On Wed, Oct 12, 2011 at 12:17 PM, V.Ravikumar <[email protected]>wrote:
> > On Wed, Oct 12, 2011 at 12:04 PM, rohan puri <[email protected]>wrote: > >> >> >> On Wed, Oct 12, 2011 at 10:50 AM, V.Ravikumar < >> [email protected]> wrote: >> >>> Other than fanotify , I can achieve my requirement through a >>> driver/module. If this can be achieved through a driver/module please >>> provide me inputs to start. >>> >>> Thanks, >>> Ravi >>> >>> On Wed, Sep 21, 2011 at 10:27 AM, rohan puri <[email protected]>wrote: >>> >>>> >>>> >>>> On Wed, Sep 21, 2011 at 10:03 AM, V.Ravikumar < >>>> [email protected]> wrote: >>>> >>>>> Hi all, >>>>> >>>>> Is it possible to write a module/driver which notifies file/directory >>>>> change asynchronously along with user name(or with uid) who modified it. >>>>> >>>>> inotify will do change notification but it will not provide uid who >>>>> modified/created the file. >>>>> >>>>> audit and inotify combination can work, but I'm looking for a better >>>>> option than this. >>>>> >>>>> Please help me. >>>>> >>>>> Thanks, >>>>> Ravi >>>>> >>>>> _______________________________________________ >>>>> Kernelnewbies mailing list >>>>> [email protected] >>>>> http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies >>>>> >>>>> You can have a look at fanotify. >>>> >>>> Refer http://lwn.net/Articles/339253/ >>>> >>>> Regards, >>>> Rohan Puri >>>> >>> >>> Hi Ravi, >> >> See, first of all if you want to notify file/dir change you need to the >> control after the invocation of that i_ops or f_ops. So there are two ways >> in which you can do : - >> >> 1. Easy & recommended : - >> >> Write a stackable file system module. The aim of this module will be to >> intercept vfs-calls on files/dirs & then call the underlying file systems >> specific operations. Now after completion of this operation, you would >> generate the change event here which will be used to notify. >> >> >> Note : - stackable file system research work was done by Professor *Erez >> Zadok. >> >> * >> refer ecryptfs for an example. >> > > This is more helpful for me. > >> >> 2. Complex & not recommended : - >> >> Instead of writing a stackable file system, you hook the mount sys call to >> get the control of the fs related structs. Then replace the original i_ops >> and f_ops pointer with your own defined ops. save the originals somewhere. >> Then when from user-space when some change operation is called, internally >> your ops will be called and now you call the original stored one and >> contruct the notification events >> >> > Even I've this in my mind. As it is not recommended,I had dropped this > choice > > 3. See if the module making use of LSM infrastructure will be able to do >> this or not. >> >> Regards, >> Rohan Puri >> > Hi Rohan > > Thank you for your valuable inputs. I will through the ecryptfs > > Thanks > Ravi > Hi Ravi, Good :). One more thing about that first approach, that stackable file system has to be mounted on the top-level directory, for which (sub-dirs and files) you need the notifications. This acts as the stacking trigger point. Was just keen to know, where are you requiring it? Regards, Rohan Puri
_______________________________________________ Kernelnewbies mailing list [email protected] http://lists.kernelnewbies.org/mailman/listinfo/kernelnewbies
