Hello! I wanted to measure speeds in the softmac driver at all rates, but it looks like I have found something more interesting in process.
The station is a BCM4312 in Dell Latitude with an antenna embedded into the laptop. The AP PCI card with Atheros 5212 running current MadWifi in 802.11g mode. Both systems sit on the same desk, about one meter apart. I'm using Michael's git repository with Larry's last patch for high rates. The speed is measured by iperf 2.0.2 in TCP mode (which is the default). rate: iperf reports: 1 928 Kbits/sec 2 1.69 Mbits/sec 5.5 3.92 Mbits/sec 6 4.78 Mbits/sec 9 6.55 Mbits/sec 11 5.97 Mbits/sec 12 8.18 Mbits/sec 18 10.4 Mbits/sec 24 13.0 Mbits/sec 36 15.1 Mbits/sec 48 187 Kbits/sec 54 N/A Testing at rate 54 caused this to appear on the console about 100 times in a row: bcm43xx: ASSERTION FAILED (!ring->suspended) at: /home/proski/src/linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_dma.c:71:request_slot() My attempts to test the UDP mode also lead to such messages. At some point, following was reported to the console: bcm43xx: Controller RESET (TX timeout) ... bcm43xx: IRQ_READY timeout bcm43xx: core_up for active 802.11 core failed (-19) bcm43xx: Controller restart failed bcm43xx: ASSERTION FAILED (!err) at: /home/proski/src/linux-2.6/drivers/net/wireless/bcm43xx/bcm43xx_main .c:4057:bcm43xx_net_stop() After that the system became very slow and I had to reboot it. Next time I got this during UDP testing: warning: many lost ticks. Your time source seems to be instable or some driver is hogging interupts rip __do_softirq+0x4d/0xd0 Falling back to HPET I could check the connection at 48M in USB mode with low 5Mbsp bandwidth limit, and the packet loss is minimal (10 out of 3211 packets are lost), so the speed drop must be explained by something else. There is a clearly a stability problem at the high rates, and perhaps there is also another problem that limits the throughput. 802.11g modes are faster relative to the rate, e.g. 9 is faster than 11 and 12 is much faster than 11. 36Mbps seems to be the fastest for TCP, although I would probably use 12Mbps to be safe. -- Regards, Pavel Roskin _______________________________________________ Bcm43xx-dev mailing list [email protected] https://lists.berlios.de/mailman/listinfo/bcm43xx-dev
