Re: staging driver s626 clashes with philips SAA7146 media/dvb based cards

[Herton Ronaldo Krzesinski]

Em Qua 17 Jun 2009, às 15:21:48, Ian Abbott escreveu:
> Herton Ronaldo Krzesinski wrote:
> > Em Quarta-feira 17 Junho 2009, às 09:26:00, Ian Abbott escreveu:
> >> Herton Ronaldo Krzesinski wrote:
> >>> Em Terça-feira 16 Junho 2009, às 17:51:21, Greg KH escreveu:
> >>>> On Tue, Jun 16, 2009 at 05:01:44PM -0300, Herton Ronaldo Krzesinski wrote:
> >>>>> Hi,
> >>>>>
> >>>>> The s626 (comedi) driver in staging conflicts with philips SAA7146
> >>>>> media/dvb based cards, because it claims the same vendor:device pci id
> >>>>> for all subdevice/subvendor ids. What happens is that for people that
> >>>>> have a philips SAA7146 based card, s626 if available gets loaded by udev
> >>>>> and makes system freeze (https://qa.mandriva.com/show_bug.cgi?id=51445).
> >>>> So a PCI device that does different things has the same device ids?
> >>>> ick, stupid vendors...
> >>>>
> >>>>> Looks like s626 shouldn't claim all 1131:7146 devices, either by
> >>>>> specifying specific subdevice/subvendor ids specific to s626 devices or
> >>>>> doing additional checks in its probe/attach function.
> >>>> If you can propose the proper sub ids, or the needed checks, please send
> >>>> a patch.
> >>> Can't propose proper sub ids here etc., as I don't know about/don't have s626
> >>> device, s626 author is CC'ed here to check this. But I could send a patch to
> >>> disable just the build of s626 if acceptable/desired for the moment.
> >> The Windows driver (<http://www.sensoray.com/downloads/sdk626.zip>) has
> >> this in the models section of the INF file:
> >>
> >> %sx26.DeviceDesc%=sxdrv.Device,PCI\VEN_1131&DEV_7146&SUBSYS_02726000
> >>
> >> And it looks like the correct device because this the strings section
> >> contains:
> >>
> >> sx26.DeviceDesc=   "Sensoray Model 626 Analog/Digital I/O"
> >>
> >> Interpreting the above information gives us:
> >>
> >> PCI Vendor ID = 0x1131
> >> PCI Device ID = 0x7146
> >> PCI Subvendor ID = 0x6000
> >> PCI Subdevice ID = 0x0272 (626)
> >>
> >> The Linux SDK for this board
> >> (<http://www.sensoray.com/downloads/s626-1.0.1.tar.gz> has the same info
> >> in the s626core.h file.
> >
> > Ok thanks. So lets limit s626 by its subvendor:subdevice id, patch follows:
> >
> > From 6f4d2430959a378ab754c5dbd3903fdcf33abe36 Mon Sep 17 00:00:00 2001
> > From: Herton Ronaldo Krzesinski <herton@xxxxxxxxxxxxxxx>
> > Date: Wed, 17 Jun 2009 13:31:15 -0300
> > Subject: [PATCH] Staging: comedi: use subvendor:subdevice ids for its pci id device table
> >
> > The current s626 comedi driver in staging conflicts with philips SAA7146
> > media/dvb based cards, because it claims the same vendor:device pci id
> > for all subdevice/subvendor ids. What happens is that for people that
> > have a philips SAA7146 based card, s626 if available gets loaded by udev
> > and makes system freeze (https://qa.mandriva.com/show_bug.cgi?id=51445).
> >
> > The s626 driver shouldn't claim all 1131:7146 devices. Fix this by
> > specifying specific known subvendor:subdevice ids in its pci id table
> > list.
> >
> > Reference: http://lkml.org/lkml/2009/6/16/552
> >
> > Signed-off-by: Herton Ronaldo Krzesinski <herton@xxxxxxxxxxxxxxx>
> > ---
> >  drivers/staging/comedi/drivers/s626.c |    8 ++++++--
> >  1 files changed, 6 insertions(+), 2 deletions(-)
> >
> > diff --git a/drivers/staging/comedi/drivers/s626.c b/drivers/staging/comedi/drivers/s626.c
> > index 30dec9d..b4b7713 100644
> > --- a/drivers/staging/comedi/drivers/s626.c
> > +++ b/drivers/staging/comedi/drivers/s626.c
> > @@ -110,9 +110,13 @@ static const struct s626_board s626_boards[] = {
> >  #define PCI_VENDOR_ID_S626 0x1131
> >  #define PCI_DEVICE_ID_S626 0x7146
> >
> > +/*
> > + * Note: always specify subvendor:subdevice ids in table below, to avoid
> > + * clash with Philips SAA7146 media/dvb based cards which have same
> > + * vendor:device ids as S626
> > + */
> >  static DEFINE_PCI_DEVICE_TABLE(s626_pci_table) = {
> > -       {PCI_VENDOR_ID_S626, PCI_DEVICE_ID_S626, PCI_ANY_ID, PCI_ANY_ID, 0, 0,
> > -               0},
> > +       {PCI_VENDOR_ID_S626, PCI_DEVICE_ID_S626, 0x6000, 0x0272, 0, 0, 0},
> >         {0}
> >  };
> >
> > --
> > 1.6.3.2
> 
> There are also a couple of calls to pci_get_device() that need changing
> to pci_get_subsys() in the s626_attach() function (so it also might be
> worth #define'ing the subsys numbers to avoid repeating these magic
> numbers).

Indeed, using pci_get_device() would still cause problems if s626 is manually
modprobed. Check new patch below, it now uses pci_get_subsys and ensures we use
subvendor or subdevice. I didn't removed the define, may be there could be more
s626 cards available in future with same id? If not, it could be cleaned up
and may be logic at s626_attach() simplified.