Skip to main content

A lot of the tweaks are deferrals because the contexts are getting too large. The FPU is a few registers on plain x86 but add in SSE and suddenly you are talking another 8 128 bit registers. On ARM's SVE the SIMD context can get up to 8kb of data! Debug registers can become similarly huge, especially if the hardware supports lots of watch and breakpoints.

#Linux #kernel 5.0-rc2 is out. I'd give it the codename 'normal' ;-) Quotes from the announcement (…) will explain: "[…] rc2 looks pretty normal […] Things look pretty normal […] As to actual changes: all looks fairly normal […] Go test, Linus […]"

That's unusual: The #Linux stable and longterm #kernel 4.20.2, 4.19.5 and 4.14.93 (released an hour ago) mark devm_memremap_pages() as well as hmm_devmem_{add, add_resource}() as EXPORT_SYMBOL_GPL. Thus proprietary drivers won't be able to use those functions anymore.

To clarify the "[…] won't be able to use […]" bit here's a quote from "[…] symbols exported by EXPORT_SYMBOL_GPL() can only be seen by modules with a MODULE_LICENSE() that specifies a GPL compatible license. […]".

For completeness: Last #LKML discussion about these patches before they got merged to mainline before #Linux version 5.0-rc1 recently: Commits for 4.19.y:

TWIMC: Patches got applied to the docs tree now \o/ Jon didn't like the commit msg and rewrote it (thx), but well, was my first non-trivial #linux #kernel patch, things like that can happen.

Finally can remove this from my todo list....…

Posted the second version of my documentation patch (which became a patch-set) that hopefully makes it easier to decode why your #Linux #kernel was tainted. This (like so often…) got way bigger than anticipated, but well, afaics it was worth the trouble.…


Why is there so little information about X1 to X10? Wikipedia at least mentions X10, but what about the previous versions?

no idea, sorry. Guess some kind of combination of "nobody really cares" and "ancient past"

Tomeu Vizoso – A #Panfrost milestone "[…] proper implementation of the winsys abstraction that makes use of ARM's kernel driver […] we can now run applications on GBM, from demos such as kmscube and glmark2 to compositors such as Weston[…]" #arm #mali…

Compile from a up2date #git checkout if you want to run a #Linux mainline #kernel. Do not rely on release patches and tarballs, as they might miss important fixes; luckily in the case of #Linux 5.0-rc1 it's only relevant for a few archs.

End of PSA ;-)…

First prerelease of #Linux #kernel 5.0 is out: "[…] The numbering change is not indicative of anything special. If you want to have an official reason, it's that I ran out of fingers and toes to count on, so 4.21 became 5.0. […] So go wild. Make up […]"…

Virtio_Blk finally supports discard and write zeroes in the #Linux mainline #kernel (thus it will be in 4.21, which might be called 5.0 instead). That will make it easier to shrink VM images by getting rid of data that got deleted (like TRIM does for SSDs)…