Re: [PATCH] nfsd: return 0 on reads of fault injection files

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

 



On May 16, 2012, at 5:58 PM, J. Bruce Fields wrote:

> On Wed, May 16, 2012 at 09:50:32PM +0000, Adamson, Dros wrote:
>> No, I haven't messed with it that much - I had other tasks take a
>> higher priority, but I'm back on it as of this afternoon.
>> 
>> I'll try nfsd-next to get CB_PATH_DOWN without fault_injection and
>> report back to you.  Also, I think we need to modify nfsd
>> (nfsd4_new_conn()) to set cl_cb_state to NFSD4_CB_UNKNOWN on a
>> successful BIND_CONN_TO_SESSION (IFF it's CB_DOWN with the right
>> direction), otherwise the CB_PATH_DOWN flag will be set on every
>> sequence OP
> 
> That's intentional:
> 
> 	http://www.ietf.org/mail-archive/web/nfsv4/current/msg10840.html

Right...

> 
>> and the client will keep sending BIND_CONN_TO_SESSION.
>> The idea is that once we call nfsd4_new_conn, we won't know if the
>> back channel is really up until a callback is attempted.
> 
> If the client hasn't given us a connection to use as the backchannel,
> then we *know* the backchannel is down, we don't need to try sending a
> callback (how could we?).

The point is that the client *has* given us a connection to use as the backchannel with the BIND_CONN_TO_SESSION call.
On the server, nfs4_new_conn() was called and was successful, but cl_cb_state is still set to CB_DOWN and will never change.

> 
>> Setting it to CB_UNKNOWN stops the loop of (sequence, bind_conn,
>> sequence, bind_conn, …), but doesn't actually mark it as "UP" until a
>> callback is successful.
> 
> If the client is attempting to clear the CB_PATH_DOWN flag by sending us
> a BIND_CONN_TO_SESSION that doesn't actually give us a backchannel, then
> it's confused.

The client *is* giving us a backchannel by sending the server a BIND_CONNN_TO_SESSION! The server accepts the call, adds the "new connection" via nfs4_new_conn(), but the problem is the internal state of nfsd never changes from CB_DOWN.

-dros

> If the client really doesn't care whether the backchannel is down or not
> then it can ignore the flag--it's a flag, not an error.
> 
> --b.

Attachment: smime.p7s
Description: S/MIME cryptographic signature


[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