Conversation

Notices

1. Embed this notice
   Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 06:19:05 JST Wolf480pl

   If you're sshing to something that's connected over wifi

   and it feels as if you were ssh-ing to another continent... from a train

   do yourself a favor and, on the remote end, set this:

   iw dev wlan0 set power_save off

   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 06:19:04 JST Wolf480pl

     in reply to

     I totally didn't just lose 2 hours chasing ghosts, comparing kernel logs, changing wifi channels, and turning off devices that might cause interference, due to not knowing about this feature.

     Haelwenn /элвэн/ :triskell: likes this.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 19:11:18 JST Wolf480pl

     in reply to

     • Alexander Monakov

     @amonakov
     I should check tomorrow at work if ssh'ing to Japan is faster without wifi power-saving on the client.

     Anyway, what is certainly faster when I disable power-saving is ssh-ing *to the laptop* (or to the mini-pc)

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 19:11:18 JST 翠星石

     in reply to

     • Alexander Monakov
     @wolf480pl @amonakov Gotta love Wi-Fi cards that run proprietary software eh?
   • Embed this notice
     Alexander Monakov (amonakov@mastodon.gamedev.place)'s status on Monday, 06-Oct-2025 19:11:20 JST Alexander Monakov

     in reply to

     @wolf480pl yeah, evidently pinging the access point from the laptop (2-4ms) is not the same as pinging the laptop from the AP (1-200ms) unless I run 'ping -i 0.1 AP' on the laptop (or disable power_save). Which seems to imply the power_save timeout is about 100ms in my case, and I randomly get its extra latency when the response doesn't arrive in that time (e.g. ssh'ing to another continent).

   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 19:11:21 JST Wolf480pl

     in reply to

     • Alexander Monakov

     @amonakov
     ping from router to my laptop

     with powersave on:
     30-200ms (highly variable)

     with powersave off:
     1.5-4ms, usually below 2.1

     on a Lenovo mini-PC with QCA6174 (ath10k) I was getting up to 800ms with wifi power-saving on

   • Embed this notice
     Alexander Monakov (amonakov@mastodon.gamedev.place)'s status on Monday, 06-Oct-2025 19:11:22 JST Alexander Monakov

     in reply to

     @wolf480pl I think it shouldn't be that bad? On my old laptop power_save is disabled by default, as I recall, due to known issues, on the new laptop I never noticed any problems (and, as I just checked, it is enabled by default).

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 19:12:33 JST 翠星石

     in reply to

     • Alexander Monakov
     @wolf480pl @amonakov Atheros 802.11n mPCIe cards have a ping round trip latency of 2-15 ms with no configuration required.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 19:12:34 JST Wolf480pl

     in reply to

     • 翠星石
     • Alexander Monakov

     @Suiseiseki
     @amonakov
     do you think a wifi card that runs free software wouldn't have that feature?
     do you know of such a card?

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 19:14:23 JST 翠星石

     in reply to

     • 翠星石
     • Alexander Monakov
     @wolf480pl @amonakov AR9281, AR9382, AR5BXB112 & AR5B22 are all cards that work fine.
   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 19:16:19 JST 翠星石

     in reply to

     • 翠星石
     • Alexander Monakov
     @amonakov @wolf480pl None of those cards run any software (all ROM) - the only software is the driver and that's free software.
     
     As for usb, AR9271 works fine (free driver and the PCIe to USB bridge software is free).
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 19:49:10 JST Wolf480pl

     in reply to

     • 翠星石
     • Alexander Monakov

     @Suiseiseki
     @amonakov
     As in, if you let the computer with that card idle, not sending any traffic, and then ping it from the router / access point / another computer on the network, it will still be 2-15ms?

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 19:49:10 JST 翠星石

     in reply to

     • Alexander Monakov
     @wolf480pl @amonakov The datasheet for a similar card says it has "Dynamic MIMO Power Save"
     
     The actual power consumption is not listed, but it seems negligible.
     
     If I let the card idle, the ping is still 2-10ms from a GNUbooted router.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 19:49:11 JST Wolf480pl

     in reply to

     • 翠星石
     • Alexander Monakov

     @Suiseiseki
     @amonakov
     yeah, ok

     Do these cards not have a power-saving feature?

   • Embed this notice
     翠星石 (suiseiseki@freesoftwareextremist.com)'s status on Monday, 06-Oct-2025 20:19:25 JST 翠星石

     in reply to

     • Alexander Monakov
     @wolf480pl @amonakov >have you measured it?
     I don't have a mPCIe power measurer, so I can't, but I can see that there's no heatsink.
     
     Generally, if something doesn't get hot enough to need a heatsink and if you have 9 18650 cells (rather than a piddly pouch cell), the power consumption of that part isn't much of a concern.
     
     >then you should be able to find out how it works by reading the source code, right?
     Reading drivers/net/wireless/ath/ath9k/main.c there's the ath9k_enable_ps() function to enable powersave.
     
     if (!(ah->caps.hw_caps & ATH9K_HW_CAP_AUTOSLEEP)) {
     if ((ah->imask & ATH9K_INT_TIM_TIMER) == 0) {
     ah->imask |= ATH9K_INT_TIM_TIMER;
     ath9k_hw_set_interrupts(ah);
     }
     ath9k_hw_setrxabort(ah, 1);
     
     It seems that the driver goes and puts the card into powersave mode if packets haven't been transmitted in a timeout period.
     
     It looks like it's enabled other than in the case where having powersave enabled uses more power;
     /*
     * We just prepare to enable PS. We have to wait until our AP has
     * ACK'd our null data frame to disable RX otherwise we'll ignore
     * those ACKs and end up retransmitting the same null data frames.
     * IEEE80211_CONF_CHANGE_PS is only passed by mac80211 for STA mode.
     */
     
     I'm actually quite pleased with that driver, as it reads like it's actually the source code - don't see magic values and any tables contain what is quite clearly just data.
     
     There's the bad Linux habit of almost no comments, but there are at least comments describing things that you can't work out from reading the function like above.
   • Embed this notice
     Wolf480pl (wolf480pl@mstdn.io)'s status on Monday, 06-Oct-2025 20:19:26 JST Wolf480pl

     in reply to

     • 翠星石
     • Alexander Monakov

     @Suiseiseki
     @amonakov
     > seems negligible

     have you measured it?

     > similar card says it has Dynamic MIMO Power save

     if this one does too, then you should be able to find out how it works by reading the source code, right?