1907505 - Frequent Random Tab Crashes (most of the times Reddit, Youtube, LinkedIn, Google etc.)

Status: UNCONFIRMED

(Core :: JavaScript: GC, defect, P5)

Description

[sonyvlt25isonyvlt25i]

User Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:128.0) Gecko/20100101 Firefox/128.0

Steps to reproduce:

It is hard to reproduce as it happens randomly.
It can happen on active tab, or background tab.
It does not matter if I have a lot of tabs open or just a few.
It can happen instantly after I have just opened a tab, minutes or even hours after.
It happens if FF running without extensions or on a new profile.

(Heck, I have just had a tab crash while writing this bug report... glad the text was saved, so I do not need to re-write all over again)

My system is: Win10, 13900k, 4090, 64Gb DDR5, 2Tb SSD. No OC, no XMP, No MCE (Enforce all limits is ON) in BIOS, Performance Profile, SVID - Intel Failsafe - but I have tried many other profiles, all possible combinations and settings, and power limits, like 253W 253W +400A or 253W 253W +307A for CPU.

Memtest86 does not give any errors.
SFC is also OK.
Firefox was very stable for me and no such issue happened before 2 months ago.
No changes to the system were made since then, apart a couple of installed games, that run normally without crashes.

I am at my wits, help, please, I love Firfox.

Actual results:

Nothing happened, the issue, random tab crash still persist and it is very frustrating.

Expected results:

No tabs crash at all. I have a NASA PC that can run everything, from Office programs to 4K VR and Emulation.

Comment 1

[Ryan VanderMeulen [:RyanVM]]

Can you please share links to any recent crash reports you might have in about:crashes?

Comment 2

[sonyvlt25isonyvlt25i]

(In reply to Ryan VanderMeulen [:RyanVM] from comment #1)

Can you please share links to any recent crash reports you might have in about:crashes?

Will be happy to:

https://crash-stats.mozilla.org/report/index/a3563b32-955f-4221-9ef3-36dcf0240712
https://crash-stats.mozilla.org/report/index/288936b2-6a87-45e1-bc33-b85ec0240712
https://crash-stats.mozilla.org/report/index/ebb1b727-d6e6-4767-bbef-39b680240712
https://crash-stats.mozilla.org/report/index/b058ccdd-e827-4972-9f66-cf7870240703
https://crash-stats.mozilla.org/report/index/56351747-927b-458c-8bd7-508610240703

Thank you for looking into it. FF was and is my primary browser on any device I use, for the last 20+ years.

Comment 3

[Ryan VanderMeulen [:RyanVM]]

Blech, Raptor Lake strikes again :(

That CPU has known stability issues acknowledged by Intel (lots of tech press articles about it) and we've had a number of different crash bugs that tie back to it. Looks like they're trying to roll out BIOS updates to help.
https://community.intel.com/t5/Processors/June-2024-Guidance-regarding-Intel-Core-13th-and-14th-Gen-K-KF/m-p/1607807#M73544
https://www.tomshardware.com/pc-components/cpus/intel-offers-new-guidance-on-13th-and-14th-gen-cpu-instability-but-no-definitive-fix-yet

Comment 4

[BugBot [:suhaib / :marco/ :calixte]]

The Bugbug bot thinks this bug should belong to the 'Firefox::Tabbed Browser' component, and is moving the bug to that component. Please correct in case you think the bot is wrong.

Comment 5

[Nicolas B. Pierron [:nbp]]

Thanks a lot for the detailed description.

I will still CC'ed some person who are working on the Garbage Collector, in case they have any insight that I do not have.

However, I suspect RyanVM suggestion is unfortunately likely, as the GC would hit hard on the memory while Firefox might be asking a lot from the CPU, and the mix of temperature and high memory usage might be a source of bit-flips localized in the GC.

In the mean time, if you are able to find which version of Firefox would be the last to not crash, this might help us finding the root cause as well as being a temporary fallback for you.

https://archive.mozilla.org/pub/firefox/releases/

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Also FYI, Intel announced yesterday that they expect to rollout CPU microcode updates in August that are intended to address this instability. Those are delivered via Windows Update also, so with any luck things will improve soon.
https://community.intel.com/t5/Processors/July-2024-Update-on-Instability-Reports-on-Intel-Core-13th-and/m-p/1617113

Comment 7

[Steve Fink [:sfink] [:s:]]

(In reply to Nicolas B. Pierron [:nbp] from comment #5)

However, I suspect RyanVM suggestion is unfortunately likely, as the GC would hit hard on the memory while Firefox might be asking a lot from the CPU, and the mix of temperature and high memory usage might be a source of bit-flips localized in the GC.

"localized in the GC" might be a little confusing. If hardware corrupts memory anywhere, then by far the most likely place to see a crash is in GC-related code, for a simple reason: the GC by its nature scans through lots of memory and follows pointers found in that memory, which is how you'll get a crash. With a lot of other code, even if it accesses incorrect memory values, it will often just result in a pixel being the wrong color or some internal text being a little distorted (and usually not shown on the screen). The GC's job is to look at everything to figure out what needs to be kept, and looking at everything is a good way to stumble across invalid memory and do things with it that the OS or CPU can tell are invalid.

Still, it's a bad idea for us to assume crashes are due to hardware. We've had lots of bugs in the GC and undoubtedly have more. But the nature of the crashes you listed is varied enough to make it less likely to be a result of a bug. Normally bugs trigger crashes in the same place or places related in some understandable way. Those crash locations are somewhat related but not enough, and together with the Raptor Lake issue that RyanVM called out, my money is on the known instability.

But as nbp said, if there is an older version of Firefox that is more stable, that would be a much stronger indication of a bug.

(GC = Garbage Collector, as you probably already know)

Comment 8

[sonyvlt25isonyvlt25i]

I have just received new 13900k after RMA, going to use it for a couple of days and then report back. Highly likely as Ryan said, Intel dropped a ball.