On 04/05/2022 11:22, Arnd Bergmann wrote:
On Wed, May 4, 2022 at 11:09 AM Naresh Kamboju naresh.kamboju@linaro.org wrote:
Following kernel DEBUG_LOCKS_WARN_ON(lock->magic != lock) noticed on arm64 dragonboard 410c device while booting Linux next-20220502 tag kernel.
warning:
[ 6.052275] ------------[ cut here ]------------ [ 6.052289] DEBUG_LOCKS_WARN_ON(lock->magic != lock) [ 6.052314] WARNING: CPU: 3 PID: 57 at kernel/locking/mutex.c:582 __mutex_lock+0x1dc/0x420 [ 6.052345] Modules linked in: [ 6.052361] CPU: 3 PID: 57 Comm: kworker/u8:2 Not tainted 5.18.0-rc5-next-20220502 #1 [ 6.052374] Hardware name: Qualcomm Technologies, Inc. APQ 8016 SBC (DT) [ 6.052384] Workqueue: events_unbound deferred_probe_work_func [ 6.052408] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--) [ 6.052421] pc : __mutex_lock+0x1dc/0x420 [ 6.052434] lr : __mutex_lock+0x1dc/0x420 [ 6.052446] sp : ffff80000c42b9a0 [ 6.052453] x29: ffff80000c42b9a0 x28: ffff80000b314860 x27: ffff00000408810d [ 6.052481] x26: ffff80000adb4bb8 x25: ffff000005220a74 x24: 0000000000000000 [ 6.052506] x23: 0000000000000000 x22: ffff800008bbef8c x21: 0000000000000002 [ 6.052533] x20: 0000000000000000 x19: ffff00000c1a5110 x18: 0000000000000000 [ 6.052558] x17: ffff800008bbd594 x16: ffff800008bc10f4 x15: ffff80000927b6c4 [ 6.052585] x14: 0000000000000000 x13: 284e4f5f4e524157 x12: 5f534b434f4c5f47 [ 6.052611] x11: 656820747563205b x10: 0000000000000029 x9 : ffff8000081cc7b8 [ 6.052637] x8 : 0000000000000029 x7 : 0000000000000013 x6 : 0000000000000001 [ 6.267204] x5 : ffff80000adb5000 x4 : ffff80000adb54f0 x3 : 0000000000000000 [ 6.274321] x2 : 0000000000000000 x1 : 0000000000000000 x0 : ffff000005265080 [ 6.281439] Call trace: [ 6.288524] __mutex_lock+0x1dc/0x420 [ 6.290789] mutex_lock_nested+0x4c/0x90 [ 6.294610] driver_set_override+0x12c/0x160 [ 6.298605] qcom_smd_register_edge+0x2d8/0x52c
The problem is in
static int qcom_smd_create_chrdev(struct qcom_smd_edge *edge) { struct qcom_smd_device *qsdev;
qsdev = kzalloc(sizeof(*qsdev), GFP_KERNEL); if (!qsdev) return -ENOMEM; qsdev->edge = edge; qsdev->rpdev.ops = &qcom_smd_device_ops; qsdev->rpdev.dev.parent = &edge->dev; qsdev->rpdev.dev.release = qcom_smd_release_device; return rpmsg_ctrldev_register_device(&qsdev->rpdev);}
static inline int rpmsg_ctrldev_register_device(struct rpmsg_device *rpdev) { int ret;
strcpy(rpdev->id.name, "rpmsg_ctrl"); ret = driver_set_override(&rpdev->dev, &rpdev->driver_override, rpdev->id.name, strlen(rpdev->id.name)); if (ret) return ret; ret = rpmsg_register_device(rpdev); if (ret) kfree(rpdev->driver_override); return ret;}
This allocates an uninitialized device structure that is then passed driver_set_override() before calling device_register(), so the mutex is not initialized yet.
Folks, this was already reported by Marek and fixed:
https://lore.kernel.org/all/20220429195946.1061725-1-krzysztof.kozlowski@lin...
This has to go via Greg's tree.
Best regards, Krzysztof