https://bugs.kde.org/show_bug.cgi?id=502411

Clément Dieperink <[email protected]> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |[email protected]

--- Comment #3 from Clément Dieperink <[email protected]> ---
I am also encountering this issue. My network card is an Aquantia AQC113CS, and
I am running Fedora 44.

>From my understanding, it is related to the atlantic driver, which updates
statistics every second. However, ksystemstats backend reads network statistics
(rx/tx bytes count) every 0.5 seconds. Meaning, that it will read the same
value twice, before reading the new value. So, the rate alternates between 0
B/s and double the actual value.

Running `watch -n 0.1 cat /sys/class/net/<network
interface>/statistics/rx_bytes`, to read the stats every 100ms, allows to see
the difference. On the same setup, with the Aquantia card, the stats update
only every second, but with a Intel card, they update every 100ms.

Changing the ksystemstats NetworkManager backend update rate to be 1000 fixes
the issue (this line:
https://invent.kde.org/plasma/ksystemstats/-/blob/91de7278bb88bdbc5ce57371afcaf23ac65345e0/plugins/network/NetworkManagerBackend.cpp#L20).
However, this was set to 500 to avoid reporting the same value for two
"ticks"(see:
https://invent.kde.org/plasma/ksystemstats/-/merge_requests/6/diffs?commit_id=1b8d4ee95ddb1797b315bd3470efe1b7a97fa346)

I'm not sure what would be the best fix, it was more of a curiosity
investigation.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to