Re: [RFC PATCH kernel] serial_core: Check for port state when tty is in error state

[Greg Kroah-Hartman <gregkh@linuxfoundation.org>]

On Tue, Jul 28, 2020 at 10:43:59PM +1000, Alexey Kardashevskiy wrote:
> At the moment opening a serial device node (such as /dev/ttyS3)
> succeeds even if there is no actual serial device behind it.
> Reading/writing/ioctls (most) expectantly fail as the uart port is not
> initialized (the type is PORT_UNKNOWN) and the TTY_IO_ERROR error state
> bit is set fot the tty.

That is only if there is no ldisc set for the port, right?  I don't
think that always will be the case if the port is not initialized.

Yes, we do clear this on port open, but we clear it before the
->activate() callback happens.

Why not check for initialized instead?  That would seem to be what you
want to do here instead of checking for an io error.

> However syzkaller (a syscall fuzzer) found that setting line discipline
> does not have these checks all the way down to io_serial_out() in
> 8250_port.c (8250 is the default choice made by univ8250_console_init()).
> As the result of PORT_UNKNOWN, uart_port::iobase is NULL which
> a platform translates onto some address accessing which produces a crash
> like below.
> 
> This adds tty_io_error() to uart_set_ldisc() to prevent the crash.
> 
> The example of crash on PPC64/pseries:
> 
> BUG: Unable to handle kernel data access on write at 0xc00a000000000001
> Faulting instruction address: 0xc000000000c9c9cc
> cpu 0x0: Vector: 300 (Data Access) at [c00000000c6d7800]
>     pc: c000000000c9c9cc: io_serial_out+0xcc/0xf0
>     lr: c000000000c9c9b4: io_serial_out+0xb4/0xf0
>     sp: c00000000c6d7a90
>    msr: 8000000000009033
>    dar: c00a000000000001
>  dsisr: 42000000
>   current = 0xc00000000cd22500
>   paca    = 0xc0000000035c0000   irqmask: 0x03   irq_happened: 0x01
>     pid   = 1371, comm = syz-executor.0
> Linux version 5.8.0-rc7-le-guest_syzkaller_a+fstn1 (aik@fstn1-p1) (gcc (Ubunt
> untu) 2.30) #660 SMP Tue Jul 28 22:29:22 AEST 2020
> enter ? for help
> [c00000000c6d7a90] c0000000018a8cc0 _raw_spin_lock_irq+0xb0/0xe0 (unreliable)
> [c00000000c6d7ad0] c000000000c9bdc0 serial8250_do_set_ldisc+0x140/0x180
> [c00000000c6d7b10] c000000000c9bea4 serial8250_set_ldisc+0xa4/0xb0
> [c00000000c6d7b50] c000000000c91138 uart_set_ldisc+0xb8/0x160
> [c00000000c6d7b90] c000000000c5a22c tty_set_ldisc+0x23c/0x330
> [c00000000c6d7c20] c000000000c4c220 tty_ioctl+0x990/0x12f0
> [c00000000c6d7d20] c00000000056357c ksys_ioctl+0x14c/0x180
> [c00000000c6d7d70] c0000000005635f0 sys_ioctl+0x40/0x60
> [c00000000c6d7db0] c00000000003b814 system_call_exception+0x1a4/0x330
> [c00000000c6d7e20] c00000000000d368 system_call_common+0xe8/0x214
> 
> Signed-off-by: Alexey Kardashevskiy <aik@ozlabs.ru>
> ---
> 
> While looking at it, I noticed that a bunch of callbacks are prone to
> this bug and since I wanted to fix them all with minimum effort,
> I tried checking for PORT_UNKNOWN in uart_port_check() but it breaks
> device opening. Another approach could be checking for uart_port::iobase
> in 8250 (and probably uart_port::membase as well) but this will make
> the rest of the code to think the device is ok while there is no device
> at all.
> 
> What would the correct approach be and what is the expectation?

We should probably check tty_port_initialized() on these code paths
better, care to fix that up?

thanks,

greg k-h