Re: [2.6.33-rc5] tty: possible irq lock inversion dependency in tty_fasync

[Eric W. Biederman]

AmÃrico Wang <xiyou.wangcong@xxxxxxxxx> writes:

> On Sat, Feb 06, 2010 at 10:31:30PM -0800, Linus Torvalds wrote:
>>
>>
>>On Sun, 7 Feb 2010, Tetsuo Handa wrote:
>>> 
>>> Below problem (which was introduced between 2.6.33-rc4 and 2.6.33-rc5) is
>>> not yet fixed as of 2.6.33-rc7.
>>> "git bisect start v2.6.33-rc5 v2.6.33-rc4" reported that
>>> 703625118069f9f8960d356676662d3db5a9d116 tty: fix race in tty_fasync
>>> is first bad commit.
>>
>>Yeah. I think we need to just revert that commit.
>>
>>Or maybe we could just do the following, rather than revert it outright: 
>>just get a ref to the 'struct pid' while holding the spinlock, and then 
>>releasing it after doing the __f_setown() call.
>
> We already fixed this, a better fix:
>
> http://lkml.org/lkml/2010/1/26/338
>
> I sent a same fix with Greg's.

That fix is present. See below.  Why does lockdep still warn?
Do we have lockdep bug?

Given that there is the only place we take f_owner.lock for write I
don't see how f_owner.lock can be unsafe to be called with irqs
disabled.

commit b04da8bfdfbbd79544cab2fadfdc12e87eb01600
Author: Greg Kroah-Hartman <gregkh@xxxxxxx>
Date:   Tue Jan 26 15:04:02 2010 -0800

    fnctl: f_modown should call write_lock_irqsave/restore
    
    Commit 703625118069f9f8960d356676662d3db5a9d116 exposed that f_modown()
    should call write_lock_irqsave instead of just write_lock_irq so that
    because a caller could have a spinlock held and it would not be good to
    renable interrupts.
    
    Cc: Eric W. Biederman <ebiederm@xxxxxxxxxxxx>
    Cc: Al Viro <viro@xxxxxxxxxxxxxxxxxx>
    Cc: Alan Cox <alan@xxxxxxxxxxxxxxxxxxx>
    Cc: Tavis Ormandy <taviso@xxxxxxxxxx>
    Cc: stable <stable@xxxxxxxxxx>
    Signed-off-by: Greg Kroah-Hartman <gregkh@xxxxxxx>
    Signed-off-by: Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx>

diff --git a/fs/fcntl.c b/fs/fcntl.c
index 97e01dc..5ef953e 100644
--- a/fs/fcntl.c
+++ b/fs/fcntl.c
@@ -199,7 +199,9 @@ static int setfl(int fd, struct file * filp, unsigned long arg)
 static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
                      int force)
 {
-       write_lock_irq(&filp->f_owner.lock);
+       unsigned long flags;
+
+       write_lock_irqsave(&filp->f_owner.lock, flags);
        if (force || !filp->f_owner.pid) {
                put_pid(filp->f_owner.pid);
                filp->f_owner.pid = get_pid(pid);
@@ -211,7 +213,7 @@ static void f_modown(struct file *filp, struct pid *pid, enum pid_type type,
                        filp->f_owner.euid = cred->euid;
                }
        }
-       write_unlock_irq(&filp->f_owner.lock);
+       write_unlock_irqrestore(&filp->f_owner.lock, flags);
 }
 
 int __f_setown(struct file *filp, struct pid *pid, enum pid_type type,


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/