Bug 81571

Summary:	iwlwifi: Lack of TX AMPDU severely impacts performance
Product:	Drivers	Reporter:	haarp (main.haarp)
Component:	network-wireless	Assignee:	drivers_network-wireless (drivers_network-wireless)
Status:	CLOSED WILL_NOT_FIX
Severity:	normal	CC:	ilw, linville, pobrn
Priority:	P1
Hardware:	All
OS:	Linux
Kernel Version:	3.12	Subsystem:
Regression:	No	Bisected commit-id:

Description haarp 2014-08-02 17:37:04 UTC

Hello,

I've recently been battling a performance regression on my Wifi, until I figured out that it was a driver issue. To be precise, it's a regression introduced by this commit:

iwlwifi: disable TX AMPDU by default for iwldvm, 205e2210daa975d92ace485a65a31ccc4077fe1a

Predictably, setting 11n_disable=8 restores performance. Without it, my throughput falls by almost an order of magnitude to 802.11g levels!

This can't be intended. Even as a trade-off for increased stability (which is what I understand this commit is for), this seems like a bad default.

Additionally, it's very hard to diagnose (took me many hours).

Can there be a better way?

Comment 1 Emmanuel Grumbach 2014-08-04 17:11:45 UTC

I can't see another way.
I can't really debug the FW issues and most people care more for reliability than performance.

Comment 2 Barnabás Pőcze 2024-08-13 10:19:33 UTC

For the record, TX AMPDU has been enabled by default for all devices since linux 5.1 due to the inadvertent reversion of commit https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=205e2210daa975d92ace485a65a31ccc4077fe1a in commit https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=e78da25e19c2dd31018c358d0717e4f6cfe317c8.