Re: [PATCH] mm: Make swapoff interruptible when unusing mms/shmem

[Chris Down]

Hi Baolin, Andrew,

Sorry, I missed this.

Yes, I think -ERESTARTSYS is appropriate here. I used -EINTR to match the 
existing signal exit in try_to_unuse(), but that also means a freezer induced 
interruption becomes visible to userspace after thaw. Let me use ERESTARTSYS 
instead.

Re: Andrew's earlier questions: before b56a2d8af914, try_to_unuse() checked for 
signals between swap entries, rather than after walking an entire mm.  
Individual I/O waits can still delay it but certainly it's a lot better than 
this situation.

So yes, I think this probably warrants a stable backport given the suspend 
timeout is reproducible and it's a reasonable usability regression even though 
it has been present for a long time. I'll add Cc: stable for v2.

Thanks!