3.16.57-rc1 review patch. If anyone has any objections, please let me know.
------------------
From: Jason Yan yanaijie@huawei.com
commit 6f54120e17e311fd7ac42b9ec2a0611caa5b46ad upstream.
We've got a kernel panic when using sata disk with sas controller:
[115946.152283] Unable to handle kernel NULL pointer dereference at virtual address 000007d8 [115946.223963] CPU: 0 PID: 22175 Comm: kworker/0:1 Tainted: G W OEL 4.14.0 #1 [115946.232925] Workqueue: events ata_scsi_hotplug [115946.237938] task: ffff8021ee50b180 task.stack: ffff00000d5d0000 [115946.244717] PC is at sas_find_dev_by_rphy+0x44/0x114 [115946.250224] LR is at sas_find_dev_by_rphy+0x3c/0x114 ...... [115946.355701] Process kworker/0:1 (pid: 22175, stack limit = 0xffff00000d5d0000) [115946.363369] Call trace: [115946.456356] [<ffff000008878a9c>] sas_find_dev_by_rphy+0x44/0x114 [115946.462908] [<ffff000008878b8c>] sas_target_alloc+0x20/0x5c [115946.469408] [<ffff00000885a31c>] scsi_alloc_target+0x250/0x308 [115946.475781] [<ffff00000885ba30>] __scsi_add_device+0xb0/0x154 [115946.481991] [<ffff0000088b520c>] ata_scsi_scan_host+0x180/0x218 [115946.488367] [<ffff0000088b53d8>] ata_scsi_hotplug+0xb0/0xcc [115946.494801] [<ffff0000080ebd70>] process_one_work+0x144/0x390 [115946.501115] [<ffff0000080ec100>] worker_thread+0x144/0x418 [115946.507093] [<ffff0000080f2c98>] kthread+0x10c/0x138 [115946.512792] [<ffff0000080855dc>] ret_from_fork+0x10/0x18
We found that Ding Xiang has reported a similar bug before: https://patchwork.kernel.org/patch/9179817/
And this bug still exists in mainline. Since libsas handles hotplug and device adding/removing itself, do not need to schedule ata hot plug task here if it is a sas host.
Signed-off-by: Jason Yan yanaijie@huawei.com Cc: Ding Xiang dingxiang@huawei.com Signed-off-by: Tejun Heo tj@kernel.org Signed-off-by: Ben Hutchings ben@decadent.org.uk --- drivers/ata/libata-eh.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
--- a/drivers/ata/libata-eh.c +++ b/drivers/ata/libata-eh.c @@ -815,7 +815,8 @@ void ata_scsi_port_error_handler(struct
if (ap->pflags & ATA_PFLAG_LOADING) ap->pflags &= ~ATA_PFLAG_LOADING; - else if (ap->pflags & ATA_PFLAG_SCSI_HOTPLUG) + else if ((ap->pflags & ATA_PFLAG_SCSI_HOTPLUG) && + !(ap->flags & ATA_FLAG_SAS_HOST)) schedule_delayed_work(&ap->hotplug_task, 0);
if (ap->pflags & ATA_PFLAG_RECOVERED)