Google Uncovers CPU Bug For Geminilake, Affecting At Least Firefox & Chrome

[JonathanM]

They also added special options in the UEFI setup to disable the most likely source of the crashes (some deeper C-states, C6 or something), so that you can still live with an affected system if you can't just call AMD and send over the CPU.

That is a separate issue, though. The issue amd replaced cpus for was the segfault bug (can also be worked around by disabling the opcache) . The c6 lockup bug is still precent in replacement cpus and in zen+ cpus like r5 2400g. On my motherboard both the opcache and c6 (idle current) options were present in early versions, were subsequently removed and recently added again.

[milkylainen]

That's a juvenile dream.

You know full well that you can't just disable the program for something you can work around in software. This is the real world for like 1000% of actual software development.

Heh. I know. I work around buggy hardware all day long, most of the days.
But sometimes I have "hizzyfits/juvenile dreams" of telling a hardware vendor to "shove it".
There isn't much liability in software or hardware for most use cases.

I think that CPUs can be considered static in their presentation.
And as such, if their bugs affect your usage and cannot be fixed without deteriorating other parameters, vendors should be liable.
Ie, taking it back/compensating/replacing it.

[milkylainen]

I think that it's not so much only browser triggering the bug, but only browsers have so large widespread use and (for Chrome) so extensive telemetry info that allows them to actually be notified of crashes and their probable cause.

Most other software won't do that, and people will blame Windows, or the software, or the Gods that didn't smile back at them or something.

This is a prime reason why I strongly believe that telemetry about technical and performance information is a very important thing to have.

Quite likely. But doesn't Windows come with pretty nice bug reports/automatic telemetry that can be amassed into nice data points?

[timofonic]

Sounds more like it's a hard to reproduce bug that depends (amidst other things) on where the linker places stuff.

Hard to blame Google if a cpu can be singled out. Especially if at the same time Mozilla did the "don't care/works for me" instead of investigating.

Typical Mozilla attitude.

Mozilla sucks these days.

Intel is having a he'l of CPU bugs these days. I'm getting nostalgia of the old Pentium days.

[F.Ultra]

Typical Mozilla attitude.

Mozilla sucks these days.

Intel is having a he'l of CPU bugs these days. I'm getting nostalgia of the old Pentium days.

Mozilla does not have a microscopic fraction of the resources that Google have. So when faced with an extremely rare non-reproduceable bug there is not much that a smaller player such as Mozilla can do. If they had blamed Intel then everyone would be up in arms about how silly Mozilla is, but when some one as large as Google is doing it then every one is listening.

[coder]

What's so special about the browsers that they keep (apparently?) triggering the bug?

Just a wild guess, but browsers are continuously JIT-compiling loads of code - quite likely more than fits in L2 cache (and perhaps invalidating Instruction Cache lines, as they do it). Meanwhile, other code running on the CPU might be less likely to generate cache misses. Just two ways in which browsers might differ from most software.

[Guest]

What's so special about the browsers that they keep (apparently?) triggering the bug?

JavaScript is JITed these days, so with it's constant code generation you run a greater risk of creating a stream of instructions that causes this bug to trigger.

Software developers should not fix this issue themselves, neither should compilers or linkers work around it. People should just pressure Intel to roll out a microcode patch ASAP.

[Guest]

Typical Mozilla attitude.

Mozilla sucks these days.

Mozilla is striving to be the underdog that stays around and develops a web browser out of spite. They aren't a major corporation that has taking control in their agenda. The fact that they take the right approach and don't clean up the mess others make is right either way. Just report a bug to Intel, and tell the person that reported it to you to resolve the issue with them. Why would they pander to a major corporation that can't even keep up with chips in architecture they own?

Intel is having a he'l of CPU bugs these days. I'm getting nostalgia of the old Pentium days.

f00F, maybe they had bugs back in these days, but a crash is better than breaking memory safety.

[Palu Macil]

What's so special about the browsers that they keep (apparently?) triggering the bug?

Browsers are special enough that you can find ARM ("reduced" instruction set, ha) instructions specific to Javascript! FJCVTZS (v8.3-A chips and later) is "Floating-point Javascript Convert to Signed fixed-point, rounding toward Zero". Arguably, the browser is becoming the most important tool after whatever kernel you're using, so it makes sense that one of the richest tech companies on the planet would be able to find a complex bug affecting their own browser.

[starshipeleven]

Quite likely. But doesn't Windows come with pretty nice bug reports/automatic telemetry that can be amassed into nice data points?

Only for Microsoft products and services, and maybe partially for Apps from the store (to mimic Google's handling of Android apps).

Which means everything else must use its own telemetry