[PATCH] nvmet: pci-epf: Move DMA initialization to EPC init callback

[PATCH] nvmet: pci-epf: Move DMA initialization to EPC init callback

[Niklas Cassel]

From: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>

For DMA initialization to work across all EPC drivers, the DMA
initialization has to be done in the .init() callback.

This is because not all EPC drivers will have a refclock (which is often
needed to access registers of a DMA controller embedded in a PCIe
controller) at the time the .bind() callback is called.

However, all EPC drivers are guaranteed to have a refclock by the time
the .init() callback is called.

Thus, move the DMA initialization to the .init() callback.

This change was already done for other EPF drivers in
commit 60bd3e039aa2 ("PCI: endpoint: pci-epf-{mhi/test}: Move DMA
initialization to EPC init callback").

Cc: stable@vger.kernel.org
Fixes: 0faa0fe6f90e ("nvmet: New NVMe PCI endpoint function target driver")
Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
Signed-off-by: Niklas Cassel <cassel@kernel.org>
---
 drivers/nvme/target/pci-epf.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/nvme/target/pci-epf.c b/drivers/nvme/target/pci-epf.c
index 2e78397a7373a..9c5b0f78ce8df 100644
--- a/drivers/nvme/target/pci-epf.c
+++ b/drivers/nvme/target/pci-epf.c
@@ -2325,6 +2325,8 @@ static int nvmet_pci_epf_epc_init(struct pci_epf *epf)
 		return ret;
 	}
 
+	nvmet_pci_epf_init_dma(nvme_epf);
+
 	/* Set device ID, class, etc. */
 	epf->header->vendorid = ctrl->tctrl->subsys->vendor_id;
 	epf->header->subsys_vendor_id = ctrl->tctrl->subsys->subsys_vendor_id;
@@ -2422,8 +2424,6 @@ static int nvmet_pci_epf_bind(struct pci_epf *epf)
 	if (ret)
 		return ret;
 
-	nvmet_pci_epf_init_dma(nvme_epf);
-
 	return 0;
 }
 
-- 
2.51.0

Re: [PATCH] nvmet: pci-epf: Move DMA initialization to EPC init callback

[Damien Le Moal]

On 9/9/25 8:21 PM, Niklas Cassel wrote:
> From: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> 
> For DMA initialization to work across all EPC drivers, the DMA
> initialization has to be done in the .init() callback.
> 
> This is because not all EPC drivers will have a refclock (which is often
> needed to access registers of a DMA controller embedded in a PCIe
> controller) at the time the .bind() callback is called.
> 
> However, all EPC drivers are guaranteed to have a refclock by the time
> the .init() callback is called.
> 
> Thus, move the DMA initialization to the .init() callback.
> 
> This change was already done for other EPF drivers in
> commit 60bd3e039aa2 ("PCI: endpoint: pci-epf-{mhi/test}: Move DMA
> initialization to EPC init callback").
> 
> Cc: stable@vger.kernel.org
> Fixes: 0faa0fe6f90e ("nvmet: New NVMe PCI endpoint function target driver")
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> Signed-off-by: Niklas Cassel <cassel@kernel.org>

Looks good.

Reviewed-by: Damien Le Moal <dlemoal@kernel.org>


-- 
Damien Le Moal
Western Digital Research

Re: [PATCH] nvmet: pci-epf: Move DMA initialization to EPC init callback

[Niklas Cassel]

On Tue, Sep 09, 2025 at 01:21:22PM +0200, Niklas Cassel wrote:
> From: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> 
> For DMA initialization to work across all EPC drivers, the DMA
> initialization has to be done in the .init() callback.
> 
> This is because not all EPC drivers will have a refclock (which is often
> needed to access registers of a DMA controller embedded in a PCIe
> controller) at the time the .bind() callback is called.
> 
> However, all EPC drivers are guaranteed to have a refclock by the time
> the .init() callback is called.
> 
> Thus, move the DMA initialization to the .init() callback.
> 
> This change was already done for other EPF drivers in
> commit 60bd3e039aa2 ("PCI: endpoint: pci-epf-{mhi/test}: Move DMA
> initialization to EPC init callback").
> 
> Cc: stable@vger.kernel.org
> Fixes: 0faa0fe6f90e ("nvmet: New NVMe PCI endpoint function target driver")
> Signed-off-by: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> Signed-off-by: Niklas Cassel <cassel@kernel.org>
> ---
>  drivers/nvme/target/pci-epf.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/nvme/target/pci-epf.c b/drivers/nvme/target/pci-epf.c
> index 2e78397a7373a..9c5b0f78ce8df 100644
> --- a/drivers/nvme/target/pci-epf.c
> +++ b/drivers/nvme/target/pci-epf.c
> @@ -2325,6 +2325,8 @@ static int nvmet_pci_epf_epc_init(struct pci_epf *epf)
>  		return ret;
>  	}
>  
> +	nvmet_pci_epf_init_dma(nvme_epf);
> +
>  	/* Set device ID, class, etc. */
>  	epf->header->vendorid = ctrl->tctrl->subsys->vendor_id;
>  	epf->header->subsys_vendor_id = ctrl->tctrl->subsys->subsys_vendor_id;
> @@ -2422,8 +2424,6 @@ static int nvmet_pci_epf_bind(struct pci_epf *epf)
>  	if (ret)
>  		return ret;
>  
> -	nvmet_pci_epf_init_dma(nvme_epf);
> -
>  	return 0;
>  }
>  
> -- 
> 2.51.0
> 
> 

Gentle ping


Kind regards,
Niklas

Re: [PATCH] nvmet: pci-epf: Move DMA initialization to EPC init callback

[Keith Busch]

On Tue, Sep 09, 2025 at 01:21:22PM +0200, Niklas Cassel wrote:
> From: Shin'ichiro Kawasaki <shinichiro.kawasaki@wdc.com>
> 
> For DMA initialization to work across all EPC drivers, the DMA
> initialization has to be done in the .init() callback.
> 
> This is because not all EPC drivers will have a refclock (which is often
> needed to access registers of a DMA controller embedded in a PCIe
> controller) at the time the .bind() callback is called.
> 
> However, all EPC drivers are guaranteed to have a refclock by the time
> the .init() callback is called.
> 
> Thus, move the DMA initialization to the .init() callback.
> 
> This change was already done for other EPF drivers in
> commit 60bd3e039aa2 ("PCI: endpoint: pci-epf-{mhi/test}: Move DMA
> initialization to EPC init callback").

Thanks, applied to nvme-6.19.