From: Mohammed Shafi Shajakhan Date: Fri, 16 Nov 2012 12:52:40 +0000 (+0530) Subject: ath6kl: Fix kernel panic on continuous driver load/unload X-Git-Url: https://git.stricted.de/?a=commitdiff_plain;h=de2070fc4aa7c0205348010f500f5abce012e67b;p=GitHub%2Fmoto-9609%2Fandroid_kernel_motorola_exynos9610.git ath6kl: Fix kernel panic on continuous driver load/unload On continuous loading and unloading of AR6004 ath6kl USB driver it triggers a panic due to NULL pointer dereference of 'target' pointer. while true; do sudo modprobe -v ath6kl_core; sudo modprobe -v ath6kl_usb; sudo modprobe -r usb; sudo modprobe -r ath6kl_core; done ar->htc_target can be NULL due to a race condition that can occur during driver initialization(we do 'ath6kl_hif_power_on' before initializing 'ar->htc_target' via 'ath6kl_htc_create'). 'ath6kl_hif_power_on' assigns 'ath6kl_recv_complete' as usb_complete_t/callback function for 'usb_fill_bulk_urb'. Thus the possibility of ar->htc_target being NULL via ath6kl_recv_complete -> ath6kl_usb_io_comp_work before even 'ath6kl_htc_create' is finished to initialize ar->htc_create. Worth noting is the obvious solution of doing 'ath6kl_hif_power_on' later(i.e after we are done with 'ath6kl_htc_create', causes a h/w bring up failure in AR6003 SDIO, as 'ath6kl_hif_power_on' is a pre-requisite to get the target version 'ath6kl_bmi_get_target_info'. So simply check for NULL pointer for 'ar->htc_target' and bail out. [23614.518282] BUG: unable to handle kernel NULL pointer dereference at 00000904 [23614.518463] IP: [] __ticket_spin_trylock+0x6/0x30 [23614.518570] *pde = 00000000 [23614.518664] Oops: 0000 [#1] SMP [23614.518795] Modules linked in: ath6kl_usb(O+) ath6kl_core(O) [23614.520012] EIP: 0060:[] EFLAGS: 00010286 CPU: 0 [23614.520012] EIP is at __ticket_spin_trylock+0x6/0x30 Call Trace: [] do_raw_spin_trylock+0x14/0x40 [] _raw_spin_lock_bh+0x52/0x80 [] ? ath6kl_htc_pipe_rx_complete+0x3b4/0x4c0 [ath6kl_core] [] ath6kl_htc_pipe_rx_complete+0x3b4/0x4c0 [ath6kl_core] [] ? skb_dequeue+0x22/0x70 [] ? skb_dequeue+0x22/0x70 [] ath6kl_core_rx_complete+0x12/0x20 [ath6kl_core] [] ath6kl_usb_io_comp_work+0xaa/0xb0 [ath6kl_usb] [] process_one_work+0x1a3/0x5e0 [] ? process_one_work+0x127/0x5e0 [] ? ath6kl_usb_reset_resume+0x30/0x30 [ath6kl_usb] [] worker_thread+0x11e/0x3f0 Kernel panic - not syncing: Fatal exception in interrupt Signed-off-by: Mohammed Shafi Shajakhan Signed-off-by: Kalle Valo --- diff --git a/drivers/net/wireless/ath/ath6kl/htc_pipe.c b/drivers/net/wireless/ath/ath6kl/htc_pipe.c index 73a38f9392f9..281390178e3d 100644 --- a/drivers/net/wireless/ath/ath6kl/htc_pipe.c +++ b/drivers/net/wireless/ath/ath6kl/htc_pipe.c @@ -967,6 +967,22 @@ static int ath6kl_htc_pipe_rx_complete(struct ath6kl *ar, struct sk_buff *skb, u16 payload_len; int status = 0; + /* + * ar->htc_target can be NULL due to a race condition that can occur + * during driver initialization(we do 'ath6kl_hif_power_on' before + * initializing 'ar->htc_target' via 'ath6kl_htc_create'). + * 'ath6kl_hif_power_on' assigns 'ath6kl_recv_complete' as + * usb_complete_t/callback function for 'usb_fill_bulk_urb'. + * Thus the possibility of ar->htc_target being NULL + * via ath6kl_recv_complete -> ath6kl_usb_io_comp_work. + */ + if (WARN_ON_ONCE(!target)) { + ath6kl_err("Target not yet initialized\n"); + status = -EINVAL; + goto free_skb; + } + + netdata = skb->data; netlen = skb->len;