From: Oliver Endriss <o.endriss@gmx.de>
To: Mauro Carvalho Chehab <mchehab@redhat.com>
Cc: linux-media@vger.kernel.org, Ralph Metzler <rjkm@metzlerbros.de>
Subject: Re: [PATCH 0/5] Driver support for cards based on Digital Devices bridge (ddbridge)
Date: Fri, 15 Jul 2011 04:11:43 +0200 [thread overview]
Message-ID: <201107150411.45222@orion.escape-edv.de> (raw)
In-Reply-To: <4E1F8E1F.3000008@redhat.com>
On Friday 15 July 2011 02:47:27 Mauro Carvalho Chehab wrote:
> Em 14-07-2011 20:45, Oliver Endriss escreveu:
> > On Monday 04 July 2011 02:17:52 Mauro Carvalho Chehab wrote:
> >> Em 03-07-2011 20:24, Oliver Endriss escreveu:
> > ...
> >>> Anyway, I spent the whole weekend to re-format the code carefully
> >>> and create both patch series, trying not to break anything.
> >>> I simply cannot go through the driver code and verify everything.
> >>
> >> As the changes on CHK_ERROR were done via script, it is unlikely that it
> >> introduced any problems (well, except if some function is returning
> >> a positive value as an error code, but I think that this is not the
> >> case).
> >>
> >> I did the same replacement when I've cleanup the drx-d driver (well, the
> >> script were not the same, but it used a similar approach), and the changes
> >> didn't break anything, but it is safer to have a test, to be sure that no
> >> functional changes were introduced.
> >>
> >> A simple test with the code and some working board is probably enough
> >> to verify that nothing broke.
> >
> > Finally I found some time to do this 'simple' test.
>
> Thanks for testing it. Big changes on complex driver require testing.
>
> > Congratulations! You completely broke the DRXK for ngene and ddbridge:
> > - DVB-T tuning does not work anymore.
>
> I don't have any DVB-T signal here. I'll double check what changed there
> and see if I can identify a possible cause for it, but eventually I may
> not discover what's wrong.
>
> Before I start bisecting, I need to know if the starting point is working.
> So, had you test that DVB-T was working after your cleanup patches?
Yes, it worked.
And now I double checked with media_build of July 3th + my patch series:
It works as expected.
Well, I did not test DVB-C, but people reported that DVB-C was working
before I applied my cleanups. So I assume it worked.
> > - Module unloading fails as well. drxk is 'in use' due to bad reference count.
>
> I was eventually expecting a feedback about that. The patch that changed the
> behavior is this one:
>
> http://git.linuxtv.org/media_tree.git?a=commitdiff;h=f087cdd6f4fa8bef0e7588b124f52649d3cebe67
>
> What happens is that drxk_attach() allocates one state struct each time it is called,
> and it initializes both frontends with the state:
drxk_attach() was meant to be called once.
See dvb_input_attach() in ddbridge-core.c:
case DDB_TUNER_DVBCT_TR:
if (demod_attach_drxk(input) < 0)
return -ENODEV;
if (tuner_attach_tda18271(input) < 0)
return -ENODEV;
if (input->fe) {
if (dvb_register_frontend(adap, input->fe) < 0)
return -ENODEV;
}
if (input->fe2) {
if (dvb_register_frontend(adap, input->fe2) < 0)
return -ENODEV;
input->fe2->tuner_priv=input->fe->tuner_priv;
memcpy(&input->fe2->ops.tuner_ops,
&input->fe->ops.tuner_ops,
sizeof(struct dvb_tuner_ops));
}
break;
demod_attach_drxk() returns two frontends: fe and fe2.
> state = kzalloc(sizeof(struct drxk_state), GFP_KERNEL);
> ...
> state->c_frontend.demodulator_priv = state;
> state->t_frontend.demodulator_priv = state;
> ...
> *fe_t = &state->t_frontend;
> ...
> return &state->c_frontend;
>
> If you call it twice, the driver will loose the references for the first struct, and you'll
> end by having a memory leak at driver removal, if both DVB-T and DVB-C are registered.
See above.
> On my code, I've made sure to call it only once, and then I ended by having a reference = (u32) -1
> for device removal, preventing me to remove the driver, as .
>
> So, clearly, that function should be called just once. However, at DVB unregister,
> dvb_frontend_detach() will be called twice:
>
> static void unregister_dvb(struct em28xx_dvb *dvb)
> {
> dvb_net_release(&dvb->net);
> dvb->demux.dmx.remove_frontend(&dvb->demux.dmx, &dvb->fe_mem);
> dvb->demux.dmx.remove_frontend(&dvb->demux.dmx, &dvb->fe_hw);
> dvb_dmxdev_release(&dvb->dmxdev);
> dvb_dmx_release(&dvb->demux);
> if (dvb->fe[1])
> dvb_unregister_frontend(dvb->fe[1]);
> dvb_unregister_frontend(dvb->fe[0]);
> if (dvb->fe[1])
> dvb_frontend_detach(dvb->fe[1]);
> dvb_frontend_detach(dvb->fe[0]);
> dvb_unregister_adapter(&dvb->adapter);
> }
>From ddbridge-core.c:
if (input->fe2)
dvb_unregister_frontend(input->fe2);
if (input->fe) {
dvb_unregister_frontend(input->fe);
dvb_frontend_detach(input->fe);
input->fe = NULL;
}
So dvb_frontend_detach() was called exactly once.
Mauro, you must not change the logic of the frontend driver, unless you
verified what the bridge driver does...
> Or we would need to write a special unregister function at the driver just to
> cover the cases where DRX-K is used, as, on all other drivers, the attach
> function is called twice, even when the frontend supports two different types.
>
> See cxd2820r_attach, for example: instead of initializing both frontends
> with just one call, it can be called twice. If called a second time, it will
> just use the previously allocated data.
>
> I think that the better is to revert my patch and apply a solution similar
> to cxd2820r_attach. It should work fine if called just once (like ngene/ddbridge)
> or twice (like em28xx).
>
> > (DVB-C not tested: I currently do not have access to a DVB-C signal.)
>
> Hmm... are you sure that DVB-C used to work? I found an error on DVB-C setup for
> the device I used for test, fixed on this patch:
>
> http://git.linuxtv.org/media_tree.git?a=commitdiff;h=21ff98772327ff182f54d2fcca69448e440e23d3
>
> Basically, on the device I tested, scu command:
> SCU_RAM_COMMAND_STANDARD_QAM | SCU_RAM_COMMAND_CMD_DEMOD_SET_PARAM
> requires 2 parameters, instead of 4.
>
> I've preserved the old behavior there, assuming that your code was working, but I suspect that
> at least you need to do this:
>
> + setParamParameters[0] = QAM_TOP_ANNEX_A;
> + if (state->m_OperationMode == OM_QAM_ITU_C)
> + setEnvParameters[0] = QAM_TOP_ANNEX_C; /* Annex */
> + else
> + setEnvParameters[0] = 0;
> +
> + status = scu_command(state, SCU_RAM_COMMAND_STANDARD_QAM | SCU_RAM_COMMAND_CMD_DEMOD_SET_ENV, 1, setEnvParameters, 1, &cmdResult);
>
> Due to this logic there, at SetQAM:
>
> /* Env parameters */
> setEnvParameters[2] = QAM_TOP_ANNEX_A; /* Annex */
> if (state->m_OperationMode == OM_QAM_ITU_C)
> setEnvParameters[2] = QAM_TOP_ANNEX_C; /* Annex */
>
> This var is filled, but there's no call to SCU_RAM_COMMAND_CMD_DEMOD_SET_ENV. Also,
> iti initializes it as parameters[2], instead of parameters[0].
Sorry, I can't test it. Maybe Ralph can comment on this.
CU
Oliver
--
----------------------------------------------------------------
VDR Remote Plugin 0.4.0: http://www.escape-edv.de/endriss/vdr/
4 MByte Mod: http://www.escape-edv.de/endriss/dvb-mem-mod/
Full-TS Mod: http://www.escape-edv.de/endriss/dvb-full-ts-mod/
----------------------------------------------------------------
next prev parent reply other threads:[~2011-07-15 2:12 UTC|newest]
Thread overview: 45+ messages / expand[flat|nested] mbox.gz Atom feed top
2011-07-03 21:21 [PATCH 0/5] Driver support for cards based on Digital Devices bridge (ddbridge) Oliver Endriss
2011-07-03 21:23 ` PATCH 1/5] ddbridge: Initial check-in Oliver Endriss
2011-07-03 21:24 ` [PATCH 2/5] ddbridge: Codingstyle fixes Oliver Endriss
2011-07-03 21:25 ` [PATCH 3/5] ddbridge: Allow compiling of the driver Oliver Endriss
2011-07-03 21:26 ` [PATCH 4/5] cxd2099: Fix compilation of ngene/ddbridge for DVB_CXD2099=n Oliver Endriss
2011-07-04 10:14 ` Bjørn Mork
2011-07-03 21:27 ` [PATCH 5/5] cxd2099: Update Kconfig description (ddbridge support) Oliver Endriss
2011-07-04 0:06 ` Walter Van Eetvelt
2011-07-03 22:27 ` [PATCH 0/5] Driver support for cards based on Digital Devices bridge (ddbridge) Mauro Carvalho Chehab
2011-07-03 23:24 ` Oliver Endriss
2011-07-04 0:17 ` Mauro Carvalho Chehab
2011-07-14 23:45 ` Oliver Endriss
2011-07-15 0:47 ` Mauro Carvalho Chehab
2011-07-15 2:11 ` Oliver Endriss [this message]
2011-07-15 4:01 ` Mauro Carvalho Chehab
2011-07-15 3:56 ` Mauro Carvalho Chehab
2011-07-15 5:17 ` Oliver Endriss
2011-07-15 8:26 ` Ralph Metzler
2011-07-15 13:25 ` Mauro Carvalho Chehab
2011-07-15 17:01 ` Andreas Oberritter
2011-07-15 17:34 ` Mauro Carvalho Chehab
2011-07-15 23:41 ` Antti Palosaari
2011-07-16 12:25 ` Mauro Carvalho Chehab
2011-07-16 14:16 ` Antti Palosaari
2011-07-16 14:54 ` Mauro Carvalho Chehab
2011-07-16 15:40 ` Andreas Oberritter
2011-07-16 15:44 ` Antti Palosaari
2011-07-16 15:53 ` Andreas Oberritter
2011-07-16 15:59 ` Antti Palosaari
2011-07-16 16:37 ` Rémi Denis-Courmont
2011-07-17 2:51 ` Andreas Oberritter
2011-07-17 7:51 ` Rémi Denis-Courmont
2011-07-17 0:56 ` Mauro Carvalho Chehab
2011-07-17 3:02 ` Andreas Oberritter
2011-07-17 3:59 ` Mauro Carvalho Chehab
2011-07-17 7:39 ` Rémi Denis-Courmont
2011-07-17 8:01 ` Mauro Carvalho Chehab
2011-07-17 1:07 ` Mauro Carvalho Chehab
2011-07-16 15:40 ` Oliver Endriss
2011-11-03 7:49 ` Steffen Barszus
2011-11-03 17:24 ` Lars Hanisch
2011-07-15 4:18 ` Mauro Carvalho Chehab
2011-07-15 5:21 ` Oliver Endriss
2011-07-15 12:40 ` Mauro Carvalho Chehab
2011-07-17 11:44 ` Oliver Endriss
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=201107150411.45222@orion.escape-edv.de \
--to=o.endriss@gmx.de \
--cc=linux-media@vger.kernel.org \
--cc=mchehab@redhat.com \
--cc=rjkm@metzlerbros.de \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).