console polling loop and inotify

Marco Pesenti Gritti mpg at redhat.com
Thu Mar 29 09:50:12 EDT 2007


On Thu, 2007-03-29 at 09:43 -0400, William Cohen wrote:
> I was running a simple script iotime.stp on the 314 build. The script monitors 
> the reads and write performed by the systemt. When the process does the close it 
> prints a summary: a time stamp in microseconds, pid, process name, type of 
> information (number of accesses or amount of time), the file name. For accesses 
> this is a "count" of the number of io operations follow by number of bytes read 
> and writen. The time ones record the accumulated number of microseconds of wall 
> clock time spent in read and write systemcalls for file by process.
> 
> When the machine was idle I saw the following entries come up again and again 
> when the alt-0 console was open:
> 
> 
> 70943178 1916 (python) access /home/olpc/.sugar/default/logs count: 0 read: 0 
> write: 0
> 70943410 1916 (python) access /home/olpc/.sugar/default/logs/shell.log count: 2 
> read: 8474 write: 0
> 70943410 1916 (python) iotime /home/olpc/.sugar/default/logs/shell.log time: 24
> 70943681 1916 (python) access /home/olpc/.sugar/default/logs/presenceservice.log 
> count: 2 read: 9052 write: 0
> 70943681 1916 (python) iotime /home/olpc/.sugar/default/logs/presenceservice.log 
> time: 25
> 70943952 1916 (python) access /home/olpc/.sugar/default/logs/clipboard.log 
> count: 2 read: 8227 write: 0
> 70943952 1916 (python) iotime /home/olpc/.sugar/default/logs/clipboard.log time: 24
> 
> 
> Why poll? Doesn't python have access to inotify? It seems like something like 
> that would be useful in a number of places to avoid having processes wake up, 
> check that nothing has changed, and then go back to sleep.

Yeah that something which needs to be fixed. A trac would be useful.

(Eduardo if you have time fixing this would be really good)

Marco




More information about the Devel mailing list