Re: general protection fault in tomoyo_socket_sendmsg_permission

[Shuah Khan]

On 1/29/21 7:25 PM, Tetsuo Handa wrote:
> On 2021/01/30 6:18, Shuah Khan wrote:
> > In this console log:
>
> It seems "this console log" refers to https://syzkaller.appspot.com/x/log.txt?x=10453034500000 .
>
> > 06:57:50 executing program 1:
> > socketpair$tipc(0x1e, 0x2, 0x0, &(0x7f00000000c0)={<r0=>0xffffffffffffffff})
> > sendmsg$BATADV_CMD_GET_TRANSTABLE_LOCAL(r0, &(0x7f00000002c0)={&(0x7f00000001c0), 0xc, &(0x7f0000000280)={0x0, 0xd001010000000000}}, 0x0)
> >
> > [ 1151.090883][T23361] vhci_hcd vhci_hcd.0: pdev(4) rhport(0) sockfd(4)
> > [ 1151.097445][T23361] vhci_hcd vhci_hcd.0: devid(0) speed(1) speed_str(low-speed)
> > 06:57:50 executing program 0:
> > r0 = syz_open_dev$binderN(&(0x7f0000000680)='/dev/binder#\x00', 0x0, 0x0)
> > ioctl$BINDER_WRITE_READ(r0, 0xc0306201, &(0x7f0000000cc0)={0x88, 0x0, &(0x7f0000000b80)=[@transaction={0x40406300, {0x2, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}}, @transaction={0x40406300, {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}}], 0x0, 0x0, 0x0})
> >
> > [ 1151.164402][T23363] vhci_hcd: connection closed
> > [ 1151.167346][  T240] vhci_hcd: stop threads
> >
> >
> > [ 1151.178329][T26761] usb 17-1: new low-speed USB device number 2 using vhci_hcd
> >
> >
> > SK: Looking at the console log, it looks like while connection is being
> >      torn down,
>
> Excuse me, but it looks like (what comes here) while connection is being torn down ?
> I'm not familiar with driver code.
>
> > [ 1151.181245][  T240] vhci_hcd: release socket
> >
> >
> > Can you share your your test code for this program:
> > "executing program 1"
>
> I don't think program 1 is relevant. I think program 4
>
>    06:57:50 executing program 4:
>    r0 = socket$tipc(0x1e, 0x2, 0x0)
>    syz_usbip_server_init(0x1)
>    close_range(r0, 0xffffffffffffffff, 0x0)
>
> which calls syz_usbip_server_init() as with other duplicates is relevant.
>
> > Also your setup? Do you run usbip_host and vhci_hcd both?
>
> Who are you referring to with "you/your" ? I'm not running syzkaller in my setup
> and I don't have test code.
>
> I'm just proposing printing more messages in order to confirm the ordering of
> events and member values in structures.

I am looking to understand the syzbot configuration and a reproducer
to be able to debug and fix the problem. How is syzbot triggering the
vhci_hcd attach and detach sequence?

This helps me determine all these fix suggestions that are coming in
are fixes or papering over a real problem.

thanks,
-- Shuah