Re: [PATCH v4 11/27] vfio/cxl: Own the whole component register BAR

[Alex Williamson]

On Thu, 13 Aug 2026 15:06:15 +0530
<mhonap@xxxxxxxxxx> wrote:

> From: Manish Honap <mhonap@xxxxxxxxxx>
> 
> devm_cxl_probe_mem() maps the component registers through the CXL core.
> By default the core claims each HDM/RAS sub-block, which then collides
> with the full-BAR request that vfio-pci makes when the guest opens the
> device.
> 
> Discover the component registers into the device reg_map and pass
> bar_owned to cxl_pci_setup_regs(), so the core maps the sub-blocks
> without claiming them and vfio-pci owns the whole BAR.
> 
> Signed-off-by: Manish Honap <mhonap@xxxxxxxxxx>
> ---
>  drivers/vfio/pci/cxl/vfio_cxl_core.c | 11 +++++++++++
>  1 file changed, 11 insertions(+)
> 
> diff --git a/drivers/vfio/pci/cxl/vfio_cxl_core.c b/drivers/vfio/pci/cxl/vfio_cxl_core.c
> index 7edc53b25576..966b7c22a6ac 100644
> --- a/drivers/vfio/pci/cxl/vfio_cxl_core.c
> +++ b/drivers/vfio/pci/cxl/vfio_cxl_core.c
> @@ -10,6 +10,7 @@
>  #include <linux/range.h>
>  #include <linux/vfio_pci_core.h>
>  #include <cxl/cxl.h>
> +#include <cxl/pci.h>
>  
>  /**
>   * struct vfio_cxl_state - per-device state for a vfio-cxl device
> @@ -70,6 +71,16 @@ static int vfio_cxl_init_device(struct vfio_pci_core_device *vdev)
>  	if (!cxl)
>  		return -ENOMEM;
>  
> +	/*
> +	 * vfio-pci requests the whole component BAR when the guest opens the
> +	 * device. Declare the BAR owned so the CXL core maps the HDM/RAS
> +	 * sub-blocks without claiming them and that request does not collide.
> +	 */
> +	ret = cxl_pci_setup_regs(pdev, CXL_REGLOC_RBI_COMPONENT,
> +				 &cxl->cxlds.reg_map, true);
> +	if (ret)
> +		return ret;
> +
>  	ret = cxl_set_capacity(&cxl->cxlds, hdm_size);
>  	if (ret)
>  		return ret;

Previous comments on patch 2 suggest a different interface here, but do
note that cxl_pci_set_regs() can return -EPROBE_DEFER and if we do
actually want to defer the probe (until when?) we'd need to split this
errno out as a legitimate probe error case versus the fallback mechanism
for other errnos.  Thanks,

Alex