On Mon, Jul 04, 2022 at 02:55:01PM +0800, Liang He wrote:
> Commit 3a87cb8f6a ("Fix refcount leak in brcm_pcie_probe()") adds
> of_node_put() for of_parse_phandle() in fail path but not adds it
> correctly in normal path. We should move the second of_node_put()
> out of the 'if(pci_msi_enabled() && msi_np == pcie->np)'.
>
> Fixes: 3a87cb8f6a ("Fix refcount leak in brcm_pcie_probe()")
> Co-authored-by: Miaoqian Lin <linmq006@xxxxxxxxx>
> Signed-off-by: Liang He <windhl@xxxxxxx>
> ---
> Patched file has been compiled test in next branch.
>
> drivers/pci/controller/pcie-brcmstb.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/pci/controller/pcie-brcmstb.c b/drivers/pci/controller/pcie-brcmstb.c
> index 48a7148376d4..80e19d053e9f 100755
> --- a/drivers/pci/controller/pcie-brcmstb.c
> +++ b/drivers/pci/controller/pcie-brcmstb.c
> @@ -1413,8 +1413,8 @@ static int brcm_pcie_probe(struct platform_device *pdev)
> of_node_put(msi_np);
> goto fail;
> }
> - of_node_put(msi_np);
> }
> + of_node_put(msi_np);
Can we just move the of_parse_phandle() and related checking into
brcm_pcie_enable_msi()? It can return success without doing anything
if !pci_msi_enabled() or msi_np != pcie_np.
If you don't want to do that, please just send a revised version of
3a87cb8f6a72 ("PCI: brcmstb: Fix refcount leak in brcm_pcie_probe()").
That's not upstream yet, and I don't want to clutter the git history
with a fix of a fix.
> bridge->ops = pcie->type == BCM7425 ? &brcm_pcie_ops32 : &brcm_pcie_ops;
> bridge->sysdata = pcie;
> --
> 2.25.1
>
