Hi, On 10/13/2015 01:22 PM, Felipe Balbi wrote:
Yegor Yefremov <yegorslists@xxxxxxxxxxxxxx> writes:On Mon, Oct 12, 2015 at 11:34 AM, Yegor Yefremov <yegorslists@xxxxxxxxxxxxxx> wrote:We have a problem, when using more than 12 FTDI ports. Kernels tried: 3.18.1, 4.2.3 and 4.3-rc5. SoC am335x 600MHz Below the USB topology: # lsusb -t /: Bus 02.Port 1: Dev 1, Class=root_hub, Driver=musb-hdrc/1p, 480M /: Bus 01.Port 1: Dev 1, Class=root_hub, Driver=musb-hdrc/1p, 480M |__ Port 1: Dev 2, If 0, Class=, Driver=hub/4p, 480M |__ Port 1: Dev 9, If 0, Class=, Driver=hub/4p, 480M |__ Port 1: Dev 10, If 0, Class=, Driver=ftdi_sio, 12M |__ Port 2: Dev 11, If 0, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 11, If 1, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 11, If 2, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 11, If 3, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 12, If 0, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 12, If 1, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 12, If 2, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 12, If 3, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 4, If 0, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 4, If 1, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 4, If 2, Class=, Driver=ftdi_sio, 480M |__ Port 2: Dev 4, If 3, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 7, If 0, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 7, If 1, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 7, If 2, Class=, Driver=ftdi_sio, 480M |__ Port 3: Dev 7, If 3, Class=, Driver=ftdi_sio, 480M
How many EPs does each FTDI device require? at least one INT EP, right? If I read it right, the topology above has 2 hubs, and 16 high-speed FTDI and 1 full-speed FTDI. So it requires at least 18 high-speed INT EPs. MUSB driver only has 11 high-speed EPs for mode-4 which is the EP configuration used by default. I am wondering how those devices got enumerated properly.
When using 12 ports and performing serial test (a pair of ports is connected via null-modem cable and a rather short string ca. 90 characters will be sent alternating at 1200 and 115200b/s, testing scripts are written in Python and running as own processes per a pair of ports) there are no timeouts, i.e. all sent characters will be received. As soon as I open ports 13 and 14 I start to get arbitrary timeouts (from test software point of view) on all ports. In order to check, if ftdi_sio has primary to do with this issue, I've performed the same test on a PC and PandaBoard Rev. A2 (EHCI port) and there were no issues with 16 ports. So it seems to have something to do with am335x + musb + number of end points. Any idea? Let me know, if you need our test script.From time to time I get following warnings (4.3.0-rc5): musb_host_rx 1915: RX1 dma busy, csr 2020 musb_host_rx 1915: RX4 dma busy, csr 2020 musb_host_rx 1915: RX7 dma busy, csr 2220 musb_host_rx 1915: RX1 dma busy, csr 2020 Though they are not timely related to serial test timeouts.yeah, I don't think MUSB can easily handle that. IIRC, endpoint scheduling in MUSB is rather bad. While we have enough endpoints to handle this case, you might be running into some IP (or driver) issues. Bin, have you ever tested this many serial devices on AM335x ?
No, I never tested this many devices. It could be resource limitation, but the log above is about CPPI, so I recommend to test with CPPI disabled to isolate if this is MUSB issue or CPPI.
Regards, -Bin.
-- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
