On Mon, Sep 11, 2017 at 04:25:30PM +0000, Bart Van Assche wrote: > On Sat, 2017-09-09 at 18:39 +0800, Ming Lei wrote: > > - the issue is not suspend/resume only, it is about SCSI quiesce vs. > > RQF_PREEMPT. > > The only in-kernel user of the SCSI quiesce/resume functionality next to > runtime power management I know of is the SPI transport, namely > spi_dv_device(). Nobody has ever reported a hang in spi_dv_device() as far > as I know. Since the number of users of the SCSI parallel transport is very > small I propose to wait with fixing spi_dv_device() until someone shows that > it is possible to trigger that hang. If we have to address a hang in > spi_dv_device() I propose to address it in the same way as my patch series > addresses power management requests, namely by letting previously submitted > requests complete and by making blk_get_request() callers wait until the > quiesced SCSI device has been resumed. As I mentioned in V4 cover letter, Cathy saw this I/O hang issue, and she has verified that my revised patchset V3 can fix the issue. > > > I don't see the issue above addressed by this patchset, so maybe > > you are trying to fix another PM specific issue(not reported by > > Oleksandr), I am just confused. > > That's correct - I'm only trying to address the PM issues. I will mention > that explicitly when I repost this patch series. Unfortunately it isn't enough, and the approach you took is wrong too, I will comment it in your patch 4. -- Ming