linux-kernel - Re: [PATCH v4 2/3] soc: qcom: qcom_stats: Add QMP support for syncing ddr stats

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
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <tqmugtxmauwo5o3lqilb3q5czpyl6uvs4wjcmq6pthni6eq7ky@ndzn6ovhnzdq>
Date: Fri, 30 May 2025 23:49:29 +0300
From: Dmitry Baryshkov <dmitry.baryshkov@....qualcomm.com>
To: Konrad Dybcio <konrad.dybcio@....qualcomm.com>
Cc: Maulik Shah <maulik.shah@....qualcomm.com>,
        Bjorn Andersson <andersson@...nel.org>,
        Konrad Dybcio <konradybcio@...nel.org>, Rob Herring <robh@...nel.org>,
        Krzysztof Kozlowski <krzk+dt@...nel.org>,
        Conor Dooley <conor+dt@...nel.org>, linux-arm-msm@...r.kernel.org,
        linux-kernel@...r.kernel.org, devicetree@...r.kernel.org,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        Doug Anderson <dianders@...omium.org>
Subject: Re: [PATCH v4 2/3] soc: qcom: qcom_stats: Add QMP support for
 syncing ddr stats

On Thu, May 29, 2025 at 09:07:49PM +0200, Konrad Dybcio wrote:
> On 5/28/25 1:02 PM, Dmitry Baryshkov wrote:
> > On Wed, May 28, 2025 at 02:51:32PM +0530, Maulik Shah wrote:
> >> Recent SoCs (SM8450 onwards) require QMP command to be sent before reading
> >> ddr stats. The duration field of ddr stats will get populated only if QMP
> >> command is sent.
> >>
> >> Add support to send ddr stats freqsync QMP command.
> >>
> >> Signed-off-by: Maulik Shah <maulik.shah@....qualcomm.com>
> >> ---
> >>  drivers/soc/qcom/qcom_stats.c | 34 +++++++++++++++++++++++++++++++++-
> >>  1 file changed, 33 insertions(+), 1 deletion(-)
> >>
> >> diff --git a/drivers/soc/qcom/qcom_stats.c b/drivers/soc/qcom/qcom_stats.c
> >> index 33fd2a1574464768bd07289e743fbb79ba415e84..0545c8cbefb8f18758d4eb51638e4ecb94e05422 100644
> >> --- a/drivers/soc/qcom/qcom_stats.c
> >> +++ b/drivers/soc/qcom/qcom_stats.c
> >> @@ -13,6 +13,7 @@
> >>  #include <linux/platform_device.h>
> >>  #include <linux/seq_file.h>
> >>  
> >> +#include <linux/soc/qcom/qcom_aoss.h>
> >>  #include <linux/soc/qcom/smem.h>
> >>  #include <clocksource/arm_arch_timer.h>
> >>  
> >> @@ -37,6 +38,8 @@
> >>  #define DDR_STATS_TYPE(data)		FIELD_GET(GENMASK(15, 8), data)
> >>  #define DDR_STATS_FREQ(data)		FIELD_GET(GENMASK(31, 16), data)
> >>  
> >> +static struct qmp *qcom_stats_qmp;
> >> +
> >>  struct subsystem_data {
> >>  	const char *name;
> >>  	u32 smem_item;
> >> @@ -188,12 +191,28 @@ static int qcom_ddr_stats_show(struct seq_file *s, void *d)
> >>  	struct ddr_stats_entry data[DDR_STATS_MAX_NUM_MODES];
> >>  	void __iomem *reg = (void __iomem *)s->private;
> >>  	u32 entry_count;
> >> -	int i;
> >> +	int i, ret;
> >>  
> >>  	entry_count = readl_relaxed(reg + DDR_STATS_NUM_MODES_ADDR);
> >>  	if (entry_count > DDR_STATS_MAX_NUM_MODES)
> >>  		return -EINVAL;
> >>  
> >> +	if (qcom_stats_qmp) {
> >> +		/*
> >> +		 * Recent SoCs (SM8450 onwards) do not have duration field
> >> +		 * populated from boot up onwards for both DDR LPM Stats
> >> +		 * and DDR Frequency Stats.
> >> +		 *
> >> +		 * Send QMP message to Always on processor which will
> >> +		 * populate duration field into MSG RAM area.
> >> +		 *
> >> +		 * Sent every time to read latest data.
> >> +		 */
> >> +		ret = qmp_send(qcom_stats_qmp, "{class: ddr, action: freqsync}");
> >> +		if (ret)
> >> +			return ret;
> >> +	}
> >> +
> >>  	reg += DDR_STATS_ENTRY_START_ADDR;
> >>  	memcpy_fromio(data, reg, sizeof(struct ddr_stats_entry) * entry_count);
> >>  
> >> @@ -304,6 +323,19 @@ static int qcom_stats_probe(struct platform_device *pdev)
> >>  
> >>  	for (i = 0; i < config->num_records; i++)
> >>  		d[i].appended_stats_avail = config->appended_stats_avail;
> >> +	/*
> >> +	 * QMP is used for DDR stats syncing to MSG RAM for recent SoCs (SM8450 onwards).
> >> +	 * The prior SoCs do not need QMP handle as the required stats are already present
> >> +	 * in MSG RAM, provided the DDR_STATS_MAGIC_KEY matches.
> >> +	 */
> >> +	qcom_stats_qmp = qmp_get(&pdev->dev);
> >> +	if (IS_ERR(qcom_stats_qmp)) {
> >> +		if (PTR_ERR(qcom_stats_qmp) == -EPROBE_DEFER)
> >> +			return -EPROBE_DEFER;
> >> +
> >> +		/* We assume any other error means it's not defined/needed */
> >> +		qcom_stats_qmp = NULL;
> > 
> > I still think that we shouldn't be ignoring actual errors here. I'd say,
> > check for of_property_present(dev->of_node, "qcom,qmp") before.
> 
> /**
>  * qmp_get() - get a qmp handle from a device
>  * @dev: client device pointer
>  *
>  * Return: handle to qmp device on success, ERR_PTR() on failure
>  */
> struct qmp *qmp_get(struct device *dev)
> {
>         struct platform_device *pdev;
>         struct device_node *np;
>         struct qmp *qmp;
> 
>         if (!dev || !dev->of_node)
>                 return ERR_PTR(-EINVAL);
> 
>         np = of_parse_phandle(dev->of_node, "qcom,qmp", 0);
>         if (!np)
>                 return ERR_PTR(-ENODEV);

So, I'd say, we need to identify whether it is this caluse or another
error. It should be enough to ignore ENODEV and return an error in all
other cases.

> 
>         pdev = of_find_device_by_node(np);
>         of_node_put(np);
>         if (!pdev)
>                 return ERR_PTR(-EINVAL);
> 
>         qmp = platform_get_drvdata(pdev);
> 
>         if (!qmp) {
>                 put_device(&pdev->dev);
>                 return ERR_PTR(-EPROBE_DEFER);
>         }
>         return qmp;
> }
> EXPORT_SYMBOL_GPL(qmp_get);
> 
> 
> Konrad

-- 
With best wishes
Dmitry