Firefox 84.0 and 78.6 ESR

[Posted December 15, 2020 by ris]

Firefox 84.0 has been released. This version includes an accelerated rendering pipeline for Linux/GNOME/X11 users and improved performance and compatibility with Docker. This is the final release to support Adobe Flash. The release notes have additional details.

Firefox 78.6.0 ESR has also been released, with various stability, functionality, and security fixes. See the release notes for more information.

Firefox 84.0 and 78.6 ESR

Posted Dec 15, 2020 20:33 UTC (Tue) by pbonzini (subscriber, #60935) [Link] (3 responses)

Anybody knows what the "more modern method for allocation of shared memory" is?

Firefox 84.0 and 78.6 ESR

Posted Dec 15, 2020 23:16 UTC (Tue) by rillian (subscriber, #11344) [Link]

Possibly memfd_create instead of /dev/shm? See Bug 1440203.

Firefox 84.0 and 78.6 ESR

Posted Dec 15, 2020 23:33 UTC (Tue) by Duncan (guest, #6647) [Link]

Since your question made me curious... "Complete list of changes" buglist link at the bottom of the release-notes page, then searching "shared memory", results in a single hit:

Use memfd_create for shared memory where available.

https://bugzilla.mozilla.org/show_bug.cgi?id=1440203

Originally filed three years ago. From that bug >>>

memfd_create is a Linux-specific system call, added in kernel 3.17 (released 2014-10-05), which creates a pseudo-file backed by anonymous memory and not associated with a filesystem, accessible only through the returned file descriptor.

It's helpful for sandboxing, because it doesn't need any access to the actual filesystem. We're currently proxying these accesses through the parent process, so avoiding that may improve performance.

More interestingly, it should also work around bug 1245239 and related bugs because these files aren't tied to any tmpfs instance's size limit; the memory is accounted for as if calling mmap with MAP_ANON (i.e., like malloc).

<<<

Apparently the old tmpfs usage is still a fallback for old kernels/glibc, etc. Based on the bug FBSD has a similar call but my scan didn't pick up whether it's implemented this release for them or if it'll be later.

The bottom of the bug has some performance benchmark metrics, 2-3% improvement, gl, svg, etc.

Firefox 84.0 and 78.6 ESR

Posted Dec 16, 2020 0:32 UTC (Wed) by flussence (guest, #85566) [Link]

Sounds like they've finally fixed the long-standing bug where it would gradually fail to load new tabs if /dev/shm isn't enormous, which is the default on systems without several gigabytes of RAM. Hopefully video playback will be more reliable too; it's been flaky and slow since the e10s project first started.