mirror of
https://github.com/AuxXxilium/linux_dsm_epyc7002.git
synced 2024-12-26 05:05:13 +07:00
75 lines
2.9 KiB
Plaintext
75 lines
2.9 KiB
Plaintext
|
|
||
|
If the box freezes hard with bttv ...
|
||
|
=====================================
|
||
|
|
||
|
It might be a bttv driver bug. It also might be bad hardware. It also
|
||
|
might be something else ...
|
||
|
|
||
|
Just mailing me "bttv freezes" isn't going to help much. This README
|
||
|
has a few hints how you can help to pin down the problem.
|
||
|
|
||
|
|
||
|
bttv bugs
|
||
|
---------
|
||
|
|
||
|
If some version works and another doesn't it is likely to be a driver
|
||
|
bug. It is very helpful if you can tell where exactly it broke
|
||
|
(i.e. the last working and the first broken version).
|
||
|
|
||
|
With a hard freeze you probably doesn't find anything in the logfiles.
|
||
|
The only way to capture any kernel messages is to hook up a serial
|
||
|
console and let some terminal application log the messages. /me uses
|
||
|
screen. See Documentation/serial-console.txt for details on setting
|
||
|
up a serial console.
|
||
|
|
||
|
Read Documentation/oops-tracing.txt to learn how to get any useful
|
||
|
information out of a register+stack dump printed by the kernel on
|
||
|
protection faults (so-called "kernel oops").
|
||
|
|
||
|
If you run into some kind of deadlock, you can try to dump a call trace
|
||
|
for each process using sysrq-t (see Documentation/sysrq.txt). ksymoops
|
||
|
will translate these dumps into kernel symbols too. This way it is
|
||
|
possible to figure where *exactly* some process in "D" state is stuck.
|
||
|
|
||
|
I've seen reports that bttv 0.7.x crashes whereas 0.8.x works rock solid
|
||
|
for some people. Thus probably a small buglet left somewhere in bttv
|
||
|
0.7.x. I have no idea where exactly, it works stable for me and alot of
|
||
|
other people. But in case you have problems with the 0.7.x versions you
|
||
|
can give 0.8.x a try ...
|
||
|
|
||
|
|
||
|
hardware bugs
|
||
|
-------------
|
||
|
|
||
|
Some hardware can't deal with PCI-PCI transfers (i.e. grabber => vga).
|
||
|
Sometimes problems show up with bttv just because of the high load on
|
||
|
the PCI bus. The bt848/878 chips have a few workarounds for known
|
||
|
incompatibilities, see README.quirks.
|
||
|
|
||
|
Some folks report that increasing the pci latency helps too,
|
||
|
althrought I'm not sure whenever this really fixes the problems or
|
||
|
only makes it less likely to happen. Both bttv and btaudio have a
|
||
|
insmod option to set the PCI latency of the device.
|
||
|
|
||
|
Some mainboard have problems to deal correctly with multiple devices
|
||
|
doing DMA at the same time. bttv + ide seems to cause this sometimes,
|
||
|
if this is the case you likely see freezes only with video and hard disk
|
||
|
access at the same time. Updating the IDE driver to get the latest and
|
||
|
greatest workarounds for hardware bugs might fix these problems.
|
||
|
|
||
|
|
||
|
other
|
||
|
-----
|
||
|
|
||
|
If you use some binary-only yunk (like nvidia module) try to reproduce
|
||
|
the problem without.
|
||
|
|
||
|
IRQ sharing is known to cause problems in some cases. It works just
|
||
|
fine in theory and many configurations. Neverless it might be worth a
|
||
|
try to shuffle around the PCI cards to give bttv another IRQ or make
|
||
|
it share the IRQ with some other piece of hardware. IRQ sharing with
|
||
|
VGA cards seems to cause trouble sometimes. I've also seen funny
|
||
|
effects with bttv sharing the IRQ with the ACPI bridge (and
|
||
|
apci-enabled kernel).
|
||
|
|