MIME-Version: 1.0
In-Reply-To: <20161108024052.6b4c15d01df5508dab02973c2a5944d9.803a60fc5b.mailapi@email21.godaddy.com>
References: <CAKfKVtGnpdBsqtV2fEg2jk5v5XZtGULx-hc5waUKqUq_xRCp9w@mail.gmail.com>
 <20161108024052.6b4c15d01df5508dab02973c2a5944d9.803a60fc5b.mailapi@email21.godaddy.com>
From: Shubhrajyoti Datta <shubhrajyoti.datta@gmail.com>
Date: Tue, 8 Nov 2016 17:26:29 +0530
Message-ID: <CAKfKVtFKp8ZnLmDh-Qq5LvHTaqVafZ3P3M+E6HfvSM+Z0qYcQA@mail.gmail.com>
Subject: Re: Re: Reset implementation for Zynq
To: iztok.jeras@redpitaya.com
Cc: Moritz Fischer <moritz.fischer@ettus.com>, p.zabel@pengutronix.de,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset=UTF-8
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1194
Lines: 36

Hi Iztok,


On Tue, Nov 8, 2016 at 3:10 PM,  <iztok.jeras@redpitaya.com> wrote:
> Hi Shubhrajyoti,
>
> By lockup state, I mean an undefined/unhandled state of the HW logic state
> machine,
> or something similar in the software driver.
> Based on symptoms I assume it is a HW lockup.
> In practice asking the driver for a new I2C transfer would result in
> nothing happening on the I2C bus (signals remain in hi-Z state).
> Driver or userspace tool reports a transfer error, but I did not copy it.
> And I did not check the state of the driver.
>
> To enter this mode it was enough to connect an oscilloscope probe to the SDA
> signal.
> The capacitance of the probe causes a deep voltage fall, which is quickly
> corrected by the pullup.
> I did not time the glitch, and the shape was partially defined by an I2C
> level shifter.
> The issue can be repeated with about 80% probability.
>
> There are two ways to get back to a working state:
> 1. create a longer zero pulse (about 50ms manual grounding) on SDA
> 2. reset the HW, this was done with a patched driver.
Thanks  for the explaination.
May be recovery mechanism  implementation could be considered.


>
> Regards,
> Iztok Jeras
>
>