User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101 Firefox/52.0 Build ID: 20171114221957 Steps to reproduce: To reproduce: Install Debian Stretch on a machine with Intel i3-2100 CPU (HD2000 graphics) Download firefox 57 64-bit unpack run Actual results: Error message printed on console: MESA-LOADER: failed to retrieve device information WebGL doesn't work at all (3d disabled in google maps, all tests fail) Expected results: No error, WebGL working as in version 52 esr.

What's about:support graphics section look like?

need info for comment 1, and you can try the https://mozilla.github.io/mozregression/ to find the regression range.

I'm having the same issue on one of my machines, WebGL not working in Firefox 58.0.2. https://get.webgl.org/ reports WebGL is disabled or unavailable. Affected laptop runs Debian 9 (stretch) on i5-8250, Intel HD Graphics 620. I also have a desktop running Debian 9 on i5-6600K and Nvidia GTX970, where WebGL works even in Firefox 58. On Firefox 52-ESR from Debian repository above link shows a rotating cube, so WebGL works with that but not 58. I attached about:support data from both versions. Using the mozregression tool I found following point where it stops working: 2018-02-20T20:29:21: INFO : Narrowed inbound regression window from [8e1e06ad, 1dea32a1] (3 builds) to [8e1e06ad, b5fa0855] (2 builds) (~1 steps left) 2018-02-20T20:29:21: DEBUG : Starting merge handling... 2018-02-20T20:29:21: DEBUG : Using url: https://hg.mozilla.org/integration/mozilla-inbound/json-pushes?changeset=b5fa08551d6e74d8300fa94f3161afdffd867764&full=1 2018-02-20T20:29:22: DEBUG : Found commit message: Bug 1308400 - Report failures in file processes too. r=jld MozReview-Commit-ID: 549WuWKaJeM 2018-02-20T20:29:22: INFO : The bisection is done. 2018-02-20T20:29:22: INFO : Stopped

Jim, a sandboxing regression to WebGL.

Can this be reproduced with MOZ_SANDBOX_LOGGING=1 set in the environment?

(In reply to Jed Davis [:jld] (⏰UTC-7) from comment #7) > Can this be reproduced with MOZ_SANDBOX_LOGGING=1 set in the environment? …let me rephrase that: If someone could reproduce this with that environment variable set, and post the logs leading up to the MESA-LOADER error message, that should help determine what's going on here. (Changing the logging shouldn't affect whether the bug *can* be reproduced.) (In reply to Simo Mattila from comment #5) > I also have a desktop running Debian 9 on i5-6600K and Nvidia GTX970, where > WebGL works even in Firefox 58. If that system is using the Nvidia proprietary driver, then that makes sense: that driver isn't part of Mesa and doesn't have the Mesa loader / libudev / sysfs sandboxing problems we've run into.

Attached full log from the start with MOZ_SANDBOX_LOGGING=1, part related to page load starts with following line, leading eventually to MESA-LOADER error and beyond: Sandbox: Failed errno -2 op 0 flags 02000000 path /home/simo/firefox/libGL.so.1 The other working system with Nvidia graphics is indeed using the proprietary driver.

Likely duplicate of bug 1416016?

(In reply to Gian-Carlo Pascutto [:gcp] from comment #10) > Likely duplicate of bug 1416016? Probably. In about:config I put /sys/class/drm/card0/device/config,/sys/class/drm/renderD128/device/config to security.sandbox.content.read_path_whitelist (Firefox 58.0.2) and now it seems to work. At least get.webgl.org and Youtube 360 videos, if it is somehow site dependant. Although I haven't noticed.

Reproduced with MOZ_SANDBOX_LOGGING=1 Relevant lines: Sandbox: SandboxBroker: denied op=open rflags=0 perms=0 path=/sys/class/drm/card0/device/config for pid=14113 Sandbox: Failed errno -13 op 0 flags 00 path /sys/class/drm/card0/device/config MESA-LOADER: failed to retrieve device information