[HELP] writting to sensors sysfs - Galaxy S 4 Developer Discussion [Developers-Only]

I want to change the poll_delay sysfs file from the gesture_sensor, I can read it fine but when I try to write to it nothing happens, I have already read the dmesg and found nothing, the source code for this driver is here:
https://github.com/gokulnatha/GT-I9505/blob/master/drivers/sensorhub/ssp.h#L98
https://github.com/gokulnatha/GT-I9505/blob/master/drivers/sensorhub/ssp_sysfs.c#L45
I am not so experienced in C so maybe someone can help me to understand what happens when I try to write this value
the file is located at "/sys/class/input/input7/poll_delay" on GT-I9505

---

frapeti said:
I want to change the poll_delay sysfs file from the gesture_sensor, I can read it fine but when I try to write to it nothing happens, I have already read the dmesg and found nothing, the source code for this driver is here:
https://github.com/gokulnatha/GT-I9505/blob/master/drivers/sensorhub/ssp.h#L98
https://github.com/gokulnatha/GT-I9505/blob/master/drivers/sensorhub/ssp_sysfs.c#L45
I am not so experienced in C so maybe someone can help me to understand what happens when I try to write this value
the file is located at "/sys/class/input/input7/poll_delay" on GT-I9505
Click to expand...
Click to collapse
I think you need to use an init.d script to write to the file.

---

is it declared in ueventd.rc in ramdisk?
like
/sys/class/input/input7/poll_delay 0664 system system

---

broodplank1337 said:
is it declared in ueventd.rc in ramdisk?
like
/sys/class/input/input7/poll_delay 0664 system system
Click to expand...
Click to collapse
Yep it's there
{
"lightbox_close": "Close",
"lightbox_next": "Next",
"lightbox_previous": "Previous",
"lightbox_error": "The requested content cannot be loaded. Please try again later.",
"lightbox_start_slideshow": "Start slideshow",
"lightbox_stop_slideshow": "Stop slideshow",
"lightbox_full_screen": "Full screen",
"lightbox_thumbnails": "Thumbnails",
"lightbox_download": "Download",
"lightbox_share": "Share",
"lightbox_zoom": "Zoom",
"lightbox_new_window": "New window",
"lightbox_toggle_sidebar": "Toggle sidebar"
}
I'm on a root shell and permissions are right, I guess the problem is in the format I write to the file or the driver code itself :/
EDIT: is that a typo on Samsung file? lol

---

frapeti said:
Yep it's there
View attachment 2675684
I'm on a root shell and permissions are right, I guess the problem is in the format I write to the file or the driver code itself :/
EDIT: is that a typo on Samsung file? lol
Click to expand...
Click to collapse
lol idd a typo, try fixing it and retry

---

doesn't it just set permissions and user/group? it's doble setted, look some lines above that typo, I'm sure I have permissions to write to the file, but I don't know the format I should write to it, Samsung services can do it fine btw

---

frapeti said:
doesn't it just set permissions and user/group? it's doble setted, look some lines above that typo, I'm sure I have permissions to write to the file, but I don't know the format I should write to it, Samsung services can do it fine btw
Click to expand...
Click to collapse
yes thats what it does, the owners that are now given are root and input. you need to add system, and maybe even put that 664 to 666. then apps should have access to it

---

broodplank1337 said:
yes thats what it does, the owners that are now given are root and input. you need to add system, and maybe even put that 664 to 666. then apps should have access to it
Click to expand...
Click to collapse
I'm trying to write it on a root shell already, where I can even chown and chmod that file, only samsung sensors service knows how to interact with its sysfs. And looking at the source code maybe some one can figure it out too
Enviado desde mi GT-I9505 mediante Tapatalk

---

frapeti said:
I'm trying to write it on a root shell already, where I can even chown and chmod that file, only samsung sensors service knows how to interact with its sysfs. And looking at the source code maybe some one can figure it out too
Enviado desde mi GT-I9505 mediante Tapatalk
Click to expand...
Click to collapse
I just tried it,
echo '200000000' > /sys/class/input/input7/poll_delay
it said
/system/bin/sh: can't create poll_delay: Read-only file system
then i did 'mount -o remount rw /' and it worked just fine.
so its just echoing the value to it, can't be easier

---

broodplank1337 said:
yes thats what it does, the owners that are now given are root and input. you need to add system, and maybe even put that 664 to 666. then apps should have access to it
Click to expand...
Click to collapse
if you use "su" in a terminal, it has access to everything. The issue isnt a permissions issue. Usually when i change /sys/class files, i have to do it in an init.d script so the kernel writes to it after others have written to it to prevent my values from being overriden. But init.d does not change it either. i really think the kernel source file holds this value.
broodplank1337 said:
I just tried it,
echo '200000000' > /sys/class/input/input7/poll_delay
it said
/system/bin/sh: can't create poll_delay: Read-only file system
then i did 'mount -o remount rw /' and it worked just fine.
so its just echoing the value to it, can't be easier
Click to expand...
Click to collapse
Try echoing a value other than whats already set, it won't hold. I think either the kernel is controlling this, and "poll_delay" is only there so we can view the speed, or that value is set and can't be changed due to hardware limitation. I don't think its possible to change on the fly so to speak. But maybe by editing the kernel source and compiling it yourself would change it.

---

Yeah that's what I mean, I want to set poll delay to 10000000 ns and the only who can do that is samsung sensors service because it knows how to write to that file
Enviado desde mi GT-I9505 mediante Tapatalk

---

elesbb said:
if you use "su" in a terminal, it has access to everything. The issue isnt a permissions issue. Usually when i change /sys/class files, i have to do it in an init.d script so the kernel writes to it after others have written to it to prevent my values from being overriden. But init.d does not change it either. i really think the kernel source file holds this value.
Try echoing a value other than whats already set, it won't hold. I think either the kernel is controlling this, and "poll_delay" is only there so we can view the speed, or that value is set and can't be changed due to hardware limitation. I don't think its possible to change on the fly so to speak. But maybe by editing the kernel source and compiling it yourself would change it.
Click to expand...
Click to collapse
I know this, just wanted to say that it is possible to write to that file, some files are not writable at all or readable, so thats a gain. and yes sysfs gets overwrited by kernel by default
And idd you're right I didn't notice it didn't hold. sorry
frapeti said:
Yeah that's what I mean, I want to set poll delay to 10000000 ns and the only who can do that is samsung sensors service because it knows how to write to that file
Enviado desde mi GT-I9505 mediante Tapatalk
Click to expand...
Click to collapse
What you can try is making a c program that regulates it, my experiences with c apps is that they do not require much permissions in order to work correctly

---

broodplank1337 said:
I know this, just wanted to say that it is possible to write to that file, some files are not writable at all or readable, so thats a gain. and yes sysfs gets overwrited by kernel by default
And idd you're right I didn't notice it didn't hold. sorry
What you can try is making a c program that regulates it, my experiences with c apps is that they do not require much permissions in order to work correctly
Click to expand...
Click to collapse
You can write apps in c code instead of Java and they will run on android!?
Sent from my SGH-M919 using Tapatalk

---

I think I don't need an app, I just need to understand the driver's source code to find out how to write to its sysfs
Enviado desde mi GT-I9505 mediante Tapatalk

---

elesbb said:
You can write apps in c code instead of Java and they will run on android!?
Sent from my SGH-M919 using Tapatalk
Click to expand...
Click to collapse
Well not as an apk, but as a binary, like everything in system/bin and xbin. like this: https://github.com/broodplank/android_external_s4utils/blob/master/s4utils.c please note that that is my first c script ever made without any tutorial, so I know, its really poor! anyways, since sensor service has permission to write to it, why shouldn't a custom binary also have that with the right permissions set.
frapeti said:
I think I don't need an app, I just need to understand the driver's source code to find out how to write to its sysfs
Enviado desde mi GT-I9505 mediante Tapatalk
Click to expand...
Click to collapse
ok sorry I wont spam your topic anymore, I just dont know how to read the usage from that code. :silly:

---

broodplank1337 said:
Well not as an apk, but as a binary, like everything in system/bin and xbin. like this: https://github.com/broodplank/android_external_s4utils/blob/master/s4utils.c please note that that is my first c script ever made without any tutorial, so I know, its really poor! anyways, since sensor service has permission to write to it, why shouldn't a custom binary also have that with the right permissions set.
ok sorry I wont spam your topic anymore, I just dont know how to read the usage from that code. :silly:
Click to expand...
Click to collapse
Thanks for helping btw, I'm sure it isn't related to permissions but format of the data written to that sysfs file, I'm going to find out by reverse engineering samsung service or reading driver's source code again many times lol -.-
Enviado desde mi GT-I9505 mediante Tapatalk

---

frapeti said:
Thanks for helping btw, I'm sure it isn't related to permissions but format of the data written to that sysfs file, I'm going to find out by reverse engineering samsung service or reading driver's source code again many times lol -.-
Enviado desde mi GT-I9505 mediante Tapatalk
Click to expand...
Click to collapse
Well, i finally had the time to look through those source codes thoroughly. And those values are set by the kernel itself. I don't even know if changing them via on the kernel level will work as limited by hardware. But those values in this file are applied in java when registering a new sensor listener. You can read the standard google example here which tells you what the meaning of those set values are used for.

---

elesbb said:
Well, i finally had the time to look through those source codes thoroughly. And those values are set by the kernel itself. I don't even know if changing them via on the kernel level will work as limited by hardware. But those values in this file are applied in java when registering a new sensor listener. You can read the standard google example here which tells you what the meaning of those set values are used for.
Click to expand...
Click to collapse
Are you sure? Look at set_gesture_delay and change_sensor_delay functions
Enviado desde mi GT-I9505 mediante Tapatalk

---

frapeti said:
Are you sure? Look at set_gesture_delay and change_sensor_delay functions
Enviado desde mi GT-I9505 mediante Tapatalk
Click to expand...
Click to collapse
Yes and from what I'm making sense of it, the kernel is calling those methods, they aren't being set by an app so to say. The kernel itself is doing it. That's what I'm getting from it anyhow.
Sent from my SGH-M919 using Tapatalk

---

look
Code:
static struct device_attribute dev_attr_gesture_poll_delay
= __ATTR([COLOR="Red"]poll_delay[/COLOR], S_IRUGO | S_IWUSR | S_IWGRP,
show_gesture_delay, [COLOR="Red"]set_gesture_delay[/COLOR]);
>>
Code:
static ssize_t set_gesture_delay(struct device *dev,
struct device_attribute *attr, const char *buf, size_t size)
{
int64_t [COLOR="blue"][B]dNewDelay[/B][/COLOR];
struct ssp_data *data = dev_get_drvdata(dev);
if (kstrtoll(buf, 10, &[COLOR="Blue"][B]dNewDelay[/B][/COLOR]) < 0)
return -1;
[COLOR="Red"]change_sensor_delay[/COLOR](data, GESTURE_SENSOR, [COLOR="Blue"][B]dNewDelay[/B][/COLOR]);
return size;
}
>>
Code:
static void change_sensor_delay(struct ssp_data *data,
int iSensorType, int64_t [COLOR="blue"][B]dNewDelay[/B][/COLOR])
{
u8 uBuf[2];
unsigned int uNewEnable = 0;
int64_t dTempDelay = data->adDelayBuf[iSensorType];
if (!(atomic_read(&data->aSensorEnable) & (1 << iSensorType))) {
data->aiCheckStatus[iSensorType] = NO_SENSOR_STATE;
return;
}
data->adDelayBuf[iSensorType] = [COLOR="blue"][B]dNewDelay[/B][/COLOR];
switch (data->aiCheckStatus[iSensorType]) {
case ADD_SENSOR_STATE:
ssp_dbg("[SSP]: %s - add %u, New = %lldns\n",
__func__, 1 << iSensorType, [COLOR="blue"][B]dNewDelay[/B][/COLOR]);
uBuf[1] = (u8)get_msdelay([COLOR="Blue"][B]dNewDelay[/B][/COLOR]);
uBuf[0] = (u8)get_delay_cmd(uBuf[1]);
if (send_instruction(data, ADD_SENSOR, iSensorType, uBuf, 2)
!= SUCCESS) {
uNewEnable =
(unsigned int)atomic_read(&data->aSensorEnable)
& (~(unsigned int)(1 << iSensorType));
atomic_set(&data->aSensorEnable, uNewEnable);
data->aiCheckStatus[iSensorType] = NO_SENSOR_STATE;
data->uMissSensorCnt++;
break;
}
data->aiCheckStatus[iSensorType] = RUNNING_SENSOR_STATE;
if (iSensorType == PROXIMITY_SENSOR) {
proximity_open_lcd_ldi(data);
proximity_open_calibration(data);
input_report_abs(data->prox_input_dev, ABS_DISTANCE, 1);
input_sync(data->prox_input_dev);
}
break;
case RUNNING_SENSOR_STATE:
if (get_msdelay(dTempDelay)
== get_msdelay(data->adDelayBuf[iSensorType]))
break;
ssp_dbg("[SSP]: %s - Change %u, New = %lldns\n",
__func__, 1 << iSensorType, [COLOR="blue"][B]dNewDelay[/B][/COLOR]);
uBuf[1] = (u8)get_msdelay([COLOR="blue"][B]dNewDelay[/B][/COLOR]);
uBuf[0] = (u8)get_delay_cmd(uBuf[1]);
send_instruction(data, CHANGE_DELAY, iSensorType, uBuf, 2);
break;
default:
data->aiCheckStatus[iSensorType] = ADD_SENSOR_STATE;
}
}