On Thursday 28 July 2005 15:25, Timothy R. Chavez wrote:
To do this we'd need two hooks. One to allocate the watch info
for the
context before the creation of the inode and then one to fill it out upon
success or free it on failure.
Right sort of like a sql transaction. Create the data structures & commit or
rollback. But you could also handle rolling back any allocations should your
function fail.
Or, we can just use audit_panic :)
No. audit_panic is for when we have overflowed the backlog or hit a rate
limit. This is for when there is no other course of action possible. With the
watches, you could have returned an error. There was a course of action.
-Steve