Linux-remoteproc Archive mirror
 help / color / mirror / Atom feed
From: neil.armstrong@linaro.org
To: Mukesh Ojha <quic_mojha@quicinc.com>,
	Bjorn Andersson <andersson@kernel.org>,
	Mathieu Poirier <mathieu.poirier@linaro.org>,
	Rob Herring <robh@kernel.org>,
	Krzysztof Kozlowski <krzk+dt@kernel.org>,
	Conor Dooley <conor+dt@kernel.org>,
	Konrad Dybcio <konradybcio@kernel.org>,
	Bartosz Golaszewski <bartosz.golaszewski@linaro.org>,
	Manivannan Sadhasivam <manivannan.sadhasivam@linaro.org>
Cc: linux-arm-msm@vger.kernel.org, linux-remoteproc@vger.kernel.org,
	devicetree@vger.kernel.org, linux-kernel@vger.kernel.org,
	Shiraz Hashim <quic_shashim@quicinc.com>
Subject: Re: [PATCH 3/6] remoteproc: qcom: Add helper function to support IOMMU devmem translation
Date: Mon, 7 Oct 2024 10:08:16 +0200	[thread overview]
Message-ID: <83e23090-0390-4c2e-91e3-e222baaa889a@linaro.org> (raw)
In-Reply-To: <20241004212359.2263502-4-quic_mojha@quicinc.com>

On 04/10/2024 23:23, Mukesh Ojha wrote:
> From: Shiraz Hashim <quic_shashim@quicinc.com>
> 
> Qualcomm SoCs runnning with Qualcomm EL2 hypervisor(QHEE), IOMMU
> translation set up for remote processors is managed by QHEE itself
> however, for a case when these remote processors has to run under KVM

This is not true, KVM is a Linux hypervisor, remote processors have
nothing to do with KVM, please rephrase.

> hypervisor, IOMMU translation need to setup from Linux remoteproc driver
> before it is brought up.
> 
> Add qcom_devmem_info and qcom_devmem_table data structure and make a
> common helper functions which caller can call if these translation need
> to be taken care by the driver to enable iommu devmem access for
> remoteproc processors.
> 
> Signed-off-by: Shiraz Hashim <quic_shashim@quicinc.com>
> Co-developed-by: Mukesh Ojha <quic_mojha@quicinc.com>
> Signed-off-by: Mukesh Ojha <quic_mojha@quicinc.com>
> ---
>   drivers/remoteproc/qcom_common.c | 96 ++++++++++++++++++++++++++++++++
>   drivers/remoteproc/qcom_common.h | 35 ++++++++++++
>   2 files changed, 131 insertions(+)
> 
> diff --git a/drivers/remoteproc/qcom_common.c b/drivers/remoteproc/qcom_common.c
> index 1c7887dc65b4..644920972b58 100644
> --- a/drivers/remoteproc/qcom_common.c
> +++ b/drivers/remoteproc/qcom_common.c
> @@ -658,5 +658,101 @@ int qcom_map_unmap_carveout(struct rproc *rproc, phys_addr_t mem_phys, size_t me
>   }
>   EXPORT_SYMBOL_GPL(qcom_map_unmap_carveout);
>   
> +/**
> + * qcom_map_devmem() - Map the device memories needed by Remoteproc using IOMMU
> + *
> + * When Qualcomm EL2 hypervisor(QHEE) present, device memories needed for remoteproc
> + * processors is managed by it and Linux remoteproc drivers should not call
> + * this and its respective unmap function in such scenario. This function
> + * should only be called if remoteproc IOMMU translation need to be managed
> + * from Linux side.
> + *
> + * @rproc: rproc handle
> + * @devmem_table: list of devmem regions to map
> + * @use_sid: decision to append sid to iova
> + * @sid: SID value
> + */
> +int qcom_map_devmem(struct rproc *rproc, struct qcom_devmem_table *devmem_table,
> +		    bool use_sid, unsigned long sid)
> +{
> +	struct qcom_devmem_info *info;
> +	unsigned long sid_def_val;
> +	int ret;
> +	int i;
> +
> +	if (!rproc->has_iommu)
> +		return 0;
> +
> +	if (!rproc->domain)
> +		return -EINVAL;
> +
> +	/* remoteproc may not have devmem data */
> +	if (!devmem_table)
> +		return 0;
> +
> +	if (use_sid && sid)
> +		sid_def_val = sid & SID_MASK_DEFAULT;
> +
> +	info = &devmem_table->entries[0];
> +	for (i = 0; i < devmem_table->num_entries; i++, info++) {
> +		/*
> +		 * Remote processor like ADSP supports upto 36 bit device
> +		 * address space and some of its clients like fastrpc uses
> +		 * upper 32-35 bits to keep lower 4 bits of its SID to use
> +		 * larger address space. To keep this consistent across other
> +		 * use cases add remoteproc SID configuration for firmware
> +		 * to IOMMU for carveouts.
> +		 */
> +		if (use_sid)
> +			info->da |= (sid_def_val << 32);
> +
> +		ret = iommu_map(rproc->domain, info->da, info->pa, info->len, info->flags, GFP_KERNEL);
> +		if (ret) {
> +			dev_err(&rproc->dev, "Unable to map devmem, ret: %d\n", ret);
> +			if (use_sid)
> +				info->da &= ~(SID_MASK_DEFAULT << 32);
> +			goto undo_mapping;
> +		}
> +	}
> +
> +	return 0;
> +
> +undo_mapping:
> +	for (i = i - 1; i >= 0; i--, info--) {
> +		iommu_unmap(rproc->domain, info->da, info->len);
> +		if (use_sid)
> +			info->da &= ~(SID_MASK_DEFAULT << 32);
> +	}
> +
> +	return ret;
> +}
> +EXPORT_SYMBOL_GPL(qcom_map_devmem);
> +
> +/**
> + * qcom_unmap_devmem() -  unmap the device memories needed by Remoteproc using IOMMU
> + *
> + * @rproc:		rproc handle
> + * @devmem_table:	list of devmem regions to unmap
> + * @use_sid:		decision to append sid to iova
> + */
> +void qcom_unmap_devmem(struct rproc *rproc, struct qcom_devmem_table *devmem_table, bool use_sid)
> +{
> +	struct qcom_devmem_info *info;
> +	int i;
> +
> +	if (!rproc->has_iommu || !rproc->domain || !devmem_table)
> +		return;
> +
> +	info = &devmem_table->entries[0];
> +	for (i = 0; i < devmem_table->num_entries; i++, info++) {
> +		iommu_unmap(rproc->domain, info->da, info->len);
> +		if (use_sid)
> +			info->da &= ~(SID_MASK_DEFAULT << 32);
> +	}
> +
> +	return;
> +}
> +EXPORT_SYMBOL_GPL(qcom_unmap_devmem);
> +
>   MODULE_DESCRIPTION("Qualcomm Remoteproc helper driver");
>   MODULE_LICENSE("GPL v2");
> diff --git a/drivers/remoteproc/qcom_common.h b/drivers/remoteproc/qcom_common.h
> index bbc41054e1ea..bbc684e1df01 100644
> --- a/drivers/remoteproc/qcom_common.h
> +++ b/drivers/remoteproc/qcom_common.h
> @@ -41,6 +41,36 @@ struct qcom_rproc_pdm {
>   	struct auxiliary_device *adev;
>   };
>   
> +/**
> + * struct qcom_devmem_info - iommu devmem region
> + * @da: device address
> + * @pa: physical address
> + * @len: length (in bytes)
> + * @flags: iommu protection flags
> + *
> + * The resource entry carries the device address to which a physical address is
> + * to be mapped with required permissions in flag. The pa, len is expected to
> + * be a physically contiguous memory region.
> + */
> +struct qcom_devmem_info {
> +	u64 da;
> +	u64 pa;
> +	u32 len;
> +	u32 flags;
> +};
> +
> +/**
> + * struct qcom_devmem_table - iommu devmem entries
> + * @num_entries: number of devmem entries
> + * @entries: devmem entries
> + *
> + * The table that carries each devmem resource entry.
> + */
> +struct qcom_devmem_table {
> +	int num_entries;
> +	struct qcom_devmem_info entries[0];
> +};
> +
>   void qcom_minidump(struct rproc *rproc, unsigned int minidump_id,
>   			void (*rproc_dumpfn_t)(struct rproc *rproc,
>   				struct rproc_dump_segment *segment, void *dest, size_t offset,
> @@ -65,6 +95,11 @@ int qcom_map_unmap_carveout(struct rproc *rproc, phys_addr_t mem_phys, size_t me
>   void qcom_add_pdm_subdev(struct rproc *rproc, struct qcom_rproc_pdm *pdm);
>   void qcom_remove_pdm_subdev(struct rproc *rproc, struct qcom_rproc_pdm *pdm);
>   
> +int qcom_map_devmem(struct rproc *rproc, struct qcom_devmem_table *table,
> +		    bool use_sid, unsigned long sid);
> +void qcom_unmap_devmem(struct rproc *rproc, struct qcom_devmem_table *table,
> +		       bool use_sid);
> +
>   #if IS_ENABLED(CONFIG_QCOM_SYSMON)
>   struct qcom_sysmon *qcom_add_sysmon_subdev(struct rproc *rproc,
>   					   const char *name,


  reply	other threads:[~2024-10-07  8:08 UTC|newest]

Thread overview: 38+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2024-10-04 21:23 [PATCH 0/6] Peripheral Image Loader support for Qualcomm SoCs Mukesh Ojha
2024-10-04 21:23 ` [PATCH 1/6] dt-bindings: remoteproc: qcom,pas-common: Introduce iommus and qcom,devmem property Mukesh Ojha
2024-10-06 19:38   ` Dmitry Baryshkov
2024-10-07 15:35     ` Mukesh Ojha
2024-10-07 16:25       ` Dmitry Baryshkov
2024-10-09 14:04         ` Shiraz Hashim
2024-10-10  7:15           ` Krzysztof Kozlowski
2024-10-10  8:30             ` Shiraz Hashim
2024-10-04 21:23 ` [PATCH 2/6] remoteproc: qcom: Add iommu map_unmap helper function Mukesh Ojha
2024-10-06  2:04   ` kernel test robot
2024-10-06  4:29   ` kernel test robot
2024-10-04 21:23 ` [PATCH 3/6] remoteproc: qcom: Add helper function to support IOMMU devmem translation Mukesh Ojha
2024-10-07  8:08   ` neil.armstrong [this message]
2024-10-07 14:37     ` Mukesh Ojha
2024-10-10  6:59       ` neil.armstrong
2024-10-17 21:25         ` Konrad Dybcio
2024-10-04 21:23 ` [PATCH 4/6] remoteproc: qcom: Add support to parse qcom,devmem property Mukesh Ojha
2024-10-04 21:23 ` [PATCH 5/6] remoteproc: qcom: Add support of SHM bridge to enable memory protection Mukesh Ojha
2024-10-05 22:26   ` kernel test robot
2024-10-25 19:03   ` Konrad Dybcio
2024-10-04 21:23 ` [PATCH 6/6] remoteproc: qcom: Enable map/unmap and SHM bridge support Mukesh Ojha
2024-10-07  8:05   ` neil.armstrong
2024-10-07 14:52     ` Mukesh Ojha
2024-10-08  6:21       ` Mukesh Ojha
2024-10-10  6:57         ` neil.armstrong
2024-10-11  5:05           ` Shiraz Hashim
2024-10-11  6:23             ` Dmitry Baryshkov
2024-10-11  7:09               ` Shiraz Hashim
2024-10-11  7:12                 ` Dmitry Baryshkov
2024-10-14 12:31                   ` Shiraz Hashim
2024-10-14 12:38                     ` Dmitry Baryshkov
2024-10-11  7:11               ` Mukesh Ojha
2024-10-11  7:09             ` neil.armstrong
2024-10-14 12:29               ` Shiraz Hashim
2024-10-25 19:07   ` Konrad Dybcio
2024-10-06 19:34 ` [PATCH 0/6] Peripheral Image Loader support for Qualcomm SoCs Dmitry Baryshkov
2024-10-07 18:39   ` Mukesh Ojha
2024-10-09 13:50   ` Shiraz Hashim

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=83e23090-0390-4c2e-91e3-e222baaa889a@linaro.org \
    --to=neil.armstrong@linaro.org \
    --cc=andersson@kernel.org \
    --cc=bartosz.golaszewski@linaro.org \
    --cc=conor+dt@kernel.org \
    --cc=devicetree@vger.kernel.org \
    --cc=konradybcio@kernel.org \
    --cc=krzk+dt@kernel.org \
    --cc=linux-arm-msm@vger.kernel.org \
    --cc=linux-kernel@vger.kernel.org \
    --cc=linux-remoteproc@vger.kernel.org \
    --cc=manivannan.sadhasivam@linaro.org \
    --cc=mathieu.poirier@linaro.org \
    --cc=quic_mojha@quicinc.com \
    --cc=quic_shashim@quicinc.com \
    --cc=robh@kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).