Re: [PATCH v5 08/10] fpga: m10bmc-sec: Differentiate rsu status from doorbell in csr map

[Ilpo Järvinen]

On Tue, 3 Jan 2023, Xu Yilun wrote:

> On 2022-12-30 at 12:23:18 +0200, Ilpo Järvinen wrote:
> > On Fri, 30 Dec 2022, Xu Yilun wrote:
> > 
> > > On 2022-12-26 at 19:58:47 +0200, Ilpo Järvinen wrote:
> > > > The rsu_status field moves from the doorbell register to the auth
> > > > result register in the PMCI implementation of the MAX10 BMC. Refactor
> > > > the sec update driver code to handle two distinct registers (rsu_status
> > > > field was added into csr map already when it was introduced but it was
> > > > unused until now).
> > > > 
> > > > Co-developed-by: Tianfei zhang <tianfei.zhang@xxxxxxxxx>
> > > > Signed-off-by: Tianfei zhang <tianfei.zhang@xxxxxxxxx>
> > > > Co-developed-by: Russ Weight <russell.h.weight@xxxxxxxxx>
> > > > Signed-off-by: Russ Weight <russell.h.weight@xxxxxxxxx>
> > > > Signed-off-by: Ilpo Järvinen <ilpo.jarvinen@xxxxxxxxxxxxxxx>
> > > > ---
> > > >  drivers/fpga/intel-m10-bmc-sec-update.c | 68 ++++++++++++++++---------
> > > >  include/linux/mfd/intel-m10-bmc.h       |  2 +-
> > > >  2 files changed, 46 insertions(+), 24 deletions(-)
> > > > 
> > > > diff --git a/drivers/fpga/intel-m10-bmc-sec-update.c b/drivers/fpga/intel-m10-bmc-sec-update.c
> > > > index 6e58a463619c..1fe8b7ff594c 100644
> > > > --- a/drivers/fpga/intel-m10-bmc-sec-update.c
> > > > +++ b/drivers/fpga/intel-m10-bmc-sec-update.c
> > > > @@ -251,7 +251,7 @@ static void log_error_regs(struct m10bmc_sec *sec, u32 doorbell)
> > > >  	const struct m10bmc_csr_map *csr_map = sec->m10bmc->info->csr_map;
> > > >  	u32 auth_result;
> > > >  
> > > > -	dev_err(sec->dev, "RSU error status: 0x%08x\n", doorbell);
> > > > +	dev_err(sec->dev, "Doorbell: 0x%08x\n", doorbell);
> > > >  
> > > >  	if (!m10bmc_sys_read(sec->m10bmc, csr_map->auth_result, &auth_result))
> > > >  		dev_err(sec->dev, "RSU auth result: 0x%08x\n", auth_result);
> > > > @@ -279,6 +279,30 @@ static bool rsu_progress_busy(u32 progress)
> > > >  		progress == RSU_PROG_PROGRAM_KEY_HASH);
> > > >  }
> > > >  
> > > > +static int m10bmc_sec_progress_status(struct m10bmc_sec *sec, u32 *doorbell,
> > > 
> > > Please try to rename the parameters, to indicate u32 *doorbell is the
> > > raw value from doorbell register, and u32 *progress & status are
> > > software managed info.
> > 
> > I'll try to do that.
> >  
> > > > +				      u32 *progress, u32 *status)
> > > > +{
> > > > +	const struct m10bmc_csr_map *csr_map = sec->m10bmc->info->csr_map;
> > > > +	u32 status_reg;
> > > > +	int ret;
> > > > +
> > > > +	ret = m10bmc_sys_read(sec->m10bmc, csr_map->doorbell, doorbell);
> > > > +	if (ret)
> > > > +		return ret;
> > > > +
> > > > +	if (csr_map->doorbell != csr_map->rsu_status) {
> > > 
> > > I prefer not to complicate the csr map filling in intel-m10-bmc, just invalid
> > > the addr value if there is no such register for the board.
> > 
> > I'm sorry but I didn't get the meaning of your comment. Could you please 
> > rephrase?
> > 
> > My guess is that you might have tried to say that if there's no register 
> > for rsu_status, mark it not existing in csr map? But the field exists in 
> 
> Yes, this is what I mean, but I see I was wrong.
> 
> > both cases, it's just part of a different register (doorbell or 
> 
> I was thinking there was no AUTH_RESULT for N3000, sorry for the
> mistake.
> 
> > auth_result) so if I use that kind of "register doesn't exist" condition, 
> > it would apply to both cases.
> > 
> > > > @@ -330,21 +350,20 @@ static enum fw_upload_err rsu_update_init(struct m10bmc_sec *sec)
> > > >  	if (ret)
> > > >  		return FW_UPLOAD_ERR_RW_ERROR;
> > > >  
> > > > -	ret = regmap_read_poll_timeout(sec->m10bmc->regmap,
> > > > -				       csr_map->base + csr_map->doorbell,
> > > > -				       doorbell,
> > > > -				       rsu_start_done(doorbell),
> > > > -				       NIOS_HANDSHAKE_INTERVAL_US,
> > > > -				       NIOS_HANDSHAKE_TIMEOUT_US);
> > > > +	ret = read_poll_timeout(m10bmc_sec_progress_status, err,
> > > > +				err < 0 || rsu_start_done(doorbell, progress, status),
> > > > +				NIOS_HANDSHAKE_INTERVAL_US,
> > > > +				NIOS_HANDSHAKE_TIMEOUT_US,
> > > > +				false,
> > > > +				sec, &doorbell, &progress, &status);
> > > >  
> > > >  	if (ret == -ETIMEDOUT) {
> > > >  		log_error_regs(sec, doorbell);
> > > >  		return FW_UPLOAD_ERR_TIMEOUT;
> > > > -	} else if (ret) {
> > > > +	} else if (err) {
> > > >  		return FW_UPLOAD_ERR_RW_ERROR;
> > > >  	}
> > > >  
> > > > -	status = rsu_stat(doorbell);
> > > >  	if (status == RSU_STAT_WEAROUT) {
> > > >  		dev_warn(sec->dev, "Excessive flash update count detected\n");
> > > >  		return FW_UPLOAD_ERR_WEAROUT;
> > > > @@ -393,7 +412,7 @@ static enum fw_upload_err rsu_prog_ready(struct m10bmc_sec *sec)
> > > >  static enum fw_upload_err rsu_send_data(struct m10bmc_sec *sec)
> > > >  {
> > > >  	const struct m10bmc_csr_map *csr_map = sec->m10bmc->info->csr_map;
> > > > -	u32 doorbell;
> > > > +	u32 doorbell, status;
> > > >  	int ret;
> > > >  
> > > >  	ret = regmap_update_bits(sec->m10bmc->regmap,
> > > > @@ -418,7 +437,10 @@ static enum fw_upload_err rsu_send_data(struct m10bmc_sec *sec)
> > > >  		return FW_UPLOAD_ERR_RW_ERROR;
> > > >  	}
> > > >  
> > > > -	if (!rsu_status_ok(rsu_stat(doorbell))) {
> > > > +	ret = m10bmc_sys_read(sec->m10bmc, csr_map->rsu_status, &status);
> > > 
> > > Same as above, please just handle the detailed register definition 
> > > differences in this driver, not in csr map.
> > 
> > Earlier you were having the exactly opposite opinion:
> > 
> > https://lore.kernel.org/linux-fpga/20221108144305.45424-1-ilpo.jarvinen@xxxxxxxxxxxxxxx/T/#me2d20e60d7feeafcdeeab4d58bd82787acf3ada9
> 
> Ah, I'm sorry. I was thinking just move one register to another addr at
> that time. I was not aware that actually the detailed register field
> definitions are changed in same registers.
> 
> > 
> > So which way you want it? Should I have the board types here in the sec 
> > update drivers as a second layer of differentiation or not?
> 
> I think the different register field definitions for the same registers
> are specific to secure driver. So please differentiate them in secure
> driver.
> 
> But with the change, enum m10bmc_type could still be removed, is it?
>
> And having the register addr differentiations in m10bmc mfd driver is good to
> me, cause with a different board type, the register offsets for all subdevs
> are often globally re-arranged. But I don't want the HW change within a
> single IP block been specified in m10bmc mfd driver.

Okay. I'll add ops for it then:

struct m10bmc_sec_ops {
       int (*rsu_status)(struct m10bmc_sec *sec);
};

Type enum won't be necessary. Those ops will be useful for other things 
too which are not included to this patch set.

-- 
 i.