Re: [PATCH 10/18] nvme-tcp: fixup send workflow for kTLS

[Hannes Reinecke <hare@suse.de>]

On 3/22/23 10:31, Sagi Grimberg wrote:
> 
> 
> On 3/21/23 14:43, Hannes Reinecke wrote:
>> kTLS does not support MSG_EOR flag for sendmsg(), and the ->sendpage()
>> call really doesn't bring any benefit as data has to be copied
>> anyway.
>> So use sock_no_sendpage() or sendmsg() instead, and ensure that the
>> MSG_EOR flag is blanked out for kTLS.
>>
>> Signed-off-by: Hannes Reinecke <hare@suse.de>
>> ---
>>   drivers/nvme/host/tcp.c | 33 +++++++++++++++++++++------------
>>   1 file changed, 21 insertions(+), 12 deletions(-)
>>
>> diff --git a/drivers/nvme/host/tcp.c b/drivers/nvme/host/tcp.c
>> index bbff1f52a167..007d457cacf9 100644
>> --- a/drivers/nvme/host/tcp.c
>> +++ b/drivers/nvme/host/tcp.c
>> @@ -1034,13 +1034,19 @@ static int nvme_tcp_try_send_data(struct 
>> nvme_tcp_request *req)
>>           bool last = nvme_tcp_pdu_last_send(req, len);
>>           int req_data_sent = req->data_sent;
>>           int ret, flags = MSG_DONTWAIT;
>> +        bool do_sendpage = sendpage_ok(page);
>> -        if (last && !queue->data_digest && !nvme_tcp_queue_more(queue))
>> +        if (!last || queue->data_digest || nvme_tcp_queue_more(queue))
>> +            flags |= MSG_MORE;
>> +        else if (!test_bit(NVME_TCP_Q_TLS, &queue->flags))
>>               flags |= MSG_EOR;
>> -        else
>> -            flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
> 
> I think its time to move the flags setting to a helper.
> 
>> -        if (sendpage_ok(page)) {
>> +        if (test_bit(NVME_TCP_Q_TLS, &queue->flags))
>> +            do_sendpage = false;
>> +
>> +        if (do_sendpage) {
> 
> The do_sendpage looks redundant to me.
> 
>> +            if (flags & MSG_MORE)
>> +                flags |= MSG_SENDPAGE_NOTLAST;
>>               ret = kernel_sendpage(queue->sock, page, offset, len,
>>                       flags);
> 
> I think that the SENDPAGE_NOLAST should be set together with MSG_MORE
> regardless.
> 
>>           } else {
>> @@ -1088,19 +1094,22 @@ static int nvme_tcp_try_send_cmd_pdu(struct 
>> nvme_tcp_request *req)
>>       bool inline_data = nvme_tcp_has_inline_data(req);
>>       u8 hdgst = nvme_tcp_hdgst_len(queue);
>>       int len = sizeof(*pdu) + hdgst - req->offset;
>> -    int flags = MSG_DONTWAIT;
>> +    struct msghdr msg = { .msg_flags = MSG_DONTWAIT };
>> +    struct kvec iov = {
>> +        .iov_base = (u8 *)req->pdu + req->offset,
>> +        .iov_len = len,
>> +    };
>>       int ret;
>>       if (inline_data || nvme_tcp_queue_more(queue))
>> -        flags |= MSG_MORE | MSG_SENDPAGE_NOTLAST;
>> -    else
>> -        flags |= MSG_EOR;
>> +        msg.msg_flags |= MSG_MORE;
>> +    else if (!test_bit(NVME_TCP_Q_TLS, &queue->flags))
>> +        msg.msg_flags |= MSG_EOR;
>>       if (queue->hdr_digest && !req->offset)
>>           nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
>> -    ret = kernel_sendpage(queue->sock, virt_to_page(pdu),
>> -            offset_in_page(pdu) + req->offset, len,  flags);
>> +    ret = kernel_sendmsg(queue->sock, &msg, &iov, 1, iov.iov_len);
> 
> I'd prefer to do kernel_sednpage/sock_no_sendpage similar to how we do
> it for data and data pdu.
> 
>>       if (unlikely(ret <= 0))
>>           return ret;
>> @@ -1131,7 +1140,7 @@ static int nvme_tcp_try_send_data_pdu(struct 
>> nvme_tcp_request *req)
>>       if (queue->hdr_digest && !req->offset)
>>           nvme_tcp_hdgst(queue->snd_hash, pdu, sizeof(*pdu));
>> -    if (!req->h2cdata_left)
>> +    if (!test_bit(NVME_TCP_Q_TLS, &queue->flags) && !req->h2cdata_left)
>>           ret = kernel_sendpage(queue->sock, virt_to_page(pdu),
>>                   offset_in_page(pdu) + req->offset, len,
>>                   MSG_DONTWAIT | MSG_MORE | MSG_SENDPAGE_NOTLAST);
> 
> Something is unclear to me. Is kernel_sendpage unsupported with tls? (I
> think it is). I understand the motivation to add more checks in the code
> for kernel_sendpage vs. sock_no_sendpage given that it should be
> perfectly fine to use either.
> 
> Did you see any regressions with using kernel_sendpage? If so, isn't
> that a bug in the tls code?

The actual issue with the tls code is the 'MSG_EOR' handling.
Problem is that tls is using MSG_EOR internally, and bails out on 
unknown MSG_ settings:

int tls_sw_sendmsg(struct sock *sk, struct msghdr *msg, size_t size)
{
[ .. ]
         if (msg->msg_flags & ~(MSG_MORE | MSG_DONTWAIT | MSG_NOSIGNAL |
                                MSG_CMSG_COMPAT))
                 return -EOPNOTSUPP;

I would _vastly_ prefer to blank out unsupported flags (like MSG_EOR) 
from the TLS code, because to all intents and purposes MSG_EOR is just 
the opposite of MSG_MORE.
Or drop MSG_EOR usage from the nvme tcp code.
But then I'm not _that_ into the networking code to make a judgement here.
And as we're using sendmsg() already I had been switching to use it for 
ktls, too (as I know that the sendmsg() flow worked).
But in the end I guess we could use sendpage going forward.

I'll check.

Cheers,

Hannes