Messages in this thread | | | Date | Mon, 14 Sep 2009 23:50:50 +0200 | From | Jiri Slaby <> | Subject | Re: V4L2 drivers: potentially dangerous and inefficient msecs_to_jiffies() calculation |
| |
On 09/14/2009 11:39 PM, Andreas Mohr wrote: > On Mon, Sep 14, 2009 at 11:34:40PM +0200, Jiri Slaby wrote: >> On 09/14/2009 11:07 PM, Andreas Mohr wrote: >>> msecs_to_jiffies(1) is quite a bit too boldly assuming >>> that all of the msecs_to_jiffies(x) implementation branches >>> always round up. >> >> They do, don't they? > > I'd hope so, but a slight risk remains, you never know, > especially with 4+ or so variants...
A potential problem here is rather that it may wait longer due to returning 1 jiffie. It's then timeout * 1000 * 1. On 250HZ system it makes a difference of multiple of 4. Don't think it's a real issue in those drivers at all, but it's worth fixing. Care to post a patch?
| |