linux-usb.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
From: Mathias Nyman <mathias.nyman@linux.intel.com>
To: Pavan Kondeti <quic_pkondeti@quicinc.com>,
	Jack Pham <quic_jackp@quicinc.com>
Cc: s.shtylyov@omp.ru, linux-usb@vger.kernel.org
Subject: Re: [RFT PATCH] xhci: make xhci_handshake timeout for xhci_reset() adjustable
Date: Tue, 1 Mar 2022 10:47:36 +0200	[thread overview]
Message-ID: <ffc9abb1-908b-1c09-cd9d-cbd83ea4e478@linux.intel.com> (raw)
In-Reply-To: <20220301040339.GA11371@hu-pkondeti-hyd.qualcomm.com>

On 1.3.2022 6.03, Pavan Kondeti wrote:
> On Mon, Feb 28, 2022 at 05:49:49PM -0800, Jack Pham wrote:
>> Hi Mathias,
>>
>> On Thu, Feb 17, 2022 at 03:56:43PM +0200, Mathias Nyman wrote:
>>> xhci_reset() timeout was increased from 250ms to 10 seconds in order to
>>> give Renesas 720201 xHC enough time to get ready in probe.
>>
>> This suggests that the only place we really need the long timeout is
>> in xhci_gen_setup().
>>
>>> @@ -1163,7 +1161,7 @@ int xhci_resume(struct xhci_hcd *xhci, bool hibernated)
>>>  		xhci_dbg(xhci, "Stop HCD\n");
>>>  		xhci_halt(xhci);
>>>  		xhci_zero_64b_regs(xhci);
>>> -		retval = xhci_reset(xhci);
>>> +		retval = xhci_reset(xhci, XHCI_RESET_LONG_USEC);
>>
>> Since preemption is disabled here (spin_lock_irq() is called near the
>> start of this function), shouldn't we also limit this to 250ms here in
>> xhci_resume() as well?
>>> The rationale of decreasing the timeout to 250 in certain places is based
> on the criticality of the operation but not on the preemption/irq state.
> Since xHC reset is critical in startup and resume, the 10 seconds timeout
> is enforced so that we don't break Renesas 720201 xHC.
> 
> Since all of our internl test reports indicate that the timeout is happening
> from stop hcd, this patch is helping.
> 

This was pretty much my reasoning as well.
I could add a comment about this to the commit message

In addition we want a targeted fix for a real world issue that we can send to
stable without changing too much, risking regressions.

I also think we should focus more on fixing the locking (preemption/irq state)
around xhci_reset() in xhci_resume() than tuning the timeout, but this needs more
thought and should be a separate patch for later.

Additionally I guess xhci_reset() is more likely to fail in xhci_stop() and
xhci_shutdown() as power or clocks for xHC may be disabled, or entire xHC removed from the
bus by then.   

Thanks
-Mathias

  reply	other threads:[~2022-03-01  8:46 UTC|newest]

Thread overview: 11+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-17 13:56 [RFT PATCH] xhci: make xhci_handshake timeout for xhci_reset() adjustable Mathias Nyman
2022-02-18  9:41 ` Pavan Kondeti
2022-02-24  8:44   ` Udipto Goswami
2022-02-28 11:39     ` Mathias Nyman
2022-02-28 12:10       ` Pavan Kondeti
2022-03-01  1:49 ` Jack Pham
2022-03-01  4:03   ` Pavan Kondeti
2022-03-01  8:47     ` Mathias Nyman [this message]
2022-03-02  3:23       ` Pavan Kondeti
2022-03-02  7:49         ` Greg Kroah-Hartman
2022-03-02  7:57           ` Pavan Kondeti

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=ffc9abb1-908b-1c09-cd9d-cbd83ea4e478@linux.intel.com \
    --to=mathias.nyman@linux.intel.com \
    --cc=linux-usb@vger.kernel.org \
    --cc=quic_jackp@quicinc.com \
    --cc=quic_pkondeti@quicinc.com \
    --cc=s.shtylyov@omp.ru \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).