* Felipe Balbi <balbi@xxxxxxxxxx> [161208 09:52]: > > Hi, > > Tony Lindgren <tony@xxxxxxxxxxx> writes: > > * Felipe Balbi <balbi@xxxxxxxxxx> [161208 01:45]: > >> > >> Hi, > >> > >> Tony Lindgren <tony@xxxxxxxxxxx> writes: > >> > Somehow starting with v4.9-rc7 there have been imprecise > >> > >> There's nothing touching dwc3 since v4.9-rc5. > > > > Right, nothing obvious has changed. I think it's just a slight timing > > change in the code that started triggering it. > > could be > > >> > external aborts on omap5-uevm dwc3 controller. I have not been > >> > able to bisect what exactly triggered this as it does not always > >> > happen. It seems that something changed with probing that > >> > now exposes the issue: > >> > > >> > Unhandled fault: imprecise external abort (0x1406) at 0x00000000 > >> > >> hmmm, clock disabled... dwc3-omap shouldn't have pm runtime at all. > > > > It does for the interconnect target module clkctrl register via PM > > runtime. That's the "usb_otg_ss" module. > > but that's all hidden in omap_device.c, we don't touch it from driver > perspective. The call to pm_runtime_get_sync() in dwc3_omap_probe() will use it. Is there also some dwc3 internal clock? If we assume the usb_otg_ss module is properly enabled it could be some dwc3 internal clock not enabled? We do have a srst_udelay needed for enabling musb controller for some SoCs, I'll check if that's the case here. > >> > ... > >> > PC is at dwc3_omap_interrupt_thread+0x20/0x80 > >> > LR is at irq_thread_fn+0x1c/0x54 > >> > ... > >> > [<c0989fa8>] (dwc3_omap_interrupt_thread) from [<c038a4d8>] > >> > (irq_thread_fn+0x1c/0x54) > >> > [<c038a4d8>] (irq_thread_fn) from [<c038a7b0>] (irq_thread+0x12c/0x1f0) > >> > [<c038a7b0>] (irq_thread) from [<c035d6f0>] (kthread+0xdc/0xf4) > >> > [<c035d6f0>] (kthread) from [<c0307d78>] (ret_from_fork+0x14/0x3c) > >> > ... > >> > > >> > Unable to handle kernel paging request at virtual address ffffffec > >> > ... > >> > Internal error: Oops: 37 [#2] SMP ARM > >> > PC is at kthread_data+0x4/0xc > >> > LR is at irq_thread_dtor+0x28/0xd0 > >> > ... > >> > [<c035e0b4>] (kthread_data) from [<c038a5dc>] (irq_thread_dtor+0x28/0xd0) > >> > [<c038a5dc>] (irq_thread_dtor) from [<c035bef0>] (task_work_run+0xb8/0xdc) > >> > [<c035bef0>] (task_work_run) from [<c03448d4>] (do_exit+0x314/0xa20) > >> > [<c03448d4>] (do_exit) from [<c030bea8>] (die+0x410/0x474) > >> > [<c030bea8>] (die) from [<c0301350>] (do_DataAbort+0xb4/0xb8) > >> > [<c0301350>] (do_DataAbort) from [<c030c578>] (__dabt_svc+0x58/0x80) > >> > Exception stack(0xee777ec8 to 0xee777f10) > >> > 7ec0: 0000014d ee6e6f10 00000034 fc020034 ee6e6f10 ee1eec00 > >> > 7ee0: 00000000 00000000 ee6ec640 c038a4bc 00000000 00000000 00000000 ee777f18 > >> > 7f00: c038a4d8 c0989fa8 60000013 ffffffff > >> > [<c030c578>] (__dabt_svc) from [<c0989fa8>] (dwc3_omap_interrupt_thread+0x20/0x80) > >> > [<c0989fa8>] (dwc3_omap_interrupt_thread) from [<c038a4d8>] (irq_thread_fn+0x1c/0x54) > >> > [<c038a4d8>] (irq_thread_fn) from [<c038a7b0>] (irq_thread+0x12c/0x1f0) > >> > [<c038a7b0>] (irq_thread) from [<c035d6f0>] (kthread+0xdc/0xf4) > >> > [<c035d6f0>] (kthread) from [<c0307d78>] (ret_from_fork+0x14/0x3c) > >> > > >> > Fix the issue by making sure the dwc3 interrupts are disabled > >> > before we call devm_request_threaded_irq(). > >> > >> that should already be the case. Are you sure that register isn't zero > >> in probe? > > > > Looks like irq0_status = 0 and irqmisc_status = 0x2121. Also just > > clearing irqmisc with dwc3_omap_write_irqmisc_clr(omap, 0xffffffff) > > stops the issue from happening. > > > > There is some deferred probing happening but irqmisc is always 0x2121. > > some spurious irq status from ROM code, perhaps? Are you not resetting > usb_otg_ss hwmod? We shouldn't have any pending interrupts when we get > to probe(), if we do have them, then we need an erratum for it. I have u-boot use it to download the kernel and that is probably using the ROM code USB support. > You could try making usb_otg_ss is reset by hwmod. It would also be good > to check other OMAP-ish devices sporting dwc3 (DRA7x, AM437x, AM57x) if > they have the same behavior. It should be reset by default by hwmod. > Adding Nishanth to the loop too for reference. Regards, Tony -- 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
