projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / commitdiff
? search:
summary | shortlog | log | commit | commitdiff | tree
raw | patch | inline | side by side (parent: cdbdda3)
scsi: ipr: Fix softlockup when rescanning devices in petitboot
authorWen Xiong <wenxiong@linux.vnet.ibm.com>
Fri, 6 Mar 2020 15:57:28 +0000 (09:57 -0600)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Thu, 2 Apr 2020 14:34:27 +0000 (16:34 +0200)
[ Upstream commit 394b61711f3ce33f75bf70a3e22938464a13b3ee ]

When trying to rescan disks in petitboot shell, we hit the following
softlockup stacktrace:

Kernel panic - not syncing: System is deadlocked on memory
[  241.223394] CPU: 32 PID: 693 Comm: sh Not tainted 5.4.16-openpower1 #1
[  241.223406] Call Trace:
[  241.223415] [c0000003f07c3180] [c000000000493fc4] dump_stack+0xa4/0xd8 (unreliable)
[  241.223432] [c0000003f07c31c0] [c00000000007d4ac] panic+0x148/0x3cc
[  241.223446] [c0000003f07c3260] [c000000000114b10] out_of_memory+0x468/0x4c4
[  241.223461] [c0000003f07c3300] [c0000000001472b0] __alloc_pages_slowpath+0x594/0x6d8
[  241.223476] [c0000003f07c3420] [c00000000014757c] __alloc_pages_nodemask+0x188/0x1a4
[  241.223492] [c0000003f07c34a0] [c000000000153e10] alloc_pages_current+0xcc/0xd8
[  241.223508] [c0000003f07c34e0] [c0000000001577ac] alloc_slab_page+0x30/0x98
[  241.223524] [c0000003f07c3520] [c0000000001597fc] new_slab+0x138/0x40c
[  241.223538] [c0000003f07c35f0] [c00000000015b204] ___slab_alloc+0x1e4/0x404
[  241.223552] [c0000003f07c36c0] [c00000000015b450] __slab_alloc+0x2c/0x48
[  241.223566] [c0000003f07c36f0] [c00000000015b754] kmem_cache_alloc_node+0x9c/0x1b4
[  241.223582] [c0000003f07c3760] [c000000000218c48] blk_alloc_queue_node+0x34/0x270
[  241.223599] [c0000003f07c37b0] [c000000000226574] blk_mq_init_queue+0x2c/0x78
[  241.223615] [c0000003f07c37e0] [c0000000002ff710] scsi_mq_alloc_queue+0x28/0x70
[  241.223631] [c0000003f07c3810] [c0000000003005b8] scsi_alloc_sdev+0x184/0x264
[  241.223647] [c0000003f07c38a0] [c000000000300ba0] scsi_probe_and_add_lun+0x288/0xa3c
[  241.223663] [c0000003f07c3a00] [c000000000301768] __scsi_scan_target+0xcc/0x478
[  241.223679] [c0000003f07c3b20] [c000000000301c64] scsi_scan_channel.part.9+0x74/0x7c
[  241.223696] [c0000003f07c3b70] [c000000000301df4] scsi_scan_host_selected+0xe0/0x158
[  241.223712] [c0000003f07c3bd0] [c000000000303f04] store_scan+0x104/0x114
[  241.223727] [c0000003f07c3cb0] [c0000000002d5ac4] dev_attr_store+0x30/0x4c
[  241.223741] [c0000003f07c3cd0] [c0000000001dbc34] sysfs_kf_write+0x64/0x78
[  241.223756] [c0000003f07c3cf0] [c0000000001da858] kernfs_fop_write+0x170/0x1b8
[  241.223773] [c0000003f07c3d40] [c0000000001621fc] __vfs_write+0x34/0x60
[  241.223787] [c0000003f07c3d60] [c000000000163c2c] vfs_write+0xa8/0xcc
[  241.223802] [c0000003f07c3db0] [c000000000163df4] ksys_write+0x70/0xbc
[  241.223816] [c0000003f07c3e20] [c00000000000b40c] system_call+0x5c/0x68

As a part of the scan process Linux will allocate and configure a
scsi_device for each target to be scanned. If the device is not present,
then the scsi_device is torn down. As a part of scsi_device teardown a
workqueue item will be scheduled and the lockups we see are because there
are 250k workqueue items to be processed.  Accoding to the specification of
SIS-64 sas controller, max_channel should be decreased on SIS-64 adapters
to 4.

The patch fixes softlockup issue.

Thanks for Oliver Halloran's help with debugging and explanation!

Link: https://lore.kernel.org/r/1583510248-23672-1-git-send-email-wenxiong@linux.vnet.ibm.com
Signed-off-by: Wen Xiong <wenxiong@linux.vnet.ibm.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
Signed-off-by: Sasha Levin <sashal@kernel.org>
drivers/scsi/ipr.c patch | blob | blame | history
drivers/scsi/ipr.h patch | blob | blame | history

diff --git a/drivers/scsi/ipr.c b/drivers/scsi/ipr.c
index 35d54ee1c5c74464526f8b56c25c46efb6c9404e..b172f0a02083491382e8c72d0edc95cc06c131cf 100644 (file)
--- a/drivers/scsi/ipr.c
+++ b/drivers/scsi/ipr.c
@@ -9962,6 +9962,7 @@ static void ipr_init_ioa_cfg(struct ipr_ioa_cfg *ioa_cfg,
        ioa_cfg->max_devs_supported = ipr_max_devs;
 
        if (ioa_cfg->sis64) {
+               host->max_channel = IPR_MAX_SIS64_BUSES;
                host->max_id = IPR_MAX_SIS64_TARGETS_PER_BUS;
                host->max_lun = IPR_MAX_SIS64_LUNS_PER_TARGET;
                if (ipr_max_devs > IPR_MAX_SIS64_DEVS)
@@ -9970,6 +9971,7 @@ static void ipr_init_ioa_cfg(struct ipr_ioa_cfg *ioa_cfg,
                                           + ((sizeof(struct ipr_config_table_entry64)
                                               * ioa_cfg->max_devs_supported)));
        } else {
+               host->max_channel = IPR_VSET_BUS;
                host->max_id = IPR_MAX_NUM_TARGETS_PER_BUS;
                host->max_lun = IPR_MAX_NUM_LUNS_PER_TARGET;
                if (ipr_max_devs > IPR_MAX_PHYSICAL_DEVS)
@@ -9979,7 +9981,6 @@ static void ipr_init_ioa_cfg(struct ipr_ioa_cfg *ioa_cfg,
                                               * ioa_cfg->max_devs_supported)));
        }
 
-       host->max_channel = IPR_VSET_BUS;
        host->unique_id = host->host_no;
        host->max_cmd_len = IPR_MAX_CDB_LEN;
        host->can_queue = ioa_cfg->max_cmds;
diff --git a/drivers/scsi/ipr.h b/drivers/scsi/ipr.h
index 085e6c90f9e6a3255469757c2d6febbaa0d66e25..89b36987ff309f6fa08e410d1d918628281445ea 100644 (file)
--- a/drivers/scsi/ipr.h
+++ b/drivers/scsi/ipr.h
@@ -1306,6 +1306,7 @@ struct ipr_resource_entry {
 #define IPR_ARRAY_VIRTUAL_BUS                  0x1
 #define IPR_VSET_VIRTUAL_BUS                   0x2
 #define IPR_IOAFP_VIRTUAL_BUS                  0x3
+#define IPR_MAX_SIS64_BUSES                    0x4
 
 #define IPR_GET_RES_PHYS_LOC(res) \
        (((res)->bus << 24) | ((res)->target << 8) | (res)->lun)