Re: [PATCH] ieee802154: Add trace events for rdev->ops

[Alexander Aring <alex.aring@gmail.com>]

On Fri, Apr 17, 2015 at 12:03:55PM +0200, Alexander Aring wrote:
> Hi,
> 
> On Sun, Apr 12, 2015 at 02:07:33PM +0200, Guido Günther wrote:
> > This is based on the rdev->ops tracing code from net/wireless/.
> > 
> 
> this patch looks good. I tested it with trace-cmd:
> 
> 1. trace-cmd record -e cfg802154:802154_rdev_set_pan_id
> 
> 2. iwpan dev wpan0 set pan_id 0xbeaa
> 
> 3. trace-cmd report
> 
> result:
> 
> version = 6
> cpus=1
>            iwpan-1266  [000]   892.921639: 802154_rdev_set_pan_id: phy0, wpan_dev(1), pan id: 0xbeaa
> 
> also test it with channel and it works pretty well. (I am not a
> trace-cmd export at the moment, my previous use-case was enough to use
> the raw ftrace interface).
> 
> 
> I looked again for the byteorder handling and detected that trace-cmd has
> a cfg802154 plugin [0]. It looks like the plugin is there, to doing

s/cfg802154/cfg80211/

> byteorder translation in userspace instead kernelspace.
> 
> Looking into "__le16_to_cpup" stuff in wireless trace.h show something
> like [1].
> 
> Do you think we can do this in the same way? Then we need some
> plugin_cfg802154 in trace-cmd tool.
> 

Please read the other mail which I also mark an issue and we should move
the byteorder translation into the format string argument. I suppose, what
wireless do with the cfg80211 plugin is to printout the handling of function
__le16_to_cpup (in our case __le16_to_cpu) more correct. So we need
both, translation byteorder in kernelspace (for the __le16 parameter)
and (to show the __le16_to_cpu in userspace correct).

- Alex