gc1024 is not detected

[moleus]

ipctool -s returns a message Error: unexpected value for SuperPix == 0x78.

Manual detection

I've followed suggestions of OpenIPC's telegram chat members and ran:

# ./ipctool i2cdetect
       0  1  2  3  4  5  6  7   8  9  a  b  c  d  e  f
     : xx xx xx xx xx xx xx xx  xx xx xx xx xx xx xx xx  |
                      ....
   70: xx xx xx xx xx xx xx xx  78 79 xx xx xx xx xx xx  |
                      ....
  f0: xx xx xx xx xx xx xx xx  xx xx xx xx xx xx xx

The sensor id can be found at f0 - 10 24. (Unnecessary lines are removed, file with the original output - i2cdump_0x78.txt)

# ./ipctool i2cdump 0x78 0 0x3ff
       0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F
    : 00 1F FF 03 98 01 6F 00  A6 00 00 00 08 02 E6 05  |  ......o.........
  10: 08 00 18 11 01 00 C0 17  0A 06 11 4F 11 10 F8 38  |  ...........O...8
...
  e0: 00 AB B6 80 9D EC F1 C7  DA 00 00 00 00 00 1D 40  |  ...............@
  f0: 10 24 0F 00 03 00 00 B9  03 0E 00 78 C4 11 00 00  |  .$.........x....
 100: FF FF FF FF FF FF FF FF  FF FF FF FF FF FF FF FF  |  ...............
...
 1f0: FF FF FF FF FF FF FF FF  FF FF FF FF FF FF FF FF  |  ................
 200: 03 03 03 03 03 03 03 03  03 03 03 03 03 03 03 03  |  ................
 ...
 2d0: 03 03 03 03 03 03 03 03  03 03 03 03 03 03 03 03  |  ................
...
 3f0: 98 98 98 98 98 98 98 98  98 98 98 98 98 98 98     |  ...............

The problem

According to the detection function source code, 3f0 or 3f1 should be FF to read the value from f0 address.

ipctool/src/sensors.c

Lines 664 to 675 in 11f8946

	static int detect_galaxycore_sensor(sensor_ctx_t *ctx, int fd,
	unsigned char i2c_addr) {
	if (i2c_change_addr(fd, i2c_addr) < 0)
	return false;
	int prod_msb = i2c_read_register(fd, i2c_addr, 0x3f0, 2, 1);
	int prod_lsb = i2c_read_register(fd, i2c_addr, 0x3f1, 2, 1);
	if (prod_msb == -1 || prod_lsb == -1) {
	prod_msb = i2c_read_register(fd, i2c_addr, 0xf0, 1, 1);
	prod_lsb = i2c_read_register(fd, i2c_addr, 0xf1, 1, 1);
	}

But it's not the case, because 3f0 and 3f1 are filled with a garbage. You can see the output of multiple executions below:

# ./ipctool i2cdump 0x78 0x3f0 0x3ff
       0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F
 3f0: CE CE CE CE CE CE CE CE  CE CE CE CE CE CE CE
# ./ipctool i2cdump 0x78 0x3f0 0x3ff
       0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F
 3f0: 50 50 50 50 50 50 50 50  50 50 50 50 50 50 50
# ./ipctool i2cdump 0x78 0x3f0 0x3ff
       0  1  2  3  4  5  6  7   8  9  A  B  C  D  E  F
 3f0: 59 59 59 59 59 59 59 59  59 59 59 59 59 59 59

Side-effects

Also, garbage in fa and in fb addresses prevents the early break in SuperPix detection function

ipctool/src/sensors.c

Lines 746 to 749 in 11f8946

	prod_msb = i2c_read_register(fd, i2c_addr, 0xfa, 1, 1);
	// early break
	if (prod_msb == -1)
	return false;

and execution falls into the default case

ipctool/src/sensors.c

Lines 768 to 769 in 11f8946

	default:
	SENSOR_ERR("SuperPix", res);

, so we get an error message Error: unexpected value for SuperPix == 0x78, but it's not a SuperPix sensor

Solutions?

I'm knew to this topic and don't have a fundamental knowledge in sensors/i2c stuff, so I can't say if it's OK to have a "garbage" in a memory and how to deal with it. But can we check the value of f0 address before 3f0?

[dimerr]

What's your soc?

reboot and try ipctool again

[moleus]

hi3518ev200

# cat cpuinfo
Processor       : ARM926EJ-S rev 5 (v5l)
BogoMIPS        : 269.10
Features        : swp half thumb fastmult edsp java
CPU implementer : 0x41
CPU architecture: 5TEJ
CPU variant     : 0x0
CPU part        : 0x926
CPU revision    : 5

Hardware        : hi3518ev200
Revision        : 0000
Serial          : 0000000000000000

The output of ipctool - ipctool_output.txt (after reboot I got the same result)

I've looked at a table with sensor identifiers provided by Paul, and saw that GC1024 has chip_id 0x1004. That's strange.

I'm tinkering with VstarCam C7824WIP and found a note about it's GC1024 sensor - code. Does this mean that it's the GC1024 with invalid id?

[dimerr]

try again

[moleus]

Got the ipctool with the patch. Still doesn't work. Same error.

I think that this condition fails, because 0x78 not listed in possible_i2c_addrs

ipctool/src/sensors.c

Line 791 in 34ef8b2

if (sdata->sensor_type == type) {

As far as I understand list of possible addresses for hisilicon is declared here:

ipctool/src/hal/hisi/hal_hisi.c

Lines 28 to 35 in 34ef8b2

	static unsigned char gc_addrs[] = {0x6e, 0x52, 0x42, 0};
	static unsigned char superpix_addrs[] = {0x79, 0};
	static sensor_addr_t my_possible_i2c_addrs[] = {
	{SENSOR_SONY, sony_addrs}, {SENSOR_SOI, soi_addrs},
	{SENSOR_ONSEMI, onsemi_addrs}, {SENSOR_SMARTSENS, ssens_addrs},
	{SENSOR_OMNIVISION, omni_addrs}, {SENSOR_GALAXYCORE, gc_addrs},
	{SENSOR_SUPERPIX, superpix_addrs}, {0, NULL}};

What if we add 0x78 to this array?

[dimerr]

it's already added

[moleus]

I see, in this 34ef8b2 commit you've only added the case statement, and might forgot to push the changes with the address. Am I wrong?

[dimerr]

Yeah, really weird, hal_hisi is missing

[moleus]

It works!

sensors:
- vendor: GalaxyCore
  model: GC1024
  control:
    bus: 0
    type: i2c
    addr: 0x78
  data:
    type: DC
  clock: 24MHz

Thank you!