Bug 4962

Summary: b2c2-flexcop stops outputting data
Product: v4l-dvb Reporter: Ruslan (br)
Component: v4l-otherAssignee: Mauro Carvalho Chehab (mchehab)
Status: REJECTED INSUFFICIENT_DATA    
Severity: normal CC: akpm, alexn, bunk, protasnb
Priority: P2    
Hardware: i386   
OS: Linux   
Kernel Version: 2.6.13-rc3 Subsystem:
Regression: --- Bisected commit-id:

Description Ruslan 2005-07-29 01:35:46 UTC
Distribution: FC4
Hardware Environment: SkyStar2 rev 2.6c
Software Environment: KDE
Problem Description:

new b2c2_flexcop module how as old skystar2.c usually (1 or 2 time in day) 
freeze and i need to reload modules. But
"modprobe -r .."
"rmmod -f .."
"dvbnet -d 0"
don't work (resource templorary unaveable)
"ifconfig dvb0_0 down" CAUSE CRASH!!

in old kernel (FC2, 2.6.5 for example) with skystar2.c
rmmod -f is work.
Comment 1 Andrew Morton 2005-07-29 01:46:36 UTC
A bunch of patches went into the flexcop driver between 2.6.13-rc2 and 2.6.13-rc3.
Are you able to test 2.6.13-rc2?

Thanks.
Comment 2 Ruslan 2005-07-29 02:02:17 UTC
ok, i will test 2.6.13-rc2.

2.6.11 kernel (standart in FedoraCore4)
skystar2.ko cause system freeze if i try to rmmod -f
Comment 3 Alexander Nyberg 2005-07-29 02:05:41 UTC
From the helptext of rmmod:
 -f (or --force) forces a module unload, and may crash your
    machine.  This requires the Forced Module Removal option
    when the kernel was compiled.

Why are you still using it? Doesn't it get removed without it? Something may
hold reference counts to it and removing it with force then is bad.
Comment 4 Ruslan 2005-07-29 02:10:04 UTC
so how then i can reload modules if skystar2 freeze?
Comment 5 Ruslan 2005-07-29 02:11:40 UTC
rmmod
modprobe -r

resulted "resource temproray unavable"
Comment 6 Alexander Nyberg 2005-07-29 02:32:27 UTC
The real problem is that the driver freezes in the first place. When that
happens the next time please press alt+sysrq+t 3-4 times, run dmesg -s 10000,
include that output in a report where you describe your problems and send it to:

DVB SUBSYSTEM AND DRIVERS
P:      LinuxTV.org Project
M:      linux-dvb-maintainer@linuxtv.org
L:      linux-dvb@linuxtv.org (subscription required)
W:      http://linuxtv.org/developer/dvb.xml
S:      Supported
Comment 7 Andrew Morton 2005-07-29 02:41:51 UTC
bugme-daemon@kernel-bugs.osdl.org wrote:
>
> The real problem is that the driver freezes in the first place. When that
> happens the next time please press alt+sysrq+t 3-4 times, run dmesg -s 10000,
> include that output in a report where you describe your problems and send it to:

Make that `dmesg -s 1000000'.

Comment 8 Johannes Stezenbach 2005-08-03 15:27:24 UTC
The initial bug report mixes two things:

- "usually (1 or 2 time in day) freeze": I guess this is a well known
  bug where the hardware stops to generate interrupts for some
  unknown reason (we believe it is a hardware issue but people
  claim that it never happens with the Windows driver).
  The flexcop-pci driver recently added a workaround which
  resets the PID filter block in the demux everytime we tune to
  a different transport stream, which helped a lot. Apparently
  this won't help if you never tune (which is likely when using
  dvb_net). As a workaround one could add a watchdog and reset
  the pid filter block when the card stops generating irqs for a
  while. Please send patches.

- one needs to stop the dvb_net interface with ifconfig and release
  the interface with the dvbnet utility before the driver can be
  unloaded; forcing an unload naturally leads to Oopses. I believe
  the module refcount handling is correct, if someone manages to
  get an Oops please post it
Comment 9 Martin J. Bligh 2005-08-05 07:25:47 UTC
I tried to put a more sensible title on this bug, but failed ... could someone
who understands what's going on here update the summary field please? Would be
nice to have something other than "?" in akpm's tracking list ...
Comment 10 Johannes Stezenbach 2005-08-05 13:26:13 UTC
summary should be changed from "b2c2-flexcop module freezes" to
"b2c2-flexcop stops outputting data"

severity should be changed to "normal" as this bug is uncritical, just annoying

however, bugzilla prevents me from doing so myself, only the owner or
submitter are allowed change these fields :-(
Comment 11 Evgeniy Manachkin 2007-03-19 07:43:11 UTC
This mistake also is present at version 2.6.20.3.
Comment 12 Natalie Protasevich 2007-10-10 22:12:23 UTC
Ruslan, Evgeniy and others who had this problem - has it been resolved for you in recent kernels, 2.6.23+ as of now?
Thanks.