Linux v4.15: Performance Goodies | Hacker News Reader

Linux v4.15: Performance Goodies

blog.stgolabs.net

179 pointsJoshTriplett43 comments

I hate to be "that guy", but that website is incredibly annoying for what amounts to a few paragraphs of text. Not everything needs to be animated.

This is a very popular Blogger theme. It might even be the default.

If anyone out there either is able to self-host their blog, or can afford a few bucks a month for it, I recommend Ghost (https://ghost.org). Example blog:

https://blog.ghost.org

Open source customizable markdown blog engine, with great themes by default. I feel like it doesn't get enough attention.

Seems to have no IPv6 support. I'd prefer to be able to access other peoples' blogs and have my own blog accessible at all times.

Send their Pro team a feature request. Either way you have the option to self-host it, which is why I'm promoting this here. If you want IPv6, it's super easy.

Hey! Ghost(Pro) is targeted mainly at business users who want a solid managed service, but there's absolutely no downside to self-hosting. We don't do any feature-gating, so you get the same software no matter where you want to run it. Easiest way to get up and running on your own env. is with Digital Ocean, who have a 1-click image: https://www.digitalocean.com/community/tutorials/how-to-set-...

If you want to just test it out locally first, that's easy too - https://docs.ghost.org/docs/install-local TLDR: npm install ghost-cli -g && ghost install local

Since you're here I just want to say that I love the operation you guys run. Ghost is great, the pro service is great and the open source, transparent development of the engine itself as well as the company/foundation is fantastic.

You guys are up there with Gitlab as a model of how open source businesses should work.

Thanks for being "that guy": it is awful & even ublock can't filter the particular issues out since the nonsensical blogger theme is embeddded left and right in both inline and external scripts..etc.

I cant tell because it requires JS to even render the text.

It's visually an empty page here, too.

Not only does it require javascript. It requires javascript hosted on a separate domain.

Blogger even blocks the "Go back one page" (Alt+Left / mouse side button) for me.

Visitor, you better stay.

It's not wrong to be that guy, if the page design is the top discussion topic then the design has failed and gotten in the way of the content.

If only it at least supported the Firefox Reading Mode I could tolerate anything, but as this doesn't ...

Printer Friendly [1] is useful in sites like this.

It removes all formatting, making it easy to read (and print, save to PDF, etc). And if you're printing or saving, you can easily delete blocks of content (like photos, diagrams, etc).

Strongly recommended. You can use either a bookmarklet or install the browser extension.

[1] https://www.printfriendly.com/

edit: seems this site traps mouse scrolls, so you can't scroll the page when using PF. What a shitty implementation! It works well saving to PDF tho.

It has scroll but inside the content box. Does the scroll bar not show for you?

Hmm, I can see the scroll bars, but seems wheel mouse scroll only works in some pages. E.g., no scroll on NYT [1], but it works on Wikipedia [2].

Interestingly, the Chrome extension is a lot more stable than the bookmarklet. For example, the NYT page scrolls just fine using the Chrome extension (probably because changes the page, while the bookmarklet just creates an overlay on top of the existing page).

Well, still has my vote as one of the most useful extensions.

[1] https://www.nytimes.com/2018/03/21/technology/personaltech/d...

[2] https://en.wikipedia.org/wiki/Jessica_Chastain

I also hate to be "that guy" but this site is completely inaccessible with JavaScript disabled and lynx -dump doesn't dump the main text. That said if you would like to restore some sanity to your web browsing, you can pull up the Google cache for a lack of these supposed animations and based off what others in this thread are saying, fully retain proper scrolling and back button behavior.

Best part is that it's a Google service for stripping the inanity of another Google service. Enjoy.

I agree. It is just a page with text, why add tons of Javascript just to render it?

Well, it is made by Google, so no wonder they cannot just serve plain HTML page.

Also they block scale change on mobile devices:

    <meta content='initial-scale=1.0, maximum-scale=1.0, user-scalable=no

It's really interesting seeing someone who has some intimate knowledge about the Linux kernel saying:

"In fact, with Linux v4.15 released, it is one of the rare times I've seen security win over performance in such a one sided way. Normally security features are tucked away under a kernel config option nobody really uses."

One just could be tricked into believing that the people suggesting that Linus doesn't care much about security could be right. ;-)

"One just could be tricked into believing that the people suggesting that Linus doesn't care much about security could be right."

I've got the impression over the years that Mr T does care about security. However, security is only one aspect of kernel development and stomping on hand waving when it is used to justify badly thought through bodges with a single purpose in mind seems fair enough to me. That is something he has done time and again and once you get through some of the rather harsh language, he invariably has a good engineer's approach to what gets released.

We don't simply use computers to "be secure" - that's nonsense. We use computers to do something and they should be secure as well. There are many, many design constraints imposed on OS kernel devs. The complexity of OS kernel development borders on the ridiculous to me sometimes (I'm just a sysadmin - I'm happy that I can compile and install the bloody thing when I need to by hand).

I did see your wink 8)

Reminds me of the security folks at my job that sport the biggest shinning smile of joy when they tell me that something can't be done because the security system blocks it.

It seems the page is entirely broken for me, even after disabling all script blockers. Console reports CSP violations.

Is there a plaintext version since the link above is only blank?

try this http://archive.is/j9Fd0

503

let's deepen the game https://app.printfriendly.com/p/g/L7s2KR

"Is there a plaintext version..."

Yes. But you have to make it yourself from some XML. Example of a script to do that (^M means insert a CR. Ctrl-V then Ctrl-M. Or just \r if you are using a GNU sed):

     #!/bin/sh 

     test $# -gt 0||exec echo usage: $0 proto://nameofBloggerblog.domain.tld;

     HTTPCLIENT=curl;# popular example only, not an endorsement of cURL;

     HTMLBROWSER=firefox;#popular example only, not an endorsement of Firefox

     a=$($HTTPCLIENT $1|exec sed 's/\\046/\&/g;s/\\46/\&/g;s/\\075/=/g;s/\\75/=/g;/targetBlogID/!d;s/.*targetBlogID=//;s/&.*//');test ${#a} -gt 0||exit;

     $HTTPCLIENT https://www.blogger.com/feeds/$a/posts/default|exec sed 's/^[0-9a-f]*^M//;s/&lt;/</g;s/&gt;/>/g;s/&amp;/\&/g;s/&quot;/\"/g;1i\
     <br><br>

     s/<name>/<br><br>name &/g;
     s/<uri>/<br>uri &/g;
     s/<generator>/<br>generator &/g;
     s/Blogger//;
     s/<id>/<br>id &/g;
     s/<published>/<br>published &/g;
     s/<email>/<br>email &/g;
     s/<title type=.text.>/<br><br>&/g;
     s/<openSearch:totalResults>/<br>total results &/g;
     s/<openSearch:startIndex>/<br>start index &/g;
     s/<openSearch:itemsPerPage>/<br>items per page &/g;
     s/<updated>/<br>updated &/g;
     s/<thr:total>/<br>thr:total &/g;
     s/<\/feed>/&<br><br><br>/;
     s/^M*/<br>/;
     ' |exec tr -cd '\12\40-\176' > 1.htm;

     $HTMLBROWSER file:///1.htm;

There is an easier method: just append AJAX crawling token to the query string: http://blog.stgolabs.net/2018/03/linux-v415-performance-good...

Same here. I had to enable everything for blogblog.com and scripts for blogger.com and apis.google.com on uMatrix.

It's amazing how this blog requires 1.6MB of data to show me a single page of text.

Ha ha, I though the same thing. Now did you mean amazing that it can use that much data and only show a single page of text instead of, say, also giving you a local weather update with an animated map displaying temperature, precipitation, and wind for the next four hours in five minute increments? Or amazing that this is only slightly on the high side of the envelope for data usage on 'modern' shitty webpages that 'do a lot of extra stuff' for no apparent reason?

You think that's bad? Just wait until you see people making 10 minutes off full 4k video on youtube to show you the same amount of content. It exists. It's terrifying.

It's almost at the point where I start longing back to the original Netscape Navigator and the simple WWW we had back then.

Enabling blogblog and blogger should do it. This is par for the course with the blogger sites.

the mentioned nested epolls is interesting, will be great if there are meaningful benchmarks on some real world applications.

check the GIT commit changes listed in the post, you will see the details there, e.g:

https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/lin...

Link with plain readable HTML: http://blog.stgolabs.net/2018/03/linux-v415-performance-good...

This should work on any blogger page.

Source: deprecate Google technology https://developers.google.com/webmasters/ajax-crawling/docs/...

This is crazy. They serve light HTML page for robots and a slow loading JS app for humans. And if an exception occurs inside their app humans get a blank page.

I would like to see these backported into RHEL7. They backported the PCID code due to the Spectre / Meltdown vulns.

Did RHEL backport full PCID or just the bare minimum for Meltdown?

Full PCID AFAIK.

    - [x86] kvm: x86: fix RSM when PCID is non-zero (Paolo Bonzini) [1531662 1530711]
    - [x86] mm/kaiser: use invpcid to flush the two kaiser PCID AISD (Josh Poimboeuf) [1519800 1519801] {CVE-2017-5754}
    - [x86] mm/kaiser: use two PCID ASIDs optimize the TLB during enter/exit kernel (Josh Poimboeuf) [1519800 1519801] {CVE-2017-5754}
    - [x86] mm/kaiser: use PCID feature to make user and kernel switches faster (Josh Poimboeuf) [1519800 1519801] {CVE-2017-5754}

This was a significant change and I did not see a formal write-up on it from Redhat. Such a write-up may exist by now.

If you have specific questions on the implementation, I can try to answer them.

Just the bare minimum. Only two PCIDs are used.

There wasn't enough time unfortunately.

Are there any plans to implement more of it?

Reply on news.ycombinator.com