vp-build/srcpkgs/dmraid/patches/15_isw_incorrect_status_fix...

323 lines
11 KiB
Diff

Patch include:
fix for "Incorrect status displayed when disk is missing" issue.
fix for "Missing supported RAID configuration in "-l" option" - Bugzilla bug
479395
fix for "Wrong disk layout in raid01" - no Bugzilla bug.
--- 1.0.0.rc16-3/dmraid/lib/format/ataraid/isw.c
+++ 1.0.0.rc16-3/dmraid/lib/format/ataraid/isw.c
@@ -4,8 +4,11 @@
* Copyright (C) 2004-2008 Heinz Mauelshagen, Red Hat GmbH.
* All rights reserved.
*
- * Copyright (C) 2007,2008 Intel Corporation. All rights reserved.
+ * Copyright (C) 2007,2009 Intel Corporation. All rights reserved.
* November, 2007 - additions for Create, Delete, Rebuild & Raid 10.
+ * March, 2008 - additions for hot-spare check
+ * August, 2008 - support for Activation, Rebuild checks
+ * January, 2009 - additions for Activation, Rebuild check
*
* See file LICENSE at the top of this source tree for license information.
*/
@@ -162,7 +165,7 @@ name(struct lib_context *lc, struct raid
while (i--) {
if (disk == isw->disk + i) {
- id = i / 2;
+ id = i % 2;
goto ok;
}
}
@@ -479,7 +482,7 @@ isw_read_extended(struct lib_context *lc
/* Read extended metadata to offset ISW_DISK_BLOCK_SIZE */
if (blocks > 1 &&
!read_file(lc, handler, di->path,
- (void *) isw_tmp + ISW_DISK_BLOCK_SIZE,
+ (void *) (((uint8_t*)isw_tmp) + ISW_DISK_BLOCK_SIZE),
*size - ISW_DISK_BLOCK_SIZE, *isw_sboffset)) {
dbg_free(isw_tmp);
isw_tmp = NULL;
@@ -708,6 +711,7 @@ super_created(struct raid_set *super, vo
super->stride = ((struct isw_dev *) private)->vol.map.blocks_per_strip;
}
+
/*
* rs_group contains the top-level group RAID set (type: t_group) on entry
* and shall be returned on success (or NULL on error).
@@ -1112,6 +1116,15 @@ isw_metadata_handler(struct lib_context
break; /* case GET_REBUILD_DRIVE_NO */
+ case ALLOW_ACTIVATE: /* same as ALLOW_REBUILD */
+ case ALLOW_REBUILD:
+ /* Do not allow activate or rebuild, if the log is non-empty */
+ isw = META (rd, isw);
+ ret = !isw->bbm_log_size; /* Is log empty */
+ if (!ret)
+ log_err(lc, "BBM entries detected!");
+ break; /* case ALLOW_REBUILD */
+
default:
LOG_ERR(lc, 0, "%u not yet supported", command);
@@ -1144,16 +1157,25 @@ check_rd(struct lib_context *lc, struct
struct raid_dev *rd, void *context)
{
struct isw_dev *dev = rd->private.ptr;
-
- /* FIXME: more status checks ? */
- if (dev->status) {
+ if (dev == NULL)
+ if (rd->type == t_spare) return 1;
+ else {
+ LOG_ERR(lc, 0, "No information about %s device on %s "
+ "in RAID set \"%s\"",
+ handler, rd->di->path, rs->name);
+ return 0;
+ }
+ /* if disk is ready to read and write return 1 */
+ if ((dev->status & ISW_DEV_READ_COALESCING)
+ && (dev->status & ISW_DEV_WRITE_COALESCING)) {
+ return 1;
+ }
+ else {
LOG_ERR(lc, 0, "%s device for volume \"%s\" broken on %s "
"in RAID set \"%s\"",
handler, dev->volume, rd->di->path, rs->name);
-
+ return 0;
}
-
- return 1;
}
static int
@@ -1622,6 +1644,7 @@ _isw_create_first_volume(struct lib_cont
isw->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
isw->num_raid_devs = (rs->type == ISW_T_SPARE) ? 0 : 1;
isw->family_num = isw->orig_family_num = _checksum(isw) + time(NULL);
+ isw->bbm_log_size = 0;
isw->check_sum = 0;
isw->check_sum = _checksum(isw);
return isw;
@@ -1719,6 +1742,7 @@ _isw_create_second_volume(struct lib_con
isw->generation_num++;
isw->attributes = MPB_ATTRIB_CHECKSUM_VERIFY;
isw->num_raid_devs++;
+ isw->bbm_log_size = 0;
isw->check_sum = 0;
isw->check_sum = _checksum(isw);
return isw;
@@ -1941,6 +1965,8 @@ _isw_log(struct lib_context *lc, struct
DP("error_log_pos: %u", isw, isw->error_log_pos);
DP("cache_size: %u", isw, isw->cache_size);
DP("orig_family_num: %u", isw, isw->orig_family_num);
+ DP ("power_cycle_count: %u", isw, isw->power_cycle_count);
+ DP ("bbm_log_size: %u", isw, isw->bbm_log_size);
for (i = 0; i < ISW_FILLERS; i++) {
if (isw->filler[i])
@@ -2296,7 +2322,7 @@ isw_delete(struct lib_context *lc, struc
static struct dmraid_format isw_format = {
.name = HANDLER,
.descr = "Intel Software RAID",
- .caps = "0,1,01",
+ .caps = "0,1,5,01",
.format = FMT_RAID,
.read = isw_read,
.write = isw_write,
--- 1.0.0.rc16-3/dmraid/lib/format/format.c
+++ 1.0.0.rc16-3/dmraid/lib/format/format.c
@@ -260,6 +260,7 @@ write_metadata(struct lib_context *lc, c
return 1;
}
+
/*
* Check devices in a RAID set.
*
@@ -277,7 +278,7 @@ _check_raid_set(struct lib_context *lc,
void *context),
void *f_check_context, const char *handler)
{
- unsigned int devs;
+ unsigned int dev_tot_qan;
uint64_t sectors;
struct raid_dev *rd;
@@ -285,36 +286,73 @@ _check_raid_set(struct lib_context *lc,
return;
sectors = total_sectors(lc, rs);
- rs->total_devs = devs = count_devs(lc, rs, ct_dev);
+ rs->total_devs = dev_tot_qan = count_devs(lc, rs, ct_dev);
list_for_each_entry(rd, &rs->devs, devs) {
- unsigned int devices = f_devices(rd, f_devices_context);
+ unsigned int dev_found_qan = f_devices(rd, f_devices_context);
/* FIXME: error if the metadatas aren't all the same? */
- rs->found_devs = devices;
+ rs->found_devs = dev_found_qan;
log_dbg(lc, "checking %s device \"%s\"", handler, rd->di->path);
- if (T_SPARE(rd) && rs->type == t_raid1 && /* FIXME: rs->type check ? */
- rd->sectors != sectors) {
- rd->status = s_inconsistent;
- log_err(lc, "%s: size mismatch in set \"%s\", spare "
- "\"%s\"", handler, rs->name, rd->di->path);
- continue;
- }
-
- if (devs != devices &&
- f_check && !f_check(lc, rs, rd, f_check_context)) {
- rd->status = s_broken;
- log_err(lc, "%s: wrong # of devices in RAID "
- "set \"%s\" [%u/%u] on %s",
- handler, rs->name, devs, devices, rd->di->path);
- } else
- rd->status = s_ok;
+ /* if disks number of found disks equals disk expected status status OK */
+ if ((dev_tot_qan == dev_found_qan) && f_check
+ && f_check(lc, rs, rd, f_check_context)) {
+ rd->status = s_ok;
+ }
+ else {
+ if(dev_tot_qan != dev_found_qan) {
+ log_err(lc, "%s: wrong # of devices in RAID "
+ "set \"%s\" [%u/%u] on %s",
+ handler, rs->name, dev_tot_qan, dev_found_qan, rd->di->path);
+ }
+ /* if number of disks is incorrect, status depends on raid type: */
+ switch(rs->type) {
+ case t_linear:
+ case t_raid0: /* raid 0 - always broken */
+ rd->status = s_broken;
+ break;
+ case t_raid1: /* raid 1 - if at least 1 disk- inconsintent */
+ if(dev_tot_qan >= 1)
+ rd->status = s_inconsistent;
+ else if (T_SPARE(rd) && rd->sectors != sectors) {
+ rd->status = s_inconsistent;
+ log_err(lc, "%s: size mismatch in set \"%s\", spare "
+ "\"%s\"", handler, rs->name, rd->di->path);
+ }
+ else rd->status = s_broken;
+ break;
+ case t_raid4:
+ case t_raid5_ls:
+ case t_raid5_rs:
+ case t_raid5_la:
+ case t_raid5_ra: /* raid 4/5 - if 1 disk missing- inconsistent */
+ if((dev_tot_qan == (dev_found_qan-1) && dev_tot_qan > 1
+ && f_check && f_check(lc, rs, rd, f_check_context))
+ || (dev_tot_qan > dev_found_qan))
+ rd->status = s_inconsistent;
+ else rd->status = s_broken;
+ break;
+ case t_raid6: /* raid 6 - if up to 2 disks missing- inconsistent */
+ if((dev_tot_qan >= (dev_found_qan-2)
+ && f_check && f_check(lc, rs, rd, f_check_context))
+ || (dev_tot_qan > dev_found_qan))
+ rd->status = s_inconsistent;
+ else rd->status = s_broken;
+ break;
+
+ case t_spare: /* spare - always broken */
+ rd->status = s_broken;
+ break;
+ default: /* other - undef */
+ rd->status = s_undef;
+ }
+ }
}
}
/*
* Update RAID set state based on operational subsets/devices.
*
- * In case of a RAID set hierachy, check availability of subsets
+ f a RAID set hierachy, check availability of subsets
* and set superset to broken in case *all* subsets are broken.
* If at least one is still available, set to inconsistent.
*
@@ -323,12 +361,19 @@ _check_raid_set(struct lib_context *lc,
*/
static void
_set_rs_status(struct lib_context *lc, struct raid_set *rs,
- unsigned int i, unsigned int operational)
+ unsigned int i, unsigned int operational,
+ unsigned int inconsist, unsigned int subsets_raid01)
{
- if (operational == i)
+ if (subsets_raid01 == 1) {
+ rs->status = s_broken;
+ return;
+ }
+ if (operational == i)
rs->status = s_ok;
else if (operational)
rs->status = s_inconsistent;
+ else if (inconsist)
+ rs->status = s_inconsistent;
else
rs->status = s_broken;
@@ -338,26 +383,30 @@ _set_rs_status(struct lib_context *lc, s
static int
set_rs_status(struct lib_context *lc, struct raid_set *rs)
{
- unsigned int i = 0, operational = 0;
+ unsigned int i = 0, operational = 0, subsets_raid01 = 0, inconsist = 0;
struct raid_set *r;
struct raid_dev *rd;
/* Set status of subsets. */
list_for_each_entry(r, &rs->sets, list) {
/* Check subsets to set status of superset. */
- i++;
+ if((rs->type == t_raid0) && (r->type == t_raid1))
+ subsets_raid01++; /* Count subsets for raid 01 */
+ i++; /* Count subsets*/
if (S_OK(r->status) || S_INCONSISTENT(r->status))
- operational++;
+ operational++; /* Count operational subsets*/
}
/* Check status of devices... */
list_for_each_entry(rd, &rs->devs, devs) {
- i++;
+ i++; /* Count disks*/
if (S_OK(rd->status))
- operational++;
+ operational++; /* Count disks in "ok" raid*/
+ if (S_INCONSISTENT(rd->status))
+ inconsist++; /* Count disks in degraded raid*/
}
- _set_rs_status(lc, rs, i, operational);
+ _set_rs_status(lc, rs, i, operational, inconsist, subsets_raid01);
return S_BROKEN(rs->status) ? 0 : 1;
}
--- 1.0.0.rc16-3/dmraid/include/dmraid/format.h
+++ 1.0.0.rc16-3/dmraid/include/dmraid/format.h
@@ -75,6 +75,8 @@ enum handler_commands {
GET_REBUILD_DRIVE,
GET_REBUILD_DRIVE_NO,
CHECK_HOT_SPARE,
+ ALLOW_ACTIVATE,
+ ALLOW_REBUILD,
/* ... */
};
--- 1.0.0.rc16-3/dmraid/lib/format/ataraid/isw.h
+++ 1.0.0.rc16-3/dmraid/lib/format/ataraid/isw.h
@@ -188,6 +188,8 @@ struct isw_dev {
struct isw {
int8_t sig[MAX_SIGNATURE_LENGTH]; /* 0x0 - 0x1F */
+ uint32_t bbm_log_size;
+ uint32_t power_cycle_count;
uint32_t check_sum; /* 0x20 - 0x23 MPB Checksum */
uint32_t mpb_size; /* 0x24 - 0x27 Size of MPB */
uint32_t family_num; /* 0x28 - 0x2B Checksum from first time this config was written */