On Wed 25-07-18 17:53:07, zhaowuyun@wingtech.com wrote:

quoted

[Please do not top post - thank you]
[CC Hugh - the original patch was http://lkml.kernel.org/r/2018072514375722198958@wingtech.com]

On Wed 25-07-18 15:57:55, zhaowuyun@wingtech.com wrote:

quoted

That is a BUG we found in mm/vmscan.c at KERNEL VERSION 4.9.82

The code is quite similar in the current tree as well.

quoted

Sumary is TASK A (normal priority) doing __remove_mapping page preempted by TASK B (RT priority) doing __read_swap_cache_async,
the TASK A preempted before swapcache_free, left SWAP_HAS_CACHE flag in the swap cache,
the TASK B which doing __read_swap_cache_async, will not success at swapcache_prepare(entry) because the swap cache was exist, then it will loop forever because it is a RT thread...
the spin lock unlocked before swapcache_free, so disable preemption until swapcache_free executed ...

OK, I see your point now. I have missed the lock is dropped before
swapcache_free. How can preemption disabling prevent this race to happen
while the code is preempted by an IRQ?
--
Michal Hocko
SUSE Labs 

Hi Michal,

The action what processes __read_swap_cache_async is on the process context, so I think disable preemption is enough.

So what you are saying is that no IRQ or other non-process contexts will
not loop in __read_swap_cache_async so the live lock is not possible?
-- 
Michal Hocko
SUSE Labs