net/fm10k: fix cleanup during init failure
Checks
Commit Message
Cleanup was not done on this PMD if a error is seen during the init:
- possible memory leak due to a missing free
- interrupt handler was not disabled: if an IRQ is received after the
init, a SIGSEGV can be seen (private data stored in
rte_eth_devices[port_id] is pointing to NULL)
Fixes: a6061d9e7075 ("fm10k: register PF driver")
Fixes: 4c287332c39a ("fm10k: add PF and VF interrupt handling")
Cc: stable@dpdk.org
Signed-off-by: Julien Meunier <julien.meunier@nokia.com>
---
drivers/net/fm10k/fm10k_ethdev.c | 39 +++++++++++++++++++++++++++-----
1 file changed, 33 insertions(+), 6 deletions(-)
Comments
On Wed, Apr 03, 2024 at 01:55:41PM +0200, Julien Meunier wrote:
> Cleanup was not done on this PMD if a error is seen during the init:
> - possible memory leak due to a missing free
> - interrupt handler was not disabled: if an IRQ is received after the
> init, a SIGSEGV can be seen (private data stored in
> rte_eth_devices[port_id] is pointing to NULL)
>
> Fixes: a6061d9e7075 ("fm10k: register PF driver")
> Fixes: 4c287332c39a ("fm10k: add PF and VF interrupt handling")
> Cc: stable@dpdk.org
>
> Signed-off-by: Julien Meunier <julien.meunier@nokia.com>
Changes LGTM
Reviewed-by: Bruce Richardson <bruce.richardson@intel.com>
On Thu, Apr 04, 2024 at 11:18:16AM +0100, Bruce Richardson wrote:
> On Wed, Apr 03, 2024 at 01:55:41PM +0200, Julien Meunier wrote:
> > Cleanup was not done on this PMD if a error is seen during the init:
> > - possible memory leak due to a missing free
> > - interrupt handler was not disabled: if an IRQ is received after the
> > init, a SIGSEGV can be seen (private data stored in
> > rte_eth_devices[port_id] is pointing to NULL)
> >
> > Fixes: a6061d9e7075 ("fm10k: register PF driver")
> > Fixes: 4c287332c39a ("fm10k: add PF and VF interrupt handling")
> > Cc: stable@dpdk.org
> >
> > Signed-off-by: Julien Meunier <julien.meunier@nokia.com>
>
> Changes LGTM
>
> Reviewed-by: Bruce Richardson <bruce.richardson@intel.com>
There are CI failures flagged in patchwork, but they appear unrelated to
this patch itself.
Patch applied to dpdk-next-net-intel.
Thanks,
/Bruce
Recheck-request: iol-intel-Functional
I see it's applied but I just want to clean up the record - checking
for lab infra failure
On Thu, Apr 4, 2024 at 6:39 AM Bruce Richardson
<bruce.richardson@intel.com> wrote:
>
> On Thu, Apr 04, 2024 at 11:18:16AM +0100, Bruce Richardson wrote:
> > On Wed, Apr 03, 2024 at 01:55:41PM +0200, Julien Meunier wrote:
> > > Cleanup was not done on this PMD if a error is seen during the init:
> > > - possible memory leak due to a missing free
> > > - interrupt handler was not disabled: if an IRQ is received after the
> > > init, a SIGSEGV can be seen (private data stored in
> > > rte_eth_devices[port_id] is pointing to NULL)
> > >
> > > Fixes: a6061d9e7075 ("fm10k: register PF driver")
> > > Fixes: 4c287332c39a ("fm10k: add PF and VF interrupt handling")
> > > Cc: stable@dpdk.org
> > >
> > > Signed-off-by: Julien Meunier <julien.meunier@nokia.com>
> >
> > Changes LGTM
> >
> > Reviewed-by: Bruce Richardson <bruce.richardson@intel.com>
>
> There are CI failures flagged in patchwork, but they appear unrelated to
> this patch itself.
>
> Patch applied to dpdk-next-net-intel.
> Thanks,
> /Bruce
@@ -3058,7 +3058,7 @@ eth_fm10k_dev_init(struct rte_eth_dev *dev)
struct fm10k_hw *hw = FM10K_DEV_PRIVATE_TO_HW(dev->data->dev_private);
struct rte_pci_device *pdev = RTE_ETH_DEV_TO_PCI(dev);
struct rte_intr_handle *intr_handle = pdev->intr_handle;
- int diag, i;
+ int diag, i, ret;
struct fm10k_macvlan_filter_info *macvlan;
PMD_INIT_FUNC_TRACE();
@@ -3147,21 +3147,24 @@ eth_fm10k_dev_init(struct rte_eth_dev *dev)
diag = fm10k_stats_reset(dev);
if (diag != 0) {
PMD_INIT_LOG(ERR, "Stats reset failed: %d", diag);
- return diag;
+ ret = diag;
+ goto err_stat;
}
/* Reset the hw */
diag = fm10k_reset_hw(hw);
if (diag != FM10K_SUCCESS) {
PMD_INIT_LOG(ERR, "Hardware reset failed: %d", diag);
- return -EIO;
+ ret = -EIO;
+ goto err_reset_hw;
}
/* Setup mailbox service */
diag = fm10k_setup_mbx_service(hw);
if (diag != FM10K_SUCCESS) {
PMD_INIT_LOG(ERR, "Failed to setup mailbox: %d", diag);
- return -EIO;
+ ret = -EIO;
+ goto err_mbx;
}
/*PF/VF has different interrupt handling mechanism */
@@ -3200,7 +3203,8 @@ eth_fm10k_dev_init(struct rte_eth_dev *dev)
if (switch_ready == false) {
PMD_INIT_LOG(ERR, "switch is not ready");
- return -1;
+ ret = -1;
+ goto err_switch_ready;
}
}
@@ -3235,7 +3239,8 @@ eth_fm10k_dev_init(struct rte_eth_dev *dev)
if (!hw->mac.default_vid) {
PMD_INIT_LOG(ERR, "default VID is not ready");
- return -1;
+ ret = -1;
+ goto err_vid;
}
}
@@ -3244,6 +3249,28 @@ eth_fm10k_dev_init(struct rte_eth_dev *dev)
MAIN_VSI_POOL_NUMBER);
return 0;
+
+err_vid:
+err_switch_ready:
+ rte_intr_disable(intr_handle);
+
+ if (hw->mac.type == fm10k_mac_pf) {
+ fm10k_dev_disable_intr_pf(dev);
+ rte_intr_callback_unregister(intr_handle,
+ fm10k_dev_interrupt_handler_pf, (void *)dev);
+ } else {
+ fm10k_dev_disable_intr_vf(dev);
+ rte_intr_callback_unregister(intr_handle,
+ fm10k_dev_interrupt_handler_vf, (void *)dev);
+ }
+
+err_mbx:
+err_reset_hw:
+err_stat:
+ rte_free(dev->data->mac_addrs);
+ dev->data->mac_addrs = NULL;
+
+ return ret;
}
static int