Skip to contentSkip to conversation

cciss: Fix problem that deleting multiple logical drives could cause a panic

PATCH In review [email protected] 2 participants 4 messages Dec 16, 2008 – Dec 18, 2008 Block layer

lore
Share & tools
lore
#0 lore raw

PATCH 1/1 cciss: Fix problem that deleting multiple logical drives could cause a panic

Fix problem that deleting multiple logical drives could cause a panic.

---

 drivers/block/cciss.c |    5 +++++
 1 file changed, 5 insertions(+)
drivers/block/cciss.c renamed from drivers/block/cciss.c~fix_rebuild_lun_table_bug +5 -0
diff -puN drivers/block/cciss.c~fix_rebuild_lun_table_bug drivers/block/cciss.c
@@ -1693,6 +1693,11 @@ static int rebuild_lun_table(ctlr_info_t
 	for (i = 0; i <= h->highest_lun; i++) {
 		int j;
 		drv_found = 0;
+
+		/* skip holes in the array from already deleted drives */
+		if (h->drv[i].raid_level == -1)
+			continue;
+
 		for (j = 0; j < num_luns; j++) {
 			memcpy(&lunid, &ld_buff->LUN[j][0], 4);
 			lunid = le32_to_cpu(lunid);

_

#1 lore raw
Quoted 2 lines · On Tue, Dec 16 2008, [email protected] wrote:

Fix problem that deleting multiple logical drives could cause a panic.

This looks like a 2.6.28 candidate, correct?

Quoted 23 lines

---

 drivers/block/cciss.c |    5 +++++
 1 file changed, 5 insertions(+)

diff -puN drivers/block/cciss.c~fix_rebuild_lun_table_bug drivers/block/cciss.c

drivers/block/cciss.c renamed from drivers/block/cciss.c~fix_rebuild_lun_table_bug +5 -0
@@ -1693,6 +1693,11 @@ static int rebuild_lun_table(ctlr_info_t
 	for (i = 0; i <= h->highest_lun; i++) {
 		int j;
 		drv_found = 0;
+
+		/* skip holes in the array from already deleted drives */
+		if (h->drv[i].raid_level == -1)
+			continue;
+
 		for (j = 0; j < num_luns; j++) {
 			memcpy(&lunid, &ld_buff->LUN[j][0], 4);
 			lunid = le32_to_cpu(lunid);

_

Signature · 1 line
Jens Axboe
#2 lore raw
Quoted 5 lines · Jens Axboe
Quoted 2 lines · On Tue, Dec 16 2008, [email protected] wrote:

Fix problem that deleting multiple logical drives could cause a panic.

This looks like a 2.6.28 candidate, correct?

Yes, and we would like to get it in anywhere else it needs to go (backport to 2.6.27.x? if that makes sense.)

It fixes a panic which can be easily reproduced in the following
way: Just create several "arrays," each with multiple logical drives via hpacucli, then delete the first array, and it will blow up in deregister_disk(), in the call to get_host() when it tries to dig the hba pointer out of a NULL queue pointer.

The problem has been present since my code to make rebuild_lun_table behave better went in.

-- steve

Quoted 28 lines
Quoted 24 lines

---

 drivers/block/cciss.c |    5 +++++
 1 file changed, 5 insertions(+)

diff -puN drivers/block/cciss.c~fix_rebuild_lun_table_bug drivers/block/cciss.c

drivers/block/cciss.c renamed from drivers/block/cciss.c~fix_rebuild_lun_table_bug +5 -0
@@ -1693,6 +1693,11 @@ static int rebuild_lun_table(ctlr_info_t
 	for (i = 0; i <= h->highest_lun; i++) {
 		int j;
 		drv_found = 0;
+
+		/* skip holes in the array from already deleted drives */
+		if (h->drv[i].raid_level == -1)
+			continue;
+
 		for (j = 0; j < num_luns; j++) {
 			memcpy(&lunid, &ld_buff->LUN[j][0], 4);
 			lunid = le32_to_cpu(lunid);

_

Signature · 1 line
Jens Axboe
#3 lore raw
Quoted 10 lines · On Tue, Dec 16 2008, [email protected] wrote:
Quoted 5 lines · Jens Axboe
Quoted 2 lines · On Tue, Dec 16 2008, [email protected] wrote:

Fix problem that deleting multiple logical drives could cause a panic.

This looks like a 2.6.28 candidate, correct?

Yes, and we would like to get it in anywhere else it needs to go (backport to 2.6.27.x? if that makes sense.)

Quoted 8 lines

It fixes a panic which can be easily reproduced in the following
way: Just create several "arrays," each with multiple logical drives via hpacucli, then delete the first array, and it will blow up in deregister_disk(), in the call to get_host() when it tries to dig the hba pointer out of a NULL queue pointer.

The problem has been present since my code to make rebuild_lun_table behave better went in.

OK, I'll make sure it goes upstream for 2.6.28. I've merged your other cciss patch for 2.6.29.

Signature · 1 line
Jens Axboe