Hi! On 00:01 Sat 11 Oct , Peter Teoh wrote: > thank you Michael, > > As far as I can tell, it is related to I/O scheduling. Still don't > quite understand, why CFQ and anticipatory will need this IO context, > and the other two it does not matter? > > On 10:50 Thu 09 Oct , Peter Teoh wrote: > >> What is CLONE_IO used for? What is the definition / meaning of I/O > >> context sharing between parent and child? > >> > >> In fact, from header file: > >> > >> #define CLONE_FS 0x00000200 /* set if fs info shared > >> between processes */ > >> #define CLONE_FILES 0x00000400 /* set if open files shared > >> between processes */ > >> #define CLONE_IO 0x80000000 /* Clone io context */ > >> > >> These three looked seemingly the same/related.....differences among them? CLONE_FS: thread 1 does chroot, chdir or umask ==> does it affect the other thread? CLONE_FILES: thread 1 calls open or close ==> does the other thread see the new file descriptor and does it see the old file descriptor closed? CLONE_IO: CFQ or some other io scheduler which lets every process have their fair share of IO time is in use. Thread 1 and a new thread created by clone() do a lot of io. If CLONE_IO is set, their fair share io bandwidth will equals the bandwidth a single process would get. If it is *not* set, both threads will have their own fair share, meaning the fair share will grow if the number of threads increases. -Michi -- programing a layer 3+4 network protocol for mesh networks see http://michaelblizek.twilightparadox.com -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
