Messages in this thread | | | Date | Wed, 10 Sep 2008 14:55:43 -0400 | From | Oren Laadan <> | Subject | Re: [RFC v4][PATCH 9/9] File descriprtors (restore) |
| |
Dave Hansen wrote: > On Tue, 2008-09-09 at 21:49 -0400, Oren Laadan wrote: >> Dave Hansen wrote: >>> On Tue, 2008-09-09 at 03:42 -0400, Oren Laadan wrote: >>>> +static int cr_close_all_fds(struct files_struct *files) >>>> +{ >>>> + int *fdtable; >>>> + int n; >>>> + >>>> + do { >>>> + n = cr_scan_fds(files, &fdtable); >>>> + if (n < 0) >>>> + return n; >>>> + while (n--) >>>> + sys_close(fdtable[n]); >>>> + kfree(fdtable); >>>> + } while (n != -1); >>>> + >>>> + return 0; >>>> +} >>> This needs to use an ERR_PTR(). It will save using the double-pointer. >> I suppose you refer to the call to cr_scan_fds(): either 'fdtable' >> or 'n' will have to pass-by-reference. Is it that you prefer it to be >> fdtable = cr_scan_fds(files, &n); >> ? > > I was misreading the use of 'n'. Can you really not use close_files() > for this operation? You'd need to add some locking around it, but I > think it does what you need here.
I thought about that. However, close_files() assumes that the files_struct will be discarded thereafter, so it does not reset ->fd_open->fd_bits[] bits, does not adjust ->next_fd field, and does not use rcu_assign_pointer(). And then, even if we adjust, we'll have to watch future differences between sys_close() and close_files(), so using sys_close() is more future-proof.
Besides, cr_scan_fds() is used by the checkpoint logic already, so it's easy to reuse for restart as well.
Oren.
| |