxt_initialize() replace white-spaces with tab for MXT_COMMS_RETRIGEN]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 51 ++++++++++++++++++++++--
 1 file changed, 48 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +48 -3
index 49bdf5cf3a0d..3f1ebe14802f 100644
@@ -20,6 +20,7 @@
 #include <linux/i2c.h>
 #include <linux/input/mt.h>
 #include <linux/interrupt.h>
+#include <linux/irq.h>
 #include <linux/of.h>
 #include <linux/property.h>
 #include <linux/slab.h>
@@ -129,6 +130,7 @@ struct t9_range {
 /* MXT_SPT_COMMSCONFIG_T18 */
 #define MXT_COMMS_CTRL		0
 #define MXT_COMMS_CMD		1
+#define MXT_COMMS_RETRIGEN	BIT(6)
 
 /* MXT_DEBUG_DIAGNOSTIC_T37 */
 #define MXT_DIAGNOSTIC_PAGEUP	0x01
@@ -308,6 +310,7 @@ struct mxt_data {
 	struct t7_config t7_cfg;
 	struct mxt_dbg dbg;
 	struct gpio_desc *reset_gpio;
+	bool use_retrigen_workaround;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -318,6 +321,7 @@ struct mxt_data {
 	u16 T71_address;
 	u8 T9_reportid_min;
 	u8 T9_reportid_max;
+	u16 T18_address;
 	u8 T19_reportid;
 	u16 T44_address;
 	u8 T100_reportid_min;
@@ -1190,9 +1194,11 @@ static int mxt_acquire_irq(struct mxt_data *data)
 
 	enable_irq(data->irq);
 
-	error = mxt_process_messages_until_invalid(data);
-	if (error)
-		return error;
+	if (data->use_retrigen_workaround) {
+		error = mxt_process_messages_until_invalid(data);
+		if (error)
+			return error;
+	}
 
 	return 0;
 }
@@ -1282,6 +1288,33 @@ static u32 mxt_calculate_crc(u8 *base, off_t start_off, off_t end_off)
 	return crc;
 }
 
+static int mxt_check_retrigen(struct mxt_data *data)
+{
+	struct i2c_client *client = data->client;
+	int error;
+	int val;
+
+	data->use_retrigen_workaround = false;
+
+	if (irq_get_trigger_type(data->irq) & IRQF_TRIGGER_LOW)
+		return 0;
+
+	if (data->T18_address) {
+		error = __mxt_read_reg(client,
+				       data->T18_address + MXT_COMMS_CTRL,
+				       1, &val);
+		if (error)
+			return error;
+
+		if (val & MXT_COMMS_RETRIGEN)
+			return 0;
+	}
+
+	dev_warn(&client->dev, "Enabling RETRIGEN workaround\n");
+	data->use_retrigen_workaround = true;
+	return 0;
+}
+
 static int mxt_prepare_cfg_mem(struct mxt_data *data, struct mxt_cfg *cfg)
 {
 	struct device *dev = &data->client->dev;
@@ -1561,6 +1594,10 @@ static int mxt_update_cfg(struct mxt_data *data, const struct firmware *fw)
 
 	mxt_update_crc(data, MXT_COMMAND_BACKUPNV, MXT_BACKUP_VALUE);
 
+	ret = mxt_check_retrigen(data);
+	if (ret)
+		goto release_mem;
+
 	ret = mxt_soft_reset(data);
 	if (ret)
 		goto release_mem;
@@ -1604,6 +1641,7 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T71_address = 0;
 	data->T9_reportid_min = 0;
 	data->T9_reportid_max = 0;
+	data->T18_address = 0;
 	data->T19_reportid = 0;
 	data->T44_address = 0;
 	data->T100_reportid_min = 0;
@@ -1678,6 +1716,9 @@ static int mxt_parse_object_table(struct mxt_data *data,
 						object->num_report_ids - 1;
 			data->num_touchids = object->num_report_ids;
 			break;
+		case MXT_SPT_COMMSCONFIG_T18:
+			data->T18_address = object->start_address;
+			break;
 		case MXT_SPT_MESSAGECOUNT_T44:
 			data->T44_address = object->start_address;
 			break;
@@ -2145,6 +2186,10 @@ static int mxt_initialize(struct mxt_data *data)
 	if (error)
 		return error;
 
+	error = mxt_check_retrigen(data);
+	if (error)
+		return error;
+
 	error = mxt_sysfs_init(data);
 	if (error)
 		return error;
Signature · 1 line
2.17.1
author
#4 lore raw

PATCH v10 4/55 Input: atmel_mxt_ts - split large i2c transfers into blocks

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

On some firmware variants, the size of the info block exceeds what can be read in a single transfer.

(cherry picked from ndyer/linux/for-upstream commit 74c4f5277cfa403d43fafc404119dc57a08677db) [gdavis: Forward port and fix conflicts due to v4.14.51 commit

	 960fe000b1d3 ("Input: atmel_mxt_ts - fix the firmware
	 update").]

[jiada: Change mxt_read_blks() to __mxt_read_reg(), original __mxt_read_reg() to

	__mxt_read_chunk()]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 28 +++++++++++++++++++++---
 1 file changed, 25 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +25 -3
index 3f1ebe14802f..7e6a66e3e1e0 100644
@@ -40,7 +40,7 @@
 #define MXT_OBJECT_START	0x07
 #define MXT_OBJECT_SIZE		6
 #define MXT_INFO_CHECKSUM_SIZE	3
-#define MXT_MAX_BLOCK_WRITE	256
+#define MXT_MAX_BLOCK_WRITE	255
 
 /* Object types */
 #define MXT_DEBUG_DIAGNOSTIC_T37	37
@@ -624,8 +624,8 @@ static int mxt_send_bootloader_cmd(struct mxt_data *data, bool unlock)
 	return 0;
 }
 
-static int __mxt_read_reg(struct i2c_client *client,
-			       u16 reg, u16 len, void *val)
+static int __mxt_read_chunk(struct i2c_client *client,
+			    u16 reg, u16 len, void *val)
 {
 	struct i2c_msg xfer[2];
 	u8 buf[2];
@@ -659,6 +659,28 @@ static int __mxt_read_reg(struct i2c_client *client,
 	return ret;
 }
 
+static int __mxt_read_reg(struct i2c_client *client,
+			  u16 reg, u16 len, void *buf)
+{
+	u16 offset = 0;
+	int error;
+	u16 size;
+
+	while (offset < len) {
+		size = min(MXT_MAX_BLOCK_WRITE, len - offset);
+
+		error = __mxt_read_chunk(client,
+					 reg + offset,
+					 size, buf + offset);
+		if (error)
+			return error;
+
+		offset += size;
+	}
+
+	return 0;
+}
+
 static int __mxt_write_reg(struct i2c_client *client, u16 reg, u16 len,
 			   const void *val)
 {
Signature · 1 line
2.17.1
author
#5 lore raw

PATCH v10 5/55 Input: atmel_mxt_ts - output status from T48 Noise Supression

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This patch outputs status from T48 Noise Supression

(cherry picked from ndyer/linux/for-upstream commit 2895a6ff150a49f27a02938f8d262be238b296d8)

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 25 ++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +25 -0
index 7e6a66e3e1e0..a53985a7736f 100644
@@ -324,6 +324,7 @@ struct mxt_data {
 	u16 T18_address;
 	u8 T19_reportid;
 	u16 T44_address;
+	u8 T48_reportid;
 	u8 T100_reportid_min;
 	u8 T100_reportid_max;
 
@@ -978,6 +979,24 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 	data->update_input = true;
 }
 
+static int mxt_proc_t48_messages(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+	u8 status, state;
+
+	status = msg[1];
+	state  = msg[4];
+
+	dev_dbg(dev, "T48 state %d status %02X %s%s%s%s%s\n", state, status,
+		status & 0x01 ? "FREQCHG " : "",
+		status & 0x02 ? "APXCHG " : "",
+		status & 0x04 ? "ALGOERR " : "",
+		status & 0x10 ? "STATCHG " : "",
+		status & 0x20 ? "NLVLCHG " : "");
+
+	return 0;
+}
+
 static int mxt_proc_message(struct mxt_data *data, u8 *message)
 {
 	u8 report_id = message[0];
@@ -987,6 +1006,8 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 
 	if (report_id == data->T6_reportid) {
 		mxt_proc_t6_messages(data, message);
+	} else if (report_id == data->T48_reportid) {
+		mxt_proc_t48_messages(data, message);
 	} else if (!data->input_dev) {
 		/*
 		 * Do not report events if input device
@@ -1666,6 +1687,7 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T18_address = 0;
 	data->T19_reportid = 0;
 	data->T44_address = 0;
+	data->T48_reportid = 0;
 	data->T100_reportid_min = 0;
 	data->T100_reportid_max = 0;
 	data->max_reportid = 0;
@@ -1747,6 +1769,9 @@ static int mxt_parse_object_table(struct mxt_data *data,
 		case MXT_SPT_GPIOPWM_T19:
 			data->T19_reportid = min_id;
 			break;
+		case MXT_PROCG_NOISESUPPRESSION_T48:
+			data->T48_reportid = min_id;
+			break;
 		case MXT_TOUCH_MULTITOUCHSCREEN_T100:
 			data->multitouch = MXT_TOUCH_MULTITOUCHSCREEN_T100;
 			data->T100_reportid_min = min_id;
Signature · 1 line
2.17.1
author
#6 lore raw

PATCH v10 6/55 Input: atmel_mxt_ts - output status from T42 Touch Suppression

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This patch outputs status from T42 touch suppression

(cherry picked from ndyer/linux/for-upstream commit ab95b5a309999d2c098daaa9f88d9fcfae7eb516)

[jiada: Replace dev_info() with dev_dbg()]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 25 ++++++++++++++++++++++++
 1 file changed, 25 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +25 -0
index a53985a7736f..f6465edaa57e 100644
@@ -155,6 +155,9 @@ struct t37_debug {
 #define MXT_RESET_VALUE		0x01
 #define MXT_BACKUP_VALUE	0x55
 
+/* Define for MXT_PROCI_TOUCHSUPPRESSION_T42 */
+#define MXT_T42_MSG_TCHSUP	BIT(0)
+
 /* T100 Multiple Touch Touchscreen */
 #define MXT_T100_CTRL		0
 #define MXT_T100_CFG1		1
@@ -323,6 +326,8 @@ struct mxt_data {
 	u8 T9_reportid_max;
 	u16 T18_address;
 	u8 T19_reportid;
+	u8 T42_reportid_min;
+	u8 T42_reportid_max;
 	u16 T44_address;
 	u8 T48_reportid;
 	u8 T100_reportid_min;
@@ -979,6 +984,17 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 	data->update_input = true;
 }
 
+static void mxt_proc_t42_messages(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+	u8 status = msg[1];
+
+	if (status & MXT_T42_MSG_TCHSUP)
+		dev_dbg(dev, "T42 suppress\n");
+	else
+		dev_dbg(dev, "T42 normal\n");
+}
+
 static int mxt_proc_t48_messages(struct mxt_data *data, u8 *msg)
 {
 	struct device *dev = &data->client->dev;
@@ -1006,6 +1022,9 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 
 	if (report_id == data->T6_reportid) {
 		mxt_proc_t6_messages(data, message);
+	} else if (report_id >= data->T42_reportid_min
+		   && report_id <= data->T42_reportid_max) {
+		mxt_proc_t42_messages(data, message);
 	} else if (report_id == data->T48_reportid) {
 		mxt_proc_t48_messages(data, message);
 	} else if (!data->input_dev) {
@@ -1686,6 +1705,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T9_reportid_max = 0;
 	data->T18_address = 0;
 	data->T19_reportid = 0;
+	data->T42_reportid_min = 0;
+	data->T42_reportid_max = 0;
 	data->T44_address = 0;
 	data->T48_reportid = 0;
 	data->T100_reportid_min = 0;
@@ -1763,6 +1784,10 @@ static int mxt_parse_object_table(struct mxt_data *data,
 		case MXT_SPT_COMMSCONFIG_T18:
 			data->T18_address = object->start_address;
 			break;
+		case MXT_PROCI_TOUCHSUPPRESSION_T42:
+			data->T42_reportid_min = min_id;
+			data->T42_reportid_max = max_id;
+			break;
 		case MXT_SPT_MESSAGECOUNT_T44:
 			data->T44_address = object->start_address;
 			break;
Signature · 1 line
2.17.1
author
#7 lore raw

PATCH v10 7/55 Input: atmel_mxt_ts - implement T9 vector/orientation support

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

The atmel touch messages contain orientation information as a byte in a packed format which can be passed straight on to Android if the input device configuration is correct.

This requires vector reports to be enabled in maXTouch config (zero DISVECT bit 3 in T9 CTRL field)

Android converts the format in InputReader.cpp, search for ORIENTATION_CALIBRATION_VECTOR.

(cherry picked from ndyer/linux/for-upstream commit a6f0ee919d2631678169b23fb18f55b6dbabcd4c)

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +8 -4
index f6465edaa57e..df2e0ba76e63 100644
@@ -817,6 +817,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 	int y;
 	int area;
 	int amplitude;
+	u8 vector;
 
 	id = message[0] - data->T9_reportid_min;
 	status = message[1];
@@ -831,9 +832,10 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 
 	area = message[5];
 	amplitude = message[6];
+	vector = message[7];
 
 	dev_dbg(dev,
-		"[%u] %c%c%c%c%c%c%c%c x: %5u y: %5u area: %3u amp: %3u\n",
+		"[%u] %c%c%c%c%c%c%c%c x: %5u y: %5u area: %3u amp: %3u vector: %02X\n",
 		id,
 		(status & MXT_T9_DETECT) ? 'D' : '.',
 		(status & MXT_T9_PRESS) ? 'P' : '.',
@@ -843,7 +845,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 		(status & MXT_T9_AMP) ? 'A' : '.',
 		(status & MXT_T9_SUPPRESS) ? 'S' : '.',
 		(status & MXT_T9_UNGRIP) ? 'U' : '.',
-		x, y, area, amplitude);
+		x, y, area, amplitude, vector);
 
 	input_mt_slot(input_dev, id);
 
@@ -868,6 +870,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 		input_report_abs(input_dev, ABS_MT_POSITION_Y, y);
 		input_report_abs(input_dev, ABS_MT_PRESSURE, amplitude);
 		input_report_abs(input_dev, ABS_MT_TOUCH_MAJOR, area);
+		input_report_abs(input_dev, ABS_MT_ORIENTATION, vector);
 	} else {
 		/* Touch no longer active, close out slot */
 		input_mt_report_slot_inactive(input_dev);
@@ -2180,8 +2183,9 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 				     0, 255, 0, 0);
 	}
 
-	if (data->multitouch == MXT_TOUCH_MULTITOUCHSCREEN_T100 &&
-	    data->t100_aux_vect) {
+	if (data->multitouch == MXT_TOUCH_MULTI_T9 ||
+	    (data->multitouch == MXT_TOUCH_MULTITOUCHSCREEN_T100 &&
+	    data->t100_aux_vect)) {
 		input_set_abs_params(input_dev, ABS_MT_ORIENTATION,
 				     0, 255, 0, 0);
 	}
Signature · 1 line
2.17.1
author
#8 lore raw

PATCH v10 8/55 Input: atmel_mxt_ts - implement T15 Key Array support

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

There is a key array object in many maXTouch chips which allows some X/Y lines to be used as a key array. This patch maps them to a series of keys which may be configured in a platform data array.

(cherry picked from ndyer/linux/for-upstream commit 15bb074b5abf3a101f7b79544213f1c110ea4cab) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[jiada: Fix compilation warning]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 85 ++++++++++++++++++++++++
 1 file changed, 85 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +85 -0
index df2e0ba76e63..d05249b02781 100644
@@ -314,6 +314,9 @@ struct mxt_data {
 	struct mxt_dbg dbg;
 	struct gpio_desc *reset_gpio;
 	bool use_retrigen_workaround;
+	unsigned long t15_keystatus;
+	int t15_num_keys;
+	const unsigned int *t15_keymap;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -324,6 +327,8 @@ struct mxt_data {
 	u16 T71_address;
 	u8 T9_reportid_min;
 	u8 T9_reportid_max;
+	u8 T15_reportid_min;
+	u8 T15_reportid_max;
 	u16 T18_address;
 	u8 T19_reportid;
 	u8 T42_reportid_min;
@@ -987,6 +992,38 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 	data->update_input = true;
 }
 
+static void mxt_proc_t15_messages(struct mxt_data *data, u8 *msg)
+{
+	struct input_dev *input_dev = data->input_dev;
+	struct device *dev = &data->client->dev;
+	int key;
+	bool curr_state, new_state;
+	bool sync = false;
+	unsigned long keystates = le32_to_cpu((__force __le32)msg[2]);
+
+	for (key = 0; key < data->t15_num_keys; key++) {
+		curr_state = test_bit(key, &data->t15_keystatus);
+		new_state = test_bit(key, &keystates);
+
+		if (!curr_state && new_state) {
+			dev_dbg(dev, "T15 key press: %u\n", key);
+			__set_bit(key, &data->t15_keystatus);
+			input_event(input_dev, EV_KEY,
+				    data->t15_keymap[key], 1);
+			sync = true;
+		} else if (curr_state && !new_state) {
+			dev_dbg(dev, "T15 key release: %u\n", key);
+			__clear_bit(key, &data->t15_keystatus);
+			input_event(input_dev, EV_KEY,
+				    data->t15_keymap[key], 0);
+			sync = true;
+		}
+	}
+
+	if (sync)
+		input_sync(input_dev);
+}
+
 static void mxt_proc_t42_messages(struct mxt_data *data, u8 *msg)
 {
 	struct device *dev = &data->client->dev;
@@ -1045,6 +1082,9 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	} else if (report_id == data->T19_reportid) {
 		mxt_input_button(data, message);
 		data->update_input = true;
+	} else if (report_id >= data->T15_reportid_min
+		   && report_id <= data->T15_reportid_max) {
+		mxt_proc_t15_messages(data, message);
 	} else {
 		mxt_dump_message(data, message);
 	}
@@ -1706,6 +1746,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T71_address = 0;
 	data->T9_reportid_min = 0;
 	data->T9_reportid_max = 0;
+	data->T15_reportid_min = 0;
+	data->T15_reportid_max = 0;
 	data->T18_address = 0;
 	data->T19_reportid = 0;
 	data->T42_reportid_min = 0;
@@ -1784,6 +1826,10 @@ static int mxt_parse_object_table(struct mxt_data *data,
 						object->num_report_ids - 1;
 			data->num_touchids = object->num_report_ids;
 			break;
+		case MXT_TOUCH_KEYARRAY_T15:
+			data->T15_reportid_min = min_id;
+			data->T15_reportid_max = max_id;
+			break;
 		case MXT_SPT_COMMSCONFIG_T18:
 			data->T18_address = object->start_address;
 			break;
@@ -2077,6 +2123,7 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 	int error;
 	unsigned int num_mt_slots;
 	unsigned int mt_flags = 0;
+	int i;
 
 	switch (data->multitouch) {
 	case MXT_TOUCH_MULTI_T9:
@@ -2190,6 +2237,15 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 				     0, 255, 0, 0);
 	}
 
+	/* For T15 Key Array */
+	if (data->T15_reportid_min) {
+		data->t15_keystatus = 0;
+
+		for (i = 0; i < data->t15_num_keys; i++)
+			input_set_capability(input_dev, EV_KEY,
+					data->t15_keymap[i]);
+	}
+
 	input_set_drvdata(input_dev, data);
 
 	error = input_register_device(input_dev);
@@ -3148,8 +3204,10 @@ static void mxt_input_close(struct input_dev *dev)
 static int mxt_parse_device_properties(struct mxt_data *data)
 {
 	static const char keymap_property[] = "linux,gpio-keymap";
+	static const char buttons_property[] = "atmel,key-buttons";
 	struct device *dev = &data->client->dev;
 	u32 *keymap;
+	u32 *buttonmap;
 	int n_keys;
 	int error;
 
@@ -3179,6 +3237,33 @@ static int mxt_parse_device_properties(struct mxt_data *data)
 		data->t19_num_keys = n_keys;
 	}
 
+	if (device_property_present(dev, buttons_property)) {
+		n_keys = device_property_read_u32_array(dev, buttons_property,
+							NULL, 0);
+		if (n_keys <= 0) {
+			error = n_keys < 0 ? n_keys : -EINVAL;
+			dev_err(dev, "invalid/malformed '%s' property: %d\n",
+				buttons_property, error);
+			return error;
+		}
+
+		buttonmap = devm_kmalloc_array(dev, n_keys, sizeof(*buttonmap),
+					       GFP_KERNEL);
+		if (!buttonmap)
+			return -ENOMEM;
+
+		error = device_property_read_u32_array(dev, buttons_property,
+						       buttonmap, n_keys);
+		if (error) {
+			dev_err(dev, "failed to parse '%s' property: %d\n",
+				buttons_property, error);
+			return error;
+		}
+
+		data->t15_keymap = buttonmap;
+		data->t15_num_keys = n_keys;
+	}
+
 	return 0;
 }
 
Signature · 1 line
2.17.1
author
#9 lore raw

PATCH v10 9/55 Input: atmel_mxt_ts - handle reports from T47 Stylus object

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This patch handles reports from T47 Stylus object

(cherry picked from ndyer/linux/for-upstream commit 56405a5ea08eb34cfe83f3121867c9de0a5c48c1)

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 12 +++++++++++-
 1 file changed, 11 insertions(+), 1 deletion(-)
drivers/input/touchscreen/atmel_mxt_ts.c +11 -1
index d05249b02781..ba58cdd5b76d 100644
@@ -823,6 +823,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 	int area;
 	int amplitude;
 	u8 vector;
+	int tool;
 
 	id = message[0] - data->T9_reportid_min;
 	status = message[1];
@@ -836,6 +837,7 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 		y >>= 2;
 
 	area = message[5];
+
 	amplitude = message[6];
 	vector = message[7];
 
@@ -865,12 +867,20 @@ static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
 			mxt_input_sync(data);
 		}
 
+		/* A size of zero indicates touch is from a linked T47 Stylus */
+		if (area == 0) {
+			area = MXT_TOUCH_MAJOR_DEFAULT;
+			tool = MT_TOOL_PEN;
+		} else {
+			tool = MT_TOOL_FINGER;
+		}
+
 		/* if active, pressure must be non-zero */
 		if (!amplitude)
 			amplitude = MXT_PRESSURE_DEFAULT;
 
 		/* Touch active */
-		input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, 1);
+		input_mt_report_slot_state(input_dev, tool, 1);
 		input_report_abs(input_dev, ABS_MT_POSITION_X, x);
 		input_report_abs(input_dev, ABS_MT_POSITION_Y, y);
 		input_report_abs(input_dev, ABS_MT_PRESSURE, amplitude);
Signature · 1 line
2.17.1
author
#10 lore raw

PATCH v10 10/55 Input: atmel_mxt_ts - implement support for T107 active stylus

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This patch implements support for T107 active stylus

(cherry picked from ndyer/linux/for-upstream commit 20e357dd9acf8c2040068c8b22d6bc1401a1893f) [gdavis: Forward port and fix conflicts due to applying upstream commit

	 96a938aa214e ("Input: atmel_mxt_ts - remove platform data
	 support").]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 117 ++++++++++++++++++++++-
 1 file changed, 113 insertions(+), 4 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +113 -4
index ba58cdd5b76d..63db8b66eb67 100644
@@ -72,6 +72,7 @@
 #define MXT_SPT_CTECONFIG_T46		46
 #define MXT_SPT_DYNAMICCONFIGURATIONCONTAINER_T71 71
 #define MXT_TOUCH_MULTITOUCHSCREEN_T100 100
+#define MXT_PROCI_ACTIVESTYLUS_T107	107
 
 /* MXT_GEN_MESSAGE_T5 object */
 #define MXT_RPTID_NOMSG		0xff
@@ -181,6 +182,7 @@ struct t37_debug {
 enum t100_type {
 	MXT_T100_TYPE_FINGER		= 1,
 	MXT_T100_TYPE_PASSIVE_STYLUS	= 2,
+	MXT_T100_TYPE_ACTIVE_STYLUS	= 3,
 	MXT_T100_TYPE_HOVERING_FINGER	= 4,
 	MXT_T100_TYPE_GLOVE		= 5,
 	MXT_T100_TYPE_LARGE_TOUCH	= 6,
@@ -192,6 +194,16 @@ enum t100_type {
 #define MXT_TOUCH_MAJOR_DEFAULT		1
 #define MXT_PRESSURE_DEFAULT		1
 
+/* Gen2 Active Stylus */
+#define MXT_T107_STYLUS_STYAUX		42
+#define MXT_T107_STYLUS_STYAUX_PRESSURE	BIT(0)
+#define MXT_T107_STYLUS_STYAUX_PEAK	BIT(4)
+
+#define MXT_T107_STYLUS_HOVER		BIT(0)
+#define MXT_T107_STYLUS_TIPSWITCH	BIT(1)
+#define MXT_T107_STYLUS_BUTTON0		BIT(2)
+#define MXT_T107_STYLUS_BUTTON1		BIT(3)
+
 /* Delay times */
 #define MXT_BACKUP_TIME		50	/* msec */
 #define MXT_RESET_GPIO_TIME	20	/* msec */
@@ -313,10 +325,12 @@ struct mxt_data {
 	struct t7_config t7_cfg;
 	struct mxt_dbg dbg;
 	struct gpio_desc *reset_gpio;
-	bool use_retrigen_workaround;
 	unsigned long t15_keystatus;
 	int t15_num_keys;
 	const unsigned int *t15_keymap;
+	u8 stylus_aux_pressure;
+	u8 stylus_aux_peak;
+	bool use_retrigen_workaround;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -337,6 +351,7 @@ struct mxt_data {
 	u8 T48_reportid;
 	u8 T100_reportid_min;
 	u8 T100_reportid_max;
+	u16 T107_address;
 
 	/* for fw update in bootloader */
 	struct completion bl_completion;
@@ -908,6 +923,8 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 	u8 major = 0;
 	u8 pressure = 0;
 	u8 orientation = 0;
+	bool active = false;
+	bool hover = false;
 
 	id = message[0] - data->T100_reportid_min - 2;
 
@@ -926,6 +943,8 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 		case MXT_T100_TYPE_HOVERING_FINGER:
 			tool = MT_TOOL_FINGER;
 			distance = MXT_DISTANCE_HOVERING;
+			hover = true;
+			active = true;
 
 			if (data->t100_aux_vect)
 				orientation = message[data->t100_aux_vect];
@@ -936,6 +955,8 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 		case MXT_T100_TYPE_GLOVE:
 			tool = MT_TOOL_FINGER;
 			distance = MXT_DISTANCE_ACTIVE_TOUCH;
+			hover = false;
+			active = true;
 
 			if (data->t100_aux_area)
 				major = message[data->t100_aux_area];
@@ -950,6 +971,9 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 
 		case MXT_T100_TYPE_PASSIVE_STYLUS:
 			tool = MT_TOOL_PEN;
+			distance = MXT_DISTANCE_ACTIVE_TOUCH;
+			hover = false;
+			active = true;
 
 			/*
 			 * Passive stylus is reported with size zero so
@@ -962,6 +986,31 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 
 			break;
 
+		case MXT_T100_TYPE_ACTIVE_STYLUS:
+			/* Report input buttons */
+			input_report_key(input_dev, BTN_STYLUS,
+					 message[6] & MXT_T107_STYLUS_BUTTON0);
+			input_report_key(input_dev, BTN_STYLUS2,
+					 message[6] & MXT_T107_STYLUS_BUTTON1);
+
+			/* stylus in range, but position unavailable */
+			if (!(message[6] & MXT_T107_STYLUS_HOVER))
+				break;
+
+			tool = MT_TOOL_PEN;
+			distance = MXT_DISTANCE_ACTIVE_TOUCH;
+			active = true;
+			major = MXT_TOUCH_MAJOR_DEFAULT;
+
+			if (!(message[6] & MXT_T107_STYLUS_TIPSWITCH)) {
+				hover = true;
+				distance = MXT_DISTANCE_HOVERING;
+			} else if (data->stylus_aux_pressure) {
+				pressure = message[data->stylus_aux_pressure];
+			}
+
+			break;
+
 		case MXT_T100_TYPE_LARGE_TOUCH:
 			/* Ignore suppressed touch */
 			break;
@@ -976,12 +1025,12 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 	 * Values reported should be non-zero if tool is touching the
 	 * device
 	 */
-	if (!pressure && type != MXT_T100_TYPE_HOVERING_FINGER)
+	if (!pressure && !hover)
 		pressure = MXT_PRESSURE_DEFAULT;
 
 	input_mt_slot(input_dev, id);
 
-	if (status & MXT_T100_DETECT) {
+	if (active) {
 		dev_dbg(dev, "[%u] type:%u x:%u y:%u a:%02X p:%02X v:%02X\n",
 			id, type, x, y, major, pressure, orientation);
 
@@ -992,6 +1041,7 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 		input_report_abs(input_dev, ABS_MT_PRESSURE, pressure);
 		input_report_abs(input_dev, ABS_MT_DISTANCE, distance);
 		input_report_abs(input_dev, ABS_MT_ORIENTATION, orientation);
+
 	} else {
 		dev_dbg(dev, "[%u] release\n", id);
 
@@ -1863,6 +1913,9 @@ static int mxt_parse_object_table(struct mxt_data *data,
 			/* first two report IDs reserved */
 			data->num_touchids = object->num_report_ids - 2;
 			break;
+		case MXT_PROCI_ACTIVESTYLUS_T107:
+			data->T107_address = object->start_address;
+			break;
 		}
 
 		end_address = object->start_address
@@ -2023,6 +2076,54 @@ static int mxt_read_t9_resolution(struct mxt_data *data)
 	return 0;
 }
 
+static int mxt_set_up_active_stylus(struct input_dev *input_dev,
+				    struct mxt_data *data)
+{
+	struct i2c_client *client = data->client;
+	int error;
+	struct mxt_object *object;
+	u8 styaux;
+	int aux;
+	u8 ctrl;
+
+	object = mxt_get_object(data, MXT_PROCI_ACTIVESTYLUS_T107);
+	if (!object)
+		return 0;
+
+	error = __mxt_read_reg(client, object->start_address, 1, &ctrl);
+	if (error)
+		return error;
+
+	/* Check enable bit */
+	if (!(ctrl & 0x01))
+		return 0;
+
+	error = __mxt_read_reg(client,
+			       object->start_address + MXT_T107_STYLUS_STYAUX,
+			       1, &styaux);
+	if (error)
+		return error;
+
+	/* map aux bits */
+	aux = 7;
+
+	if (styaux & MXT_T107_STYLUS_STYAUX_PRESSURE)
+		data->stylus_aux_pressure = aux++;
+
+	if (styaux & MXT_T107_STYLUS_STYAUX_PEAK)
+		data->stylus_aux_peak = aux++;
+
+	input_set_capability(input_dev, EV_KEY, BTN_STYLUS);
+	input_set_capability(input_dev, EV_KEY, BTN_STYLUS2);
+	input_set_abs_params(input_dev, ABS_MT_TOOL_TYPE, 0, MT_TOOL_MAX, 0, 0);
+
+	dev_dbg(&client->dev,
+		"T107 active stylus, aux map pressure:%u peak:%u\n",
+		data->stylus_aux_pressure, data->stylus_aux_peak);
+
+	return 0;
+}
+
 static int mxt_read_t100_config(struct mxt_data *data)
 {
 	struct i2c_client *client = data->client;
@@ -2229,7 +2330,7 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 
 	if (data->multitouch == MXT_TOUCH_MULTI_T9 ||
 	    (data->multitouch == MXT_TOUCH_MULTITOUCHSCREEN_T100 &&
-	     data->t100_aux_ampl)) {
+	     (data->t100_aux_ampl || data->stylus_aux_pressure))) {
 		input_set_abs_params(input_dev, ABS_MT_PRESSURE,
 				     0, 255, 0, 0);
 	}
@@ -2247,6 +2348,14 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 				     0, 255, 0, 0);
 	}
 
+	/* For T107 Active Stylus */
+	if (data->multitouch == MXT_TOUCH_MULTITOUCHSCREEN_T100 &&
+	    data->T107_address) {
+		error = mxt_set_up_active_stylus(input_dev, data);
+		if (error)
+			dev_warn(dev, "Failed to read T107 config\n");
+	}
+
 	/* For T15 Key Array */
 	if (data->T15_reportid_min) {
 		data->t15_keystatus = 0;
Signature · 1 line
2.17.1
author
#11 lore raw

PATCH v10 11/55 Input: atmel_mxt_ts - add debug for T92 gesture and T93 touch seq msgs

In reply to Jiada Wang

From: Karl Tsou <[email protected]>

output T92 gesture and T93 touch sequence messages.

(cherry picked from ndyer/linux/for-upstream commit cb98986f8342107bf4a536aed4160b20839e97c1)

[jiada: changed dev_debug() to dev_info()]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 40 ++++++++++++++++++++++++
 1 file changed, 40 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +40 -0
index 63db8b66eb67..6126bb8a7acc 100644
@@ -71,6 +71,8 @@
 #define MXT_SPT_MESSAGECOUNT_T44	44
 #define MXT_SPT_CTECONFIG_T46		46
 #define MXT_SPT_DYNAMICCONFIGURATIONCONTAINER_T71 71
+#define MXT_PROCI_SYMBOLGESTUREPROCESSOR	92
+#define MXT_PROCI_TOUCHSEQUENCELOGGER	93
 #define MXT_TOUCH_MULTITOUCHSCREEN_T100 100
 #define MXT_PROCI_ACTIVESTYLUS_T107	107
 
@@ -349,6 +351,10 @@ struct mxt_data {
 	u8 T42_reportid_max;
 	u16 T44_address;
 	u8 T48_reportid;
+	u16 T92_address;
+	u8 T92_reportid;
+	u16 T93_address;
+	u8 T93_reportid;
 	u8 T100_reportid_min;
 	u8 T100_reportid_max;
 	u16 T107_address;
@@ -1113,6 +1119,24 @@ static int mxt_proc_t48_messages(struct mxt_data *data, u8 *msg)
 	return 0;
 }
 
+static void mxt_proc_t92_messages(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+	u8 status = msg[1];
+
+	dev_info(dev, "T92 long stroke LSTR=%d %d\n",
+		 (status & 0x80) ? 1 : 0,
+		 status & 0x0F);
+}
+
+static void mxt_proc_t93_messages(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+	u8 status = msg[1];
+
+	dev_info(dev, "T93 report double tap %d\n", status);
+}
+
 static int mxt_proc_message(struct mxt_data *data, u8 *message)
 {
 	u8 report_id = message[0];
@@ -1145,6 +1169,10 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	} else if (report_id >= data->T15_reportid_min
 		   && report_id <= data->T15_reportid_max) {
 		mxt_proc_t15_messages(data, message);
+	} else if (report_id == data->T92_reportid) {
+		mxt_proc_t92_messages(data, message);
+	} else if (report_id == data->T93_reportid) {
+		mxt_proc_t93_messages(data, message);
 	} else {
 		mxt_dump_message(data, message);
 	}
@@ -1814,6 +1842,10 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T42_reportid_max = 0;
 	data->T44_address = 0;
 	data->T48_reportid = 0;
+	data->T92_reportid = 0;
+	data->T92_address = 0;
+	data->T93_reportid = 0;
+	data->T93_address = 0;
 	data->T100_reportid_min = 0;
 	data->T100_reportid_max = 0;
 	data->max_reportid = 0;
@@ -1906,6 +1938,14 @@ static int mxt_parse_object_table(struct mxt_data *data,
 		case MXT_PROCG_NOISESUPPRESSION_T48:
 			data->T48_reportid = min_id;
 			break;
+		case MXT_PROCI_SYMBOLGESTUREPROCESSOR:
+			data->T92_reportid = min_id;
+			data->T92_address = object->start_address;
+			break;
+		case MXT_PROCI_TOUCHSEQUENCELOGGER:
+			data->T93_reportid = min_id;
+			data->T93_address = object->start_address;
+			break;
 		case MXT_TOUCH_MULTITOUCHSCREEN_T100:
 			data->multitouch = MXT_TOUCH_MULTITOUCHSCREEN_T100;
 			data->T100_reportid_min = min_id;
Signature · 1 line
2.17.1
author
#12 lore raw

PATCH v10 12/55 Input: atmel_mxt_ts - release touch state during suspend

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

If fingers are down as the MXT chip goes into suspend it does not send a lift message. In addition, it may not complete its final measurement cycle immediately, which means touch messages may be received by the interrupt handler after mxt_stop() has completed.

So:
- disable irq during suspend
- flush any messages created after suspend
- tell app layer that slots were released at suspend

(cherry picked from ndyer/linux/for-upstream commit 26794433086dbc7dea18d2f6a1c8d61ab25bcfda) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[gdavis: Squash fix from Dirk Behme:
- Input: atmel_mxt_ts - remove superfluous data->suspended]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 52 ++++++++++++++++++++++--
 1 file changed, 49 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +49 -3
index 6126bb8a7acc..9aafed92db9c 100644
@@ -372,6 +372,9 @@ struct mxt_data {
 	unsigned int t19_num_keys;
 
 	enum mxt_suspend_mode suspend_mode;
+
+	/* Indicates whether device is in suspend */
+	bool suspended;
 };
 
 struct mxt_vb2_buffer {
@@ -1151,10 +1154,10 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 		mxt_proc_t42_messages(data, message);
 	} else if (report_id == data->T48_reportid) {
 		mxt_proc_t48_messages(data, message);
-	} else if (!data->input_dev) {
+	} else if (!data->input_dev || data->suspended) {
 		/*
-		 * Do not report events if input device
-		 * is not yet registered.
+		 * Do not report events if input device is not
+		 * yet registered or returning from suspend
 		 */
 		mxt_dump_message(data, message);
 	} else if (report_id >= data->T9_reportid_min &&
@@ -3135,6 +3138,11 @@ static int mxt_load_fw(struct device *dev, const char *fn)
 	if (ret)
 		goto release_firmware;
 
+	if (data->suspended) {
+		enable_irq(data->irq);
+		data->suspended = false;
+	}
+
 	if (!data->in_bootloader) {
 		/* Change to the bootloader mode */
 		data->in_bootloader = true;
@@ -3306,8 +3314,27 @@ static void mxt_sysfs_remove(struct mxt_data *data)
 	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
 }
 
+static void mxt_reset_slots(struct mxt_data *data)
+{
+	struct input_dev *input_dev = data->input_dev;
+	int id;
+
+	if (!input_dev)
+		return;
+
+	for (id = 0; id < data->num_touchids; id++) {
+		input_mt_slot(input_dev, id);
+		input_mt_report_slot_state(input_dev, MT_TOOL_FINGER, 0);
+	}
+
+	mxt_input_sync(data);
+}
+
 static void mxt_start(struct mxt_data *data)
 {
+	if (!data->suspended || data->in_bootloader)
+		return;
+
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
 		mxt_soft_reset(data);
@@ -3320,16 +3347,29 @@ static void mxt_start(struct mxt_data *data)
 
 	case MXT_SUSPEND_DEEP_SLEEP:
 	default:
+		/*
+		 * Discard any touch messages still in message buffer
+		 * from before chip went to sleep
+		 */
+		mxt_process_messages_until_invalid(data);
+
 		mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
 
 		/* Recalibrate since chip has been in deep sleep */
 		mxt_t6_command(data, MXT_COMMAND_CALIBRATE, 1, false);
+
+		mxt_acquire_irq(data);
 		break;
 	}
+
+	data->suspended = false;
 }
 
 static void mxt_stop(struct mxt_data *data)
 {
+	if (data->suspended || data->in_bootloader)
+		return;
+
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
 		/* Touch disable */
@@ -3339,9 +3379,15 @@ static void mxt_stop(struct mxt_data *data)
 
 	case MXT_SUSPEND_DEEP_SLEEP:
 	default:
+		disable_irq(data->irq);
+
 		mxt_set_t7_power_cfg(data, MXT_POWER_CFG_DEEPSLEEP);
+
+		mxt_reset_slots(data);
 		break;
 	}
+
+	data->suspended = true;
 }
 
 static int mxt_input_open(struct input_dev *dev)
Signature · 1 line
2.17.1
author
#13 lore raw

PATCH v10 13/55 dt-bindings: input: atmel: add suspend mode support

In reply to Jiada Wang

Add suspend mode support for atmel touchscreen driver

---
 .../bindings/input/atmel,maxtouch.txt         |  9 ++++++++
 MAINTAINERS                                   |  1 +
 include/dt-bindings/input/atmel_mxt_ts.h      | 23 +++++++++++++++++++
 3 files changed, 33 insertions(+)
 create mode 100644 include/dt-bindings/input/atmel_mxt_ts.h
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +9 -0
index c88919480d37..0307e7f7bb43 100644
@@ -31,6 +31,15 @@ Optional properties for main touchpad device:
 
 - reset-gpios: GPIO specifier for the touchscreen's reset pin (active low)
 
+- atmel,suspend-mode: Select method used to suspend:
+    MXT_SUSPEND_DEEP_SLEEP - use T7 to suspend the device into deep sleep
+    MXT_SUSPEND_T9_CTRL - use T9.CTRL to turn off touch processing
+    MXT_SUSPEND_REGULATOR - use regulators to power down device during suspend
+    Definitions are in <dt-bindings/input/atmel_mxt_ts.h>.
+
+- vdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
+- avdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
+
 Example:
 
 	touch@4b {
MAINTAINERS +1 -0
index e48ab79879ac..350ae664e6f0 100644
@@ -2873,6 +2873,7 @@ T:	git git://github.com/ndyer/linux.git
 S:	Maintained
 F:	Documentation/devicetree/bindings/input/atmel,maxtouch.txt
 F:	drivers/input/touchscreen/atmel_mxt_ts.c
+F:	include/dt-bindings/input/atmel_mxt_ts.h
 
 ATMEL WIRELESS DRIVER
 M:	Simon Kelley <[email protected]>
include/dt-bindings/input/atmel_mxt_ts.h new file +23 -0
new file mode 100644
index 000000000000..41ed0f8111aa
@@ -0,0 +1,23 @@
+/* SPDX-License-Identifier: GPL-2.0 */
+/*
+ * Atmel maXTouch Touchscreen driver
+ *
+ * Copyright (C) 2015 Atmel Corporation
+ * Author: Nick Dyer <[email protected]>
+ *
+ * This program is free software; you can redistribute  it and/or modify it
+ * under  the terms of  the GNU General  Public License as published by the
+ * Free Software Foundation;  either version 2 of the  License, or (at your
+ * option) any later version.
+ */
+
+#ifndef __DT_BINDINGS_ATMEL_MXT_TS_H
+#define __DT_BINDINGS_ATMEL_MXT_TS_H
+
+enum mxt_suspend_mode {
+	MXT_SUSPEND_DEEP_SLEEP = 0,
+	MXT_SUSPEND_T9_CTRL = 1,
+	MXT_SUSPEND_REGULATOR = 2,
+};
+
+#endif /* __DT_BINDINGS_ATMEL_MXT_TS_H */
Signature · 1 line
2.17.1
#14 lore raw

Re: [PATCH v10 13/55] dt-bindings: input: atmel: add suspend mode support

31.03.2020 13:50, Jiada Wang пишет:

Quoted 16 lines

Add suspend mode support for atmel touchscreen driver

---
 .../bindings/input/atmel,maxtouch.txt         |  9 ++++++++
 MAINTAINERS                                   |  1 +
 include/dt-bindings/input/atmel_mxt_ts.h      | 23 +++++++++++++++++++
 3 files changed, 33 insertions(+)
 create mode 100644 include/dt-bindings/input/atmel_mxt_ts.h
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +0 -0
index c88919480d37..0307e7f7bb43 100644
@@ -31,6 +31,15 @@ Optional properties for main touchpad device:
 

Documentation/devicetree/bindings/input/atmel,maxtouch.txt Optional properties for main touchpad device: · line 41

Quoted 10 lines
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +8 -0
 - reset-gpios: GPIO specifier for the touchscreen's reset pin (active low)
 
+- atmel,suspend-mode: Select method used to suspend:
+    MXT_SUSPEND_DEEP_SLEEP - use T7 to suspend the device into deep sleep
+    MXT_SUSPEND_T9_CTRL - use T9.CTRL to turn off touch processing
+    MXT_SUSPEND_REGULATOR - use regulators to power down device during suspend
+    Definitions are in <dt-bindings/input/atmel_mxt_ts.h>.
+
+- vdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
+- avdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend

Some of NVIDIA Tegra devices have a VDD regulator in a form of GPIO, which is turned off by default at a boot time, and thus, the power-regulator needs to be enabled by the driver at a probe time in order to power-up the touchscreen.

I think the VDD/AVDD description isn't correct because these are the essential power regulators, they aren't used only for the suspending.

What about something more generic, like this:

- vdd: phandle to Power supply regulator
- avdd: phandle to Analog Power supply regulator

#15 lore raw

Re: [PATCH v10 13/55] dt-bindings: input: atmel: add suspend mode support

Hi Dmitry

Quoted 33 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 16 lines

Add suspend mode support for atmel touchscreen driver

---
  .../bindings/input/atmel,maxtouch.txt         |  9 ++++++++
  MAINTAINERS                                   |  1 +
  include/dt-bindings/input/atmel_mxt_ts.h      | 23 +++++++++++++++++++
  3 files changed, 33 insertions(+)

create mode 100644 include/dt-bindings/input/atmel_mxt_ts.h

Documentation/devicetree/bindings/input/atmel,maxtouch.txt +0 -0
index c88919480d37..0307e7f7bb43 100644
@@ -31,6 +31,15 @@ Optional properties for main touchpad device:
 

Documentation/devicetree/bindings/input/atmel,maxtouch.txt Optional properties for main touchpad device: · line 41

Quoted 10 lines
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +8 -0
  - reset-gpios: GPIO specifier for the touchscreen's reset pin (active low)
 
+- atmel,suspend-mode: Select method used to suspend:
+    MXT_SUSPEND_DEEP_SLEEP - use T7 to suspend the device into deep sleep
+    MXT_SUSPEND_T9_CTRL - use T9.CTRL to turn off touch processing
+    MXT_SUSPEND_REGULATOR - use regulators to power down device during suspend
+    Definitions are in <dt-bindings/input/atmel_mxt_ts.h>.
+
+- vdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
+- avdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend

Some of NVIDIA Tegra devices have a VDD regulator in a form of GPIO, which is turned off by default at a boot time, and thus, the power-regulator needs to be enabled by the driver at a probe time in order to power-up the touchscreen.

Quoted 8 lines

I think the VDD/AVDD description isn't correct because these are the essential power regulators, they aren't used only for the suspending.

What about something more generic, like this:

- vdd: phandle to Power supply regulator
- avdd: phandle to Analog Power supply regulator

I agree, will update with generic descriptions

Thanks,
Jiada

author
#16 lore raw

PATCH v10 14/55 Input: atmel_mxt_ts - add regulator control support

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Allow the driver to optionally manage enabling/disable power to the touch controller itself. If the regulators are not present then use the deep sleep power mode instead.

For a correct power on sequence, it is required that we have control over the RESET line.

(cherry picked from ndyer/linux/for-upstream commit 14052b61bb66c2f2283c00e733e131be7a9b8bfc) [gdavis: Resolve forward port conflicts due to v4.14-rc1 commmit
f657b00df22e ("Input: atmel_mxt_ts - add support for reset
line") and applying upstream commit 96a938aa214e ("Input:
atmel_mxt_ts - remove platform data support").]

[gdavis: Squash fixes from Dirk Behme:
- Input: atmel_mxt_ts - in failure case disable the regulator
- Input: atmel_mxt_ts - disable only enabled regulators
- Input: atmel_mxt_ts - use devm_regulator_get()]

[jiada: Replace white-spaces with tab for MXT_CHG_DELAY
separate Documentation/ and include/dt-bindings/ portion change to another commit]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 130 +++++++++++++++++++++--
 1 file changed, 121 insertions(+), 9 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +121 -9
index 9aafed92db9c..ef8baf64659e 100644
@@ -26,10 +26,12 @@
 #include <linux/slab.h>
 #include <linux/gpio/consumer.h>
 #include <asm/unaligned.h>
+#include <linux/regulator/consumer.h>
 #include <media/v4l2-device.h>
 #include <media/v4l2-ioctl.h>
 #include <media/videobuf2-v4l2.h>
 #include <media/videobuf2-vmalloc.h>
+#include <dt-bindings/input/atmel_mxt_ts.h>
 
 /* Firmware files */
 #define MXT_FW_NAME		"maxtouch.fw"
@@ -215,6 +217,9 @@ enum t100_type {
 #define MXT_CRC_TIMEOUT		1000	/* msec */
 #define MXT_FW_RESET_TIME	3000	/* msec */
 #define MXT_FW_CHG_TIMEOUT	300	/* msec */
+#define MXT_REGULATOR_DELAY	150	/* msec */
+#define MXT_CHG_DELAY		100	/* msec */
+#define MXT_POWERON_DELAY	150	/* msec */
 
 /* Command to unlock bootloader */
 #define MXT_UNLOCK_CMD_MSB	0xaa
@@ -275,11 +280,6 @@ enum v4l_dbg_inputs {
 	MXT_V4L_INPUT_MAX,
 };
 
-enum mxt_suspend_mode {
-	MXT_SUSPEND_DEEP_SLEEP	= 0,
-	MXT_SUSPEND_T9_CTRL	= 1,
-};
-
 /* Config update context */
 struct mxt_cfg {
 	u8 *raw;
@@ -333,6 +333,8 @@ struct mxt_data {
 	u8 stylus_aux_pressure;
 	u8 stylus_aux_peak;
 	bool use_retrigen_workaround;
+	struct regulator *reg_vdd;
+	struct regulator *reg_avdd;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -2073,6 +2075,94 @@ static int mxt_read_info_block(struct mxt_data *data)
 	return error;
 }
 
+static void mxt_regulator_enable(struct mxt_data *data)
+{
+	int error;
+
+	if (!data->reg_vdd || !data->reg_avdd)
+		return;
+
+	gpiod_set_value(data->reset_gpio, 0);
+
+	error = regulator_enable(data->reg_vdd);
+	if (error)
+		return;
+
+	error = regulator_enable(data->reg_avdd);
+	if (error) {
+		regulator_disable(data->reg_vdd);
+		return;
+	}
+
+	/*
+	 * According to maXTouch power sequencing specification, RESET line
+	 * must be kept low until some time after regulators come up to
+	 * voltage
+	 */
+	msleep(MXT_REGULATOR_DELAY);
+	gpiod_set_value(data->reset_gpio, 1);
+	msleep(MXT_CHG_DELAY);
+
+retry_wait:
+	reinit_completion(&data->bl_completion);
+	data->in_bootloader = true;
+	error = mxt_wait_for_completion(data, &data->bl_completion,
+					MXT_POWERON_DELAY);
+	if (error == -EINTR)
+		goto retry_wait;
+
+	data->in_bootloader = false;
+}
+
+static void mxt_regulator_disable(struct mxt_data *data)
+{
+	if (!data->reg_vdd || !data->reg_avdd)
+		return;
+
+	if (regulator_is_enabled(data->reg_vdd))
+		regulator_disable(data->reg_vdd);
+	if (regulator_is_enabled(data->reg_avdd))
+		regulator_disable(data->reg_avdd);
+}
+
+static int mxt_probe_regulators(struct mxt_data *data)
+{
+	struct device *dev = &data->client->dev;
+	int error;
+
+	/* Must have reset GPIO to use regulator support */
+	if (!data->reset_gpio) {
+		error = -EINVAL;
+		goto fail;
+	}
+
+	data->reg_vdd = devm_regulator_get(dev, "vdd");
+	if (IS_ERR(data->reg_vdd)) {
+		error = PTR_ERR(data->reg_vdd);
+		dev_err(dev, "Error %d getting vdd regulator\n", error);
+		goto fail;
+	}
+
+	data->reg_avdd = devm_regulator_get(dev, "avdd");
+	if (IS_ERR(data->reg_avdd)) {
+		error = PTR_ERR(data->reg_avdd);
+		dev_err(dev, "Error %d getting avdd regulator\n", error);
+		goto fail_release;
+	}
+
+	mxt_regulator_enable(data);
+
+	dev_dbg(dev, "Initialised regulators\n");
+	return 0;
+
+fail_release:
+	regulator_put(data->reg_vdd);
+fail:
+	data->reg_vdd = NULL;
+	data->reg_avdd = NULL;
+	return error;
+}
+
 static int mxt_read_t9_resolution(struct mxt_data *data)
 {
 	struct i2c_client *client = data->client;
@@ -3139,7 +3229,12 @@ static int mxt_load_fw(struct device *dev, const char *fn)
 		goto release_firmware;
 
 	if (data->suspended) {
-		enable_irq(data->irq);
+		if (data->suspend_mode == MXT_SUSPEND_REGULATOR)
+			mxt_regulator_enable(data);
+
+		if (data->suspend_mode == MXT_SUSPEND_DEEP_SLEEP)
+			enable_irq(data->irq);
+
 		data->suspended = false;
 	}
 
@@ -3345,6 +3440,11 @@ static void mxt_start(struct mxt_data *data)
 				MXT_TOUCH_MULTI_T9, MXT_T9_CTRL, 0x83);
 		break;
 
+	case MXT_SUSPEND_REGULATOR:
+		enable_irq(data->irq);
+		mxt_regulator_enable(data);
+		break;
+
 	case MXT_SUSPEND_DEEP_SLEEP:
 	default:
 		/*
@@ -3377,6 +3477,12 @@ static void mxt_stop(struct mxt_data *data)
 				MXT_TOUCH_MULTI_T9, MXT_T9_CTRL, 0);
 		break;
 
+	case MXT_SUSPEND_REGULATOR:
+		disable_irq(data->irq);
+		mxt_regulator_disable(data);
+		mxt_reset_slots(data);
+		break;
+
 	case MXT_SUSPEND_DEEP_SLEEP:
 	default:
 		disable_irq(data->irq);
@@ -3469,6 +3575,8 @@ static int mxt_parse_device_properties(struct mxt_data *data)
 		data->t15_num_keys = n_keys;
 	}
 
+	device_property_read_u32(dev, "atmel,suspend-mode",
+				 &data->suspend_mode);
 	return 0;
 }
 
@@ -3555,14 +3663,18 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		return error;
 	}
 
-	disable_irq(client->irq);
-
-	if (data->reset_gpio) {
+	if (data->suspend_mode == MXT_SUSPEND_REGULATOR) {
+		error = mxt_probe_regulators(data);
+		if (error)
+			return error;
+	} else if (data->reset_gpio) {
 		msleep(MXT_RESET_GPIO_TIME);
 		gpiod_set_value(data->reset_gpio, 1);
 		msleep(MXT_RESET_INVALID_CHG);
 	}
 
+	disable_irq(data->irq);
+
 	error = mxt_initialize(data);
 	if (error)
 		return error;
Signature · 1 line
2.17.1
author
#17 lore raw

PATCH v10 15/55 Input: atmel_mxt_ts - report failures in suspend/resume

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This patch reports failures in suspend/resume

(cherry picked from ndyer/linux/for-upstream commit 93a57575403d) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform data support").]

[jiada: Fix compilation warning
Add commit description]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 51 ++++++++++++++++++------
 1 file changed, 39 insertions(+), 12 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +39 -12
index ef8baf64659e..f8783e37436f 100644
@@ -3425,10 +3425,12 @@ static void mxt_reset_slots(struct mxt_data *data)
 	mxt_input_sync(data);
 }
 
-static void mxt_start(struct mxt_data *data)
+static int mxt_start(struct mxt_data *data)
 {
+	int ret = 0;
+
 	if (!data->suspended || data->in_bootloader)
-		return;
+		return 0;
 
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
@@ -3453,28 +3455,42 @@ static void mxt_start(struct mxt_data *data)
 		 */
 		mxt_process_messages_until_invalid(data);
 
-		mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
+		ret = mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
+		if (ret)
+			return ret;
 
 		/* Recalibrate since chip has been in deep sleep */
-		mxt_t6_command(data, MXT_COMMAND_CALIBRATE, 1, false);
+		ret = mxt_t6_command(data, MXT_COMMAND_CALIBRATE, 1, false);
+		if (ret)
+			return ret;
+
+		ret = mxt_acquire_irq(data);
+		if (ret)
+			return ret;
 
-		mxt_acquire_irq(data);
 		break;
 	}
 
 	data->suspended = false;
+
+	return 0;
 }
 
-static void mxt_stop(struct mxt_data *data)
+static int mxt_stop(struct mxt_data *data)
 {
+	int ret;
+
 	if (data->suspended || data->in_bootloader)
-		return;
+		return 0;
 
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
 		/* Touch disable */
-		mxt_write_object(data,
+		ret = mxt_write_object(data,
 				MXT_TOUCH_MULTI_T9, MXT_T9_CTRL, 0);
+		if (ret)
+			return ret;
+
 		break;
 
 	case MXT_SUSPEND_REGULATOR:
@@ -3487,29 +3503,40 @@ static void mxt_stop(struct mxt_data *data)
 	default:
 		disable_irq(data->irq);
 
-		mxt_set_t7_power_cfg(data, MXT_POWER_CFG_DEEPSLEEP);
+		ret = mxt_set_t7_power_cfg(data, MXT_POWER_CFG_DEEPSLEEP);
+		if (ret)
+			return ret;
 
 		mxt_reset_slots(data);
 		break;
 	}
 
 	data->suspended = true;
+	return 0;
 }
 
 static int mxt_input_open(struct input_dev *dev)
 {
 	struct mxt_data *data = input_get_drvdata(dev);
+	int ret;
 
-	mxt_start(data);
+	ret = mxt_start(data);
 
-	return 0;
+	if (ret)
+		dev_err(&data->client->dev, "%s failed rc=%d\n", __func__, ret);
+
+	return ret;
 }
 
 static void mxt_input_close(struct input_dev *dev)
 {
 	struct mxt_data *data = input_get_drvdata(dev);
+	int ret;
 
-	mxt_stop(data);
+	ret = mxt_stop(data);
+
+	if (ret)
+		dev_err(&data->client->dev, "%s failed rc=%d\n", __func__, ret);
 }
 
 static int mxt_parse_device_properties(struct mxt_data *data)
Signature · 1 line
2.17.1
author
#18 lore raw

PATCH v10 16/55 Input: atmel_mxt_ts - allow specification of firmware file name

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

On platforms which have multiple device instances using this driver, the firmware may be different on each device. This patch makes the user give the name of the firmware file when flashing.

This also prevents accidental triggering of the firmware load process.

(cherry picked from ndyer/linux/for-upstream commit 76ebb7cee971cb42dfb0a3a9224403b8b09abcf1)

[gdavis: Forward port and fix conflicts.]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 43 ++++++++++++++++++++----
 1 file changed, 37 insertions(+), 6 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +37 -6
index f8783e37436f..0e30ff372a43 100644
@@ -33,8 +33,7 @@
 #include <media/videobuf2-vmalloc.h>
 #include <dt-bindings/input/atmel_mxt_ts.h>
 
-/* Firmware files */
-#define MXT_FW_NAME		"maxtouch.fw"
+/* Configuration file */
 #define MXT_CFG_NAME		"maxtouch.cfg"
 #define MXT_CFG_MAGIC		"OBP_RAW V1"
 
@@ -335,6 +334,7 @@ struct mxt_data {
 	bool use_retrigen_workaround;
 	struct regulator *reg_vdd;
 	struct regulator *reg_avdd;
+	char *fw_name;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -3207,7 +3207,7 @@ static int mxt_check_firmware_format(struct device *dev,
 	return -EINVAL;
 }
 
-static int mxt_load_fw(struct device *dev, const char *fn)
+static int mxt_load_fw(struct device *dev)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
 	const struct firmware *fw = NULL;
@@ -3217,9 +3217,9 @@ static int mxt_load_fw(struct device *dev, const char *fn)
 	unsigned int frame = 0;
 	int ret;
 
-	ret = request_firmware(&fw, fn, dev);
+	ret = request_firmware(&fw, data->fw_name, dev);
 	if (ret) {
-		dev_err(dev, "Unable to open firmware %s\n", fn);
+		dev_err(dev, "Unable to open firmware %s\n", data->fw_name);
 		return ret;
 	}
 
@@ -3339,6 +3339,33 @@ static int mxt_load_fw(struct device *dev, const char *fn)
 	return ret;
 }
 
+static int mxt_update_file_name(struct device *dev, char **file_name,
+				const char *buf, size_t count)
+{
+	char *file_name_tmp;
+
+	/* Simple sanity check */
+	if (count > 64) {
+		dev_warn(dev, "File name too long\n");
+		return -EINVAL;
+	}
+
+	file_name_tmp = krealloc(*file_name, count + 1, GFP_KERNEL);
+	if (!file_name_tmp)
+		return -ENOMEM;
+
+	*file_name = file_name_tmp;
+	memcpy(*file_name, buf, count);
+
+	/* Echo into the sysfs entry may append newline at the end of buf */
+	if (buf[count - 1] == '\n')
+		(*file_name)[count - 1] = '\0';
+	else
+		(*file_name)[count] = '\0';
+
+	return 0;
+}
+
 static ssize_t mxt_update_fw_store(struct device *dev,
 					struct device_attribute *attr,
 					const char *buf, size_t count)
@@ -3346,7 +3373,11 @@ static ssize_t mxt_update_fw_store(struct device *dev,
 	struct mxt_data *data = dev_get_drvdata(dev);
 	int error;
 
-	error = mxt_load_fw(dev, MXT_FW_NAME);
+	error = mxt_update_file_name(dev, &data->fw_name, buf, count);
+	if (error)
+		return error;
+
+	error = mxt_load_fw(dev);
 	if (error) {
 		dev_err(dev, "The firmware update failed(%d)\n", error);
 		count = error;
Signature · 1 line
2.17.1
author
#19 lore raw

PATCH v10 17/55 Input: atmel_mxt_ts: Rename mxt_fw_version_show to fw_version_show

In reply to Jiada Wang

Rename mxt_fw_version_show to fw_version_show to address checkpatch warning

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +3 -3
index 0e30ff372a43..b2a37a9597f3 100644
@@ -3104,8 +3104,8 @@ static int mxt_configure_objects(struct mxt_data *data,
 }
 
 /* Firmware Version is returned as Major.Minor.Build */
-static ssize_t mxt_fw_version_show(struct device *dev,
-				   struct device_attribute *attr, char *buf)
+static ssize_t fw_version_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
 	struct mxt_info *info = data->info;
@@ -3403,7 +3403,7 @@ static const struct attribute_group mxt_fw_attr_group = {
 	.attrs = mxt_fw_attrs,
 };
 
-static DEVICE_ATTR(fw_version, S_IRUGO, mxt_fw_version_show, NULL);
+static DEVICE_ATTR_RO(fw_version);
 static DEVICE_ATTR(hw_version, S_IRUGO, mxt_hw_version_show, NULL);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
 
Signature · 1 line
2.17.1
author
#20 lore raw

PATCH v10 18/55 Input: atmel_mxt_ts: Rename mxt_hw_version_show to hw_version_show

In reply to Jiada Wang

Rename mxt_hw_version_show to hw_version_show to address checkpatch warning

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +3 -3
index b2a37a9597f3..cec823de4096 100644
@@ -3114,8 +3114,8 @@ static ssize_t fw_version_show(struct device *dev,
 }
 
 /* Hardware Version is returned as FamilyID.VariantID */
-static ssize_t mxt_hw_version_show(struct device *dev,
-				   struct device_attribute *attr, char *buf)
+static ssize_t hw_version_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
 	struct mxt_info *info = data->info;
@@ -3404,7 +3404,7 @@ static const struct attribute_group mxt_fw_attr_group = {
 };
 
 static DEVICE_ATTR_RO(fw_version);
-static DEVICE_ATTR(hw_version, S_IRUGO, mxt_hw_version_show, NULL);
+static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
 
 static struct attribute *mxt_attrs[] = {
Signature · 1 line
2.17.1
#21 lore raw

Re: [PATCH v10 18/55] Input: atmel_mxt_ts: Rename mxt_hw_version_show to hw_version_show

31.03.2020 13:50, Jiada Wang пишет:

Quoted 23 lines

Rename mxt_hw_version_show to hw_version_show to address checkpatch warning

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +2 -2
index b2a37a9597f3..cec823de4096 100644
@@ -3114,8 +3114,8 @@ static ssize_t fw_version_show(struct device *dev,
 }
 
 /* Hardware Version is returned as FamilyID.VariantID */
-static ssize_t mxt_hw_version_show(struct device *dev,
-				   struct device_attribute *attr, char *buf)
+static ssize_t hw_version_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
 	struct mxt_info *info = data->info;

drivers/input/touchscreen/atmel_mxt_ts.c static const struct attribute_group mxt_fw_attr_group = { · line 3408

Quoted 6 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
@@ -3404,7 +3404,7 @@ static const struct attribute_group mxt_fw_attr_group = {
 };
 
 static DEVICE_ATTR_RO(fw_version);
-static DEVICE_ATTR(hw_version, S_IRUGO, mxt_hw_version_show, NULL);
+static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);

What about to add a patch to convert the mxt_object_show to DEVICE_ATTR_RO(object_show)?

#22 lore raw

Re: [PATCH v10 18/55] Input: atmel_mxt_ts: Rename mxt_hw_version_show to hw_version_show

Hi Dmitry

Quoted 26 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 25 lines

Rename mxt_hw_version_show to hw_version_show to address checkpatch warning

---
  drivers/input/touchscreen/atmel_mxt_ts.c | 6 +++---
  1 file changed, 3 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +2 -2
index b2a37a9597f3..cec823de4096 100644
@@ -3114,8 +3114,8 @@ static ssize_t fw_version_show(struct device *dev,
  }
 
  /* Hardware Version is returned as FamilyID.VariantID */
-static ssize_t mxt_hw_version_show(struct device *dev,
-				   struct device_attribute *attr, char *buf)
+static ssize_t hw_version_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
  {
  	struct mxt_data *data = dev_get_drvdata(dev);
  	struct mxt_info *info = data->info;
@@ -3404,7 +3404,7 @@ static const struct attribute_group mxt_fw_attr_group = {
  };
 
Quoted 8 lines

drivers/input/touchscreen/atmel_mxt_ts.c static const struct attribute_group mxt_fw_attr_group = { · line 3408

Quoted 4 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
  static DEVICE_ATTR_RO(fw_version);
-static DEVICE_ATTR(hw_version, S_IRUGO, mxt_hw_version_show, NULL);
+static DEVICE_ATTR_RO(hw_version);
  static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);

What about to add a patch to convert the mxt_object_show to DEVICE_ATTR_RO(object_show)?

sure, mxt_object, is the last attr not using DEVICE_ATTR_[RO|WO|RW], I will add another commit to update in next version

Thanks,
Jiada

author
#23 lore raw

PATCH v10 19/55 Input: atmel_mxt_ts: rename mxt_update_fw_store to update_fw_store

In reply to Jiada Wang

Rename mxt_update_fw_store to update_fw_store, to address checkpatch warning.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +4 -4
index cec823de4096..720574417219 100644
@@ -3366,9 +3366,9 @@ static int mxt_update_file_name(struct device *dev, char **file_name,
 	return 0;
 }
 
-static ssize_t mxt_update_fw_store(struct device *dev,
-					struct device_attribute *attr,
-					const char *buf, size_t count)
+static ssize_t update_fw_store(struct device *dev,
+			       struct device_attribute *attr,
+			       const char *buf, size_t count)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
 	int error;
@@ -3392,7 +3392,7 @@ static ssize_t mxt_update_fw_store(struct device *dev,
 	return count;
 }
 
-static DEVICE_ATTR(update_fw, 0200, NULL, mxt_update_fw_store);
+static DEVICE_ATTR_WO(update_fw);
 
 static struct attribute *mxt_fw_attrs[] = {
 	&dev_attr_update_fw.attr,
Signature · 1 line
2.17.1
author
#24 lore raw

PATCH v10 20/55 Input: atmel_mxt_ts - handle cfg filename via pdata/sysfs

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

There may be multiple maXTouch chips on a single device which will require different configuration files. Add a platform data value for the configuration filename.

Add sysfs entry to write configuration file if the platform data is not set.

Split out the object initialisation code from mxt_initialize() into mxt_configure_objects() to allow this.

(cherry picked from ndyer/linux/for-upstream commit 71a2a4d1954460b949a16b607f72bafab294ca79) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[gdavis: Squash fix from Vladimir Zapolskiy:
- Input: atmel_mxt_ts - fix error paths in mxt_configure_objects()]

[jiada: Separate Documentation/ portion change to another commit
Rename mxt_update_cfg_store to update_cfg_store Replace DEVICE_ATTR with DEVICE_ATTR_WO]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 109 +++++++++++++++++++----
 1 file changed, 94 insertions(+), 15 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +94 -15
index 720574417219..d1865250f492 100644
@@ -34,7 +34,6 @@
 #include <dt-bindings/input/atmel_mxt_ts.h>
 
 /* Configuration file */
-#define MXT_CFG_NAME		"maxtouch.cfg"
 #define MXT_CFG_MAGIC		"OBP_RAW V1"
 
 /* Registers */
@@ -335,6 +334,8 @@ struct mxt_data {
 	struct regulator *reg_vdd;
 	struct regulator *reg_avdd;
 	char *fw_name;
+	char *cfg_name;
+	const char *pcfg_name;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -377,6 +378,9 @@ struct mxt_data {
 
 	/* Indicates whether device is in suspend */
 	bool suspended;
+
+	/* Indicates whether device is updating configuration */
+	bool updating_config;
 };
 
 struct mxt_vb2_buffer {
@@ -833,9 +837,11 @@ static void mxt_input_button(struct mxt_data *data, u8 *message)
 
 static void mxt_input_sync(struct mxt_data *data)
 {
-	input_mt_report_pointer_emulation(data->input_dev,
-					  data->t19_num_keys);
-	input_sync(data->input_dev);
+	if (data->input_dev) {
+		input_mt_report_pointer_emulation(data->input_dev,
+						  data->t19_num_keys);
+		input_sync(data->input_dev);
+	}
 }
 
 static void mxt_proc_t9_message(struct mxt_data *data, u8 *message)
@@ -2578,13 +2584,21 @@ static int mxt_initialize(struct mxt_data *data)
 	if (error)
 		return error;
 
-	error = request_firmware_nowait(THIS_MODULE, true, MXT_CFG_NAME,
-					&client->dev, GFP_KERNEL, data,
-					mxt_config_cb);
-	if (error) {
-		dev_err(&client->dev, "Failed to invoke firmware loader: %d\n",
-			error);
-		goto err_free_sysfs;
+	if (data->cfg_name) {
+		error = request_firmware_nowait(THIS_MODULE, true,
+						data->cfg_name,
+						&client->dev,
+						GFP_KERNEL, data,
+						mxt_config_cb);
+		if (error) {
+			dev_err(&client->dev, "Failed to invoke firmware loader: %d\n",
+				error);
+			goto err_free_sysfs;
+		}
+	} else {
+		error = mxt_configure_objects(data, NULL);
+		if (error)
+			goto err_free_sysfs;
 	}
 
 	return 0;
@@ -3081,19 +3095,21 @@ static int mxt_configure_objects(struct mxt_data *data,
 	error = mxt_init_t7_power_cfg(data);
 	if (error) {
 		dev_err(dev, "Failed to initialize power cfg\n");
-		return error;
+		goto err_free_object_table;
 	}
 
 	if (cfg) {
 		error = mxt_update_cfg(data, cfg);
-		if (error)
+		if (error) {
 			dev_warn(dev, "Error %d updating config\n", error);
+			goto err_free_object_table;
+		}
 	}
 
 	if (data->multitouch) {
 		error = mxt_initialize_input_device(data);
 		if (error)
-			return error;
+			goto err_free_object_table;
 	} else {
 		dev_warn(dev, "No touch object detected\n");
 	}
@@ -3101,6 +3117,10 @@ static int mxt_configure_objects(struct mxt_data *data,
 	mxt_debug_init(data);
 
 	return 0;
+
+err_free_object_table:
+	mxt_free_object_table(data);
+	return error;
 }
 
 /* Firmware Version is returned as Major.Minor.Build */
@@ -3392,6 +3412,55 @@ static ssize_t update_fw_store(struct device *dev,
 	return count;
 }
 
+static ssize_t update_cfg_store(struct device *dev,
+				struct device_attribute *attr,
+				const char *buf, size_t count)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	const struct firmware *cfg;
+	int ret;
+
+	ret = mxt_update_file_name(dev, &data->cfg_name, buf, count);
+	if (ret)
+		return ret;
+
+	ret = request_firmware(&cfg, data->cfg_name, dev);
+	if (ret < 0) {
+		dev_err(dev, "Failure to request config file %s\n",
+			data->cfg_name);
+		ret = -ENOENT;
+		goto out;
+	}
+
+	data->updating_config = true;
+
+	mxt_free_input_device(data);
+
+	if (data->suspended) {
+		if (data->suspend_mode == MXT_SUSPEND_REGULATOR) {
+			enable_irq(data->irq);
+			mxt_regulator_enable(data);
+		} else if (data->suspend_mode == MXT_SUSPEND_DEEP_SLEEP) {
+			mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
+			mxt_acquire_irq(data);
+		}
+
+		data->suspended = false;
+	}
+
+	ret = mxt_configure_objects(data, cfg);
+	if (ret)
+		goto release;
+
+	ret = count;
+
+release:
+	release_firmware(cfg);
+out:
+	data->updating_config = false;
+	return ret;
+}
+
 static DEVICE_ATTR_WO(update_fw);
 
 static struct attribute *mxt_fw_attrs[] = {
@@ -3406,11 +3475,13 @@ static const struct attribute_group mxt_fw_attr_group = {
 static DEVICE_ATTR_RO(fw_version);
 static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
+static DEVICE_ATTR_WO(update_cfg);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
 	&dev_attr_hw_version.attr,
 	&dev_attr_object.attr,
+	&dev_attr_update_cfg.attr,
 	NULL
 };
 
@@ -3511,7 +3582,7 @@ static int mxt_stop(struct mxt_data *data)
 {
 	int ret;
 
-	if (data->suspended || data->in_bootloader)
+	if (data->suspended || data->in_bootloader || data->updating_config)
 		return 0;
 
 	switch (data->suspend_mode) {
@@ -3580,6 +3651,8 @@ static int mxt_parse_device_properties(struct mxt_data *data)
 	int n_keys;
 	int error;
 
+	device_property_read_string(dev, "atmel,cfg_name", &data->pcfg_name);
+
 	if (device_property_present(dev, keymap_property)) {
 		n_keys = device_property_count_u32(dev, keymap_property);
 		if (n_keys <= 0) {
@@ -3705,6 +3778,12 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	if (error)
 		return error;
 
+	if (data->pcfg_name)
+		mxt_update_file_name(&data->client->dev,
+				     &data->cfg_name,
+				     data->pcfg_name,
+				     strlen(data->pcfg_name));
+
 	data->reset_gpio = devm_gpiod_get_optional(&client->dev,
 						   "reset", GPIOD_OUT_LOW);
 	if (IS_ERR(data->reset_gpio)) {
Signature · 1 line
2.17.1
#25 lore raw

Re: [PATCH v10 20/55] Input: atmel_mxt_ts - handle cfg filename via pdata/sysfs

31.03.2020 13:50, Jiada Wang пишет:

Quoted 40 lines

From: Nick Dyer <[email protected]>

There may be multiple maXTouch chips on a single device which will require different configuration files. Add a platform data value for the configuration filename.

Add sysfs entry to write configuration file if the platform data is not set.

Split out the object initialisation code from mxt_initialize() into mxt_configure_objects() to allow this.

(cherry picked from ndyer/linux/for-upstream commit 71a2a4d1954460b949a16b607f72bafab294ca79) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[gdavis: Squash fix from Vladimir Zapolskiy:
- Input: atmel_mxt_ts - fix error paths in mxt_configure_objects()]

[jiada: Separate Documentation/ portion change to another commit
Rename mxt_update_cfg_store to update_cfg_store Replace DEVICE_ATTR with DEVICE_ATTR_WO]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 109 +++++++++++++++++++----
 1 file changed, 94 insertions(+), 15 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +0 -1
index 720574417219..d1865250f492 100644
@@ -34,7 +34,6 @@
 #include <dt-bindings/input/atmel_mxt_ts.h>
 
 /* Configuration file */
-#define MXT_CFG_NAME		"maxtouch.cfg"
 #define MXT_CFG_MAGIC		"OBP_RAW V1"

...

drivers/input/touchscreen/atmel_mxt_ts.c static void mxt_input_button(struct mxt_data *data, u8 *message) · line 845

Quoted 11 lines

static void mxt_input_sync(struct mxt_data *data) {

drivers/input/touchscreen/atmel_mxt_ts.c +5 -3
-	input_mt_report_pointer_emulation(data->input_dev,
-					  data->t19_num_keys);
-	input_sync(data->input_dev);
+	if (data->input_dev) {
+		input_mt_report_pointer_emulation(data->input_dev,
+						  data->t19_num_keys);
+		input_sync(data->input_dev);
+	}
 }

Looks like this hunk shouldn't belong to this patch.

...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3785

Quoted 5 lines
drivers/input/touchscreen/atmel_mxt_ts.c +5 -0
+	if (data->pcfg_name)
+		mxt_update_file_name(&data->client->dev,
+				     &data->cfg_name,
+				     data->pcfg_name,
+				     strlen(data->pcfg_name));

This looks wrong to me because I think the cfg_name should fall back to "maxtouch.cfg" if atmel,cfg_name isn't specified in device-tree, otherwise this commit may break devices that use an older device-tree.

#26 lore raw

Re: [PATCH v10 20/55] Input: atmel_mxt_ts - handle cfg filename via pdata/sysfs

Hi Dmitrij

Quoted 43 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 40 lines

From: Nick Dyer <[email protected]>

There may be multiple maXTouch chips on a single device which will require different configuration files. Add a platform data value for the configuration filename.

Add sysfs entry to write configuration file if the platform data is not set.

Split out the object initialisation code from mxt_initialize() into mxt_configure_objects() to allow this.

(cherry picked from ndyer/linux/for-upstream commit 71a2a4d1954460b949a16b607f72bafab294ca79) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[gdavis: Squash fix from Vladimir Zapolskiy:
- Input: atmel_mxt_ts - fix error paths in mxt_configure_objects()]

[jiada: Separate Documentation/ portion change to another commit
Rename mxt_update_cfg_store to update_cfg_store Replace DEVICE_ATTR with DEVICE_ATTR_WO]

---
  drivers/input/touchscreen/atmel_mxt_ts.c | 109 +++++++++++++++++++----
  1 file changed, 94 insertions(+), 15 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +0 -1
index 720574417219..d1865250f492 100644
@@ -34,7 +34,6 @@
  #include <dt-bindings/input/atmel_mxt_ts.h>
 
  /* Configuration file */
-#define MXT_CFG_NAME		"maxtouch.cfg"
  #define MXT_CFG_MAGIC		"OBP_RAW V1"

...

Quoted 13 lines

drivers/input/touchscreen/atmel_mxt_ts.c static void mxt_input_button(struct mxt_data *data, u8 *message) · line 845

Quoted 11 lines
  static void mxt_input_sync(struct mxt_data *data)
  {
drivers/input/touchscreen/atmel_mxt_ts.c +5 -3
-	input_mt_report_pointer_emulation(data->input_dev,
-					  data->t19_num_keys);
-	input_sync(data->input_dev);
+	if (data->input_dev) {
+		input_mt_report_pointer_emulation(data->input_dev,
+						  data->t19_num_keys);
+		input_sync(data->input_dev);
+	}
  }

Looks like this hunk shouldn't belong to this patch.

yes, this part of change
squash fix from ndyer/linux/for-upstream commit c909ada856861f305653b127db3ea0fa60264331
- Input: atmel_mxt_ts - check data->input_dev is not null in mxt_input_sync() seems due to some mistake, commit description about the squash,
start to be missing from v6 patch-set

I will separate it back into a single commit

Quoted 12 lines

...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3785

Quoted 5 lines
drivers/input/touchscreen/atmel_mxt_ts.c +5 -0
+	if (data->pcfg_name)
+		mxt_update_file_name(&data->client->dev,
+				     &data->cfg_name,
+				     data->pcfg_name,
+				     strlen(data->pcfg_name));

This looks wrong to me because I think the cfg_name should fall back to "maxtouch.cfg" if atmel,cfg_name isn't specified in device-tree, otherwise this commit may break devices that use an older device-tree.

I agree with you, in next version,
I will update this patch to also be compatible with legacy cfg name "maxtouch.cfg"

Thanks,
Jiada

author
#27 lore raw

PATCH v10 21/55 dt-bindings: input: atmel: provide name of configuration file

In reply to Jiada Wang

Add support to set name of configuration file

---
 Documentation/devicetree/bindings/input/atmel,maxtouch.txt | 3 +++
 1 file changed, 3 insertions(+)
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +3 -0
index 0307e7f7bb43..29bb84a920d6 100644
@@ -40,6 +40,9 @@ Optional properties for main touchpad device:
 - vdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
 - avdd: Required supply regulator when MXT_SUSPEND_REGULATOR is used to suspend
 
+- atmel,cfg_name: Provide name of configuration file in OBP_RAW format. This
+    will be downloaded from the firmware loader on probe to the device.
+
 Example:
 
 	touch@4b {
Signature · 1 line
2.17.1
author
#28 lore raw

PATCH v10 22/55 Input: atmel_mxt_ts - allow input name to be specified in platform data

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Android systems identify the input device and map to IDC file by using the input device name. To avoid unnecessary deltas to the driver file, allow this to be set from the platform data.

(cherry picked from ndyer/linux/for-upstream commit cbf94a7bda754d2e1899d9f50313a0bccc91422d) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[jiada: Separate Documentation/ portion change to another commit]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 9 ++++++++-
 1 file changed, 8 insertions(+), 1 deletion(-)
drivers/input/touchscreen/atmel_mxt_ts.c +8 -1
index d1865250f492..dfc20c0ca0a5 100644
@@ -336,6 +336,7 @@ struct mxt_data {
 	char *fw_name;
 	char *cfg_name;
 	const char *pcfg_name;
+	const char *input_name;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -2412,7 +2413,11 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 	if (!input_dev)
 		return -ENOMEM;
 
-	input_dev->name = "Atmel maXTouch Touchscreen";
+	if (data->input_name)
+		input_dev->name = data->input_name;
+	else
+		input_dev->name = "Atmel maXTouch Touchscreen";
+
 	input_dev->phys = data->phys;
 	input_dev->id.bustype = BUS_I2C;
 	input_dev->dev.parent = dev;
@@ -3653,6 +3658,8 @@ static int mxt_parse_device_properties(struct mxt_data *data)
 
 	device_property_read_string(dev, "atmel,cfg_name", &data->pcfg_name);
 
+	device_property_read_string(dev, "atmel,input_name", &data->input_name);
+
 	if (device_property_present(dev, keymap_property)) {
 		n_keys = device_property_count_u32(dev, keymap_property);
 		if (n_keys <= 0) {
Signature · 1 line
2.17.1
author
#29 lore raw

PATCH v10 23/55 dt-bindings: input: atmel: support to specify input name

In reply to Jiada Wang

Support to specify input name

---
 Documentation/devicetree/bindings/input/atmel,maxtouch.txt | 2 ++
 1 file changed, 2 insertions(+)
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +2 -0
index 29bb84a920d6..126266891cdc 100644
@@ -43,6 +43,8 @@ Optional properties for main touchpad device:
 - atmel,cfg_name: Provide name of configuration file in OBP_RAW format. This
     will be downloaded from the firmware loader on probe to the device.
 
+- atmel,input_name: Override name of input device from the default.
+
 Example:
 
 	touch@4b {
Signature · 1 line
2.17.1
author
#30 lore raw

PATCH v10 24/55 Input: atmel_mxt_ts - add config checksum attribute to sysfs

In reply to Jiada Wang

From: karl tsou <[email protected]>

Add config checksum attribute to sysfs

(cherry picked from ndyer/linux/for-upstream commit 03477477ddbe5dcad42853ab3f84166a8f807acf)

[gdavis: Forward port and fix conflicts.]

[jiada: Add commit description
Rename mxt_config_crc_show to config_crc_show Replace DEVICE_ATTR with DEVICE_ATTR_RO]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 11 +++++++++++
 1 file changed, 11 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +11 -0
index dfc20c0ca0a5..67931cf64629 100644
@@ -3128,6 +3128,15 @@ static int mxt_configure_objects(struct mxt_data *data,
 	return error;
 }
 
+/* Configuration crc check sum is returned as hex xxxxxx */
+static ssize_t config_crc_show(struct device *dev,
+			       struct device_attribute *attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+
+	return scnprintf(buf, PAGE_SIZE, "%06x\n", data->config_crc);
+}
+
 /* Firmware Version is returned as Major.Minor.Build */
 static ssize_t fw_version_show(struct device *dev,
 			       struct device_attribute *attr, char *buf)
@@ -3481,12 +3490,14 @@ static DEVICE_ATTR_RO(fw_version);
 static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
 static DEVICE_ATTR_WO(update_cfg);
+static DEVICE_ATTR_RO(config_crc);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
 	&dev_attr_hw_version.attr,
 	&dev_attr_object.attr,
 	&dev_attr_update_cfg.attr,
+	&dev_attr_config_crc.attr,
 	NULL
 };
 
Signature · 1 line
2.17.1
author
#31 lore raw

PATCH v10 25/55 Input: atmel_mxt_ts - refactor firmware flash to extract context into struct

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Refactor firmware flash to extract context into struct

(cherry picked from ndyer/linux/for-upstream commit 1bbe20ff3dcd6612e7942c495929eae5c138ece2)

[jiada: Add commit description]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 59 +++++++++++++++---------
 1 file changed, 36 insertions(+), 23 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +36 -23
index 67931cf64629..2c4b68802bc3 100644
@@ -291,6 +291,22 @@ struct mxt_cfg {
 	struct mxt_info info;
 };
 
+/* Firmware frame structure */
+struct mxt_fw_frame {
+	__be16 size;
+	u8 data[];
+};
+
+/* Firmware update context */
+struct mxt_flash {
+	const struct firmware *fw;
+	struct mxt_fw_frame *frame;
+	loff_t pos;
+	size_t frame_size;
+	unsigned int count;
+	unsigned int retry;
+};
+
 /* Each client has this additional data */
 struct mxt_data {
 	struct i2c_client *client;
@@ -3244,21 +3260,17 @@ static int mxt_check_firmware_format(struct device *dev,
 static int mxt_load_fw(struct device *dev)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
-	const struct firmware *fw = NULL;
-	unsigned int frame_size;
-	unsigned int pos = 0;
-	unsigned int retry = 0;
-	unsigned int frame = 0;
+	struct mxt_flash f = { 0, };
 	int ret;
 
-	ret = request_firmware(&fw, data->fw_name, dev);
+	ret = request_firmware(&f.fw, data->fw_name, dev);
 	if (ret) {
 		dev_err(dev, "Unable to open firmware %s\n", data->fw_name);
 		return ret;
 	}
 
 	/* Check for incorrect enc file */
-	ret = mxt_check_firmware_format(dev, fw);
+	ret = mxt_check_firmware_format(dev, f.fw);
 	if (ret)
 		goto release_firmware;
 
@@ -3312,41 +3324,42 @@ static int mxt_load_fw(struct device *dev)
 			goto disable_irq;
 	}
 
-	while (pos < fw->size) {
+	while (f.pos < f.fw->size) {
+		f.frame = (struct mxt_fw_frame *)(f.fw->data + f.pos);
+
 		ret = mxt_check_bootloader(data, MXT_WAITING_FRAME_DATA, true);
 		if (ret)
 			goto disable_irq;
 
-		frame_size = ((*(fw->data + pos) << 8) | *(fw->data + pos + 1));
-
 		/* Take account of CRC bytes */
-		frame_size += 2;
+		f.frame_size = __be16_to_cpu(f.frame->size) + 2U;
 
 		/* Write one frame to device */
-		ret = mxt_bootloader_write(data, fw->data + pos, frame_size);
+		ret = mxt_bootloader_write(data, f.fw->data + f.pos,
+					   f.frame_size);
 		if (ret)
 			goto disable_irq;
 
 		ret = mxt_check_bootloader(data, MXT_FRAME_CRC_PASS, true);
 		if (ret) {
-			retry++;
+			f.retry++;
 
 			/* Back off by 20ms per retry */
-			msleep(retry * 20);
+			msleep(f.retry * 20);
 
-			if (retry > 20) {
+			if (f.retry > 20) {
 				dev_err(dev, "Retry count exceeded\n");
 				goto disable_irq;
 			}
 		} else {
-			retry = 0;
-			pos += frame_size;
-			frame++;
+			f.retry = 0;
+			f.pos += f.frame_size;
+			f.count++;
 		}
 
-		if (frame % 50 == 0)
-			dev_dbg(dev, "Sent %d frames, %d/%zd bytes\n",
-				frame, pos, fw->size);
+		if (f.count % 50 == 0)
+			dev_dbg(dev, "Sent %u frames, %lld/%zu bytes\n",
+				f.count, f.pos, f.fw->size);
 	}
 
 	/* Wait for flash. */
@@ -3355,7 +3368,7 @@ static int mxt_load_fw(struct device *dev)
 	if (ret)
 		goto disable_irq;
 
-	dev_dbg(dev, "Sent %d frames, %d bytes\n", frame, pos);
+	dev_dbg(dev, "Sent %u frames, %lld bytes\n", f.count, f.pos);
 
 	/*
 	 * Wait for device to reset. Some bootloader versions do not assert
@@ -3369,7 +3382,7 @@ static int mxt_load_fw(struct device *dev)
 disable_irq:
 	disable_irq(data->irq);
 release_firmware:
-	release_firmware(fw);
+	release_firmware(f.fw);
 	return ret;
 }
 
Signature · 1 line
2.17.1
author
#32 lore raw

PATCH v10 26/55 Input: atmel_mxt_ts - refactor code to enter bootloader into separate func

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Refactor code to enter bootloader into separate func

(cherry picked from ndyer/linux/for-upstream commit bedd706a32522b946467e15f4f4f24de86a1b4d7) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[jiada: Squash change from ndyer/linux/for-upstream commit d691d3ee6c6de84b38464a42 3207b3e23cb9dc3a
- Input: atmel_mxt_ts - check firmware format before entering bootloader Add commit description]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 43 +++++++++++++++---------
 1 file changed, 27 insertions(+), 16 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +27 -16
index 2c4b68802bc3..11b2f133e6be 100644
@@ -3257,23 +3257,10 @@ static int mxt_check_firmware_format(struct device *dev,
 	return -EINVAL;
 }
 
-static int mxt_load_fw(struct device *dev)
+static int mxt_enter_bootloader(struct mxt_data *data)
 {
-	struct mxt_data *data = dev_get_drvdata(dev);
-	struct mxt_flash f = { 0, };
 	int ret;
 
-	ret = request_firmware(&f.fw, data->fw_name, dev);
-	if (ret) {
-		dev_err(dev, "Unable to open firmware %s\n", data->fw_name);
-		return ret;
-	}
-
-	/* Check for incorrect enc file */
-	ret = mxt_check_firmware_format(dev, f.fw);
-	if (ret)
-		goto release_firmware;
-
 	if (data->suspended) {
 		if (data->suspend_mode == MXT_SUSPEND_REGULATOR)
 			mxt_regulator_enable(data);
@@ -3291,14 +3278,14 @@ static int mxt_load_fw(struct device *dev)
 		ret = mxt_t6_command(data, MXT_COMMAND_RESET,
 				     MXT_BOOT_VALUE, false);
 		if (ret)
-			goto release_firmware;
+			return ret;
 
 		msleep(MXT_RESET_TIME);
 
 		/* Do not need to scan since we know family ID */
 		ret = mxt_lookup_bootloader_address(data, 0);
 		if (ret)
-			goto release_firmware;
+			return ret;
 
 		mxt_sysfs_remove(data);
 		mxt_free_input_device(data);
@@ -3309,6 +3296,30 @@ static int mxt_load_fw(struct device *dev)
 
 	reinit_completion(&data->bl_completion);
 
+	return 0;
+}
+
+static int mxt_load_fw(struct device *dev)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	struct mxt_flash f = { 0, };
+	int ret;
+
+	ret = request_firmware(&f.fw, data->fw_name, dev);
+	if (ret) {
+		dev_err(dev, "Unable to open firmware %s\n", data->fw_name);
+		return ret;
+	}
+
+	/* Check for incorrect enc file */
+	ret = mxt_check_firmware_format(dev, f.fw);
+	if (ret)
+		goto release_firmware;
+
+	ret = mxt_enter_bootloader(data);
+	if (ret)
+		goto release_firmware;
+
 	ret = mxt_check_bootloader(data, MXT_WAITING_BOOTLOAD_CMD, false);
 	if (ret) {
 		/* Bootloader may still be unlocked from previous attempt */
Signature · 1 line
2.17.1
author
#33 lore raw

PATCH v10 27/55 Input: atmel_mxt_ts - combine bootloader version query with probe

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

This removes some complexity from the bootloader state machine, and means that we always output some debug about the version as soon as we start talking to the bootloader.

(cherry picked from ndyer/linux/for-upstream commit a2d141f170c80fea6663af98aab0be32abc0ddb0)

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 45 +++++++-----------------
 1 file changed, 13 insertions(+), 32 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +13 -32
index 11b2f133e6be..71d5781d3539 100644
@@ -558,47 +558,31 @@ static int mxt_probe_bootloader(struct mxt_data *data, bool alt_address)
 {
 	struct device *dev = &data->client->dev;
 	int error;
-	u8 val;
-	bool crc_failure;
+	u8 buf[3];
+	bool crc_failure, extended_id;
 
 	error = mxt_lookup_bootloader_address(data, alt_address);
 	if (error)
 		return error;
 
-	error = mxt_bootloader_read(data, &val, 1);
+	/* Check bootloader status and version information */
+	error = mxt_bootloader_read(data, buf, sizeof(buf));
 	if (error)
 		return error;
 
-	/* Check app crc fail mode */
-	crc_failure = (val & ~MXT_BOOT_STATUS_MASK) == MXT_APP_CRC_FAIL;
+	crc_failure = (buf[0] & ~MXT_BOOT_STATUS_MASK) == MXT_APP_CRC_FAIL;
+	extended_id = buf[0] & MXT_BOOT_EXTENDED_ID;
 
-	dev_err(dev, "Detected bootloader, status:%02X%s\n",
-			val, crc_failure ? ", APP_CRC_FAIL" : "");
+	dev_info(dev, "Found bootloader addr:%02x ID:%u%s%u%s\n",
+		 data->bootloader_addr,
+		 extended_id ? (buf[1] & MXT_BOOT_ID_MASK) : buf[0],
+		 extended_id ? " version:" : "",
+		 extended_id ? buf[2] : 0,
+		 crc_failure ? ", APP_CRC_FAIL" : "");
 
 	return 0;
 }
 
-static u8 mxt_get_bootloader_version(struct mxt_data *data, u8 val)
-{
-	struct device *dev = &data->client->dev;
-	u8 buf[3];
-
-	if (val & MXT_BOOT_EXTENDED_ID) {
-		if (mxt_bootloader_read(data, &buf[0], 3) != 0) {
-			dev_err(dev, "%s: i2c failure\n", __func__);
-			return val;
-		}
-
-		dev_dbg(dev, "Bootloader ID:%d Version:%d\n", buf[1], buf[2]);
-
-		return buf[0];
-	} else {
-		dev_dbg(dev, "Bootloader ID:%d\n", val & MXT_BOOT_ID_MASK);
-
-		return val;
-	}
-}
-
 static int mxt_check_bootloader(struct mxt_data *data, unsigned int state,
 				bool wait)
 {
@@ -632,9 +616,6 @@ static int mxt_check_bootloader(struct mxt_data *data, unsigned int state,
 	if (ret)
 		return ret;
 
-	if (state == MXT_WAITING_BOOTLOAD_CMD)
-		val = mxt_get_bootloader_version(data, val);
-
 	switch (state) {
 	case MXT_WAITING_BOOTLOAD_CMD:
 	case MXT_WAITING_FRAME_DATA:
@@ -3283,7 +3264,7 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 		msleep(MXT_RESET_TIME);
 
 		/* Do not need to scan since we know family ID */
-		ret = mxt_lookup_bootloader_address(data, 0);
+		ret = mxt_probe_bootloader(data, 0);
 		if (ret)
 			return ret;
 
Signature · 1 line
2.17.1
author
#34 lore raw

PATCH v10 28/55 Input: atmel_mxt_ts - improve bootloader state machine handling

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

The code is much clearer if we switch on the actual state the bootloader is in, rather than the state we want it to be in, and allows the removal of a goto retry tangle.

(cherry picked from ndyer/linux/for-upstream commit 463e15ee95ee6e6274017ff645839dbe34d75c99) [gdavis: Squash fix from George G. Davis:
- input: atmel_mxt_ts - Fix 'mxt_send_bootloader_cmd' was not declared warning]

[jiada: only wait on some status change,
cleanup code style]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 165 +++++++++++++----------
 1 file changed, 95 insertions(+), 70 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +95 -70
index 71d5781d3539..0f1c0e5c6169 100644
@@ -305,6 +305,9 @@ struct mxt_flash {
 	size_t frame_size;
 	unsigned int count;
 	unsigned int retry;
+	u8 previous;
+	bool complete;
+	bool wait;
 };
 
 /* Each client has this additional data */
@@ -583,15 +586,27 @@ static int mxt_probe_bootloader(struct mxt_data *data, bool alt_address)
 	return 0;
 }
 
-static int mxt_check_bootloader(struct mxt_data *data, unsigned int state,
-				bool wait)
+static int mxt_send_bootloader_cmd(struct mxt_data *data, bool unlock);
+
+static int mxt_write_firmware_frame(struct mxt_data *data, struct mxt_flash *f)
+{
+	f->frame = (struct mxt_fw_frame *)(f->fw->data + f->pos);
+
+	/* Take account of CRC bytes */
+	f->frame_size = __be16_to_cpu(f->frame->size) + 2U;
+
+	/* Write one frame to device */
+	return mxt_bootloader_write(data, f->fw->data + f->pos,
+				   f->frame_size);
+}
+
+static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 {
 	struct device *dev = &data->client->dev;
-	u8 val;
+	u8 state;
 	int ret;
 
-recheck:
-	if (wait) {
+	if (f->wait) {
 		/*
 		 * In application update mode, the interrupt
 		 * line signals state transitions. We must wait for the
@@ -607,40 +622,96 @@ static int mxt_check_bootloader(struct mxt_data *data, unsigned int state,
 			 * by writing length 0x000 to device (iff we are in
 			 * WAITING_FRAME_DATA state).
 			 */
-			dev_err(dev, "Update wait error %d\n", ret);
+			dev_warn(dev, "Update wait error %d\n", ret);
 			return ret;
 		}
 	}
 
-	ret = mxt_bootloader_read(data, &val, 1);
+	f->wait = false;
+
+	ret = mxt_bootloader_read(data, &state, 1);
 	if (ret)
 		return ret;
 
+	/* Remove don't care bits */
+	if (state & ~MXT_BOOT_STATUS_MASK)
+		state &= ~MXT_BOOT_STATUS_MASK;
+
 	switch (state) {
 	case MXT_WAITING_BOOTLOAD_CMD:
+		dev_info(dev, "Unlocking bootloader\n");
+		ret = mxt_send_bootloader_cmd(data, true);
+		if (ret)
+			return ret;
+		f->wait = true;
+
+		break;
+
 	case MXT_WAITING_FRAME_DATA:
-	case MXT_APP_CRC_FAIL:
-		val &= ~MXT_BOOT_STATUS_MASK;
+		if (f->previous != MXT_WAITING_BOOTLOAD_CMD &&
+		    f->previous != MXT_FRAME_CRC_PASS &&
+		    f->previous != MXT_FRAME_CRC_FAIL)
+			goto unexpected;
+
+		ret = mxt_write_firmware_frame(data, f);
+		if (ret)
+			return ret;
+
+		f->wait = true;
+
+		break;
+
+	case MXT_FRAME_CRC_CHECK:
+		if (f->previous != MXT_WAITING_FRAME_DATA)
+			goto unexpected;
+		f->wait = true;
 		break;
+
 	case MXT_FRAME_CRC_PASS:
-		if (val == MXT_FRAME_CRC_CHECK) {
-			goto recheck;
-		} else if (val == MXT_FRAME_CRC_FAIL) {
-			dev_err(dev, "Bootloader CRC fail\n");
-			return -EINVAL;
+		if (f->previous != MXT_FRAME_CRC_CHECK)
+			goto unexpected;
+
+		/* Next frame */
+		f->retry = 0;
+		f->pos += f->frame_size;
+		f->count++;
+		f->wait = true;
+
+		if (f->pos >= f->fw->size) {
+			f->complete = true;
+			dev_info(dev, "Sent %u frames, %zu bytes\n",
+				f->count, f->fw->size);
+		} else if (f->count % 50 == 0) {
+			dev_dbg(dev, "Sent %u frames, %lld/%zu bytes\n",
+				f->count, f->pos, f->fw->size);
+		}
+
+		break;
+
+	case MXT_FRAME_CRC_FAIL:
+		if (f->retry > 20) {
+			dev_err(dev, "Retry count exceeded\n");
+			return -EIO;
 		}
+
+		/* Back off by 20ms per retry */
+		dev_dbg(dev, "Bootloader frame CRC failure\n");
+		f->retry++;
+		f->wait = true;
+		msleep(f->retry * 20);
 		break;
+
 	default:
 		return -EINVAL;
 	}
 
-	if (val != state) {
-		dev_err(dev, "Invalid bootloader state %02X != %02X\n",
-			val, state);
-		return -EINVAL;
-	}
+	f->previous = state;
 
 	return 0;
+
+unexpected:
+	dev_err(dev, "Unexpected state transition\n");
+	return -EINVAL;
 }
 
 static int mxt_send_bootloader_cmd(struct mxt_data *data, bool unlock)
@@ -3301,57 +3372,13 @@ static int mxt_load_fw(struct device *dev)
 	if (ret)
 		goto release_firmware;
 
-	ret = mxt_check_bootloader(data, MXT_WAITING_BOOTLOAD_CMD, false);
-	if (ret) {
-		/* Bootloader may still be unlocked from previous attempt */
-		ret = mxt_check_bootloader(data, MXT_WAITING_FRAME_DATA, false);
-		if (ret)
-			goto disable_irq;
-	} else {
-		dev_info(dev, "Unlocking bootloader\n");
-
-		/* Unlock bootloader */
-		ret = mxt_send_bootloader_cmd(data, true);
+	while (true) {
+		ret = mxt_check_bootloader(data, &f);
 		if (ret)
-			goto disable_irq;
-	}
-
-	while (f.pos < f.fw->size) {
-		f.frame = (struct mxt_fw_frame *)(f.fw->data + f.pos);
-
-		ret = mxt_check_bootloader(data, MXT_WAITING_FRAME_DATA, true);
-		if (ret)
-			goto disable_irq;
-
-		/* Take account of CRC bytes */
-		f.frame_size = __be16_to_cpu(f.frame->size) + 2U;
-
-		/* Write one frame to device */
-		ret = mxt_bootloader_write(data, f.fw->data + f.pos,
-					   f.frame_size);
-		if (ret)
-			goto disable_irq;
-
-		ret = mxt_check_bootloader(data, MXT_FRAME_CRC_PASS, true);
-		if (ret) {
-			f.retry++;
-
-			/* Back off by 20ms per retry */
-			msleep(f.retry * 20);
-
-			if (f.retry > 20) {
-				dev_err(dev, "Retry count exceeded\n");
-				goto disable_irq;
-			}
-		} else {
-			f.retry = 0;
-			f.pos += f.frame_size;
-			f.count++;
-		}
+			return ret;
 
-		if (f.count % 50 == 0)
-			dev_dbg(dev, "Sent %u frames, %lld/%zu bytes\n",
-				f.count, f.pos, f.fw->size);
+		if (f.complete)
+			break;
 	}
 
 	/* Wait for flash. */
@@ -3360,7 +3387,6 @@ static int mxt_load_fw(struct device *dev)
 	if (ret)
 		goto disable_irq;
 
-	dev_dbg(dev, "Sent %u frames, %lld bytes\n", f.count, f.pos);
 
 	/*
 	 * Wait for device to reset. Some bootloader versions do not assert
@@ -3370,7 +3396,6 @@ static int mxt_load_fw(struct device *dev)
 	mxt_wait_for_completion(data, &data->bl_completion, MXT_FW_RESET_TIME);
 
 	data->in_bootloader = false;
-
 disable_irq:
 	disable_irq(data->irq);
 release_firmware:
Signature · 1 line
2.17.1
author
#35 lore raw

PATCH v10 29/55 Input: atmel_mxt_ts - rename bl_completion to chg_completion

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Rename bl_completion to chg_completion

(cherry picked from ndyer/linux/for-upstream commit dda8453bfb44216645ede798918a314d4fca2481) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[jiada: call complete(&data->chg_completion) only when in_bootloader is TRUE
Add commit description]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 23 +++++++++++------------
 1 file changed, 11 insertions(+), 12 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +11 -12
index 0f1c0e5c6169..ed142b24d12d 100644
@@ -382,9 +382,6 @@ struct mxt_data {
 	u8 T100_reportid_max;
 	u16 T107_address;
 
-	/* for fw update in bootloader */
-	struct completion bl_completion;
-
 	/* for reset handling */
 	struct completion reset_completion;
 
@@ -396,6 +393,9 @@ struct mxt_data {
 
 	enum mxt_suspend_mode suspend_mode;
 
+	/* for power up handling */
+	struct completion chg_completion;
+
 	/* Indicates whether device is in suspend */
 	bool suspended;
 
@@ -613,7 +613,7 @@ static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 		 * CHG assertion before reading the status byte.
 		 * Once the status byte has been read, the line is deasserted.
 		 */
-		ret = mxt_wait_for_completion(data, &data->bl_completion,
+		ret = mxt_wait_for_completion(data, &data->chg_completion,
 					      MXT_FW_CHG_TIMEOUT);
 		if (ret) {
 			/*
@@ -1414,8 +1414,7 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 	struct mxt_data *data = dev_id;
 
 	if (data->in_bootloader) {
-		/* bootloader state transition completion */
-		complete(&data->bl_completion);
+		complete(&data->chg_completion);
 		return IRQ_HANDLED;
 	}
 
@@ -2179,9 +2178,9 @@ static void mxt_regulator_enable(struct mxt_data *data)
 	msleep(MXT_CHG_DELAY);
 
 retry_wait:
-	reinit_completion(&data->bl_completion);
+	reinit_completion(&data->chg_completion);
 	data->in_bootloader = true;
-	error = mxt_wait_for_completion(data, &data->bl_completion,
+	error = mxt_wait_for_completion(data, &data->chg_completion,
 					MXT_POWERON_DELAY);
 	if (error == -EINTR)
 		goto retry_wait;
@@ -3346,7 +3345,7 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 		enable_irq(data->irq);
 	}
 
-	reinit_completion(&data->bl_completion);
+	reinit_completion(&data->chg_completion);
 
 	return 0;
 }
@@ -3382,7 +3381,7 @@ static int mxt_load_fw(struct device *dev)
 	}
 
 	/* Wait for flash. */
-	ret = mxt_wait_for_completion(data, &data->bl_completion,
+	ret = mxt_wait_for_completion(data, &data->chg_completion,
 				      MXT_FW_RESET_TIME);
 	if (ret)
 		goto disable_irq;
@@ -3393,7 +3392,7 @@ static int mxt_load_fw(struct device *dev)
 	 * the CHG line after bootloading has finished, so ignore potential
 	 * errors.
 	 */
-	mxt_wait_for_completion(data, &data->bl_completion, MXT_FW_RESET_TIME);
+	mxt_wait_for_completion(data, &data->chg_completion, MXT_FW_RESET_TIME);
 
 	data->in_bootloader = false;
 disable_irq:
@@ -3815,7 +3814,7 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	data->irq = client->irq;
 	i2c_set_clientdata(client, data);
 
-	init_completion(&data->bl_completion);
+	init_completion(&data->chg_completion);
 	init_completion(&data->reset_completion);
 	init_completion(&data->crc_completion);
 
Signature · 1 line
2.17.1
author
#36 lore raw

PATCH v10 30/55 Input: atmel_mxt_ts - make bootloader interrupt driven

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Make bootloader interrupt driven

(cherry picked from ndyer/linux/for-upstream commit 67a3eea0cfc724c3c2a7410ac064f74227c7c6ef) [gdavis: Resolve forward port conflicts due to applying upstream
commit 96a938aa214e ("Input: atmel_mxt_ts - remove platform
data support").]

[jiada: Replace two use msecs_to_jiffies() instead of HZ,

	remove check of &data->flash->work
	don't poll to call mxt_check_bootloader() in mxt_check_bootloader()]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 127 ++++++++++-------------
 1 file changed, 57 insertions(+), 70 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +57 -70
index ed142b24d12d..62f11afb26ae 100644
@@ -27,6 +27,7 @@
 #include <linux/gpio/consumer.h>
 #include <asm/unaligned.h>
 #include <linux/regulator/consumer.h>
+#include <linux/workqueue.h>
 #include <media/v4l2-device.h>
 #include <media/v4l2-ioctl.h>
 #include <media/videobuf2-v4l2.h>
@@ -218,6 +219,7 @@ enum t100_type {
 #define MXT_REGULATOR_DELAY	150	/* msec */
 #define MXT_CHG_DELAY		100	/* msec */
 #define MXT_POWERON_DELAY	150	/* msec */
+#define MXT_BOOTLOADER_WAIT	36E5	/* 1 minute */
 
 /* Command to unlock bootloader */
 #define MXT_UNLOCK_CMD_MSB	0xaa
@@ -299,6 +301,7 @@ struct mxt_fw_frame {
 
 /* Firmware update context */
 struct mxt_flash {
+	struct mxt_data *data;
 	const struct firmware *fw;
 	struct mxt_fw_frame *frame;
 	loff_t pos;
@@ -306,8 +309,8 @@ struct mxt_flash {
 	unsigned int count;
 	unsigned int retry;
 	u8 previous;
-	bool complete;
-	bool wait;
+	struct completion flash_completion;
+	struct delayed_work work;
 };
 
 /* Each client has this additional data */
@@ -356,6 +359,7 @@ struct mxt_data {
 	char *cfg_name;
 	const char *pcfg_name;
 	const char *input_name;
+	struct mxt_flash *flash;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -600,35 +604,19 @@ static int mxt_write_firmware_frame(struct mxt_data *data, struct mxt_flash *f)
 				   f->frame_size);
 }
 
-static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
+static int mxt_check_bootloader(struct mxt_data *data)
 {
 	struct device *dev = &data->client->dev;
+	struct mxt_flash *f = data->flash;
 	u8 state;
 	int ret;
 
-	if (f->wait) {
-		/*
-		 * In application update mode, the interrupt
-		 * line signals state transitions. We must wait for the
-		 * CHG assertion before reading the status byte.
-		 * Once the status byte has been read, the line is deasserted.
-		 */
-		ret = mxt_wait_for_completion(data, &data->chg_completion,
-					      MXT_FW_CHG_TIMEOUT);
-		if (ret) {
-			/*
-			 * TODO: handle -ERESTARTSYS better by terminating
-			 * fw update process before returning to userspace
-			 * by writing length 0x000 to device (iff we are in
-			 * WAITING_FRAME_DATA state).
-			 */
-			dev_warn(dev, "Update wait error %d\n", ret);
-			return ret;
-		}
+	/* Handle interrupt after download/flash process */
+	if (f->pos >= f->fw->size) {
+		complete(&f->flash_completion);
+		return 0;
 	}
 
-	f->wait = false;
-
 	ret = mxt_bootloader_read(data, &state, 1);
 	if (ret)
 		return ret;
@@ -643,7 +631,6 @@ static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 		ret = mxt_send_bootloader_cmd(data, true);
 		if (ret)
 			return ret;
-		f->wait = true;
 
 		break;
 
@@ -657,14 +644,11 @@ static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 		if (ret)
 			return ret;
 
-		f->wait = true;
-
 		break;
 
 	case MXT_FRAME_CRC_CHECK:
 		if (f->previous != MXT_WAITING_FRAME_DATA)
 			goto unexpected;
-		f->wait = true;
 		break;
 
 	case MXT_FRAME_CRC_PASS:
@@ -675,16 +659,13 @@ static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 		f->retry = 0;
 		f->pos += f->frame_size;
 		f->count++;
-		f->wait = true;
 
-		if (f->pos >= f->fw->size) {
-			f->complete = true;
+		if (f->pos >= f->fw->size)
 			dev_info(dev, "Sent %u frames, %zu bytes\n",
 				f->count, f->fw->size);
-		} else if (f->count % 50 == 0) {
+		else if (f->count % 50 == 0)
 			dev_dbg(dev, "Sent %u frames, %lld/%zu bytes\n",
 				f->count, f->pos, f->fw->size);
-		}
 
 		break;
 
@@ -697,7 +678,6 @@ static int mxt_check_bootloader(struct mxt_data *data, struct mxt_flash *f)
 		/* Back off by 20ms per retry */
 		dev_dbg(dev, "Bootloader frame CRC failure\n");
 		f->retry++;
-		f->wait = true;
 		msleep(f->retry * 20);
 		break;
 
@@ -1415,7 +1395,11 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 
 	if (data->in_bootloader) {
 		complete(&data->chg_completion);
-		return IRQ_HANDLED;
+
+		if (data->flash)
+			cancel_delayed_work_sync(&data->flash->work);
+
+		return IRQ_RETVAL(mxt_check_bootloader(data));
 	}
 
 	if (!data->object_table)
@@ -3316,16 +3300,13 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 		if (data->suspend_mode == MXT_SUSPEND_REGULATOR)
 			mxt_regulator_enable(data);
 
-		if (data->suspend_mode == MXT_SUSPEND_DEEP_SLEEP)
-			enable_irq(data->irq);
-
 		data->suspended = false;
 	}
 
 	if (!data->in_bootloader) {
-		/* Change to the bootloader mode */
-		data->in_bootloader = true;
+		disable_irq(data->irq);
 
+		/* Change to the bootloader mode */
 		ret = mxt_t6_command(data, MXT_COMMAND_RESET,
 				     MXT_BOOT_VALUE, false);
 		if (ret)
@@ -3338,67 +3319,73 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 		if (ret)
 			return ret;
 
+		data->in_bootloader = true;
 		mxt_sysfs_remove(data);
 		mxt_free_input_device(data);
 		mxt_free_object_table(data);
-	} else {
-		enable_irq(data->irq);
 	}
 
-	reinit_completion(&data->chg_completion);
+	dev_dbg(&data->client->dev, "Entered bootloader\n");
 
 	return 0;
 }
 
+static void mxt_fw_work(struct work_struct *work)
+{
+	struct mxt_flash *f =
+		container_of(work, struct mxt_flash, work.work);
+
+	mxt_check_bootloader(f->data);
+}
+
 static int mxt_load_fw(struct device *dev)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
-	struct mxt_flash f = { 0, };
 	int ret;
 
-	ret = request_firmware(&f.fw, data->fw_name, dev);
+	data->flash = devm_kzalloc(dev, sizeof(struct mxt_flash), GFP_KERNEL);
+	if (!data->flash)
+		return -ENOMEM;
+
+	data->flash->data = data;
+
+	ret = request_firmware(&data->flash->fw, data->fw_name, dev);
 	if (ret) {
 		dev_err(dev, "Unable to open firmware %s\n", data->fw_name);
-		return ret;
+		goto free;
 	}
 
 	/* Check for incorrect enc file */
-	ret = mxt_check_firmware_format(dev, f.fw);
+	ret = mxt_check_firmware_format(dev, data->flash->fw);
 	if (ret)
 		goto release_firmware;
 
-	ret = mxt_enter_bootloader(data);
-	if (ret)
-		goto release_firmware;
+	init_completion(&data->flash->flash_completion);
+	INIT_DELAYED_WORK(&data->flash->work, mxt_fw_work);
+	reinit_completion(&data->flash->flash_completion);
 
-	while (true) {
-		ret = mxt_check_bootloader(data, &f);
+	if (!data->in_bootloader) {
+		ret = mxt_enter_bootloader(data);
 		if (ret)
-			return ret;
-
-		if (f.complete)
-			break;
+			goto release_firmware;
 	}
 
-	/* Wait for flash. */
-	ret = mxt_wait_for_completion(data, &data->chg_completion,
-				      MXT_FW_RESET_TIME);
-	if (ret)
-		goto disable_irq;
+	enable_irq(data->irq);
 
+	/* Poll after 0.1s if no interrupt received */
+	schedule_delayed_work(&data->flash->work, msecs_to_jiffies(100));
 
-	/*
-	 * Wait for device to reset. Some bootloader versions do not assert
-	 * the CHG line after bootloading has finished, so ignore potential
-	 * errors.
-	 */
-	mxt_wait_for_completion(data, &data->chg_completion, MXT_FW_RESET_TIME);
+	/* Wait for flash. */
+	ret = mxt_wait_for_completion(data, &data->flash->flash_completion,
+				      MXT_BOOTLOADER_WAIT);
 
-	data->in_bootloader = false;
-disable_irq:
 	disable_irq(data->irq);
+	cancel_delayed_work_sync(&data->flash->work);
+	data->in_bootloader = false;
 release_firmware:
-	release_firmware(f.fw);
+	release_firmware(data->flash->fw);
+free:
+	devm_kfree(dev, data->flash);
 	return ret;
 }
 
Signature · 1 line
2.17.1
author
#37 lore raw

PATCH v10 31/55 Input: atmel_mxt_ts - delay enabling IRQ when not using regulators

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

The path of enabling the IRQ in the probe function is not safe in level triggered operation, if it was already powered up and there is a message waiting on the device (eg finger down) because the object table has not yet been read. This forces the ISR into a hard loop.

Delay enabling the interrupt until it is first needed.

(cherry picked from ndyer/linux/for-upstream commit 64c9dadc4a3250a185baf06ab0f628be45d5d9a0) [gdavis: Resolve forward port conflicts due to v4.14-rc1 commit
8cc8446b9b62 ("Input: atmel_mxt_ts - use more managed
resources") and applying upstream commit 96a938aa214e ("Input:
atmel_mxt_ts - remove platform data support").]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 40 +++++++++++++++---------
 1 file changed, 26 insertions(+), 14 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +26 -14
index 62f11afb26ae..c26470bcb38b 100644
@@ -1448,9 +1448,24 @@ static int mxt_acquire_irq(struct mxt_data *data)
 {
 	int error;
 
-	enable_irq(data->irq);
+	if (!data->irq) {
+		error = devm_request_threaded_irq(&data->client->dev,
+						  data->client->irq,
+						  NULL, mxt_interrupt,
+						  IRQF_ONESHOT,
+						  data->client->name, data);
+		if (error) {
+			dev_err(&data->client->dev, "Error requesting irq\n");
+			return error;
+		}
 
-	if (data->use_retrigen_workaround) {
+		/* Presence of data->irq means IRQ initialised */
+		data->irq = data->client->irq;
+	} else {
+		enable_irq(data->irq);
+	}
+
+	if (data->object_table && data->use_retrigen_workaround) {
 		error = mxt_process_messages_until_invalid(data);
 		if (error)
 			return error;
@@ -3370,7 +3385,9 @@ static int mxt_load_fw(struct device *dev)
 			goto release_firmware;
 	}
 
-	enable_irq(data->irq);
+	ret = mxt_acquire_irq(data);
+	if (ret)
+		goto release_firmware;
 
 	/* Poll after 0.1s if no interrupt received */
 	schedule_delayed_work(&data->flash->work, msecs_to_jiffies(100));
@@ -3798,7 +3815,6 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		 client->adapter->nr, client->addr);
 
 	data->client = client;
-	data->irq = client->irq;
 	i2c_set_clientdata(client, data);
 
 	init_completion(&data->chg_completion);
@@ -3826,26 +3842,22 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		return error;
 	}
 
-	error = devm_request_threaded_irq(&client->dev, client->irq,
-					  NULL, mxt_interrupt, IRQF_ONESHOT,
-					  client->name, data);
-	if (error) {
-		dev_err(&client->dev, "Failed to register interrupt\n");
-		return error;
-	}
-
 	if (data->suspend_mode == MXT_SUSPEND_REGULATOR) {
+		error = mxt_acquire_irq(data);
+		if (error)
+			return error;
+
 		error = mxt_probe_regulators(data);
 		if (error)
 			return error;
+
+		disable_irq(data->irq);
 	} else if (data->reset_gpio) {
 		msleep(MXT_RESET_GPIO_TIME);
 		gpiod_set_value(data->reset_gpio, 1);
 		msleep(MXT_RESET_INVALID_CHG);
 	}
 
-	disable_irq(data->irq);
-
 	error = mxt_initialize(data);
 	if (error)
 		return error;
Signature · 1 line
2.17.1
#38 lore raw

Re: [PATCH v10 31/55] Input: atmel_mxt_ts - delay enabling IRQ when not using regulators

31.03.2020 13:50, Jiada Wang пишет:

Quoted 11 lines

From: Nick Dyer <[email protected]>

The path of enabling the IRQ in the probe function is not safe in level triggered operation, if it was already powered up and there is a message waiting on the device (eg finger down) because the object table has not yet been read. This forces the ISR into a hard loop.

Delay enabling the interrupt until it is first needed.

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_acquire_irq(struct mxt_data *data) · line 1463

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +2 -1
-	if (data->use_retrigen_workaround) {
+		/* Presence of data->irq means IRQ initialised */
+		data->irq = data->client->irq;

IIRC, IRQ=0 could be a valid interrupt since this is a "virtual" interrupt number.

...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3843

Quoted 4 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -0
 	init_completion(&data->chg_completion);
@@ -3826,26 +3842,22 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		return error;
 	}

What about just to tell the IRQ core not to enable the interrupt handling during of the devm_request_threaded_irq()?

To achieve that, add this line here:

	irq_set_status_flags(client->irq, IRQ_NOAUTOEN);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3831

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -3
-	error = devm_request_threaded_irq(&client->dev, client->irq,
-					  NULL, mxt_interrupt, IRQF_ONESHOT,
-					  client->name, data);

...

Then the interrupt will be requested in the disabled state and it will be enabled only after the first enable_irq() invocation.

#39 lore raw

Re: [PATCH v10 31/55] Input: atmel_mxt_ts - delay enabling IRQ when not using regulators

Hi Dmitry

Quoted 32 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 11 lines

From: Nick Dyer <[email protected]>

The path of enabling the IRQ in the probe function is not safe in level triggered operation, if it was already powered up and there is a message waiting on the device (eg finger down) because the object table has not yet been read. This forces the ISR into a hard loop.

Delay enabling the interrupt until it is first needed.

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_acquire_irq(struct mxt_data *data) · line 1463

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +2 -1
-	if (data->use_retrigen_workaround) {
+		/* Presence of data->irq means IRQ initialised */
+		data->irq = data->client->irq;

IIRC, IRQ=0 could be a valid interrupt since this is a "virtual" interrupt number.

...

Review static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3843

Quoted 4 lines
  	init_completion(&data->chg_completion);
(quoted diff) +0 -0
@@ -3826,26 +3842,22 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
  		return error;
  	}

What about just to tell the IRQ core not to enable the interrupt handling during of the devm_request_threaded_irq()?

To achieve that, add this line here:

	irq_set_status_flags(client->irq, IRQ_NOAUTOEN);
Quoted 8 lines

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id) · line 3831

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -3
-	error = devm_request_threaded_irq(&client->dev, client->irq,
-					  NULL, mxt_interrupt, IRQF_ONESHOT,
-					  client->name, data);

...

Then the interrupt will be requested in the disabled state and it will be enabled only after the first enable_irq() invocation.

thanks, I think your solution makes more sense,
I will replace with your suggested solution in next version

Thanks,
Jiada

#40 lore raw

Re: [PATCH v10 31/55] Input: atmel_mxt_ts - delay enabling IRQ when not using regulators

15.04.2020 17:44, Wang, Jiada пишет:
..

Quoted 2 lines

thanks, I think your solution makes more sense,
I will replace with your suggested solution in next version

Please feel free to CC me on the next version.

author
#41 lore raw

PATCH v10 32/55 Input: atmel_mxt_ts - implement I2C retries

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Some maXTouch chips (eg mXT1386) will not respond on the first I2C request when they are in a sleep state. It must be retried after a delay for the chip to wake up.

(cherry picked from ndyer/linux/for-upstream commit 63fd7a2cd03c3a572a5db39c52f4856819e1835d)

[gdavis: Forward port and fix conflicts.]
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 45 ++++++++++++++++--------
 1 file changed, 30 insertions(+), 15 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +30 -15
index c26470bcb38b..daa77eb9561b 100644
@@ -216,6 +216,7 @@ enum t100_type {
 #define MXT_CRC_TIMEOUT		1000	/* msec */
 #define MXT_FW_RESET_TIME	3000	/* msec */
 #define MXT_FW_CHG_TIMEOUT	300	/* msec */
+#define MXT_WAKEUP_TIME		25	/* msec */
 #define MXT_REGULATOR_DELAY	150	/* msec */
 #define MXT_CHG_DELAY		100	/* msec */
 #define MXT_POWERON_DELAY	150	/* msec */
@@ -720,6 +721,7 @@ static int __mxt_read_chunk(struct i2c_client *client,
 	struct i2c_msg xfer[2];
 	u8 buf[2];
 	int ret;
+	bool retry = false;
 
 	buf[0] = reg & 0xff;
 	buf[1] = (reg >> 8) & 0xff;
@@ -736,17 +738,22 @@ static int __mxt_read_chunk(struct i2c_client *client,
 	xfer[1].len = len;
 	xfer[1].buf = val;
 
-	ret = i2c_transfer(client->adapter, xfer, 2);
-	if (ret == 2) {
-		ret = 0;
-	} else {
-		if (ret >= 0)
-			ret = -EIO;
-		dev_err(&client->dev, "%s: i2c transfer failed (%d)\n",
-			__func__, ret);
+retry_read:
+	ret = i2c_transfer(client->adapter, xfer, ARRAY_SIZE(xfer));
+	if (ret != ARRAY_SIZE(xfer)) {
+		if (!retry) {
+			dev_dbg(&client->dev, "%s: i2c retry\n", __func__);
+			msleep(MXT_WAKEUP_TIME);
+			retry = true;
+			goto retry_read;
+		} else {
+			dev_err(&client->dev, "%s: i2c transfer failed (%d)\n",
+				__func__, ret);
+			return -EIO;
+		}
 	}
 
-	return ret;
+	return 0;
 }
 
 static int __mxt_read_reg(struct i2c_client *client,
@@ -777,6 +784,7 @@ static int __mxt_write_reg(struct i2c_client *client, u16 reg, u16 len,
 	u8 *buf;
 	size_t count;
 	int ret;
+	bool retry = false;
 
 	count = len + 2;
 	buf = kmalloc(count, GFP_KERNEL);
@@ -787,14 +795,21 @@ static int __mxt_write_reg(struct i2c_client *client, u16 reg, u16 len,
 	buf[1] = (reg >> 8) & 0xff;
 	memcpy(&buf[2], val, len);
 
+retry_write:
 	ret = i2c_master_send(client, buf, count);
-	if (ret == count) {
-		ret = 0;
-	} else {
-		if (ret >= 0)
+	if (ret != count) {
+		if (!retry) {
+			dev_dbg(&client->dev, "%s: i2c retry\n", __func__);
+			msleep(MXT_WAKEUP_TIME);
+			retry = true;
+			goto retry_write;
+		} else {
+			dev_err(&client->dev, "%s: i2c send failed (%d)\n",
+				__func__, ret);
 			ret = -EIO;
-		dev_err(&client->dev, "%s: i2c send failed (%d)\n",
-			__func__, ret);
+		}
+	} else {
+		ret = 0;
 	}
 
 	kfree(buf);
Signature · 1 line
2.17.1
author
#42 lore raw

PATCH v10 33/55 Input: atmel_mxt_ts - orientation is not present in hover

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

When in hover, the orientation information is not sent

(cherry picked from ndyer/linux/for-upstream commit 0c885d5bd276bd9240c43aa046fc407cbe2ae864)

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 4 ----
 1 file changed, 4 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +0 -4
index daa77eb9561b..b1f54bf207b8 100644
@@ -1026,10 +1026,6 @@ static void mxt_proc_t100_message(struct mxt_data *data, u8 *message)
 			distance = MXT_DISTANCE_HOVERING;
 			hover = true;
 			active = true;
-
-			if (data->t100_aux_vect)
-				orientation = message[data->t100_aux_vect];
-
 			break;
 
 		case MXT_T100_TYPE_FINGER:
Signature · 1 line
2.17.1
author
#43 lore raw

PATCH v10 34/55 Input: atmel_mxt_ts - implement debug output for messages

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Add a debug switch which causes all messages from the touch controller to be dumped to the dmesg log with a set prefix "MXT MSG:". This is used by Atmel user-space utilities to debug touch operation. Enabling this output does impact touch performance.

(cherry picked from ndyer/linux/for-upstream commit 3c3fcfdd4889dfeb1c80ae8cd94a622c6342b06a)

[gdavis: Forward port and fix conflicts.]

[jiada: Rename mxt_debug_enable_store to debug_enable_store
Rename mxt_debug_enable_show to debug_enable_show Replace DEVICE_ATTR with DEVICE_ATTR_RW]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 46 ++++++++++++++++++++++--
 1 file changed, 43 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +43 -3
index b1f54bf207b8..656d956f9c93 100644
@@ -335,6 +335,7 @@ struct mxt_data {
 	u8 t100_aux_ampl;
 	u8 t100_aux_area;
 	u8 t100_aux_vect;
+	bool debug_enabled;
 	u8 max_reportid;
 	u32 config_crc;
 	u32 info_crc;
@@ -460,8 +461,11 @@ static bool mxt_object_readable(unsigned int type)
 
 static void mxt_dump_message(struct mxt_data *data, u8 *message)
 {
-	dev_dbg(&data->client->dev, "message: %*ph\n",
-		data->T5_msg_size, message);
+	/* debug message with prefix 'MXT MSG:' used by
+	 * Atmel user-space utilities to debug touch operation
+	 */
+	dev_dbg(&data->client->dev, "MXT MSG: %*ph\n",
+		       data->T5_msg_size, message);
 }
 
 static int mxt_wait_for_completion(struct mxt_data *data,
@@ -1211,6 +1215,7 @@ static void mxt_proc_t93_messages(struct mxt_data *data, u8 *msg)
 static int mxt_proc_message(struct mxt_data *data, u8 *message)
 {
 	u8 report_id = message[0];
+	bool dump = data->debug_enabled;
 
 	if (report_id == MXT_RPTID_NOMSG)
 		return 0;
@@ -1245,9 +1250,12 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	} else if (report_id == data->T93_reportid) {
 		mxt_proc_t93_messages(data, message);
 	} else {
-		mxt_dump_message(data, message);
+		dump = true;
 	}
 
+	if (dump)
+		mxt_dump_message(data, message);
+
 	return 1;
 }
 
@@ -3519,6 +3527,36 @@ static ssize_t update_cfg_store(struct device *dev,
 	return ret;
 }
 
+static ssize_t debug_enable_show(struct device *dev,
+	struct device_attribute *attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	char c;
+
+	c = data->debug_enabled ? '1' : '0';
+	return scnprintf(buf, PAGE_SIZE, "%c\n", c);
+}
+
+static ssize_t debug_enable_store(struct device *dev,
+	struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	u8 i;
+	ssize_t ret;
+
+	if (kstrtou8(buf, 0, &i) == 0 && i < 2) {
+		data->debug_enabled = (i == 1);
+
+		dev_dbg(dev, "%s\n", i ? "debug enabled" : "debug disabled");
+		ret = count;
+	} else {
+		dev_dbg(dev, "debug_enabled write error\n");
+		ret = -EINVAL;
+	}
+
+	return ret;
+}
+
 static DEVICE_ATTR_WO(update_fw);
 
 static struct attribute *mxt_fw_attrs[] = {
@@ -3535,6 +3573,7 @@ static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
 static DEVICE_ATTR_WO(update_cfg);
 static DEVICE_ATTR_RO(config_crc);
+static DEVICE_ATTR_RW(debug_enable);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
@@ -3542,6 +3581,7 @@ static struct attribute *mxt_attrs[] = {
 	&dev_attr_object.attr,
 	&dev_attr_update_cfg.attr,
 	&dev_attr_config_crc.attr,
+	&dev_attr_debug_enable.attr,
 	NULL
 };
 
Signature · 1 line
2.17.1
author
#44 lore raw

PATCH v10 35/55 Input: atmel_mxt_ts - implement improved debug message interface

In reply to Jiada Wang

From: Nick Dyer <[email protected]>

Implement improved debug message interface

(cherry picked from ndyer/linux/for-upstream commit 078569c13c88dcb6f8d882bfe17168587712df7d) [gdavis: Resolve forward port conflicts due to v4.14.51 commit

	 960fe000b1d3 ("Input: atmel_mxt_ts - fix the firmware
	 update").]

[gdavis: Squash fixes from Dirk Behme:
- Input: atmel_mxt_ts - add missing unlock in error path
- Input: atmel_mxt_ts - add missing unlock in error path
- Input: atmel_mxt_ts - call mxt_debug_msg_remove() in error path
- Input: atmel_mxt_ts - protect debug_v2_enabled by mutex

[gdavis: Squash fix from Vladimir Zapolskiy:
- Input: atmel_mxt_ts - simplify debug_msg binary attribute handling]

[jiada: Rename mxt_debug_notify_show to debug_notify_show
Rename mxt_debug_v2_enable_store to debug_v2_enable_store Add debug_v2_enable_show
Replace DEVICE_ATTR with DEVICE_ATTR_[RW|RO] ]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 203 ++++++++++++++++++++++-
 1 file changed, 201 insertions(+), 2 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +201 -2
index 656d956f9c93..9476326596d9 100644
@@ -242,6 +242,8 @@ enum t100_type {
 
 #define MXT_PIXELS_PER_MM	20
 
+#define DEBUG_MSG_MAX		200
+
 struct mxt_info {
 	u8 family_id;
 	u8 variant_id;
@@ -336,6 +338,11 @@ struct mxt_data {
 	u8 t100_aux_area;
 	u8 t100_aux_vect;
 	bool debug_enabled;
+	bool debug_v2_enabled;
+	u8 *debug_msg_data;
+	u16 debug_msg_count;
+	struct bin_attribute *debug_msg_attr;
+	struct mutex debug_msg_lock;
 	u8 max_reportid;
 	u32 config_crc;
 	u32 info_crc;
@@ -468,6 +475,144 @@ static void mxt_dump_message(struct mxt_data *data, u8 *message)
 		       data->T5_msg_size, message);
 }
 
+static void mxt_debug_msg_enable(struct mxt_data *data)
+{
+	struct device *dev = &data->client->dev;
+
+	if (data->debug_v2_enabled)
+		return;
+
+	mutex_lock(&data->debug_msg_lock);
+
+	data->debug_msg_data = kcalloc(DEBUG_MSG_MAX,
+				data->T5_msg_size, GFP_KERNEL);
+	if (!data->debug_msg_data) {
+		mutex_unlock(&data->debug_msg_lock);
+		return;
+	}
+
+	data->debug_v2_enabled = true;
+	mutex_unlock(&data->debug_msg_lock);
+
+	dev_dbg(dev, "Enabled message output\n");
+}
+
+static void mxt_debug_msg_disable(struct mxt_data *data)
+{
+	struct device *dev = &data->client->dev;
+
+	if (!data->debug_v2_enabled)
+		return;
+
+	mutex_lock(&data->debug_msg_lock);
+
+	data->debug_v2_enabled = false;
+
+	kfree(data->debug_msg_data);
+	data->debug_msg_data = NULL;
+	data->debug_msg_count = 0;
+	mutex_unlock(&data->debug_msg_lock);
+	dev_dbg(dev, "Disabled message output\n");
+}
+
+static void mxt_debug_msg_add(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+
+	mutex_lock(&data->debug_msg_lock);
+
+	if (!data->debug_msg_data) {
+		mutex_unlock(&data->debug_msg_lock);
+		dev_err(dev, "No buffer!\n");
+		return;
+	}
+
+	if (data->debug_msg_count < DEBUG_MSG_MAX) {
+		memcpy(data->debug_msg_data +
+		       data->debug_msg_count * data->T5_msg_size,
+		       msg,
+		       data->T5_msg_size);
+		data->debug_msg_count++;
+	} else {
+		dev_dbg(dev, "Discarding %u messages\n", data->debug_msg_count);
+		data->debug_msg_count = 0;
+	}
+
+	mutex_unlock(&data->debug_msg_lock);
+
+	sysfs_notify(&data->client->dev.kobj, NULL, "debug_notify");
+}
+
+static ssize_t mxt_debug_msg_read(struct file *filp, struct kobject *kobj,
+	struct bin_attribute *bin_attr, char *buf, loff_t off, size_t bytes)
+{
+	struct device *dev = container_of(kobj, struct device, kobj);
+	struct mxt_data *data = dev_get_drvdata(dev);
+	int count;
+	size_t bytes_read;
+
+	if (!data->debug_msg_data) {
+		dev_err(dev, "No buffer!\n");
+		return 0;
+	}
+
+	count = bytes / data->T5_msg_size;
+
+	if (count > DEBUG_MSG_MAX)
+		count = DEBUG_MSG_MAX;
+
+	mutex_lock(&data->debug_msg_lock);
+
+	if (count > data->debug_msg_count)
+		count = data->debug_msg_count;
+
+	bytes_read = count * data->T5_msg_size;
+
+	memcpy(buf, data->debug_msg_data, bytes_read);
+	data->debug_msg_count = 0;
+
+	mutex_unlock(&data->debug_msg_lock);
+
+	return bytes_read;
+}
+
+static struct bin_attribute debug_msg_attr = {
+	.attr = {
+		.name = "debug_msg",
+		.mode = 0444,
+	},
+	.read = mxt_debug_msg_read,
+};
+
+static int mxt_debug_msg_init(struct mxt_data *data)
+{
+	/*
+	 * Binary attribute is not used in callback, removal is done by name,
+	 * so it is safe to update a single struct bin_attribute entity
+	 */
+	debug_msg_attr.size = data->T5_msg_size * DEBUG_MSG_MAX;
+
+	if (sysfs_create_bin_file(&data->client->dev.kobj,
+				  &debug_msg_attr) < 0) {
+		dev_err(&data->client->dev, "Failed to create %s\n",
+			debug_msg_attr.attr.name);
+		return -EINVAL;
+	}
+
+	data->debug_msg_attr = &debug_msg_attr;
+
+	return 0;
+}
+
+static void mxt_debug_msg_remove(struct mxt_data *data)
+{
+	if (data->debug_msg_attr) {
+		sysfs_remove_bin_file(&data->client->dev.kobj,
+				      data->debug_msg_attr);
+		data->debug_msg_attr = NULL;
+	}
+}
+
 static int mxt_wait_for_completion(struct mxt_data *data,
 				   struct completion *comp,
 				   unsigned int timeout_ms)
@@ -1256,6 +1401,9 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	if (dump)
 		mxt_dump_message(data, message);
 
+	if (data->debug_v2_enabled)
+		mxt_debug_msg_add(data, message);
+
 	return 1;
 }
 
@@ -2674,6 +2822,10 @@ static int mxt_initialize(struct mxt_data *data)
 	if (error)
 		return error;
 
+	error = mxt_debug_msg_init(data);
+	if (error)
+		goto err_free_sysfs;
+
 	if (data->cfg_name) {
 		error = request_firmware_nowait(THIS_MODULE, true,
 						data->cfg_name,
@@ -2683,16 +2835,18 @@ static int mxt_initialize(struct mxt_data *data)
 		if (error) {
 			dev_err(&client->dev, "Failed to invoke firmware loader: %d\n",
 				error);
-			goto err_free_sysfs;
+			goto err_free_dbg_msg;
 		}
 	} else {
 		error = mxt_configure_objects(data, NULL);
 		if (error)
-			goto err_free_sysfs;
+			goto err_free_dbg_msg;
 	}
 
 	return 0;
 
+err_free_dbg_msg:
+	mxt_debug_msg_remove(data);
 err_free_sysfs:
 	mxt_sysfs_remove(data);
 	return error;
@@ -3354,6 +3508,7 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 			return ret;
 
 		data->in_bootloader = true;
+		mxt_debug_msg_remove(data);
 		mxt_sysfs_remove(data);
 		mxt_free_input_device(data);
 		mxt_free_object_table(data);
@@ -3527,6 +3682,44 @@ static ssize_t update_cfg_store(struct device *dev,
 	return ret;
 }
 
+static ssize_t debug_notify_show(struct device *dev,
+	struct device_attribute *attr, char *buf)
+{
+	return sprintf(buf, "0\n");
+}
+
+static ssize_t debug_v2_enable_store(struct device *dev,
+	struct device_attribute *attr, const char *buf, size_t count)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	u8 i;
+	ssize_t ret;
+
+	if (kstrtou8(buf, 0, &i) == 0 && i < 2) {
+		if (i == 1)
+			mxt_debug_msg_enable(data);
+		else
+			mxt_debug_msg_disable(data);
+
+		ret = count;
+	} else {
+		dev_dbg(dev, "debug_enabled write error\n");
+		ret = -EINVAL;
+	}
+
+	return ret;
+}
+
+static ssize_t debug_v2_enable_show(struct device *dev,
+	struct device_attribute *attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	char c;
+
+	c = data->debug_v2_enabled ? '1' : '0';
+	return scnprintf(buf, PAGE_SIZE, "%c\n", c);
+}
+
 static ssize_t debug_enable_show(struct device *dev,
 	struct device_attribute *attr, char *buf)
 {
@@ -3574,6 +3767,8 @@ static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
 static DEVICE_ATTR_WO(update_cfg);
 static DEVICE_ATTR_RO(config_crc);
 static DEVICE_ATTR_RW(debug_enable);
+static DEVICE_ATTR_RW(debug_v2_enable);
+static DEVICE_ATTR_RO(debug_notify);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
@@ -3582,6 +3777,8 @@ static struct attribute *mxt_attrs[] = {
 	&dev_attr_update_cfg.attr,
 	&dev_attr_config_crc.attr,
 	&dev_attr_debug_enable.attr,
+	&dev_attr_debug_v2_enable.attr,
+	&dev_attr_debug_notify.attr,
 	NULL
 };
 
@@ -3871,6 +4068,7 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	init_completion(&data->chg_completion);
 	init_completion(&data->reset_completion);
 	init_completion(&data->crc_completion);
+	mutex_init(&data->debug_msg_lock);
 
 	data->suspend_mode = dmi_check_system(chromebook_T9_suspend_dmi) ?
 		MXT_SUSPEND_T9_CTRL : MXT_SUSPEND_DEEP_SLEEP;
@@ -3929,6 +4127,7 @@ static int mxt_remove(struct i2c_client *client)
 
 	disable_irq(data->irq);
 	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
+	mxt_debug_msg_remove(data);
 	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
Signature · 1 line
2.17.1
author
#45 lore raw

PATCH v10 36/55 Input: atmel_mxt_ts - eliminate data->raw_info_block

In reply to Jiada Wang

Dynamically allocated in mxt_read_info_block() buffer buf is assigned both to data->info and data->raw_info_block, having both data->info and data->raw_info_block is redundant and confusing.

This patch eliminates data->raw_info_block.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 7 ++-----
 1 file changed, 2 insertions(+), 5 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +2 -5
index 9476326596d9..36a87d9b9c83 100644
@@ -323,7 +323,6 @@ struct mxt_data {
 	char phys[64];		/* device physical location */
 	struct mxt_object *object_table;
 	struct mxt_info *info;
-	void *raw_info_block;
 	unsigned int irq;
 	unsigned int max_x;
 	unsigned int max_y;
@@ -2067,9 +2066,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	v4l2_device_unregister(&data->dbg.v4l2);
 #endif
 	data->object_table = NULL;
+	kfree(data->info);
 	data->info = NULL;
-	kfree(data->raw_info_block);
-	data->raw_info_block = NULL;
 	kfree(data->msg_buf);
 	data->msg_buf = NULL;
 	data->T5_address = 0;
@@ -2238,7 +2236,7 @@ static int mxt_read_info_block(struct mxt_data *data)
 	u8 *crc_ptr;
 
 	/* If info block already allocated, free it */
-	if (data->raw_info_block)
+	if (data->info)
 		mxt_free_object_table(data);
 
 	/* Read 7-byte ID information block starting at address 0 */
@@ -2289,7 +2287,6 @@ static int mxt_read_info_block(struct mxt_data *data)
 		goto err_free_mem;
 	}
 
-	data->raw_info_block = id_buf;
 	data->info = (struct mxt_info *)id_buf;
 
 	dev_info(&client->dev,
Signature · 1 line
2.17.1
author
#46 lore raw

PATCH v10 37/55 Input: atmel_mxt_ts - Change call-points of mxt_free_* functions

In reply to Jiada Wang

From: Kautuk Consul <[email protected]>

Revamping the code to call mxt_free_object_table and mxt_free_input_device functions only in the following scenarios and code paths:
1) The error path of the mxt_probe() entry point
2) The mxt_remove de-init path entry point
3) All paths which definitely expect to populate the object table like:
- the mxt_update_fw_store path which first calls mxt_load_fw and then resorts to calling mxt_initialize itself.
- the mxt_read_info_block function which attempts to fill in the object table itself as the main non-error part of the logic.
4) All paths in the code expected to definitely allocate and register the input device such as:
- the mxt_update_fw_store path which first calls mxt_load_fw and then resorts to calling mxt_initialize itself.
- the mxt_update_cfg_store function which will call mxt_configure_objects.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 19 ++++++++++---------
 1 file changed, 10 insertions(+), 9 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +10 -9
index 36a87d9b9c83..860e8ddd4c6b 100644
@@ -3336,21 +3336,21 @@ static int mxt_configure_objects(struct mxt_data *data,
 	error = mxt_init_t7_power_cfg(data);
 	if (error) {
 		dev_err(dev, "Failed to initialize power cfg\n");
-		goto err_free_object_table;
+		return error;
 	}
 
 	if (cfg) {
 		error = mxt_update_cfg(data, cfg);
 		if (error) {
 			dev_warn(dev, "Error %d updating config\n", error);
-			goto err_free_object_table;
+			return error;
 		}
 	}
 
 	if (data->multitouch) {
 		error = mxt_initialize_input_device(data);
 		if (error)
-			goto err_free_object_table;
+			return error;
 	} else {
 		dev_warn(dev, "No touch object detected\n");
 	}
@@ -3358,10 +3358,6 @@ static int mxt_configure_objects(struct mxt_data *data,
 	mxt_debug_init(data);
 
 	return 0;
-
-err_free_object_table:
-	mxt_free_object_table(data);
-	return error;
 }
 
 /* Configuration crc check sum is returned as hex xxxxxx */
@@ -4106,16 +4102,21 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 
 	error = mxt_initialize(data);
 	if (error)
-		return error;
+		goto err_free_object;
 
 	error = sysfs_create_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (error) {
 		dev_err(&client->dev, "Failure %d creating fw sysfs group\n",
 			error);
-		return error;
+		goto err_free_object;
 	}
 
 	return 0;
+
+err_free_object:
+	mxt_free_input_device(data);
+	mxt_free_object_table(data);
+	return error;
 }
 
 static int mxt_remove(struct i2c_client *client)
Signature · 1 line
2.17.1
author
#47 lore raw

PATCH v10 38/55 Input: atmel_mxt_ts - rely on calculated_crc rather than file config_crc

In reply to Jiada Wang

From: Kautuk Consul <[email protected]>

We now prefer to rely on the calculated CRC and not on the CRC stored in the file.

The new logic is as follows:
1) stored CRC of file != calculated CRC of file, then refuse the possible corrupted file
2) calculated CRC of file != CRC of configuration in controller, then update configuration in controller
3) calculated CRC of file == CRC of configuration in controller, then ignore configuration file

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 65 +++++++++++++-----------
 1 file changed, 36 insertions(+), 29 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +36 -29
index 860e8ddd4c6b..11e65f32b1a2 100644
@@ -1905,7 +1905,7 @@ static int mxt_update_cfg(struct mxt_data *data, const struct firmware *fw)
 	int ret;
 	int offset;
 	int i;
-	u32 info_crc, config_crc, calculated_crc;
+	u32 info_crc, config_crc, calculated_crc = 0;
 	u16 crc_start = 0;
 
 	/* Make zero terminated copy of the OBP_RAW file */
@@ -1968,30 +1968,6 @@ static int mxt_update_cfg(struct mxt_data *data, const struct firmware *fw)
 	}
 	cfg.raw_pos += offset;
 
-	/*
-	 * The Info Block CRC is calculated over mxt_info and the object
-	 * table. If it does not match then we are trying to load the
-	 * configuration from a different chip or firmware version, so
-	 * the configuration CRC is invalid anyway.
-	 */
-	if (info_crc == data->info_crc) {
-		if (config_crc == 0 || data->config_crc == 0) {
-			dev_info(dev, "CRC zero, attempting to apply config\n");
-		} else if (config_crc == data->config_crc) {
-			dev_dbg(dev, "Config CRC 0x%06X: OK\n",
-				 data->config_crc);
-			ret = 0;
-			goto release_raw;
-		} else {
-			dev_info(dev, "Config CRC 0x%06X: does not match file 0x%06X\n",
-				 data->config_crc, config_crc);
-		}
-	} else {
-		dev_warn(dev,
-			 "Warning: Info CRC error - device=0x%06X file=0x%06X\n",
-			 data->info_crc, info_crc);
-	}
-
 	/* Malloc memory to store configuration */
 	cfg.start_ofs = MXT_OBJECT_START +
 			data->info->object_num * sizeof(struct mxt_object) +
@@ -2015,14 +1991,45 @@ static int mxt_update_cfg(struct mxt_data *data, const struct firmware *fw)
 	else
 		dev_warn(dev, "Could not find CRC start\n");
 
-	if (crc_start > cfg.start_ofs) {
+	if (crc_start > cfg.start_ofs)
 		calculated_crc = mxt_calculate_crc(cfg.mem,
 						   crc_start - cfg.start_ofs,
 						   cfg.mem_size);
 
-		if (config_crc > 0 && config_crc != calculated_crc)
-			dev_warn(dev, "Config CRC in file inconsistent, calculated=%06X, file=%06X\n",
-				 calculated_crc, config_crc);
+	/* If the CRC stored in the file is not the same as what
+	 * was calculated by mxt_calculate_crc, this means we
+	 * have to refuse the config file and abort download.
+	 */
+	if (config_crc != calculated_crc) {
+		dev_warn(dev,
+			 "Config CRC in file inconsistent, calculated=%06X, file=%06X\n",
+			 calculated_crc, config_crc);
+		ret = 0;
+		goto release_mem;
+	}
+
+	/*
+	 * The Info Block CRC is calculated over mxt_info and the object
+	 * table. If it does not match then we are trying to load the
+	 * configuration from a different chip or firmware version, so
+	 * the configuration CRC is invalid anyway.
+	 */
+	if (info_crc == data->info_crc) {
+		if (config_crc == 0 || data->config_crc == 0) {
+			dev_info(dev, "CRC zero, attempting to apply config\n");
+		} else if (config_crc == data->config_crc) {
+			dev_dbg(dev, "Config CRC 0x%06X: OK\n",
+				data->config_crc);
+			ret = 0;
+			goto release_mem;
+		} else {
+			dev_info(dev, "Config CRC 0x%06X: does not match file 0x%06X\n",
+				 data->config_crc, config_crc);
+		}
+	} else {
+		dev_warn(dev,
+			 "Warning: Info CRC error - device=0x%06X file=0x%06X\n",
+			 data->info_crc, info_crc);
 	}
 
 	ret = mxt_upload_cfg_mem(data, &cfg);
Signature · 1 line
2.17.1
author
#48 lore raw

PATCH v10 39/55 input: atmel_mxt_ts: export GPIO reset line via sysfs

In reply to Jiada Wang

From: "George G. Davis" <[email protected]>

N.B. Modifying the atmel_mxt_ts GPIO reset line during operation will cause problems with normal driver operation. This feature is provided as a diagnostic debug aid. It does not take into consideration any pending operations which may be in progress. Modifying the atmel_mxt_ts GPIO reset line at any time will inevitably cause the driver to fail.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 21 +++++++++++++++++++++
 1 file changed, 21 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +21 -0
index 11e65f32b1a2..cfeeacdd4084 100644
@@ -4091,6 +4091,19 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		return error;
 	}
 
+	if (data->reset_gpio) {
+		error = gpiod_export(data->reset_gpio, 0);
+		if (error)
+			return error;
+
+		error = gpiod_export_link(&client->dev, "reset",
+					  data->reset_gpio);
+		if (error) {
+			gpiod_unexport(data->reset_gpio);
+			return error;
+		}
+	}
+
 	if (data->suspend_mode == MXT_SUSPEND_REGULATOR) {
 		error = mxt_acquire_irq(data);
 		if (error)
@@ -4123,6 +4136,10 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 err_free_object:
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
+	if (data->reset_gpio) {
+		sysfs_remove_link(&client->dev.kobj, "reset");
+		gpiod_unexport(data->reset_gpio);
+	}
 	return error;
 }
 
@@ -4132,6 +4149,10 @@ static int mxt_remove(struct i2c_client *client)
 
 	disable_irq(data->irq);
 	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
+	if (data->reset_gpio) {
+		sysfs_remove_link(&client->dev.kobj, "reset");
+		gpiod_unexport(data->reset_gpio);
+	}
 	mxt_debug_msg_remove(data);
 	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
Signature · 1 line
2.17.1
author
#49 lore raw

PATCH v10 40/55 input: atmel_mxt_ts: Add Missing Delay for reset handling of Atmel touch panel controller in detachable displays.

In reply to Jiada Wang

From: keerthikumarp <[email protected]>

In case of attached display, the touchpanel reset is controlled via imx gpio's from atmel driver and the delay between touchpanel reset and the time at which the chip becomes capable to communicate with the host processor, has be taken care.

However in case of detachable displays, the touchpanel reset is controlled via a deserializer gpio which is triggered just before the atmel driver is probed.The delay between touchpanel reset and the time at which the chip becomes capable to communicate (as specified in datasheet) was not being accounted for. This patch introduces that delay.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 4 ++++
 1 file changed, 4 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +4 -0
index cfeeacdd4084..fb7cd06cab53 100644
@@ -4118,6 +4118,10 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		msleep(MXT_RESET_GPIO_TIME);
 		gpiod_set_value(data->reset_gpio, 1);
 		msleep(MXT_RESET_INVALID_CHG);
+	} else {
+		dev_dbg(&client->dev,
+			"atmel reset pin not found in device tree");
+		msleep(MXT_RESET_TIME);
 	}
 
 	error = mxt_initialize(data);
Signature · 1 line
2.17.1
author
#50 lore raw

PATCH v10 41/55 Input: atmel_mxt_ts: Add support for run self-test routine.

In reply to Jiada Wang

From: Nikhil Ravindran <[email protected]>

The self test object T25 runs self test routines in device to find faults Sysfs entry add to start self test routine and read back the test results for atmel touchcontrollers.The feature will be used for A-IVI and CAF projects.

[jiada: Rename mxt_t25_selftest_show to t25_selftest_show
Rename mxt_t25_selftest_store to t25_selftest_show Rename attr t25 to t25_selftest
Replace DEVICE_ATTR with DEVICE_ATTR_RW]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 113 +++++++++++++++++++++++
 1 file changed, 113 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +113 -0
index fb7cd06cab53..54cd6219cbaa 100644
@@ -336,6 +336,9 @@ struct mxt_data {
 	u8 t100_aux_ampl;
 	u8 t100_aux_area;
 	u8 t100_aux_vect;
+	u16 T25_address;
+	u8  T25_reportid;
+	u8  t25_msg[6];
 	bool debug_enabled;
 	bool debug_v2_enabled;
 	u8 *debug_msg_data;
@@ -413,6 +416,8 @@ struct mxt_data {
 
 	/* Indicates whether device is updating configuration */
 	bool updating_config;
+
+	bool t25_status;
 };
 
 struct mxt_vb2_buffer {
@@ -1356,6 +1361,24 @@ static void mxt_proc_t93_messages(struct mxt_data *data, u8 *msg)
 	dev_info(dev, "T93 report double tap %d\n", status);
 }
 
+static void mxt_proc_t25_messages(struct mxt_data *data, u8 *msg)
+{
+	struct device *dev = &data->client->dev;
+
+	/* Output debug if status has changed */
+	dev_dbg(dev, "T25 Status 0x%x Info: %x %x %x %x %x\n",
+		msg[1],
+		msg[2],
+		msg[3],
+		msg[4],
+		msg[5],
+		msg[6]);
+
+	/* Save current status */
+	memcpy(&data->t25_msg[0], &msg[1], sizeof(data->t25_msg));
+	data->t25_status = false;
+}
+
 static int mxt_proc_message(struct mxt_data *data, u8 *message)
 {
 	u8 report_id = message[0];
@@ -1386,6 +1409,8 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	} else if (report_id == data->T19_reportid) {
 		mxt_input_button(data, message);
 		data->update_input = true;
+	} else if (report_id == data->T25_reportid) {
+		mxt_proc_t25_messages(data, message);
 	} else if (report_id >= data->T15_reportid_min
 		   && report_id <= data->T15_reportid_max) {
 		mxt_proc_t15_messages(data, message);
@@ -1610,6 +1635,86 @@ static int mxt_t6_command(struct mxt_data *data, u16 cmd_offset,
 	return 0;
 }
 
+static int mxt_t25_command(struct mxt_data *data, u8 cmd, bool wait)
+{
+	u16 reg;
+	int timeout_counter = 0;
+	int ret;
+	u8  val[2];
+
+	reg = data->T25_address;
+	val[0] = 0x3;
+	val[1] = cmd;
+
+	data->t25_status = true;
+	ret = __mxt_write_reg(data->client, reg, sizeof(val), val);
+	if (ret) {
+		data->t25_status = false;
+		return ret;
+	}
+
+	if (!wait)
+		return 0;
+
+	do {
+		msleep(MXT_WAKEUP_TIME);
+		ret = __mxt_read_reg(data->client, reg + 1, 1, &val[1]);
+		if (ret)
+			return ret;
+	} while ((val[1] != 0) && (timeout_counter++ <= 100));
+
+	if (timeout_counter > 100) {
+		dev_err(&data->client->dev, "Command failed!\n");
+		data->t25_status = false;
+		return -EIO;
+	}
+	return 0;
+}
+
+/* Firmware Version is returned as Major.Minor.Build */
+static ssize_t t25_selftest_show(struct device *dev, struct
+				 device_attribute * attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	ssize_t offset = 0;
+
+	if (data->t25_status)
+		return -EAGAIN;
+
+	if (data->t25_msg[0] == 0xFE)
+		offset += scnprintf(buf, PAGE_SIZE, "PASS\n");
+	else
+		offset += scnprintf(buf, PAGE_SIZE, "FAILED\n");
+
+	offset += scnprintf(buf + offset, PAGE_SIZE, "%x %x %x %x %x %x\n",
+		 data->t25_msg[0],
+		 data->t25_msg[1],
+		 data->t25_msg[2],
+		 data->t25_msg[3],
+		 data->t25_msg[4],
+		 data->t25_msg[5]);
+	return offset;
+}
+
+static ssize_t t25_selftest_store(struct device *dev, struct
+				  device_attribute * attr, const char *buf,
+				  size_t count)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	u32 cmd;
+	int ret;
+
+	ret = kstrtou32(buf, 0, &cmd);
+	if (ret)
+		return ret;
+
+	if (mxt_t25_command(data, (u8)cmd, 1) == 0)
+		return count;
+
+	dev_dbg(dev, "mxt_t25_cmd_store write cmd %x error\n", cmd);
+	return -EINVAL;
+}
+
 static int mxt_acquire_irq(struct mxt_data *data)
 {
 	int error;
@@ -2088,6 +2193,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T15_reportid_max = 0;
 	data->T18_address = 0;
 	data->T19_reportid = 0;
+	data->T25_address = 0;
+	data->T25_reportid = 0;
 	data->T42_reportid_min = 0;
 	data->T42_reportid_max = 0;
 	data->T44_address = 0;
@@ -2175,6 +2282,10 @@ static int mxt_parse_object_table(struct mxt_data *data,
 		case MXT_SPT_COMMSCONFIG_T18:
 			data->T18_address = object->start_address;
 			break;
+		case MXT_SPT_SELFTEST_T25:
+			data->T25_address = object->start_address;
+			data->T25_reportid = min_id;
+			break;
 		case MXT_PROCI_TOUCHSUPPRESSION_T42:
 			data->T42_reportid_min = min_id;
 			data->T42_reportid_max = max_id;
@@ -3769,6 +3880,7 @@ static DEVICE_ATTR_RO(config_crc);
 static DEVICE_ATTR_RW(debug_enable);
 static DEVICE_ATTR_RW(debug_v2_enable);
 static DEVICE_ATTR_RO(debug_notify);
+static DEVICE_ATTR_RW(t25_selftest);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
@@ -3779,6 +3891,7 @@ static struct attribute *mxt_attrs[] = {
 	&dev_attr_debug_enable.attr,
 	&dev_attr_debug_v2_enable.attr,
 	&dev_attr_debug_notify.attr,
+	&dev_attr_t25_selftest.attr,
 	NULL
 };
 
Signature · 1 line
2.17.1
author
#51 lore raw

PATCH v10 42/55 Input: atmel_mxt_ts: Limit the max bytes transferred in an i2c transaction

In reply to Jiada Wang

From: Balasubramani Vivekanandan <[email protected]>

In mxt_process_messages_until_invalid() function, driver tries to read all possible reportid in a single i2c transaction. Number of bytes read is limited by the max_reportid parameter.
If the max_reportid is a very large value, then a large chunk of bytes will be requested from the controller in a single i2c transaction. This transaction can fail due to timeout. This is visible when the Atmel controller is connected to the SOC via a i2c mux hardware.

mxt_process_messages_t44() reads the T44 message which contains the pending T5 message count. If the number of pending T5 messages returned by T44 message is too high then there is a risk of i2c transaction timeout while reading T5 messages in mxt_process_messages_t44().

New property 'atmel,mtu' is created. This property limits the maximum number of bytes that can read/transferred in an i2c transcation

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 65 +++++++++++++++++++-----
 1 file changed, 51 insertions(+), 14 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +51 -14
index 54cd6219cbaa..384b53717139 100644
@@ -417,6 +417,7 @@ struct mxt_data {
 	/* Indicates whether device is updating configuration */
 	bool updating_config;
 
+	unsigned int mtu;
 	bool t25_status;
 };
 
@@ -1462,11 +1463,32 @@ static int mxt_read_and_process_messages(struct mxt_data *data, u8 count)
 	return num_valid;
 }
 
+static u8 mxt_max_msg_read_count(struct mxt_data *data, u8 max_T5_msg_count)
+{
+	u8 T5_msg_count_limit = data->mtu / data->T5_msg_size;
+
+	if (!data->mtu)
+		return max_T5_msg_count;
+
+	if (data->mtu < data->T5_msg_size) {
+		WARN(1, "mtu set is lesser than the T5 message size\n");
+		/* Return count of 1, as fallback */
+		return 1;
+	}
+	/*
+	 * Return maximum number of T5 messages in single i2c transaction
+	 * based on "atmel,mtu" property.
+	 */
+	return min(T5_msg_count_limit, max_T5_msg_count);
+}
+
 static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 {
 	struct device *dev = &data->client->dev;
 	int ret;
-	u8 count, num_left;
+	u8 T5_msg_count, total_pending;
+	u8 total_processed = 0;
+	u8 processed_valid = 0;
 
 	/* Read T44 and T5 together */
 	ret = __mxt_read_reg(data->client, data->T44_address,
@@ -1476,18 +1498,19 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 		return IRQ_NONE;
 	}
 
-	count = data->msg_buf[0];
+	T5_msg_count = data->msg_buf[0];
 
 	/*
 	 * This condition may be caused by the CHG line being configured in
 	 * Mode 0. It results in unnecessary I2C operations but it is benign.
 	 */
-	if (count == 0)
+	if (!T5_msg_count)
 		return IRQ_NONE;
 
-	if (count > data->max_reportid) {
-		dev_warn(dev, "T44 count %d exceeded max report id\n", count);
-		count = data->max_reportid;
+	if (T5_msg_count > data->max_reportid) {
+		dev_warn(dev, "T44 count %d exceeded max report id\n",
+			 T5_msg_count);
+		T5_msg_count = data->max_reportid;
 	}
 
 	/* Process first message */
@@ -1497,16 +1520,25 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 		return IRQ_NONE;
 	}
 
-	num_left = count - 1;
+	total_pending = T5_msg_count - 1;
+	if (!total_pending)
+		goto end;
 
 	/* Process remaining messages if necessary */
-	if (num_left) {
-		ret = mxt_read_and_process_messages(data, num_left);
+	T5_msg_count = mxt_max_msg_read_count(data, total_pending);
+
+	do {
+		if ((total_pending - total_processed) < T5_msg_count)
+			T5_msg_count = total_pending - total_processed;
+		ret = mxt_read_and_process_messages(data, T5_msg_count);
 		if (ret < 0)
 			goto end;
-		else if (ret != num_left)
-			dev_warn(dev, "Unexpected invalid message\n");
-	}
+		total_processed += T5_msg_count;
+		processed_valid += ret;
+	} while (total_processed < total_pending);
+
+	if (processed_valid != total_pending)
+		dev_warn(dev, "Unexpected invalid message\n");
 
 end:
 	if (data->update_input) {
@@ -1521,9 +1553,10 @@ static int mxt_process_messages_until_invalid(struct mxt_data *data)
 {
 	struct device *dev = &data->client->dev;
 	int count, read;
-	u8 tries = 2;
+	int tries;
 
-	count = data->max_reportid;
+	count = mxt_max_msg_read_count(data, data->max_reportid);
+	tries = (data->max_reportid / count) + 1;
 
 	/* Read messages until we force an invalid */
 	do {
@@ -4120,6 +4153,10 @@ static int mxt_parse_device_properties(struct mxt_data *data)
 
 	device_property_read_u32(dev, "atmel,suspend-mode",
 				 &data->suspend_mode);
+	device_property_read_u32(dev, "atmel,mtu", &data->mtu);
+	if (data->mtu)
+		dev_dbg(dev, "mtu is set as %d\n", data->mtu);
+
 	return 0;
 }
 
Signature · 1 line
2.17.1
author
#52 lore raw

PATCH v10 43/55 dt-bindings: input: atmel: support to set max bytes transferred

In reply to Jiada Wang

Add support to set max bytes transferred in an i2c transaction

---
 Documentation/devicetree/bindings/input/atmel,maxtouch.txt | 3 +++
 1 file changed, 3 insertions(+)
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +3 -0
index 126266891cdc..37a798cab0dd 100644
@@ -45,6 +45,8 @@ Optional properties for main touchpad device:
 
 - atmel,input_name: Override name of input device from the default.
 
+- atmel,mtu: Maximum number of bytes that can read/transferred in an i2c transaction
+
 Example:
 
 	touch@4b {
@@ -52,4 +54,5 @@ Example:
 		reg = <0x4b>;
 		interrupt-parent = <&gpio>;
 		interrupts = <TEGRA_GPIO(W, 3) IRQ_TYPE_LEVEL_LOW>;
+		atmel,mtu = <200>
 	};
Signature · 1 line
2.17.1
#53 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

31.03.2020 13:50, Jiada Wang пишет:

Quoted 22 lines

Add support to set max bytes transferred in an i2c transaction

---
 Documentation/devicetree/bindings/input/atmel,maxtouch.txt | 3 +++
 1 file changed, 3 insertions(+)
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +2 -0
index 126266891cdc..37a798cab0dd 100644
@@ -45,6 +45,8 @@ Optional properties for main touchpad device:
 
 - atmel,input_name: Override name of input device from the default.
 
+- atmel,mtu: Maximum number of bytes that can read/transferred in an i2c transaction
+
 Example:
 
 	touch@4b {

Documentation/devicetree/bindings/input/atmel,maxtouch.txt Example: · line 58

Quoted 5 lines
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +1 -0
@@ -52,4 +54,5 @@ Example:
 		reg = <0x4b>;
 		interrupt-parent = <&gpio>;
 		interrupts = <TEGRA_GPIO(W, 3) IRQ_TYPE_LEVEL_LOW>;
+		atmel,mtu = <200>
 	};

Is this a software (firmware) limitation which varies from version to version?

If the limitation is per-hardware version, then perhaps should be better to have a hardcoded per-hardware table of limits in the driver?

#54 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

Hi Dmitry

Quoted 20 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 19 lines

Add support to set max bytes transferred in an i2c transaction

---
  Documentation/devicetree/bindings/input/atmel,maxtouch.txt | 3 +++
  1 file changed, 3 insertions(+)
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +2 -0
index 126266891cdc..37a798cab0dd 100644
@@ -45,6 +45,8 @@ Optional properties for main touchpad device:
 
  - atmel,input_name: Override name of input device from the default.
 
+- atmel,mtu: Maximum number of bytes that can read/transferred in an i2c transaction
+
  Example:
 
Quoted 12 lines

Documentation/devicetree/bindings/input/atmel,maxtouch.txt Example: · line 58

Quoted 8 lines
Documentation/devicetree/bindings/input/atmel,maxtouch.txt +1 -0
  	touch@4b {
@@ -52,4 +54,5 @@ Example:
  		reg = <0x4b>;
  		interrupt-parent = <&gpio>;
  		interrupts = <TEGRA_GPIO(W, 3) IRQ_TYPE_LEVEL_LOW>;
+		atmel,mtu = <200>
  	};

Is this a software (firmware) limitation which varies from version to version?

the timeout issue trying to be addressed in this patch is from software,
one of our board a Serializer/Deserializer bridge exists between the SoC (imx6) and the Atmel touch controller. imx6 i2c controller driver has a timeout value(100ms) for each i2c transaction,
Large i2c read transaction failed to complete within this timeout value and therefore imx6 i2c controller driver aborts the transaction and returns failure.

Therefore this patch was created to split the large i2c transaction into smaller chunks which can complete within the timeout defined by i2c controller driver.

Thanks,
Jiada

Quoted 3 lines

If the limitation is per-hardware version, then perhaps should be better to have a hardcoded per-hardware table of limits in the driver?

#55 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

07.04.2020 12:27, Wang, Jiada пишет:
..

Quoted 13 lines
Quoted 3 lines

Is this a software (firmware) limitation which varies from version to version?

the timeout issue trying to be addressed in this patch is from software, one of our board a Serializer/Deserializer bridge exists between the SoC (imx6) and the Atmel touch controller.
imx6 i2c controller driver has a timeout value(100ms) for each i2c transaction,
Large i2c read transaction failed to complete within this timeout value and therefore imx6 i2c controller driver aborts the transaction and returns failure.

Quoted 3 lines

Therefore this patch was created to split the large i2c transaction into smaller chunks which can complete
within the timeout defined by i2c controller driver.

Isn't it possible to use the max_read/write_len of the generic struct i2c_adapter_quirks for limiting the transfer size?

BTW, it looks like the i.MX I2C driver doesn't specify the i2c_adapter_quirks, which probably needs to be fixed.

#56 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

Hi Dmitry

Quoted 19 lines · Dmitry Osipenko

07.04.2020 12:27, Wang, Jiada пишет:
..

Quoted 13 lines
Quoted 3 lines

Is this a software (firmware) limitation which varies from version to version?

the timeout issue trying to be addressed in this patch is from software, one of our board a Serializer/Deserializer bridge exists between the SoC (imx6) and the Atmel touch controller.
imx6 i2c controller driver has a timeout value(100ms) for each i2c transaction,
Large i2c read transaction failed to complete within this timeout value and therefore imx6 i2c controller driver aborts the transaction and returns failure.

Quoted 3 lines

Therefore this patch was created to split the large i2c transaction into smaller chunks which can complete
within the timeout defined by i2c controller driver.

Quoted 6 lines

Isn't it possible to use the max_read/write_len of the generic struct i2c_adapter_quirks for limiting the transfer size?

BTW, it looks like the i.MX I2C driver doesn't specify the i2c_adapter_quirks, which probably needs to be fixed.

yes, i.MX I2C driver can specify i2c_adapter_quirks to limit the size be transferred in one transaction.

But even in this case, mxt_process_messages_t44() fails when it tries to transfer data count larger than max_read/write_len set in i.MX I2C driver, which we would like to avoid.

Thanks,
Jiada

#57 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

09.04.2020 09:25, Wang, Jiada пишет:

Quoted 28 lines

Hi Dmitry

Quoted 19 lines · Dmitry Osipenko

07.04.2020 12:27, Wang, Jiada пишет:
..

Quoted 13 lines
Quoted 3 lines

Is this a software (firmware) limitation which varies from version to version?

the timeout issue trying to be addressed in this patch is from software, one of our board a Serializer/Deserializer bridge exists between the SoC (imx6) and the Atmel touch controller.
imx6 i2c controller driver has a timeout value(100ms) for each i2c transaction,
Large i2c read transaction failed to complete within this timeout value and therefore imx6 i2c controller driver aborts the transaction and returns failure.

Quoted 3 lines

Therefore this patch was created to split the large i2c transaction into smaller chunks which can complete
within the timeout defined by i2c controller driver.

Quoted 6 lines

Isn't it possible to use the max_read/write_len of the generic struct i2c_adapter_quirks for limiting the transfer size?

BTW, it looks like the i.MX I2C driver doesn't specify the i2c_adapter_quirks, which probably needs to be fixed.

Quoted 6 lines

yes, i.MX I2C driver can specify i2c_adapter_quirks to limit the size be transferred in one transaction.

But even in this case, mxt_process_messages_t44() fails when it tries to transfer data count larger than max_read/write_len set in i.MX I2C driver, which we would like to avoid.

IIUC, the transfer's limitation is a part of I2C controller hardware and not the touch controller, so it should be wrong to describe that limitation in the maxtouch's DT node.

I meant that we probably could set the data->mtu based on i2c_client->adapter->quirks->max_read and then the DT property shouldn't be needed, couldn't this be done?

The I2C core only rejects transfers that don't fit into the max_read/write_len and nothing more.

#58 lore raw

Re: [PATCH v10 43/55] dt-bindings: input: atmel: support to set max bytes transferred

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

09.04.2020 09:25, Wang, Jiada пишет:

Quoted 28 lines

Hi Dmitry

Quoted 19 lines · Dmitry Osipenko

07.04.2020 12:27, Wang, Jiada пишет:
..

Quoted 13 lines
Quoted 3 lines

Is this a software (firmware) limitation which varies from version to version?

the timeout issue trying to be addressed in this patch is from software, one of our board a Serializer/Deserializer bridge exists between the SoC (imx6) and the Atmel touch controller.
imx6 i2c controller driver has a timeout value(100ms) for each i2c transaction,
Large i2c read transaction failed to complete within this timeout value and therefore imx6 i2c controller driver aborts the transaction and returns failure.

Quoted 3 lines

Therefore this patch was created to split the large i2c transaction into smaller chunks which can complete
within the timeout defined by i2c controller driver.

Quoted 6 lines

Isn't it possible to use the max_read/write_len of the generic struct i2c_adapter_quirks for limiting the transfer size?

BTW, it looks like the i.MX I2C driver doesn't specify the i2c_adapter_quirks, which probably needs to be fixed.

Quoted 6 lines

yes, i.MX I2C driver can specify i2c_adapter_quirks to limit the size be transferred in one transaction.

But even in this case, mxt_process_messages_t44() fails when it tries to transfer data count larger than max_read/write_len set in i.MX I2C driver, which we would like to avoid.

Quoted 8 lines

IIUC, the transfer's limitation is a part of I2C controller hardware and not the touch controller, so it should be wrong to describe that limitation in the maxtouch's DT node.

I meant that we probably could set the data->mtu based on i2c_client->adapter->quirks->max_read and then the DT property shouldn't be needed, couldn't this be done?
Thanks, now I understand your point,

and yes, by this way, we can address the I2C controller limitation issue by its own configuration.

I will replace this commit with your proposed solution

Thanks,
jiada

Quoted 3 lines

The I2C core only rejects transfers that don't fit into the max_read/write_len and nothing more.

author
#59 lore raw

PATCH v10 44/55 Input: atmel_mxt_ts: return error from mxt_process_messages_until_invalid()

In reply to Jiada Wang

From: Dean Jenkins <[email protected]>

mxt_process_messages_until_invalid() failed to propagate the error code from mxt_read_and_process_messages() so return the error code.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 2 ++
 1 file changed, 2 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +2 -0
index 384b53717139..bbec5a8e0bb6 100644
@@ -1561,6 +1561,8 @@ static int mxt_process_messages_until_invalid(struct mxt_data *data)
 	/* Read messages until we force an invalid */
 	do {
 		read = mxt_read_and_process_messages(data, count);
+		if (read < 0)
+			return read;
 		if (read < count)
 			return 0;
 	} while (--tries);
Signature · 1 line
2.17.1
author
#60 lore raw

PATCH v10 45/55 Input: Atmel: improve error handling in mxt_start()

In reply to Jiada Wang

From: Deepak Das <[email protected]>

mxt_start() does not return error in any of the failure cases which will allow input_dev->open() to return success even in case of any failure.

This commit modifies mxt_start() to return error in failure cases.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 31 ++++++++++++------------
 1 file changed, 16 insertions(+), 15 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +16 -15
index bbec5a8e0bb6..0f7ee987a9c5 100644
@@ -3981,12 +3981,13 @@ static int mxt_start(struct mxt_data *data)
 
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
-		mxt_soft_reset(data);
-
+		ret = mxt_soft_reset(data);
+		if (ret)
+			break;
 		/* Touch enable */
 		/* 0x83 = SCANEN | RPTEN | ENABLE */
-		mxt_write_object(data,
-				MXT_TOUCH_MULTI_T9, MXT_T9_CTRL, 0x83);
+		ret = mxt_write_object(data,
+				       MXT_TOUCH_MULTI_T9, MXT_T9_CTRL, 0x83);
 		break;
 
 	case MXT_SUSPEND_REGULATOR:
@@ -4000,27 +4001,26 @@ static int mxt_start(struct mxt_data *data)
 		 * Discard any touch messages still in message buffer
 		 * from before chip went to sleep
 		 */
-		mxt_process_messages_until_invalid(data);
+		ret = mxt_process_messages_until_invalid(data);
+		if (ret)
+			break;
 
 		ret = mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
 		if (ret)
-			return ret;
+			break;
 
 		/* Recalibrate since chip has been in deep sleep */
 		ret = mxt_t6_command(data, MXT_COMMAND_CALIBRATE, 1, false);
 		if (ret)
-			return ret;
+			break;
 
 		ret = mxt_acquire_irq(data);
-		if (ret)
-			return ret;
-
-		break;
 	}
 
-	data->suspended = false;
+	if (!ret)
+		data->suspended = false;
 
-	return 0;
+	return ret;
 }
 
 static int mxt_stop(struct mxt_data *data)
@@ -4343,6 +4343,7 @@ static int __maybe_unused mxt_resume(struct device *dev)
 	struct i2c_client *client = to_i2c_client(dev);
 	struct mxt_data *data = i2c_get_clientdata(client);
 	struct input_dev *input_dev = data->input_dev;
+	int ret = 0;
 
 	if (!input_dev)
 		return 0;
@@ -4352,11 +4353,11 @@ static int __maybe_unused mxt_resume(struct device *dev)
 	mutex_lock(&input_dev->mutex);
 
 	if (input_dev->users)
-		mxt_start(data);
+		ret = mxt_start(data);
 
 	mutex_unlock(&input_dev->mutex);
 
-	return 0;
+	return ret;
 }
 
 static SIMPLE_DEV_PM_OPS(mxt_pm_ops, mxt_suspend, mxt_resume);
Signature · 1 line
2.17.1
author
#61 lore raw

PATCH v10 46/55 Input: Atmel: improve error handling in mxt_initialize()

In reply to Jiada Wang

From: Deepak Das <[email protected]>

Currently mxt_initialize() tries to probe bootloader mode even if valid bootloader address is not specified.

This commit modifies mxt_initialize() to return error if Device is not in appmode and bootloader address is not specified.

This commit also returns error code from mxt_send_bootloader_cmd() in mxt_initialize().

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 58 +++++++++++++++++-------
 1 file changed, 41 insertions(+), 17 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +41 -17
index 0f7ee987a9c5..258dcd8a68e5 100644
@@ -716,17 +716,13 @@ static int mxt_lookup_bootloader_address(struct mxt_data *data, bool retry)
 	return 0;
 }
 
-static int mxt_probe_bootloader(struct mxt_data *data, bool alt_address)
+static int mxt_probe_bootloader(struct mxt_data *data)
 {
 	struct device *dev = &data->client->dev;
 	int error;
 	u8 buf[3];
 	bool crc_failure, extended_id;
 
-	error = mxt_lookup_bootloader_address(data, alt_address);
-	if (error)
-		return error;
-
 	/* Check bootloader status and version information */
 	error = mxt_bootloader_read(data, buf, sizeof(buf));
 	if (error)
@@ -2922,6 +2918,32 @@ static void mxt_config_cb(const struct firmware *cfg, void *ctx)
 	release_firmware(cfg);
 }
 
+static int mxt_bootloader_status(struct mxt_data *data)
+{
+	struct i2c_client *client = data->client;
+	int error;
+
+	error = mxt_lookup_bootloader_address(data, false);
+	if (error) {
+		dev_info(&client->dev,
+			 "Bootloader address is not specified\n");
+		return error;
+	}
+	/* Check bootloader state */
+	error = mxt_probe_bootloader(data);
+	if (error) {
+		dev_info(&client->dev, "Trying alternate bootloader address\n");
+		mxt_lookup_bootloader_address(data, true);
+		error = mxt_probe_bootloader(data);
+		if (error) {
+			dev_err(&client->dev,
+				"Chip is not in appmode or bootloader mode\n");
+			return error;
+		}
+	}
+	return 0;
+}
+
 static int mxt_initialize(struct mxt_data *data)
 {
 	struct i2c_client *client = data->client;
@@ -2933,16 +2955,13 @@ static int mxt_initialize(struct mxt_data *data)
 		if (!error)
 			break;
 
-		/* Check bootloader state */
-		error = mxt_probe_bootloader(data, false);
-		if (error) {
-			dev_info(&client->dev, "Trying alternate bootloader address\n");
-			error = mxt_probe_bootloader(data, true);
-			if (error) {
-				/* Chip is not in appmode or bootloader mode */
-				return error;
-			}
-		}
+		dev_info(&client->dev,
+			 "info block read failed (%d), so try bootloader method\n",
+			 error);
+
+		error = mxt_bootloader_status(data);
+		if (error)
+			return error;
 
 		/* OK, we are in bootloader, see if we can recover */
 		if (++recovery_attempts > 1) {
@@ -2956,7 +2975,9 @@ static int mxt_initialize(struct mxt_data *data)
 		}
 
 		/* Attempt to exit bootloader into app mode */
-		mxt_send_bootloader_cmd(data, false);
+		error = mxt_send_bootloader_cmd(data, false);
+		if (error)
+			return error;
 		msleep(MXT_FW_RESET_TIME);
 	}
 
@@ -3648,8 +3669,11 @@ static int mxt_enter_bootloader(struct mxt_data *data)
 
 		msleep(MXT_RESET_TIME);
 
+		ret = mxt_lookup_bootloader_address(data, false);
+		if (ret)
+			return ret;
 		/* Do not need to scan since we know family ID */
-		ret = mxt_probe_bootloader(data, 0);
+		ret = mxt_probe_bootloader(data);
 		if (ret)
 			return ret;
 
Signature · 1 line
2.17.1
author
#62 lore raw

PATCH v10 47/55 Input: Atmel: improve error handling in mxt_update_cfg()

In reply to Jiada Wang

From: Deepak Das <[email protected]>

mxt_update_cfg() failed to propagate the error code from mxt_init_t7_power_cfg() so return the error code.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)
drivers/input/touchscreen/atmel_mxt_ts.c +3 -1
index 258dcd8a68e5..069a532bd0b2 100644
@@ -2185,7 +2185,9 @@ static int mxt_update_cfg(struct mxt_data *data, const struct firmware *fw)
 	dev_info(dev, "Config successfully updated\n");
 
 	/* T7 config may have changed */
-	mxt_init_t7_power_cfg(data);
+	ret = mxt_init_t7_power_cfg(data);
+	if (ret)
+		dev_warn(dev, "Power Config failed to update\n");
 
 release_mem:
 	kfree(cfg.mem);
Signature · 1 line
2.17.1
author
#63 lore raw

PATCH v10 48/55 Input: Atmel: Improve error handling in mxt_initialize_input_device()

In reply to Jiada Wang

From: Deepak Das <[email protected]>

Currently Driver probe continues with a warning message when it fails to get the proper multitouch object configurations like TouchScreen resolution.

But Driver probe should fail in case of above scneario because it will not behave as expected without the proper touchscreen configurations.

This commit modifies mxt_initialize_input_device() to return error when it fails to get the proper touch screen configurations.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +8 -4
index 069a532bd0b2..ed357e958f07 100644
@@ -2761,15 +2761,19 @@ static int mxt_initialize_input_device(struct mxt_data *data)
 	case MXT_TOUCH_MULTI_T9:
 		num_mt_slots = data->T9_reportid_max - data->T9_reportid_min + 1;
 		error = mxt_read_t9_resolution(data);
-		if (error)
-			dev_warn(dev, "Failed to initialize T9 resolution\n");
+		if (error) {
+			dev_err(dev, "Failed to initialize T9 resolution\n");
+			return error;
+		}
 		break;
 
 	case MXT_TOUCH_MULTITOUCHSCREEN_T100:
 		num_mt_slots = data->num_touchids;
 		error = mxt_read_t100_config(data);
-		if (error)
-			dev_warn(dev, "Failed to read T100 config\n");
+		if (error) {
+			dev_err(dev, "Failed to read T100 config\n");
+			return error;
+		}
 		break;
 
 	default:
Signature · 1 line
2.17.1
author
#64 lore raw

PATCH v10 49/55 Input: Atmel: handle ReportID "0x00" while processing T5 messages

In reply to Jiada Wang

From: Deepak Das <[email protected]>

ReportID "0x00" is reserved by Atmel and should not be used by any Atmel touch controller.

reportID is the first byte retrieved from T5 message payload. Currently Atmel driver continues to process the T5 messages even if the reportID "0x00" is returned by Touch Controller.

This commit modifies Atmel touch driver to return -EINVAL if ReportID "0x00" is received while processing T5 messages.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 8 ++++++++
 1 file changed, 8 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +8 -0
index ed357e958f07..496b1f67d9a5 100644
@@ -76,6 +76,7 @@
 #define MXT_PROCI_TOUCHSEQUENCELOGGER	93
 #define MXT_TOUCH_MULTITOUCHSCREEN_T100 100
 #define MXT_PROCI_ACTIVESTYLUS_T107	107
+#define MXT_RPTID_RESERVED		0
 
 /* MXT_GEN_MESSAGE_T5 object */
 #define MXT_RPTID_NOMSG		0xff
@@ -1381,6 +1382,11 @@ static int mxt_proc_message(struct mxt_data *data, u8 *message)
 	u8 report_id = message[0];
 	bool dump = data->debug_enabled;
 
+	if (report_id == MXT_RPTID_RESERVED) {
+		dev_err(&data->client->dev,
+			"Received Reserved ReportID 0x00\n");
+		return -EINVAL;
+	}
 	if (report_id == MXT_RPTID_NOMSG)
 		return 0;
 
@@ -1451,6 +1457,8 @@ static int mxt_read_and_process_messages(struct mxt_data *data, u8 count)
 		ret = mxt_proc_message(data,
 			data->msg_buf + data->T5_msg_size * i);
 
+		if (ret < 0)
+			return ret;
 		if (ret == 1)
 			num_valid++;
 	}
Signature · 1 line
2.17.1
author
#65 lore raw

PATCH v10 50/55 Input: Atmel: use T44 object to process T5 messages

In reply to Jiada Wang

From: Deepak Das <[email protected]>

T44 object returns the count of valid T5 messages in the buffer. According to atmel, this count should be the main criteria to read the number of T5 messages.

Following is the statement from atmel confirming the same :- "For the readout of messages we recommend to stop after the last message is read out from the buffer. One way to identify the amount of new messages is to read T44. The other way is to monitor the /CHG line which indicates independent of mode 0 or mode 1 if there are still data in the buffer. 0xFF indicates that there is no message pending anymore, but it is not recommended to use this as the main criteria to control the data transfer."

This commit modifies the logic to readout the T5 messages on the basis of T44 object.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 55 +++++++++++++++---------
 1 file changed, 35 insertions(+), 20 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +35 -20
index 496b1f67d9a5..1cabe75df562 100644
@@ -1486,7 +1486,7 @@ static u8 mxt_max_msg_read_count(struct mxt_data *data, u8 max_T5_msg_count)
 	return min(T5_msg_count_limit, max_T5_msg_count);
 }
 
-static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
+static int mxt_process_messages_t44(struct mxt_data *data)
 {
 	struct device *dev = &data->client->dev;
 	int ret;
@@ -1499,7 +1499,7 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 		data->T5_msg_size + 1, data->msg_buf);
 	if (ret) {
 		dev_err(dev, "Failed to read T44 and T5 (%d)\n", ret);
-		return IRQ_NONE;
+		return ret;
 	}
 
 	T5_msg_count = data->msg_buf[0];
@@ -1509,7 +1509,7 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 	 * Mode 0. It results in unnecessary I2C operations but it is benign.
 	 */
 	if (!T5_msg_count)
-		return IRQ_NONE;
+		return processed_valid;
 
 	if (T5_msg_count > data->max_reportid) {
 		dev_warn(dev, "T44 count %d exceeded max report id\n",
@@ -1521,12 +1521,14 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 	ret = mxt_proc_message(data, data->msg_buf + 1);
 	if (ret < 0) {
 		dev_warn(dev, "Unexpected invalid message\n");
-		return IRQ_NONE;
+		return ret;
 	}
 
 	total_pending = T5_msg_count - 1;
-	if (!total_pending)
+	if (!total_pending) {
+		processed_valid = 1;
 		goto end;
+	}
 
 	/* Process remaining messages if necessary */
 	T5_msg_count = mxt_max_msg_read_count(data, total_pending);
@@ -1550,7 +1552,7 @@ static irqreturn_t mxt_process_messages_t44(struct mxt_data *data)
 		data->update_input = false;
 	}
 
-	return IRQ_HANDLED;
+	return processed_valid;
 }
 
 static int mxt_process_messages_until_invalid(struct mxt_data *data)
@@ -1580,7 +1582,7 @@ static int mxt_process_messages_until_invalid(struct mxt_data *data)
 	return -EBUSY;
 }
 
-static irqreturn_t mxt_process_messages(struct mxt_data *data)
+static int mxt_process_messages(struct mxt_data *data)
 {
 	int total_handled, num_handled;
 	u8 count = data->last_message_count;
@@ -1591,7 +1593,7 @@ static irqreturn_t mxt_process_messages(struct mxt_data *data)
 	/* include final invalid message */
 	total_handled = mxt_read_and_process_messages(data, count + 1);
 	if (total_handled < 0)
-		return IRQ_NONE;
+		return total_handled;
 	/* if there were invalid messages, then we are done */
 	else if (total_handled <= count)
 		goto update_count;
@@ -1600,7 +1602,7 @@ static irqreturn_t mxt_process_messages(struct mxt_data *data)
 	do {
 		num_handled = mxt_read_and_process_messages(data, 2);
 		if (num_handled < 0)
-			return IRQ_NONE;
+			return num_handled;
 
 		total_handled += num_handled;
 
@@ -1616,12 +1618,13 @@ static irqreturn_t mxt_process_messages(struct mxt_data *data)
 		data->update_input = false;
 	}
 
-	return IRQ_HANDLED;
+	return total_handled;
 }
 
 static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 {
 	struct mxt_data *data = dev_id;
+	int ret;
 
 	if (data->in_bootloader) {
 		complete(&data->chg_completion);
@@ -1629,17 +1632,22 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 		if (data->flash)
 			cancel_delayed_work_sync(&data->flash->work);
 
-		return IRQ_RETVAL(mxt_check_bootloader(data));
+		ret = mxt_check_bootloader(data);
+		return IRQ_RETVAL(ret);
 	}
 
 	if (!data->object_table)
 		return IRQ_HANDLED;
 
-	if (data->T44_address) {
-		return mxt_process_messages_t44(data);
-	} else {
-		return mxt_process_messages(data);
-	}
+	if (data->T44_address)
+		ret = mxt_process_messages_t44(data);
+	else
+		ret = mxt_process_messages(data);
+
+	if (ret <= 0)
+		return IRQ_NONE;
+	else
+		return IRQ_HANDLED;
 }
 
 static int mxt_t6_command(struct mxt_data *data, u16 cmd_offset,
@@ -1776,8 +1784,11 @@ static int mxt_acquire_irq(struct mxt_data *data)
 	}
 
 	if (data->object_table && data->use_retrigen_workaround) {
-		error = mxt_process_messages_until_invalid(data);
-		if (error)
+		if (data->T44_address)
+			error = mxt_process_messages_t44(data);
+		else
+			error = mxt_process_messages_until_invalid(data);
+		if (error < 0)
 			return error;
 	}
 
@@ -4039,8 +4050,12 @@ static int mxt_start(struct mxt_data *data)
 		 * Discard any touch messages still in message buffer
 		 * from before chip went to sleep
 		 */
-		ret = mxt_process_messages_until_invalid(data);
-		if (ret)
+
+		if (data->T44_address)
+			ret = mxt_process_messages_t44(data);
+		else
+			ret = mxt_process_messages_until_invalid(data);
+		if (ret < 0)
 			break;
 
 		ret = mxt_set_t7_power_cfg(data, MXT_POWER_CFG_RUN);
Signature · 1 line
2.17.1
author
#66 lore raw

PATCH v10 51/55 Input: atmel_mxt_ts: use gpiod_set_value_cansleep for reset pin

In reply to Jiada Wang

From: Balasubramani Vivekanandan <[email protected]>

In case of remote display, touch controller will be also remote. In such cases, the reset pin of the touch controller will be controlled through bridging ICs like Deserilizer and Serializer. Therefore accessing the gpio pins require transactions with the external IC. Using the function gpiod_set_value will print a warning like below

WARNING: CPU: 0 PID: 576 at drivers/gpio/gpiolib.c:1441 gpiod_set_value+0x34/0x60()
CPU: 0 PID: 576 Comm: modprobe Not tainted 3.14.79-08377-g84ea22f-dirty #4
Backtrace:
[<80011c58>] (dump_backtrace) from [<80011e60>] (show_stack+0x18/0x1c)
[<80011e48>] (show_stack) from [<8052d7ac>] (dump_stack+0x7c/0x9c)
[<8052d730>] (dump_stack) from [<800241bc>] (warn_slowpath_common+0x74/0x9c)
[<80024148>] (warn_slowpath_common) from [<80024288>] (warn_slowpath_null+0x24/0x2c)
[<80024264>] (warn_slowpath_null) from [<8029e070>] (gpiod_set_value+0x34/0x60)
[<8029e03c>] (gpiod_set_value) from [<7f492e98>] (mxt_probe+0x1e0/0x718 [atmel_mxt_ts])
[<7f492cb8>] (mxt_probe [atmel_mxt_ts]) from [<803c4d34>] (i2c_device_probe+0xcc/0xec)
[<803c4c68>] (i2c_device_probe) from [<803252a0>] (driver_probe_device+0xc0/0x200)
---
 drivers/input/touchscreen/atmel_mxt_ts.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +3 -3
index 1cabe75df562..14bd64d194b0 100644
@@ -2489,7 +2489,7 @@ static void mxt_regulator_enable(struct mxt_data *data)
 	if (!data->reg_vdd || !data->reg_avdd)
 		return;
 
-	gpiod_set_value(data->reset_gpio, 0);
+	gpiod_set_value_cansleep(data->reset_gpio, 0);
 
 	error = regulator_enable(data->reg_vdd);
 	if (error)
@@ -2507,7 +2507,7 @@ static void mxt_regulator_enable(struct mxt_data *data)
 	 * voltage
 	 */
 	msleep(MXT_REGULATOR_DELAY);
-	gpiod_set_value(data->reset_gpio, 1);
+	gpiod_set_value_cansleep(data->reset_gpio, 1);
 	msleep(MXT_CHG_DELAY);
 
 retry_wait:
@@ -4321,7 +4321,7 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		disable_irq(data->irq);
 	} else if (data->reset_gpio) {
 		msleep(MXT_RESET_GPIO_TIME);
-		gpiod_set_value(data->reset_gpio, 1);
+		gpiod_set_value_cansleep(data->reset_gpio, 1);
 		msleep(MXT_RESET_INVALID_CHG);
 	} else {
 		dev_dbg(&client->dev,
Signature · 1 line
2.17.1
author
#67 lore raw

PATCH v10 52/55 input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

In reply to Jiada Wang

From: Naveen Chakka <[email protected]>

To know the current communication status of the touch controller during runtime, sysfs interface is added

sysfs interface: /sys/class/i2c-dev/i2c-*/device/*/touch_dev_stat Executing the above sysfs interface provides two output values

1)Status of the touch device
value 0 represents device is inactive
value 1 represents device is active
2)Error counter
value represents the number of times device in inactive since last read

New module_param "debug_state" is introduced, by set its value, a watchdog work is scheduled to periodically check device state. default value is false.

[jiada: Rename mxt_touch_device_status to touch_dev_stat_show
Replace DEVICE_ATTR to DEVICE_ATTR_RO]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 90 ++++++++++++++++++++++--
 1 file changed, 83 insertions(+), 7 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +83 -7
index 14bd64d194b0..dda10bb51cb3 100644
@@ -222,6 +222,7 @@ enum t100_type {
 #define MXT_CHG_DELAY		100	/* msec */
 #define MXT_POWERON_DELAY	150	/* msec */
 #define MXT_BOOTLOADER_WAIT	36E5	/* 1 minute */
+#define MXT_WATCHDOG_TIMEOUT	1000	/* msec */
 
 /* Command to unlock bootloader */
 #define MXT_UNLOCK_CMD_MSB	0xaa
@@ -245,6 +246,9 @@ enum t100_type {
 
 #define DEBUG_MSG_MAX		200
 
+#define MXT_DEBUG_STATE		false
+static bool debug_state = MXT_DEBUG_STATE;
+
 struct mxt_info {
 	u8 family_id;
 	u8 variant_id;
@@ -317,6 +321,12 @@ struct mxt_flash {
 	struct delayed_work work;
 };
 
+struct mxt_statusinfo {
+	bool dev_status;
+	bool intp_triggered;
+	u32 error_count;
+};
+
 /* Each client has this additional data */
 struct mxt_data {
 	struct i2c_client *client;
@@ -372,6 +382,8 @@ struct mxt_data {
 	const char *pcfg_name;
 	const char *input_name;
 	struct mxt_flash *flash;
+	struct delayed_work watchdog_work;
+	struct mxt_statusinfo mxt_status;
 
 	/* Cached parameters from object table */
 	u16 T5_address;
@@ -1626,6 +1638,8 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 	struct mxt_data *data = dev_id;
 	int ret;
 
+	data->mxt_status.intp_triggered = true;
+
 	if (data->in_bootloader) {
 		complete(&data->chg_completion);
 
@@ -1633,21 +1647,25 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 			cancel_delayed_work_sync(&data->flash->work);
 
 		ret = mxt_check_bootloader(data);
-		return IRQ_RETVAL(ret);
+		ret = IRQ_RETVAL(ret);
+		goto exit;
 	}
 
-	if (!data->object_table)
-		return IRQ_HANDLED;
+	if (!data->object_table) {
+		ret = IRQ_HANDLED;
+		goto exit;
+	}
 
 	if (data->T44_address)
 		ret = mxt_process_messages_t44(data);
 	else
 		ret = mxt_process_messages(data);
 
-	if (ret <= 0)
-		return IRQ_NONE;
-	else
-		return IRQ_HANDLED;
+	ret = (ret <= 0) ? IRQ_NONE : IRQ_HANDLED;
+
+exit:
+	data->mxt_status.intp_triggered = false;
+	return ret;
 }
 
 static int mxt_t6_command(struct mxt_data *data, u16 cmd_offset,
@@ -2969,6 +2987,31 @@ static int mxt_bootloader_status(struct mxt_data *data)
 	return 0;
 }
 
+static void mxt_watchdog_work(struct work_struct *work)
+{
+	struct mxt_data *data =
+		container_of(work, struct mxt_data, watchdog_work.work);
+	u16 info_buf;
+	int ret;
+
+	if (data->suspended || data->in_bootloader ||
+	    data->mxt_status.intp_triggered)
+		goto sched_work;
+
+	ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);
+
+	if (ret) {
+		data->mxt_status.error_count++;
+		data->mxt_status.dev_status = false;
+	} else {
+		data->mxt_status.dev_status = true;
+	}
+
+sched_work:
+	schedule_delayed_work(&data->watchdog_work,
+			      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}
+
 static int mxt_initialize(struct mxt_data *data)
 {
 	struct i2c_client *client = data->client;
@@ -3956,6 +3999,22 @@ static const struct attribute_group mxt_fw_attr_group = {
 	.attrs = mxt_fw_attrs,
 };
 
+static ssize_t touch_dev_stat_show(struct device *dev, struct
+				   device_attribute * attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	int ret = 0;
+
+	if (data->mxt_status.dev_status)
+		data->mxt_status.error_count = 0;
+
+	ret =  snprintf(buf, PAGE_SIZE, "%d %d\n", data->mxt_status.dev_status,
+			data->mxt_status.error_count);
+	/* clear the error counter once it is read */
+	data->mxt_status.error_count = 0;
+	return ret;
+}
+
 static DEVICE_ATTR_RO(fw_version);
 static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
@@ -3965,6 +4024,7 @@ static DEVICE_ATTR_RW(debug_enable);
 static DEVICE_ATTR_RW(debug_v2_enable);
 static DEVICE_ATTR_RO(debug_notify);
 static DEVICE_ATTR_RW(t25_selftest);
+static DEVICE_ATTR_RO(touch_dev_stat);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
@@ -3976,6 +4036,7 @@ static struct attribute *mxt_attrs[] = {
 	&dev_attr_debug_v2_enable.attr,
 	&dev_attr_debug_notify.attr,
 	&dev_attr_t25_selftest.attr,
+	&dev_attr_touch_dev_stat.attr,
 	NULL
 };
 
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		msleep(MXT_RESET_TIME);
 	}
 
+	if (debug_state) {
+		INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+		schedule_delayed_work(&data->watchdog_work,
+				      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+	}
+
 	error = mxt_initialize(data);
 	if (error)
 		goto err_free_object;
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	return 0;
 
 err_free_object:
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 	if (data->reset_gpio) {
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);
+
 	return 0;
 }
 
@@ -4463,3 +4535,7 @@ module_i2c_driver(mxt_driver);
 MODULE_AUTHOR("Joonyoung Shim <[email protected]>");
 MODULE_DESCRIPTION("Atmel maXTouch Touchscreen driver");
 MODULE_LICENSE("GPL");
+
+module_param(debug_state, bool, 0);
+MODULE_PARM_DESC(debug_state, "Enable/Disable watchdog work to check device state (default="
+		 __MODULE_STRING(MXT_DEBUG_STATE) ")");
Signature · 1 line
2.17.1
#68 lore raw

Re: [PATCH v10 52/55] input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

31.03.2020 13:50, Jiada Wang пишет:
...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_bootloader_status(struct mxt_data *data) · line 2999

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +10 -0
+static void mxt_watchdog_work(struct work_struct *work)
+{
+	struct mxt_data *data =
+		container_of(work, struct mxt_data, watchdog_work.work);
+	u16 info_buf;
+	int ret;
+
+	if (data->suspended || data->in_bootloader ||
+	    data->mxt_status.intp_triggered)
+		goto sched_work;

Won't it become a problem if other thread puts device into suspended / bootloader state in the same time?

Quoted 13 lines
drivers/input/touchscreen/atmel_mxt_ts.c +13 -0
+	ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);
+
+	if (ret) {
+		data->mxt_status.error_count++;
+		data->mxt_status.dev_status = false;
+	} else {
+		data->mxt_status.dev_status = true;
+	}
+
+sched_work:
+	schedule_delayed_work(&data->watchdog_work,
+			      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}

...

Quoted 23 lines
drivers/input/touchscreen/atmel_mxt_ts.c +8 -0
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 		msleep(MXT_RESET_TIME);
 	}
 
+	if (debug_state) {
+		INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+		schedule_delayed_work(&data->watchdog_work,
+				      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+	}
+
 	error = mxt_initialize(data);
 	if (error)
 		goto err_free_object;
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	return 0;
 
 err_free_object:
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 	if (data->reset_gpio) {

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client) · line 4440

Quoted 5 lines
drivers/input/touchscreen/atmel_mxt_ts.c +2 -0
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);

What will happen if debug_state was false during of mxt_probe() and then the debug_state parameter was changed to true via sysfs?

I think the INIT_DELAYED_WORK() and cancel_delayed_work_sync() should be invoked unconditionally.

drivers/input/touchscreen/atmel_mxt_ts.c module_i2c_driver(mxt_driver); · line 4541

Quoted 12 lines
drivers/input/touchscreen/atmel_mxt_ts.c +4 -0
 	return 0;
 }
 
@@ -4463,3 +4535,7 @@ module_i2c_driver(mxt_driver);
 MODULE_AUTHOR("Joonyoung Shim <[email protected]>");
 MODULE_DESCRIPTION("Atmel maXTouch Touchscreen driver");
 MODULE_LICENSE("GPL");
+
+module_param(debug_state, bool, 0);
+MODULE_PARM_DESC(debug_state, "Enable/Disable watchdog work to check device state (default="
+		 __MODULE_STRING(MXT_DEBUG_STATE) ")");

The "default=..." part of MODULE_PARM_DESC() probably isn't really useful and could be omitted, don't you think so?

#69 lore raw

Re: [PATCH v10 52/55] input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

Hi Dmitry

Thanks for your comments

Quoted 16 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:
...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_bootloader_status(struct mxt_data *data) · line 2999

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +10 -0
+static void mxt_watchdog_work(struct work_struct *work)
+{
+	struct mxt_data *data =
+		container_of(work, struct mxt_data, watchdog_work.work);
+	u16 info_buf;
+	int ret;
+
+	if (data->suspended || data->in_bootloader ||
+	    data->mxt_status.intp_triggered)
+		goto sched_work;

Won't it become a problem if other thread puts device into suspended / bootloader state in the same time?

right, I will use mutex lock to prevent such case. also I think data->mxt_status.intp_triggered isn't necessary, when lock is used.

Quoted 41 lines
Quoted 13 lines
drivers/input/touchscreen/atmel_mxt_ts.c +13 -0
+	ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);
+
+	if (ret) {
+		data->mxt_status.error_count++;
+		data->mxt_status.dev_status = false;
+	} else {
+		data->mxt_status.dev_status = true;
+	}
+
+sched_work:
+	schedule_delayed_work(&data->watchdog_work,
+			      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}

...

Quoted 26 lines
drivers/input/touchscreen/atmel_mxt_ts.c +8 -0
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
  		msleep(MXT_RESET_TIME);
  	}
 
+	if (debug_state) {
+		INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+		schedule_delayed_work(&data->watchdog_work,
+				      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+	}
+
  	error = mxt_initialize(data);
  	if (error)
  		goto err_free_object;
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
  	return 0;
 
  err_free_object:
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);
  	mxt_free_input_device(data);
  	mxt_free_object_table(data);
  	if (data->reset_gpio) {
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
  	mxt_free_input_device(data);
  	mxt_free_object_table(data);
 
Quoted 5 lines

Review static int mxt_remove(struct i2c_client *client) · line 4440

Quoted 2 lines
(quoted diff) +2 -0
+	if (debug_state)
+		cancel_delayed_work_sync(&data->watchdog_work);

What will happen if debug_state was false during of mxt_probe() and then the debug_state parameter was changed to true via sysfs?

module_param debug_state is added with permission 0, so it's value won't change during driver operation

Quoted 7 lines

I think the INIT_DELAYED_WORK() and cancel_delayed_work_sync() should be invoked unconditionally.

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -0
  	return 0;
  }
 
Quoted 13 lines

drivers/input/touchscreen/atmel_mxt_ts.c module_i2c_driver(mxt_driver); · line 4541

Quoted 9 lines
drivers/input/touchscreen/atmel_mxt_ts.c +4 -0
@@ -4463,3 +4535,7 @@ module_i2c_driver(mxt_driver);
  MODULE_AUTHOR("Joonyoung Shim <[email protected]>");
  MODULE_DESCRIPTION("Atmel maXTouch Touchscreen driver");
  MODULE_LICENSE("GPL");
+
+module_param(debug_state, bool, 0);
+MODULE_PARM_DESC(debug_state, "Enable/Disable watchdog work to check device state (default="
+		 __MODULE_STRING(MXT_DEBUG_STATE) ")");

The "default=..." part of MODULE_PARM_DESC() probably isn't really useful and could be omitted, don't you think so?

since debug_state can't be updated via sysfs, so I think add it's default in MODULE_PARM_DESC() is useful? what's your opinion?

Thanks,
Jiada

#70 lore raw

Re: [PATCH v10 52/55] input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

01.04.2020 15:51, Wang, Jiada пишет:

Quoted 71 lines

Hi Dmitry

Thanks for your comments

Quoted 16 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:
...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_bootloader_status(struct mxt_data *data) · line 2999

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +10 -0
+static void mxt_watchdog_work(struct work_struct *work)
+{
+    struct mxt_data *data =
+        container_of(work, struct mxt_data, watchdog_work.work);
+    u16 info_buf;
+    int ret;
+
+    if (data->suspended || data->in_bootloader ||
+        data->mxt_status.intp_triggered)
+        goto sched_work;

Won't it become a problem if other thread puts device into suspended / bootloader state in the same time?

right, I will use mutex lock to prevent such case. also I think data->mxt_status.intp_triggered isn't necessary, when lock is used.

Quoted 37 lines
Quoted 13 lines

+    ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);

drivers/input/touchscreen/atmel_mxt_ts.c +12 -0
+
+    if (ret) {
+        data->mxt_status.error_count++;
+        data->mxt_status.dev_status = false;
+    } else {
+        data->mxt_status.dev_status = true;
+    }
+
+sched_work:
+    schedule_delayed_work(&data->watchdog_work,
+                  msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}

...

Quoted 22 lines
drivers/input/touchscreen/atmel_mxt_ts.c +5 -0
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client
 *client, const struct i2c_device_id *id)
           msleep(MXT_RESET_TIME);
       }
   +    if (debug_state) {
+        INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+        schedule_delayed_work(&data->watchdog_work,
+                      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+    }
+
       error = mxt_initialize(data);
       if (error)

          goto err_free_object;

(quoted diff) +0 -0
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client,
 const struct i2c_device_id *id)
       return 0;

    err_free_object:
+    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work); mxt_free_input_device(data); mxt_free_object_table(data);
      if (data->reset_gpio) {

Quoted 8 lines

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client) · line 4439

Quoted 5 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -0
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
       mxt_free_input_device(data);
       mxt_free_object_table(data);
   +    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work);

What will happen if debug_state was false during of mxt_probe() and then the debug_state parameter was changed to true via sysfs?

Quoted 2 lines

module_param debug_state is added with permission 0, so it's value won't change during driver operation

Thank you for the clarification, I didn't realize that setting permission to 0 hides the parameter completely in sysfs.

Quoted 19 lines
Quoted 19 lines

I think the INIT_DELAYED_WORK() and cancel_delayed_work_sync() should be invoked unconditionally.

Quoted 9 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -0
       return 0;
   }
   @@ -4463,3 +4535,7 @@ module_i2c_driver(mxt_driver);

  MODULE_AUTHOR("Joonyoung Shim <[email protected]>"); MODULE_DESCRIPTION("Atmel maXTouch Touchscreen driver"); MODULE_LICENSE("GPL");

drivers/input/touchscreen/atmel_mxt_ts.c +3 -0
+
+module_param(debug_state, bool, 0);
+MODULE_PARM_DESC(debug_state, "Enable/Disable watchdog work to check
Quoted 3 lines

device state (default="
+         __MODULE_STRING(MXT_DEBUG_STATE) ")");

The "default=..." part of MODULE_PARM_DESC() probably isn't really useful and could be omitted, don't you think so?

Quoted 2 lines

since debug_state can't be updated via sysfs, so I think add it's default in MODULE_PARM_DESC() is useful? what's your opinion?

If you want to keep the "default" kept in the description, then perhaps a simple "default=false" should be enough because nothing would be able to change it to true.

#71 lore raw

Re: [PATCH v10 52/55] input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

01.04.2020 17:33, Dmitry Osipenko пишет:

Quoted 25 lines

01.04.2020 15:51, Wang, Jiada пишет:

Quoted 24 lines

Hi Dmitry

Thanks for your comments

Quoted 16 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:
...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_bootloader_status(struct mxt_data *data) · line 2999

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +10 -0
+static void mxt_watchdog_work(struct work_struct *work)
+{
+    struct mxt_data *data =
+        container_of(work, struct mxt_data, watchdog_work.work);
+    u16 info_buf;
+    int ret;
+
+    if (data->suspended || data->in_bootloader ||
+        data->mxt_status.intp_triggered)
+        goto sched_work;

Won't it become a problem if other thread puts device into suspended / bootloader state in the same time?

right, I will use mutex lock to prevent such case. also I think data->mxt_status.intp_triggered isn't necessary, when lock is used.

Won't it be cleaner to stop/start the watchdog instead of messing with the locks?

Quoted 46 lines
Quoted 46 lines
Quoted 45 lines
Quoted 13 lines

+    ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);

drivers/input/touchscreen/atmel_mxt_ts.c +12 -0
+
+    if (ret) {
+        data->mxt_status.error_count++;
+        data->mxt_status.dev_status = false;
+    } else {
+        data->mxt_status.dev_status = true;
+    }
+
+sched_work:
+    schedule_delayed_work(&data->watchdog_work,
+                  msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}

...

Quoted 23 lines
drivers/input/touchscreen/atmel_mxt_ts.c +5 -0
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client
 *client, const struct i2c_device_id *id)
           msleep(MXT_RESET_TIME);
       }
   +    if (debug_state) {
+        INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+        schedule_delayed_work(&data->watchdog_work,
+                      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+    }
+
       error = mxt_initialize(data);
       if (error)

          goto err_free_object;

(quoted diff) +0 -0
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client,
 const struct i2c_device_id *id)
       return 0;

    err_free_object:
+    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work); mxt_free_input_device(data); mxt_free_object_table(data);
      if (data->reset_gpio) {

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client) · line 4439

Quoted 4 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -0
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
       mxt_free_input_device(data);
       mxt_free_object_table(data);
   +    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work);

What will happen if debug_state was false during of mxt_probe() and then the debug_state parameter was changed to true via sysfs?

Quoted 5 lines
Quoted 2 lines

module_param debug_state is added with permission 0, so it's value won't change during driver operation

Thank you for the clarification, I didn't realize that setting permission to 0 hides the parameter completely in sysfs.

Anyways, I'm still thinking that the condition removal will make code cleaner a tad.

#72 lore raw

Re: [PATCH v10 52/55] input: touchscreen: atmel_mxt_ts: Added sysfs entry for touchscreen status

Hi Dmitry

Quoted 23 lines · Dmitry Osipenko

01.04.2020 17:33, Dmitry Osipenko пишет:

Quoted 22 lines

01.04.2020 15:51, Wang, Jiada пишет:

Quoted 21 lines

Hi Dmitry

Thanks for your comments

Quoted 16 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:
...

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_bootloader_status(struct mxt_data *data) · line 2999

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +10 -0
+static void mxt_watchdog_work(struct work_struct *work)
+{
+    struct mxt_data *data =
+        container_of(work, struct mxt_data, watchdog_work.work);
+    u16 info_buf;
+    int ret;
+
+    if (data->suspended || data->in_bootloader ||
+        data->mxt_status.intp_triggered)
+        goto sched_work;

Won't it become a problem if other thread puts device into suspended / bootloader state in the same time?

Quoted 7 lines
Quoted 3 lines
Quoted 3 lines

right, I will use mutex lock to prevent such case. also I think data->mxt_status.intp_triggered isn't necessary, when lock is used.

Won't it be cleaner to stop/start the watchdog instead of messing with the locks?

will stop/start watchdog work when necessary in next version

Thanks,
Jiada

Quoted 55 lines
Quoted 46 lines
Quoted 46 lines
Quoted 45 lines
Quoted 13 lines

+    ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);

drivers/input/touchscreen/atmel_mxt_ts.c +12 -0
+
+    if (ret) {
+        data->mxt_status.error_count++;
+        data->mxt_status.dev_status = false;
+    } else {
+        data->mxt_status.dev_status = true;
+    }
+
+sched_work:
+    schedule_delayed_work(&data->watchdog_work,
+                  msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+}

...

Quoted 23 lines
drivers/input/touchscreen/atmel_mxt_ts.c +7 -0
@@ -4329,6 +4390,12 @@ static int mxt_probe(struct i2c_client
 *client, const struct i2c_device_id *id)
           msleep(MXT_RESET_TIME);
       }
   +    if (debug_state) {
+        INIT_DELAYED_WORK(&data->watchdog_work, mxt_watchdog_work);
+        schedule_delayed_work(&data->watchdog_work,
+                      msecs_to_jiffies(MXT_WATCHDOG_TIMEOUT));
+    }
+
       error = mxt_initialize(data);
       if (error)
           goto err_free_object;
@@ -4343,6 +4410,8 @@ static int mxt_probe(struct i2c_client *client,
 const struct i2c_device_id *id)
       return 0;
     err_free_object:
+    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work);
       mxt_free_input_device(data);
       mxt_free_object_table(data);
       if (data->reset_gpio) {

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client) · line 4439

Quoted 4 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -0
@@ -4367,6 +4436,9 @@ static int mxt_remove(struct i2c_client *client)
       mxt_free_input_device(data);
       mxt_free_object_table(data);
   +    if (debug_state)
+        cancel_delayed_work_sync(&data->watchdog_work);

What will happen if debug_state was false during of mxt_probe() and then the debug_state parameter was changed to true via sysfs?

Quoted 5 lines
Quoted 2 lines

module_param debug_state is added with permission 0, so it's value won't change during driver operation

Thank you for the clarification, I didn't realize that setting permission to 0 hides the parameter completely in sysfs.

Anyways, I'm still thinking that the condition removal will make code cleaner a tad.

author
#73 lore raw

PATCH v10 53/55 input: atmel_mxt_ts: added sysfs interface to update atmel T38 data

In reply to Jiada Wang

From: Naveen Chakka <[email protected]>

Atmel touch controller contains T38 object where a user can store its own data of length 64 bytes. T38 data will not be part of checksum calculation on executing T6 BACKUP command.

format used to update the T38 data is given below:

<offset> <length> <actual_data>

offset: offset address of the data to be written in the t38 object
(in decimal)

length: length of the data to be written into the t38 object(in decimal)

data: actual data bytes to be written into the t38 object
(values should be in hex)

Ex:
1. 0 2 10 20
updates first two bytes of the t38 data with values 10 and 20

2. 19 6 10 2f 30 4a 50 60
updates 6 bytes of t38 data from the index 19-24 with hex values

[jiada: Rename mxt_t38_data_show to t38_data_show
Rename mxt_t38_data_store to t38_data_store Replace DEVICE_ATTR with DEVICE_ATTR_RW]

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 102 +++++++++++++++++++++++
 1 file changed, 102 insertions(+)
drivers/input/touchscreen/atmel_mxt_ts.c +102 -0
index dda10bb51cb3..10d6726de9c4 100644
@@ -4015,6 +4015,106 @@ static ssize_t touch_dev_stat_show(struct device *dev, struct
 	return ret;
 }
 
+static ssize_t t38_data_show(struct device *dev,
+			     struct device_attribute *attr, char *buf)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	struct mxt_object *object;
+	size_t count = 0, size;
+	u8 i, *t38_buf;
+
+	if (!data->object_table)
+		return -ENXIO;
+
+	object = mxt_get_object(data, MXT_SPT_USERDATA_T38);
+	size = mxt_obj_size(object);
+
+	/* Pre-allocate buffer large enough to hold max size of t38 object.*/
+	t38_buf = kmalloc(size, GFP_KERNEL);
+	if (!t38_buf)
+		return -ENOMEM;
+
+	count = __mxt_read_reg(data->client, object->start_address,
+			       size, t38_buf);
+	if (count)
+		goto end;
+
+	for (i = 0; i < size; i++)
+		count += scnprintf(buf + count, PAGE_SIZE - count,
+				   "[%2u]: %02x\n", i, t38_buf[i]);
+	count += scnprintf(buf + count, PAGE_SIZE - count, "\n");
+end:
+	kfree(t38_buf);
+	return count;
+}
+
+static ssize_t t38_data_store(struct device *dev,
+			      struct device_attribute *attr,
+			      const char *buf, size_t count)
+{
+	struct mxt_data *data = dev_get_drvdata(dev);
+	struct mxt_object *object;
+	ssize_t ret = 0, pos, offset;
+	unsigned int i, len, index;
+	u8 *t38_buf;
+
+	if (!data->object_table)
+		return -ENXIO;
+
+	object = mxt_get_object(data, MXT_SPT_USERDATA_T38);
+
+	/* Pre-allocate buffer large enough to hold max size of t38 object.*/
+	t38_buf = kmalloc(mxt_obj_size(object), GFP_KERNEL);
+	if (!t38_buf)
+		return -ENOMEM;
+
+	ret = sscanf(buf, "%zd %d%zd", &offset, &len, &pos);
+	if (ret != 2) {
+		dev_err(dev, "Bad format: Invalid parameter to update t38\n");
+		ret = -EINVAL;
+		goto end;
+	}
+
+	if (len == 0) {
+		dev_err(dev,
+			"Bad format: Data length should not be equal to 0\n");
+		ret = -EINVAL;
+		goto end;
+	}
+
+	if (offset < 0 || ((offset + len) > 64)) {
+		dev_err(dev, "Invalid offset value to update t38\n");
+		ret = -EINVAL;
+		goto end;
+	}
+
+	index = pos;
+	for (i = 0; i < len; i++) {
+		ret = sscanf(buf + index, "%hhx%zd", t38_buf + i, &pos);
+		if (ret != 1) {
+			dev_err(dev, "Bad format: Invalid Data\n");
+			ret = -EINVAL;
+			goto end;
+		}
+		index += pos;
+	}
+
+	ret = __mxt_write_reg(data->client, object->start_address + offset,
+			      len, t38_buf);
+	if (ret)
+		goto end;
+
+	ret = mxt_t6_command(data, MXT_COMMAND_BACKUPNV, MXT_BACKUP_VALUE,
+			     true);
+	if (ret)
+		dev_err(dev, "backup command failed\n");
+	else
+		ret = count;
+end:
+	kfree(t38_buf);
+	return ret;
+}
+
 static DEVICE_ATTR_RO(fw_version);
 static DEVICE_ATTR_RO(hw_version);
 static DEVICE_ATTR(object, S_IRUGO, mxt_object_show, NULL);
@@ -4025,6 +4125,7 @@ static DEVICE_ATTR_RW(debug_v2_enable);
 static DEVICE_ATTR_RO(debug_notify);
 static DEVICE_ATTR_RW(t25_selftest);
 static DEVICE_ATTR_RO(touch_dev_stat);
+static DEVICE_ATTR_RW(t38_data);
 
 static struct attribute *mxt_attrs[] = {
 	&dev_attr_fw_version.attr,
@@ -4037,6 +4138,7 @@ static struct attribute *mxt_attrs[] = {
 	&dev_attr_debug_notify.attr,
 	&dev_attr_t25_selftest.attr,
 	&dev_attr_touch_dev_stat.attr,
+	&dev_attr_t38_data.attr,
 	NULL
 };
 
Signature · 1 line
2.17.1
author
#74 lore raw

PATCH v10 54/55 Input: atmel_mxt_ts: Implement synchronization during various operation

In reply to Jiada Wang

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

Thread Thread 2:

=========================                       =========================
mxt_update_fw_store()                           mxt_remove()
mutex_lock(&data->lock)                         ...
mxt_initialize()                                //Tries to acquire lock
  request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock()
...                                             .
...                                             .
mutex_unlock(&data->lock)                       .
                                                //Gets lock and proceeds
                                                mxt_free_input_device();
                                                ...
                                                mutex_unlock(&data->lock)
                                                //Frees atmel driver data
                                                kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

This commit adds state machine to serialize such scenarios.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 223 ++++++++++++++++++++---
 1 file changed, 198 insertions(+), 25 deletions(-)
drivers/input/touchscreen/atmel_mxt_ts.c +198 -25
index 10d6726de9c4..d244d97d134e 100644
@@ -223,6 +223,7 @@ enum t100_type {
 #define MXT_POWERON_DELAY	150	/* msec */
 #define MXT_BOOTLOADER_WAIT	36E5	/* 1 minute */
 #define MXT_WATCHDOG_TIMEOUT	1000	/* msec */
+#define MXT_CONFIG_TIMEOUT	36E5	/* msec */
 
 /* Command to unlock bootloader */
 #define MXT_UNLOCK_CMD_MSB	0xaa
@@ -249,6 +250,20 @@ enum t100_type {
 #define MXT_DEBUG_STATE		false
 static bool debug_state = MXT_DEBUG_STATE;
 
+enum device_state {
+	MXT_STATE_READY,
+	MXT_STATE_UPDATING_CONFIG,
+	MXT_STATE_UPDATING_CONFIG_ASYNC,
+	MXT_STATE_START,
+	MXT_STATE_STOP,
+	MXT_STATE_GOING_AWAY
+};
+
+enum mxt_cmd {
+	UPDATE_CFG,
+	UPDATE_FW
+};
+
 struct mxt_info {
 	u8 family_id;
 	u8 variant_id;
@@ -427,11 +442,15 @@ struct mxt_data {
 	/* Indicates whether device is in suspend */
 	bool suspended;
 
-	/* Indicates whether device is updating configuration */
-	bool updating_config;
+	struct mutex lock;
 
 	unsigned int mtu;
 	bool t25_status;
+
+	/* State handling for probe/remove, open/close and config update */
+	enum device_state e_state;
+
+	struct completion update_cfg_completion;
 };
 
 struct mxt_vb2_buffer {
@@ -1638,6 +1657,7 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 	struct mxt_data *data = dev_id;
 	int ret;
 
+	mutex_lock(&data->lock);
 	data->mxt_status.intp_triggered = true;
 
 	if (data->in_bootloader) {
@@ -1665,6 +1685,8 @@ static irqreturn_t mxt_interrupt(int irq, void *dev_id)
 
 exit:
 	data->mxt_status.intp_triggered = false;
+	mutex_unlock(&data->lock);
+
 	return ret;
 }
 
@@ -2247,6 +2269,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	video_unregister_device(&data->dbg.vdev);
 	v4l2_device_unregister(&data->dbg.v4l2);
 #endif
+	mutex_lock(&data->lock);
+
 	data->object_table = NULL;
 	kfree(data->info);
 	data->info = NULL;
@@ -2276,6 +2300,8 @@ static void mxt_free_object_table(struct mxt_data *data)
 	data->T100_reportid_min = 0;
 	data->T100_reportid_max = 0;
 	data->max_reportid = 0;
+
+	mutex_unlock(&data->lock);
 }
 
 static int mxt_parse_object_table(struct mxt_data *data,
@@ -2957,8 +2983,15 @@ static int mxt_configure_objects(struct mxt_data *data,
 
 static void mxt_config_cb(const struct firmware *cfg, void *ctx)
 {
+	struct mxt_data *data = ctx;
+
 	mxt_configure_objects(ctx, cfg);
 	release_firmware(cfg);
+	complete(&data->update_cfg_completion);
+	mutex_lock(&data->lock);
+	if (data->e_state != MXT_STATE_GOING_AWAY)
+		data->e_state = MXT_STATE_READY;
+	mutex_unlock(&data->lock);
 }
 
 static int mxt_bootloader_status(struct mxt_data *data)
@@ -2994,9 +3027,12 @@ static void mxt_watchdog_work(struct work_struct *work)
 	u16 info_buf;
 	int ret;
 
+	mutex_lock(&data->lock);
 	if (data->suspended || data->in_bootloader ||
-	    data->mxt_status.intp_triggered)
+	    data->mxt_status.intp_triggered) {
+		mutex_unlock(&data->lock);
 		goto sched_work;
+	}
 
 	ret = __mxt_read_reg(data->client, 0, sizeof(info_buf), &info_buf);
 
@@ -3066,6 +3102,15 @@ static int mxt_initialize(struct mxt_data *data)
 		goto err_free_sysfs;
 
 	if (data->cfg_name) {
+		mutex_lock(&data->lock);
+		if (data->e_state != MXT_STATE_GOING_AWAY) {
+			data->e_state = MXT_STATE_UPDATING_CONFIG_ASYNC;
+		} else {
+			mutex_unlock(&data->lock);
+			return -EBUSY;
+		}
+		reinit_completion(&data->update_cfg_completion);
+		mutex_unlock(&data->lock);
 		error = request_firmware_nowait(THIS_MODULE, true,
 						data->cfg_name,
 						&client->dev,
@@ -3845,30 +3890,58 @@ static int mxt_update_file_name(struct device *dev, char **file_name,
 	return 0;
 }
 
+static int mxt_process_operation(struct mxt_data *data,
+				 enum mxt_cmd cmd,
+				 void *cmd_data);
+
 static ssize_t update_fw_store(struct device *dev,
 			       struct device_attribute *attr,
 			       const char *buf, size_t count)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
+	char *filename = NULL;
+	int ret;
+
+	ret = mxt_update_file_name(dev, &filename, buf, count);
+	if (ret)
+		goto out;
+
+	ret = mxt_process_operation(data, UPDATE_FW, filename);
+	kfree(filename);
+
+	if (ret)
+		goto out;
+
+	return count;
+out:
+	return ret;
+}
+
+static int mxt_fw_update(struct mxt_data *data,
+			 const char *filename)
+{
+	struct device *dev = &data->client->dev;
+	unsigned int len = 0;
 	int error;
 
-	error = mxt_update_file_name(dev, &data->fw_name, buf, count);
+	len = strlen(filename);
+	error = mxt_update_file_name(dev, &data->fw_name, filename, len);
 	if (error)
 		return error;
 
 	error = mxt_load_fw(dev);
 	if (error) {
 		dev_err(dev, "The firmware update failed(%d)\n", error);
-		count = error;
-	} else {
-		dev_info(dev, "The firmware update succeeded\n");
-
-		error = mxt_initialize(data);
-		if (error)
-			return error;
+		return error;
 	}
 
-	return count;
+	error = mxt_initialize(data);
+	if (error)
+		return error;
+
+	dev_info(dev, "The firmware update succeeded\n");
+
+	return error;
 }
 
 static ssize_t update_cfg_store(struct device *dev,
@@ -3876,14 +3949,38 @@ static ssize_t update_cfg_store(struct device *dev,
 				const char *buf, size_t count)
 {
 	struct mxt_data *data = dev_get_drvdata(dev);
+	char *filename = NULL;
+	int ret;
+
+	ret = mxt_update_file_name(dev, &filename, buf, count);
+	if (ret)
+		goto out;
+
+	ret = mxt_process_operation(data, UPDATE_CFG, filename);
+	kfree(filename);
+
+	if (ret)
+		goto out;
+
+	return count;
+out:
+	return ret;
+}
+
+static int mxt_cfg_update(struct mxt_data *data,
+			  char *filename)
+{
+	struct device *dev = &data->client->dev;
 	const struct firmware *cfg;
+	unsigned int len = 0;
 	int ret;
 
-	ret = mxt_update_file_name(dev, &data->cfg_name, buf, count);
+	len = strlen(filename);
+	ret = mxt_update_file_name(dev, &data->cfg_name, filename, len);
 	if (ret)
 		return ret;
 
-	ret = request_firmware(&cfg, data->cfg_name, dev);
+	ret = request_firmware(&cfg, data->cfg_name, &data->client->dev);
 	if (ret < 0) {
 		dev_err(dev, "Failure to request config file %s\n",
 			data->cfg_name);
@@ -3891,8 +3988,6 @@ static ssize_t update_cfg_store(struct device *dev,
 		goto out;
 	}
 
-	data->updating_config = true;
-
 	mxt_free_input_device(data);
 
 	if (data->suspended) {
@@ -3908,15 +4003,8 @@ static ssize_t update_cfg_store(struct device *dev,
 	}
 
 	ret = mxt_configure_objects(data, cfg);
-	if (ret)
-		goto release;
-
-	ret = count;
-
-release:
 	release_firmware(cfg);
 out:
-	data->updating_config = false;
 	return ret;
 }
 
@@ -4188,8 +4276,17 @@ static int mxt_start(struct mxt_data *data)
 {
 	int ret = 0;
 
-	if (!data->suspended || data->in_bootloader)
+	mutex_lock(&data->lock);
+	if (!data->suspended) {
+		mutex_unlock(&data->lock);
 		return 0;
+	}
+	if (data->in_bootloader || data->e_state != MXT_STATE_READY) {
+		mutex_unlock(&data->lock);
+		return -EBUSY;
+	}
+	data->e_state = MXT_STATE_START;
+	mutex_unlock(&data->lock);
 
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
@@ -4233,8 +4330,12 @@ static int mxt_start(struct mxt_data *data)
 		ret = mxt_acquire_irq(data);
 	}
 
+	mutex_lock(&data->lock);
 	if (!ret)
 		data->suspended = false;
+	if (data->e_state != MXT_STATE_GOING_AWAY)
+		data->e_state = MXT_STATE_READY;
+	mutex_unlock(&data->lock);
 
 	return ret;
 }
@@ -4243,8 +4344,19 @@ static int mxt_stop(struct mxt_data *data)
 {
 	int ret;
 
-	if (data->suspended || data->in_bootloader || data->updating_config)
+	mutex_lock(&data->lock);
+	if (data->suspended || data->e_state == MXT_STATE_UPDATING_CONFIG) {
+		mutex_unlock(&data->lock);
 		return 0;
+	}
+	if ((data->e_state != MXT_STATE_READY &&
+	     data->e_state != MXT_STATE_GOING_AWAY)) {
+		mutex_unlock(&data->lock);
+		return -EBUSY;
+	}
+	if (data->e_state != MXT_STATE_GOING_AWAY)
+		data->e_state = MXT_STATE_STOP;
+	mutex_unlock(&data->lock);
 
 	switch (data->suspend_mode) {
 	case MXT_SUSPEND_T9_CTRL:
@@ -4274,8 +4386,15 @@ static int mxt_stop(struct mxt_data *data)
 		break;
 	}
 
+	mutex_lock(&data->lock);
 	data->suspended = true;
+
+	if (data->e_state != MXT_STATE_GOING_AWAY)
+		data->e_state = MXT_STATE_READY;
+	mutex_unlock(&data->lock);
+
 	return 0;
+
 }
 
 static int mxt_input_open(struct input_dev *dev)
@@ -4430,12 +4549,15 @@ static int mxt_probe(struct i2c_client *client, const struct i2c_device_id *id)
 	snprintf(data->phys, sizeof(data->phys), "i2c-%u-%04x/input0",
 		 client->adapter->nr, client->addr);
 
+	mutex_init(&data->lock);
+
 	data->client = client;
 	i2c_set_clientdata(client, data);
 
 	init_completion(&data->chg_completion);
 	init_completion(&data->reset_completion);
 	init_completion(&data->crc_completion);
+	init_completion(&data->update_cfg_completion);
 	mutex_init(&data->debug_msg_lock);
 
 	data->suspend_mode = dmi_check_system(chromebook_T9_suspend_dmi) ?
@@ -4527,6 +4649,18 @@ static int mxt_remove(struct i2c_client *client)
 {
 	struct mxt_data *data = i2c_get_clientdata(client);
 
+	mutex_lock(&data->lock);
+	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
+	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
+		data->e_state = MXT_STATE_GOING_AWAY;
+		mutex_unlock(&data->lock);
+		mxt_wait_for_completion(data, &data->update_cfg_completion,
+					MXT_CONFIG_TIMEOUT);
+	} else {
+		data->e_state = MXT_STATE_GOING_AWAY;
+		mutex_unlock(&data->lock);
+	}
+
 	disable_irq(data->irq);
 	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (data->reset_gpio) {
@@ -4587,6 +4721,45 @@ static int __maybe_unused mxt_resume(struct device *dev)
 	return ret;
 }
 
+static int mxt_process_operation(struct mxt_data *data,
+				 enum mxt_cmd cmd,
+				 void *cmd_data)
+{
+	int ret = 0;
+
+	mutex_lock(&data->lock);
+	if (data->e_state != MXT_STATE_READY) {
+		mutex_unlock(&data->lock);
+		dev_err(&data->client->dev, "Atmel touch device is shutting down\n");
+		return -EBUSY;
+	}
+	data->e_state = MXT_STATE_UPDATING_CONFIG;
+	reinit_completion(&data->update_cfg_completion);
+	mutex_unlock(&data->lock);
+
+	switch (cmd) {
+	case UPDATE_CFG:
+	case UPDATE_FW:
+		if (cmd == UPDATE_CFG)
+			ret = mxt_cfg_update(data, (char *)cmd_data);
+		else
+			ret = mxt_fw_update(data, (char *)cmd_data);
+		break;
+
+	default:
+		break;
+	}
+	mutex_lock(&data->lock);
+	if (data->e_state != MXT_STATE_UPDATING_CONFIG_ASYNC) {
+		complete(&data->update_cfg_completion);
+		if (data->e_state != MXT_STATE_GOING_AWAY)
+			data->e_state = MXT_STATE_READY;
+	}
+	mutex_unlock(&data->lock);
+
+	return ret;
+}
+
 static SIMPLE_DEV_PM_OPS(mxt_pm_ops, mxt_suspend, mxt_resume);
 
 static const struct of_device_id mxt_of_match[] = {
Signature · 1 line
2.17.1
#75 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

31.03.2020 13:50, Jiada Wang пишет:

Quoted 26 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

Thread Thread 2:

=========================                       =========================
mxt_update_fw_store()                           mxt_remove()
mutex_lock(&data->lock)                         ...
mxt_initialize()                                //Tries to acquire lock
  request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock()
...                                             .
...                                             .
mutex_unlock(&data->lock)                       .
                                                //Gets lock and proceeds
                                                mxt_free_input_device();
                                                ...
                                                mutex_unlock(&data->lock)
                                                //Frees atmel driver data
                                                kfree(data)
Quoted 5 lines

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()?

#76 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

Hi Dmitry

Quoted 27 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 26 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

Thread Thread 2:

=========================                       =========================
mxt_update_fw_store()                           mxt_remove()
mutex_lock(&data->lock)                         ...
mxt_initialize()                                //Tries to acquire lock
   request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock()
...                                             .
...                                             .
mutex_unlock(&data->lock)                       .
                                                 //Gets lock and proceeds
                                                 mxt_free_input_device();
                                                 ...
                                                 mutex_unlock(&data->lock)
                                                 //Frees atmel driver data
                                                 kfree(data)
Quoted 8 lines
Quoted 5 lines

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

Thanks,
Jiada

#77 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock request_firmware_nowait()                     mutex_lock(&data->lock) ...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

(quoted diff) +0 -0

b/drivers/input/touchscreen/atmel_mxt_ts.c
index b2edf51e1595..4e66106feeb9 100644

drivers/input/touchscreen/atmel_mxt_ts.c +3 -14
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
 	struct i2c_client *client = data->client;
 
 	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 }
 
 static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
 {
 	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (data->reset_gpio) {
 		sysfs_remove_link(&client->dev.kobj, "reset");
 		gpiod_unexport(data->reset_gpio);
 	}
+
 	mxt_debug_msg_remove(data);
-	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
	if (debug_state)
		cancel_delayed_work_sync(&data->watchdog_work);

+ disable_irq(data->irq);

return 0;
}

#78 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

02.04.2020 16:24, Dmitry Osipenko пишет:

Quoted 97 lines

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock request_firmware_nowait()                     mutex_lock(&data->lock) ...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +3 -13
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
 	struct i2c_client *client = data->client;
 
 	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 }
 
 static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
 {
 	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (data->reset_gpio) {
 		sysfs_remove_link(&client->dev.kobj, "reset");
 		gpiod_unexport(data->reset_gpio);
 	}
+
 	mxt_debug_msg_remove(data);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
-	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
 	return 0;
 }

I'm also wondering why dev_attr_update_fw needs a separate attribute_group, couldn't it be moved into mxt_attrs[]?

#79 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

Hi Dmitrij

Quoted 105 lines · Dmitry Osipenko

02.04.2020 16:24, Dmitry Osipenko пишет:

Quoted 104 lines

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock
   request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and
proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver
data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +4 -14
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
  	struct i2c_client *client = data->client;
 
  	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  }
 
  static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
  {
  	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  	if (data->reset_gpio) {
  		sysfs_remove_link(&client->dev.kobj, "reset");
  		gpiod_unexport(data->reset_gpio);
  	}
+
  	mxt_debug_msg_remove(data);
-	mxt_sysfs_remove(data);
  	mxt_free_input_device(data);
  	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
Quoted 6 lines

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -0
  	return 0;
  }

I'm also wondering why dev_attr_update_fw needs a separate attribute_group, couldn't it be moved into mxt_attrs[]

Separate sysfs into different groups are done by commit "Input: atmel_mxt_ts - rework sysfs init/remove"
I think the main purpose,
is to remove other sysfs entries when firmware is being updated.

Thanks,
Jiada

#80 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

In reply to Dmitry Osipenko

02.04.2020 16:24, Dmitry Osipenko пишет:

Quoted 97 lines

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock request_firmware_nowait()                     mutex_lock(&data->lock) ...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +3 -13
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
 	struct i2c_client *client = data->client;
 
 	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 }
 
 static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
 {
 	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (data->reset_gpio) {
 		sysfs_remove_link(&client->dev.kobj, "reset");
 		gpiod_unexport(data->reset_gpio);
 	}
+
 	mxt_debug_msg_remove(data);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
-	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
 	return 0;
 }

I'm looking at this again and the original tear-down order of the mxt_remove() looks okay, so no need to change it.

Reading the commit message, it says that request_firmware_nowait() races with kfree(data), but that can't happen because the data is resource-managed and request_firmware_nowait() bumps device's use-count.

https://elixir.bootlin.com/linux/v5.6.2/source/drivers/base/firmware_loader/main.c#L1043

I think this patch was ported from some very old kernel version and it's simply not applicable to the upstream anymore.

#81 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

02.04.2020 17:00, Dmitry Osipenko пишет:

Quoted 112 lines

02.04.2020 16:24, Dmitry Osipenko пишет:

Quoted 97 lines

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock request_firmware_nowait()                     mutex_lock(&data->lock) ...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +3 -13
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
 	struct i2c_client *client = data->client;
 
 	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 }
 
 static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
 {
 	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
 	if (data->reset_gpio) {
 		sysfs_remove_link(&client->dev.kobj, "reset");
 		gpiod_unexport(data->reset_gpio);
 	}
+
 	mxt_debug_msg_remove(data);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
-	mxt_sysfs_remove(data);
 	mxt_free_input_device(data);
 	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
 	return 0;
 }

I'm looking at this again and the original tear-down order of the mxt_remove() looks okay, so no need to change it.

Quoted 9 lines

Reading the commit message, it says that request_firmware_nowait() races with kfree(data), but that can't happen because the data is resource-managed and request_firmware_nowait() bumps device's use-count.

https://elixir.bootlin.com/linux/v5.6.2/source/drivers/base/firmware_loader/main.c#L1043

I think this patch was ported from some very old kernel version and it's simply not applicable to the upstream anymore.

Moreover, request_firmware_nowait() uses try_module_get(), hence driver's module can't be unloaded until firmware is loaded. This patch is wrong, please drop it.

#82 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

In reply to Dmitry Osipenko

Hi Dmitry

Quoted 112 lines · Dmitry Osipenko

02.04.2020 16:24, Dmitry Osipenko пишет:

Quoted 97 lines

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock
   request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and
proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver
data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +3 -13
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
  	struct i2c_client *client = data->client;
 
  	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  }
 
  static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
  {
  	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  	if (data->reset_gpio) {
  		sysfs_remove_link(&client->dev.kobj, "reset");
  		gpiod_unexport(data->reset_gpio);
  	}
+
  	mxt_debug_msg_remove(data);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
-	mxt_sysfs_remove(data);
  	mxt_free_input_device(data);
  	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
  	return 0;
  }

I'm looking at this again and the original tear-down order of the mxt_remove() looks okay, so no need to change it.

Quoted 9 lines

Reading the commit message, it says that request_firmware_nowait() races with kfree(data), but that can't happen because the data is resource-managed and request_firmware_nowait() bumps device's use-count.

https://elixir.bootlin.com/linux/v5.6.2/source/drivers/base/firmware_loader/main.c#L1043

I think this patch was ported from some very old kernel version and it's simply not applicable to the upstream anymore.

I had some test,
and confirmed you are right,
this commit is no longer applicable to upstream,
but as discussed in another patch,
disable_irq() need to be moved after remove of mxt_fw_attr_group. I will add this change in a new commit.

Thanks,
Jiada

#83 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

06.04.2020 07:18, Wang, Jiada пишет:
...

Quoted 6 lines

I had some test,
and confirmed you are right,
this commit is no longer applicable to upstream,
but as discussed in another patch,
disable_irq() need to be moved after remove of mxt_fw_attr_group. I will add this change in a new commit.

Sounds good.

#84 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

In reply to Dmitry Osipenko

Hi Dmitry

Quoted 48 lines · Dmitry Osipenko

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 39 lines

Hi Dmitry

Quoted 36 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 35 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

     Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock
   request_firmware_nowait()                     mutex_lock(&data->lock)
...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                 //Gets lock and
proceeds

mxt_free_input_device();
                                                 ...

mutex_unlock(&data->lock)
                                                 //Frees atmel driver
data
                                                 kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

Quoted 6 lines

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

Yes, you are correct,
this commit isn't addressing the real root cause

Quoted 43 lines

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +3 -13
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
  	struct i2c_client *client = data->client;
 
  	sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  }
 
  static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
  {
  	struct mxt_data *data = i2c_get_clientdata(client);
 
-	mutex_lock(&data->lock);
-	if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-	    data->e_state == MXT_STATE_UPDATING_CONFIG) {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-		mxt_wait_for_completion(data, &data->update_cfg_completion,
-					MXT_CONFIG_TIMEOUT);
-	} else {
-		data->e_state = MXT_STATE_GOING_AWAY;
-		mutex_unlock(&data->lock);
-	}
+	mxt_sysfs_remove(data);
 
-	disable_irq(data->irq);
-	sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
  	if (data->reset_gpio) {
  		sysfs_remove_link(&client->dev.kobj, "reset");
  		gpiod_unexport(data->reset_gpio);
  	}
+
  	mxt_debug_msg_remove(data);

drivers/input/touchscreen/atmel_mxt_ts.c static int mxt_remove(struct i2c_client *client)

Quoted 10 lines
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
-	mxt_sysfs_remove(data);
  	mxt_free_input_device(data);
  	mxt_free_object_table(data);
 
 	if (debug_state)
 		cancel_delayed_work_sync(&data->watchdog_work);
+	disable_irq(data->irq);
 
  	return 0;
  }

yes, I confirmed, the root cause is because irq is disabled while firmware is being updated, this cause update of firmware can't proceed. by move disable irq after sysfs entry removal can fix the issue

Thanks,
Jiada

#85 lore raw

Re: [PATCH v10 54/55] Input: atmel_mxt_ts: Implement synchronization during various operation

03.04.2020 11:39, Wang, Jiada пишет:

Quoted 113 lines

Hi Dmitry

Quoted 50 lines · Dmitry Osipenko

02.04.2020 14:50, Wang, Jiada пишет:

Quoted 41 lines

Hi Dmitry

Quoted 38 lines · Dmitry Osipenko

31.03.2020 13:50, Jiada Wang пишет:

Quoted 37 lines

From: Sanjeev Chugh <[email protected]>

There could be scope of race conditions when sysfs is being handled and at the same time, device removal is occurring. For example, we don't want the device removal to begin if the Atmel device cfg update is going on or firmware update is going on. In such cases, wait for device update to be completed before the removal continues.

      Thread                                          Thread 2:
=========================
=========================
mxt_update_fw_store()                           mxt_remove() mutex_lock(&data->lock)                         ... mxt_initialize()                                //Tries to acquire lock
    request_firmware_nowait()
mutex_lock(&data->lock)
...                                             ==>waits for lock() ...                                             .
...                                             . mutex_unlock(&data->lock)                       .
                                                  //Gets lock and proceeds

mxt_free_input_device();
                                                  ...

mutex_unlock(&data->lock)
                                                  //Frees atmel driver data
                                                  kfree(data)

If the request_firmware_nowait() completes after the driver removal, and callback is triggered. But kernel crashes since the module is already removed.

Quoted 7 lines
Quoted 4 lines
Quoted 1 line

This commit adds state machine to serialize such scenarios.

Won't it be easier to bump driver's module use-count by __module_get() while firmware is updating? Or remove sysfs during of mxt_remove()? >

thanks for your inspiration, I will replace state machine with module use-count.

Quoted 6 lines

I'm actually now thinking that the suggestion about the module-count wasn't very correct because this won't really help in regards to mxt_update_fw_store() / mxt_remove() racing.

I see that mxt_remove() already invokes the mxt_sysfs_remove(), which should block until mxt_update_fw_store() is completed, shouldn't it?

Yes, you are correct,
this commit isn't addressing the real root cause

Quoted 50 lines

I guess the kfree(data) isn't the real cause of the problem and something like this should help:

drivers/input/touchscreen/atmel_mxt_ts.c +4 -14
index b2edf51e1595..4e66106feeb9 100644
@@ -4254,6 +4254,7 @@ static void mxt_sysfs_remove(struct mxt_data *data)
       struct i2c_client *client = data->client;
 
       sysfs_remove_group(&client->dev.kobj, &mxt_attr_group);
+    sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
   }
 
   static void mxt_reset_slots(struct mxt_data *data)
@@ -4649,31 +4650,19 @@ static int mxt_remove(struct i2c_client *client)
   {
       struct mxt_data *data = i2c_get_clientdata(client);
 
-    mutex_lock(&data->lock);
-    if (data->e_state == MXT_STATE_UPDATING_CONFIG_ASYNC ||
-        data->e_state == MXT_STATE_UPDATING_CONFIG) {
-        data->e_state = MXT_STATE_GOING_AWAY;
-        mutex_unlock(&data->lock);
-        mxt_wait_for_completion(data, &data->update_cfg_completion,
-                    MXT_CONFIG_TIMEOUT);
-    } else {
-        data->e_state = MXT_STATE_GOING_AWAY;
-        mutex_unlock(&data->lock);
-    }
+    mxt_sysfs_remove(data);
 
-    disable_irq(data->irq);
-    sysfs_remove_group(&client->dev.kobj, &mxt_fw_attr_group);
       if (data->reset_gpio) {
           sysfs_remove_link(&client->dev.kobj, "reset");
           gpiod_unexport(data->reset_gpio);
       }
+
       mxt_debug_msg_remove(data);
-    mxt_sysfs_remove(data);
       mxt_free_input_device(data);
       mxt_free_object_table(data);
 
     if (debug_state)
         cancel_delayed_work_sync(&data->watchdog_work);
+    disable_irq(data->irq);
 
Quoted 6 lines
Quoted 3 lines
drivers/input/touchscreen/atmel_mxt_ts.c +0 -0
       return 0;

  }

yes, I confirmed, the root cause is because irq is disabled while firmware is being updated, this cause update of firmware can't proceed. by move disable irq after sysfs entry removal can fix the issue

Okay :)

author
#86 lore raw

PATCH v10 55/55 Input: atmel_mxt_ts - Fix compilation warning

In reply to Jiada Wang

fix "make W=1" compilation warnings from Atmel driver as per the compilation logs.

---
 drivers/input/touchscreen/atmel_mxt_ts.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/input/touchscreen/atmel_mxt_ts.c +1 -1
index d244d97d134e..b2edf51e1595 100644
@@ -2032,7 +2032,7 @@ static int mxt_prepare_cfg_mem(struct mxt_data *data, struct mxt_cfg *cfg)
 
 			byte_offset = reg + i - cfg->start_ofs;
 
-			if (byte_offset >= 0 && byte_offset < cfg->mem_size) {
+			if (byte_offset < cfg->mem_size) {
 				*(cfg->mem + byte_offset) = val;
 			} else {
 				dev_err(dev, "Bad object: reg:%d, T%d, ofs=%d\n",
Signature · 1 line
2.17.1