Chris Down

Hi there! I work on the Linux kernel at @Meta and help maintain systemd. @[email protected]

[Chris Down @unixchris]

Big thanks to @hnakamur2 for diligently translating two of my articles — "In defence of swap" (https://t.co/YHCrNIS5hD) and "Linux memory management at scale" (https://t.co/gj8PAHURIP) — to Japanese! ご苦労様です！🇯🇵

[Chris Down @unixchris]

@theodorvaryag Anecdotally speaking, many of the distros which have moved towards zram did so because they believed it would be simpler and less error prone. But having spoken to maintainers, many have been surprised by the edge cases. I would expect more zswap adoption over the next few years.

[Chris Down @unixchris]

@theodorvaryag Thank you for reading! The main one is Arch, which has it by default with zstd:

% zgrep -e ZSWAP= -e ZSWAP_COMPRESSOR_DEFAULT= /proc/config.gz
CONFIG_ZSWAP=y
CONFIG_ZSWAP_COMPRESSOR_DEFAULT="zstd"

Some Ubuntu flavours also enable(d?) it by default, some have moved towards zram.

[Chris Down @unixchris]

@kaihendry It's a bit rough around the edges and missing some things (like CM_OUTPUT_CLIP and rofi-script support), but clipmenu-git in AUR now builds the C version :-)

[Chris Down @unixchris]

And many thanks to @samwhoo and @javierhonduco for all their suggestions during writing this!

[Chris Down @unixchris]

Ever wondered how Linux manages process states, especially uninterruptible sleep ("D" in ps/top/etc)? What if I told you that -- despite what most Linux users think -- you can often actually interrupt or kill them?

https://t.co/OQjYfccmtw

[Chris Down @unixchris]

@hnakamur2 Nice to bump into you again in another context after the swap post. Hope you are well!

[Chris Down @unixchris]

@ianpatterson99 For this particular one, I sent https://t.co/Rd9L7Txv0W, but then there was some back and forth about just removing the message and I had other patches, so I forgot about it. I can rebase and resend it if you would find it useful.

[Chris Down @unixchris]

@samwhoo Great article, congrats on finishing it up! I know how easy it is to get in the mire of being half way done on these and having to get the context back to finish it, and having one with so many animations to work out must have been even more taxing :-)

[Chris Down @unixchris]

@MortenLinderud

[Chris Down @unixchris]

@erwanaliasr1 @__dxu Yeah, netconsole is a good answer and the emergency path tends to still work pretty well even in panic/other dire scenarios. If you don't want to have to flood other consoles as well, you can apply https://t.co/OkUY3Ng0nD which adds per-console loglevels and should be merged soon

[Meta Open Source @MetaOpenSource]

Signals make up the plumbing in Unix-like operating environments, but they come with a significant number of pitfalls as a project grows. Learn about some of the alternative methods and mitigations to use from @unixchris: https://t.co/rRvLDbGDvR

[Chris Down @unixchris]

@kernellogger Sadly the T14s Gen2 reliably still has interrupt spam on the i2c and connected Touchpad after resume from S3, so that still needs mem_sleep_default=shallow to avoid halving the battery life. I've been meaning to look into what's going on there but haven't had time yet.

[Chris Down @unixchris]

@MortenLinderud Congrats mate! With all of the hard work you've been putting in to drive the community forward especially it's very well deserved.

[Chris Down @unixchris]

@juliusvolz @lclarkmichalek @vorsprungbike @PrometheusIO Sorry, missed this. Senpai can actually look at any cgroup (eg. /sys/fs/cgroup/.../memory.pressure), and as you found there's globally scoped pressure information at /proc/pressure :-)

[Chris Down @unixchris]

@juliusvolz @lclarkmichalek @vorsprungbike @PrometheusIO It's certainly cumbersome to calculate this properly, but if you are serious about it, the only meaningful way to do it right now is with cgroup v2 and Senpai: https://t.co/jmuUSuY6nK

[Chris Down @unixchris]

@juliusvolz @lclarkmichalek @vorsprungbike @PrometheusIO The problem with MemAvailable is that its estimations are not uniformly too high or too low, and in some cases it can be very far off the mark in either direction (eg. when forward progress of the system is largely memory bound), but people take it almost at face value. (2/n)

[Chris Down @unixchris]

@rufusmairo Rahul isn't a kernel contributor, so that's why it doesn't link there, but it make sense to update TJ's link as well. I made it a bit more uniform in https://t.co/yV3ibiq2o4. Thanks!

[Chris Down @unixchris]

@rufusmairo Hi! This is from when Patchwork went: https://t.co/H6aCVtrJYW -- the link used to point to Patchwork. In the old days commit author search didn't work on the Linux repo on GitHub because it was too big, but it works now and Johannes' link was only updated after that :-) (1/2)

[Chris Down @unixchris]

@defineSayak It's a mistake in the commit message I sent for v4 (it should say cgroup_disable=memory, not cgroup_memory=disable). That said, that message is for the patch version and shouldn't be in the commit though, the tree maintainer should have taken it out. No idea why they didn't.