Re: [PATCH] PM: hibernate: Get test_resume hibernation mode working again

[Pavan Kondeti <quic_pkondeti@quicinc.com>]

On Wed, Apr 05, 2023 at 12:43:03PM +0800, Chen Yu wrote:
> Hi Pavankumar,
> On 2023-04-05 at 09:15:02 +0530, Pavankumar Kondeti wrote:
> > Commit 39fbef4b0f77 ("PM: hibernate: Get block device exclusively in
> > swsusp_check()") changed the opening mode of the block device to
> > (FMODE_READ | FMODE_EXCL) during resume. This breaks the test_resume
> > hibernation mode as the block device is not available for exclusive
> > open. Because the block device is configured as swap during hibernation.
> > 
> > Fix this issue by opening the device in FMODE_READ only in test_resume
> > restore path. Cache the flags used in opening the block device so that
> > callers of swsusp_close() does not need to worry about test_resume vs
> > normal restore.
> > 
> > Fixes: 39fbef4b0f77 ("PM: hibernate: Get block device exclusively in swsusp_check()")
> > Signed-off-by: Pavankumar Kondeti <quic_pkondeti@quicinc.com>
> > ---
> >  kernel/power/hibernate.c | 21 +++++++++++-----
> >  kernel/power/power.h     |  4 +--
> >  kernel/power/swap.c      | 53 +++++++++++++++++++++++-----------------
> >  3 files changed, 48 insertions(+), 30 deletions(-)
> >
> Thanks for taking a look at this, would the following fix make sense?
> https://lore.kernel.org/lkml/20230401165540.322665-1-yu.c.chen@intel.com/
> 

Thanks Chenyu for pointing it out.

Yes, your patch works. It looks good also. I left a minor comment. You
can probably clarify what "safe" means there either by a comment or
changing the variable name. I also left an alternative approach, I initially
started with.

Thanks,
Pavan