Re: [Qemu-devel] [PATCH v5] net: Allocating Large sized arrays to heap

[Pooja Dhannawat]

On Thu, Mar 31, 2016 at 3:28 PM, Stefan Hajnoczi <address@hidden> wrote:

On Mon, Mar 28, 2016 at 06:04:11PM +0530, Pooja Dhannawat wrote:
> nc_sendv_compat has a huge stack usage of 69680 bytes approx.
> Moving large arrays to heap to reduce stack usage.
>
> Reviewed-by: Stefan Hajnoczi <address@hidden>
> Signed-off-by: Pooja Dhannawat <address@hidden>
> ---
>  net/net.c | 13 +++++++++----
>  1 file changed, 9 insertions(+), 4 deletions(-)
>
> diff --git a/net/net.c b/net/net.c
> index b0c832e..663da13 100644
> --- a/net/net.c
> +++ b/net/net.c
> @@ -709,23 +709,28 @@ ssize_t qemu_send_packet_raw(NetClientState *nc, const uint8_t *buf, int size)
>  static ssize_t nc_sendv_compat(NetClientState *nc, const struct iovec *iov,
>                                 int iovcnt, unsigned flags)
>  {
> -    uint8_t buf[NET_BUFSIZE];
> +    uint8_t *buf = NULL;
>      uint8_t *buffer;
>      size_t offset;
> +    ssize_t ret;
>
>      if (iovcnt == 1) {
>          buffer = iov[0].iov_base;
>          offset = iov[0].iov_len;
>      } else {
> +        buf = g_new(uint8_t, NET_BUFSIZE);
>          buffer = buf;
> -        offset = iov_to_buf(iov, iovcnt, 0, buf, sizeof(buf));
> +        offset = iov_to_buf(iov, iovcnt, 0, buf, NET_BUFSIZE);
>      }
>
>      if (flags & QEMU_NET_PACKET_FLAG_RAW && nc->info->receive_raw) {
> -        return nc->info->receive_raw(nc, buffer, offset);
> +        ret = nc->info->receive_raw(nc, buffer, offset);
>      } else {
> -        return nc->info->receive(nc, buffer, offset);
> +        ret = nc->info->receive(nc, buffer, offset);
>      }
> +
> +    g_free(buf);
> +    return ret;
>  }
>
>  ssize_t qemu_deliver_packet_iov(NetClientState *sender,
> --
> 2.5.0

CCing Jason Wang, net subsystem maintainer.  Please use
scripts/get_maintainer.pl to find the right people to CC in future
patches.

Yes, I am sorry about that. Fam also pointed out this thing but I forgot this time also.

I will keep that mind from onwards.
 

We may also want to keep a smaller stack buffer so that reasonably-sized
packets (e.g. up to 2 KB) can be send without the performance cost of
g_malloc().