Re: [PATCH] arm64: PCI: free all allocated memory in case of failure
From: Lorenzo Pieralisi
Date: Mon May 22 2017 - 08:59:40 EST
On Mon, May 22, 2017 at 08:32:09PM +0800, Timmy Li wrote:
> Hi Lorenzo,
>
> Thanks for your review and sorry for the late reply.
>
> On 2017/5/12 18:28, Lorenzo Pieralisi wrote:
> > On Fri, May 12, 2017 at 05:57:47PM +0800, Timmy Li wrote:
> >> There are some memory allocations in pci_acpi_scan_root(). But
> >> ri, root_ops and ri->cfg are not freed properly in failure cases,
> >> which results in memory leaks. This patch fixes the potential
> >> memory leaks.
> >>
> >> Signed-off-by: Timmy Li <lixiaoping3@xxxxxxxxxx>
> >> ---
> >> arch/arm64/kernel/pci.c | 22 ++++++++++++++--------
> >> 1 file changed, 14 insertions(+), 8 deletions(-)
> >>
> >> diff --git a/arch/arm64/kernel/pci.c b/arch/arm64/kernel/pci.c
> >> index 4f0e3eb..e7e88ce 100644
> >> --- a/arch/arm64/kernel/pci.c
> >> +++ b/arch/arm64/kernel/pci.c
> >> @@ -188,25 +188,22 @@ struct pci_bus *pci_acpi_scan_root(struct acpi_pci_root *root)
> >>
> >> ri = kzalloc_node(sizeof(*ri), GFP_KERNEL, node);
> >> if (!ri)
> >> - return NULL;
> >> + goto err_allocri;
> >>
> >> root_ops = kzalloc_node(sizeof(*root_ops), GFP_KERNEL, node);
> >> if (!root_ops)
> >> - return NULL;
> >> + goto err_allocops;
>
> I think it is still needed to free ri here When ri is allocated
> successfully and root_ops is not. Considering this is the only remain
> place that need to fix, a simple way to do this would be:
> if (!root_ops) {
> + kfree(ri);
> return NULL;
> }
Ok, I will send a fix for that, thanks.
Lorenzo
>
> >>
> >> ri->cfg = pci_acpi_setup_ecam_mapping(root);
> >> - if (!ri->cfg) {
> >> - kfree(ri);
> >> - kfree(root_ops);
> >> - return NULL;
> >> - }
> >> + if (!ri->cfg)
> >> + goto err_ecam;
> >>
> >> root_ops->release_info = pci_acpi_generic_release_info;
> >
> > You are missing this ^^^^^^^^^
> >
> >> root_ops->prepare_resources = pci_acpi_root_prepare_resources;
> >> root_ops->pci_ops = &ri->cfg->ops->pci_ops;
> >> bus = acpi_pci_root_create(root, root_ops, &ri->common, ri->cfg);
> >> if (!bus)
> >
> > And how it works if (bus == NULL) here.
>
> You are correct. There is no need to do anything When acpi_pci_root_create() fails.
>
> >
> > Lorenzo
> >
> >> - return NULL;
> >> + goto err_rootcreate;
> >>
> >> pci_bus_size_bridges(bus);
> >> pci_bus_assign_resources(bus);
> >> @@ -215,6 +212,15 @@ struct pci_bus *pci_acpi_scan_root(struct acpi_pci_root *root)
> >> pcie_bus_configure_settings(child);
> >>
> >> return bus;
> >> +
> >> +err_rootcreate:
> >> + pci_ecam_free(ri->cfg);
> >> +err_ecam:
> >> + kfree(root_ops);
> >> +err_allocops:
> >> + kfree(ri);
> >> +err_allocri:
> >> + return NULL;
> >> }
> >>
> >> void pcibios_add_bus(struct pci_bus *bus)
> >> --
> >> 1.9.1
> >>
> >>
> >
> > .
> >
>