*BUG 1:
In rtl_pci_probe(), when rtlpriv->cfg->ops->init_sw_vars() fails,
rtl_deinit_core() in the error handling code is executed.
rtl_deinit_core() calls rtl_free_entries_from_scan_list(), which uses
rtlpriv->scan_list.list in list_for_each_entry_safe(), but it has been
initialized. Thus a null-pointer dereference occurs.
The reason is that rtlpriv->scan_list.list is initialized by
INIT_LIST_HEAD() in rtl_init_core(), which has not been called.
To fix this bug, rtl_deinit_core() should not be called when
rtlpriv->cfg->ops->init_sw_vars() fails.
*BUG 2:
In rtl_pci_probe(), rtl_init_core() can fail when rtl_regd_init() in
this function fails, and rtlpriv->scan_list.list has not been
initialized by INIT_LIST_HEAD(). Then, rtl_deinit_core() in the error
handling code of rtl_pci_probe() is executed. Finally, a null-pointer
dereference occurs due to the same reason of the above bug.
To fix this bug, the initialization of lists in rtl_init_core() are
performed before the call to rtl_regd_init().
These bugs are found by a runtime fuzzing tool named FIZZER written by
us.
Signed-off-by: Jia-Ju Bai <[email protected]>
---
drivers/net/wireless/realtek/rtlwifi/base.c | 15 ++++++++-------
drivers/net/wireless/realtek/rtlwifi/pci.c | 4 ++--
2 files changed, 10 insertions(+), 9 deletions(-)
diff --git a/drivers/net/wireless/realtek/rtlwifi/base.c b/drivers/net/wireless/realtek/rtlwifi/base.c
index 217d2a7a43c7..b3f341ec3710 100644
--- a/drivers/net/wireless/realtek/rtlwifi/base.c
+++ b/drivers/net/wireless/realtek/rtlwifi/base.c
@@ -526,8 +526,14 @@ int rtl_init_core(struct ieee80211_hw *hw)
/* <2> rate control register */
hw->rate_control_algorithm = "rtl_rc";
+ /* <3> init list */
+ INIT_LIST_HEAD(&rtlpriv->entry_list);
+ INIT_LIST_HEAD(&rtlpriv->scan_list.list);
+ skb_queue_head_init(&rtlpriv->tx_report.queue);
+ skb_queue_head_init(&rtlpriv->c2hcmd_queue);
+
/*
- * <3> init CRDA must come after init
+ * <4> init CRDA must come after init
* mac80211 hw in _rtl_init_mac80211.
*/
if (rtl_regd_init(hw, rtl_reg_notifier)) {
@@ -535,7 +541,7 @@ int rtl_init_core(struct ieee80211_hw *hw)
return 1;
}
- /* <4> locks */
+ /* <5> locks */
mutex_init(&rtlpriv->locks.conf_mutex);
mutex_init(&rtlpriv->locks.ips_mutex);
mutex_init(&rtlpriv->locks.lps_mutex);
@@ -550,11 +556,6 @@ int rtl_init_core(struct ieee80211_hw *hw)
spin_lock_init(&rtlpriv->locks.cck_and_rw_pagea_lock);
spin_lock_init(&rtlpriv->locks.fw_ps_lock);
spin_lock_init(&rtlpriv->locks.iqk_lock);
- /* <5> init list */
- INIT_LIST_HEAD(&rtlpriv->entry_list);
- INIT_LIST_HEAD(&rtlpriv->scan_list.list);
- skb_queue_head_init(&rtlpriv->tx_report.queue);
- skb_queue_head_init(&rtlpriv->c2hcmd_queue);
rtlmac->link_state = MAC80211_NOLINK;
diff --git a/drivers/net/wireless/realtek/rtlwifi/pci.c b/drivers/net/wireless/realtek/rtlwifi/pci.c
index 48ca52102cef..864cb76230c4 100644
--- a/drivers/net/wireless/realtek/rtlwifi/pci.c
+++ b/drivers/net/wireless/realtek/rtlwifi/pci.c
@@ -2267,7 +2267,7 @@ int rtl_pci_probe(struct pci_dev *pdev,
if (rtlpriv->cfg->ops->init_sw_vars(hw)) {
pr_err("Can't init_sw_vars\n");
err = -ENODEV;
- goto fail3;
+ goto fail2;
}
rtlpriv->cfg->ops->init_sw_leds(hw);
--
2.17.0
Jia-Ju Bai <[email protected]> wrote:
> *BUG 1:
> In rtl_pci_probe(), when rtlpriv->cfg->ops->init_sw_vars() fails,
> rtl_deinit_core() in the error handling code is executed.
> rtl_deinit_core() calls rtl_free_entries_from_scan_list(), which uses
> rtlpriv->scan_list.list in list_for_each_entry_safe(), but it has been
> initialized. Thus a null-pointer dereference occurs.
> The reason is that rtlpriv->scan_list.list is initialized by
> INIT_LIST_HEAD() in rtl_init_core(), which has not been called.
>
> To fix this bug, rtl_deinit_core() should not be called when
> rtlpriv->cfg->ops->init_sw_vars() fails.
>
> *BUG 2:
> In rtl_pci_probe(), rtl_init_core() can fail when rtl_regd_init() in
> this function fails, and rtlpriv->scan_list.list has not been
> initialized by INIT_LIST_HEAD(). Then, rtl_deinit_core() in the error
> handling code of rtl_pci_probe() is executed. Finally, a null-pointer
> dereference occurs due to the same reason of the above bug.
>
> To fix this bug, the initialization of lists in rtl_init_core() are
> performed before the call to rtl_regd_init().
>
> These bugs are found by a runtime fuzzing tool named FIZZER written by
> us.
>
> Signed-off-by: Jia-Ju Bai <[email protected]>
Ping & Larry, is this ok to take?
--
https://patchwork.kernel.org/patch/10942971/
https://wireless.wiki.kernel.org/en/developers/documentation/submittingpatches
On 5/28/19 6:55 AM, Kalle Valo wrote:
> Jia-Ju Bai <[email protected]> wrote:
>
>> *BUG 1:
>> In rtl_pci_probe(), when rtlpriv->cfg->ops->init_sw_vars() fails,
>> rtl_deinit_core() in the error handling code is executed.
>> rtl_deinit_core() calls rtl_free_entries_from_scan_list(), which uses
>> rtlpriv->scan_list.list in list_for_each_entry_safe(), but it has been
>> initialized. Thus a null-pointer dereference occurs.
>> The reason is that rtlpriv->scan_list.list is initialized by
>> INIT_LIST_HEAD() in rtl_init_core(), which has not been called.
>>
>> To fix this bug, rtl_deinit_core() should not be called when
>> rtlpriv->cfg->ops->init_sw_vars() fails.
>>
>> *BUG 2:
>> In rtl_pci_probe(), rtl_init_core() can fail when rtl_regd_init() in
>> this function fails, and rtlpriv->scan_list.list has not been
>> initialized by INIT_LIST_HEAD(). Then, rtl_deinit_core() in the error
>> handling code of rtl_pci_probe() is executed. Finally, a null-pointer
>> dereference occurs due to the same reason of the above bug.
>>
>> To fix this bug, the initialization of lists in rtl_init_core() are
>> performed before the call to rtl_regd_init().
>>
>> These bugs are found by a runtime fuzzing tool named FIZZER written by
>> us.
>>
>> Signed-off-by: Jia-Ju Bai <[email protected]>
>
> Ping & Larry, is this ok to take?
>
Kalle,
Not at the moment. In reviewing the code, I was unable to see how this situation
could develop, and his backtrace did not mention any rtlwifi code. For that
reason, I asked him to add printk stat4ements to show the last part of rtl_pci
that executed correctly. In
https://marc.info/?l=linux-wireless&m=155788322631134&w=2, he promised to do
that, but I have not seen the result.
Larry
Larry Finger <[email protected]> writes:
> On 5/28/19 6:55 AM, Kalle Valo wrote:
>> Jia-Ju Bai <[email protected]> wrote:
>>
>>> *BUG 1:
>>> In rtl_pci_probe(), when rtlpriv->cfg->ops->init_sw_vars() fails,
>>> rtl_deinit_core() in the error handling code is executed.
>>> rtl_deinit_core() calls rtl_free_entries_from_scan_list(), which uses
>>> rtlpriv->scan_list.list in list_for_each_entry_safe(), but it has been
>>> initialized. Thus a null-pointer dereference occurs.
>>> The reason is that rtlpriv->scan_list.list is initialized by
>>> INIT_LIST_HEAD() in rtl_init_core(), which has not been called.
>>>
>>> To fix this bug, rtl_deinit_core() should not be called when
>>> rtlpriv->cfg->ops->init_sw_vars() fails.
>>>
>>> *BUG 2:
>>> In rtl_pci_probe(), rtl_init_core() can fail when rtl_regd_init() in
>>> this function fails, and rtlpriv->scan_list.list has not been
>>> initialized by INIT_LIST_HEAD(). Then, rtl_deinit_core() in the error
>>> handling code of rtl_pci_probe() is executed. Finally, a null-pointer
>>> dereference occurs due to the same reason of the above bug.
>>>
>>> To fix this bug, the initialization of lists in rtl_init_core() are
>>> performed before the call to rtl_regd_init().
>>>
>>> These bugs are found by a runtime fuzzing tool named FIZZER written by
>>> us.
>>>
>>> Signed-off-by: Jia-Ju Bai <[email protected]>
>>
>> Ping & Larry, is this ok to take?
>>
>
> Kalle,
>
> Not at the moment. In reviewing the code, I was unable to see how this
> situation could develop, and his backtrace did not mention any rtlwifi
> code. For that reason, I asked him to add printk stat4ements to show
> the last part of rtl_pci that executed correctly. In
> https://marc.info/?l=linux-wireless&m=155788322631134&w=2, he promised
> to do that, but I have not seen the result.
Ok, thanks. I'll then drop this, please resubmit once everything is
understood and the patch is ready to be applied.
--
Kalle Valo