Re: Bugs in multipath scsi in 4.3-rc2

[Paul Mackerras <paulus@ozlabs.org>]

On Wed, Sep 30, 2015 at 05:14:49PM +0200, Christoph Hellwig wrote:
> 
> Paul, can you try the trivial one liner below?
> 
> diff --git a/drivers/scsi/scsi_dh.c b/drivers/scsi/scsi_dh.c
> index edb044a..fbc9502 100644
> --- a/drivers/scsi/scsi_dh.c
> +++ b/drivers/scsi/scsi_dh.c
> @@ -391,7 +391,7 @@ int scsi_dh_attach(struct request_queue *q, const char *name)
>  	if (!sdev)
>  		return -ENODEV;
>  
> -	scsi_dh = scsi_dh_lookup(name);
> +	scsi_dh = __scsi_dh_lookup(name);
>  	if (!scsi_dh) {
>  		err = -EINVAL;
>  		goto out_put_device;

I still get the warning:

WARNING: at /home/paulus/kernel/kvm/kernel/kmod.c:140
Modules linked in: radeon(+) i2c_algo_bit drm_kms_helper syscopyarea sysfillrect
 sysimgblt fb_sys_fops ttm drm tg3(+) ptp ipr cxgb3 pps_core i2c_core mdio dm_multipath
CPU: 11 PID: 7 Comm: kworker/u321:0 Not tainted 4.3.0-rc2-kvm+ #73
Workqueue: events_unbound async_run_entry_fn
task: c000000ff0545080 ti: c000000ff0610000 task.ti: c000000ff0610000

...

NIP [c0000000000d390c] __request_module+0x21c/0x380
LR [c0000000000d38f4] __request_module+0x204/0x380
Call Trace:
[c000000ff0613920] [c0000000000d38f4] __request_module+0x204/0x380 (unreliable)
[c000000ff06139e0] [c0000000006369d4] scsi_dh_lookup+0x64/0x80
[c000000ff0613a50] [c000000000636fcc] scsi_dh_add_device+0x13c/0x170
[c000000ff0613aa0] [c000000000630ea4] scsi_sysfs_add_sdev+0x114/0x380
[c000000ff0613b30] [c00000000062e040] do_scan_async+0xf0/0x240
[c000000ff0613bc0] [c0000000000e6bc0] async_run_entry_fn+0xa0/0x200
[c000000ff0613c50] [c0000000000d9750] process_one_work+0x1a0/0x4b0
[c000000ff0613ce0] [c0000000000d9bf0] worker_thread+0x190/0x5f0
[c000000ff0613d80] [c0000000000e21b0] kthread+0x110/0x130
[c000000ff0613e30] [c0000000000095b0] ret_from_kernel_thread+0x5c/0xac

So it's the call to scsi_dh_lookup() in scsi_dh_add_device() that is
causing the warning this time.  Do you want me to add the "__" to that
one too?

Paul.