I cold booted my number cruncher after a hiatus of a couple of weeks, applied a couple of extra dev_dbg printouts in the si2168_cmd_execute and installed the newly built module. The results: [ 663.147757] si2168 2-0066: Silicon Labs Si2168 successfully attached [ 663.151735] si2157 1-0060: Silicon Labs Si2147/2148/2157/2158 successfully attached [ 663.152436] DVB: registering new adapter (saa7164) [ 663.152441] saa7164 0000:07:00.0: DVB: registering adapter 1 frontend 0 (Silicon Labs Si2168)... [ 678.690104] si2168:si2168_init: si2168 2-0064: [ 678.690111] si2168:si2168_cmd_execute: si2168 2-0064: wlen: 13, rlen: 0 [ 678.690115] si2168:si2168_cmd_execute: si2168 2-0064: i2c write: c0 12 00 0c 00 0d 16 00 00 00 00 00 00 [ 678.693331] si2168:si2168_cmd_execute: si2168 2-0064: wlen: 8, rlen: 1 [ 678.693337] si2168:si2168_cmd_execute: si2168 2-0064: i2c write: c0 06 01 0f 00 20 20 01 [ 678.701914] si2168:si2168_cmd_execute: si2168 2-0064: i2c read: 80 [ 678.701920] si2168:si2168_cmd_execute: si2168 2-0064: cmd execution took 6 ms [ 678.701923] si2168:si2168_cmd_execute: si2168 2-0064: wlen: 1, rlen: 13 [ 678.701926] si2168:si2168_cmd_execute: si2168 2-0064: i2c write: 02 [ 678.708631] si2168:si2168_cmd_execute: si2168 2-0064: i2c read: 80 00 44 34 30 02 00 00 00 00 00 00 00 [ 678.708636] si2168:si2168_cmd_execute: si2168 2-0064: cmd execution took 2 ms [ 678.708639] si2168 2-0064: unknown chip version Si2168- [ 678.714777] si2168:si2168_init: si2168 2-0064: failed=-22 [ 678.727424] si2157 0-0060: found a 'Silicon Labs Si2157-A30' [ 678.783587] si2157 0-0060: firmware version: 3.0.5 The answer to the 02 command seems really odd. You can see it is a Si2168, version 40, but I'd expect the second octet to say 42 instead of 00. Cheers, -olli On 3 June 2015 at 17:34, Antti Palosaari <crope@xxxxxx> wrote: > On 06/03/2015 12:29 PM, Olli Salonen wrote: >> >> I'm seeing the same issue as well. I thought that maybe some recent >> Si2168 changes did impact this, but it does not seem to be the case. >> >> I made a quick test myself. I reverted the latest si2168 patches one >> by one, but that did not remedy the situation. Anyway, the kernel log >> does not seem to indicate that the si2168_cmd_execute itself would >> fail (which is what happens after the I2C error handling patch in case >> the demod sets the error bit). >> >> olli@dl160:~/src/media_tree/drivers/media/dvb-frontends$ git log >> --oneline si2168.c >> >> d4b3830 Revert "[media] si2168: add support for gapped clock" >> eb62eb1 Revert "[media] si2168: add I2C error handling" >> 7adf99d [media] si2168: add I2C error handling >> 8117a31 [media] si2168: add support for gapped clock >> 17d4d6a [media] si2168: add support for 1.7MHz bandwidth >> 683e98b [media] si2168: return error if set_frontend is called with >> invalid para >> c32b281 [media] si2168: change firmware variable name and type >> 9b7839c [media] si2168: print chip version >> >> dmesg lines when it fails (this is with a card that has worked before): >> >> [66661.336898] saa7164[0]: registered device video0 [mpeg] >> [66661.567295] saa7164[0]: registered device video1 [mpeg] >> [66661.778660] saa7164[0]: registered device vbi0 [vbi] >> [66661.778817] saa7164[0]: registered device vbi1 [vbi] >> [66675.175508] si2168:si2168_init: si2168 2-0064: >> [66675.187299] si2168:si2168_cmd_execute: si2168 2-0064: cmd execution >> took 6 ms >> [66675.194105] si2168:si2168_cmd_execute: si2168 2-0064: cmd execution >> took 2 ms [OLLI: The result of this I2C cmd must be bogus] >> [66675.194110] si2168 2-0064: unknown chip version Si2168- >> [66675.200244] si2168:si2168_init: si2168 2-0064: failed=-22 >> [66675.213020] si2157 0-0060: found a 'Silicon Labs Si2157-A30' >> [66675.242856] si2157 0-0060: firmware version: 3.0.5 > > > Okei, so it has been working earlier... Could you enable I2C debugs to see > what kind of data that command returns? > > What I suspect in first hand is that Windows driver has downloaded firmware > to chip and linux driver does it again, but with incompatible firmware, > which leads to situation it starts failing. But if that is issue you likely > already noted it. > > regards > Antti > > -- > http://palosaari.fi/ -- 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
