Troy, the patch Pete mentions should fix that assertion.
-sam

On Feb 12, 2008, at 3:59 PM, Pete Wyckoff wrote:

[EMAIL PROTECTED] wrote on Tue, 12 Feb 2008 15:44 -0600:
I am running current CVS (last change being this:
user:        slang
date:        Mon Feb 11 13:55:08 2008 -0600
summary:     fixes for building request unify commit.
)

on 2 64 bit intel servers, with 1 ehca power5 client using native mode
NetPipe-pvfs, and I am getting the following on one of the servers:

[D 02/12 15:25] [BMI CONTROL]: BMI_set_info: set_info: 6680848 option: 6 [D 02/12 15:25] encourage_recv_incoming: recv eager unexpected len 208. [D 02/12 15:25] [BMI CONTROL]: BMI_set_info: searching for ref 6680848 [D 02/12 15:25] [BMI CONTROL]: BMI_set_info: decremented ref 6680848 to: 0
[D 02/12 15:25] encourage_recv_incoming: new rq 0x2aaaaac32740 now
RQ_EAGER_WAITING_USER_TESTUNEXPECTED.
[D 02/12 15:25] BMI_ib_testunexpected: found waiting testunexpected.
[D 02/12 15:25] [BMI CONTROL]: BMI_set_info: set_info: 6680848 option: 6 [D 02/12 15:25] [BMI CONTROL]: BMI_set_info: searching for ref 6680848 [D 02/12 15:25] [BMI CONTROL]: BMI_set_info: decremented ref 6680848 to: -1
pvfs2-server: ../src/io/bmi/bmi.c:1290: BMI_set_info: Assertion
`tmp_ref->ref_count >= 0' failed.

The full log is at:

http://www.scl.ameslab.gov/~troy/pvfs/pvfs2-server-da13.log <-- failing
server
http://www.scl.ameslab.gov/~troy/pvfs/pvfs2-server-da12.log <-- other
server

There's one more critical-looking patch that slang checked in
recently.  Without looking too much further into it, I hope this
fixes it for you.

                -- Pete


commit ccbda10a129933b12ec31c00350e95ae36eb2d70
Author: slang <slang>
Date:   Mon Feb 11 23:27:38 2008 +0000

   thought I removed this already...

diff --git a/src/server/io.sm b/src/server/io.sm
index 0c16391..c82d19f 100644
--- a/src/server/io.sm
+++ b/src/server/io.sm
@@ -306,13 +306,6 @@ static PINT_sm_action io_cleanup(
        PINT_encode_release(&s_op->encoded, PINT_ENCODE_RESP);
    }

-    /* NOTE: this would happen in final_response nested state machine
-     * for most operations, but io.sm is the one state machine which
-     * does not use final_response -- decrement reference count for
-     * this bmi address
-     */
-    BMI_set_info(s_op->addr, BMI_DEC_ADDR_REF, NULL);
-
    return(server_state_machine_complete(smcb));
}

_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers


_______________________________________________
Pvfs2-developers mailing list
[email protected]
http://www.beowulf-underground.org/mailman/listinfo/pvfs2-developers

Reply via email to