Hi Jim,

On Thu, Oct 29, 2026 at 03:05:10PM +0000, Jim MacArthur wrote:

Please add a commit message to this patch. Even for small changes, a few lines
of context help reviewers now and anyone reading the history later. For example:
why a DVSEC helper is needed, what this capability allows, which upcoming patch
uses it, DVSEC meaning, PCIe spec reference, etc.


> Signed-off-by: Jim MacArthur <[email protected]>
> ---
>  hw/pci/pcie.c         | 16 ++++++++++++++++
>  include/hw/pci/pcie.h |  4 ++++
>  2 files changed, 20 insertions(+)
> 
> diff --git a/hw/pci/pcie.c b/hw/pci/pcie.c
> index 42bcb9206d..42c4997b42 100644
> --- a/hw/pci/pcie.c
> +++ b/hw/pci/pcie.c
> @@ -1157,6 +1157,22 @@ void pcie_add_capability(PCIDevice *dev,
>      memset(dev->cmask + offset, 0xFF, size);
>  }
>  
> +/*
> + * Add a DVSEC capability. As per pcie_add_capability, the caller must
> + * ensure (offset, size) does not overlap any other capability or extend
> + * past the extended configuration space.
> + */
> +void pcie_add_dvsec_capability(PCIDevice *device, uint16_t capability_id,
> +                               uint8_t version, uint32_t offset,
> +                               uint16_t size_bytes, uint16_t vendor_id,
> +                               uint8_t dvsec_version, uint16_t dvsec_id)
> +{
> +    const uint32_t header1 = size_bytes << 20 | version << 16 | vendor_id;

Header 1 should have 'dvsec_version' in it, not 'version'. As a side effect
'dvsec_version' is left unused here. Also I think it should be 'dvsec_revision',
not 'dvsec_version'.

I also think it's better s/size_bytes/dvsec_length/ and
s/vendor_id/dvsec_vendor_id for consistence.

It's also better to add range checks for 'dvsec_length' and 'dvsec_revision'
since their size don't match, respectively, size 16b and 8b, like:

      assert((dvsec_length & ~0xfff) == 0); 
      assert((dvsec_revision & ~0xf) == 0);


> +    pcie_add_capability(device, capability_id, version, offset, size_bytes);

To me, capability_id and version should actually be hardecoded because a DVSEC
is by definition PCI_EXT_CAP_ID_DVSEC with Capability Version 1, so:

pcie_add_capability(dev, PCI_EXT_CAP_ID_DVSEC, 1, offset, dvsec_length);

I'm looking at PCIe spec rev. 7.0.


> +    pci_set_long(device->config + offset + 0x4, header1);

Could we use PCI_DVSEC_HEADER1 here instead of 0x4?


> +    pci_set_word(device->config + offset + 0x8, dvsec_id);

Could we use PCI_DVSEC_HEADER2 here instead of 0x8?


Cheers,
Gustavo

> +}
> +
>  /*
>   * Sync the PCIe Link Status negotiated speed and width of a bridge with the
>   * downstream device.  If downstream device is not present, re-write with the
> diff --git a/include/hw/pci/pcie.h b/include/hw/pci/pcie.h
> index ec25e7a7de..86c29f7d8b 100644
> --- a/include/hw/pci/pcie.h
> +++ b/include/hw/pci/pcie.h
> @@ -133,6 +133,10 @@ uint16_t pcie_find_capability(PCIDevice *dev, uint16_t 
> cap_id);
>  void pcie_add_capability(PCIDevice *dev,
>                           uint16_t cap_id, uint8_t cap_ver,
>                           uint16_t offset, uint16_t size);
> +void pcie_add_dvsec_capability(PCIDevice *device, uint16_t capability_id,
> +                               uint8_t version, uint32_t offset,
> +                               uint16_t size_bytes, uint16_t vendor_id,
> +                               uint8_t dvsec_version, uint16_t dvsec_id);
>  bool pcie_insert_capability(PCIDevice *dev, uint16_t cap_id, uint8_t cap_ver,
>                              uint16_t offset, uint16_t size);
>  void pcie_sync_bridge_lnk(PCIDevice *dev);
> 
> -- 
> 2.43.0
> 
> 


Reply via email to