Petr Tesařík

Linux kernel hacker.
Pronouns: he/him

Public Key

npub1nh0xem054c0vrpgf3xw4ldrxulvfzr9dpzfdjssnsre650edxxtscneru4

NIP-05 Address

Profile Code

nprofile1qqsfmhnvah62u8kps5ycn82lk3nw0ky3pjks3ykeggfcpua28uknr9cpzemhxue69uhhyetvv9ujuurjd9kkzmpwdejhgqg5waehxw309aex2mrp0yhxgctdw4eju6t0ngnr63

Publishing to

relay.primal.net

relay.damus.io

nos.lol

Author Public Key

npub1nh0xem054c0vrpgf3xw4ldrxulvfzr9dpzfdjssnsre650edxxtscneru4

Show more details

Last Notes

2026-08-11 12:19:24 GMT

- reply

@nprofile…vv83 Dang, it might be my fault. I quoted from your blog and included a link:
https://lists.opensuse.org/archives/list/[email protected]/thread/WDWQ3354UXTZPWQ7LFTBPERHKCNTF7KB/

2026-06-16 06:02:42 GMT

- reply

@nprofile…0hde I like the sentiment, but really, STOP USING greed-powered social media.

2026-05-15 12:36:16 GMT

If you ask me, the current rate of Linux #kernel zero-day exploits is unsustainable, so while it's hard at the moment, it's likely to be much calmer in a few weeks.
However, it's unlikely to be the case for some other (closed-source) operating systems.
Just saying.

2026-05-11 14:39:34 GMT

Note to self: kexec into a new kernel works much better if you use an actual initrd for the --initrd argument rather than the bzImage binary…

2026-03-25 13:55:46 GMT

- reply

@nprofile…z8u4 It's complicated. Short answer: Yes, libvirt can work just fine with nft.
Long answer: Read this:
https://libvirt.org/firewall.html

2026-03-25 13:44:48 GMT

- reply

@nprofile…z8u4 By now, docker has left my system and will never make a comeback. But why did nobody warn me before I broke my system?
Besides, why didn't the #opensuse docker package revert those changes to iptables at unistall time?

2026-03-25 13:21:56 GMT

If your firewall starts behaving strangely after installing #docker on #opensuse #slowroll, the reason is that firewalld has switched to nft, but docker still uses iptables. You may have to install iptables (the CLI tool) to fix the damage.

2026-01-08 16:52:00 GMT

Idea: A keyboard with a big lever on the side. You type a #LLM prompt and pull the lever to spin the wheels of the slop machine (technically, to send the ENTER key).

2026-01-08 15:15:02 GMT

Achievement unlocked: Review a patch by a subsystem maintainer, who then finds my idea good and takes it for v2.
https://lore.kernel.org/linux-iommu/[email protected]/

2025-12-08 13:58:52 GMT

- reply

@nprofile…00p2 @nprofile…mxzg That has never been my feeling. Quite to the contrary, many of the former X11 developers (= current Wayland developers) have complained about the scope of X11 and promised they would never repeat that mistake.
BTW Have you ever heard about Xprint, for example? No, we surely don't need that in Wayland. Ever.

2025-12-08 11:20:42 GMT

- reply

@nprofile…mxzg Lucky you! ❤️

2025-12-08 08:02:40 GMT

Oh, great! This morning (after a #Slowroll update), @kde has finally reached feature parity between #Wayland and #X11 for me: When I try to log in, I get only black screen in both cases.

2025-12-03 17:11:32 GMT

And now #konsole even crashed on me. This never happened to me with KDE and #x11, but it's the third (and most serious) issue I have with KDE and #Wayland. Let me repeat, @nprofile…z87z, I have no objections against #Wayland as such. It's fighting with the most basic functionality that puts me off.

2025-11-28 09:54:48 GMT

Dear @nprofile…z87z, let me take this one from another thread. This! Just that it was 15 years ago!
Needless to say, it works (mostly) fine with X11. So much for all the #Wayland excitement.
https://bugs.kde.org/show_bug.cgi?id=15329
https://media.infosec.exchange/infosec.exchange/media_attachments/files/115/626/706/436/383/857/original/23914bdadd9e71e1.png

2025-11-20 07:38:16 GMT

- reply

@nprofile…00p2 @nprofile…w3ml @nprofile…z68c But, wait, then you might forget you need to launch Excel. It's better if it's right in front of your eyes all the time, so you can't forget!

2025-11-19 06:47:27 GMT

@nprofile…yrqu IIRC you like Ode to Joy. Did you go to #Kyiv on Sunday? It was performed at the railway station for free.
https://youtu.be/as_vMdL4PFI?si=_IOtRC1RCdKy7Fn1

2025-11-06 07:51:54 GMT

- reply

@nprofile…00p2 @nprofile…yrqu Now I can report my weight as 12 stones 1 kg. Enjoy.

2025-11-06 06:59:15 GMT

I'm sure @nprofile…yrqu has already seen it, but this is a genius contribution to the unit debate.
https://xkcd.com/3164/

2025-11-04 12:52:00 GMT

- reply

@nprofile…7645 My approach is this: Do not store private data on an Android device (or any other mobile device). It works, but I agree it's a bit extreme.

2025-10-07 13:42:01 GMT

I've been debugging an approx. 10% regression in iperf3. It's not fully analysed yet, but I can already share this neat trick to revert about one third of it (IOW 3 percent perf boost):
echo 0 > /sys/devices/virtual/graphics/fbcon/cursor_blink

2025-09-04 11:08:31 GMT

- reply

@nprofile…36ma even Grieg? have you already tried Mussorgsky?

2025-08-06 13:29:59 GMT

South Bohemian sky looks a bit better today.
https://media.infosec.exchange/infosec.exchange/media_attachments/files/114/982/056/980/914/184/original/823efc6bbae84eeb.jpg

2025-07-03 12:29:41 GMT

- reply

@nprofile…36ma That's the problem: Computers were not invented for people with average computer literacy. But then there was this Mr. Bill Gates who made a fortune on making people believe they were…

2025-06-23 18:38:40 GMT

I've just had a dinner conversation with my eldest son, and he discovered that we do our jobs all backwards. In principle, we take a variable X (some existing kernel code), apply function f (a benchmark) to arrive at some result Y (performance metric). But it would be much more useful to have the inverse function f' which takes the desired performance Y and returns the corresponding code X (or produces a proof that no such code exists). Genius! 😂

2025-06-17 18:57:18 GMT

It seems like #SUSE celebrates #pridemonth by hiring. This time a position in the Performance Team, and yes, that's my team (though I'm not the manager). Let's work together to make the kernel fast again!
#FediHire
#DEI guaranteed
https://suse.wd3.myworkdayjobs.com/Jobsatsuse/job/Czech-Republic-EMEA/Senior-Kernel-Engineer_71006927

2025-06-10 05:00:11 GMT

In case you want to join my company (#SUSE), we're looking for an #infrastructure engineer!
#FediHire
https://suse.wd3.myworkdayjobs.com/Jobsatsuse/job/Czech-Republic-EMEA/Senior-Software-Engineer--Core-Platforms-_71006946

2025-06-05 16:55:57 GMT

Just an idea:
LLM operators are extremely afraid of feeding AI output back into the model.
AI-generated texts are believed to be recognizable by the use of em-dash.
Are texts which contain em-dash filtered out now?
So—let me use em-dash in my texts lest they be stolen. 🖕

2025-05-30 09:40:15 GMT

It's not why pull-at-expiry timers are slow, but this code in kernel/time/timer_migration.c is pathetic:
new_migr_bit = find_first_bit(&active, BIT_CNT);
if (new_migr_bit != BIT_CNT) {
newstate.migrator = BIT(new_migr_bit);
} else {
Why not simply:
if (active) {
newstate.migrator = active & -active;
} else {
Is this a little-known trick?

2025-05-08 05:32:16 GMT

With 6.15-rc5, I'm now getting a page fault (non-present page) on in __srcu_read_unlock() on suspend. It has happened twice, both times while executing hidinput_led_worker.
Anyone else?

2025-04-29 12:18:22 GMT

Wow! The migration from #fosstodon was surprisingly smooth. I'm an infosec guy now. 😎