diff mbox series

[v7,2/4] i2c: npcm: Modify the client address assignment

Message ID 20241021062732.5592-3-kfting@nuvoton.com
State Under Review
Delegated to: Andi Shyti
Headers show
Series i2c: npcm: read/write operation, checkpatch | expand

Commit Message

Tyrone Ting Oct. 21, 2024, 6:27 a.m. UTC
From: Tyrone Ting <kfting@nuvoton.com>

Store the client address earlier since it might get called in
the i2c_recover_bus() logic flow at the early stage of
npcm_i2c_master_xfer().

Signed-off-by: Tyrone Ting <kfting@nuvoton.com>
Reviewed-by: Tali Perry <tali.perry1@gmail.com>
---
 drivers/i2c/busses/i2c-npcm7xx.c | 20 +++++++++++++++++++-
 1 file changed, 19 insertions(+), 1 deletion(-)

Comments

Paul Menzel Oct. 21, 2024, 7:01 a.m. UTC | #1
Dear Tyrone,


Thank you for your patch.

Am 21.10.24 um 08:27 schrieb Tyrone Ting:
> From: Tyrone Ting <kfting@nuvoton.com>
> 
> Store the client address earlier since it might get called in
> the i2c_recover_bus() logic flow at the early stage of
> npcm_i2c_master_xfer().

Thank you for the description. For the summary/title it’d be great, if 
you were more specific. For example:

i2c: npcm: Assign client address earlier for `i2c_recover_bus()`

It’d be great if you noted the commit, your patch fixes, so it’s clear 
since when the problem has been present.

> Signed-off-by: Tyrone Ting <kfting@nuvoton.com>
> Reviewed-by: Tali Perry <tali.perry1@gmail.com>
> ---
>   drivers/i2c/busses/i2c-npcm7xx.c | 20 +++++++++++++++++++-
>   1 file changed, 19 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/i2c/busses/i2c-npcm7xx.c b/drivers/i2c/busses/i2c-npcm7xx.c
> index c96a25d37c14..a9a9b21f1f0b 100644
> --- a/drivers/i2c/busses/i2c-npcm7xx.c
> +++ b/drivers/i2c/busses/i2c-npcm7xx.c
> @@ -2155,6 +2155,16 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
>   
>   	} while (time_is_after_jiffies(time_left) && bus_busy);
>   
> +	/*
> +	 * Previously, the 7-bit address was stored and being converted to
> +	 * the address of event in the following call to npcm_i2c_master_start_xmit().
> +	 *
> +	 * Since there are cases that the i2c_recover_bus() gets called at the
> +	 * early stage of npcm_i2c_master_xfer(), the address of event is stored
> +	 * and then used in the i2c_recover_bus().
> +	 */
> +	bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> +
>   	/*
>   	 * Check the BER (bus error) state, when ber_state is true, it means that the module
>   	 * detects the bus error which is caused by some factor like that the electricity
> @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
>   	 * bus is busy.
>   	 */
>   	if (bus_busy || bus->ber_state) {
> +		/*
> +		 * Since the transfer might be a read operation, remove the I2C_M_RD flag
> +		 * from the bus->dest_addr for the i2c_recover_bus() call later.
> +		 *
> +		 * The i2c_recover_bus() uses the address in a write direction to recover
> +		 * the i2c bus if some error condition occurs.
> +		 */
> +		bus->dest_addr &= ~I2C_M_RD;
> +
>   		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
>   		npcm_i2c_reset(bus);
>   		i2c_recover_bus(adap);
> @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
>   	}
>   
>   	npcm_i2c_init_params(bus);
> -	bus->dest_addr = slave_addr;
>   	bus->msgs = msgs;
>   	bus->msgs_num = num;
>   	bus->cmd_err = 0;


Kind regards,

Paul
Tyrone Ting Oct. 22, 2024, 8:08 a.m. UTC | #2
Hi Paul:

Thank you for your comment.

It'll be addressed in the next patch set.

Paul Menzel <pmenzel@molgen.mpg.de> 於 2024年10月21日 週一 下午3:01寫道:
>
> Dear Tyrone,
>
>
> Thank you for your patch.
>
> Am 21.10.24 um 08:27 schrieb Tyrone Ting:
> > From: Tyrone Ting <kfting@nuvoton.com>
> >
> > Store the client address earlier since it might get called in
> > the i2c_recover_bus() logic flow at the early stage of
> > npcm_i2c_master_xfer().
>
> Thank you for the description. For the summary/title it’d be great, if
> you were more specific. For example:
>
> i2c: npcm: Assign client address earlier for `i2c_recover_bus()`
>
> It’d be great if you noted the commit, your patch fixes, so it’s clear
> since when the problem has been present.
>
> > Signed-off-by: Tyrone Ting <kfting@nuvoton.com>
> > Reviewed-by: Tali Perry <tali.perry1@gmail.com>
> > ---
> >   drivers/i2c/busses/i2c-npcm7xx.c | 20 +++++++++++++++++++-
> >   1 file changed, 19 insertions(+), 1 deletion(-)
> >
> > diff --git a/drivers/i2c/busses/i2c-npcm7xx.c b/drivers/i2c/busses/i2c-npcm7xx.c
> > index c96a25d37c14..a9a9b21f1f0b 100644
> > --- a/drivers/i2c/busses/i2c-npcm7xx.c
> > +++ b/drivers/i2c/busses/i2c-npcm7xx.c
> > @@ -2155,6 +2155,16 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> >
> >       } while (time_is_after_jiffies(time_left) && bus_busy);
> >
> > +     /*
> > +      * Previously, the 7-bit address was stored and being converted to
> > +      * the address of event in the following call to npcm_i2c_master_start_xmit().
> > +      *
> > +      * Since there are cases that the i2c_recover_bus() gets called at the
> > +      * early stage of npcm_i2c_master_xfer(), the address of event is stored
> > +      * and then used in the i2c_recover_bus().
> > +      */
> > +     bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> > +
> >       /*
> >        * Check the BER (bus error) state, when ber_state is true, it means that the module
> >        * detects the bus error which is caused by some factor like that the electricity
> > @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> >        * bus is busy.
> >        */
> >       if (bus_busy || bus->ber_state) {
> > +             /*
> > +              * Since the transfer might be a read operation, remove the I2C_M_RD flag
> > +              * from the bus->dest_addr for the i2c_recover_bus() call later.
> > +              *
> > +              * The i2c_recover_bus() uses the address in a write direction to recover
> > +              * the i2c bus if some error condition occurs.
> > +              */
> > +             bus->dest_addr &= ~I2C_M_RD;
> > +
> >               iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
> >               npcm_i2c_reset(bus);
> >               i2c_recover_bus(adap);
> > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> >       }
> >
> >       npcm_i2c_init_params(bus);
> > -     bus->dest_addr = slave_addr;
> >       bus->msgs = msgs;
> >       bus->msgs_num = num;
> >       bus->cmd_err = 0;
>
>
> Kind regards,
>
> Paul

Have a nice day.

Regards,
Tyrone
Andi Shyti Oct. 24, 2024, 9:28 a.m. UTC | #3
Hi Tyrone,

On Tue, Oct 22, 2024 at 04:08:46PM +0800, Tyrone Ting wrote:
> Hi Paul:
> 
> Thank you for your comment.
> 
> It'll be addressed in the next patch set.

No need to resend, I can take care of it.

Andi
Andi Shyti Oct. 24, 2024, 10:03 a.m. UTC | #4
Hi Tyrone,

...

> +	/*
> +	 * Previously, the 7-bit address was stored and being converted to
> +	 * the address of event in the following call to npcm_i2c_master_start_xmit().

Do we care how it was done previously? I think this is not a
useful information as the code readers will se the code the way
it is now, not the way it was done "previously".

(there is a related comment at the end)

> +	 * Since there are cases that the i2c_recover_bus() gets called at the
> +	 * early stage of npcm_i2c_master_xfer(), the address of event is stored
> +	 * and then used in the i2c_recover_bus().

I could rephrase this sentence to something like:

/*
 * Store the address early in a global position to ensure it is
 * accessible for a potential call to i2c_recover_bus().
 */

> +	 */
> +	bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> +
>  	/*
>  	 * Check the BER (bus error) state, when ber_state is true, it means that the module
>  	 * detects the bus error which is caused by some factor like that the electricity
> @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
>  	 * bus is busy.
>  	 */
>  	if (bus_busy || bus->ber_state) {
> +		/*
> +		 * Since the transfer might be a read operation, remove the I2C_M_RD flag
> +		 * from the bus->dest_addr for the i2c_recover_bus() call later.
> +		 *
> +		 * The i2c_recover_bus() uses the address in a write direction to recover
> +		 * the i2c bus if some error condition occurs.
> +		 */
> +		bus->dest_addr &= ~I2C_M_RD;
> +
>  		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
>  		npcm_i2c_reset(bus);
>  		i2c_recover_bus(adap);
> @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
>  	}
>  
>  	npcm_i2c_init_params(bus);
> -	bus->dest_addr = slave_addr;

We can now get rid of slave_addr. It's just used in
npcm_i2c_master_start_xmit(). Right?

Andi

>  	bus->msgs = msgs;
>  	bus->msgs_num = num;
>  	bus->cmd_err = 0;
> -- 
> 2.34.1
>
Tyrone Ting Oct. 25, 2024, 1:36 a.m. UTC | #5
HI Andi:

Thank you for your help on this.

Andi Shyti <andi.shyti@kernel.org> 於 2024年10月24日 週四 下午5:28寫道:
>
> Hi Tyrone,
>
> On Tue, Oct 22, 2024 at 04:08:46PM +0800, Tyrone Ting wrote:
> > Hi Paul:
> >
> > Thank you for your comment.
> >
> > It'll be addressed in the next patch set.
>
> No need to resend, I can take care of it.
>
> Andi

Have a nice day.

Regards,
Tyrone
Tyrone Ting Oct. 25, 2024, 1:43 a.m. UTC | #6
Hi Andi:

Thank you for your comments.

Andi Shyti <andi.shyti@kernel.org> 於 2024年10月24日 週四 下午6:04寫道:
>
> Hi Tyrone,
>
> ...
>
> > +     /*
> > +      * Previously, the 7-bit address was stored and being converted to
> > +      * the address of event in the following call to npcm_i2c_master_start_xmit().
>
> Do we care how it was done previously? I think this is not a
> useful information as the code readers will se the code the way
> it is now, not the way it was done "previously".
>

Yes, it's not a useful information anymore.

> (there is a related comment at the end)
>
> > +      * Since there are cases that the i2c_recover_bus() gets called at the
> > +      * early stage of npcm_i2c_master_xfer(), the address of event is stored
> > +      * and then used in the i2c_recover_bus().
>
> I could rephrase this sentence to something like:
>
> /*
>  * Store the address early in a global position to ensure it is
>  * accessible for a potential call to i2c_recover_bus().
>  */

Understood. Thank you for your help on this.

>
> > +      */
> > +     bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> > +
> >       /*
> >        * Check the BER (bus error) state, when ber_state is true, it means that the module
> >        * detects the bus error which is caused by some factor like that the electricity
> > @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> >        * bus is busy.
> >        */
> >       if (bus_busy || bus->ber_state) {
> > +             /*
> > +              * Since the transfer might be a read operation, remove the I2C_M_RD flag
> > +              * from the bus->dest_addr for the i2c_recover_bus() call later.
> > +              *
> > +              * The i2c_recover_bus() uses the address in a write direction to recover
> > +              * the i2c bus if some error condition occurs.
> > +              */
> > +             bus->dest_addr &= ~I2C_M_RD;
> > +
> >               iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
> >               npcm_i2c_reset(bus);
> >               i2c_recover_bus(adap);
> > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> >       }
> >
> >       npcm_i2c_init_params(bus);
> > -     bus->dest_addr = slave_addr;
>
> We can now get rid of slave_addr. It's just used in
> npcm_i2c_master_start_xmit(). Right?

Yes, slave_addr is just used as the link
https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
suggests with this patch.

>
> Andi
>
> >       bus->msgs = msgs;
> >       bus->msgs_num = num;
> >       bus->cmd_err = 0;
> > --
> > 2.34.1
> >

Thank you.

Regards,
Tyrone
Tyrone Ting Oct. 29, 2024, 8:50 a.m. UTC | #7
Hi Andi:

Sorry to bother you.

May I have your comments about my feedback on these patches?

It'll be great to know if I need to prepare the next patch set for reviewing.

Tyrone Ting <warp5tw@gmail.com> 於 2024年10月25日 週五 上午9:43寫道:
>
> Hi Andi:
>
> Thank you for your comments.
>
> Andi Shyti <andi.shyti@kernel.org> 於 2024年10月24日 週四 下午6:04寫道:
> >
> > Hi Tyrone,
> >
> > ...
> >
> > > +     /*
> > > +      * Previously, the 7-bit address was stored and being converted to
> > > +      * the address of event in the following call to npcm_i2c_master_start_xmit().
> >
> > Do we care how it was done previously? I think this is not a
> > useful information as the code readers will se the code the way
> > it is now, not the way it was done "previously".
> >
>
> Yes, it's not a useful information anymore.
>
> > (there is a related comment at the end)
> >
> > > +      * Since there are cases that the i2c_recover_bus() gets called at the
> > > +      * early stage of npcm_i2c_master_xfer(), the address of event is stored
> > > +      * and then used in the i2c_recover_bus().
> >
> > I could rephrase this sentence to something like:
> >
> > /*
> >  * Store the address early in a global position to ensure it is
> >  * accessible for a potential call to i2c_recover_bus().
> >  */
>
> Understood. Thank you for your help on this.
>
> >
> > > +      */
> > > +     bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> > > +
> > >       /*
> > >        * Check the BER (bus error) state, when ber_state is true, it means that the module
> > >        * detects the bus error which is caused by some factor like that the electricity
> > > @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > >        * bus is busy.
> > >        */
> > >       if (bus_busy || bus->ber_state) {
> > > +             /*
> > > +              * Since the transfer might be a read operation, remove the I2C_M_RD flag
> > > +              * from the bus->dest_addr for the i2c_recover_bus() call later.
> > > +              *
> > > +              * The i2c_recover_bus() uses the address in a write direction to recover
> > > +              * the i2c bus if some error condition occurs.
> > > +              */
> > > +             bus->dest_addr &= ~I2C_M_RD;
> > > +
> > >               iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
> > >               npcm_i2c_reset(bus);
> > >               i2c_recover_bus(adap);
> > > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > >       }
> > >
> > >       npcm_i2c_init_params(bus);
> > > -     bus->dest_addr = slave_addr;
> >
> > We can now get rid of slave_addr. It's just used in
> > npcm_i2c_master_start_xmit(). Right?
>
> Yes, slave_addr is just used as the link
> https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
> suggests with this patch.
>
> >
> > Andi
> >
> > >       bus->msgs = msgs;
> > >       bus->msgs_num = num;
> > >       bus->cmd_err = 0;
> > > --
> > > 2.34.1
> > >
>
> Thank you.
>
> Regards,
> Tyrone

Have a nice day.

Thank you.

Regards,
Tyrone
Tyrone Ting Nov. 5, 2024, 1:52 a.m. UTC | #8
Hi Andi:

May I have your comments about my feedback on these patches?

Tyrone Ting <warp5tw@gmail.com> 於 2024年10月29日 週二 下午4:50寫道:
>
> Hi Andi:
>
> Sorry to bother you.
>
> May I have your comments about my feedback on these patches?
>
> It'll be great to know if I need to prepare the next patch set for reviewing.
>
> Tyrone Ting <warp5tw@gmail.com> 於 2024年10月25日 週五 上午9:43寫道:
> >
> > Hi Andi:
> >
> > Thank you for your comments.
> >
> > Andi Shyti <andi.shyti@kernel.org> 於 2024年10月24日 週四 下午6:04寫道:
> > >
> > > Hi Tyrone,
> > >
> > > ...
> > >
> > > > +     /*
> > > > +      * Previously, the 7-bit address was stored and being converted to
> > > > +      * the address of event in the following call to npcm_i2c_master_start_xmit().
> > >
> > > Do we care how it was done previously? I think this is not a
> > > useful information as the code readers will se the code the way
> > > it is now, not the way it was done "previously".
> > >
> >
> > Yes, it's not a useful information anymore.
> >
> > > (there is a related comment at the end)
> > >
> > > > +      * Since there are cases that the i2c_recover_bus() gets called at the
> > > > +      * early stage of npcm_i2c_master_xfer(), the address of event is stored
> > > > +      * and then used in the i2c_recover_bus().
> > >
> > > I could rephrase this sentence to something like:
> > >
> > > /*
> > >  * Store the address early in a global position to ensure it is
> > >  * accessible for a potential call to i2c_recover_bus().
> > >  */
> >
> > Understood. Thank you for your help on this.
> >
> > >
> > > > +      */
> > > > +     bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
> > > > +
> > > >       /*
> > > >        * Check the BER (bus error) state, when ber_state is true, it means that the module
> > > >        * detects the bus error which is caused by some factor like that the electricity
> > > > @@ -2165,6 +2175,15 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > > >        * bus is busy.
> > > >        */
> > > >       if (bus_busy || bus->ber_state) {
> > > > +             /*
> > > > +              * Since the transfer might be a read operation, remove the I2C_M_RD flag
> > > > +              * from the bus->dest_addr for the i2c_recover_bus() call later.
> > > > +              *
> > > > +              * The i2c_recover_bus() uses the address in a write direction to recover
> > > > +              * the i2c bus if some error condition occurs.
> > > > +              */
> > > > +             bus->dest_addr &= ~I2C_M_RD;
> > > > +
> > > >               iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
> > > >               npcm_i2c_reset(bus);
> > > >               i2c_recover_bus(adap);
> > > > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > > >       }
> > > >
> > > >       npcm_i2c_init_params(bus);
> > > > -     bus->dest_addr = slave_addr;
> > >
> > > We can now get rid of slave_addr. It's just used in
> > > npcm_i2c_master_start_xmit(). Right?
> >
> > Yes, slave_addr is just used as the link
> > https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
> > suggests with this patch.
> >
> > >
> > > Andi
> > >
> > > >       bus->msgs = msgs;
> > > >       bus->msgs_num = num;
> > > >       bus->cmd_err = 0;
> > > > --
> > > > 2.34.1
> > > >
> >
> > Thank you.
> >
> > Regards,
> > Tyrone
>
> Have a nice day.
>
> Thank you.
>
> Regards,
> Tyrone

Thank you.

Regards,
Tyrone
Andi Shyti Nov. 19, 2024, 10:22 p.m. UTC | #9
Hi Tyrone,

Sorry for the late reply!

...

> > > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > >       }
> > >
> > >       npcm_i2c_init_params(bus);
> > > -     bus->dest_addr = slave_addr;
> >
> > We can now get rid of slave_addr. It's just used in
> > npcm_i2c_master_start_xmit(). Right?
> 
> Yes, slave_addr is just used as the link
> https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
> suggests with this patch.

What I mean is that slave_addr now is completely unused. We
declare it, we initialize it to msg0->addr and it stays like
this.

What I'm suggesting is to remove it completely.

Andi
Tyrone Ting Nov. 21, 2024, 9:21 a.m. UTC | #10
Hi Andi:

Thank you for your reply.

Andi Shyti <andi.shyti@kernel.org> 於 2024年11月20日 週三 上午6:22寫道:
>
> Hi Tyrone,
>
> Sorry for the late reply!
>
> ...
>
> > > > @@ -2172,7 +2191,6 @@ static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
> > > >       }
> > > >
> > > >       npcm_i2c_init_params(bus);
> > > > -     bus->dest_addr = slave_addr;
> > >
> > > We can now get rid of slave_addr. It's just used in
> > > npcm_i2c_master_start_xmit(). Right?
> >
> > Yes, slave_addr is just used as the link
> > https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
> > suggests with this patch.
>
> What I mean is that slave_addr now is completely unused. We
> declare it, we initialize it to msg0->addr and it stays like
> this.
>
> What I'm suggesting is to remove it completely.
>

Sorry that I might not describe it clearly in previous discussion
thread. slave_addr is passed to npcm_i2c_master_start_xmit() as
https://github.com/torvalds/linux/blob/master/drivers/i2c/busses/i2c-npcm7xx.c#L2182
suggests. If slave_addr is removed, may I use "(bus->dest_addr &=
~I2C_M_RD)" to replace the slave_addr parameter to
npcm_i2c_master_start_xmit()?

> Andi

Thank you.

Regards,
Tyrone
diff mbox series

Patch

diff --git a/drivers/i2c/busses/i2c-npcm7xx.c b/drivers/i2c/busses/i2c-npcm7xx.c
index c96a25d37c14..a9a9b21f1f0b 100644
--- a/drivers/i2c/busses/i2c-npcm7xx.c
+++ b/drivers/i2c/busses/i2c-npcm7xx.c
@@ -2155,6 +2155,16 @@  static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 
 	} while (time_is_after_jiffies(time_left) && bus_busy);
 
+	/*
+	 * Previously, the 7-bit address was stored and being converted to
+	 * the address of event in the following call to npcm_i2c_master_start_xmit().
+	 *
+	 * Since there are cases that the i2c_recover_bus() gets called at the
+	 * early stage of npcm_i2c_master_xfer(), the address of event is stored
+	 * and then used in the i2c_recover_bus().
+	 */
+	bus->dest_addr = i2c_8bit_addr_from_msg(msg0);
+
 	/*
 	 * Check the BER (bus error) state, when ber_state is true, it means that the module
 	 * detects the bus error which is caused by some factor like that the electricity
@@ -2165,6 +2175,15 @@  static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 	 * bus is busy.
 	 */
 	if (bus_busy || bus->ber_state) {
+		/*
+		 * Since the transfer might be a read operation, remove the I2C_M_RD flag
+		 * from the bus->dest_addr for the i2c_recover_bus() call later.
+		 *
+		 * The i2c_recover_bus() uses the address in a write direction to recover
+		 * the i2c bus if some error condition occurs.
+		 */
+		bus->dest_addr &= ~I2C_M_RD;
+
 		iowrite8(NPCM_I2CCST_BB, bus->reg + NPCM_I2CCST);
 		npcm_i2c_reset(bus);
 		i2c_recover_bus(adap);
@@ -2172,7 +2191,6 @@  static int npcm_i2c_master_xfer(struct i2c_adapter *adap, struct i2c_msg *msgs,
 	}
 
 	npcm_i2c_init_params(bus);
-	bus->dest_addr = slave_addr;
 	bus->msgs = msgs;
 	bus->msgs_num = num;
 	bus->cmd_err = 0;