On Thu, Sep 17, 2020 at 02:01:33PM +0200, Oleg Nesterov wrote:
> IIUC, file_end_write() was never IRQ safe (at least if !CONFIG_SMP), even
> before 8129ed2964 ("change sb_writers to use percpu_rw_semaphore"), but this
> doesn't matter...
>
> Perhaps we can change aio.c, io_uring.c and fs/overlayfs/file.c to avoid
> file_end_write() in IRQ context, but I am not sure it's worth the trouble.
If we change bio_endio to invoke the ->bi_end_io callbacks in softirq
context instead of hardirq context, we can change the pagecache to take
BH-safe locks instead of IRQ-safe locks. I believe the only reason the
lock needs to be IRQ-safe is for the benefit of paths like:
mpage_end_io
page_endio
end_page_writeback
test_clear_page_writeback
Admittedly, I haven't audited all the places that call end_page_writeback;
there might be others called from non-BIO contexts (network filesystems?).
That was the point where I gave up my investigation of why we use an
IRQ-safe spinlock when basically all page cache operations are done
from user context.
