2019-09-20 14:35:18

by Larry Finger

[permalink] [raw]
Subject: [PATCH] rtlwifi: rtl_pci: Fix errors when CONFIG_RTLWIFI_DEBUG is not set

When rtlwifi debugging is disabled, missing globals occur due to the
attempt execution of code that is never generated.

Signed-off-by: Larry Finger <[email protected]>
---
drivers/net/wireless/realtek/rtlwifi/pci.c | 4 ++++
1 file changed, 4 insertions(+)

diff --git a/drivers/net/wireless/realtek/rtlwifi/pci.c b/drivers/net/wireless/realtek/rtlwifi/pci.c
index 6087ec7a90a6..2f8aad28f1c0 100644
--- a/drivers/net/wireless/realtek/rtlwifi/pci.c
+++ b/drivers/net/wireless/realtek/rtlwifi/pci.c
@@ -2298,8 +2298,10 @@ int rtl_pci_probe(struct pci_dev *pdev,
}
rtlpriv->mac80211.mac80211_registered = 1;

+#ifdef CONFIG_RTLWIFI_DEBUG
/* add for debug */
rtl_debug_add_one(hw);
+#endif

/*init rfkill */
rtl_init_rfkill(hw); /* Init PCI sw */
@@ -2349,8 +2351,10 @@ void rtl_pci_disconnect(struct pci_dev *pdev)
wait_for_completion(&rtlpriv->firmware_loading_complete);
clear_bit(RTL_STATUS_INTERFACE_START, &rtlpriv->status);

+#ifdef CONFIG_RTLWIFI_DEBUG
/* remove form debug */
rtl_debug_remove_one(hw);
+#endif

/*ieee80211_unregister_hw will call ops_stop */
if (rtlmac->mac80211_registered == 1) {
--
2.23.0


2019-09-20 17:12:14

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] rtlwifi: rtl_pci: Fix errors when CONFIG_RTLWIFI_DEBUG is not set

Larry Finger <[email protected]> writes:

> When rtlwifi debugging is disabled, missing globals occur due to the
> attempt execution of code that is never generated.
>
> Signed-off-by: Larry Finger <[email protected]>

Is this is a new regression or an old problem?

--
Kalle Valo

2019-09-22 19:15:17

by Kalle Valo

[permalink] [raw]
Subject: Re: [PATCH] rtlwifi: rtl_pci: Fix errors when CONFIG_RTLWIFI_DEBUG is not set

Larry Finger <[email protected]> writes:

> When rtlwifi debugging is disabled, missing globals occur due to the
> attempt execution of code that is never generated.

Do you have the exact error message? I don't know why this is failing.

> Signed-off-by: Larry Finger <[email protected]>
> ---
> drivers/net/wireless/realtek/rtlwifi/pci.c | 4 ++++
> 1 file changed, 4 insertions(+)
>
> diff --git a/drivers/net/wireless/realtek/rtlwifi/pci.c b/drivers/net/wireless/realtek/rtlwifi/pci.c
> index 6087ec7a90a6..2f8aad28f1c0 100644
> --- a/drivers/net/wireless/realtek/rtlwifi/pci.c
> +++ b/drivers/net/wireless/realtek/rtlwifi/pci.c
> @@ -2298,8 +2298,10 @@ int rtl_pci_probe(struct pci_dev *pdev,
> }
> rtlpriv->mac80211.mac80211_registered = 1;
>
> +#ifdef CONFIG_RTLWIFI_DEBUG
> /* add for debug */
> rtl_debug_add_one(hw);
> +#endif
>
> /*init rfkill */
> rtl_init_rfkill(hw); /* Init PCI sw */
> @@ -2349,8 +2351,10 @@ void rtl_pci_disconnect(struct pci_dev *pdev)
> wait_for_completion(&rtlpriv->firmware_loading_complete);
> clear_bit(RTL_STATUS_INTERFACE_START, &rtlpriv->status);
>
> +#ifdef CONFIG_RTLWIFI_DEBUG
> /* remove form debug */
> rtl_debug_remove_one(hw);
> +#endif

Then I checked debug.h I see that it's supposed to be handled already:

#ifdef CONFIG_RTLWIFI_DEBUG
void rtl_debug_add_one(struct ieee80211_hw *hw);
void rtl_debug_remove_one(struct ieee80211_hw *hw);
void rtl_debugfs_add_topdir(void);
void rtl_debugfs_remove_topdir(void);
#else
#define rtl_debug_add_one(hw)
#define rtl_debug_remove_one(hw)
#define rtl_debugfs_add_topdir()
#define rtl_debugfs_remove_topdir()
#endif

But the else part looks weird, I'm used to using inline functions
instead of defines:

static inline void rtl_debug_add_one(struct ieee80211_hw *hw) {}
static inline void rtl_debug_remove_one(struct ieee80211_hw *hw) {}
static inline void rtl_debugfs_add_topdir(void) {}
static inline void rtl_debugfs_remove_topdir(void) {}

Inline functions also make it possible for the compiler to check the
types are correct.

--
Kalle Valo