Re: [PATCH net-next v6 2/2] page_pool: Track DMA-mapped pages and unmap them when destroying the pool

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Pavel Begunkov <asml.silence@gmail.com> writes:

>>>> +	if (err) {
>>>> +		WARN_ONCE(1, "couldn't track DMA mapping, please report to netdev@");
>>>
>>> That can happen with enough memory pressure, I don't think
>>> it should be a warning. Maybe some pr_info?
>> 
>> So my reasoning here was that this code is only called in the alloc
>> path, so if we're under memory pressure, the page allocation itself
>> should fail before the xarray alloc does. And if it doesn't (i.e., if
>> the use of xarray itself causes allocation failures), we really want to
>> know about it so we can change things. Hence the loud warning.
>
> There is a gap between allocations, one doesn't guarantee
> another. I'd say the mental test here is whether we can reasonably
> cause it from user space (including by abusive users), because crash
> on warning setups exist, and it'll let you know about itself too
> loudly, when it could've been tolerated just fine. Not going to
> insist though.

Right, I do see what you mean - it's not guaranteed to be coupled.
However, my feeling is nonetheless that it's better for this to be loud
to weed out any new issues that may arise from this, so I'm inclined to
keep it as-is.

-Toke