Re: [PATCH] RDMA/srp: Check dev_set_name() return value

[Jason Gunthorpe]

On Fri, Aug 05, 2022 at 01:34:34AM -0400, Bo Liu wrote:
> It's possible that dev_set_name() returns -ENOMEM, catch and handle this.
> 
> Signed-off-by: Bo Liu <liubo03@xxxxxxxxxx>
> ---
>  drivers/infiniband/ulp/srp/ib_srp.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)

Ah, while you are here can you fix this:

>  	if (device_register(&host->dev))
>  		goto free_host;
[..]
> free_host:
>         kfree(host);

That isn't allowed, you have to do put_device():

/**
 * device_register - register a device with the system.
 * @dev: pointer to the device structure
 *
 * This happens in two clean steps - initialize the device
 * and add it to the system. The two steps can be called
 * separately, but this is the easiest and most common.
 * I.e. you should only call the two helpers separately if
 * have a clearly defined need to use and refcount the device
 * before it is added to the hierarchy.
 *
 * For more information, see the kerneldoc for device_initialize()
 * and device_add().
 *
 * NOTE: _Never_ directly free @dev after calling this function, even
 * if it returned an error! Always use put_device() to give up the
 * reference initialized in this function instead.
 */
int device_register(struct device *dev)

Everyone get this wrong, it is why I think device_register is a
terrible interface. Write it as device_initialize()/device_add() as
seperate steps and never call kfree().

Jason