On Mon, Apr 09, 2018 at 09:03:54AM -0700, Matthew Wilcox wrote:
> > @@ -499,7 +499,7 @@ int sg_scsi_ioctl(struct request_queue *q, struct
> > gendisk *disk, fmode_t mode,
> > break;
> > }
> >
> > - if (bytes && blk_rq_map_kern(q, rq, buffer, bytes, __GFP_RECLAIM)) {
> > + if (bytes && blk_rq_map_kern(q, rq, buffer, bytes, GFP_NOIO)) {
>
> We don't seem to have grabbed any locks between the line which allocates
> memory using GFP_USER (line 446) and here, so I don't see why we should
> prohibit I/O?
No change of behavior for this patch. If you care deeply about using
GFP_KERNEL here send an incremental patch that I can add to the end of
the series.
