2022-09-08 07:45:23

by Xiaoke Wang

Subject: [PATCH v3] staging: rtl8723bs: fix potential memory leak in rtw_init_drv_sw()

From: Xiaoke Wang <[email protected]>

In rtw_init_drv_sw(), there are various init functions are called to
populate the padapter structure and some checks for their return value.
However, except for the first one error path, the other five error paths
do not properly release the previous allocated resources, which leads to
various memory leaks.

This patch fixes them and keeps the success and error separate. As there
is no FooBar device to test with, no runtime testing was performed.

Signed-off-by: Xiaoke Wang <[email protected]>
---
ChangeLog:
v1->v2 update the description.
v2->v3 update the description.
drivers/staging/rtl8723bs/os_dep/os_intfs.c | 60 +++++++++++----------
1 file changed, 31 insertions(+), 29 deletions(-)

diff --git a/drivers/staging/rtl8723bs/os_dep/os_intfs.c b/drivers/staging/rtl8723bs/os_dep/os_intfs.c
index 0a8c24e..1bf6fc0 100644
--- a/drivers/staging/rtl8723bs/os_dep/os_intfs.c
+++ b/drivers/staging/rtl8723bs/os_dep/os_intfs.c
@@ -664,51 +664,36 @@ void rtw_reset_drv_sw(struct adapter *padapter)

u8 rtw_init_drv_sw(struct adapter *padapter)
{
- u8 ret8 = _SUCCESS;
-
rtw_init_default_value(padapter);

rtw_init_hal_com_default_value(padapter);

- if (rtw_init_cmd_priv(&padapter->cmdpriv)) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (rtw_init_cmd_priv(&padapter->cmdpriv))
+ return _FAIL;

padapter->cmdpriv.padapter = padapter;

- if (rtw_init_evt_priv(&padapter->evtpriv)) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (rtw_init_evt_priv(&padapter->evtpriv))
+ goto free_cmd_priv;

-
- if (rtw_init_mlme_priv(padapter) == _FAIL) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (rtw_init_mlme_priv(padapter) == _FAIL)
+ goto free_evt_priv;

init_mlme_ext_priv(padapter);

- if (_rtw_init_xmit_priv(&padapter->xmitpriv, padapter) == _FAIL) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (_rtw_init_xmit_priv(&padapter->xmitpriv, padapter) == _FAIL)
+ goto free_mlme_ext;

- if (_rtw_init_recv_priv(&padapter->recvpriv, padapter) == _FAIL) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (_rtw_init_recv_priv(&padapter->recvpriv, padapter) == _FAIL)
+ goto free_xmit_priv;
/* add for CONFIG_IEEE80211W, none 11w also can use */
spin_lock_init(&padapter->security_key_mutex);

/* We don't need to memset padapter->XXX to zero, because adapter is allocated by vzalloc(). */
/* memset((unsigned char *)&padapter->securitypriv, 0, sizeof (struct security_priv)); */

- if (_rtw_init_sta_priv(&padapter->stapriv) == _FAIL) {
- ret8 = _FAIL;
- goto exit;
- }
+ if (_rtw_init_sta_priv(&padapter->stapriv) == _FAIL)
+ goto free_recv_priv;

padapter->stapriv.padapter = padapter;
padapter->setband = GHZ24_50;
@@ -719,9 +704,26 @@ u8 rtw_init_drv_sw(struct adapter *padapter)

rtw_hal_dm_init(padapter);

-exit:
+ return _SUCCESS;
+
+free_recv_priv:
+ _rtw_free_recv_priv(&padapter->recvpriv);
+
+free_xmit_priv:
+ _rtw_free_xmit_priv(&padapter->xmitpriv);
+
+free_mlme_ext:
+ free_mlme_ext_priv(&padapter->mlmeextpriv);

- return ret8;
+ rtw_free_mlme_priv(&padapter->mlmepriv);
+
+free_evt_priv:
+ rtw_free_evt_priv(&padapter->evtpriv);
+
+free_cmd_priv:
+ rtw_free_cmd_priv(&padapter->cmdpriv);
+
+ return _FAIL;
}

void rtw_cancel_all_timer(struct adapter *padapter)
--

2022-09-09 08:03:24

by Greg Kroah-Hartman

Subject: Re: [PATCH v3] staging: rtl8723bs: fix potential memory leak in rtw_init_drv_sw()

On Thu, Sep 08, 2022 at 03:31:25PM +0800, [email protected] wrote:
> From: Xiaoke Wang <[email protected]>
>
> In rtw_init_drv_sw(), there are various init functions are called to
> populate the padapter structure and some checks for their return value.
> However, except for the first one error path, the other five error paths
> do not properly release the previous allocated resources, which leads to

You have trailing whitespace on your text :(

> various memory leaks.
>
> This patch fixes them and keeps the success and error separate. As there
> is no FooBar device to test with, no runtime testing was performed.

What do you mean by "FooBar"? Please fix that up.

thanks,

greg k-h