lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 27 Nov 2017 08:37:46 +0000
From:   "Zhang, Sean C. (NSB - CN/Hangzhou)" <sean.c.zhang@...ia-sbell.com>
To:     Jan Glauber <jan.glauber@...iumnetworks.com>
CC:     "david.daney@...ium.com" <david.daney@...ium.com>,
        "wsa@...-dreams.de" <wsa@...-dreams.de>,
        "linux-i2c@...r.kernel.org" <linux-i2c@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>
Subject: RE: [Bug fix] octeon-i2c driver updates

Hi Jan,

There are two points in this patch.

Point 1. As you see, replaced octeon_i2c_init_lowlevel() by recover bus status if TWSI controller is not IDLE.
Please take a scenario like this: when system soft reset without I2C slave reset, maybe make this I2C bus
dead lock occurred (I2C slave device stuck low SCL) in chance. Then during system goes up and I2C slave 
device creating process, if this I2C slave device has a register with less than 8 bytes to read, but I2C bus was
still stuck low SCL by last system reset, then the read will failed and this I2C slave device cannot be created.
If bus recovered before the reading process, this failure can be fixed.

Function flow explanation shown as below:

a. System reset without I2C slave device reset
--make SCL stuck low by I2C slave device
......
b. octeon_i2c_probe()
-- octeon_i2c_init_lowlevel  //reset TWSI core, but SCL still stuck low by.
......

c. Another I2C slave device creating process
octeon_i2c_xfer()
-- octeon_i2c_hlc_comp_read() //failed due to SCL stuck low.

If full recovery executed in octeon_i2c_probe(), above failure can be avoided.


Point 2. octeon_i2c_recovery() is used in octeon_i2c_start() error branch, in the case of octeon_i2c_recovery()
successful, octeon_i2c_start() will return -EAGAIN, and then octeon_i2c_xfer() return with error. I understand this like
this: if octeon_i2c_recovery() successful, then i2c START signal can be sent again, and all following step can be continue,
octeon_i2c_xfer() should not return error from this condition.

BR,
Sean Zhang

-----Original Message-----
From: Jan Glauber [mailto:jan.glauber@...iumnetworks.com] 
Sent: Friday, November 24, 2017 9:10 PM
To: Zhang, Sean C. (NSB - CN/Hangzhou) <sean.c.zhang@...ia-sbell.com>
Cc: david.daney@...ium.com; wsa@...-dreams.de; linux-i2c@...r.kernel.org; linux-kernel@...r.kernel.org
Subject: Re: [Bug fix] octeon-i2c driver updates

On Thu, Nov 23, 2017 at 11:42:36AM +0000, Zhang, Sean C. (NSB - CN/Hangzhou) wrote:
> Dear Maintainer,
> 
> For octeon TWSI controller, I found below two cases, maybe can be improved.

Hi Sean,

form the description below this looks like you're fixing a bug. Can you
elaborate on when the I2C bus dead lock occurs. Is it always happening?

What I don't like about the patch is that you're removing
octeon_i2c_init_lowlevel() from the probe and replacing it by _always_
going through a full recovery. Why is this neccessary?

Regards,
Jan

> 
> >From 09d9f0ce658d7f6a50d1af352dde619c29bc8bcf Mon Sep 17 00:00:00 2001
> From: hgt463 <sean.c.zhang@...ia.com>
> Date: Thu, 23 Nov 2017 18:46:09 +0800
> Subject: [PATCH] Driver updates:
>  - In the case of I2C bus dead lock occurred during driver probing,
>    it is better try to recovery it. so added bus recovery step in
>    octeon_i2c_probe();
>  - The purpose of function octeon_i2c_start() is to send START, so after
>    bus recovery, also need try to send START again.
> 
> Signed-off-by: hgt463 <sean.c.zhang@...ia.com>
> ---
>  drivers/i2c/busses/i2c-octeon-core.c    |   31 ++++++++++++++++++++++++++++++-
>  drivers/i2c/busses/i2c-octeon-platdrv.c |   15 +++++++++------
>  2 files changed, 39 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/i2c/busses/i2c-octeon-core.c b/drivers/i2c/busses/i2c-octeon-core.c
> index 1d87757..3ae1e03 100644
> --- a/drivers/i2c/busses/i2c-octeon-core.c
> +++ b/drivers/i2c/busses/i2c-octeon-core.c
> @@ -255,6 +255,35 @@ static int octeon_i2c_recovery(struct octeon_i2c *i2c)
>         return ret;
>  }
>  
> +/*
> + * octeon_i2c_start_retry - send START to the bus after bus recovery.
> + * @i2c: The struct octeon_i2c
> + *
> + * Returns 0 on success, otherwise a negative errno.
> + */
> +static int octeon_i2c_start_retry(struct octeon_i2c *i2c)
> +{
> +       int ret;
> +       u8 stat;
> +
> +       ret = octeon_i2c_recovery(i2c);
> +       if (ret)
> +               goto error;
> +
> +       octeon_i2c_ctl_write(i2c, TWSI_CTL_ENAB | TWSI_CTL_STA);
> +       ret = octeon_i2c_wait(i2c);
> +       if (ret)
> +               goto error;
> +
> +       stat = octeon_i2c_stat_read(i2c);
> +       if (stat == STAT_START || stat == STAT_REP_START)
> +               /* START successful, bail out */
> +               return 0;
> +
> +error:
> +       return (ret) ? ret : -EAGAIN;
> +}
> +
>  /**
>   * octeon_i2c_start - send START to the bus
>   * @i2c: The struct octeon_i2c
> @@ -280,7 +309,7 @@ static int octeon_i2c_start(struct octeon_i2c *i2c)
>  
>  error:
>         /* START failed, try to recover */
> -       ret = octeon_i2c_recovery(i2c);
> +       ret = octeon_i2c_start_retry(i2c);
>         return (ret) ? ret : -EAGAIN;
>  }
>  
> diff --git a/drivers/i2c/busses/i2c-octeon-platdrv.c b/drivers/i2c/busses/i2c-octeon-platdrv.c
> index 64bda83..98063af 100644
> --- a/drivers/i2c/busses/i2c-octeon-platdrv.c
> +++ b/drivers/i2c/busses/i2c-octeon-platdrv.c
> @@ -228,12 +228,6 @@ static int octeon_i2c_probe(struct platform_device *pdev)
>         if (OCTEON_IS_MODEL(OCTEON_CN38XX))
>                 i2c->broken_irq_check = true;
>  
> -       result = octeon_i2c_init_lowlevel(i2c);
> -       if (result) {
> -               dev_err(i2c->dev, "init low level failed\n");
> -               goto  out;
> -       }
> -
>         octeon_i2c_set_clock(i2c);
>  
>         i2c->adap = octeon_i2c_ops;
> @@ -245,6 +239,15 @@ static int octeon_i2c_probe(struct platform_device *pdev)
>         i2c_set_adapdata(&i2c->adap, i2c);
>         platform_set_drvdata(pdev, i2c);
>  
> +       stat = octeon_i2c_stat_read(i2c);
> +       if (stat != STAT_IDLE) {
> +               result = octeon_i2c_recovery(i2c);
> +               if (result) {
> +                       dev_err(i2c->dev, "octeon i2c recovery failed\n");
> +                       goto  out;
> +               }
> +       }
> +
>         result = i2c_add_adapter(&i2c->adap);
>         if (result < 0)
>                 goto out;
> 
> 
> Attached patch for you review. Thanks in advance.
> 
> BR,
> Sean Zhang


Powered by blists - more mailing lists