On 1/26/25 03:46, Wolfram Sang wrote:
Hi Guenter,
I don't know if my rationale is correct, but the attached patch fixes
the problem for me.
Ah, yes. I didn't think twice when seeing that the debugfs entry for
adapters is created in i2c_register_adapter(). But, of course, this
function is called from the probe() function of the parent device.
Your patch is definitely one solution, but give me a day to think more
about it...
Sure.
In this context, it would be great if client->debugfs was cleared if creating
the debugfs directory failed. While debugfs functions check if the dentry
pointer is valid, that is not the case for other kernel functions. Knowing
that the pointer is either NULL or valid would simplify drivers (such as the
PMBus code) which need to make that check.
Thanks,
Guenter
