On Wed, May 15, 2019 at 12:50:32PM +0200, Bart Van Assche wrote: > On 5/15/19 11:50 AM, Leon Romanovsky wrote: > > @@ -4089,11 +4093,15 @@ static struct srp_host *srp_add_port(struct srp_device *device, u8 port) > > > > host->dev.class = &srp_class; > > host->dev.parent = device->dev->dev.parent; > > - dev_set_name(&host->dev, "srp-%s-%d", dev_name(&device->dev->dev), > > - port); > > + devnum = find_first_zero_bit(dev_map, SRP_MAX_DEVICES); > > + if (devnum >= SRP_MAX_DEVICES) > > + goto free_host; > > + set_bit(devnum, dev_map); > > + host->devnum = devnum; > > + dev_set_name(&host->dev, "srp%d", devnum); > > > > if (device_register(&host->dev)) > > - goto free_host; > > + goto free_num; > > if (device_create_file(&host->dev, &dev_attr_add_target)) > > goto err_class; > > if (device_create_file(&host->dev, &dev_attr_ibdev)) > > Hi Leon, > > Thank you for having root-caused this issue. However, this patch > modifies the ABI between kernel and user space and hence breaks at least > srp_daemon and blktests. Are you aware it is considered completely > unacceptable in the Linux community to break user space? You may have > noticed that the SRP sysfs ABI has been documented in > Documentation/ABI/stable/sysfs-driver-ib_srp. Are you aware that ib_srp is broken for systems with 2+ adapters? Any suggestions on what should be done in ib_srp to fix it? Thanks > > Bart. >
