On 05/21, Mina Almasry wrote:
> On Tue, May 20, 2025 at 1:30 PM Stanislav Fomichev <stfomic...@gmail.com>
> wrote:
> >
> > Add new -z argument to specify max IOV size. By default, use
> > single large IOV.
> >
> > Signed-off-by: Stanislav Fomichev <stfomic...@gmail.com>
> > ---
> > .../selftests/drivers/net/hw/ncdevmem.c | 49 +++++++++++--------
> > 1 file changed, 29 insertions(+), 20 deletions(-)
> >
> > diff --git a/tools/testing/selftests/drivers/net/hw/ncdevmem.c
> > b/tools/testing/selftests/drivers/net/hw/ncdevmem.c
> > index ca723722a810..fc7ba7d71502 100644
> > --- a/tools/testing/selftests/drivers/net/hw/ncdevmem.c
> > +++ b/tools/testing/selftests/drivers/net/hw/ncdevmem.c
> > @@ -82,6 +82,9 @@
> > #define MSG_SOCK_DEVMEM 0x2000000
> > #endif
> >
> > +#define MAX_IOV 1024
> > +
> > +static size_t max_chunk;
> > static char *server_ip;
> > static char *client_ip;
> > static char *port;
> > @@ -834,10 +837,10 @@ static int do_client(struct memory_buffer *mem)
> > struct sockaddr_in6 server_sin;
> > struct sockaddr_in6 client_sin;
> > struct ynl_sock *ys = NULL;
> > + struct iovec iov[MAX_IOV];
> > struct msghdr msg = {};
> > ssize_t line_size = 0;
> > struct cmsghdr *cmsg;
> > - struct iovec iov[2];
> > char *line = NULL;
> > unsigned long mid;
> > size_t len = 0;
> > @@ -893,27 +896,29 @@ static int do_client(struct memory_buffer *mem)
> > if (line_size < 0)
> > break;
> >
> > - mid = (line_size / 2) + 1;
> > -
> > - iov[0].iov_base = (void *)1;
> > - iov[0].iov_len = mid;
> > - iov[1].iov_base = (void *)(mid + 2);
> > - iov[1].iov_len = line_size - mid;
> > + if (max_chunk) {
> > + msg.msg_iovlen =
> > + (line_size + max_chunk - 1) / max_chunk;
> > + if (msg.msg_iovlen > MAX_IOV)
> > + error(1, 0,
> > + "can't partition %zd bytes into
> > maximum of %d chunks",
> > + line_size, MAX_IOV);
> >
> > - provider->memcpy_to_device(mem, (size_t)iov[0].iov_base,
> > line,
> > - iov[0].iov_len);
> > - provider->memcpy_to_device(mem, (size_t)iov[1].iov_base,
> > - line + iov[0].iov_len,
> > - iov[1].iov_len);
> > + for (int i = 0; i < msg.msg_iovlen; i++) {
> > + iov[i].iov_base = (void *)(i * max_chunk);
> > + iov[i].iov_len = max_chunk;
>
> Isn't the last iov going to be truncated in the case where line_size
> is not exactly divisible with max_chunk?
I have this for the last iov entry:
iov[msg.msg_iovlen - 1].iov_len =
line_size - (msg.msg_iovlen - 1) * max_chunk;
I think that should correctly adjust it to the remaining 1..max_chunk
len?
> > + }
> >
> > - fprintf(stderr,
> > - "read line_size=%ld iov[0].iov_base=%lu,
> > iov[0].iov_len=%lu, iov[1].iov_base=%lu, iov[1].iov_len=%lu\n",
> > - line_size, (unsigned long)iov[0].iov_base,
> > - iov[0].iov_len, (unsigned long)iov[1].iov_base,
> > - iov[1].iov_len);
> > + iov[msg.msg_iovlen - 1].iov_len =
> > + line_size - (msg.msg_iovlen - 1) *
> > max_chunk;
> > + } else {
> > + iov[0].iov_base = 0;
> > + iov[0].iov_len = line_size;
> > + msg.msg_iovlen = 1;
> > + }
>
> Do you need to special case this? Shouldn't this be the same as max_chunk==1?
I might need a better name. max_chunk is the max size of the iov entry
(max iov_len), not the max number of IOVs. And I use max_chunk==0 as
"max size of the iov is not provided -> use line_size", so not sure
I can't make it work without a special case?
