nfp: bpf: don't depend on high order allocations for program image

The translator pre-allocates a buffer of maximal program size.
Due to HW/FW limitations the program buffer can't currently be
longer than 128Kb, so we used to kmalloc() it, and then map for
DMA directly.

Now that the late branch resolution is copying the program image
anyway, we can just kvmalloc() the buffer.  While at it, after
translation reallocate the buffer to save space.

Signed-off-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: Daniel Borkmann <daniel@iogearbox.net>

drivers/net/ethernet/netronome/nfp/bpf/jit.c

@@ -2676,6 +2676,20 @@ static int nfp_bpf_ustore_calc(u64 *prog, unsigned int len)
 	return 0;
 }
 
+static void nfp_bpf_prog_trim(struct nfp_prog *nfp_prog)
+{
+	void *prog;
+
+	prog = kvmalloc_array(nfp_prog->prog_len, sizeof(u64), GFP_KERNEL);
+	if (!prog)
+		return;
+
+	nfp_prog->__prog_alloc_len = nfp_prog->prog_len * sizeof(u64);
+	memcpy(prog, nfp_prog->prog, nfp_prog->__prog_alloc_len);
+	kvfree(nfp_prog->prog);
+	nfp_prog->prog = prog;
+}
+
 int nfp_bpf_jit(struct nfp_prog *nfp_prog)
 {
 	int ret;
@@ -2691,6 +2705,8 @@ int nfp_bpf_jit(struct nfp_prog *nfp_prog)
 		return -EINVAL;
 	}
 
+	nfp_bpf_prog_trim(nfp_prog);
+
 	return ret;
 }
 

drivers/net/ethernet/netronome/nfp/bpf/offload.c

@@ -42,6 +42,7 @@
 #include <linux/jiffies.h>
 #include <linux/timer.h>
 #include <linux/list.h>
+#include <linux/mm.h>
 
 #include <net/pkt_cls.h>
 #include <net/tc_act/tc_gact.h>
@@ -135,7 +136,7 @@ int nfp_bpf_translate(struct nfp_app *app, struct nfp_net *nn,
 	max_instr = nn_readw(nn, NFP_NET_CFG_BPF_MAX_LEN);
 	nfp_prog->__prog_alloc_len = max_instr * sizeof(u64);
 
-	nfp_prog->prog = kmalloc(nfp_prog->__prog_alloc_len, GFP_KERNEL);
+	nfp_prog->prog = kvmalloc(nfp_prog->__prog_alloc_len, GFP_KERNEL);
 	if (!nfp_prog->prog)
 		return -ENOMEM;
 
@@ -147,7 +148,7 @@ int nfp_bpf_destroy(struct nfp_app *app, struct nfp_net *nn,
 {
 	struct nfp_prog *nfp_prog = prog->aux->offload->dev_priv;
 
-	kfree(nfp_prog->prog);
+	kvfree(nfp_prog->prog);
 	nfp_prog_free(nfp_prog);
 
 	return 0;