bus/pci: fix bus info memleak during PCI scan
Checks
Commit Message
During pci_scan_one() for devices that were already registered
the pci_common_set() is called to set some of the fields again.
This resulted in bus_info allocation leaking, so this patch
ensures they are always freed beforehand.
Fixes: 8f4de2dba9b9 ("bus/pci: fill bus specific information")
Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
---
drivers/bus/pci/pci_common.c | 1 +
1 file changed, 1 insertion(+)
Comments
On Fri, Nov 18, 2022 at 2:36 PM Tomasz Zawadzki
<tomasz.zawadzki@intel.com> wrote:
>
> During pci_scan_one() for devices that were already registered
> the pci_common_set() is called to set some of the fields again.
>
> This resulted in bus_info allocation leaking, so this patch
> ensures they are always freed beforehand.
>
> Fixes: 8f4de2dba9b9 ("bus/pci: fill bus specific information")
>
> Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
> ---
> drivers/bus/pci/pci_common.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/drivers/bus/pci/pci_common.c b/drivers/bus/pci/pci_common.c
> index 9901c34f4e..9a866055e8 100644
> --- a/drivers/bus/pci/pci_common.c
> +++ b/drivers/bus/pci/pci_common.c
> @@ -114,6 +114,7 @@ pci_common_set(struct rte_pci_device *dev)
> /* Otherwise, it uses the internal, canonical form. */
> dev->device.name = dev->name;
>
> + free(dev->bus_info);
> if (asprintf(&dev->bus_info, "vendor_id=%"PRIx16", device_id=%"PRIx16,
> dev->id.vendor_id, dev->id.device_id) != -1)
> dev->device.bus_info = dev->bus_info;
Indeed, good catch.
The bus_info content is constant for a given device, there is no need
to free and reallocate.
WDYT of:
@@ -114,8 +114,9 @@ pci_common_set(struct rte_pci_device *dev)
/* Otherwise, it uses the internal, canonical form. */
dev->device.name = dev->name;
- if (asprintf(&dev->bus_info, "vendor_id=%"PRIx16", device_id=%"PRIx16,
- dev->id.vendor_id, dev->id.device_id) != -1)
+ if (dev->bus_info != NULL ||
+ asprintf(&dev->bus_info, "vendor_id=%"PRIx16",
device_id=%"PRIx16,
+ dev->id.vendor_id, dev->id.device_id) != -1)
dev->device.bus_info = dev->bus_info;
}
On Mon, Nov 21, 2022 at 9:44 AM David Marchand
<david.marchand@redhat.com> wrote:
>
> On Fri, Nov 18, 2022 at 2:36 PM Tomasz Zawadzki
> <tomasz.zawadzki@intel.com> wrote:
> >
> > During pci_scan_one() for devices that were already registered
> > the pci_common_set() is called to set some of the fields again.
> >
> > This resulted in bus_info allocation leaking, so this patch
> > ensures they are always freed beforehand.
> >
> > Fixes: 8f4de2dba9b9 ("bus/pci: fill bus specific information")
> >
> > Signed-off-by: Tomasz Zawadzki <tomasz.zawadzki@intel.com>
I sent https://patchwork.dpdk.org/project/dpdk/patch/20221121111209.2396341-1-david.marchand@redhat.com/
as an alternative.
Can you have a look?
Thanks.
@@ -114,6 +114,7 @@ pci_common_set(struct rte_pci_device *dev)
/* Otherwise, it uses the internal, canonical form. */
dev->device.name = dev->name;
+ free(dev->bus_info);
if (asprintf(&dev->bus_info, "vendor_id=%"PRIx16", device_id=%"PRIx16,
dev->id.vendor_id, dev->id.device_id) != -1)
dev->device.bus_info = dev->bus_info;