| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |
| Who | What | When |
|---|---|---|
| buxy81AtWork | yesterday night's experience: compiling a kernel driver module from scratch ain't easy.
i could not manage to place media_tree and media_build correctly checked out like https://www.linuxtv.org/repo/ said and then i tried multiple versions of https://www.linuxtv.org/wiki/index.php/How_to_Obtain,_Build_and_Install_V4L-DVB_Device_Drivers maybe i should start from scratch with the latter only? | [07:43] |
| ............................. (idle for 2h22mn) | ||
| hi again. is there any way to see the channel log from web irc client? | [10:09] | |
| ................. (idle for 1h21mn) | ||
| wow, compile has finished, modprobe-ing em28xx was successful. i'm so excited! :) but i have to wait to get home and try with the actual hardware... | [11:30] | |
| iive | buxy81AtWork: iirc the patch had tuner disabled, so it might not be able to get a program, without you recompiling again.
you said that you had opened your device, do you have a photo, have you written down the labels of the big chips? | [11:34] |
| buxy81AtWork | iive: now that i have the toolchain, i can recompile anytime, if needed.
the chips are marked EM2884, DRX3926, TDA18271 | [11:43] |
| iive | it looks like tuner absent is used for other dongles that use same tda, and it is called from the init source code. | [11:57] |
| buxy81AtWork | iive: i'm not really sure that i know what that means. if there's anything i should modify in the source code, please tell me. (the terratec h6 will be available to connect in about 6 hours from now.) | [12:06] |
| iive | the same if is present in H5 code, so it might work just as well. | [12:17] |
| buxy81AtWork | i really really hope that it will work. (the device identifies itself as H5 MKII) | [12:20] |
| ......................................................... (idle for 4h41mn) | ||
| buxy81 | mchehab: SUCCESS! :)
thank you very much for this little patch! here is the dmesg output: https://pastebin.com/WK3a95Lm the next step would be some higher level dvb tests, or just smash it into tvheadend somehow. | [17:01] |
| .................... (idle for 1h35mn) | ||
| yahal | I am using librelec on a Raspberry PI 3, with a Terratec H7 DVB-T/DVB-C USB-Box with external power supply and over vdr-osd gui it founds no channels and in dmesg I found the error: "drxk: Error -22 on get_lock_status" - Any idea how to fix this, DVB-Card seems to be correctly recognized. | [18:40] |
| ................. (idle for 1h22mn) | ||
| mchehab | buxy81: good news! | [20:02] |
| buxy81 | is the dmesg log normal? | [20:02] |
| mchehab | [ Â 489.934521] em28xx 1-1.5:1.0: The frontend of your DVB/ATSC card isn't supported yet
you likely need to modify em28xx-dvb too probably an one line change is enough to include the new board model if drx init is identical to HTC the remaining lines at the log looks OK right now you have: case EM2884_BOARD_CINERGY_HTC_STICK: | [20:03] |
| buxy81 | checking... | [20:06] |
| mchehab | you would need to add a new case below it, with the TERRATEC_H6 board name
(or below H5) something like: | [20:06] |
| buxy81 | i also have a case EM3884_BOARD_TERRATEC_H5 | [20:08] |
| mchehab | case EM2884_BOARD_TERRATEC_H5:
case EM2884_BOARD_TERRATEC_H6: I've no idea what model is closer to the one you have | [20:08] |
| buxy81 | i'll check the net for a quick comparison
and decide where to insert the case H6 row | [20:09] |
| mchehab | ok, I checked at the code... | [20:11] |
| buxy81 | weird. there are actually TWO htc sticks, one with and one without analog components | [20:12] |
| mchehab | the difference between HTC, HTC_XS and H5 are the init sequences
see terratec_h5_init(), terratec_htc_stick_init() and terratec_htc_usb_xs_init() the only way to be sure would be to get the GPIO sequence that the original driver does... by either sniffing the USB commands from the original driver... or looking at the .inf (with might contain something useful) | [20:12] |
| buxy81 | i'll try to get the .inf file first... | [20:13] |
| mchehab | basically, GPIOs are pins inside em28xx that are used to control other devices
(GPIO is Generic Pin I/O) each manufacturer can do whatever they want with those pins | [20:14] |
| buxy81 | whoa, i have a driver set. i can provide .inf files for terratec H6, H7, S7, T1, T3, T6. | [20:15] |
| mchehab | they're typically used to power down/power up other chips...
and to reset chips | [20:15] |
| buxy81 | right. so we need the actual pinout | [20:15] |
| mchehab | yes | [20:15] |
| buxy81 | well, searching for gpio in the inf file got:
; GPIO for AM DEMOD HKR,settings\OEMSettings,AMDEMOD,0x00010001,2 ; 2=GPIO2, 6=GPIO6 | [20:16] |
| mchehab | you could get it from the hardware, if you have its circuit... | [20:16] |
| buxy81 | damn that it's two lines | [20:16] |
| mchehab | (or if you're capable of discovering it by looking at the circuit lines) | [20:16] |
| buxy81 | ; GPIO for AM DEMOD
HKR,settings\OEMSettings,AMDEMOD,0x00010001,2 ; 2=GPIO2, 6=GPIO6 opening... | [20:16] |
| mchehab | I don't have the em28xx pinup handy, but I guess you can get it from the internet
in the specific case of DVB, it is used to control DRX-K power/reset lines and some lines at the tda18271 tuner | [20:17] |
| buxy81 | it's impossible to tell. there are TWO pcbs on each other like a sandwich. on one side is the em2884, the other pcb the other two chips... | [20:18] |
| mchehab | (typically: power, reset and AGC control)
yeah, getting it from the hardware is not trivial | [20:18] |
| buxy81 | there is a 22-pin connector between the two panels. ...and a 10-pin too.
can i do any harm with experimenting with the 3 init sequences as trial-and-error? ...or should i set up a windows usb sniffer instead... | [20:18] |
| mchehab | trial and error usually don't damage
sniffing is safer | [20:21] |
| buxy81 | i found this page https://www.linuxtv.org/wiki/index.php/Usbsnoop | [20:21] |
| mchehab | you can even sniff using QEMU
yep, that's it | [20:21] |
| buxy81 | which do you prefer? | [20:21] |
| mchehab | whatever works for you | [20:22] |
| buxy81 | right. :) | [20:22] |
| mchehab | In this specific case, we're insterested only on writes to registers 0x08 and 0x80 | [20:22] |
| buxy81 | as i have a 64-bit system, it should be sniffusb. hope it works.
cannot download. site is down. searching for another mirror... | [20:22] |
| mchehab | the problem is that several those sniffers don't work with newer USB stacks
found on newer OSes | [20:23] |
| buxy81 | h6 is not a very new stick. i have win7
do you know this sw? http://desowin.org/usbpcap/tour.html | [20:24] |
| mchehab | https://www.linuxtv.org/wiki/index.php/Bus_snooping/sniffing#Snooping_Procedures:
the above is the main link with regards to sniffing I don't use win sniffers for a **long** time | [20:25] |
| crope | use wireshark as a sniffer | [20:25] |
| buxy81 | alright, np | [20:25] |
| mchehab | nowadays, I just run windows under a QEMU and I run a perl script I wrote
VM eventually crashes, but GPIO setting happen early at the code | [20:26] |
| buxy81 | (i thought wireshark is for tcp/ip stuff, but i'll give it a shot) | [20:26] |
| mchehab | newer versions can sniff USB too | [20:27] |
| buxy81 | aham | [20:27] |
| mchehab | (not sure if this is supported at its window version) | [20:27] |
| crope | i think wireshark is ~only one usb sniffer that works with new windows versions. usbsnoop supports only xp
free sniffer I mean, there is some commercial available too | [20:27] |
| mchehab | crope: good to know!
buxy81: if you succeed using it, perhaps you could consider updating bus sniffing pages at our wiki in order to help others with similar issues | [20:28] |
| buxy81 | okay. usbpcap wants to restart my pc, wireshark download is really slow. | [20:30] |
| whew, i didn't know that h6 was so "rare" it is not supported yet. :/ but i'm happy to help now. :) | [20:41] | |
| ..... (idle for 21mn) | ||
| hi, i'm back.
i've made a usbpcap binary file as i'm plugging in the h6 strick and i have a wireshark that can open that log file (and i've lost my irc history by rebooting) | [21:02] | |
| mchehab | I have an H6 stick here
but it is different from yours on mine, eeprom is at I2C bus 0 very likely, the hardware vendor changed (again) the harware manufacturer (it is OEM) | [21:03] |
| buxy81 | yes, i already wanted to ask about your h6
my stick is made in taiwan barcode 9301249975 i'm trying to follow that guide you showed me there is no hex sequence like "40 02 00 00 ba 00 03 00" | [21:04] |
| mchehab | the better is to use a parser to filter it
there are some em28xx parsers at v4l-utils contrib/ dir | [21:08] |
| buxy81 | i'm new to wireshark... | [21:09] |
| mchehab | but the capture should match the format the tool expects
(it is a perl script) | [21:09] |
| buxy81 | did not find a v4l-utils dir on my linux installation... | [21:10] |
| mchehab | git.linuxtv.org | [21:11] |
| buxy81 | got it. now to get the snoop file to the reach of the util...
winscp rulez well i ran "parse_em28xx.pl terratec.pcap" and......nothing happened. | [21:14] |
| mchehab | you'll likely need to modify it
it is not prepared to handle wireshark's format | [21:16] |
| buxy81 | it's usbpcap's format
but i can save it using wireshark in several formats for example, is "suse 6.3 tcpdump" supported? or "redhat 6.1 tcpdump"? | [21:17] |
| mchehab | maybe you can parse the tcpdump format and feed the em28xx parser | [21:23] |
| buxy81 | ehh, trying all tcpdump exports, but they're all binary. em28xx expects a hex list in ascii... | [21:24] |
| mchehab | parse_tcpdump_log.pl | [21:26] |
| crope | mchehab: that parser didnt know used payload. I tried it very recently for one windows sniff... | [21:26] |
| buxy81 | oh :)
let's give it a try "Can't locate Net/Pcap.pm in @INC (you may need to install the Net::Pcap module)" | [21:26] |
| mchehab | you need to install it
that's actually not really needed when just parsing, but I never cared to make it optional :-P | [21:28] |
| buxy81 | :D
i'll have to search that package for my distro | [21:28] |
| mchehab | (that parse script actually can do the same as tcpdump)
Pcap.pm provides such functionality | [21:28] |
| buxy81 | Link type 249 ERROR: Link type is not USB at ./parse_tcpdump_log.pl line 603. | [21:30] |
| mchehab | no... just do something like: | [21:30] |
| crope | yep. thats the same error I saw...
it does not know that link type. iirc it was 220 script knows | [21:30] |
| mchehab | cat file.dump | ./parse_tcpdump_log.pl | ./em28xx/parse_em28xx_drxk.pl | [21:31] |
| crope | http://www.tcpdump.org/linktypes.html | [21:32] |
| buxy81 | i get the same error
i'll try with a different wireshark export format | [21:32] |
| crope | LINKTYPE_USBPCAP vs. LINKTYPE_USB_LINUX_MMAPPED
I tested all the formats and it was same for all | [21:32] |
| mchehab | ah
yeah, I never used it with a windows-parsed stuff | [21:33] |
| buxy81 | err, i'm lost a bit. :) please tell me what i should do :) | [21:34] |
| mchehab | the script handles one specific type of output... | [21:34] |
| crope | buxy81: use wireshark gui to examine data | [21:34] |
| mchehab | with is used by Linux TCP dump
apparently, wireshark on windows produce a different type of output | [21:34] |
| buxy81 | (perl scripts failed with suse and redhat style tcpdump files too) | [21:34] |
| mchehab | the script assumes this specific type: LINKTYPE_USB_LINUX_MMAPPED
if wireshark produces a different format, the script won't recognize | [21:35] |
| buxy81 | okay, so i have wireshark. according to https://www.linuxtv.org/wiki/index.php/Bus_snooping/sniffing#Snooping_Procedures: i could not find hex values "40 02 00 00 ba 00 03 00" | [21:35] |
| mchehab | I guess it shouldn't be that hard to add a new type there, but you need to be a perl programmer | [21:36] |
| buxy81 | but there are human-readable, parsed lines | [21:36] |
| mchehab | (I don't have any time to do that - at least not this week) | [21:36] |
| crope | what I did was I saved sniff to text format using wireshark and then made own dissector using python | [21:37] |
| mchehab | that parser wants a real tcpdump binary output file
# Decode an USB header mapped frame. The frame is defined at libpcap as: # #typedef struct _usb_header_mmapped { # u_int64_t id; # u_int8_t event_type; # u_int8_t transfer_type; ... buxy81: yeah, crope solution works | [21:38] |
| buxy81 | wireshark could export a verbose, too-human-friendly text format, but not an ascii hex (at least, i did not find it yet) | [21:39] |
| mchehab | it would also work if you could convert the output of wireshark to this format:
000000000 ms 000000 ms (000127 us EP=80) 80 06 00 01 00 00 28 00 >>> 12 01 00 02 00 00 00 40 40 20 13 65 10 01 00 01 02 01 000000000 ms 000000 ms (000002 us EP=80) 80 06 00 01 00 00 28 00 >>> 12 01 00 02 09 00 00 40 6b 1d 02 00 06 02 03 02 01 01 000000006 ms 000005 ms (000239 us EP=80) c0 00 00 00 45 00 03 00 <<< 00 00 10 000001006 ms 001000 ms (000112 us EP=80) c0 00 00 00 45 00 03 00 <<< 00 00 10 the em28xx/parse_em28xx.pl actually ignores the timing information at the beginning... parsing just: 80 06 00 01 00 00 28 00 >>> 12 01 00 02 00 00 00 40 40 20 13 65 10 01 00 01 02 01 | [21:39] |
| buxy81 | yeah right. now i have to fight with Mr. Wireshark to export that stuff for me | [21:40] |
| crope | buxy81: there is that kind of format:
126 0.405601 host 3.3.2 USB 39 URB_BULK out 0000 1b 00 20 e5 de 07 80 fa ff ff 80 f8 ff ff 09 00 .. ............. 0010 00 03 00 03 00 02 03 0c 00 00 00 0b 00 00 28 01 ..............(. 0020 02 00 00 12 22 b3 ec ....".. | [21:41] |
| buxy81 | i see the hex stuff, but i could not batch export it to file :( | [21:41] |
| crope | buxy81: export as "plain text" file
export packet dissections -> export as "plain text" file | [21:44] |
| buxy81 | Export Packet Dissections -> As Plain Text
oooooh if i resize the export window there are lots of options to tweak | [21:45] |
| crope | check packet bytes | [21:46] |
| buxy81 | ok now i have like
0000 1c 00 10 a0 aa 07 80 fa ff ff 00 00 00 00 0b 00 ................ 0010 00 04 00 01 00 00 02 08 00 00 00 00 80 06 00 01 ................ stupid multiline anyway, you get it let's feed it... unknown magic in header, cannot parse this file, make sure it is pcap and not a pcapng (run file <filename> to find out) and then convert with wireshark. at ./parse_tcpdump_log.pl line 185. oh | [21:46] |
| crope | buxy81: you have to make parser that converts it to format: "00:39 < mchehab> 000000000 ms 000000 ms (000127 us EP=80) 80 06 00 01 00 00 28 00 >>> 12 01 00 02 00 00 00 40 40 20 13 65 10 01 00 01 02 01"
or analyze bytes. it should not be hard to find gpios | [21:49] |
| buxy81 | maybe i was too strict on the export parameters
no luck yet. experimenting further... | [21:50] |
| guys, my hex dump does not have ">>>" anywhere. what am i doing wrong? | [22:00] | |
| mchehab | >>> (or <<<) gives the direction: output or input
that is what the old usbsnoop tools used to produce on USB, there are actually two packets... one on each direction they need to be merged, in order to understand what's really going on | [22:01] |
| buxy81 | hm, in this dump, the USB URB is also included as hex. there is an IRP information byte at 0x10 there is a direction byte 00 for FDO->PDO and 01 for PDO->FDO | [22:05] |
| crope | just make simple comversion script | [22:07] |
| buxy81 | maybe that's what i should do...... | [22:09] |
| .... (idle for 17mn) | ||
| iive | buxy81: why don't you just give them the wireshark file... they can extend their tools :D | [22:26] |
| buxy81 | iive: that's the easy way ...for me! :D
i could write a converter by myself, but i just cannot match the data fields of usbpcap and usbsnoop. currently only the direction bit is sure for me :D and packet data length of course it's 4 bytes in one, and 2 bytes in the other, but that's minor i don't understand "URB Request", "URB Value", "URB Index" in the context of the existing usbpcap data fields i have like "URB Function", "URB transfer type", and "Control transfer stage" on the other side | [22:26] |
| anyway, it's 00:30 local time, i must leave and go sleep.
thanks everyone, will continue tomorrow! | [22:37] | |
| ↑back Search ←Prev date Next date→ Show only urls | (Click on time to select a line by its url) |