[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [PATCH] af9035: add remote control support

On 18.04.2012 23:42, Hans-Frieder Vogt wrote:
Am Mittwoch, 18. April 2012 schrieb Mauro Carvalho Chehab:
Em 18-04-2012 11:57, Antti Palosaari escreveu:
I haven't tried to and not commented it. But I see clearly few problems.

On 18.04.2012 17:17, Mauro Carvalho Chehab wrote:
Em 07-04-2012 14:24, Hans-Frieder Vogt escreveu:
af9035: support remote controls. Currently, for remotes using the NEC
protocol, the map of the TERRATEC_CINERGY_XS remote is loaded, for RC6
the map of RC_MAP_RC6_MCE.

Signed-off-by: Hans-Frieder Vogt<hfvogt@xxxxxxx>

   drivers/media/dvb/dvb-usb/af9035.c |   72
   drivers/media/dvb/dvb-usb/af9035.h |    3 +
   2 files changed, 75 insertions(+)

diff -Nupr a/drivers/media/dvb/dvb-usb/af9035.c
b/drivers/media/dvb/dvb-usb/af9035.c ---
a/drivers/media/dvb/dvb-usb/af9035.c    2012-04-07 15:59:56.000000000
+0200 +++ b/drivers/media/dvb/dvb-usb/af9035.c    2012-04-07
19:17:55.044874329 +0200 @@ -313,6 +313,41 @@ static struct
i2c_algorithm af9035_i2c_a

       .functionality = af9035_i2c_functionality,


+#define AF9035_POLL 250
+static int af9035_rc_query(struct dvb_usb_device *d)
+    unsigned int key;
+    unsigned char b[4];
+    int ret;
+    struct usb_req req = { CMD_IR_GET, 0, 0, NULL, 4, b };
+    if (!af9035_config.raw_ir)
+        return 0;
+    ret = af9035_ctrl_msg(d->udev,&req);
+    if (ret<   0)
+        goto err;
+    if ((b[2] + b[3]) == 0xff) {
+        if ((b[0] + b[1]) == 0xff) {
+            /* NEC */
+            key = b[0]<<   8 | b[2];
+        } else {
+            /* ext. NEC */
+            key = b[0]<<   16 | b[1]<<   8 | b[2];
+        }
+    } else {
+        key = b[0]<<   24 | b[1]<<   16 | b[2]<<   8 | b[3];
+    }
+    if (d->rc_dev != NULL)
+        rc_keydown(d->rc_dev, key, 0);

Is that checking needed and why? If there is no rc_device why we even
call poll for it? Better to fix some core routines if that is true.

Also rc_keydown() takes 2nd param as int, but in that case it does not
matter. Anyhow, 3rd param is toggle which is used by RC5/6. IIRC I have
never implemented RC5 or RC6 remote receiver, so I am not sure if it is
needed and in which case.

It is better to implement the toggle, when it is available/known, as the
core will use it to detect when the same key was pressed quickly twice, or
if someone just kept it pressed by a long time.

When this is not implemented and someone presses the same key quickly twice
(a "double click"), the second click will be ignored, if the time is lower
than REP_DELAY (by default, 500 ms).

The IR_GET command only delivers 4 bytes, which give no indication of a
repeated key.

Then it is good idea to add comment why it is not used, as normally RC5/RC6 uses it.

As a side note, I implemented AF9015 remote reading IR codes directly from the RAM and leaving firmware command unused. That was mainly due to HID problems (starts repeating wildly due to chip sets wrong USB polling interval) and also gain full control to remote. Similar solution could be nice for that driver too.

Not all protocols/decoders can detect it though. NEC protocol can't.
RC-5/RC-6 can do it. Yet, not all hardware reports the toggle big on RC-5.

+    /* ignore errors */
+    return 0;

   static int af9035_init(struct dvb_usb_device *d)

       int ret, i;

@@ -627,6 +662,34 @@ static int af9035_read_mac_address(struc

       for (i = 0; i<   af9035_properties[0].num_adapters; i++)

           af9035_af9033_config[i].clock = clock_lut[tmp];

+    ret = af9035_rd_reg(d, EEPROM_IR_MODE,&tmp);
+    if (ret<   0)
+        goto err;
+    pr_debug("%s: ir_mode=%02x\n", __func__, tmp);
+    af9035_config.raw_ir = tmp == 5;

This looks odd for my eyes. Maybe x = (y == z); is better. Checkpatch
didn't complain it?

I think checkpatch will accept that. I generally prefer to use:

	foo = (tmp == 5) = true : false;

shouldn't it rather be
         foo = (tmp == 5) ? true : false;

as some source code analyzers complain about statements like the above.

Maybe whole boolean conversion can be dropped. I suspect there is more possible values than than 5 and !5.

+    if (af9035_config.raw_ir) {
+        ret = af9035_rd_reg(d, EEPROM_IR_TYPE,&tmp);

No space between x,y, IIRC checkpatch reports that.

the only errors that checkpatch is reporting is ERROR: trailing whitespace,
but that seems to be normal for lines in the patch that are unchanged (I run
checkpatch.pl --no-tree --file ...patch).

I still like idea to add one space after comma always

+        if (ret<   0)
+            goto err;
+        pr_debug("%s: ir_type=%02x\n", __func__, tmp);
+        switch (tmp) {
+        case 0: /* NEC */
+        default:
+            af9035_config.ir_rc6 = false;

unused variable


+            d->props.rc.core.protocol = RC_TYPE_NEC;
+            d->props.rc.core.rc_codes =
+            break;
+        case 1: /* RC6 */
+            af9035_config.ir_rc6 = true;
+            d->props.rc.core.protocol = RC_TYPE_RC6;
+            d->props.rc.core.rc_codes = RC_MAP_RC6_MCE;
+            break;
+        }

I hate to default some random remote controller keytable. Use EMPTY map,
it is for that.

Good idea.

+    }

       return 0;

@@ -1003,6 +1066,15 @@ static struct dvb_usb_device_properties

           .i2c_algo =&af9035_i2c_algo,

+        .rc.core = {
+            .protocol       = RC_TYPE_NEC,
+            .module_name    = "af9035",
+            .rc_query       = af9035_rc_query,
+            .rc_interval    = AF9035_POLL,
+            .allowed_protos = RC_TYPE_NEC | RC_TYPE_RC6,

Does this mean we promise userspace we can do both NEC and RC6? Does it
mean we should offer method to change protocol in that case? I suspect
it is not even possible to switch from remote protocol to other unless
eeprom change or firmware hack.

Yes, that assumes a callback to allow to switch the protocol, OR that the
device can automatically recognize both protocols (there are a few that
are able to handle both NEC and RC-5 or RC-6 without any specific command).
The RC and NEC timings are very different, so, auto-detecting it is quite

If this is the case for af9035, all that it is needed test the protocol
auto-detection is to replace the table from one protocol to the other and
use an IR compatible with the new table.

I think the af9035 doesn't autodetect the protocol. I tested a device which is
configured for RC6 (as layed down in the eeprom) and it doesn't read any raw
code from a NEC rc.

In the way this code was written, it leaves the reviewer without any af9035
device to believe that auto-detection is supported by af9035 (and also
because there's no command sent to the device in order to switch the mode).

It is easy to check if the device accepts both automatically: just load
a different table with ir-keycode and test the remote with a different

If this is not the case, then rc.core.allowed_protocols should be equal to


+            .rc_codes       = RC_MAP_EMPTY, /* may be changed in
+                           af9035_read_mac_address */

Commented that earlier. But RC_MAP_EMPTY is correct choice for default.

This is just a minor thing, but the comment here seems to be outdated,
as this is actually set at af9035_init().

+        },

           .num_device_descs = 5,
           .devices = {


diff -Nupr a/drivers/media/dvb/dvb-usb/af9035.h
b/drivers/media/dvb/dvb-usb/af9035.h ---
a/drivers/media/dvb/dvb-usb/af9035.h    2012-04-07 15:58:43.000000000
+0200 +++ b/drivers/media/dvb/dvb-usb/af9035.h    2012-04-07
17:35:08.517840044 +0200 @@ -49,6 +49,8 @@ struct usb_req {

   struct config {

       bool dual_mode;

+    bool raw_ir;
+    bool ir_rc6;

Both of these new configs are unused and not needed. Please do not add
new configuration option unless needed (to pass config data from
function to other inside driver).

raw_ir is indeed used (see af9035_rc_query). However I agree that I could
implement this switch in a different way without the need of an extra config

Hmmm, OK, I may be then wrong, it is a little bit hard to review patches using mailer.

But of course it is wise to some configuration values if needed continuously. Very good reason is for example avoid I/O (register read / write) when polling remote.

But for the remote, most info is already stored to device properties structure and existing info should used instead of duplicating same info to multiple locations.

       bool hw_not_supported;


@@ -96,6 +98,7 @@ u32 clock_lut_it9135[] = {

   #define CMD_MEM_WR                  0x01
   #define CMD_I2C_RD                  0x02
   #define CMD_I2C_WR                  0x03

+#define CMD_IR_GET                  0x18

   #define CMD_FW_DL                   0x21
   #define CMD_FW_QUERYINFO            0x22
   #define CMD_FW_BOOT                 0x23

Hans-Frieder Vogt                       e-mail: hfvogt<at>   gmx .dot.

Except for that minor mistake at the comment above, the rest looks fine
on my eyes.

I added some comments. And there was some basic remote controller issues
- I didn't checked those, but those were commented as what is my
understanding and some may be even wrong. In all cases please fix
properly or explain I was wrong.


Thanks, Mauro and Antti, for your comments. Expect an improved patch soon.


Hans-Frieder Vogt                       e-mail: hfvogt<at>  gmx .dot. net


To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[Linux Input]     [Video for Linux]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Photos]     [Yosemite Photos]     [Linux Kernel]     [Linux SCSI]     [XFree86]     [Devices]     [Yosemite Backpacking]

Add to Google Powered by Linux