Re: [BUG] usb: cdc-wdm: Missing barriers in ad-hoc lockless buffer

[Oliver Neukum <oneukum@suse.com>]

On 05.03.26 14:26, Gui-Dong Han wrote:

Hi,

> Based on my shallow understanding, reordering issues typically happen
> between different memory addresses, not within the same one.

Nevertheless, you've found the issue, hence I will ask you :-)

Is that something we can depend on or is that just how it works
on the architectures we are currently running on? If I go to the effort
of checking for reordering effects, I want to do it right in all cases.
   
> The real danger of weak memory architectures lies in accessing
> associated variables. For instance, if we write 1 to int a and then 2
> to int b, another CPU might observe b == 2 before a == 1. This is
> exactly the situation I pointed out in my original report regarding
> the lack of barriers between desc->ubuf and desc->length.

Yes. Hence I was looking. The results of a completed IO can be

a) data
b) an error
c) a buffer overflow

thus there must be ordering between recording any of these results
and changing WDM_READ, right?

> Honestly, lockless algorithm design is incredibly hard, which is why
> drivers should probably just rely on well-tested libraries instead of
> rolling their own. I am definitely no expert in this dark art, just
> know enough to be dangerous :)

I agree. The issue is that lockless IO is also error handling, not
just the buffer.

	Regards
		Oliver