Lustre devel - Dec 2006 - [Bug 11203] MDS failing to send precreate requests due to OSCC_FLAG_RECOVERING

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11203



Created an attachment (id=9182)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9182&action=view)
No resend patch

This wasn''t proven to be a fix for this bug, but it was an issue
identified
through code inspection.

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11203



Created an attachment (id=9183)
Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
 --> (https://bugzilla.lustre.org/attachment.cgi?id=9183&action=view)
Create State Diagram

Please don''t reply to lustre-devel. Instead, comment in Bugzilla by
using the following link:
https://bugzilla.lustre.org/show_bug.cgi?id=11203

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Attachment #9182|review?(green@clusterfs.com)|
               Flag|                            |


(From update of attachment 9182)
>Index: lustre/lustre/ptlrpc/recover.c
>==================================================================>---
lustre.orig/lustre/ptlrpc/recover.c
>+++ lustre/lustre/ptlrpc/recover.c
>@@ -245,7 +245,8 @@ int ptlrpc_resend(struct obd_import *imp
>                 LASSERTF((long)req > PAGE_SIZE && req !=
LP_POISON,
>                          "req %p bad\n", req);
>                 LASSERTF(req->rq_type != LI_POISON, "req %p
freed\n", req);
>-                ptlrpc_resend_req(req);
>+                if (!req->rq_no_resend)
>+                        ptlrpc_resend_req(req);
>         }
Will the no_resend request hang in imp_sending_list until timeout hits? It
there much sense in that as opposed to setting some sort of "timed
out" or
otherwise mark request as invalid to be dropped immediatelly?