rinato: Warnings related to USB serial gadget

[casept]

Kernel splat:

Oct 11 07:05:45 rinato kernel: ------------[ cut here ]------------
Oct 11 07:05:45 rinato kernel: WARNING: CPU: 1 PID: 233 at drivers/usb/gadget/function/u_serial.c:695 gs_close (drivers/usb/gadget/function/u_serial.c:695 (discriminator 3)) 
Oct 11 07:05:45 rinato kernel: Modules linked in:
Oct 11 07:05:45 rinato kernel: Tainted: [W]=WARN
Oct 11 07:05:45 rinato kernel: Hardware name: Samsung Exynos (Flattened Device Tree)
Oct 11 07:05:45 rinato kernel: Call trace: 
Oct 11 07:05:45 rinato kernel: (unwind_backtrace) from show_stack (arch/arm/kernel/traps.c:259) 
Oct 11 07:05:45 rinato kernel: (show_stack) from dump_stack_lvl (lib/dump_stack.c:122) 
Oct 11 07:05:45 rinato kernel: (dump_stack_lvl) from __warn (./arch/arm/include/asm/current.h:30 kernel/panic.c:753) 
Oct 11 07:05:45 rinato kernel: (__warn) from warn_slowpath_fmt (./include/linux/context_tracking.h:159 kernel/panic.c:777) 
Oct 11 07:05:45 rinato kernel: (warn_slowpath_fmt) from gs_close (drivers/usb/gadget/function/u_serial.c:695 (discriminator 3)) 
Oct 11 07:05:45 rinato kernel: (gs_close) from tty_release (drivers/tty/tty_io.c:1791) 
Oct 11 07:05:45 rinato kernel: (tty_release) from __fput (fs/file_table.c:451 (discriminator 1)) 
Oct 11 07:05:45 rinato kernel: (__fput) from sys_close (fs/open.c:1582 fs/open.c:1564) 
Oct 11 07:05:45 rinato kernel: (sys_close) from ret_fast_syscall (arch/arm/kernel/entry-common.S:68) 
Oct 11 07:05:45 rinato kernel: Exception stack(0xe091dfa8 to 0xe091dff0)
Oct 11 07:05:45 rinato kernel: dfa0:                   00000000 b63da828 00000004 0002b448 00000000 00000001
Oct 11 07:05:45 rinato kernel: dfc0: 00000000 b63da828 00000003 00000006 00000003 be8e75d4 be8e7778 be8e772c
Oct 11 07:05:45 rinato kernel: dfe0: b6d5cc90 be8e7448 b6be6b68 b69d779c
Oct 11 07:05:45 rinato kernel: irq event stamp: 11790
Oct 11 07:05:45 rinato kernel: hardirqs last enabled at (11789): finish_task_switch (./arch/arm/include/asm/irqflags.h:39 kernel/sched/sched.h:1533 kernel/sched/core.c:5129 kernel/sched/core.c:5247) 
Oct 11 07:05:46 rinato kernel: hardirqs last disabled at (11790): _raw_spin_lock_irq (./include/linux/spinlock_api_smp.h:117 (discriminator 1) kernel/locking/spinlock.c:170 (discriminator 1)) 
Oct 11 07:05:46 rinato kernel: softirqs last enabled at (11622): vfp_support_entry (arch/arm/vfp/vfpmodule.c:781) 
Oct 11 07:05:46 rinato kernel: softirqs last disabled at (11620): vfp_support_entry (./include/linux/bottom_half.h:20 arch/arm/vfp/vfpmodule.c:73 arch/arm/vfp/vfpmodule.c:717) 
Oct 11 07:05:46 rinato kernel: ---[ end trace 0000000000000000 ]---

[casept]

u_serial.c:695 should only be reachable if gs_open() is called while gs_close() is in progress, in a narrow window where gs_close() releases the lock to permit data to be drained. According to commit c6561082b833a this shouldn't be possible, as the portmaster should itself be holding a lock to prevent this.

next step: Figure out what the portmaster is in our case and inspect it. Also add some printk's in the serial gadget driver to confirm our theory.

[casept]

This theory seems wrong, as the warning is not actually triggered as a result of the goto jump, but rather because gs_close is called one time too many from the tty subsystem.