Re: [RFC][PATCH] inotify 0.9.2

[John McCutchan <ttb@tentacle.dhs.org>]

On Tue, 2004-09-21 at 14:56, Robert Love wrote:
> On Tue, 2004-09-21 at 12:04 -0400, Robert Love wrote:
> 
> > Hey, John.
> > 
> > We are seeing an oops when monitoring a large number of directories.
> > The system keeps running, but I/O gets flaky and eventually processes
> > start getting stuck.
> > 
> > Also, the ioctl() stops returning new WD after 1024.  Thereafter, it
> > keeps returning the same value.
> > 
> > I have attached the relevant bits from the syslog.  I will debug it, but
> > I thought that perhaps you would immediately see the issue.
> 
> OK.  I fixed the problem with ioctl() failing after 1024 WD's.  This may
> also fix the oopses.  Still checking on that.
> 

I hope it fixes the oopses, I have only just started looking at the oops
you sent me, and nothing jumped out at me.

> The problem was that we were passing the size of dev->bitmask in _bytes_
> to find_first_zero_bit().  But find_first_zero_bit()'s second parameter
> is the size in _bits_.
>

Good to know, I wasn't sure when I wrote this code.

> I then went ahead and just made dev->bitmask an array, since we know the
> size at compile time.
> 
> Comments?

Sounds good.

John