After setting up functionfs for adb w/ 4.1-rc7, I noticed some flakey behavior. I enabled some lock debugging and got the following:
[ 91.648093] read strings [ 91.650264] g_ffs gadget: g_ffs ready [ 91.652551] ci_hdrc ci_hdrc.0: CI_HDRC_CONTROLLER_RESET_EVENT received [ 96.068693] BUG: spinlock lockup suspected on CPU#0, adbd/2791 [ 96.068751] lock: 0xe7764880, .magic: e7764880, .owner: <none>/-1, .owner_cpu: -407539900 [ 96.073448] CPU: 0 PID: 2791 Comm: adbd Not tainted 4.1.0-rc1-00032-g359b12f #147 [ 96.081688] Hardware name: Qualcomm (Flattened Device Tree) [ 96.089266] [<c0216ac8>] (unwind_backtrace) from [<c02136a8>] (show_stack+0x10/0x14) [ 96.094635] [<c02136a8>] (show_stack) from [<c075d9fc>] (dump_stack+0x70/0xbc) [ 96.102627] [<c075d9fc>] (dump_stack) from [<c026ef90>] (do_raw_spin_lock+0x114/0x1a0) [ 96.109661] [<c026ef90>] (do_raw_spin_lock) from [<c0764cb8>] (_raw_spin_lock_irqsave+0x50/0x5c) [ 96.117560] [<c0764cb8>] (_raw_spin_lock_irqsave) from [<c037c1a0>] (kiocb_set_cancel_fn+0x1c/0x60) [ 96.126519] [<c037c1a0>] (kiocb_set_cancel_fn) from [<c05ae568>] (ffs_epfile_read_iter+0x8c/0x140) [ 96.135289] [<c05ae568>] (ffs_epfile_read_iter) from [<c0332018>] (__vfs_read+0xb0/0xd4) [ 96.144290] [<c0332018>] (__vfs_read) from [<c0332ef8>] (vfs_read+0x7c/0x100) [ 96.152535] [<c0332ef8>] (vfs_read) from [<c0332fbc>] (SyS_read+0x40/0x8c) [ 96.159571] [<c0332fbc>] (SyS_read) from [<c020ff20>] (ret_fast_syscall+0x0/0x4c) [ 117.678633] INFO: rcu_preempt detected stalls on CPUs/tasks: [ 117.683069] 0: (1 GPs behind) idle=805/140000000000000/0 softirq=7187/7189 fqs=2601 [ 117.683316] (detected by 3, t=2603 jiffies, g=3028, c=3027, q=474) [ 117.692345] Task dump for CPU 0: [ 117.697202] adbd R running 0 2791 1 0x00000002 [ 117.704296] [<c075f234>] (__schedule) from [<ffffffff>] (0xffffffff) It seems we're stuck on the kioctx.ctx_lock, and reviewing that lock usage I don't see any problems in fs/aio.c right off. So I'm guessing the f_fs.c code is somehow not initializing the lock structure, or maybe calling kiocb_set_cancel_fn() from the wrong context? Anyway, I was curious if anyone else has seen similar issues or had suggestions for further debugging. I've seen this issue as well with 4.1-rc1. thanks -john -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majord...@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
