Messages in this thread | | | Date | Tue, 20 Jan 2009 12:07:35 -0800 | From | Mike Waychison <> | Subject | Re: [PATCH v1 3/8] Deferred batching of iput() |
| |
Evgeniy Polyakov wrote: > On Fri, Jan 16, 2009 at 06:29:52PM -0800, Mike Waychison (mikew@google.com) wrote: >> +static void postpone_iput(struct inode *inode) >> +{ >> + struct postponed_inodes *ppi, *new_ppi; >> + >> +again: >> + ppi = get_cpu_var(postponed_inodes); >> + if (!pending_iput_full(ppi)) { >> + add_pending_iput(ppi, inode); >> + put_cpu_var(postponed_inodes); >> + return; >> + } >> + >> + /* Need to flush out existing pending inodes */ >> + put_cpu_var(postponed_inodes); >> + /* Allocate more space.. */ >> + new_ppi = new_postponed_inodes(); >> + if (!new_ppi) { >> + struct postponed_inodes_onstack ppi_onstack; >> + >> + ppi = init_ppi_onstack(&ppi_onstack); >> + add_pending_iput(ppi, inode); >> + process_postponed_inodes(ppi); >> + return; >> + } > > The same here: what just not to call a real_iput() without on-stack > allocation and the line? >
Agreed. Will fix in the next version.
| |