On 2018/05/09 21:00, Petr Mladek wrote:
>>>> But we first need a real reason. Right now it looks to me like
>>>> we have "a solution" to a problem which we have never witnessed.
>>>
>>> I am trying to find a "simple" and generic solution for the problem
>>> reported by Tejun:
>> [..]
>>> 1. Console is IPMI emulated serial console. Super slow. Also
>>> netconsole is in use.
>>> 2. System runs out of memory, OOM triggers.
>>> 3. OOM handler is printing out OOM debug info.
>>> 4. While trying to emit the messages for netconsole, the network stack
>>> / driver tries to allocate memory and then fail, which in turn
>>> triggers allocation failure or other warning messages. printk was
>>> already flushing, so the messages are queued on the ring.
>>> 5. OOM handler keeps flushing but 4 repeats and the queue is never
>>> shrinking. Because OOM handler is trapped in printk flushing, it
>>> never manages to free memory and no one else can enter OOM path
>>> either, so the system is trapped in this state.
>>> </paste>
>
> IMHO, we do not need to chase down this particular problem. It was
> already "solved" by the commit 400e22499dd92613821 ("mm: don't warn
> about allocations which stall for too long").
Does memory allocation by network stack / driver while trying to emit
the messages include __GFP_DIRECT_RECLAIM flag (e.g. GFP_KERNEL) ?
Commit 400e22499dd92613821 handles only memory allocations with
__GFP_DIRECT_RECLAIM flag. If memory allocation when trying to emit
the messages does not include __GFP_DIRECT_RECLAIM flag (e.g.
GFP_ATOMIC / GFP_NOWAIT), doesn't this particular problem still exist?
