On 12/27/2017 2:37 AM, Marc Zyngier wrote: > On Tue, 26 Dec 2017 21:57:58 +0000, > Troy Kisky wrote: >> >> On 12/26/2017 1:52 PM, Troy Kisky wrote: >>> On 12/25/2017 2:10 AM, Marc Zyngier wrote: >>>> On Sun, 24 Dec 2017 23:01:33 +0000, >>>> Troy Kisky wrote: >>>>> >>>>> commit 8466489ef5ba48272ba4fa4ea9f8f403306de4c7 >>>>> Author: Marc Zyngier <marc.zyngier@xxxxxxx> >>>>> Date: Tue Aug 1 20:11:08 2017 -0500 >>>>> >>>>> xhci: Reset Renesas uPD72020x USB controller for 32-bit DMA issue >>>>> ... >>>>> ... >>>>> ... >>>>> --- a/drivers/usb/host/xhci-pci.c >>>>> +++ b/drivers/usb/host/xhci-pci.c >>>>> @@ -284,6 +284,13 @@ static int xhci_pci_probe(struct pci_dev *dev, const struct pci_device_id *id) >>>>> >>>>> driver = (struct hc_driver *)id->driver_data; >>>>> >>>>> + /* For some HW implementation, a XHCI reset is just not enough... */ >>>>> + if (usb_xhci_needs_pci_reset(dev)) { >>>>> + dev_info(&dev->dev, "Resetting\n"); >>>>> + if (pci_reset_function_locked(dev)) >>>>> + dev_warn(&dev->dev, "Reset failed"); >>>>> + } >>>>> + >>>>> ________ >>>>> >>>>> >>>>> This pci_reset_function_locked call, causes my i.mx6qp processor to >>>>> hang. It no longer responds to the MAGIC_SYSRQ key (break on serial >>>>> port). >>>>> >>>>> >>>>> If I comment it out, things return to normal when testing on >>>>> Linux-next (20171222). >>>>> >>>>> If you need more info, let me know. >>>> >>>> Well, for a start: >>>> >>>> - Does it fail with 4.13 or 4.14 too? >>> >>> v4.13-rc4 - good >>> v4.13-rc5 - fail >>> v4.13 - fail >>> v4.14 - fail >>> >>> >>> >>>> - Can you work out where it is locking up exactly (going slightly >>>> deeper than the pci_reset_function_locked() call)? >>> >>> With this patch >>> diff --git a/drivers/usb/host/xhci.c b/drivers/usb/host/xhci.c >>> index 05104bd2b611..8782945a9f70 100644 >>> --- a/drivers/usb/host/xhci.c >>> +++ b/drivers/usb/host/xhci.c >>> @@ -4808,8 +4808,10 @@ int xhci_gen_setup(struct usb_hcd *hcd, xhci_get_quirks_t get_quirks) >>> >>> mutex_init(&xhci->mutex); >>> xhci->cap_regs = hcd->regs; >>> + pr_err("%s: 1\n", __func__); >>> xhci->op_regs = hcd->regs + >>> HC_LENGTH(readl(&xhci->cap_regs->hc_capbase)); >>> + pr_err("%s: 2\n", __func__); >>> xhci->run_regs = hcd->regs + >>> (readl(&xhci->cap_regs->run_regs_off) & RTSOFF_MASK); >>> /* Cache read-only capability registers */ >>> _____________________ >>> >>> The last thing printed is >>> [ OK ] Reached target System Initialization. >>> [ OK ] Listening on D-Bus System Message Bus Socket. >>> [ OK ] Reached target Sockets. >>> [ OK ] Started Daily Cleanup of Temporary Directories. >>> [ 15.770319] xhci_gen_setup: 1 > > OK, so it looks like the CPU is blocked reading from the > device. Almost feels like the device is not clocked (that's about the > only reason why the CPU would be stuck on such a read). > >>> >>> >>>> - Do you get the symptom at boot time? >>> >>> Boot >>> >>>> On resume? Is the USB driver >>>> compiled as a module or built-in? >>> >>> INSTALL drivers/usb/host/xhci-hcd.ko >>> INSTALL drivers/usb/host/xhci-pci.ko >>> INSTALL drivers/usb/host/xhci-renesas.ko >>> >>>> - What is your exact platform (something a bit more precise than just >>>> i.MX6...)? >>> >>> It is a custom board not in mainline. It does need to load firmware. >>> >>> I tried moving pci_reset_function_locked before renesas_check_if_fw_dl_is_needed >>> but that did not help. >>> >> >> >> >> Which I should say I also have this patch >> >> >> commit 01ac5cfd811088b5fe8c97f28f96086e30393c66 >> Author: Christian Lamparter <chunkeey@xxxxxxxxx> >> Date: Wed Jun 8 00:14:57 2016 +0200 >> >> usb: xhci: handle uPD720201 and uPD720202 w/o ROM >> >> This patch adds a firmware check for the uPD720201K8-711-BAC-A >> and uPD720202K8-711-BAA-A variant. Both of these chips are listed >> in Renesas' R19UH0078EJ0500 Rev.5.00 "User's Manual: Hardware" as >> devices which need a firmware in order to work as they do not have >> support to load the firmware from an external ROM. >> >> __________ >> >> Sorry for not saying earlier. > > That's a very interesting piece of information. Looking at that > patch[1], it doesn't provide any FW loader, and simply detects whether > the firmware is correctly present and running. > > What is probably happening is that your bootloader loads the XHCI > firmware, the driver resets the XHCI device, and the firmware is > lost. Assuming my hunch is right, what would be required here is for > the firmware to be loaded after the reset, but that'd require writing > a firmware loader for xhci-pci (probably not a huge undertaking) > > Could you please investigate whether my hunch is correct or not? > > Thanks, > > M. > > [1]: https://patchwork.kernel.org/patch/9162693/ > Hi Marc It already had a firmware downloader, and I tried reset both before and after that. No change. So, now I've tried the same thing with 0 patches to linux-next. A nitrogen6_max board with a Renesas USB3.0 PCIe card that has a firmware ROM, and so does not need the extra patch. 01:00.0 USB controller: Renesas Technology Corp. uPD720201 USB 3.0 Host Controller (rev 03) It behaves identically to the other board. So, the firmware issue is a red herring. [ OK ] Found device /dev/ttymxc0. [ OK ] Found device /dev/ttymxc1. [ 11.916718] xhci_gen_setup: 1 BR Troy -- 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