[PATCH v2] libsas: fix bug for vacant phy

["Jack Wang" <jack_wang@usish.com>]

[-- Attachment #1: Type: text/plain, Size: 6766 bytes --]


Hi, James

Please drop the previous patch and apply this new one.
Attached patch fix following bugs reported by Chuck.

Chuck
Could you test if this solve your problem ?

Jack -
Signed-off-by: Jack Wang <jack_wang@usish.com>

Signed-off-by: Lindar <lindar_liu@usish.com>
---
 sas_expander.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/sas_expander.c b/sas_expander.c
index d1d86a6..f63dbea 100644
--- a/sas_expander.c
+++ b/sas_expander.c
@@ -175,10 +175,10 @@ static void sas_set_ex_phy(struct domain_device *dev,
int phy_id,
 	switch (resp->result) {
 	case SMP_RESP_PHY_VACANT:
 		phy->phy_state = PHY_VACANT;
-		return;
+		break;
 	default:
 		phy->phy_state = PHY_NOT_PRESENT;
-		return;
+		break;
 	case SMP_RESP_FUNC_ACC:
 		phy->phy_state = PHY_EMPTY; /* do not know yet */
 		break;
@@ -209,7 +209,8 @@ static void sas_set_ex_phy(struct domain_device *dev,
int phy_id,
 	phy->phy->negotiated_linkrate = phy->linkrate;
 
 	if (!rediscover)
-		sas_phy_add(phy->phy);
+		if (sas_phy_add(phy->phy))
+			sas_phy_free(phy->phy);
 
 	SAS_DPRINTK("ex %016llx phy%02d:%c attached: %016llx\n",
 		    SAS_ADDR(dev->sas_addr), phy->phy_id,
-- 
1.7.2.3.msysgit.0


I finally had a chance to try something more recent (2.6.34) and I still see
the problem. I posted my findings to linux-scsi
(http://marc.info/?l=linux-scsi&m=128254243405363&w=2), but no one has
commented. Do you have any suggestions for approaches to fix this? I'm
willing to do the work, but am a little unclear where to look. Thanks!

-----Original Message-----
From: jack wang [mailto:jack_wang@usish.com] 
Sent: Thursday, July 08, 2010 6:35 PM
To: Chuck Tuffli
Cc: 'lindar_liu'; 'aoqingy'; 'roy'
Subject: RE: BUG reported during pm8001 driver rmmod

Hi Jack.

We have been using your Linux driver for testing and it has been working
great (thanks!). Today I tested against a new JBOD (HP D2700) and am
hitting an error when unloading the driver. Note that I don't see this
error with other JBODs (IBM, USI, etc), only the new one. Have you seen
anything like this before?

Chuck

[510] uname -srv
Linux 2.6.31-22-server #60-Ubuntu SMP Thu May 27 03:42:09 UTC 2010
[511] sudo insmod ./pm8001.ko 
[512] sudo rmmod pm8001 
Segmentation fault
[513] dmesg
...
[14131.624620] pm8001 0000:09:00.0: pm8001: driver version 0.1.36
[14131.630619] pm8001 0000:09:00.0: PCI INT A -> GSI 16 (level, low) ->
IRQ 16
[14131.637752] pm8001 0000:09:00.0: setting latency timer to 64
[14132.541239] scsi4 : pm8001
[14132.544347]   alloc irq_desc for 97 on node 0
[14132.548799]   alloc kstat_irqs on node 0
[14132.552851] pm8001 0000:09:00.0: irq 97 for MSI/MSI-X
[14248.581889] scsi 4:0:0:0: Direct-Access     HP       DG0300FARVV
HPD6 PQ: 0 ANSI: 5
[14248.590706] sd 4:0:0:0: Attached scsi generic sg2 type 0
[14248.596738] sd 4:0:0:0: [sdb] 585937500 512-byte logical blocks: (300
GB/279 GiB)
[14248.605469] sd 4:0:0:0: [sdb] Write Protect is off
[14248.610371] sd 4:0:0:0: [sdb] Mode Sense: eb 00 10 08
[14248.616130] sd 4:0:0:0: [sdb] Write cache: disabled, read cache:
enabled, supports DPO and FUA
[14248.627080]  sdb: unknown partition table
[14248.634138] sd 4:0:0:0: [sdb] Attached SCSI disk
[14248.656076] scsi 4:0:1:0: Enclosure         HP       D2700 SAS AJ941A
0052 PQ: 0 ANSI: 5
[14248.667852] scsi 4:0:1:0: Attached scsi generic sg3 type 13
[14248.945688] ses 4:0:1:0: Attached Enclosure device
[14288.328616] pm8001 0000:09:00.0: PCI INT A disabled
[14288.333712] ------------[ cut here ]------------
[14288.338429] kernel BUG at
/build/buildd/linux-2.6.31/include/linux/transport_class.h:92!
[14288.343639] invalid opcode: 0000 [#1] SMP 
[14288.343639] last sysfs file:
/sys/devices/pci0000:00/0000:00:1c.0/0000:09:00.0/host4/port-4:0/expande
r-4:0/port-4:0:36/end_device-4:0:36/target4:0:1/4:0:1:0/type
[14288.362505] CPU 0 
[14288.362505] Modules linked in: ses enclosure pm8001(-) nfs lockd
nfs_acl auth_rpcgss sunrpc radeon ttm drm libsas i2c_algo_bit
scsi_transport_sas iptable_filter psmouse ip_tables i5400_edac edac_core
lp x_tables serio_raw i5k_amb shpchp parport floppy igb dca
[14288.391254] Pid: 1204, comm: rmmod Not tainted 2.6.31-22-server
#60-Ubuntu X7DW3
[14288.392505] RIP: 0010:[<ffffffffa00c10c8>]  [<ffffffffa00c10c8>]
sas_release_transport+0x88/0x90 [scsi_transport_sas]
[14288.402505] RSP: 0018:ffff88003cdd5eb8  EFLAGS: 00010286
[14288.412505] RAX: 00000000fffffff0 RBX: ffff88003b698000 RCX:
01000000000000c1
[14288.422505] RDX: ffff88003b698700 RSI: ffffffff812782b0 RDI:
ffffffff817d8fa0
[14288.422505] RBP: ffff88003cdd5ec8 R08: 0000000000000000 R09:
0000000000000000
[14288.432505] R10: 0000000000000000 R11: ffff88003d94d9f4 R12:
ffffffffa02a6fa0
[14288.442505] R13: 0000000000000000 R14: 00007fff3e63a700 R15:
0000000000000001
[14288.451254] FS:  00007f8dd6c6d6f0(0000) GS:ffff8800019f3000(0000)
knlGS:0000000000000000
[14288.452505] CS:  0010 DS: 0000 ES: 0000 CR0: 000000008005003b
[14288.462505] CR2: 00007fed69a640a0 CR3: 000000003c8ef000 CR4:
00000000000006f0
[14288.472505] DR0: 0000000000000000 DR1: 0000000000000000 DR2:
0000000000000000
[14288.472505] DR3: 0000000000000000 DR6: 00000000ffff0ff0 DR7:
0000000000000400
[14288.482505] Process rmmod (pid: 1204, threadinfo ffff88003cdd4000,
task ffff88003ca516b0)
[14288.492505] Stack:
[14288.492505]  0000000000000000 0000000000000880 ffff88003cdd5ed8
ffffffffa029dd20
[14288.502505] <0> ffff88003cdd5f78 ffffffff8108ed38 ffff88003cdd5ef8
ffffffff8107c909
[14288.512505] <0> ffffffffa02a6fa0 ffffffff00000880 ffff88003cdd5f14
0000000000000014
[14288.521254] Call Trace:
[14288.522505]  [<ffffffffa029dd20>] pm8001_exit+0x1c/0x1e [pm8001]
[14288.522505]  [<ffffffff8108ed38>] sys_delete_module+0x1a8/0x280
[14288.532505]  [<ffffffff8107c909>] ? up_read+0x9/0x10
[14288.541254]  [<ffffffff81012042>] system_call_fastpath+0x16/0x1b
[14288.542505] Code: 0f 6c 26 e1 85 c0 75 13 48 89 df e8 d3 31 05 e1 48
83 c4 08 5b c9 c3 0f 0b eb fe 0f 0b eb fe 0f 0b eb fe 0f 0b eb fe 0f 0b
eb fe <0f> 0b eb fe 0f 1f 40 00 55 48 89 e5 41 56 41 55 41 54 53 4c 8b 
[14288.562505] RIP  [<ffffffffa00c10c8>] sas_release_transport+0x88/0x90
[scsi_transport_sas]
[14288.572505]  RSP <ffff88003cdd5eb8>
[14288.580975] ---[ end trace 0436c237fa6eeca0 ]---

Hi， Chuck
I haven't seen this bug before and We don't have the HP JBOD to test. It
seams you hit the bug in linux-2.6.31/include/linux/transport_class.h:92!,
the problem is transport unresgister the HP Enclosure HP D2700,
the bug appears. Maybe there something wrong with the ses && enclosure
modules , could you update to newer kernel to see whether the bug 
still exists?

[-- Attachment #2: 0001-fix-bug-for-vacant-phy.patch --]
[-- Type: application/octet-stream, Size: 1251 bytes --]

>From d53c99693154be14f72c5a843ae4b7a1951c1b7c Mon Sep 17 00:00:00 2001
From: Jack Wang <jack_wang@usish.com>
Date: Wed, 15 Sep 2010 16:43:03 +0800
Subject: [PATCH] fix bug for vacant phy

Signed-off-by: Jack Wang <jack_wang@usish.com>
Signed-off-by: Lindar Liu <lindar_liu@usish.com>
---
 sas_expander.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)

diff --git a/sas_expander.c b/sas_expander.c
index d1d86a6..f63dbea 100644
--- a/sas_expander.c
+++ b/sas_expander.c
@@ -175,10 +175,10 @@ static void sas_set_ex_phy(struct domain_device *dev, int phy_id,
 	switch (resp->result) {
 	case SMP_RESP_PHY_VACANT:
 		phy->phy_state = PHY_VACANT;
-		return;
+		break;
 	default:
 		phy->phy_state = PHY_NOT_PRESENT;
-		return;
+		break;
 	case SMP_RESP_FUNC_ACC:
 		phy->phy_state = PHY_EMPTY; /* do not know yet */
 		break;
@@ -209,7 +209,8 @@ static void sas_set_ex_phy(struct domain_device *dev, int phy_id,
 	phy->phy->negotiated_linkrate = phy->linkrate;
 
 	if (!rediscover)
-		sas_phy_add(phy->phy);
+		if (sas_phy_add(phy->phy))
+			sas_phy_free(phy->phy);
 
 	SAS_DPRINTK("ex %016llx phy%02d:%c attached: %016llx\n",
 		    SAS_ADDR(dev->sas_addr), phy->phy_id,
-- 
1.7.2.3.msysgit.0