Re: [PATCH 0/2] pipe: Fixes [ver #2]

[David Sterba <dsterba@suse.cz>]

On Thu, Dec 05, 2019 at 10:30:22PM +0000, David Howells wrote:
> David Howells (2):
>       pipe: Remove assertion from pipe_poll()
>       pipe: Fix missing mask update after pipe_wait()

For reference, I've retested current master (b0d4beaa5a4b7d), that
incldes the 2 pipe fixes, the test still hangs.

The stack now points to pipe_wait but otherwise seems to be the same:

[<0>] pipe_wait+0x72/0xc0
[<0>] pipe_write+0x217/0x4b0
[<0>] new_sync_write+0x11e/0x1c0
[<0>] vfs_write+0xc1/0x1d0
[<0>] kernel_write+0x2c/0x40
[<0>] send_cmd+0x78/0xf0 [btrfs]
[<0>] send_extent_data+0x4af/0x52a [btrfs]
[<0>] process_extent+0xe5d/0xeb4 [btrfs]
[<0>] changed_cb+0xcf5/0xd2f [btrfs]
[<0>] send_subvol+0x8af/0xc0d [btrfs]
[<0>] btrfs_ioctl_send+0xe2b/0xef0 [btrfs]
[<0>] _btrfs_ioctl_send+0x80/0x110 [btrfs]
[<0>] btrfs_ioctl+0x14b8/0x3120 [btrfs]
[<0>] do_vfs_ioctl+0xa1/0x750
[<0>] ksys_ioctl+0x70/0x80
[<0>] __x64_sys_ioctl+0x16/0x20
[<0>] do_syscall_64+0x56/0x240
[<0>] entry_SYSCALL_64_after_hwframe+0x49/0xbe

(gdb) l *(pipe_wait+0x72)
0x622 is in pipe_wait (fs/pipe.c:120).
115              * is considered a noninteractive wait:
116              */
117             prepare_to_wait(&pipe->wait, &wait, TASK_INTERRUPTIBLE);
118             pipe_unlock(pipe);
119             schedule();
120             finish_wait(&pipe->wait, &wait);
121             pipe_lock(pipe);
122     }
123
124     static void anon_pipe_buf_release(struct pipe_inode_info *pipe,