1335925 - Bobcat APU hw errata resulted in loads of AMD gpus false positives (trim/remove atiumd64.dll blacklist)

Status: UNCONFIRMED

(Core :: Graphics, defect, P3)

Description

[mirh]

Follow bug 1265281, bug 1292183 and bug 1320579. 

Are you sure considerations in issues above weren't related to bug 1198815 and bug 1267970?
Even because you blocked 2011/2012 dlls that have no reason to suddenly start to have problems now. 

I made some quick testing my 9.14.10.1171 6310 Graphics (with no bug 728533 or bug 839270 btw) and it all went smooth, both 1080p yt and vimeo.

Comment 1

[mirh]

I can confirm what had been hypothesized in bug 772330, comment 22 (which in turn totally confirm bug 839270, comment 24)

I put 19b5 into a Firefox Portable installation, then started to experiment with phantasy (because unfortunately all the remotely working test ideas of the past are no longer actual in 2017 changed web)
Long story short, what absolutely could reproduce crash reliably (albeit crash reporter was always 'null') is:

* override all possible direct2d/layers/drivers blacklisting for maximum attack surface
* open just about any page (e.g. search on google)
* open this in the same tab https://addons.mozilla.org/it/firefox/addon/reloadevery/versions/
* go apeshit, scroll, reload, do just anything
* (if after some tries this doesn't work, install the addon and set auto reload time to 1sec)
(using 7-zip built-in benchmark to keep cpu stressed might or might not have helped)

I tried all the possibly conceivable combinations of drivers (13.1, latest unofficial ones, and even windows VGA fallback) and bios (RTM release, and latest one, among other things fixed for erratum 688). 
100% of times crashing after some seconds with older BIOS, stable for minutes with newest one. 
(guess like I could have tested this under linux, where toggling just the specific fix is feasible.. but boy the difficulty of setting everything up there)

All of this with an asus K53U W7x64 machine (last updated sometime this summer) w/ E-350 stepping ON-B0 cpu. 
I took the liberty of raising importance (especially since I'm not sure how you'd be supposed to check for this crazy condition)

Regards.

Comment 2

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

Redirect a needinfo that is pending on an inactive user to the triage owner.
:bhood, since the bug has high severity, could you have a look please?

For more information, please visit auto_nag documentation.

Comment 3

[Bob Hood [:bhood]]

All of the references to other bugs sited within this report are marked as resolved/closed. Could you please weigh in on whether or not this report remains viable?

Comment 4

[mirh]

The blocklist I was talking about in the title was removed after bug 1678684. So.. I guess that was the biggest problem here (assuming you aren't having some other big catch all net elsewhere).

Still, given bug 1281759 comment 62, I'm left wondering if now it couldn't be rules number 25 and 44 to be redundant.
https://firefox.settings.services.mozilla.com/v1/buckets/blocklists/collections/gfx/records

Comment 5

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

In the process of migrating remaining bugs to the new severity system, the severity for this bug cannot be automatically determined. Please retriage this bug using the new severity system.