On Thu, Oct 29, 2009 at 12:30 PM, Dirk Behme <dirk.behme@xxxxxxxxxxxxxx> wrote: > Dirk Behme wrote: > > I promised to come back with test results. As mentioned in this thread > already, I can't test on my own yet, instead Steve (many, many thanks!) > tests it on Overo air. Overo air uses Marvell's 88W8686 connected to MMC > port 2 in 4 bit configuration. > > First, my 'revised' version of Phani's patch doesn't work at all. There seem > to be some tricks in Phani's original patch we (I?) missed in above review. > I tried to incorporate all review comments, result is a non working patch. > As i mentioned earlier, the code changes you made will result in loss of sdio interrupts which will effect the host driver functionality. > Second, Steve tested Phani's original patch, too (with manually fixed > wrapped lines). Let me copy his comments: > > -- cut -- > this is looking better! > > libertas_sdio: Libertas SDIO driver > libertas_sdio: Copyright Pierre Ossman > libertas_sdio mmc1:0001:1: firmware: requesting sd8686_helper.bin > libertas_sdio mmc1:0001:1: firmware: requesting sd8686.bin > libertas: 00:19:88:20:fa:23, fw 9.70.3p24, cap 0x00000303 > libertas: eth1: Marvell WLAN 802.11 adapter > udev: renamed network interface eth1 to wlan0 > > (with the other patches we always had timeouts and errors) > > But performance is really bad (and flakey!) > > When it does work, I get 5.88K/s as compared to 120K/s with the unpatched > code. It also dies fairly quickly. > > No error messages, so I am really not sure what is going on! > -- cut -- Iam not sure about what might be causing such a drastically low performance with steve's testing. The initial SDIO interrupts occur once the firmware gets loaded properly. From the log it looks the firmware is loading successfully and interface is up. In which case it means the interrupt mechanism is proper as it should be. So the reduction in performance cant be that worse when handling SDIO interrupts with irq support. Even with the SDIO interrupt mechanism in polling mode i observed throughput in the range of 1 to 2 Mbps. Given below are details of my test environment: Target boards: OMAP3530EVM and AM3517EVM Linux kernel version: 2.6.31-rc7. The host operating frequency is 25MHz and in sdio 4bit mode of operation. Libertas firmware<9.70.3.p24-26409.P45-GPL> from marvell <http://www.marvell.com/drivers/driverDisplay.do?driverId=203> Mode: Infrastructure mode Channel 11 Phy Mode: G Observed throughput [using ttcp utility]: 13Mbps (Downlink), 10.5 Mbps(Uplink) The throughput measurements are taken with Marvell 88W8686 card on one end connected to an access point and on the other end the access point connected to a test system through ethernet. The WiFi performance would enhance under proper test conditions. Regards, Phani > Any ideas? > > Thanks to Phani and Steve, best regards > > Dirk > -- > To unsubscribe from this list: send the line "unsubscribe linux-omap" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > > -- To unsubscribe from this list: send the line "unsubscribe linux-omap" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
