| 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 linux-hardening linux-cve-announce PHC | |
|
Open Source and information security mailing list archives
| ||
|
Message-ID: <202312271350242a208426@mail.local>
Date: Wed, 27 Dec 2023 14:50:24 +0100
From: Alexandre Belloni <alexandre.belloni@...tlin.com>
To: Jingbao Qiu <qiujingbao.dlmu@...il.com>
Cc: a.zummo@...ertech.it, robh+dt@...nel.org,
krzysztof.kozlowski+dt@...aro.org, conor@...nel.org,
conor+dt@...nel.org, chao.wei@...hgo.com, unicorn_wang@...look.com,
paul.walmsley@...ive.com, palmer@...belt.com, aou@...s.berkeley.edu,
linux-rtc@...r.kernel.org, devicetree@...r.kernel.org,
linux-kernel@...r.kernel.org, dlan@...too.org,
inochiama@...look.com
Subject: Re: [PATCH v3 3/4] rtc: sophgo: add rtc support for Sophgo CV1800 SoC
On 27/12/2023 16:03:56+0800, Jingbao Qiu wrote:
> On Tue, Dec 26, 2023 at 8:37 PM Alexandre Belloni
> <alexandre.belloni@...tlin.com> wrote:
> >
> > Hello,
> >
> > please run checkpatch.pl --strict, there are a few issues.
> >
> > On 26/12/2023 18:04:30+0800, Jingbao Qiu wrote:
> > > +struct cv1800_rtc_priv {
> > > + struct rtc_device *rtc_dev;
> > > + struct device *dev;
> > > + struct regmap *rtc_map;
> > > + struct clk *clk;
> > > + spinlock_t rtc_lock;
> >
> > This lock seems unnecessary, please check
> >
>
> you are right. I will fix it.
>
> > > + int irq;
> > > +};
> > > +
> > > +static int cv1800_rtc_alarm_irq_enable(struct device *dev, unsigned int enabled)
> > > +{
> > > + struct cv1800_rtc_priv *info = dev_get_drvdata(dev);
> > > +
> > > + if (enabled)
> > > + regmap_write(info->rtc_map, ALARM_ENABLE, REG_ENABLE_FUN);
> > > + else
> > > + regmap_write(info->rtc_map, ALARM_ENABLE, REG_DISABLE_FUN);
> > > +
> >
> > This could be:
> > regmap_write(info->rtc_map, ALARM_ENABLE, enabled);
>
> you are right, i will fix it.
>
> >
> > > + return 0;
> > > +}
> > > +
> > > +static int cv1800_rtc_set_alarm(struct device *dev, struct rtc_wkalrm *alrm)
> > > +{
> > > + struct cv1800_rtc_priv *info = dev_get_drvdata(dev);
> > > + unsigned long alarm_time;
> > > +
> > > + alarm_time = rtc_tm_to_time64(&alrm->time);
> > > +
> > > + if (alarm_time > SEC_MAX_VAL)
> > > + return -EINVAL;
> >
> > The core is already checking fr this.
>
> Thanks, I will remove it.
>
> >
> > > +
> > > + regmap_write(info->rtc_map, ALARM_ENABLE, REG_DISABLE_FUN);
> > > +
> > > + udelay(DEALY_TIME_PREPARE);
> >
> > Why is this needed?
>
> This doesn't seem to require waiting, I will check it.
>
> >
> > > +
> > > + regmap_write(info->rtc_map, ALARM_TIME, alarm_time);
> > > + regmap_write(info->rtc_map, ALARM_ENABLE, REG_ENABLE_FUN);
> >
> > You must follow alrm->enabled instead of unconditionally enabling the
> > alarm.
>
> ok,i will fix it.
>
> >
> > > +
> > > + return 0;
> > > +}
> > > +
> >
> >
> > > +static int cv1800_rtc_32k_coarse_val_calib(struct cv1800_rtc_priv *info)
> >
> > Please explain those two calibration functions. I don't think you can
> > achieve what you want to do.
>
> The goal of these two calibration functions is to achieve calibration
> of RTC time.
> The code is written according to the data manual.
>
> The calibration circuit uses 25MHz crystal clock to sample 32KHz
> clock. In coarse
> tune mode, the 25MHz crystal clock samples one 32KHz clock cycle period and
> report the counting results.
>
> the datasheet link:
> Link: https://github.com/milkv-duo/duo-files/blob/main/duo/datasheet/CV1800B-CV1801B-Preliminary-Datasheet-full-en.pdf
> page:195
I'm really curious as to why this is calibrated using a 25MHz crystal as
it may be as imprecise as the 32kHz one. I'm asking because we have an
interface to get calibration done properly so you can use a precise clock
like GPS, NTP or PTP. This is what you should probably implement
instead or on top of it.
> >
> > > +}
> > > +
> > > +static int cv1800_rtc_read_time(struct device *dev, struct rtc_time *tm)
> > > +{
> > > + struct cv1800_rtc_priv *info = dev_get_drvdata(dev);
> > > + unsigned int sec;
> > > + unsigned int sec_ro_t;
> > > + unsigned long flag;
> > > +
> > > + spin_lock_irqsave(&info->rtc_lock, flag);
> > > +
> > > + regmap_read(info->rtc_map, SEC_CNTR_VAL, &sec);
> > > + regmap_read(info->rtc_map, MACRO_RO_T, &sec_ro_t);
> > > +
> > > + if (sec_ro_t > (SET_SEC_CNTR_VAL_UPDATE)) {
> > > + sec = sec_ro_t;
> > > + regmap_write(info->rtc_map, SET_SEC_CNTR_VAL, sec);
> > > + regmap_write(info->rtc_map, SET_SEC_CNTR_TRIG, REG_ENABLE_FUN);
> >
> > What does this do?
>
> the sec_ro_t be considered to have high accuracy after calibration.
> So every time read the time, update the RTC time.
So why don't you always use sec_ro_t instead of sec?
Also, why is this done conditionally on a arbitrary value? As it stands,
it will happen if the date is after 1995-07-09T16:12:48 for no good
reason.
This is awful because the alarm is matching SEC_CNTR_VAL with ALARM_TIME
so if this means the calibration doesn't affect SEC_CNTR_VAL (which I
seriously doubt), the alarm will end up being imprecise anyway
> > > +static int cv1800_rtc_probe(struct platform_device *pdev)
> > > +{
> > > + struct cv1800_rtc_priv *rtc;
> > > + uint32_t ctrl_val;
> > > + int ret;
> > > +
> > > + rtc = devm_kzalloc(&pdev->dev, sizeof(struct cv1800_rtc_priv),
> > > + GFP_KERNEL);
> > > + if (!rtc)
> > > + return -ENOMEM;
> > > +
> > > + rtc->dev = &pdev->dev;
> > > +
> > > + rtc->rtc_map = syscon_node_to_regmap(rtc->dev->of_node->parent);
> > > + if (IS_ERR(rtc->rtc_map))
> > > + return PTR_ERR(rtc->rtc_map);
> > > +
> > > + rtc->irq = platform_get_irq(pdev, 0);
> > > + if (rtc->irq < 0)
> > > + return rtc->irq;
> > > +
> > > + ret = devm_request_irq(&pdev->dev, rtc->irq, cv1800_rtc_irq_handler,
> > > + IRQF_TRIGGER_HIGH, "alarm", &pdev->dev);
> > > + if (ret)
> > > + return dev_err_probe(&pdev->dev, ret,
> > > + "cannot register interrupt handler\n");
> > > +
> > > + rtc->clk = devm_clk_get(rtc->dev, NULL);
> > > + if (IS_ERR(rtc->clk))
> > > + return PTR_ERR(rtc->clk);
> > > +
> >
> > You are going to leak rtc->clk after the next call.
>
> I will release him at the appropriate time. And add the remove
> function to release.
>
> >
> > > + rtc->clk = devm_clk_get_enabled(&pdev->dev, NULL);
> > > + if (IS_ERR(rtc->clk))
> > > + return dev_err_probe(&pdev->dev, PTR_ERR(rtc->clk),
> > > + "clk not found\n");
> > > +
> > > + platform_set_drvdata(pdev, rtc);
> > > +
> > > + spin_lock_init(&rtc->rtc_lock);
> > > +
> > > + rtc->rtc_dev = devm_rtc_device_register(&pdev->dev,
> > > + dev_name(&pdev->dev),
> > > + &cv800b_rtc_ops,
> > > + THIS_MODULE);
> > > + if (IS_ERR(rtc->rtc_dev))
> > > + return dev_err_probe(&pdev->dev, PTR_ERR(rtc->rtc_dev),
> > > + "can't register rtc device\n");
> >
> > Please use devm_rtc_allocate_device and devm_rtc_register_device
>
> ok,I will use it.
Also you have to set the RTC range properly.
--
Alexandre Belloni, co-owner and COO, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com
Powered by blists - more mailing lists