[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20210614073942.GB25299@jackp-linux.qualcomm.com>
Date: Mon, 14 Jun 2021 00:39:42 -0700
From: Jack Pham <jackp@...eaurora.org>
To: Felipe Balbi <balbi@...nel.org>
Cc: Wesley Cheng <wcheng@...eaurora.org>, gregkh@...uxfoundation.org,
linux-usb@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] usb: dwc3: gadget: Disable gadget IRQ during pullup
disable
On Fri, Jun 11, 2021 at 10:23:43AM +0300, Felipe Balbi wrote:
> Wesley Cheng <wcheng@...eaurora.org> writes:
<snip>
> > Hi Felipe,
> >
> > Let me share the ftrace snippets as well:
> >
> > USB FTRACE:
> >
> > <idle>-0 [002] 304.567900: dwc3_gadget_ep_cmd name=ep0out cmd=1030 param0=0 param1=4026523648 param2=0 cmd_status=0
> > <idle>-0 [004] 304.583225: dwc3_gadget_ep_cmd name=ep1out cmd=134152 param0=0 param1=0 param2=0 cmd_status=0
> >
> > //We don't have any logging in pullup disable, but this is when that
> > occurs, as we start to see -ESHUTDOWN statuses due to stop active
> > transfer.
>
> argh, maybe we should add :-)
>
> I noticed that the logs here look different. Did you modify dwc3 trace
> events or is it a special tracer (just curious)? Anyway...
These logs come from an internal tool that extracts the traces from a
complete RAM dump from a crashed device. One of the limitations is that
the script doesn't (yet) understand the event printk-formats so we're
left with just the raw field contents of the TP_STRUCT__entry. Not
pretty at all but it's *lot* better than nothing especially for
post-mortem analysis.
Yes there is crash-utility [1] which does all of this already but
logistically our internal tooling is not very well equipped (i.e. is
Windows-based) to make use of this, not without going through some
hoops. Been meaning to play around with it some more though to reduce
the need for our home-rolled solution.
[1] https://github.com/crash-utility
> > <idle>-0 [004] 304.583237: dwc3_gadget_giveback name=ep1out req=1243650560 actual=0 length=16384 status=4294967188 zero=0 short_not_ok=0 no_interrupt=0
> > <idle>-0 [004] 304.583275: dwc3_gadget_giveback name=ep1out req=1243645440 actual=0 length=16384 status=4294967188 zero=0 short_not_ok=0 no_interrupt=0
> > <idle>-0 [004] 304.583282: dwc3_gadget_giveback name=ep1out req=1243648256 actual=0 length=16384 status=4294967188 zero=0 short_not_ok=0 no_interrupt=0
> > <idle>-0 [004] 304.583312: dwc3_gadget_giveback name=ep1out req=4075146240 actual=0 length=16384 status=4294967188 zero=0 short_not_ok=0 no_interrupt=0
> > ...
> > //USB gadget disconnect is printed AFTER the pullup(0) routine is complete.
> > <idle>-0 [004] 304.584240: usb_gadget_disconnect speed=3 max_speed=6 state=7 mA=500 sg_supported=1 is_otg=0 is_a_peripheral=0 b_hnp_enable=0 a_hnp_support=0 hnp_polling_support=0 host_request_flag=0 quirk_ep_out_aligned_size=0 quirk_altset_not_supp=0 quirk_stall_not_supp=0 quirk_zlp_not_supp=0 is_selfpowered=0 deactivated=0 connected=1 ret=4294967186
> >
> > ---------------------------------------------------------
> > CPU2 SCHED FTRACE:
> >
> > kworker/u16:5-192 [002] 304.583354: irq_handler_entry irq=343 name=dwc3-msm
> > kworker/u16:5-192 [002] 304.583356: irq_handler_exit irq=343 ret=0
> > kworker/u16:5-192 [002] 304.583358: irq_handler_entry irq=343 name=dwc3
> > kworker/u16:5-192 [002] 304.583366: irq_handler_exit irq=343 ret=2
> > kworker/u16:5-192 [002] 304.583377: sched_wakeup comm=irq/343-dwc3 pid=20174 prio=100 success=1 target_cpu=2
> > kworker/u16:5-192 [002] 304.583411: sched_switch: prev_comm=kworker/u16:5 prev_pid=192 prev_prio=120 prev_state=R ==> next_comm=irq/343-dwc3 next_pid=20174 next_prio=100
> > irq/343-dwc3-20174 [002] 304.583454: irq_handler_entry irq=343 name=dwc3-msm
> > irq/343-dwc3-20174 [002] 304.583455: irq_handler_exit irq=343 ret=0
> > irq/343-dwc3-20174 [002] 304.583458: irq_handler_entry irq=343 name=dwc3
> > irq/343-dwc3-20174 [002] 304.583465: irq_handler_exit irq=343 ret=2
Jack
--
The Qualcomm Innovation Center, Inc. is a member of Code Aurora Forum,
a Linux Foundation Collaborative Project
Powered by blists - more mailing lists