On Thursday, June 30, 2011 07:18:54 PM you wrote: > On 06/30/2011 11:39 AM, Chun Yan Liu wrote: > > On Thursday, June 30, 2011 03:58:57 PM Alexander Graf wrote: > >> On 30.06.2011, at 09:08, Chunyan Liu wrote: > >>> Add code to support logging xen-domU console, as what xenconsoled does. > >>> To enable logging, set environment variable XENCONSOLED_TRACE=guest and > >>> XENCONSOLED_LOGDIR=<specified directory>, log file will be saved in > >>> <specified directory>. > >> > >> In fact, this whole thing looks as if you're merely trying to > >> reimplement "tee" on the xenconsole output. Wouldn't it make more sense > >> to do this in > >> > >> the char layer? So we could do: > >> -xenconsole tee:stdio,file:/tmp/xen.log > >> > >> or similar? That's probably a lot more useful than a random Xen specific > >> hack. > > > > Thanks, Alex. It IS something like "tee". But IMO, change in > > xen_console.c and change in char layer are just different time points, > > do not have essential difference. Change in xen_console.c is trying to > > backup output data into log file before sending to char device, change > > in char device is trying to dupicate data from char device to log file. > > Correct me if I'm wrong. > > Sure, the outcome is the same though, no? We get the output data in both > a file and the char backend. > Char device in qemu is not only used by console. Compared with the benefits that brings, I still doubt if it is proper to adding this functionality to char layer. Stefano, how do you think?
Thanks, Chunyan