Re: WARNING in dummy_free_request

[Andrey Konovalov]

On Fri, Feb 21, 2020 at 5:41 PM Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
>
> On Fri, 21 Feb 2020, syzbot wrote:
>
> > Hello,
> >
> > syzbot found the following crash on:
> >
> > HEAD commit:    7f0cd6c7 usb: gadget: add raw-gadget interface
> > git tree:       https://github.com/google/kasan.git usb-fuzzer
> > console output: https://syzkaller.appspot.com/x/log.txt?x=17b58e5ee00000
> > kernel config:  https://syzkaller.appspot.com/x/.config?x=f10b12ae04e03319
> > dashboard link: https://syzkaller.appspot.com/bug?extid=55ae006e0a1feae5aeab
> > compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> >
> > Unfortunately, I don't have any reproducer for this crash yet.
> >
> > IMPORTANT: if you fix the bug, please add the following tag to the commit:
> > Reported-by: syzbot+55ae006e0a1feae5aeab@xxxxxxxxxxxxxxxxxxxxxxxxx
> >
> > ------------[ cut here ]------------
> > WARNING: CPU: 1 PID: 19332 at drivers/usb/gadget/udc/dummy_hcd.c:679 dummy_free_request+0x6c/0x80 drivers/usb/gadget/udc/dummy_hcd.c:679
> > Kernel panic - not syncing: panic_on_warn set ...
> > CPU: 1 PID: 19332 Comm: syz-executor.5 Not tainted 5.6.0-rc1-syzkaller #0
> > Hardware name: Google Google Compute Engine/Google Compute Engine, BIOS Google 01/01/2011
> > Call Trace:
> >  __dump_stack lib/dump_stack.c:77 [inline]
> >  dump_stack+0xef/0x16e lib/dump_stack.c:118
> >  panic+0x2aa/0x6e1 kernel/panic.c:221
> >  __warn.cold+0x2f/0x30 kernel/panic.c:582
> >  report_bug+0x27b/0x2f0 lib/bug.c:195
> >  fixup_bug arch/x86/kernel/traps.c:174 [inline]
> >  fixup_bug arch/x86/kernel/traps.c:169 [inline]
> >  do_error_trap+0x12b/0x1e0 arch/x86/kernel/traps.c:267
> >  do_invalid_op+0x32/0x40 arch/x86/kernel/traps.c:286
> >  invalid_op+0x23/0x30 arch/x86/entry/entry_64.S:1027
> > RIP: 0010:dummy_free_request+0x6c/0x80 drivers/usb/gadget/udc/dummy_hcd.c:679
> > Code: c5 75 22 e8 26 06 96 fd 48 89 ef e8 3e 54 be fd 5b 5d e9 17 06 96 fd e8 12 06 96 fd 0f 0b 5b 5d e9 09 06 96 fd e8 04 06 96 fd <0f> 0b eb d5 48 89 ef e8 08 ae be fd eb c2 66 0f 1f 44 00 00 41 56
> > RSP: 0018:ffff8881c9eafdd8 EFLAGS: 00010016
> > RAX: 0000000000040000 RBX: ffff8881d0d5aa10 RCX: ffffc900012a0000
> > RDX: 000000000000012f RSI: ffffffff83a95c1c RDI: ffff8881d4d501c8
> > RBP: ffff8881d0d5aa00 R08: ffff8881d0f9b100 R09: ffffed103b666a84
> > R10: ffffed103b666a83 R11: ffff8881db33541b R12: 0000000000000212
> > R13: ffff8881ca832008 R14: 0000000000000000 R15: ffff8881ca832180
> >  raw_ioctl_ep_disable drivers/usb/gadget/legacy/raw_gadget.c:814 [inline]
> >  raw_ioctl+0x1281/0x19e0 drivers/usb/gadget/legacy/raw_gadget.c:1031
>
> Andrey:
>
> This could be a bug in your raw_gadget driver.  This particular WARN is
> triggered when a gadget driver tries to call usb_ep_free_request() for
> a request that is still in flight.

Hi Alan,

It might. It's not obvious what's wrong though. raw_ioctl_ep_disable()
calls usb_ep_disable()->dummy_disable() first, that supposedly does
list_del_init(&req->queue). And then it calls
usb_ep_free_request()->dummy_free_request(), which for some reason
still causes the warning. And a racy queueing of another request
should be prevented by the disabling flag.

I'll take a closer look on Monday.

Thanks!