Bug 8623
Summary: | (fw_core) Kernel ooops and system freeze when trying to access a Sony HandyCam using a VIA chipset FireWire card | ||
---|---|---|---|
Product: | Drivers | Reporter: | Mauro M. (mmkernel) |
Component: | IEEE1394 | Assignee: | drivers_ieee1394 |
Status: | CLOSED CODE_FIX | ||
Severity: | blocking | CC: | excessory, mail, stefanr |
Priority: | P1 | ||
Hardware: | All | ||
OS: | Linux | ||
Kernel Version: | 2.6.21-1.3194.fc7 | Subsystem: | |
Regression: | --- | Bisected commit-id: | |
Attachments: | firewire: Only set client->iso_context if allocation was successful. |
Description
Mauro M.
2007-06-13 11:27:44 UTC
(Hmm, I sent an e-mail reply; I wonder if it will ever show up here. What I wrote was:) This Fedora kernel bug is tracked at https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=240771 . You could add your Cc there to get notified when it is fixed in a Fedora update. If you know how to configure and install kernels from source, please also check if the latest 2.6.22-rc kernel from kernel.org is affected. When you configure the new kernel, you can select the new experimental FireWire drivers to test for presence of the bug, and after that you can select the old drivers to (hopefully) get back to a working setup. OK, I will rebuild a custom kernel with the old driver stack, thank you. I have seen also the bug filed with RedHat. However I reckon that this bug report should be left open until resolved because if there are kernel drivers causing the system to freeze, these have either to be fixed or removed. What I hoped you could confirm is whether vanilla 2.6.22-rc has the bug too. But there is a posting on linux1394-devel just now saying that Fedora's libraw1394 won't actually work with the slightly newer fw-core in kernel.org's sources. For those who need to capture from their camera and do not want to wait for Fedora, here is a fix that will restore kernel, libraries and kino former and working FireWire stack: http://www.ezplanetone.com/xwiki/bin/view/KnowledgeBase/BrokenFC7FireWire About your replacement Fedora packages: I don't know what you and Fedora precisely have in their packages, but I suspect that Fedora 7's kino and libavc1394 do not have any updates regarding the new kernel drivers, and that Fedora 7's libraw1394 works with both the old and the new drivers. Also, the intro text is not 100% precise: The new FireWire drivers are available in Linus' kernel since linux-2.6.22-rc1. A difference to Fedora 7's kernel is that Linus does not distribute prebuilt kernels and the user can chose between old or new or both driver stacks when he configures the kernel. I spent a little time to inspect the code regarding the bug but didn't recognize a cause. My problem is that I don't have any AV/C device to reproduce the bug, but I plan to get one sooner or later. If you are somewhat familiar with C, know how to build kernels from source, and are willing to afford the spare time and downtime to crash kernels, you could help to get closer to the cause of the bug. As a primitive debug help, you can for example add printk(KERN_INFO "format string\n", ...); at interesting places in the driver sources to get messages out via dmesg. The entry point for userland calls to the new FireWire drivers is linux/drivers/firewire/fw-cdev.c. PS: Confirmation whether this bug is in Linus' tree would still be appreciated, because this is bugzilla.kernel.org, not bugzilla.redhat.com. PPS: Also, a test with an untainted kernel would be nice. Some maintainers delete bug reports against tainted kernels right away. Created attachment 11834 [details]
firewire: Only set client->iso_context if allocation was successful.
Does this fix it?
The patch in comment #6 has been merged to Linus' tree. If you know how to configure and install kernels from source, try the latest 2.6.22-rc (-rc7 at the moment) or 2.6.22 which will be released soon. Fedora users see also https://bugzilla.redhat.com/bugzilla/show_bug.cgi?id=240771#c14 Isochronous I/O is still not implemented in firewire-ohci for OHCI 1.0 controllers such as VIA VT6306 rev 46. Iso I/O requires an OHCI 1.1 controller. Kristian plans to implement an OHCI 1.0 compatible mode. However I assume that at least the oops was fixed in kernel.org's 2.6.22. Please reopen this bug if you still get the oops, but please test 2.6.23-rc first. Open a new bug if you need tracking for the OHCI 1.0 compatibility feature. Thanks. |