kernel crash on ARM after ifconfig up

[sporockel]

Hi, I've tried to get a RTL8192Du running on a ARM platform, kernel 3.14.0-rc6. The same with 3.12.0. Any idea where to start searching?

insmod /lib/modules/3.14.0-rc6/kernel/drivers/net/wireless/8192du.ko
[ 1507.776697] r8192du: EEPROM type is E-FUSE, E-CUT chip
[ 1508.023634] r8192du: MacPhyMode: SINGLEMAC_SINGLEPHY
[ 1508.055153] usbcore: registered new interface driver rtl8192du

ifconfig wlan1 up

[ 1517.392721] 
[ 1517.394324] =============================================
[ 1517.399964] [ INFO: possible recursive locking detected ]
[ 1517.405609] 3.14.0-rc6 #1 Tainted: G           O
[ 1517.410427] ---------------------------------------------
[ 1517.416065] ifconfig/184 is trying to acquire lock:
[ 1517.421159]  (pmutex){+.+...}, at: [<bf0a9ec8>] usbctrl_vendorreq+0xac/0x208 [8192du]
[ 1517.429727] 
[ 1517.429727] but task is already holding lock:
[ 1517.435824]  (pmutex){+.+...}, at: [<bf0ac6dc>] netdev_open+0x1c/0x3c [8192du]
[ 1517.443565] 
[ 1517.443565] other info that might help us debug this:
[ 1517.450384]  Possible unsafe locking scenario:
[ 1517.450384] 
[ 1517.456564]        CPU0
[ 1517.459117]        ----
[ 1517.461669]   lock(pmutex);
[ 1517.464600]   lock(pmutex);
[ 1517.467531] 
[ 1517.467531]  *** DEADLOCK ***
[ 1517.467531] 
[ 1517.473716]  May be due to missing lock nesting notation
[ 1517.473716] 
[ 1517.480807] 2 locks held by ifconfig/184:
[ 1517.484991]  #0:  (rtnl_mutex){+.+.+.}, at: [<c045cb58>] devinet_ioctl+0xd0/0x764
[ 1517.492867]  #1:  (pmutex){+.+...}, at: [<bf0ac6dc>] netdev_open+0x1c/0x3c [8192du]
[ 1517.501043] 
[ 1517.501043] stack backtrace:
[ 1517.505606] CPU: 0 PID: 184 Comm: ifconfig Tainted: G           O 3.14.0-rc6 #1
[ 1517.513279] [<c00137b8>] (unwind_backtrace) from [<c00115bc>] (show_stack+0x10/0x14)
[ 1517.521386] [<c00115bc>] (show_stack) from [<c0068bb8>] (validate_chain.isra.33+0xd6c/0x1178)
[ 1517.530302] [<c0068bb8>] (validate_chain.isra.33) from [<c006a1a4>] (__lock_acquire+0x3cc/0xae4)
[ 1517.539485] [<c006a1a4>] (__lock_acquire) from [<c006b060>] (lock_acquire+0x68/0x7c)
[ 1517.547589] [<c006b060>] (lock_acquire) from [<c051fd6c>] (mutex_lock_nested+0x54/0x3a0)
[ 1517.556189] [<c051fd6c>] (mutex_lock_nested) from [<bf0a9ec8>] (usbctrl_vendorreq+0xac/0x208 [8192du])
[ 1517.566158] [<bf0a9ec8>] (usbctrl_vendorreq [8192du]) from [<bf0aa08c>] (usb_write_reg+0x68/0x70 [8192du])
[ 1517.576486] [<bf0aa08c>] (usb_write_reg [8192du]) from [<bf0aa15c>] (usb_write8+0x20/0x28 [8192du])
[ 1517.586134] [<bf0aa15c>] (usb_write8 [8192du]) from [<bf06c708>] (_rtw_write8+0x1c/0x24 [8192du])
[ 1517.595597] [<bf06c708>] (_rtw_write8 [8192du]) from [<bf0a4bfc>] (rtl8192du_hal_init+0x164/0xb38 [8192du])
[ 1517.605995] [<bf0a4bfc>] (rtl8192du_hal_init [8192du]) from [<bf0922c8>] (rtw_hal_init+0x30/0x114 [8192du])
[ 1517.616396] [<bf0922c8>] (rtw_hal_init [8192du]) from [<bf0ac330>] (_netdev_open+0x98/0x428 [8192du])
[ 1517.626266] [<bf0ac330>] (_netdev_open [8192du]) from [<bf0ac6e4>] (netdev_open+0x24/0x3c [8192du])
[ 1517.635853] [<bf0ac6e4>] (netdev_open [8192du]) from [<c0407008>] (__dev_open+0xb8/0x11c)
[ 1517.644409] [<c0407008>] (__dev_open) from [<c0407290>] (__dev_change_flags+0x94/0x168)
[ 1517.652780] [<c0407290>] (__dev_change_flags) from [<c040737c>] (dev_change_flags+0x18/0x48)
[ 1517.661602] [<c040737c>] (dev_change_flags) from [<c045d11c>] (devinet_ioctl+0x694/0x764)
[ 1517.670169] [<c045d11c>] (devinet_ioctl) from [<c03ee2cc>] (sock_ioctl+0x1d8/0x2a8)
[ 1517.678181] [<c03ee2cc>] (sock_ioctl) from [<c00e9c68>] (do_vfs_ioctl+0x3fc/0x5f0)
[ 1517.686093] [<c00e9c68>] (do_vfs_ioctl) from [<c00e9ec8>] (SyS_ioctl+0x6c/0x7c)
[ 1517.693736] [<c00e9ec8>] (SyS_ioctl) from [<c000e340>] (ret_fast_syscall+0x0/0x48)
[ 1517.708668] r8192du: MacPhyMode: SINGLEMAC_SINGLEPHY
[ 1517.771356] r8192du: Loaded firmware file rtlwifi/rtl8192dufw.bin of 32302 bytes
[ 1519.031042] pdmpriv->TxPowerTrackControl = 1
[ 1519.049144] IPv6: ADDRCONF(NETDEV_UP): wlan1: link is not ready

[lwfinger]

What crash? That lock dependency warning does not interrupt operations. If you do not ewant to see it, turn off lock dependency checking in your kernel.

If I knew what to fix, I would do it; however, I have more important thingd to look after.

[sporockel]

ok, it's not a crash, however iw list does not show a wireless interface and I thought it may have to do with the output. Maybe I'm overlooking something.

[lwfinger]

I do not see anything either.

Which branch of the repo are you running? If you are not using the kernel-version branch, please switch to it.

The only difference I see in your device is that yours is SINGLEMAC_SINGLEPHY while mine is DOUBLEMAC_DOUBLEPHY. That means I have both 2.4 GHz and 5 GHz radios, and both can work at the same time. It shouldn't make a difference, but who knows.

[sporockel]

it tried both, but there is no difference. I'm going to do further investigations in April, now it's vacation time :-). Thanks for reply

[lwfinger]

Have a good vacation. When you return, it will be better if you use the keernel-version branch as that is the one I am currently working on.

[mun1r]

hi lwfinger.how to compile the kernel-version branch.I have an error when make process.my kernel version is 3.2.0-60-generic

[lwfinger]

The kernel-version branch is only for kernel 3.14, although 3.13 may work. For old kernels such as 3.0, you must use the master branch!

[mun1r]

Oh i see thank you

[sporockel]

Hello,

I can confirm with latest kernel-version (#489e933c98f45ca793a995f747282d764c020b5e) the issue is gone.

[okligaofeng]

我也遇到了相同的问题最后解决了，我这边的原因：1.内核抢占模式为Low-Latency Desktop，改为自主强占 2.DEBUG_SPINLOCK需要禁掉，满足1.2才可以实现unlock and unlock_irq functions are inlined，说到底都是因为IO Schedulers不同造成的

[lwfinger]

I used Google Translate as I do not read Chinese to get the following:

"I also encountered the same problem finally solved, my side of reasons: 1. kernel preemption mode as Low-Latency Desktop, need to customize the seizure 2.DEBUG_SPINLOCK ban to meet before they can achieve 1.2 unlock and unlock_irq functions are inlined in the final analysis are caused because different IO Schedulers"

Sorry, but I do not understand what you are saying. Is there a patch that solves the problem? If so, perhaps I can understand it.

[okligaofeng]

kernel .config patch: @@ -219,7 +214,11 @@ CONFIG_IOSCHED_CFQ=y CONFIG_DEFAULT_CFQ=y

CONFIG_DEFAULT_NOOP is not set

CONFIG_DEFAULT_IOSCHED="cfq" -CONFIG_UNINLINE_SPIN_UNLOCK=y +CONFIG_INLINE_SPIN_UNLOCK_IRQ=y +CONFIG_INLINE_READ_UNLOCK=y +CONFIG_INLINE_READ_UNLOCK_IRQ=y +CONFIG_INLINE_WRITE_UNLOCK=y +CONFIG_INLINE_WRITE_UNLOCK_IRQ=y CONFIG_FREEZER=y

-# CONFIG_PREEMPT_VOLUNTARY is not set -CONFIG_PREEMPT=y -CONFIG_PREEMPT_COUNT=y +CONFIG_PREEMPT_VOLUNTARY=y +# CONFIG_PREEMPT is not set CONFIG_HZ_FIXED=0

Sorry, My English poll. I have found the solution . The issues to be further . I just know the issues to be caused by spinlock.