Changes to Userspace multipath code to support lsi-rdac

PATCH In review Chandra Seetharaman 3 participants 7 messages Mar 21, 2007 – Mar 22, 2007 Device mapper

lore

Share & tools

lore

Chandra Seetharaman

author

2007-03-21 19:38 UTC #0 lore raw

cover Changes to Userspace multipath code to support lsi-rdac

Hi All,

While working on lsi-rdac hardware handler, I found that the checkers "tur" and "readsector0" are not serving the lsi-rdac properly. Also, the storage device has a active/passive state which need to be handled properly.

So, I wrote a new path checker for lsi-rdac, to use the C4 inquiry page to get the current owner of the lun and set the state as UP(active) or GHOST(passive). Changes are in patch 1/3.

After applying the above patch, I realized the priorities are not set properly. Made some changes to use the GHOST state while setting priorities. Changes are in patch 2/3.

Debugging the above patches found that the wait time (5 seconds) in multipathd was not sufficient for the devices to appear in /dev/ It was taking about 20 seconds. Changed the wait time to be 60 seconds. Change in patch 3/3.

Please provide me with your feedback/comment on these changes regarding correctness, completeness etc.,

Thanks & regards,

chandra

Signature · 4 lines

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [email protected]   |      .......you may get it.
----------------------------------------------------------------------

Chandra Seetharaman

author

2007-03-21 19:38 UTC #1 lore raw

PATCH 1/3 C4 Inquiry based path checker

libcheckers/Makefile +1 -1

Index: multipath-tools-0.4.7/libcheckers/Makefile
===================================================================
@@ -6,7 +6,7 @@
 
 include ../Makefile.inc
 
-OBJS = checkers.o readsector0.o tur.o directio.o emc_clariion.o hp_sw.o
+OBJS = checkers.o readsector0.o tur.o directio.o emc_clariion.o hp_sw.o rdac.o
 
 all: $(BUILD)

libcheckers/rdac.c new file +109 -0

109 lines not on this page · raw message

libcheckers/checkers.c +10 -0

Index: multipath-tools-0.4.7/libcheckers/checkers.c
===================================================================
@@ -7,6 +7,7 @@
 #include "tur.h"
 #include "hp_sw.h"
 #include "emc_clariion.h"
+#include "rdac.h"
 #include "readsector0.h"
 
 static struct checker checkers[] = {

@@ -48,6 +49,15 @@
 	},
 	{
 		.fd         = 0,
+		.name       = RDAC,
+		.message    = "",
+		.context    = NULL,
+		.check      = rdac,
+		.init       = rdac_init,
+		.free       = rdac_free
+	},
+	{
+		.fd         = 0,
 		.name       = READSECTOR0,
 		.message    = "",
 		.context    = NULL,

libcheckers/rdac.h new file +8 -0

Index: multipath-tools-0.4.7/libcheckers/rdac.h
===================================================================
@@ -0,0 +1,8 @@
+#ifndef _RDAC_H
+#define _RDAC_H
+
+int rdac(struct checker *);
+int rdac_init(struct checker *);
+void rdac_free(struct checker *);
+
+#endif /* _RDAC_H */

libcheckers/checkers.h +1 -0

Index: multipath-tools-0.4.7/libcheckers/checkers.h
===================================================================
@@ -14,6 +14,7 @@
 #define DIRECTIO     "directio"
 #define TUR          "tur"
 #define HP_SW        "hp_sw"
+#define RDAC         "rdac"
 #define EMC_CLARIION "emc_clariion"
 #define READSECTOR0  "readsector0"

Signature · 4 lines

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [email protected]   |      .......you may get it.
----------------------------------------------------------------------

Chandra Seetharaman

author

2007-03-21 19:38 UTC #2 lore raw

PATCH 2/3 Fix the priority calculation to use PATH_GHOST state

In reply to Chandra Seetharaman

libmultipath/discovery.c +1 -1

Index: multipath-tools-0.4.7/libmultipath/discovery.c
===================================================================
@@ -699,7 +699,7 @@
 	if (mask & DI_CHECKER && get_state(pp))
 		goto blank;
 	
-	if (mask & DI_PRIO && pp->state != PATH_DOWN)
+	if (mask & DI_PRIO && pp->state == PATH_UP)
 		get_prio(pp);
 
 	if (mask & DI_WWID && !strlen(pp->wwid))

libmultipath/switchgroup.c +1 -1

Index: multipath-tools-0.4.7/libmultipath/switchgroup.c
===================================================================
@@ -28,7 +28,7 @@
 		priority = 0;
 
 		vector_foreach_slot (pgp->paths, pp, j) {
-			if (pp->state != PATH_DOWN)
+			if (pp->state == PATH_UP)
 				priority += pp->priority;
 		}
 		pgp->priority = priority;

Signature · 4 lines

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [email protected]   |      .......you may get it.
----------------------------------------------------------------------

Chandra Seetharaman

author

2007-03-21 19:38 UTC #3 lore raw

PATCH 3/3 Increase the wait time for device to appear in /dev

In reply to Chandra Seetharaman

libmultipath/discovery.c +1 -1

Index: multipath-tools-0.4.7/libmultipath/discovery.c
===================================================================
@@ -112,7 +112,7 @@
  * not multipath(8), ran by udev
  */
 #if DAEMON
-#define WAIT_MAX_SECONDS 5
+#define WAIT_MAX_SECONDS 60
 #define WAIT_LOOP_PER_SECOND 5
 
 static int

Signature · 4 lines

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [email protected]   |      .......you may get it.
----------------------------------------------------------------------

Christophe Varoqui

2007-03-21 23:06 UTC #4 lore raw

In reply to Chandra Seetharaman

Quoted 14 lines · Chandra Seetharaman

Hi All,
While working on lsi-rdac hardware handler, I found that the checkers "tur" and "readsector0" are not serving the lsi-rdac properly. Also, the storage device has a active/passive state which need to be handled properly.
So, I wrote a new path checker for lsi-rdac, to use the C4 inquiry page to get the current owner of the lun and set the state as UP(active) or GHOST(passive). Changes are in patch 1/3.
After applying the above patch, I realized the priorities are not set properly. Made some changes to use the GHOST state while setting priorities. Changes are in patch 2/3.

Quoted 7 lines

Debugging the above patches found that the wait time (5 seconds) in multipathd was not sufficient for the devices to appear in /dev/ It was taking about 20 seconds. Changed the wait time to be 60 seconds. Change in patch 3/3.
Please provide me with your feedback/comment on these changes regarding correctness, completeness etc.,

Looks very good. I merged the 3 patches.

In 1/3, I changed the DEF_TIMEOUT name to avoid surcharging the one in checker.h

I'm having a hard time push to kernel.org, so don't expect to see the git tree updated soon.

Thanks,
cvaroqui

Bernd Zeimetz

2007-03-22 22:52 UTC #5 lore raw

Hi,

Quoted 3 lines

I'm having a hard time push to kernel.org, so don't expect to see the git tree updated soon.

hmm. git.kernel.org seems to work well, at least to get stuff - I have nothing to push.

Is there any other place where one could grab the latest multipath-tools, and probably recent kernel modules to test?

Thanks,

Signature · 2 lines

Bernd Zeimetz
<[email protected]>                         <http://bzed.de/>

Chandra Seetharaman

author

2007-03-22 23:45 UTC #6 lore raw

In reply to Christophe Varoqui

Cool,

Thanks.

chandra

Quoted 33 lines · Christophe Varoqui

Quoted 14 lines · Chandra Seetharaman
Hi All,
While working on lsi-rdac hardware handler, I found that the checkers "tur" and "readsector0" are not serving the lsi-rdac properly. Also, the storage device has a active/passive state which need to be handled properly.
So, I wrote a new path checker for lsi-rdac, to use the C4 inquiry page to get the current owner of the lun and set the state as UP(active) or GHOST(passive). Changes are in patch 1/3.
After applying the above patch, I realized the priorities are not set properly. Made some changes to use the GHOST state while setting priorities. Changes are in patch 2/3.
Quoted 7 lines
Debugging the above patches found that the wait time (5 seconds) in multipathd was not sufficient for the devices to appear in /dev/ It was taking about 20 seconds. Changed the wait time to be 60 seconds. Change in patch 3/3.
Please provide me with your feedback/comment on these changes regarding correctness, completeness etc.,
Looks very good. I merged the 3 patches.
In 1/3, I changed the DEF_TIMEOUT name to avoid surcharging the one in checker.h
I'm having a hard time push to kernel.org, so don't expect to see the git tree updated soon.
Thanks,
cvaroqui

Signature · 4 lines

----------------------------------------------------------------------
    Chandra Seetharaman               | Be careful what you choose....
              - [email protected]   |      .......you may get it.
----------------------------------------------------------------------