On Mon, Jul 06, 2015 at 06:58:52PM +0200, Paul Jones wrote:
> Hi,
>
> On 05 Jul 2015, at 20:43, Felipe Balbi <balbi@xxxxxx> wrote:
>
> > Hi,
> >
> > On Sun, Jul 05, 2015 at 06:56:29PM +0200, Paul Jones wrote:
> >> Ricardo,
> >>
> >> I’m trying to get the 3380 to work in UAS mode on the 4.0.1 and/or 4.1.0-rc8 kernels.
> >> I’m using the following script (derived from Sebastian’s post):
> >> modprobe tcm_usb_gadget
> >> mount -t configfs none /sys/kernel/config
> >> CONFIGFS=/sys/kernel/config/
> >> TARGET=$CONFIGFS/target/core/
> >> FABRIC=$CONFIGFS/target/usb_gadget/
> >>
> >> mkdir -p $TARGET/rd_mcp_0/ramdisk
> >> echo -n rd_pages=32768 > $TARGET/rd_mcp_0/ramdisk/control
> >> echo -n 1 > $TARGET/rd_mcp_0/ramdisk/enable
> >>
> >> mkdir -p $FABRIC/naa.6001405c3214b06a/tpgt_1
> >> mkdir $FABRIC/naa.6001405c3214b06a/tpgt_1/lun/lun_0
> >> echo -n naa.6001405c3214b06b > $FABRIC/naa.6001405c3214b06a/tpgt_1/nexus
> >>
> >> ln -sv $TARGET/rd_mcp_0/ramdisk $FABRIC/naa.6001405c3214b06a/tpgt_1/lun/lun_0/virtual_scsi_port
> >> echo 1 > $FABRIC/naa.6001405c3214b06a/tpgt_1/enable
> >>
> >> On either of the kernels, once I setup UAS I get an error from tcm_usb_gadget: "Can't claim all required eps”.
> >> This seems to be due to the fact that max_streams = 0 on every EP for the 3380.
> >
> > yeah, that needs to be initialized.
> >
> >> If I patch around that:
> >> diff --git a/drivers/usb/gadget/epautoconf.c b/drivers/usb/gadget/epautoconf.c
> >> index 0567cca..94ade78 100644
> >> --- a/drivers/usb/gadget/epautoconf.c
> >> +++ b/drivers/usb/gadget/epautoconf.c
> >> @@ -117,8 +117,10 @@ ep_matches (
> >> if (usb_endpoint_xfer_bulk(desc)) {
> >> if (ep_comp && gadget->max_speed >= USB_SPEED_SUPER) {
> >> num_req_streams = ep_comp->bmAttributes & 0x1f;
> >> - if (num_req_streams > ep->max_streams)
> >> + printk("%s: requested streams %d < max streams %d",ep->name, num_req_streams, ep->max_streams);
> >> +/* if (num_req_streams > ep->max_streams)
> >> return 0;
> >> +*/
> >
> > no, you should make sure that max_streams is initialized corrent when
> > runing on 3380.
>
> The above temporary patch was just to see if more is needed besides setting max_streams and to help Ricardo to reproduce the issue.
>
> >
> >> }
> >>
> >> }
> >>
> >> then I can at least get UAS to load, but as soon as I connect a linux box to the 3380 I then get errors:
> >> [ 50.954080] ep1in: requested streams 4 < max streams 0
> >> [ 50.954089] ep2out: requested streams 4 < max streams 0
> >> [ 50.954097] ep3in: requested streams 4 < max streams 0
> >> [ 50.954105] ep4out: requested streams 0 < max streams 0
> >> [ 50.954116] g_target gadget: g_target ready
> >> [ 50.954129] net2280 0000:05:00.0: Operate Defect 7374 workaround soft this time
> >> [ 50.954130] net2280 0000:05:00.0: It will operate on cold-reboot and SS connect
> >> [ 51.198283] net2280 0000:05:00.0: INFO: Defect 7374 workaround waited about
> >> [ 51.198283] 0uSec for Control Read Data Phase ACK
> >> [ 51.199366] g_target gadget: super-speed config #1: Linux Target
> >> [ 51.199394] Using the BOT protocol
> >> [ 51.199903] Wrong signature on CBW
> >> [ 51.199913] bot_cmd_complete(313): -22
> >> [ 51.199955] Using the UAS protocol
> >> [ 72.789041] Unsupported type 0
> >> [ 72.789063] g_target gadget: super-speed config #1: Linux Target
> >> [ 72.789114] Using the BOT protocol
> >> [ 72.789406] Wrong signature on CBW
> >> [ 72.789415] bot_cmd_complete(313): -22
> >> [ 72.789456] Using the UAS protocol
> >> [ 72.921156] Unsupported type 0
> >> [ 72.921174] g_target gadget: super-speed config #1: Linux Target
> >> [ 72.921222] Using the BOT protocol
> >> [ 72.921514] Wrong signature on CBW
> >> [ 72.921523] bot_cmd_complete(313): -22
> >> [ 72.921563] Using the UAS protocol
> >>
> >> On the connected Linux box (running 4.1.0-rc8 as well), I get:
> >> [ 4563.476084] usb 2-3: new SuperSpeed USB device number 5 using xhci_hcd
> >> [ 4563.492900] usb 2-3: New USB device found, idVendor=0525, idProduct=a4a5
> >> [ 4563.492903] usb 2-3: New USB device strings: Mfr=1, Product=2, SerialNumber=3
> >> [ 4563.492904] usb 2-3: Product: Target Product
> >> [ 4563.492906] usb 2-3: Manufacturer: Target Manufactor
> >> [ 4563.492907] usb 2-3: SerialNumber: 000000000001
> >> [ 4563.494193] scsi host9: uas
> >> [ 4563.494298] xhci_hcd 0000:00:14.0: ERROR Transfer event for disabled endpoint or incorrect stream ring
> >> [ 4563.494324] xhci_hcd 0000:00:14.0: @00000000d43e2500 d3921000 00000000 0a000000 04078000
> >> [ 4563.494346] xhci_hcd 0000:00:14.0: ERROR Transfer event for disabled endpoint or incorrect stream ring
> >> [ 4563.494371] xhci_hcd 0000:00:14.0: @00000000d43e2510 d3921400 00000000 0a000000 04038000
> >> [ 4584.936226] scsi 9:0:0:0: tag#0 uas_eh_abort_handler 0 uas-tag 1 inflight: CMD IN
> >> [ 4584.936230] scsi 9:0:0:0: tag#0 CDB: Inquiry 12 00 00 00 24 00
> >> [ 4584.936286] scsi host9: uas_eh_bus_reset_handler start
> >> [ 4585.048559] usb 2-3: reset SuperSpeed USB device number 5 using xhci_hcd
> >> [ 4585.065687] scsi host9: uas_eh_bus_reset_handler success
> >> [ 4585.065707] xhci_hcd 0000:00:14.0: ERROR Transfer event for disabled endpoint or incorrect stream ring
> >> [ 4585.065736] xhci_hcd 0000:00:14.0: @00000000d43e2650 d3921800 00000000 0a000000 04078000
> >> [ 4585.068219] scsi 9:0:0:0: tag#0 uas_eh_abort_handler 0 uas-tag 1 inflight: CMD
> >> [ 4585.068223] scsi 9:0:0:0: tag#0 CDB: Test Unit Ready 00 00 00 00 00 00
> >> [ 4585.068224] scsi host9: uas_eh_bus_reset_handler start
> >> [ 4585.180561] usb 2-3: reset SuperSpeed USB device number 5 using xhci_hcd
> >> [ 4585.197711] scsi host9: uas_eh_bus_reset_handler success
> >> [ 4585.197717] scsi 9:0:0:0: Device offlined - not ready after error recovery
> >>
> >> I presume that the net2280 driver should set max_stream for each EP
> >> when in enhanced mode (3380) but I haven’t been able to figure out yet
> >> where to do that.
> >> I am starting to get myself familiarised with the code so I can help
> >> out and provide some patches.
> >
> > Have a look at dwc3, bdc and dummy_hcd drivers.
> >
> >> Sebastian: any ideas what else might be wrong that could cause the CBW
> >> signature to be wrong?
> >
> > yeah, you're lying to your host. You're basically accepting stream
> > transfers without initializing any streams in HW. When XHCI tries to
> > move that in a stream, your 3380 controller complains because it hasn't
> > been programed to do any stream transfers.
> I presumed (maybe incorrectly) that tcm_usb_gadget is setting
> up/handling the streams support?
it really can't. Streams is a function of the underlying HW. Some HW
support 1 streams and some support 16.
> Does that mean that currently none of the drivers in
> drivers/usb/gadget/udc support UAS (besides dummy_hcd)?
dwc3 and bcd do.
--
balbi
Attachment:
signature.asc
Description: Digital signature
