Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Jason Gunthorpe]

On Thu, Jan 05, 2023 at 05:37:10PM -0500, Zhu Yanjun wrote:
> From: Zhu Yanjun <yanjun.zhu@linux.dev>
> 
> This is a followup to the EFA dmabuf[1]. Irdma driver currently does
> not support on-demand-paging(ODP). So it uses habanalabs as the
> dmabuf exporter, and irdma as the importer to allow for peer2peer
> access through libibverbs.
> 
> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
> This function is introduced in EFA dmabuf[1] which allows the driver
> to get a dmabuf umem which is pinned and does not require move_notify
> callback implementation. The returned umem is pinned and DMA mapped
> like standard cpu umems, and is released through ib_umem_release().
> 
> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
> 
> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
> ---
> V2->V3: Simplify the function by removing QP and CQ handling;
> V1->V2: Fix the build warning by adding a static;
> ---
>  drivers/infiniband/hw/irdma/verbs.c | 97 +++++++++++++++++++++++++++++
>  1 file changed, 97 insertions(+)
> 
> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
> index f6973ea55eda..7028b8af87b9 100644
> --- a/drivers/infiniband/hw/irdma/verbs.c
> +++ b/drivers/infiniband/hw/irdma/verbs.c
> @@ -2912,6 +2912,102 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
>  	return ERR_PTR(err);
>  }
>  
> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
> +					      u64 len, u64 virt,
> +					      int fd, int access,
> +					      struct ib_udata *udata)
> +{
> +	struct irdma_device *iwdev = to_iwdev(pd->device);
> +	struct irdma_pble_alloc *palloc;
> +	struct irdma_pbl *iwpbl;
> +	struct irdma_mr *iwmr;
> +	u32 stag = 0;
> +	bool use_pbles = false;
> +	int err = -EINVAL;
> +	struct ib_umem_dmabuf *umem_dmabuf;
> +
> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
> +		return ERR_PTR(-EINVAL);
> +
> +	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd,
> +						access);
> +	if (IS_ERR(umem_dmabuf)) {
> +		err = PTR_ERR(umem_dmabuf);
> +		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n", err);
> +		return ERR_PTR(err);
> +	}
> +
> +	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
> +	if (!iwmr) {
> +		ib_umem_release(&umem_dmabuf->umem);
> +		return ERR_PTR(-ENOMEM);
> +	}

again, please don't duplicate all this code, refactor the mr code so
it can be shared.

Jason

[PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Zhu Yanjun]

From: Zhu Yanjun <yanjun.zhu@linux.dev>

This is a followup to the EFA dmabuf[1]. Irdma driver currently does
not support on-demand-paging(ODP). So it uses habanalabs as the
dmabuf exporter, and irdma as the importer to allow for peer2peer
access through libibverbs.

In this commit, the function ib_umem_dmabuf_get_pinned() is used.
This function is introduced in EFA dmabuf[1] which allows the driver
to get a dmabuf umem which is pinned and does not require move_notify
callback implementation. The returned umem is pinned and DMA mapped
like standard cpu umems, and is released through ib_umem_release().

[1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/

Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
---
V2->V3: Simplify the function by removing QP and CQ handling;
V1->V2: Fix the build warning by adding a static;
---
 drivers/infiniband/hw/irdma/verbs.c | 97 +++++++++++++++++++++++++++++
 1 file changed, 97 insertions(+)

diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
index f6973ea55eda..7028b8af87b9 100644
--- a/drivers/infiniband/hw/irdma/verbs.c
+++ b/drivers/infiniband/hw/irdma/verbs.c
@@ -2912,6 +2912,102 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
 	return ERR_PTR(err);
 }
 
+static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
+					      u64 len, u64 virt,
+					      int fd, int access,
+					      struct ib_udata *udata)
+{
+	struct irdma_device *iwdev = to_iwdev(pd->device);
+	struct irdma_pble_alloc *palloc;
+	struct irdma_pbl *iwpbl;
+	struct irdma_mr *iwmr;
+	u32 stag = 0;
+	bool use_pbles = false;
+	int err = -EINVAL;
+	struct ib_umem_dmabuf *umem_dmabuf;
+
+	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
+		return ERR_PTR(-EINVAL);
+
+	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
+		return ERR_PTR(-EINVAL);
+
+	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd,
+						access);
+	if (IS_ERR(umem_dmabuf)) {
+		err = PTR_ERR(umem_dmabuf);
+		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n", err);
+		return ERR_PTR(err);
+	}
+
+	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
+	if (!iwmr) {
+		ib_umem_release(&umem_dmabuf->umem);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	iwpbl = &iwmr->iwpbl;
+	iwpbl->iwmr = iwmr;
+	iwmr->region = &umem_dmabuf->umem;
+	iwmr->ibmr.pd = pd;
+	iwmr->ibmr.device = pd->device;
+	iwmr->ibmr.iova = virt;
+
+	iwmr->page_size = ib_umem_find_best_pgsz(iwmr->region,
+						 iwdev->rf->sc_dev.hw_attrs.page_size_cap,
+						 virt);
+	if (unlikely(!iwmr->page_size)) {
+		kfree(iwmr);
+		ib_umem_release(iwmr->region);
+		return ERR_PTR(-EOPNOTSUPP);
+	}
+
+	iwmr->len = iwmr->region->length;
+	iwpbl->user_base = virt;
+	palloc = &iwpbl->pble_alloc;
+	iwmr->type = IRDMA_MEMREG_TYPE_MEM;
+	iwmr->page_cnt = ib_umem_num_dma_blocks(iwmr->region, iwmr->page_size);
+
+	use_pbles = (iwmr->page_cnt != 1);
+
+	err = irdma_setup_pbles(iwdev->rf, iwmr, use_pbles, false);
+	if (err)
+		goto error;
+
+	if (use_pbles) {
+		err = irdma_check_mr_contiguous(palloc,	iwmr->page_size);
+		if (err) {
+			irdma_free_pble(iwdev->rf->pble_rsrc, palloc);
+			iwpbl->pbl_allocated = false;
+		}
+	}
+
+	stag = irdma_create_stag(iwdev);
+	if (!stag) {
+		err = -ENOMEM;
+		goto error;
+	}
+
+	iwmr->stag = stag;
+	iwmr->ibmr.rkey = stag;
+	iwmr->ibmr.lkey = stag;
+	err = irdma_hwreg_mr(iwdev, iwmr, access);
+	if (err) {
+		irdma_free_stag(iwdev, stag);
+		goto error;
+	}
+
+	return &iwmr->ibmr;
+
+error:
+	if (palloc->level != PBLE_LEVEL_0 && iwpbl->pbl_allocated)
+		irdma_free_pble(iwdev->rf->pble_rsrc, palloc);
+	kfree(iwmr);
+	ib_umem_release(&umem_dmabuf->umem);
+
+	return ERR_PTR(err);
+}
+
 /**
  * irdma_reg_phys_mr - register kernel physical memory
  * @pd: ibpd pointer
@@ -4418,6 +4514,7 @@ static const struct ib_device_ops irdma_dev_ops = {
 	.query_port = irdma_query_port,
 	.query_qp = irdma_query_qp,
 	.reg_user_mr = irdma_reg_user_mr,
+	.reg_user_mr_dmabuf = irdma_reg_user_mr_dmabuf,
 	.req_notify_cq = irdma_req_notify_cq,
 	.resize_cq = irdma_resize_cq,
 	INIT_RDMA_OBJ_SIZE(ib_pd, irdma_pd, ibpd),
-- 
2.34.1

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Yanjun Zhu]

在 2023/1/5 21:37, Jason Gunthorpe 写道:
> On Thu, Jan 05, 2023 at 05:37:10PM -0500, Zhu Yanjun wrote:
>> From: Zhu Yanjun <yanjun.zhu@linux.dev>
>>
>> This is a followup to the EFA dmabuf[1]. Irdma driver currently does
>> not support on-demand-paging(ODP). So it uses habanalabs as the
>> dmabuf exporter, and irdma as the importer to allow for peer2peer
>> access through libibverbs.
>>
>> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
>> This function is introduced in EFA dmabuf[1] which allows the driver
>> to get a dmabuf umem which is pinned and does not require move_notify
>> callback implementation. The returned umem is pinned and DMA mapped
>> like standard cpu umems, and is released through ib_umem_release().
>>
>> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
>>
>> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
>> ---
>> V2->V3: Simplify the function by removing QP and CQ handling;
>> V1->V2: Fix the build warning by adding a static;
>> ---
>>   drivers/infiniband/hw/irdma/verbs.c | 97 +++++++++++++++++++++++++++++
>>   1 file changed, 97 insertions(+)
>>
>> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
>> index f6973ea55eda..7028b8af87b9 100644
>> --- a/drivers/infiniband/hw/irdma/verbs.c
>> +++ b/drivers/infiniband/hw/irdma/verbs.c
>> @@ -2912,6 +2912,102 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
>>   	return ERR_PTR(err);
>>   }
>>   
>> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
>> +					      u64 len, u64 virt,
>> +					      int fd, int access,
>> +					      struct ib_udata *udata)
>> +{
>> +	struct irdma_device *iwdev = to_iwdev(pd->device);
>> +	struct irdma_pble_alloc *palloc;
>> +	struct irdma_pbl *iwpbl;
>> +	struct irdma_mr *iwmr;
>> +	u32 stag = 0;
>> +	bool use_pbles = false;
>> +	int err = -EINVAL;
>> +	struct ib_umem_dmabuf *umem_dmabuf;
>> +
>> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
>> +		return ERR_PTR(-EINVAL);
>> +
>> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
>> +		return ERR_PTR(-EINVAL);
>> +
>> +	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd,
>> +						access);
>> +	if (IS_ERR(umem_dmabuf)) {
>> +		err = PTR_ERR(umem_dmabuf);
>> +		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n", err);
>> +		return ERR_PTR(err);
>> +	}
>> +
>> +	iwmr = kzalloc(sizeof(*iwmr), GFP_KERNEL);
>> +	if (!iwmr) {
>> +		ib_umem_release(&umem_dmabuf->umem);
>> +		return ERR_PTR(-ENOMEM);
>> +	}
> 
> again, please don't duplicate all this code, refactor the mr code so
> it can be shared.
Got it. I will refactor the mr code and split the shared code into 
several helper functions. After the commits are merged, I will continue 
to add the support of dmabuf with the helper functions.

Zhu Yanjun
> 
> Jason

[PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Zhu Yanjun]

From: Zhu Yanjun <yanjun.zhu@linux.dev>

This is a followup to the EFA dmabuf[1]. Irdma driver currently does
not support on-demand-paging(ODP). So it uses habanalabs as the
dmabuf exporter, and irdma as the importer to allow for peer2peer
access through libibverbs.

In this commit, the function ib_umem_dmabuf_get_pinned() is used.
This function is introduced in EFA dmabuf[1] which allows the driver
to get a dmabuf umem which is pinned and does not require move_notify
callback implementation. The returned umem is pinned and DMA mapped
like standard cpu umems, and is released through ib_umem_release().

[1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/

Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
---
V2->V3: Remove unnecessary variable initialization;
        Use error handler;
V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
        This commit is based on the shared functions from refactored
        irdma_reg_user_mr.
---
 drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
 1 file changed, 45 insertions(+)

diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
index 6982f38596c8..7525f4cdf6fb 100644
--- a/drivers/infiniband/hw/irdma/verbs.c
+++ b/drivers/infiniband/hw/irdma/verbs.c
@@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
 	return ERR_PTR(err);
 }
 
+static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
+					      u64 len, u64 virt,
+					      int fd, int access,
+					      struct ib_udata *udata)
+{
+	struct irdma_device *iwdev = to_iwdev(pd->device);
+	struct ib_umem_dmabuf *umem_dmabuf;
+	struct irdma_mr *iwmr;
+	int err;
+
+	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
+		return ERR_PTR(-EINVAL);
+
+	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
+		return ERR_PTR(-EINVAL);
+
+	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd, access);
+	if (IS_ERR(umem_dmabuf)) {
+		err = PTR_ERR(umem_dmabuf);
+		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n", err);
+		return ERR_PTR(err);
+	}
+
+	iwmr = irdma_alloc_iwmr(&umem_dmabuf->umem, pd, virt, IRDMA_MEMREG_TYPE_MEM);
+	if (IS_ERR(iwmr)) {
+		err = PTR_ERR(iwmr);
+		goto err_release;
+	}
+
+	err = irdma_reg_user_mr_type_mem(iwmr, access);
+	if (err)
+		goto err_iwmr;
+
+	return &iwmr->ibmr;
+
+err_iwmr:
+	irdma_free_iwmr(iwmr);
+
+err_release:
+	ib_umem_release(&umem_dmabuf->umem);
+
+	return ERR_PTR(err);
+}
+
 /**
  * irdma_reg_phys_mr - register kernel physical memory
  * @pd: ibpd pointer
@@ -4483,6 +4527,7 @@ static const struct ib_device_ops irdma_dev_ops = {
 	.query_port = irdma_query_port,
 	.query_qp = irdma_query_qp,
 	.reg_user_mr = irdma_reg_user_mr,
+	.reg_user_mr_dmabuf = irdma_reg_user_mr_dmabuf,
 	.req_notify_cq = irdma_req_notify_cq,
 	.resize_cq = irdma_resize_cq,
 	INIT_RDMA_OBJ_SIZE(ib_pd, irdma_pd, ibpd),
-- 
2.27.0

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Zhu Yanjun]

在 2023/2/1 11:21, Zhu Yanjun 写道:
> From: Zhu Yanjun <yanjun.zhu@linux.dev>
> 
> This is a followup to the EFA dmabuf[1]. Irdma driver currently does
> not support on-demand-paging(ODP). So it uses habanalabs as the
> dmabuf exporter, and irdma as the importer to allow for peer2peer
> access through libibverbs.
> 
> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
> This function is introduced in EFA dmabuf[1] which allows the driver
> to get a dmabuf umem which is pinned and does not require move_notify
> callback implementation. The returned umem is pinned and DMA mapped
> like standard cpu umems, and is released through ib_umem_release().
> 
> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
> 
> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
> ---
> V2->V3: Remove unnecessary variable initialization;
>          Use error handler;
> V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
>          This commit is based on the shared functions from refactored
>          irdma_reg_user_mr.

Shiraz Saleem is on vacation and returning to office on Feb.13, 2023.
We can wait for him.

Zhu Yanjun

> ---
>   drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
>   1 file changed, 45 insertions(+)
> 
> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
> index 6982f38596c8..7525f4cdf6fb 100644
> --- a/drivers/infiniband/hw/irdma/verbs.c
> +++ b/drivers/infiniband/hw/irdma/verbs.c
> @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
>   	return ERR_PTR(err);
>   }
>   
> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
> +					      u64 len, u64 virt,
> +					      int fd, int access,
> +					      struct ib_udata *udata)
> +{
> +	struct irdma_device *iwdev = to_iwdev(pd->device);
> +	struct ib_umem_dmabuf *umem_dmabuf;
> +	struct irdma_mr *iwmr;
> +	int err;
> +
> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
> +		return ERR_PTR(-EINVAL);
> +
> +	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd, access);
> +	if (IS_ERR(umem_dmabuf)) {
> +		err = PTR_ERR(umem_dmabuf);
> +		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n", err);
> +		return ERR_PTR(err);
> +	}
> +
> +	iwmr = irdma_alloc_iwmr(&umem_dmabuf->umem, pd, virt, IRDMA_MEMREG_TYPE_MEM);
> +	if (IS_ERR(iwmr)) {
> +		err = PTR_ERR(iwmr);
> +		goto err_release;
> +	}
> +
> +	err = irdma_reg_user_mr_type_mem(iwmr, access);
> +	if (err)
> +		goto err_iwmr;
> +
> +	return &iwmr->ibmr;
> +
> +err_iwmr:
> +	irdma_free_iwmr(iwmr);
> +
> +err_release:
> +	ib_umem_release(&umem_dmabuf->umem);
> +
> +	return ERR_PTR(err);
> +}
> +
>   /**
>    * irdma_reg_phys_mr - register kernel physical memory
>    * @pd: ibpd pointer
> @@ -4483,6 +4527,7 @@ static const struct ib_device_ops irdma_dev_ops = {
>   	.query_port = irdma_query_port,
>   	.query_qp = irdma_query_qp,
>   	.reg_user_mr = irdma_reg_user_mr,
> +	.reg_user_mr_dmabuf = irdma_reg_user_mr_dmabuf,
>   	.req_notify_cq = irdma_req_notify_cq,
>   	.resize_cq = irdma_resize_cq,
>   	INIT_RDMA_OBJ_SIZE(ib_pd, irdma_pd, ibpd),

RE: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Saleem, Shiraz]

> Subject: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin
> memory regions
> 
> From: Zhu Yanjun <yanjun.zhu@linux.dev>
> 
> This is a followup to the EFA dmabuf[1]. Irdma driver currently does not support
> on-demand-paging(ODP). So it uses habanalabs as the dmabuf exporter, and
> irdma as the importer to allow for peer2peer access through libibverbs.
> 
> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
> This function is introduced in EFA dmabuf[1] which allows the driver to get a
> dmabuf umem which is pinned and does not require move_notify callback
> implementation. The returned umem is pinned and DMA mapped like standard cpu
> umems, and is released through ib_umem_release().
> 
> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
> 
> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
> ---
> V2->V3: Remove unnecessary variable initialization;
>         Use error handler;
> V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
>         This commit is based on the shared functions from refactored
>         irdma_reg_user_mr.
> ---
>  drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
>  1 file changed, 45 insertions(+)
> 
> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
> index 6982f38596c8..7525f4cdf6fb 100644
> --- a/drivers/infiniband/hw/irdma/verbs.c
> +++ b/drivers/infiniband/hw/irdma/verbs.c
> @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd
> *pd, u64 start, u64 len,
>  	return ERR_PTR(err);
>  }
> 
> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
> +					      u64 len, u64 virt,
> +					      int fd, int access,
> +					      struct ib_udata *udata)
> +{
> +	struct irdma_device *iwdev = to_iwdev(pd->device);
> +	struct ib_umem_dmabuf *umem_dmabuf;
> +	struct irdma_mr *iwmr;
> +	int err;
> +
> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
> +		return ERR_PTR(-EINVAL);

Do we need this? we don't copy anything from udata. There is no info passed via ABI struct irdma_mem_reg_req.

> +
> +	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd,
> access);
> +	if (IS_ERR(umem_dmabuf)) {
> +		err = PTR_ERR(umem_dmabuf);
> +		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n",
> err);
> +		return ERR_PTR(err);
> +	}
> +
> +	iwmr = irdma_alloc_iwmr(&umem_dmabuf->umem, pd, virt,
> IRDMA_MEMREG_TYPE_MEM);
> +	if (IS_ERR(iwmr)) {
> +		err = PTR_ERR(iwmr);
> +		goto err_release;
> +	}
> +
> +	err = irdma_reg_user_mr_type_mem(iwmr, access);
> +	if (err)
> +		goto err_iwmr;
> +
> +	return &iwmr->ibmr;
> +
> +err_iwmr:
> +	irdma_free_iwmr(iwmr);
> +
> +err_release:
> +	ib_umem_release(&umem_dmabuf->umem);
> +
> +	return ERR_PTR(err);
> +}
> +
>  /**
>   * irdma_reg_phys_mr - register kernel physical memory
>   * @pd: ibpd pointer
> @@ -4483,6 +4527,7 @@ static const struct ib_device_ops irdma_dev_ops = {
>  	.query_port = irdma_query_port,
>  	.query_qp = irdma_query_qp,
>  	.reg_user_mr = irdma_reg_user_mr,
> +	.reg_user_mr_dmabuf = irdma_reg_user_mr_dmabuf,
>  	.req_notify_cq = irdma_req_notify_cq,
>  	.resize_cq = irdma_resize_cq,
>  	INIT_RDMA_OBJ_SIZE(ib_pd, irdma_pd, ibpd),
> --

Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Jason Gunthorpe]

On Wed, Feb 01, 2023 at 11:21:15AM +0800, Zhu Yanjun wrote:
> From: Zhu Yanjun <yanjun.zhu@linux.dev>
> 
> This is a followup to the EFA dmabuf[1]. Irdma driver currently does
> not support on-demand-paging(ODP). So it uses habanalabs as the
> dmabuf exporter, and irdma as the importer to allow for peer2peer
> access through libibverbs.
> 
> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
> This function is introduced in EFA dmabuf[1] which allows the driver
> to get a dmabuf umem which is pinned and does not require move_notify
> callback implementation. The returned umem is pinned and DMA mapped
> like standard cpu umems, and is released through ib_umem_release().
> 
> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
> 
> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
> Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
> ---
> V2->V3: Remove unnecessary variable initialization;
>         Use error handler;
> V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
>         This commit is based on the shared functions from refactored
>         irdma_reg_user_mr.
> ---
>  drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
>  1 file changed, 45 insertions(+)
> 
> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
> index 6982f38596c8..7525f4cdf6fb 100644
> --- a/drivers/infiniband/hw/irdma/verbs.c
> +++ b/drivers/infiniband/hw/irdma/verbs.c
> @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
>  	return ERR_PTR(err);
>  }
>  
> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
> +					      u64 len, u64 virt,
> +					      int fd, int access,
> +					      struct ib_udata *udata)
> +{
> +	struct irdma_device *iwdev = to_iwdev(pd->device);
> +	struct ib_umem_dmabuf *umem_dmabuf;
> +	struct irdma_mr *iwmr;
> +	int err;
> +
> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
> +		return ERR_PTR(-EINVAL);
> +
> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
> +		return ERR_PTR(-EINVAL);

Shiraz is correct, I'm wondering how this even works. This is a new
style uAPI without UVERBS_ATTR_UHW so inlen should always be 0.

How did you manage to test this??

Jason

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Zhu Yanjun]

在 2023/2/16 23:03, Jason Gunthorpe 写道:
> On Wed, Feb 01, 2023 at 11:21:15AM +0800, Zhu Yanjun wrote:
>> From: Zhu Yanjun <yanjun.zhu@linux.dev>
>>
>> This is a followup to the EFA dmabuf[1]. Irdma driver currently does
>> not support on-demand-paging(ODP). So it uses habanalabs as the
>> dmabuf exporter, and irdma as the importer to allow for peer2peer
>> access through libibverbs.
>>
>> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
>> This function is introduced in EFA dmabuf[1] which allows the driver
>> to get a dmabuf umem which is pinned and does not require move_notify
>> callback implementation. The returned umem is pinned and DMA mapped
>> like standard cpu umems, and is released through ib_umem_release().
>>
>> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
>>
>> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
>> Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
>> ---
>> V2->V3: Remove unnecessary variable initialization;
>>          Use error handler;
>> V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
>>          This commit is based on the shared functions from refactored
>>          irdma_reg_user_mr.
>> ---
>>   drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
>>   1 file changed, 45 insertions(+)
>>
>> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
>> index 6982f38596c8..7525f4cdf6fb 100644
>> --- a/drivers/infiniband/hw/irdma/verbs.c
>> +++ b/drivers/infiniband/hw/irdma/verbs.c
>> @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
>>   	return ERR_PTR(err);
>>   }
>>   
>> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
>> +					      u64 len, u64 virt,
>> +					      int fd, int access,
>> +					      struct ib_udata *udata)
>> +{
>> +	struct irdma_device *iwdev = to_iwdev(pd->device);
>> +	struct ib_umem_dmabuf *umem_dmabuf;
>> +	struct irdma_mr *iwmr;
>> +	int err;
>> +
>> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
>> +		return ERR_PTR(-EINVAL);
>> +
>> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
>> +		return ERR_PTR(-EINVAL);
> Shiraz is correct, I'm wondering how this even works. This is a new
> style uAPI without UVERBS_ATTR_UHW so inlen should always be 0.

Got it. Thanks Shiraz and Jason.

I will remove the test of inlen in the latest commit.

Best Regards,

Zhu Yanjun

>
> How did you manage to test this??
>
> Jason

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Zhu Yanjun]

在 2023/2/15 21:38, Saleem, Shiraz 写道:
>> Subject: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin
>> memory regions
>>
>> From: Zhu Yanjun <yanjun.zhu@linux.dev>
>>
>> This is a followup to the EFA dmabuf[1]. Irdma driver currently does not support
>> on-demand-paging(ODP). So it uses habanalabs as the dmabuf exporter, and
>> irdma as the importer to allow for peer2peer access through libibverbs.
>>
>> In this commit, the function ib_umem_dmabuf_get_pinned() is used.
>> This function is introduced in EFA dmabuf[1] which allows the driver to get a
>> dmabuf umem which is pinned and does not require move_notify callback
>> implementation. The returned umem is pinned and DMA mapped like standard cpu
>> umems, and is released through ib_umem_release().
>>
>> [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
>>
>> Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
>> ---
>> V2->V3: Remove unnecessary variable initialization;
>>          Use error handler;
>> V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
>>          This commit is based on the shared functions from refactored
>>          irdma_reg_user_mr.
>> ---
>>   drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
>>   1 file changed, 45 insertions(+)
>>
>> diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
>> index 6982f38596c8..7525f4cdf6fb 100644
>> --- a/drivers/infiniband/hw/irdma/verbs.c
>> +++ b/drivers/infiniband/hw/irdma/verbs.c
>> @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd
>> *pd, u64 start, u64 len,
>>   	return ERR_PTR(err);
>>   }
>>
>> +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
>> +					      u64 len, u64 virt,
>> +					      int fd, int access,
>> +					      struct ib_udata *udata)
>> +{
>> +	struct irdma_device *iwdev = to_iwdev(pd->device);
>> +	struct ib_umem_dmabuf *umem_dmabuf;
>> +	struct irdma_mr *iwmr;
>> +	int err;
>> +
>> +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
>> +		return ERR_PTR(-EINVAL);
>> +
>> +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
>> +		return ERR_PTR(-EINVAL);
> Do we need this? we don't copy anything from udata. There is no info passed via ABI struct irdma_mem_reg_req.


Got it. I will remove the inlen test and send a new commit out.


>
>> +
>> +	umem_dmabuf = ib_umem_dmabuf_get_pinned(pd->device, start, len, fd,
>> access);
>> +	if (IS_ERR(umem_dmabuf)) {
>> +		err = PTR_ERR(umem_dmabuf);
>> +		ibdev_dbg(&iwdev->ibdev, "Failed to get dmabuf umem[%d]\n",
>> err);
>> +		return ERR_PTR(err);
>> +	}
>> +
>> +	iwmr = irdma_alloc_iwmr(&umem_dmabuf->umem, pd, virt,
>> IRDMA_MEMREG_TYPE_MEM);
>> +	if (IS_ERR(iwmr)) {
>> +		err = PTR_ERR(iwmr);
>> +		goto err_release;
>> +	}
>> +
>> +	err = irdma_reg_user_mr_type_mem(iwmr, access);
>> +	if (err)
>> +		goto err_iwmr;
>> +
>> +	return &iwmr->ibmr;
>> +
>> +err_iwmr:
>> +	irdma_free_iwmr(iwmr);
>> +
>> +err_release:
>> +	ib_umem_release(&umem_dmabuf->umem);
>> +
>> +	return ERR_PTR(err);
>> +}
>> +
>>   /**
>>    * irdma_reg_phys_mr - register kernel physical memory
>>    * @pd: ibpd pointer
>> @@ -4483,6 +4527,7 @@ static const struct ib_device_ops irdma_dev_ops = {
>>   	.query_port = irdma_query_port,
>>   	.query_qp = irdma_query_qp,
>>   	.reg_user_mr = irdma_reg_user_mr,
>> +	.reg_user_mr_dmabuf = irdma_reg_user_mr_dmabuf,
>>   	.req_notify_cq = irdma_req_notify_cq,
>>   	.resize_cq = irdma_resize_cq,
>>   	INIT_RDMA_OBJ_SIZE(ib_pd, irdma_pd, ibpd),
>> --
> Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>

Thanks.

Zhu Yanjun

>

Re: [PATCHv3 for-next 1/1] RDMA/irdma: Add support for dmabuf pin memory regions

[Jason Gunthorpe]

On Fri, Feb 17, 2023 at 08:46:52AM +0800, Zhu Yanjun wrote:
> 
> 在 2023/2/16 23:03, Jason Gunthorpe 写道:
> > On Wed, Feb 01, 2023 at 11:21:15AM +0800, Zhu Yanjun wrote:
> > > From: Zhu Yanjun <yanjun.zhu@linux.dev>
> > > 
> > > This is a followup to the EFA dmabuf[1]. Irdma driver currently does
> > > not support on-demand-paging(ODP). So it uses habanalabs as the
> > > dmabuf exporter, and irdma as the importer to allow for peer2peer
> > > access through libibverbs.
> > > 
> > > In this commit, the function ib_umem_dmabuf_get_pinned() is used.
> > > This function is introduced in EFA dmabuf[1] which allows the driver
> > > to get a dmabuf umem which is pinned and does not require move_notify
> > > callback implementation. The returned umem is pinned and DMA mapped
> > > like standard cpu umems, and is released through ib_umem_release().
> > > 
> > > [1]https://lore.kernel.org/lkml/20211007114018.GD2688930@ziepe.ca/t/
> > > 
> > > Signed-off-by: Zhu Yanjun <yanjun.zhu@linux.dev>
> > > Reviewed-by: Shiraz Saleem <shiraz.saleem@intel.com>
> > > ---
> > > V2->V3: Remove unnecessary variable initialization;
> > >          Use error handler;
> > > V1->V2: Thanks Shiraz Saleem, he gave me a lot of good suggestions.
> > >          This commit is based on the shared functions from refactored
> > >          irdma_reg_user_mr.
> > > ---
> > >   drivers/infiniband/hw/irdma/verbs.c | 45 +++++++++++++++++++++++++++++
> > >   1 file changed, 45 insertions(+)
> > > 
> > > diff --git a/drivers/infiniband/hw/irdma/verbs.c b/drivers/infiniband/hw/irdma/verbs.c
> > > index 6982f38596c8..7525f4cdf6fb 100644
> > > --- a/drivers/infiniband/hw/irdma/verbs.c
> > > +++ b/drivers/infiniband/hw/irdma/verbs.c
> > > @@ -2977,6 +2977,50 @@ static struct ib_mr *irdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 len,
> > >   	return ERR_PTR(err);
> > >   }
> > > +static struct ib_mr *irdma_reg_user_mr_dmabuf(struct ib_pd *pd, u64 start,
> > > +					      u64 len, u64 virt,
> > > +					      int fd, int access,
> > > +					      struct ib_udata *udata)
> > > +{
> > > +	struct irdma_device *iwdev = to_iwdev(pd->device);
> > > +	struct ib_umem_dmabuf *umem_dmabuf;
> > > +	struct irdma_mr *iwmr;
> > > +	int err;
> > > +
> > > +	if (len > iwdev->rf->sc_dev.hw_attrs.max_mr_size)
> > > +		return ERR_PTR(-EINVAL);
> > > +
> > > +	if (udata->inlen < IRDMA_MEM_REG_MIN_REQ_LEN)
> > > +		return ERR_PTR(-EINVAL);
> > Shiraz is correct, I'm wondering how this even works. This is a new
> > style uAPI without UVERBS_ATTR_UHW so inlen should always be 0.
> 
> Got it. Thanks Shiraz and Jason.
> 
> I will remove the test of inlen in the latest commit.

Please answer how did you manage to test this?

Jason