Re: [PATCH v7 04/15] net: prepare for non devmem TCP memory providers

[Pavel Begunkov <asml.silence@gmail.com>]

On 11/4/24 20:20, Mina Almasry wrote:
> On Fri, Nov 1, 2024 at 10:41 AM Pavel Begunkov <asml.silence@gmail.com> wrote:
...
>> io_uring originated netmem that are marked unreadable as well
>> and so will end up in tcp_recvmsg_dmabuf(), then we reject and
>> fail since they should not be fed to devmem TCP. It should be
>> fine from correctness perspective.
>>
>> We need to check frags, and that's the place where we iterate
>> frags. Another option is to add a loop in tcp_recvmsg_locked
>> walking over all frags of an skb and doing the checks, but
>> that's an unnecessary performance burden to devmem.
>>
> 
> Checking each frag in tcp_recvmsg_dmabuf (and the equivalent io_uring
> function) is not ideal really. Especially when you're dereferencing
> nio->pp to do the check which IIUC will pull a cache line not normally
> needed in this code path and may have a performance impact.

Even if it's cold, all net_iov processed are expected to come
from the same page pool and would be cached. And it should be of
a comparable hotness as dmabuf_genpool_chunk_owner accesses below,
considering that there is enough of processing happening around,
it should be of a worry and can be improve upon later.

-- 
Pavel Begunkov