nfs_getpages: error 4

[Dmitry Sivachenko]

> On 05 Mar 2016, at 23:10, Dmitry Sivachenko <trtrmitya at gmail.com> wrote:
> 
> 
>> On 05 Mar 2016, at 21:35, Konstantin Belousov <kostikbel at gmail.com> wrote:
>> 
>> But I suspect that you do have enough free or reclamaible pages for OOM
>> to not trigger, e.g. because you demonstrated commands output from the
>> live system after the situation occured.  It more likely was a temporal
>> free page shortage, after which the system recovered.
>> 
>> I more believe in a bug in the handling of killed process in vm_fault().
>> Could you get the p_flag value for the hung process ?  Like
>> 	ps -o flags <pid>
> 
> 
> Unfortunately I already rebooted this machine because our developers needed it and processes did not stop after kill -9.
> 
> When this repeats, I will try to keep this server up for longer time and provide any necessary information.


So far I got the same error:

Mar  8 07:13:08 skazka4 kernel: nfs_getpages: error 4
Mar  8 07:13:08 skazka4 kernel: vm_fault: pager read error, pid 58483 (decodcmd)

But the process in question finished successfully.