OK #Firefox users, this is a special request. I'm looking for someone with a high-end Raptor Lake machine that is experiencing instability in Firefox (or anything else for the matter). The best candidate is a CPU from the 13900* SKUs. K or not doesn't matter as long as it's one with the 8P+16E cores.
We have a potential workaround for one of the most common bugs of this CPU, but we need a broken machine to test it.
@gabrielesvelto if need be, i have an i9-13950HX under my paws with the 8P+16E config; no instability issues as far as i can tell (at least no crashes and nothing that resembles what's described in the thread), but i can be a control machine to determine what is affecting some users. i'm pretty rough with firefox (intensive CPU stuff, hw acceleration for video, audio, lots of tabs open), if that could help make things worse.
@gabrielesvelto I have an 13900K which seems (?) to be getting SIGSEGV in random processes when the clock speed is above the base clock and the core voltage is set below 1.4V
The Firefox crash count is in the hundreds I think ?
@tranquillity that sounds like you're one of the users affected by this. If you have submitted some of the crashes you'll have links in the about:crashes page to the crash reports. Can you check if the crash reports in those links look like this one: https://crash-stats.mozilla.org/report/index/ea67a961-6afa-4043-ba30-bc6cd0260522
In particular if you click on the Bugzilla tab of those crash reports they should have a link to this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1950764
@gabrielesvelto is there a way to dump the crashes somehow without submitting them? The machine is currently serving as my server and I only have ssh access to it
I might also have some coredumps still lying around depending on how systemd was feeling
After reading thru the bugzilla thread, I really wanna go dig up the crash logs from FF somehow
If it's just
potentially because the generated code is issuing instructions that manipulate x86 8-bit subregisters to compose the 16-bit value and that's one of the codepaths that are more likely to trigger the hardware bug.
I can convince the codegen of a compiler on my system to just not do that and probably (?) make it at least less bad system wide? The kpanics every 1-3 days were pretty bad so getting it patched in the kernel would be nice... I had zero leads back when I tried to diagnose this in 2023/2024 and gave up
@tranquillity not that I know of. And it isn't the only issue, just the most common. We have an entire collection of bugs on the topic: https://bugzilla.mozilla.org/show_bug.cgi?id=1975808
... and we stopped collecting them because they were too many, and we couldn't do anything to mitigate them.
@gabrielesvelto I have a i9-13900K at work. Not using FF there, but I could.
@jisotupa could you try Firefox Nightly on it? Just regular browsing. If the machine is affected it should crash a lot, submit the crash report and send me the links that will be left in the about:crashes page. If it's the bug we're looking for they'll look like this: https://crash-stats.mozilla.org/report/index/162ca882-9961-4533-8572-a56dd0260522
And the "Bugzilla" tab in the crash report will point to this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1950764
@gabrielesvelto yes sir, installed! I have a couple of hours remaining today, then I'll continue on Monday. Let's see what happens. 鈽猴笍
@jisotupa thanks a lot 馃檹
Maybe contact some of the PC hardware channels such as @gamersnexus , @jayztwocents or @paulshardware and see if their viewers can help?
They would be the target audience for such hardware specs and more likely that most to use Firefox 馃檪
If you get stuck then I can pass word to various podcasts who would be most interested including @SGgrc on the Security Now! Podcast.
@simonzerafa @gamersnexus@bird.makeup @jayztwocents@bird.makeup @paulshardware @SGgrc that's an excellent idea! Thank you!
@gabrielesvelto @paulshardware @SGgrc
Security Now! did mention your work about system instability related to environmental heat 馃檪
Maybe contact Steve directly via email at Security@grc.com and see if he can ask for help / volunteers on the next episode 馃檪
@Rairii yes, that would work too and is likely affected by the bug. Can you double-check with a tool like CPU-Z that your CPU has the following CPU ID?
Family: 6 (06h)
Model: 183 (0B7h)
Stepping: 1 (01h)
It would also be interesting to know on which microcode version you are on. And yes, downclocking the CPU is definitely a way to fix this problem, because it's caused by a race in a circuit that's on the critical path. A slower clock will not hit it.
@Rairii if it's not a bother could you try removing the fix and trying to run Firefox nightly? You can find it at the bottom of this page: https://www.firefox.com/it/channel/desktop/
Just browse normally, on a machine with the bug tabs will crash rather often. Submit the crash report and check it out in about:crashes. If it is the one we're looking for it will look like this: https://crash-stats.mozilla.org/report/index/b645e7f3-e6a6-4626-9e29-7961d0260522
Clicking on the bugzilla tab will lead you to this bug: https://bugzilla.mozilla.org/show_bug.cgi?id=1950764
Which would confirm it.
Install bonfire.cafe
Get the full app experience
