RE: [PATCH v3 3/3] serial: 8250_em: Add serial_out() to struct serial8250_em_hw_info

["Ilpo Järvinen" <ilpo.jarvinen@linux.intel.com>]

On Mon, 13 Feb 2023, Biju Das wrote:

> HI Ilpo,
> 
> Thanks for the feedback.
> 
> > Subject: Re: [PATCH v3 3/3] serial: 8250_em: Add serial_out() to struct
> > serial8250_em_hw_info
> > 
> > On Fri, 10 Feb 2023, Biju Das wrote:
> > 
> > > As per HW manual section 40.6.1, we need to perform FIFO reset + SW
> > > reset before updating the below registers.
> > >
> > > FCR[7:5], FCR[3:0], LCR[7][5:0], MCR[6:4], DLL[7:0], DLM[7:0] and
> > > HCR0[6:5][3:2].
> > >
> > > This patch adds serial_out() to struct serial8250_em_hw_info to handle
> > > this difference between emma mobile and rz/v2m.
> > >
> > > DLL/DLM register can be updated only by setting LCR[7]. So the
> > > updation of LCR[7] will perform reset for DLL/DLM register changes.
> > >
> > > Signed-off-by: Biju Das <biju.das.jz@bp.renesas.com>
> > > ---
> > > v2->v3:
> > >  * Replaced readl/writel()->serial8250_em_serial_in/out() in
> > serial8250_rzv2m_
> > >    reg_update() to avoid duplication of offset trickery.
> > >  * Added support for HCR0 read/write in serial8250_em_{serial_in,
> > > serial_out}
> > >  * Added UART_LCR macro support in serial8250_em_serial_in() to read
> > > LCR
> > >  * Reordered serial8250_em_{serial_in, serial_out} above serial8250_rzv2m_
> > >    reg_update().
> > >  * Replaced priv->info->serial_out to info->serial_out.
> > > v1->v2:
> > >  * Added serial_out to struct serial8250_em_hw_info
> > > ---
> > >  drivers/tty/serial/8250/8250_em.c | 67
> > > ++++++++++++++++++++++++++++++-
> > >  1 file changed, 66 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/tty/serial/8250/8250_em.c
> > > b/drivers/tty/serial/8250/8250_em.c
> > > index 69cd3b611501..a90a5cc4234a 100644
> > > --- a/drivers/tty/serial/8250/8250_em.c
> > > +++ b/drivers/tty/serial/8250/8250_em.c
> > > @@ -19,10 +19,14 @@
> > >
> > >  #define UART_DLL_EM 9
> > >  #define UART_DLM_EM 10
> > > +#define UART_HCR0 11
> > > +
> > > +#define UART_HCR0_SW_RESET	BIT(7) /* SW Reset */
> > >
> > >  struct serial8250_em_hw_info {
> > >  	unsigned int type;
> > >  	upf_t flags;
> > > +	void (*serial_out)(struct uart_port *p, int off, int value);
> > >  };
> > >
> > >  struct serial8250_em_priv {
> > > @@ -46,6 +50,7 @@ static void serial8250_em_serial_out(struct uart_port
> > *p, int offset, int value)
> > >  		fallthrough;
> > >  	case UART_DLL_EM: /* DLL @ 0x24 (+9) */
> > >  	case UART_DLM_EM: /* DLM @ 0x28 (+9) */
> > > +	case UART_HCR0: /* HCR0 @ 0x2c */
> > >  		writel(value, p->membase + (offset << 2));
> > >  	}
> > >  }
> > > @@ -55,6 +60,7 @@ static unsigned int serial8250_em_serial_in(struct
> > uart_port *p, int offset)
> > >  	switch (offset) {
> > >  	case UART_RX: /* RX @ 0x00 */
> > >  		return readb(p->membase);
> > > +	case UART_LCR: /* LCR @ 0x10 (+1) */
> > >  	case UART_MCR: /* MCR @ 0x14 (+1) */
> > >  	case UART_LSR: /* LSR @ 0x18 (+1) */
> > >  	case UART_MSR: /* MSR @ 0x1c (+1) */ @@ -64,11 +70,68 @@ static
> > > unsigned int serial8250_em_serial_in(struct uart_port *p, int offset)
> > >  	case UART_IIR: /* IIR @ 0x08 */
> > >  	case UART_DLL_EM: /* DLL @ 0x24 (+9) */
> > >  	case UART_DLM_EM: /* DLM @ 0x28 (+9) */
> > > +	case UART_HCR0: /* HCR0 @ 0x2c */
> > >  		return readl(p->membase + (offset << 2));
> > >  	}
> > >  	return 0;
> > >  }
> > >
> > > +static void serial8250_rzv2m_reg_update(struct uart_port *p, int off,
> > > +int value) {
> > > +	unsigned int ier, fcr, lcr, mcr, hcr0;
> > > +
> > > +	ier = serial8250_em_serial_in(p, UART_IER);
> > > +	lcr = serial8250_em_serial_in(p, UART_LCR);
> > > +	mcr = serial8250_em_serial_in(p, UART_MCR);
> > > +	hcr0 = serial8250_em_serial_in(p, UART_HCR0);
> > > +	/*
> > > +	 * The value of UART_IIR and UART_FCR are 2, but corresponding
> > > +	 * RZ/V2M address offset are different(0x08 and 0x0c). So we need to
> > > +	 * use readl() here.
> > > +	 */
> > > +	fcr = readl(p->membase + ((UART_FCR + 1) << 2));
> > 
> > I don't get the meaning of that comment. It doesn't seem to match what your
> > code does as the code seemingly has nothing to do with IIR (and none of you
> > changelogs refer to IIR either)?
> 
> The generic macro UART_IIR and UART_FCR in linux/serial_reg.h has a value of 2.
> 
> The RZ/V2M register offset don't match with standard ones, we are doing remapping
> inside case statements in serial8250_em_serial_in() and serial8250_em_serial_out().
> 
> Currently IIR is already present in serial8250_em_serial_in(), so no issue.
> 
> But when I tried to add UART_FCR in case, it gives the compilation error.
> You cannot have same case values inside switch.
> 
> > 
> > Is the only reason for this that serial8250_em_serial_in() currently lacks
> > case UART_FCR: ? Why cannot that just be added there?
> 
> Yes, It gives build error. 

Ok, thanks for explaining. I guess adding case UART_FCR_EM: with 
the correct value has a risk colliding with UART_LCR reads? I'm actually a 
bit surprised serial8250_em_serial_in() doesn't map LCR at all.

One option would be to just invent some big (bogus) number for UART_FCR_EM 
and correct the offset in serial8250_em_serial_in() into 0x0c. And just
add a comment to the define that you cannot use the correct value to avoid
overlapping with existing register defines.

-- 
 i.