Re: [PATCH v2 1/3] dma/idxd: fix memory leak in pci close

[Bruce Richardson <bruce.richardson@intel.com>]

On Sun, Jul 03, 2022 at 01:22:41PM +0100, Kevin Laatz wrote:
> ASAN reports a memory leak for the 'pci' pointer in the 'idxd_dmadev'
> struct.
> 
> This is fixed by free'ing the struct when the last queue on the PCI
> device is being closed.
> 
> Fixes: 9449330a8458 ("dma/idxd: create dmadev instances on PCI probe")
> Cc: stable@dpdk.org
> Cc: bruce.richardson@intel.com
> 
> Reported-by: Xingguang He <xingguang.he@intel.com>
> Signed-off-by: Kevin Laatz <kevin.laatz@intel.com>
> ---
>  drivers/dma/idxd/idxd_common.c   |  2 ++
>  drivers/dma/idxd/idxd_internal.h |  2 ++
>  drivers/dma/idxd/idxd_pci.c      | 34 +++++++++++++++++++++++++-------
>  3 files changed, 31 insertions(+), 7 deletions(-)
> 

Some comments inline below.

/Bruce

> diff --git a/drivers/dma/idxd/idxd_common.c b/drivers/dma/idxd/idxd_common.c
> index c77200a457..d347bbed21 100644
> --- a/drivers/dma/idxd/idxd_common.c
> +++ b/drivers/dma/idxd/idxd_common.c
> @@ -620,6 +620,8 @@ idxd_dmadev_create(const char *name, struct rte_device *dev,
>  	dmadev->fp_obj->dev_private = idxd;
>  
>  	idxd->dmadev->state = RTE_DMA_DEV_READY;
> +	if (idxd->u.pci != NULL)
> +		rte_atomic16_inc(&idxd->u.pci->ref_count);
>  
>  	return 0;
>  

I don't think this belongs in the common code. Can it be put somewhere in
the pci-specific driver code to avoid issues, e.g. after idxd_dmadev_create
returns in probe_pci() function.


> diff --git a/drivers/dma/idxd/idxd_internal.h b/drivers/dma/idxd/idxd_internal.h
> index 3375600217..180a8587c6 100644
> --- a/drivers/dma/idxd/idxd_internal.h
> +++ b/drivers/dma/idxd/idxd_internal.h
> @@ -7,6 +7,7 @@
>  
>  #include <rte_dmadev_pmd.h>
>  #include <rte_spinlock.h>
> +#include <rte_atomic.h>
>  
>  #include "idxd_hw_defs.h"
>  
> @@ -33,6 +34,7 @@ struct idxd_pci_common {
>  	rte_spinlock_t lk;
>  
>  	uint8_t wq_cfg_sz;
> +	rte_atomic16_t ref_count;
>  	volatile struct rte_idxd_bar0 *regs;
>  	volatile uint32_t *wq_regs_base;
>  	volatile struct rte_idxd_grpcfg *grp_regs;
> diff --git a/drivers/dma/idxd/idxd_pci.c b/drivers/dma/idxd/idxd_pci.c
> index 65c6bbf4c1..918981f2ea 100644
> --- a/drivers/dma/idxd/idxd_pci.c
> +++ b/drivers/dma/idxd/idxd_pci.c
> @@ -6,6 +6,7 @@
>  #include <rte_devargs.h>
>  #include <rte_dmadev_pmd.h>
>  #include <rte_malloc.h>
> +#include <rte_atomic.h>
>  
>  #include "idxd_internal.h"
>  
> @@ -115,20 +116,38 @@ idxd_pci_dev_close(struct rte_dma_dev *dev)
>  {
>  	struct idxd_dmadev *idxd = dev->fp_obj->dev_private;
>  	uint8_t err_code;
> +	int is_last_wq;
>  
> -	/* disable the device */
> -	err_code = idxd_pci_dev_command(idxd, idxd_disable_dev);
> -	if (err_code) {
> -		IDXD_PMD_ERR("Error disabling device: code %#x", err_code);
> -		return err_code;
> +	if (idxd_is_wq_enabled(idxd)) {
> +		/* disable the wq */
> +		err_code = idxd_pci_dev_command(idxd, idxd_disable_wq);
> +		if (err_code) {
> +			IDXD_PMD_ERR("Error disabling wq: code %#x", err_code);
> +			return err_code;
> +		}
> +		IDXD_PMD_DEBUG("IDXD WQ disabled OK");
>  	}
> -	IDXD_PMD_DEBUG("IDXD Device disabled OK");
>  
>  	/* free device memory */
>  	IDXD_PMD_DEBUG("Freeing device driver memory");
>  	rte_free(idxd->batch_idx_ring);
>  	rte_free(idxd->desc_ring);
>  
> +	/* if this is the last WQ on the device, disable the device and free
> +	 * the PCI struct
> +	 */
> +	is_last_wq = rte_atomic16_dec_and_test(&idxd->u.pci->ref_count);
> +	if (is_last_wq) {
> +		/* disable the device */
> +		err_code = idxd_pci_dev_command(idxd, idxd_disable_dev);
> +		if (err_code) {
> +			IDXD_PMD_ERR("Error disabling device: code %#x", err_code);
> +			return err_code;
> +		}
> +		IDXD_PMD_DEBUG("IDXD device disabled OK");
> +		rte_free(idxd->u.pci);
> +	}
> +
>  	return 0;
>  }
>  
> @@ -159,12 +178,13 @@ init_pci_device(struct rte_pci_device *dev, struct idxd_dmadev *idxd,
>  	uint8_t lg2_max_batch, lg2_max_copy_size;
>  	unsigned int i, err_code;
>  
> -	pci = malloc(sizeof(*pci));
> +	pci = rte_malloc(NULL, sizeof(*pci), 0);

Any particular reason for the change from regular malloc to rte_malloc?

>  	if (pci == NULL) {
>  		IDXD_PMD_ERR("%s: Can't allocate memory", __func__);
>  		err_code = -1;
>  		goto err;
>  	}
> +	memset(pci, 0, sizeof(*pci));
>  	rte_spinlock_init(&pci->lk);
>  
>  	/* assign the bar registers, and then configure device */
> -- 
> 2.31.1
>