Summary: | nvidia-kernel: bug with NVIDIA_kernel-1.0-7167-disable-preempt-on-smp_processor_id.patch | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | BlaisorBlade <blaisorblade_spam> |
Component: | [OLD] Core system | Assignee: | X11 External Driver Maintainers <x11-drivers> |
Status: | VERIFIED FIXED | ||
Severity: | normal | CC: | azarah, blaisorblade_spam, eradicator |
Priority: | High | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- |
Description
BlaisorBlade
2006-03-07 08:49:51 UTC
Please assign this bug to azarah which I cc'ed, since he introduced the patch. Do you have a link to Zander's NVIDIA_kernel-1.0-6629-1182399.patch that I can compare to ours? That link is /usr/portage/media-video/nvidia-kernel/files/1.0.6629/NVIDIA_kernel-1.0-6629-1182399.patch (I wouldn't search for the web-link now, it's much more awkward, but the patch is the same). You (Gentoo) fetched that from the forum, in fact. See my first comment for the description of the supposed patch history, where I say: == The current version of 1.0.6629/nv-disable-preempt-on-smp_processor_id.patch is therefore dummy; however, I can assume it made at least a bit of sense and fixed the warnings (while still having the imbalance bug) before Zander rolled his patch, and that you updated it wrongly. == Bye (btw, I tried to send this by answering email, but it seems this doesn't work here - if it does this will be duplicated, sorry for that). Sorry, but any news on this? On my side I'm just waiting a response, and I'm still confident on what I said. Need any further info or explaination? I rather wanted to abstain from replying, as I haven't really touched the nvidia drivers since I added a _fixed_ version of the patch _before_ Zander released his version, and some other reasons below. Whoever incorporated Zander's fix, and then reverted to the broken version of the patch (mismatched preempt_enable()) - see files/1.0.6629/nv-disable-preempt-on-smp_processor_id-2.patch in viewcvs. So basically, the patch is faulty as pointed out, and also unneeded as pointed out. I should point out that the reporter could actually have checked what really happened instead of trying to throw all the blame at me. Whatever the patch looks like now, it is entirely the work of other people, and in fact did not even have get_cpu() call back then (and yes, I did a thinko with the first one, but it was fixed with the second version of the patch). Nobody checked with me about the patch back then, and unfortunately I get stretches that I really cannot do much in regards to Gentoo (like now) - as was the problem back then. Should I just remove the patch then and all will be corrected? I'm answering to both Kris (for his question) and Martin (to excuse myself). For Kris: yes, everything I described will be resolved by removing the patch, and no further action is needed. I think that Martin also agrees, by reading his comment (I guess your doubt was only about whether another fix was needed). Indeed, every version except 6111 contains the patch we're going to remove; Zander's patch is applied explicitly in 6629 ebuild and should be already included in Nvidia sources for subsequent releases. So we can indeed remove it from all nvidia-kernel ebuilds - note there are 2 copies of the patch, both can be removed: $ ls /usr/portage/media-video/nvidia-kernel/files/1.0.*/*-disable-preempt-on-smp_processor_id.patch /usr/portage/media-video/nvidia-kernel/files/1.0.6629/nv-disable-preempt-on-smp_processor_id.patch /usr/portage/media-video/nvidia-kernel/files/1.0.7167/NVIDIA_kernel-1.0-7167-disable-preempt-on-smp_processor_id.patch === For Martin: First, thanks for replying, even if I've maybe been a bit rude - now the bug can be solved easily. I didn't want to throw all the blame onto you, I thought you were involved since I saw the entry in the ChangeLog (or whatever, I don't recall). I'm very sorry if I've been a bit rude in describing the bug (and I must admit I've been). I didn't want to fully track it down, I've just reported a bug; yes I could, but since I wanted mainly to discuss the merit of the code and not to judge anybody, I didn't. (In reply to comment #6) > Should I just remove the patch then and all will be corrected? > Yup Sorry for the delay. I have committed the suggested change to all ebuilds. I didn't bump any versions because I didn't think it would be necessary for everyone to rebuild. If the bug triggers we get a potential instability (I think you need to be an SMP box at least to experience it); so I'm not sure that not rebuilding is a good idea. However I'm not sure either,so feel free to decide. |