Em 22-06-2011 10:13, Andreas Oberritter escreveu: > On 06/22/2011 03:03 PM, Mauro Carvalho Chehab wrote: >> Em 22-06-2011 09:37, HoP escreveu: >>> 2011/6/22 Mauro Carvalho Chehab <mche...@redhat.com>: >>>> Em 21-06-2011 14:38, HoP escreveu: >>>>> 2011/6/21 Mauro Carvalho Chehab <mche...@redhat.com>: >>>>>> Em 21-06-2011 12:09, Andreas Oberritter escreveu: >>>>>>> On 06/21/2011 04:35 PM, Mauro Carvalho Chehab wrote: >>>>>>>> Em 21-06-2011 11:15, Andreas Oberritter escreveu: >>>>>>>>> On 06/21/2011 03:44 PM, Devin Heitmueller wrote: >>>>>>>>>> On Tue, Jun 21, 2011 at 7:04 AM, Andreas Oberritter >>>>>>>>>> <o...@linuxtv.org> wrote: >>>>>>>>>>> Mauro and Devin, I think you're missing the point. This is not about >>>>>>>>>>> creating drivers in userspace. This is not about open or closed >>>>>>>>>>> source. >>>>>>>>>>> The "vtuner" interface, as implemented for the Dreambox, is used to >>>>>>>>>>> access remote tuners: Put x tuners into y boxes and access them from >>>>>>>>>>> another box as if they were local. It's used in conjunction with >>>>>>>>>>> further >>>>>>>>>>> software to receive the transport stream over a network connection. >>>>>>>>>>> Honza's code does the same thing. >>>>>>>>>> >>>>>>>>>> I'm not missing the point at all. I realize exactly what Honza is >>>>>>>>>> trying to accomplish (and from a purely technical standpoint, it's >>>>>>>>>> not >>>>>>>>>> a bad approach) - but I'm talking about the effects of such a driver >>>>>>>>>> being introduced which changes the kernel/userland licensing boundary >>>>>>>>>> and has very real implications with how the in-kernel code is >>>>>>>>>> accessed. >>>>>>>>>> >>>>>>>>>>> You don't need it in order to create closed source drivers. You can >>>>>>>>>>> already create closed kernel drivers now. Also, you can create tuner >>>>>>>>>>> drivers in userspace using the i2c-dev interface. If you like to >>>>>>>>>>> connect >>>>>>>>>>> a userspace driver to a DVB API device node, you can distribute a >>>>>>>>>>> small >>>>>>>>>>> (open or closed) wrapper with it. So what are you arguing about? >>>>>>>>>>> Everything you're feared of can already be done since virtually >>>>>>>>>>> forever. >>>>>>>>>> >>>>>>>>>> I disagree. There is currently no API which allows applications to >>>>>>>>>> issue tuning requests into the DVB core, and have those requests >>>>>>>>>> proxied back out to userland where an application can then use >>>>>>>>>> i2c-dev >>>>>>>>>> to tune the actual device. Meaning if somebody wants to write a >>>>>>>>>> closed source userland application which controls the tuner, he/she >>>>>>>>>> can do that (while not conforming to the DVB API). But if if he >>>>>>>>>> wants >>>>>>>>>> to reuse the GPL licensed DVB core, he has to replace the entire DVB >>>>>>>>>> core. >>>>>>>>>> >>>>>>>>>> The introduction of this patch makes it trivial for a third party to >>>>>>>>>> provide closed-source userland support for tuners while reusing all >>>>>>>>>> the existing GPL driver code that makes up the framework. >>>>>>>>>> >>>>>>>>>> I used to work for a vendor that makes tuners, and they do a bunch of >>>>>>>>>> Linux work. And that work has resulted in a bunch of open source >>>>>>>>>> drivers. I can tell you though that *every* conversation I've had >>>>>>>>>> regarding a new driver goes something like this: >>>>>>>>>> >>>>>>>>>> === >>>>>>>>>> "Devin, we need to support tuner X under Linux." >>>>>>>>>> >>>>>>>>>> "Great! I'll be happy to write a new GPL driver for the >>>>>>>>>> tuner/demodulator/whatever for that device" >>>>>>>>>> >>>>>>>>>> "But to save time/money, we just want to reuse the Windows driver >>>>>>>>>> code >>>>>>>>>> (or reference code from the vendor)." >>>>>>>>>> >>>>>>>>>> "Ok. Well, what is the licensing for that code? Is it GPL >>>>>>>>>> compatible?" >>>>>>>>>> >>>>>>>>>> "Not currently. So can we just make our driver closed source?" >>>>>>>>>> >>>>>>>>>> "Well, you can't reuse any of the existing DVB core functionality or >>>>>>>>>> any of the other GPL drivers (tuners, bridges, demods), so you would >>>>>>>>>> have rewrite all that from scratch." >>>>>>>>>> >>>>>>>>>> "Oh, that would be a ton of work. Can we maybe write some userland >>>>>>>>>> stuff that controls the demodulator which we can keep closed source? >>>>>>>>>> Since it's not in the kernel, the GPL won't apply". >>>>>>>>>> >>>>>>>>>> "Well, you can't really do that because there is no way for the DVB >>>>>>>>>> core to call back out to userland when the application makes the >>>>>>>>>> tuning request to the DVB core." >>>>>>>>>> >>>>>>>>>> "Oh, ok then. I guess we'll have to talk to the vendor and get them >>>>>>>>>> to give us the reference driver code under the GPL." >>>>>>>>>> === >>>>>>>>>> >>>>>>>>>> I can tell you without a doubt that if this driver were present in >>>>>>>>>> the >>>>>>>>>> kernel, that going forward that vendor would have *zero* interest in >>>>>>>>>> doing any GPL driver work. Why would they? Why give away the code >>>>>>>>>> which could potentially help their competitors if they can keep it >>>>>>>>>> safe and protected while still being able to reuse everybody else's >>>>>>>>>> contributions? >>>>>>>>>> >>>>>>>>>> Companies don't contribute GPL code out of "good will". They do it >>>>>>>>>> because they are compelled to by licenses or because there is no >>>>>>>>>> economically viable alternative. >>>>>>>>>> >>>>>>>>>> Mauro, ultimately it is your decision as the maintainer which drivers >>>>>>>>>> get accepted in to the kernel. I can tell you though that this will >>>>>>>>>> be a very bad thing for the driver ecosystem as a whole - it will >>>>>>>>>> essentially make it trivial for vendors (some of which who are doing >>>>>>>>>> GPL work now) to provide solutions that reuse the GPL'd DVB core >>>>>>>>>> without having to make any of their stuff open source. >>>>>>>>>> >>>>>>>>>> Anyway, I said in my last email that would be my last email on the >>>>>>>>>> topic. I guess I lied. >>>>>>>>> >>>>>>>>> Yes, and you did lie to your vendor, too, as you did not mention the >>>>>>>>> possibilities to create >>>>>>>>> 1.) closed source modules derived from existing vendor drivers while >>>>>>>>> still being able to use other drivers (c.f. EXPORT_SYMBOL vs. >>>>>>>>> EXPORT_SYMBOL_GPL). >>>>>>>> >>>>>>>> AFAIK, the legal issues on writing a closed source driver using >>>>>>>> EXPORT_SYMBOL >>>>>>>> are not proofed legally in any court. While EXPORT_SYMBOL_GPL >>>>>>>> explicitly >>>>>>>> adds a restriction, not using it doesn't necessarily mean that the >>>>>>>> symbol >>>>>>>> can be used by a closed source driver. >>>>>>>> >>>>>>>> If you take a look at Kernel's COPYING file, the only exception to GPL >>>>>>>> license >>>>>>>> allowed there is: >>>>>>>> >>>>>>>> NOTE! This copyright does *not* cover user programs that use >>>>>>>> kernel >>>>>>>> services by normal system calls - this is merely considered >>>>>>>> normal use >>>>>>>> of the kernel, and does *not* fall under the heading of "derived >>>>>>>> work". >>>>>>>> >>>>>>>> IANAL, but, as EXPORT_SYMBOL is not a "normal system call", my >>>>>>>> understanding is that >>>>>>>> it is also covered by GPL. >>>>>>> >>>>>>> Of course. But as you should know, the GPL only covers derived work. >>>>>>> Whether or not a driver is a derived work of the kernel can only be >>>>>>> decided individually. It is my understanding that a Windows driver >>>>>>> ported to Linux is unlikely to be a derived work of Linux. >>>>>>> >>>>>>>> I was told that several lawyers defend the idea that all software >>>>>>>> inside the >>>>>>>> kernel tree is covered by GPL, even the aggregated ones. That was the >>>>>>>> rationale >>>>>>>> used to split the firmware packages from the kernel itself. >>>>>>> >>>>>>> However, I wasn't referring to the kernel tree at all. >>>>>>> >>>>>>>>> 2.) a simple wrapper that calls userspace, therefore not having to >>>>>>>>> open >>>>>>>>> up any "secrets" at all. >>>>>>>> >>>>>>>> A wrapper for a closed source driver is illegal, as it is trying to >>>>>>>> circumvent >>>>>>>> the GPL license. >>>>>>> >>>>>>> Is it? First, you are not a lawyer. Second, a wrapper is unlikely to be >>>>>>> illegal by its pure existence and a wrapper does usually not try to do >>>>>>> anything by itself. Third, you can implement a wrapper using normal >>>>>>> system calls (read, write, mmap, ioctl ...). That's what vtuner does, >>>>>>> too, to accomplish a totally different goal. Do you think vtuner is >>>>>>> illegal? I would be very surprised if it was. It perfectly matches the >>>>>>> license exception cited above. And even without the exception, a closed >>>>>>> driver in userspace would only very unlikely be a derived work of the >>>>>>> kernel. >>>>>> >>>>>> I think we're diverging from the subject. Most of those discussions are >>>>>> interesting on some lawyers forum, not here. >>>>>> >>>>>> My view about this subject is that vtuner can't give any additional >>>>>> permissions >>>>>> to the kernel GPL'd code, as vtuner were not made by the Kernel >>>>>> Copyright owners, >>>>>> nor were approved by them. So, the extra permission at the COPYING clause >>>>>> from kernel doesn't apply here, while the code is not merged into the >>>>>> Kernel. >>>>>> >>>>>> So, while it should be legal to use vtuner with a GPL'd client >>>>>> application, >>>>>> using it by a closed source application violates GPL. >>>>>> >>>>>> My understanding is that an addition of a code that exposes the internal >>>>>> DVB core API to userspace like that will require that all dvb developers >>>>>> that have copyright rights at the dvb core should explicitly ack with >>>>>> such >>>>>> change, otherwise adding such code will violate the original license. >>>>>> >>>>>> On the other hand, if vtunerc won't act as a proxy to userspace, it >>>>>> should >>>>>> probably be ok. >>>>> >>>>> Are you serious? Why there is not same violation on NFS? Or even beter >>>>> example NBD (network block device)? It sits in kernel for ages and nobody >>>>> cares. It looks for me like you should send some patch for removal such >>>>> "weak" places in kernel which allow to violate GPL. >>>>> >>>>> Do you really think that it is possible (in real, no in threory) to create >>>>> any networked subsystem for sharing anything over net the way >>>>> when it is not exposed (somehow) to the userspace? How will be >>>>> such system managable? Why there is usually companion daemon >>>>> there, which is responsible for managing connections etc? >>>>> >>>>> I think it is very evident you want find the way how to get yours word >>>>> back and return to your original position = such code is not acceptable. >>>>> Even if you still are not able to give anything clear. >>>>> >>>>> If I understand your last few mails, you won't accept such driver, isn't >>>>> it? >>>> >>>> You got wrong. You can't change someone's else license without their acks. >>> >>> That I understand very well. I never want to force anybody to change >>> his licence. >>> >>> I simply don't believe you that it is necessary. Why the same was not needed >>> with USBIP driver? If you check theirs nice big picture on >>> http://usbip.sourceforge.net/ >>> you see that it is exactly same technology like vtunerc, but for USB >>> subsystem. >>> Why such driver exists at all? >>> >>> And I'm sure I can find more examples inside kernel tree. What about NBD >>> (http://nbd.sourceforge.net)? Do you want find me more examples? >>> >>>> It is as simple as that. Getting everybody's ack is not that hard, if they >>>> accept that what you're doing is the right thing. We've got everybody's >>>> ack in the past to change the licensing for videodev2.h for example, to >>>> allow >>>> using the V4L2 API under BSD license (just the license API was changed, >>>> not the >>>> code itself). >>>> >>>>>> If people have different understandings, then we'll likely need to ask >>>>>> some >>>>>> support from Open source lawyers about this subject. >>>>> >>>>> My very little opinion is that waving GPL is way to the hell. Nobody told >>>>> me >>>>> why similar technologies, in different kernel parts are acceptable, >>>>> but not here. >>>> >>>> If you want to do the networking code at userspace, why do you need a >>>> kernel >>>> driver after all? The proper solution is to write an userspace library for >>>> that, >>>> and either enclose such library inside the applications, or use LD_PRELOAD >>>> to >>>> bind the library to handle the open/close/ioctl glibc calls. libv4l does >>>> that. >>>> As it proofed to be a good library, now almost all V4L applications are >>>> using >>>> it. >>> >>> LD_PELOAD is out of bussiness for normal work. It is technique for >>> development >>> and/or debugging. >> >> Well, libv4l successfully uses LD_PRELOAD in order to support all >> applications >> that weren't ported to it yet. It offers two ways: >> 1) you can use it as a normal library; >> 2) you can use it with LD_PRELOAD. >> >> >>> Library would be possible, but then you kill main advantage >>> - totally independece of changes inside userland DVB applications. >> >> Why? if you write a "dvb_open", "dvb_ioctl", ... methods with the same >> syntax of >> glibc open, ioctl, ..., the efforts to migrate an userspace application to >> use it >> is to just run: >> sed s,open,dvb_open,g >> sed s,ioctl,dvb_ioctl,g >> >> >> The library and the application will be completely independent. > > How do you transparently set up the network parameters? By using > environment variables? How do you pass existing sockets to the library? > How do you intercept an open() that won't ever happen, because no > virtual device to be opened exists?
Sorry, but I failed to see at the vtunerc driver anything network-related. Also, the picture shows that it is just acting as a proxy to an userspace code that it is actually handling the network conversion. The complete solution seems to have a kernel driver and an userspace client/daemon. Technically, doing such proxy in kernel is not a good idea, due to several reasons: 1) The proxy code and the userspace network client will need to be tightly coupled: if you add a new feature at the proxy, the same feature will need to be supported by the userspace daemon; 2) Data will need to be using copy_from_user/copy_to_user for every data access; 3) There's no good reason to write such code inside kernelspace. On a library based approach, what you'll have, instead is a library. The same userspace client/daemon will be needed. However, as both can be shipped together (the library proxy code and the userspace client/daemon), there are several advantages, like: 1) The library and the userspace client will be in sync: there's no need to check for version differences at the api, or providing any sort of backport support; 2) There's no need to recompile the kernel when someone wants to use the proxy; 3) The userspace won't be bound to the Kernel release schedule: When the code is stable enough, both libraries and userspace can be released at the same time. Mauro. -- To unsubscribe from this list: send the line "unsubscribe linux-media" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
