[External] Re: [PATCH] i2c: aspeed: Fix i2c bus hang in slave read

Jian Zhang zhangjian.3032 at bytedance.com
Sat Sep 23 00:39:13 AEST 2023


> From: "Joel Stanley"<joel at jms.id.au>
> Date:  Fri, Sep 22, 2023, 14:48
> Subject:  [External] Re: [PATCH] i2c: aspeed: Fix i2c bus hang in slave read
> To: "Jian Zhang"<zhangjian.3032 at bytedance.com>, "Tommy Huang"<tommy_huang at aspeedtech.com>
> Cc: <brendan.higgins at linux.dev>, <benh at kernel.crashing.org>, <andrew at aj.id.au>, <zhangjian3032 at gmail.com>, <yulei.sh at bytedance.com>, <xiexinnan at bytedance.com>, "open list:ARM/ASPEED I2C DRIVER"<linux-i2c at vger.kernel.org>, "moderated list:ARM/ASPEED I2C DRIVER"<openbmc at lists.ozlabs.org>, "moderated list:ARM/ASPEED MACHINE SUPPORT"<linux-arm-kernel at lists.infradead.org>, "moderated list:ARM/ASPEED MACHINE SUPPORT"<linux-aspeed at lists.ozlabs.org>, "open list"<linux-kernel at vger.kernel.org>
> On Thu, 10 Aug 2023 at 07:22, Jian Zhang <zhangjian.3032 at bytedance.com> wrote:
> >
> > When the `CONFIG_I2C_SLAVE` option is enabled and the device operates
> > as a slave, a situation arises where the master sends a START signal
> > without the accompanying STOP signal. This action results in a
> > persistent I2C bus timeout. The core issue stems from the fact that
> > the i2c controller remains in a slave read state without a timeout
> > mechanism. As a consequence, the bus perpetually experiences timeouts.
> >
> > This proposed patch addresses this problem by introducing a status
> > check during i2c transmit timeouts. In the event that the controller
> > is in a slave read state, the i2c controller will be reset to restore
> > proper functionality and allow the I2C bus to resume normal operation.
> >
> > Signed-off-by: Jian Zhang <zhangjian.3032 at bytedance.com>
>
> Tommy has submitted a similar fix:
>
>  https://lore.kernel.org/linux-i2c/20230906004910.4157305-1-tommy_huang@aspeedtech.com/
>
> His change is very heavy handed; it reinitialises the bus including
> re-parsing the device tree (!).
>
> Should we have merged this fix instead? If not, are you able to
> confirm that his change fixes your issue?

I feel it's for solving the same issue, but I think this patch is
missing the action
`bus->slave_state = ASPEED_I2C_SLAVE_INACTIVE;`,
 which means it can't resolve my problem. @Tommy, can you help confirm this?

Thanks,

Jian

>
> Cheers,
>
> Joel
>
> > ---
> >  drivers/i2c/busses/i2c-aspeed.c | 17 +++++++++++++++++
> >  1 file changed, 17 insertions(+)
> >
> > diff --git a/drivers/i2c/busses/i2c-aspeed.c b/drivers/i2c/busses/i2c-aspeed.c
> > index e76befe3f60f..1a95205fc946 100644
> > --- a/drivers/i2c/busses/i2c-aspeed.c
> > +++ b/drivers/i2c/busses/i2c-aspeed.c
> > @@ -113,6 +113,7 @@
> >                  ASPEED_I2CD_M_RX_CMD |                                        \
> >                  ASPEED_I2CD_M_TX_CMD |                                        \
> >                  ASPEED_I2CD_M_START_CMD)
> > +#define ASPEED_I2CD_SLAVE_CMDS_MASK                    GENMASK(31, 29)
> >
> >  /* 0x18 : I2CD Slave Device Address Register   */
> >  #define ASPEED_I2CD_DEV_ADDR_MASK                      GENMASK(6, 0)
> > @@ -706,6 +707,22 @@ static int aspeed_i2c_master_xfer(struct i2c_adapter *adap,
> >                      ASPEED_I2CD_BUS_BUSY_STS))
> >                         aspeed_i2c_recover_bus(bus);
> >
> > +#if IS_ENABLED(CONFIG_I2C_SLAVE)
> > +               /*
> > +                * If master timed out and bus is in slave mode.
> > +                * reset the slave mode.
> > +                */
> > +               if (readl(bus->base + ASPEED_I2C_CMD_REG) & ASPEED_I2CD_SLAVE_CMDS_MASK) {
> > +                       spin_lock_irqsave(&bus->lock, flags);
> > +                       u32 func_ctrl_reg_val = readl(bus->base + ASPEED_I2C_FUN_CTRL_REG);
> > +
> > +                       writel(0, bus->base + ASPEED_I2C_FUN_CTRL_REG);
> > +                       writel(func_ctrl_reg_val, bus->base + ASPEED_I2C_FUN_CTRL_REG);
> > +                       bus->slave_state = ASPEED_I2C_SLAVE_INACTIVE;
> > +                       spin_unlock_irqrestore(&bus->lock, flags);
> > +               }
> > +#endif
> > +
> >                 /*
> >                  * If timed out and the state is still pending, drop the pending
> >                  * master command.
> > --
> > 2.30.2
> >


More information about the Linux-aspeed mailing list