I'm wondering why the 8.09 RC1 announcement was not made here. It was made on the "forums" at http://forum.openwrt.org/viewtopic.php?pid=75990. Do we really have to subject ourselves to the far inferior interface of silly "forums" in order to be up to date on this stuff?
Mailing lists are where the heart of the Internet is, not
johnny-come-lately web forums. It's disappointing to see the OpenWrt
team treat forums with such preference and the mailing lists with such
prejudice.
In any case, the release announcement made on the forums is as such:
It is well known fact that it takes time to build quality, and sometimes
more time than expected. The OpenWrt developers would like to announce
the availability of OpenWrt Kamikaze 8.09 RC1.
You can find it at http://downloads.openwrt.org/kamikaze/8.09_RC1
Release Candidate 1 for Kamikaze 8.09
New changes/features include:
* Busybox upgraded to 1.11.2
* Kernel upgrades for all targets
* uClibc upgraded to 0.9.29
* Firewall rewrite, fully modular netfilter/iptables support
- Syntax of /etc/config/firewall converted to UCI
* ipkg package manager replaced with opkg
- package lists moved to ramdisk, lists are not retained
between reboots.
- `opkg update' must be run after each reboot before new
packages are installed
* Broadcom 47xx running reliably with the new Kernel, including b43
wireless driver
- WEP, WDS and multi-SSID not yet supported
- AP mode and STA mode are supported with WPA, WPA2, or no
encryption
* IMQ and Traffic shaping fixed with newer kernels (2.6.25+)
* Sysupgrade for x86, broadcom and adm5120
* New web interface (LuCI, Lua Configuration Interface)
* New HAL and driver for Atheros based WiFi cards
* Attention towards the integration of security updates
* Build system cleanup
* Package maintaining and updates between releases
Supported targets:
* AMD Alchemy (MeshCube)
* Atheros AR531x/231x based routers
* Atheros AR71xx based routers
* Atmel AVR32
* Broadcom 47xx based routers (Linux 2.4 and 2.6)
* Infineon/ADMtek ADM5120
* Infineon Danube/TwinPass
* Intel IXP42x/43x
* MagicBox
* MikroTik RouterBoard 532
* RDC SoC
* Texas Instruments AR7
* User-Mode Linux (source code release only)
* x86 based boards
Known Issues:
* New uClibc version breaks ABI compatibility with older releases.
* Wireless no longer supported on older Broadcom devices
(bcm4710/bcm94702)
- Support was dropped in upstream Broadcom wl driver and b43
- b43legacy (which adds support for such hardware) might be added
in a future release
The OpenWrt Team
I think it's really bad to be advertising the support of b43 on Broadcom
47xx based systems given that it works so badly in both my own
experience and the experiences of others on the above announcement
forum.
The throughput is dismal due to the amount of softirq:
Mem: 14404K used, 15524K free, 0K shrd, 1008K buff, 6800K cached
CPU: 0% usr 0% sys 0% nice 0% idle 0% io 0% irq 99% softirq
Load average: 1.21 0.33 0.31
PID PPID USER STAT VSZ %MEM %CPU COMMAND
3 2 root RW< 0 0% 57% [ksoftirqd/0]
This is without any encryption (i.e. WEP, WPA, etc.), just plain,
clear-text wireless. Resulting throughput is 1.4MB/s on an 802.11g
connection showing a link quality of 100/100. So out of a theoretical
54Mb/s actual throughput with all that softirq activity is only 10Mb/s
or 20% with the CPU of the router pegged at full.
As well, the wireless connection only works for a while after which it
just stops passing any traffic and then you have to issue
an /etc/init.d/network restart in order to have it working again.
Many people are reporting these results, not just me.
The user that posted the release also said subsequently in the same
forum thread (http://forum.openwrt.org/viewtopic.php?pid=76039#p76039):
broadcom 47xx needs to be migrated to 2.6.27 before we can fully
take advantage of the new mac80211 stuff.
this won't happen in 8.09, but probably in the release after
that.
So it's known that b43 and broadcom 47xx are known to not work until at
least 2.6.27 yet it's still being included in the 8.09 release
announcements?
Maybe a bit of work on why the softirq is so high and solving that could
be done before 8.09 is final and that would be at least a big step
closer to an actual working b43 in 8.09.
IMHO, if you wanted to keep the b43/bcm47xx in the release announcement,
this softirq problem has to be solved at minimum. Maybe whatever is
causing that is also causing the wifi interface to just stop working.
b.
signature.asc
Description: This is a digitally signed message part
_______________________________________________ openwrt-devel mailing list [email protected] http://lists.openwrt.org/cgi-bin/mailman/listinfo/openwrt-devel
