Hi, On Tue, Jun 29, 2010 at 9:33 PM, Michael Blizek <michi1@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx> wrote: > Hi! > > On 11:00 Tue 29 Jun , Sudheer wrote: >> One basic doubt regarding block layer & write system call. When a >> process writes to a file in blocking mode, when the 'write' system >> call returns, >> >> A. will the process be returned only after actually writing the data >> on to the target block device? In this case we will get the benefit of >> elevator only if multiple processes are writing to the adjacent >> locations in the device. Is this true? > > Usually not, but you might get this behaviour by specifying the mount option > "sync" on some file systems. > >> >> >> OR >> >> B. will the process be returned after scheduling the writing (say >> after passing the data to the elevator)? > > Yes, this is what usually happens. > >> If this case is true, how >> would the process ensure that the data has been actually written on to >> the block device? > > By calling fsync() . Thanks for the comments. In this case, is there any difference in invoking 'write' with O_NONBLOCK flag set and without O_NONBLOCK?. >From the man page it looks like that the O_NONBLOCK is used to determine whether or not to block the process if it tries to access a 'locked' region. Means it doesn't have any effect in 'write' system call? -- Thanks Sudheer -- To unsubscribe from this list: send an email with "unsubscribe kernelnewbies" to ecartis@xxxxxxxxxxxx Please read the FAQ at http://kernelnewbies.org/FAQ
