Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 10 Jan 2023 12:07:34 -0600
From:   Bjorn Andersson <andersson@kernel.org>
To:     Sibi Sankar <quic_sibis@quicinc.com>
Cc:     agross@kernel.org, linux-arm-msm@vger.kernel.org,
        devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
        krzysztof.kozlowski+dt@linaro.org, robh+dt@kernel.org,
        konrad.dybcio@somainline.org, robimarko@gmail.com,
        quic_gurus@quicinc.com
Subject: Re: [PATCH V7 2/2] firmware: qcom: scm: Add wait-queue handling logic
Message-ID: <20230110180734.gwjaant3sd45ljdd@builder.lan>
References: <20230110063745.16739-1-quic_sibis@quicinc.com>
 <20230110063745.16739-3-quic_sibis@quicinc.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=utf-8
Content-Disposition: inline
Content-Transfer-Encoding: 8bit
In-Reply-To: <20230110063745.16739-3-quic_sibis@quicinc.com>
Precedence: bulk

On Tue, Jan 10, 2023 at 12:07:45PM +0530, Sibi Sankar wrote:
> From: Guru Das Srinagesh <quic_gurus@quicinc.com>
> 
> When the firmware (FW) supports multiple requests per VM, multiple requests
> from the same/different VM can reach the firmware at the same time. Since
> the firmware currently being used has limited resources, it guards them
> with a resource lock and puts requests on a wait-queue internally and
> signals to HLOS that it is doing so. It does this by returning a new return
> value in addition to success or error: SCM_WAITQ_SLEEP. A sleeping SCM call
> can be woken up by an interrupt that the FW raises.
> 
>   1) SCM_WAITQ_SLEEP:
> 
>   	When an SCM call receives this return value instead of success
>   	or error, FW has placed this call on a wait-queue and has signalled
> 	HLOS to put it to non-interruptible sleep.
> 
> 	Along with this return value, FW also passes to HLOS `wq_ctx` -
> 	a unique number (UID) identifying the wait-queue that it has put
> 	the call on, internally. This is to help HLOS with its own
> 	bookkeeping to wake this sleeping call later.
> 
> 	Additionally, FW also passes to HLOS `smc_call_ctx` - a UID
> 	identifying the SCM call thus being put to sleep. This is also
> 	for HLOS' bookkeeping to wake this call up later.
> 
> 	These two additional values are passed via the a1 and a2
> 	registers.
> 
> 	N.B.: The "ctx" in the above UID names = "context".
> 
> The handshake mechanism that HLOS uses to talk to FW about wait-queue
> operations involves two new SMC calls.
> 
>   1) get_wq_ctx():
> 
>     	Arguments: 	None
>     	Returns:	wq_ctx, flags, more_pending
> 
>     	Get the wait-queue context, and wake up either one or all of the
>     	sleeping SCM calls associated with that wait-queue.
> 
>     	Additionally, repeat this if there are more wait-queues that are
>     	ready to have their requests woken up (`more_pending`).
> 
>   2) wq_resume(smc_call_ctx):
> 
>   	Arguments:	smc_call_ctx
> 
>   	HLOS needs to issue this in response to receiving an
>   	IRQ, passing to FW the same smc_call_ctx that FW
>   	receives from HLOS via the get_wq_ctx() call.
> 
> (The mechanism to wake a SMC call back up is described in detail below)
> 
>  VM_1                     VM_2                            Firmware
>    │                        │                                 │
>    │                        │                                 │
>    │                        │                                 │
>    │                        │                                 │
>    │      REQUEST_1         │                                 │
>    ├────────────────────────┼─────────────────────────────────┤
>    │                        │                                 │
>    │                        │                              ┌──┼──┐
>    │                        │                              │  │  │
>    │                        │     REQUEST_2                │  │  │
>    │                        ├──────────────────────────────┼──┤  │
>    │                        │                              │  │  │Resource
>    │                        │                              │  │  │is busy
>    │                        │       {WQ_SLEEP}             │  │  │
>    │                        │◄─────────────────────────────┼──┤  │
>    │                        │  wq_ctx, smc_call_ctx        │  │  │
>    │                        │                              └──┼──┘
>    │   REQUEST_1 COMPLETE   │                                 │
>    │◄───────────────────────┼─────────────────────────────────┤
>    │                        │                                 │
>    │                        │         IRQ                     │
>    │                        │◄─-------------------------------│
>    │                        │                                 │
>    │                        │      get_wq_ctx()               │
>    │                        ├────────────────────────────────►│
>    │                        │                                 │
>    │                        │                                 │
>    │                        │◄────────────────────────────────┤
>    │                        │   wq_ctx, flags, and            │
>    │                        │        more_pending             │
>    │                        │                                 │
>    │                        │                                 │
>    │                        │ wq_resume(smc_call_ctx)         │
>    │                        ├────────────────────────────────►│
>    │                        │                                 │
>    │                        │                                 │
>    │                        │      REQUEST_2 COMPLETE         │
>    │                        │◄────────────────────────────────┤
>    │                        │                                 │
>    │                        │                                 │
> 
> With the exception of get_wq_ctx(), the other SMC call wq_resume() can
> return WQ_SLEEP (these nested rounds of WQ_SLEEP are not shown in the
> above diagram for the sake of simplicity). Therefore, introduce a new
> do-while loop to handle multiple WQ_SLEEP return values for the same
> parent SCM call.
> 
> Request Completion in the above diagram refers to either a success
> return value (zero) or error (and not SMC_WAITQ_SLEEP)
> 
> Also add the interrupt handler that wakes up a sleeping SCM call.
> 
> Signed-off-by: Guru Das Srinagesh <quic_gurus@quicinc.com>
> Co-developed-by: Sibi Sankar <quic_sibis@quicinc.com>
> Signed-off-by: Sibi Sankar <quic_sibis@quicinc.com>
> ---
> 
> v7:
> - Move lookup + wait_for_completion into a single function in qcom_scm [Bjorn]
> - Simplify completion retrieval [Bjorn]
> 
> v6:
> - Fix misc. nits in the scm driver [Krzysztof]
> 
> v5:
> - Handle the wake_one/wake_all flags [Guru]
> - Rename flag handler to qcom_scm_waitq_wakeup [Bjorn]
> - Resume scm call can return ebusy as well handle that scenario by retrying
>   the original smc call and not the resume call
> 
> v4:
> - platform_set_drvdata will be used by __scm_smc_do_quirk_handle_waitq to
>   get access to scm struct from device so retain it
> - Use a single completion as it satisfies all of the current usecases [Bjorn]
> - Inline scm_get_wq_ctx [Bjorn]
> - Convert all pr_err to dev_err [Bjorn]
> - Handle idr_destroy in a thread safe manner [Bjorn]
> - Misc. Style fixes [Bjorn]
> 
> v3:
> - Fixup irq handling so as not to affect SoCs without the interrupt.
> - Fix warnings reported by kernel test-bot.
> 
>  drivers/firmware/qcom_scm-smc.c | 90 ++++++++++++++++++++++++++++++---
>  drivers/firmware/qcom_scm.c     | 89 +++++++++++++++++++++++++++++++-
>  drivers/firmware/qcom_scm.h     |  8 +++
>  3 files changed, 179 insertions(+), 8 deletions(-)
> 
> diff --git a/drivers/firmware/qcom_scm-smc.c b/drivers/firmware/qcom_scm-smc.c
> index d111833364ba..30999f04749c 100644
> --- a/drivers/firmware/qcom_scm-smc.c
> +++ b/drivers/firmware/qcom_scm-smc.c
> @@ -52,29 +52,101 @@ static void __scm_smc_do_quirk(const struct arm_smccc_args *smc,
>  	} while (res->a0 == QCOM_SCM_INTERRUPTED);
>  }
>  
> -static void __scm_smc_do(const struct arm_smccc_args *smc,
> -			 struct arm_smccc_res *res, bool atomic)
> +static void fill_wq_resume_args(struct arm_smccc_args *resume, u32 smc_call_ctx)
>  {
> -	int retry_count = 0;
> +	memset(resume->args, 0, sizeof(resume->args[0]) * ARRAY_SIZE(resume->args));
> +
> +	resume->args[0] = ARM_SMCCC_CALL_VAL(ARM_SMCCC_STD_CALL,
> +					ARM_SMCCC_SMC_64, ARM_SMCCC_OWNER_SIP,
> +					SCM_SMC_FNID(QCOM_SCM_SVC_WAITQ, QCOM_SCM_WAITQ_RESUME));
> +
> +	resume->args[1] = QCOM_SCM_ARGS(1);
> +
> +	resume->args[2] = smc_call_ctx;
> +}
> +
> +int scm_get_wq_ctx(u32 *wq_ctx, u32 *flags, u32 *more_pending)
> +{
> +	int ret;
> +	struct arm_smccc_args get_wq_ctx = {0};
> +	struct arm_smccc_res get_wq_res;
> +
> +	get_wq_ctx.args[0] = ARM_SMCCC_CALL_VAL(ARM_SMCCC_STD_CALL,
> +				ARM_SMCCC_SMC_64, ARM_SMCCC_OWNER_SIP,
> +				SCM_SMC_FNID(QCOM_SCM_SVC_WAITQ, QCOM_SCM_WAITQ_GET_WQ_CTX));
> +
> +	/* Guaranteed to return only success or error, no WAITQ_* */
> +	__scm_smc_do_quirk(&get_wq_ctx, &get_wq_res);
> +	ret = get_wq_res.a0;
> +	if (ret)
> +		return ret;
> +
> +	*wq_ctx = get_wq_res.a1;
> +	*flags  = get_wq_res.a2;
> +	*more_pending = get_wq_res.a3;
> +
> +	return 0;
> +}
> +
> +static int __scm_smc_do_quirk_handle_waitq(struct device *dev, struct arm_smccc_args *waitq,
> +					   struct arm_smccc_res *res)
> +{
> +	int ret;
> +	struct arm_smccc_args resume;
> +	u32 wq_ctx, smc_call_ctx, flags;
> +	struct arm_smccc_args *smc = waitq;
> +
> +	do {
> +		__scm_smc_do_quirk(smc, res);
> +
> +		if (res->a0 == QCOM_SCM_WAITQ_SLEEP) {
> +			wq_ctx = res->a1;
> +			smc_call_ctx = res->a2;
> +			flags = res->a3;
> +
> +			if (!dev)
> +				return -EPROBE_DEFER;
> +
> +			ret = qcom_scm_lookup_completion(wq_ctx);
> +			if (ret)
> +				return ret;
> +
> +			fill_wq_resume_args(&resume, smc_call_ctx);
> +			smc = &resume;
> +		}
> +	} while (res->a0 == QCOM_SCM_WAITQ_SLEEP);
> +
> +	return 0;
> +}
> +
> +static int __scm_smc_do(struct device *dev, struct arm_smccc_args *smc,
> +			struct arm_smccc_res *res, bool atomic)
> +{
> +	int ret, retry_count = 0;
>  
>  	if (atomic) {
>  		__scm_smc_do_quirk(smc, res);
> -		return;
> +		return 0;
>  	}
>  
>  	do {
>  		mutex_lock(&qcom_scm_lock);
>  
> -		__scm_smc_do_quirk(smc, res);
> +		ret = __scm_smc_do_quirk_handle_waitq(dev, smc, res);
>  
>  		mutex_unlock(&qcom_scm_lock);
>  
> +		if (ret)
> +			return ret;
> +
>  		if (res->a0 == QCOM_SCM_V2_EBUSY) {
>  			if (retry_count++ > QCOM_SCM_EBUSY_MAX_RETRY)
>  				break;
>  			msleep(QCOM_SCM_EBUSY_WAIT_MS);
>  		}
>  	}  while (res->a0 == QCOM_SCM_V2_EBUSY);
> +
> +	return 0;
>  }
>  
>  
> @@ -83,7 +155,7 @@ int __scm_smc_call(struct device *dev, const struct qcom_scm_desc *desc,
>  		   struct qcom_scm_res *res, bool atomic)
>  {
>  	int arglen = desc->arginfo & 0xf;
> -	int i;
> +	int i, ret;
>  	dma_addr_t args_phys = 0;
>  	void *args_virt = NULL;
>  	size_t alloc_len;
> @@ -135,13 +207,17 @@ int __scm_smc_call(struct device *dev, const struct qcom_scm_desc *desc,
>  		smc.args[SCM_SMC_LAST_REG_IDX] = args_phys;
>  	}
>  
> -	__scm_smc_do(&smc, &smc_res, atomic);
> +	/* ret error check follows after args_virt cleanup*/
> +	ret = __scm_smc_do(dev, &smc, &smc_res, atomic);
>  
>  	if (args_virt) {
>  		dma_unmap_single(dev, args_phys, alloc_len, DMA_TO_DEVICE);
>  		kfree(args_virt);
>  	}
>  
> +	if (ret)
> +		return ret;
> +
>  	if (res) {
>  		res->result[0] = smc_res.a1;
>  		res->result[1] = smc_res.a2;
> diff --git a/drivers/firmware/qcom_scm.c b/drivers/firmware/qcom_scm.c
> index cdbfe54c8146..19ac506a9b1f 100644
> --- a/drivers/firmware/qcom_scm.c
> +++ b/drivers/firmware/qcom_scm.c
> @@ -4,6 +4,7 @@
>   */
>  #include <linux/platform_device.h>
>  #include <linux/init.h>
> +#include <linux/interrupt.h>
>  #include <linux/cpumask.h>
>  #include <linux/export.h>
>  #include <linux/dma-mapping.h>
> @@ -13,6 +14,7 @@
>  #include <linux/qcom_scm.h>
>  #include <linux/of.h>
>  #include <linux/of_address.h>
> +#include <linux/of_irq.h>
>  #include <linux/of_platform.h>
>  #include <linux/clk.h>
>  #include <linux/reset-controller.h>
> @@ -33,6 +35,7 @@ struct qcom_scm {
>  	struct clk *iface_clk;
>  	struct clk *bus_clk;
>  	struct icc_path *path;
> +	struct completion waitq_comp;
>  	struct reset_controller_dev reset;
>  
>  	/* control access to the interconnect path */
> @@ -63,6 +66,9 @@ static const u8 qcom_scm_cpu_warm_bits[QCOM_SCM_BOOT_MAX_CPUS] = {
>  	BIT(2), BIT(1), BIT(4), BIT(6)
>  };
>  
> +#define QCOM_SMC_WAITQ_FLAG_WAKE_ONE	BIT(0)
> +#define QCOM_SMC_WAITQ_FLAG_WAKE_ALL	BIT(1)
> +
>  static const char * const qcom_scm_convention_names[] = {
>  	[SMC_CONVENTION_UNKNOWN] = "unknown",
>  	[SMC_CONVENTION_ARM_32] = "smc arm 32",
> @@ -1325,11 +1331,79 @@ bool qcom_scm_is_available(void)
>  }
>  EXPORT_SYMBOL(qcom_scm_is_available);
>  
> +static struct completion *qcom_scm_lookup_wq(struct qcom_scm *scm, u32 wq_ctx)

This could either be seen as a remnant of the older versions of this
patch, or a signalling of "place implementation here". I dislike both...

Please rename it qcom_scm_assert_valid_wq_ctx() and just reference
&scm->waitq_comp directly in the two places below.

> +{
> +	/* assert wq_ctx is zero */
> +	if (wq_ctx != 0) {
> +		dev_err(scm->dev, "No waitqueue found for wq_ctx %d\n", wq_ctx);

I think you should be more specific here, perhaps:

	"Firmware unexpectedly passed non-zero wq_ctx\n"

And then as suggested a comment stating what needs to be done when that
happens.

> +		return ERR_PTR(-EINVAL);
> +	}
> +
> +	return &scm->waitq_comp;
> +}
> +
> +int qcom_scm_lookup_completion(u32 wq_ctx)
> +{
> +	struct completion *wq = NULL;

You assign it on the very next line, no need to do so here.

> +
> +	wq = qcom_scm_lookup_wq(__scm, wq_ctx);
> +	if (IS_ERR(wq))
> +		return PTR_ERR(wq);
> +
> +	wait_for_completion(wq);
> +
> +	return 0;
> +}
> +
> +static int qcom_scm_waitq_wakeup(struct qcom_scm *scm, unsigned int wq_ctx, bool wake_all)
> +{
> +	struct completion *wq_to_wake;

"wq" would be sufficient.

Thanks,
Bjorn

> +
> +	wq_to_wake = qcom_scm_lookup_wq(scm, wq_ctx);
> +	if (IS_ERR(wq_to_wake))
> +		return PTR_ERR(wq_to_wake);
> +
> +	if (wake_all)
> +		complete_all(wq_to_wake);
> +	else
> +		complete(wq_to_wake);
> +
> +	return 0;
> +}
> +
> +static irqreturn_t qcom_scm_irq_handler(int irq, void *data)
> +{
> +	int ret;
> +	struct qcom_scm *scm = data;
> +	u32 wq_ctx, flags, more_pending = 0;
> +
> +	do {
> +		ret = scm_get_wq_ctx(&wq_ctx, &flags, &more_pending);
> +		if (ret) {
> +			dev_err(scm->dev, "GET_WQ_CTX SMC call failed: %d\n", ret);
> +			goto out;
> +		}
> +
> +		if (flags != QCOM_SMC_WAITQ_FLAG_WAKE_ONE &&
> +		    flags != QCOM_SMC_WAITQ_FLAG_WAKE_ALL) {
> +			dev_err(scm->dev, "Invalid flags found for wq_ctx: %u\n", flags);
> +			goto out;
> +		}
> +
> +		ret = qcom_scm_waitq_wakeup(scm, wq_ctx, !!(flags & QCOM_SMC_WAITQ_FLAG_WAKE_ALL));
> +		if (ret)
> +			goto out;
> +	} while (more_pending);
> +
> +out:
> +	return IRQ_HANDLED;
> +}
> +
>  static int qcom_scm_probe(struct platform_device *pdev)
>  {
>  	struct qcom_scm *scm;
>  	unsigned long clks;
> -	int ret;
> +	int irq, ret;
>  
>  	scm = devm_kzalloc(&pdev->dev, sizeof(*scm), GFP_KERNEL);
>  	if (!scm)
> @@ -1402,6 +1476,19 @@ static int qcom_scm_probe(struct platform_device *pdev)
>  	__scm = scm;
>  	__scm->dev = &pdev->dev;
>  
> +	init_completion(&__scm->waitq_comp);
> +
> +	irq = platform_get_irq(pdev, 0);
> +	if (irq < 0) {
> +		if (irq != -ENXIO)
> +			return irq;
> +	} else {
> +		ret = devm_request_threaded_irq(__scm->dev, irq, NULL, qcom_scm_irq_handler,
> +						IRQF_ONESHOT, "qcom-scm", __scm);
> +		if (ret < 0)
> +			return dev_err_probe(scm->dev, ret, "Failed to request qcom-scm irq\n");
> +	}
> +
>  	__get_convention();
>  
>  	/*
> diff --git a/drivers/firmware/qcom_scm.h b/drivers/firmware/qcom_scm.h
> index db3d08a01209..018e9867d55a 100644
> --- a/drivers/firmware/qcom_scm.h
> +++ b/drivers/firmware/qcom_scm.h
> @@ -60,6 +60,9 @@ struct qcom_scm_res {
>  	u64 result[MAX_QCOM_SCM_RETS];
>  };
>  
> +int qcom_scm_lookup_completion(u32 wq_ctx);
> +int scm_get_wq_ctx(u32 *wq_ctx, u32 *flags, u32 *more_pending);
> +
>  #define SCM_SMC_FNID(s, c)	((((s) & 0xFF) << 8) | ((c) & 0xFF))
>  extern int __scm_smc_call(struct device *dev, const struct qcom_scm_desc *desc,
>  			  enum qcom_scm_convention qcom_convention,
> @@ -129,6 +132,10 @@ extern int scm_legacy_call(struct device *dev, const struct qcom_scm_desc *desc,
>  #define QCOM_SCM_SMMU_CONFIG_ERRATA1		0x03
>  #define QCOM_SCM_SMMU_CONFIG_ERRATA1_CLIENT_ALL	0x02
>  
> +#define QCOM_SCM_SVC_WAITQ			0x24
> +#define QCOM_SCM_WAITQ_RESUME			0x02
> +#define QCOM_SCM_WAITQ_GET_WQ_CTX		0x03
> +
>  /* common error codes */
>  #define QCOM_SCM_V2_EBUSY	-12
>  #define QCOM_SCM_ENOMEM		-5
> @@ -137,6 +144,7 @@ extern int scm_legacy_call(struct device *dev, const struct qcom_scm_desc *desc,
>  #define QCOM_SCM_EINVAL_ARG	-2
>  #define QCOM_SCM_ERROR		-1
>  #define QCOM_SCM_INTERRUPTED	1
> +#define QCOM_SCM_WAITQ_SLEEP	2
>  
>  static inline int qcom_scm_remap_error(int err)
>  {
> -- 
> 2.17.1
>