Re: Is "unmount -f" worked as expected?

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hi:  Bruce:

> On 5/31/18, 10:34 AM, "bfields@xxxxxxxxxxxx" <bfields@xxxxxxxxxxxx> wrote:

>  On Thu, May 31, 2018 at 04:51:49PM +0000, Shawn Lu (shawlu) wrote:
>   > >On 5/31/18, 8:18 AM, "bfields@xxxxxxxxxxxx" <bfields@xxxxxxxxxxxx> wrote:
>   >  >  > One case is involved with  cross mounted nfs from different host.
>   >    
>    >  >  Note: in theory I think there are some deadlocks possible if client and
>   >  >   server mount each other.  (Each host could be waiting on the other one
>   >  >   to process writes before it can free memory that it needs to make
>   >  >   progress.)
>  >
>   > Yes.  I hope aggressive umount  will  help by setting a very short grace period and timeout before umount.
    
>   No, the deadlock isn't an unmounting problem, it's a problem that could
>  happen during normal operation.  I don't know if it's easy to reproduce.
>   You'd probably need one or both hosts to be writing to NFS very quickly.
  
Yes, This will a big issue for cross mount.  Luckily, we have not experience it yet due to low traffic.  

By the way, what is your suggestion on the road to enhance “umount –f” ?   Will add kernel config option to support guaranteed force unmount helps?
  >  --b.

Thanks,
Shawn
On 5/31/18, 10:34 AM, "bfields@xxxxxxxxxxxx" <bfields@xxxxxxxxxxxx> wrote:

    On Thu, May 31, 2018 at 04:51:49PM +0000, Shawn Lu (shawlu) wrote:
    > >On 5/31/18, 8:18 AM, "bfields@xxxxxxxxxxxx" <bfields@xxxxxxxxxxxx> wrote:
    >  >  > One case is involved with  cross mounted nfs from different host.
    >     
    >  >  Note: in theory I think there are some deadlocks possible if client and
    >  >   server mount each other.  (Each host could be waiting on the other one
    >  >   to process writes before it can free memory that it needs to make
    >  >   progress.)
    > 
    > Yes.  I hope aggressive umount  will  help by setting a very short grace period and timeout before umount.
    
    No, the deadlock isn't an unmounting problem, it's a problem that could
    happen during normal operation.  I don't know if it's easy to reproduce.
    You'd probably need one or both hosts to be writing to NFS very quickly.
    
    --b.
    

��.n��������+%������w��{.n�����{��w���jg��������ݢj����G�������j:+v���w�m������w�������h�����٥




[Index of Archives]     [Linux Filesystem Development]     [Linux USB Development]     [Linux Media Development]     [Video for Linux]     [Linux NILFS]     [Linux Audio Users]     [Yosemite Info]     [Linux SCSI]

  Powered by Linux