From da63a6e1b2fcbb780271548a898f00acb5f50cf5 Mon Sep 17 00:00:00 2001
From: bobijam <bobijam>
Date: Wed, 7 May 2008 01:50:50 +0000
Subject: [PATCH] Branch b1_6 b=15575 i=wangdi, johann
Description: Stack overflow during MDS log replay
Details : ease stack pressure by using a thread dealing llog_process.
---
lustre/ChangeLog | 5 ++
lustre/obdclass/llog.c | 82 ++++++++++++++++++++++++++-------
lustre/obdclass/llog_internal.h | 11 +++++
3 files changed, 81 insertions(+), 17 deletions(-)
diff --git a/lustre/ChangeLog b/lustre/ChangeLog
index 0967934c99..cb5df0aeaa 100644
--- a/lustre/ChangeLog
+++ b/lustre/ChangeLog
@@ -25,6 +25,11 @@ tbd Sun Microsystems, Inc.
'tunefs.lustre --param="mdt.quota_type=ug1" $MDTDEV'.
For more information, please refer to bugzilla 13904.
+Severity : major
+Bugzilla : 15575
+Description: Stack overflow during MDS log replay
+Details : ease stack pressure by using a thread dealing llog_process.
+
Severity : normal
Bugzilla : 15278
Description: fix build on ppc32
diff --git a/lustre/obdclass/llog.c b/lustre/obdclass/llog.c
index 977b489367..ec44fde1fb 100644
--- a/lustre/obdclass/llog.c
+++ b/lustre/obdclass/llog.c
@@ -42,6 +42,7 @@
#include <obd_class.h>
#include <lustre_log.h>
#include <libcfs/list.h>
+#include "llog_internal.h"
/* Allocate a new log or catalog handle */
struct llog_handle *llog_alloc_handle(void)
@@ -204,22 +205,30 @@ int llog_close(struct llog_handle *loghandle)
}
EXPORT_SYMBOL(llog_close);
-int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
- void *data, void *catdata)
+static int llog_process_thread(void *arg)
{
- struct llog_log_hdr *llh = loghandle->lgh_hdr;
- struct llog_process_cat_data *cd = catdata;
- char *buf;
- __u64 cur_offset = LLOG_CHUNK_SIZE, last_offset;
- int rc = 0, index = 1, last_index;
- int saved_index = 0, last_called_index = 0;
- ENTRY;
+ struct llog_process_info *lpi = (struct llog_process_info *)arg;
+ struct llog_handle *loghandle = lpi->lpi_loghandle;
+ struct llog_log_hdr *llh = loghandle->lgh_hdr;
+ struct llog_process_cat_data *cd = lpi->lpi_catdata;
+ char *buf;
+ __u64 cur_offset = LLOG_CHUNK_SIZE;
+ __u64 last_offset;
+ int rc = 0, index = 1, last_index;
+ int saved_index = 0, last_called_index = 0;
LASSERT(llh);
OBD_ALLOC(buf, LLOG_CHUNK_SIZE);
- if (!buf)
- RETURN(-ENOMEM);
+ if (!buf) {
+ lpi->lpi_rc = -ENOMEM;
+#ifdef __KERNEL__
+ complete(&lpi->lpi_completion);
+#endif
+ return 0;
+ }
+
+ cfs_daemonize("llog_process");
if (cd != NULL) {
last_called_index = cd->first_idx;
@@ -267,7 +276,7 @@ int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
CDEBUG(D_OTHER, "after swabbing, type=%#x idx=%d\n",
rec->lrh_type, rec->lrh_index);
-
+
if (rec->lrh_index == 0)
GOTO(out, 0); /* no more records */
@@ -284,18 +293,18 @@ int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
continue;
}
- CDEBUG(D_OTHER,
+ CDEBUG(D_OTHER,
"lrh_index: %d lrh_len: %d (%d remains)\n",
rec->lrh_index, rec->lrh_len,
(int)(buf + LLOG_CHUNK_SIZE - (char *)rec));
- loghandle->lgh_cur_idx = rec->lrh_index;
+ loghandle->lgh_cur_idx = rec->lrh_index;
loghandle->lgh_cur_offset = (char *)rec - (char *)buf +
- last_offset;
+ last_offset;
/* if set, process the callback on this record */
if (ext2_test_bit(index, llh->llh_bitmap)) {
- rc = cb(loghandle, rec, data);
+ rc = lpi->lpi_cb(loghandle, rec, lpi->lpi_cbdata);
last_called_index = index;
if (rc == LLOG_PROC_BREAK) {
CDEBUG(D_HA, "recovery from log: "LPX64
@@ -304,7 +313,8 @@ int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
loghandle->lgh_id.lgl_ogen);
GOTO(out, rc);
} else if (rc == LLOG_DEL_RECORD) {
- llog_cancel_rec(loghandle, rec->lrh_index);
+ llog_cancel_rec(loghandle,
+ rec->lrh_index);
rc = 0;
}
if (rc)
@@ -325,6 +335,44 @@ int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
cd->last_idx = last_called_index;
if (buf)
OBD_FREE(buf, LLOG_CHUNK_SIZE);
+ lpi->lpi_rc = rc;
+#ifdef __KERNEL__
+ complete(&lpi->lpi_completion);
+#endif
+ return 0;
+}
+
+int llog_process(struct llog_handle *loghandle, llog_cb_t cb,
+ void *data, void *catdata)
+{
+ struct llog_process_info *lpi;
+ int rc;
+ ENTRY;
+
+ OBD_ALLOC_PTR(lpi);
+ if (lpi == NULL) {
+ CERROR("cannot alloc pointer\n");
+ RETURN(-ENOMEM);
+ }
+ lpi->lpi_loghandle = loghandle;
+ lpi->lpi_cb = cb;
+ lpi->lpi_cbdata = data;
+ lpi->lpi_catdata = catdata;
+
+#ifdef __KERNEL__
+ init_completion(&lpi->lpi_completion);
+ rc = cfs_kernel_thread(llog_process_thread, lpi, CLONE_VM | CLONE_FILES);
+ if (rc < 0) {
+ CERROR("cannot start thread: %d\n", rc);
+ OBD_FREE_PTR(lpi);
+ RETURN(rc);
+ }
+ wait_for_completion(&lpi->lpi_completion);
+#else
+ llog_process_thread(lpi);
+#endif
+ rc = lpi->lpi_rc;
+ OBD_FREE_PTR(lpi);
RETURN(rc);
}
EXPORT_SYMBOL(llog_process);
diff --git a/lustre/obdclass/llog_internal.h b/lustre/obdclass/llog_internal.h
index d4be19c282..82bb2e3e99 100644
--- a/lustre/obdclass/llog_internal.h
+++ b/lustre/obdclass/llog_internal.h
@@ -1,6 +1,17 @@
#ifndef __LLOG_INTERNAL_H__
#define __LLOG_INTERNAL_H__
+#include <lustre_log.h>
+
+struct llog_process_info {
+ struct llog_handle *lpi_loghandle;
+ llog_cb_t lpi_cb;
+ void *lpi_cbdata;
+ void *lpi_catdata;
+ int lpi_rc;
+ struct completion lpi_completion;
+};
+
int llog_put_cat_list(struct obd_device *obd, struct obd_device *disk_obd,
char *name, int count, struct llog_catid *idarray);
int llog_cat_id2handle(struct llog_handle *cathandle, struct llog_handle **res,
--
GitLab