[RFC] Removing VLA usage in l1oip_core

[Gustavo A. R. Silva]

Hi Karsten,

I'm trying to figure out the best way to fix the following VLA warning:

drivers/isdn/mISDN/l1oip_core.c:282:2: warning: ISO C90 forbids variable length 
array ‘frame’ [-Wvla]
  u8 frame[len + 32];
  ^~

So while doing some research I've found the following.

Based on this code at include/linux/mISDNhw.h:38: #define MAX_DFRAME_LEN_L1 300 
and the following at drivers/isdn/mISDN/l1oip_core.c:1115:


                if (skb->len > MAX_DFRAME_LEN_L1 || skb->len > L1OIP_MAX_LEN) {
                        printk(KERN_WARNING "%s: skb too large\n",
                               __func__);
                        break;
                }
                /* check for AIS / ulaw-silence */
                l = skb->len;
                if (!memchr_inv(skb->data, 0xff, l)) {
                        if (debug & DEBUG_L1OIP_MSG)
                                printk(KERN_DEBUG "%s: got AIS, not sending, "
                                       "but counting\n", __func__);
                        hc->chan[bch->slot].tx_counter += l;
                        skb_trim(skb, 0);
                        queue_ch_frame(ch, PH_DATA_CNF, hh->id, skb);
                        return 0;
                }
                /* check for silence */
                l = skb->len;
                if (!memchr_inv(skb->data, 0x2a, l)) {
                        if (debug & DEBUG_L1OIP_MSG)
                                printk(KERN_DEBUG "%s: got silence, not sending"
                                       ", but counting\n", __func__);
                        hc->chan[bch->slot].tx_counter += l;
                        skb_trim(skb, 0);
                        queue_ch_frame(ch, PH_DATA_CNF, hh->id, skb);
                        return 0;
                }

                /* send frame */
                p = skb->data;
                l = skb->len;
                while (l) {
                        ll = (l < L1OIP_MAX_PERFRAME) ? l : L1OIP_MAX_PERFRAME;
                        l1oip_socket_send(hc, hc->codec, bch->slot, 0,
                                          hc->chan[bch->slot].tx_counter, p, 
ll);
                        hc->chan[bch->slot].tx_counter += ll;
                        p += ll;
                        l -= ll;
                }


it seems that the maximum value 'len' can take at 
drivers/isdn/mISDN/l1oip_core.c:282 is 300


/*
 * send a frame via socket, if open and restart timer
 */
static int
l1oip_socket_send(struct l1oip *hc, u8 localcodec, u8 channel, u32 chanmask,
                  u16 timebase, u8 *buf, int len)
{
        u8 *p;
        u8 frame[len + 32];


If this is correct, I could send the following patch to fix the VLA warning:

diff --git a/drivers/isdn/mISDN/l1oip_core.c b/drivers/isdn/mISDN/l1oip_core.c
index 21d50e4..31e3cd5 100644
--- a/drivers/isdn/mISDN/l1oip_core.c
+++ b/drivers/isdn/mISDN/l1oip_core.c
@@ -279,7 +279,7 @@ l1oip_socket_send(struct l1oip *hc, u8 localcodec, u8 
channel, u32 chanmask,
                  u16 timebase, u8 *buf, int len)
 {
        u8 *p;
-       u8 frame[len + 32];
+       u8 frame[332];
        struct socket *socket = NULL;
 
        if (debug & DEBUG_L1OIP_MSG)

But I wanted to ask for your feedback first, in case I may be missing something.

Another solution is to use dynamic memory allocation, but if the maximum size 
for 'frame' is in the hundreds of bytes, it might not be worth the performace 
penalty.

What do you think?

Thanks!
--
Gustavo