1/* bpf_jit_comp.c: BPF JIT compiler
2 *
3 * Copyright 2011 Matt Evans <matt@ozlabs.org>, IBM Corporation
4 *
5 * Based on the x86 BPF compiler, by Eric Dumazet (eric.dumazet@gmail.com)
6 * Ported to ppc32 by Denis Kirjanov <kda@linux-powerpc.org>
7 *
8 * This program is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU General Public License
10 * as published by the Free Software Foundation; version 2
11 * of the License.
12 */
13#include <linux/moduleloader.h>
14#include <asm/cacheflush.h>
15#include <linux/netdevice.h>
16#include <linux/filter.h>
17#include <linux/if_vlan.h>
18
19#include "bpf_jit.h"
20
21int bpf_jit_enable __read_mostly;
22
23static inline void bpf_flush_icache(void *start, void *end)
24{
25	smp_wmb();
26	flush_icache_range((unsigned long)start, (unsigned long)end);
27}
28
29static void bpf_jit_build_prologue(struct bpf_prog *fp, u32 *image,
30				   struct codegen_context *ctx)
31{
32	int i;
33	const struct sock_filter *filter = fp->insns;
34
35	if (ctx->seen & (SEEN_MEM | SEEN_DATAREF)) {
36		/* Make stackframe */
37		if (ctx->seen & SEEN_DATAREF) {
38			/* If we call any helpers (for loads), save LR */
39			EMIT(PPC_INST_MFLR | __PPC_RT(R0));
40			PPC_BPF_STL(0, 1, PPC_LR_STKOFF);
41
42			/* Back up non-volatile regs. */
43			PPC_BPF_STL(r_D, 1, -(REG_SZ*(32-r_D)));
44			PPC_BPF_STL(r_HL, 1, -(REG_SZ*(32-r_HL)));
45		}
46		if (ctx->seen & SEEN_MEM) {
47			/*
48			 * Conditionally save regs r15-r31 as some will be used
49			 * for M[] data.
50			 */
51			for (i = r_M; i < (r_M+16); i++) {
52				if (ctx->seen & (1 << (i-r_M)))
53					PPC_BPF_STL(i, 1, -(REG_SZ*(32-i)));
54			}
55		}
56		PPC_BPF_STLU(1, 1, -BPF_PPC_STACKFRAME);
57	}
58
59	if (ctx->seen & SEEN_DATAREF) {
60		/*
61		 * If this filter needs to access skb data,
62		 * prepare r_D and r_HL:
63		 *  r_HL = skb->len - skb->data_len
64		 *  r_D	 = skb->data
65		 */
66		PPC_LWZ_OFFS(r_scratch1, r_skb, offsetof(struct sk_buff,
67							 data_len));
68		PPC_LWZ_OFFS(r_HL, r_skb, offsetof(struct sk_buff, len));
69		PPC_SUB(r_HL, r_HL, r_scratch1);
70		PPC_LL_OFFS(r_D, r_skb, offsetof(struct sk_buff, data));
71	}
72
73	if (ctx->seen & SEEN_XREG) {
74		/*
75		 * TODO: Could also detect whether first instr. sets X and
76		 * avoid this (as below, with A).
77		 */
78		PPC_LI(r_X, 0);
79	}
80
81	/* make sure we dont leak kernel information to user */
82	if (bpf_needs_clear_a(&filter[0]))
83		PPC_LI(r_A, 0);
84}
85
86static void bpf_jit_build_epilogue(u32 *image, struct codegen_context *ctx)
87{
88	int i;
89
90	if (ctx->seen & (SEEN_MEM | SEEN_DATAREF)) {
91		PPC_ADDI(1, 1, BPF_PPC_STACKFRAME);
92		if (ctx->seen & SEEN_DATAREF) {
93			PPC_BPF_LL(0, 1, PPC_LR_STKOFF);
94			PPC_MTLR(0);
95			PPC_BPF_LL(r_D, 1, -(REG_SZ*(32-r_D)));
96			PPC_BPF_LL(r_HL, 1, -(REG_SZ*(32-r_HL)));
97		}
98		if (ctx->seen & SEEN_MEM) {
99			/* Restore any saved non-vol registers */
100			for (i = r_M; i < (r_M+16); i++) {
101				if (ctx->seen & (1 << (i-r_M)))
102					PPC_BPF_LL(i, 1, -(REG_SZ*(32-i)));
103			}
104		}
105	}
106	/* The RETs have left a return value in R3. */
107
108	PPC_BLR();
109}
110
111#define CHOOSE_LOAD_FUNC(K, func) \
112	((int)K < 0 ? ((int)K >= SKF_LL_OFF ? func##_negative_offset : func) : func##_positive_offset)
113
114/* Assemble the body code between the prologue & epilogue. */
115static int bpf_jit_build_body(struct bpf_prog *fp, u32 *image,
116			      struct codegen_context *ctx,
117			      unsigned int *addrs)
118{
119	const struct sock_filter *filter = fp->insns;
120	int flen = fp->len;
121	u8 *func;
122	unsigned int true_cond;
123	int i;
124
125	/* Start of epilogue code */
126	unsigned int exit_addr = addrs[flen];
127
128	for (i = 0; i < flen; i++) {
129		unsigned int K = filter[i].k;
130		u16 code = bpf_anc_helper(&filter[i]);
131
132		/*
133		 * addrs[] maps a BPF bytecode address into a real offset from
134		 * the start of the body code.
135		 */
136		addrs[i] = ctx->idx * 4;
137
138		switch (code) {
139			/*** ALU ops ***/
140		case BPF_ALU | BPF_ADD | BPF_X: /* A += X; */
141			ctx->seen |= SEEN_XREG;
142			PPC_ADD(r_A, r_A, r_X);
143			break;
144		case BPF_ALU | BPF_ADD | BPF_K: /* A += K; */
145			if (!K)
146				break;
147			PPC_ADDI(r_A, r_A, IMM_L(K));
148			if (K >= 32768)
149				PPC_ADDIS(r_A, r_A, IMM_HA(K));
150			break;
151		case BPF_ALU | BPF_SUB | BPF_X: /* A -= X; */
152			ctx->seen |= SEEN_XREG;
153			PPC_SUB(r_A, r_A, r_X);
154			break;
155		case BPF_ALU | BPF_SUB | BPF_K: /* A -= K */
156			if (!K)
157				break;
158			PPC_ADDI(r_A, r_A, IMM_L(-K));
159			if (K >= 32768)
160				PPC_ADDIS(r_A, r_A, IMM_HA(-K));
161			break;
162		case BPF_ALU | BPF_MUL | BPF_X: /* A *= X; */
163			ctx->seen |= SEEN_XREG;
164			PPC_MUL(r_A, r_A, r_X);
165			break;
166		case BPF_ALU | BPF_MUL | BPF_K: /* A *= K */
167			if (K < 32768)
168				PPC_MULI(r_A, r_A, K);
169			else {
170				PPC_LI32(r_scratch1, K);
171				PPC_MUL(r_A, r_A, r_scratch1);
172			}
173			break;
174		case BPF_ALU | BPF_MOD | BPF_X: /* A %= X; */
175		case BPF_ALU | BPF_DIV | BPF_X: /* A /= X; */
176			ctx->seen |= SEEN_XREG;
177			PPC_CMPWI(r_X, 0);
178			if (ctx->pc_ret0 != -1) {
179				PPC_BCC(COND_EQ, addrs[ctx->pc_ret0]);
180			} else {
181				PPC_BCC_SHORT(COND_NE, (ctx->idx*4)+12);
182				PPC_LI(r_ret, 0);
183				PPC_JMP(exit_addr);
184			}
185			if (code == (BPF_ALU | BPF_MOD | BPF_X)) {
186				PPC_DIVWU(r_scratch1, r_A, r_X);
187				PPC_MUL(r_scratch1, r_X, r_scratch1);
188				PPC_SUB(r_A, r_A, r_scratch1);
189			} else {
190				PPC_DIVWU(r_A, r_A, r_X);
191			}
192			break;
193		case BPF_ALU | BPF_MOD | BPF_K: /* A %= K; */
194			PPC_LI32(r_scratch2, K);
195			PPC_DIVWU(r_scratch1, r_A, r_scratch2);
196			PPC_MUL(r_scratch1, r_scratch2, r_scratch1);
197			PPC_SUB(r_A, r_A, r_scratch1);
198			break;
199		case BPF_ALU | BPF_DIV | BPF_K: /* A /= K */
200			if (K == 1)
201				break;
202			PPC_LI32(r_scratch1, K);
203			PPC_DIVWU(r_A, r_A, r_scratch1);
204			break;
205		case BPF_ALU | BPF_AND | BPF_X:
206			ctx->seen |= SEEN_XREG;
207			PPC_AND(r_A, r_A, r_X);
208			break;
209		case BPF_ALU | BPF_AND | BPF_K:
210			if (!IMM_H(K))
211				PPC_ANDI(r_A, r_A, K);
212			else {
213				PPC_LI32(r_scratch1, K);
214				PPC_AND(r_A, r_A, r_scratch1);
215			}
216			break;
217		case BPF_ALU | BPF_OR | BPF_X:
218			ctx->seen |= SEEN_XREG;
219			PPC_OR(r_A, r_A, r_X);
220			break;
221		case BPF_ALU | BPF_OR | BPF_K:
222			if (IMM_L(K))
223				PPC_ORI(r_A, r_A, IMM_L(K));
224			if (K >= 65536)
225				PPC_ORIS(r_A, r_A, IMM_H(K));
226			break;
227		case BPF_ANC | SKF_AD_ALU_XOR_X:
228		case BPF_ALU | BPF_XOR | BPF_X: /* A ^= X */
229			ctx->seen |= SEEN_XREG;
230			PPC_XOR(r_A, r_A, r_X);
231			break;
232		case BPF_ALU | BPF_XOR | BPF_K: /* A ^= K */
233			if (IMM_L(K))
234				PPC_XORI(r_A, r_A, IMM_L(K));
235			if (K >= 65536)
236				PPC_XORIS(r_A, r_A, IMM_H(K));
237			break;
238		case BPF_ALU | BPF_LSH | BPF_X: /* A <<= X; */
239			ctx->seen |= SEEN_XREG;
240			PPC_SLW(r_A, r_A, r_X);
241			break;
242		case BPF_ALU | BPF_LSH | BPF_K:
243			if (K == 0)
244				break;
245			else
246				PPC_SLWI(r_A, r_A, K);
247			break;
248		case BPF_ALU | BPF_RSH | BPF_X: /* A >>= X; */
249			ctx->seen |= SEEN_XREG;
250			PPC_SRW(r_A, r_A, r_X);
251			break;
252		case BPF_ALU | BPF_RSH | BPF_K: /* A >>= K; */
253			if (K == 0)
254				break;
255			else
256				PPC_SRWI(r_A, r_A, K);
257			break;
258		case BPF_ALU | BPF_NEG:
259			PPC_NEG(r_A, r_A);
260			break;
261		case BPF_RET | BPF_K:
262			PPC_LI32(r_ret, K);
263			if (!K) {
264				if (ctx->pc_ret0 == -1)
265					ctx->pc_ret0 = i;
266			}
267			/*
268			 * If this isn't the very last instruction, branch to
269			 * the epilogue if we've stuff to clean up.  Otherwise,
270			 * if there's nothing to tidy, just return.  If we /are/
271			 * the last instruction, we're about to fall through to
272			 * the epilogue to return.
273			 */
274			if (i != flen - 1) {
275				/*
276				 * Note: 'seen' is properly valid only on pass
277				 * #2.	Both parts of this conditional are the
278				 * same instruction size though, meaning the
279				 * first pass will still correctly determine the
280				 * code size/addresses.
281				 */
282				if (ctx->seen)
283					PPC_JMP(exit_addr);
284				else
285					PPC_BLR();
286			}
287			break;
288		case BPF_RET | BPF_A:
289			PPC_MR(r_ret, r_A);
290			if (i != flen - 1) {
291				if (ctx->seen)
292					PPC_JMP(exit_addr);
293				else
294					PPC_BLR();
295			}
296			break;
297		case BPF_MISC | BPF_TAX: /* X = A */
298			PPC_MR(r_X, r_A);
299			break;
300		case BPF_MISC | BPF_TXA: /* A = X */
301			ctx->seen |= SEEN_XREG;
302			PPC_MR(r_A, r_X);
303			break;
304
305			/*** Constant loads/M[] access ***/
306		case BPF_LD | BPF_IMM: /* A = K */
307			PPC_LI32(r_A, K);
308			break;
309		case BPF_LDX | BPF_IMM: /* X = K */
310			PPC_LI32(r_X, K);
311			break;
312		case BPF_LD | BPF_MEM: /* A = mem[K] */
313			PPC_MR(r_A, r_M + (K & 0xf));
314			ctx->seen |= SEEN_MEM | (1<<(K & 0xf));
315			break;
316		case BPF_LDX | BPF_MEM: /* X = mem[K] */
317			PPC_MR(r_X, r_M + (K & 0xf));
318			ctx->seen |= SEEN_MEM | (1<<(K & 0xf));
319			break;
320		case BPF_ST: /* mem[K] = A */
321			PPC_MR(r_M + (K & 0xf), r_A);
322			ctx->seen |= SEEN_MEM | (1<<(K & 0xf));
323			break;
324		case BPF_STX: /* mem[K] = X */
325			PPC_MR(r_M + (K & 0xf), r_X);
326			ctx->seen |= SEEN_XREG | SEEN_MEM | (1<<(K & 0xf));
327			break;
328		case BPF_LD | BPF_W | BPF_LEN: /*	A = skb->len; */
329			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, len) != 4);
330			PPC_LWZ_OFFS(r_A, r_skb, offsetof(struct sk_buff, len));
331			break;
332		case BPF_LDX | BPF_W | BPF_LEN: /* X = skb->len; */
333			PPC_LWZ_OFFS(r_X, r_skb, offsetof(struct sk_buff, len));
334			break;
335
336			/*** Ancillary info loads ***/
337		case BPF_ANC | SKF_AD_PROTOCOL: /* A = ntohs(skb->protocol); */
338			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff,
339						  protocol) != 2);
340			PPC_NTOHS_OFFS(r_A, r_skb, offsetof(struct sk_buff,
341							    protocol));
342			break;
343		case BPF_ANC | SKF_AD_IFINDEX:
344		case BPF_ANC | SKF_AD_HATYPE:
345			BUILD_BUG_ON(FIELD_SIZEOF(struct net_device,
346						ifindex) != 4);
347			BUILD_BUG_ON(FIELD_SIZEOF(struct net_device,
348						type) != 2);
349			PPC_LL_OFFS(r_scratch1, r_skb, offsetof(struct sk_buff,
350								dev));
351			PPC_CMPDI(r_scratch1, 0);
352			if (ctx->pc_ret0 != -1) {
353				PPC_BCC(COND_EQ, addrs[ctx->pc_ret0]);
354			} else {
355				/* Exit, returning 0; first pass hits here. */
356				PPC_BCC_SHORT(COND_NE, ctx->idx * 4 + 12);
357				PPC_LI(r_ret, 0);
358				PPC_JMP(exit_addr);
359			}
360			if (code == (BPF_ANC | SKF_AD_IFINDEX)) {
361				PPC_LWZ_OFFS(r_A, r_scratch1,
362				     offsetof(struct net_device, ifindex));
363			} else {
364				PPC_LHZ_OFFS(r_A, r_scratch1,
365				     offsetof(struct net_device, type));
366			}
367
368			break;
369		case BPF_ANC | SKF_AD_MARK:
370			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, mark) != 4);
371			PPC_LWZ_OFFS(r_A, r_skb, offsetof(struct sk_buff,
372							  mark));
373			break;
374		case BPF_ANC | SKF_AD_RXHASH:
375			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, hash) != 4);
376			PPC_LWZ_OFFS(r_A, r_skb, offsetof(struct sk_buff,
377							  hash));
378			break;
379		case BPF_ANC | SKF_AD_VLAN_TAG:
380		case BPF_ANC | SKF_AD_VLAN_TAG_PRESENT:
381			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff, vlan_tci) != 2);
382			BUILD_BUG_ON(VLAN_TAG_PRESENT != 0x1000);
383
384			PPC_LHZ_OFFS(r_A, r_skb, offsetof(struct sk_buff,
385							  vlan_tci));
386			if (code == (BPF_ANC | SKF_AD_VLAN_TAG)) {
387				PPC_ANDI(r_A, r_A, ~VLAN_TAG_PRESENT);
388			} else {
389				PPC_ANDI(r_A, r_A, VLAN_TAG_PRESENT);
390				PPC_SRWI(r_A, r_A, 12);
391			}
392			break;
393		case BPF_ANC | SKF_AD_QUEUE:
394			BUILD_BUG_ON(FIELD_SIZEOF(struct sk_buff,
395						  queue_mapping) != 2);
396			PPC_LHZ_OFFS(r_A, r_skb, offsetof(struct sk_buff,
397							  queue_mapping));
398			break;
399		case BPF_ANC | SKF_AD_PKTTYPE:
400			PPC_LBZ_OFFS(r_A, r_skb, PKT_TYPE_OFFSET());
401			PPC_ANDI(r_A, r_A, PKT_TYPE_MAX);
402			PPC_SRWI(r_A, r_A, 5);
403			break;
404		case BPF_ANC | SKF_AD_CPU:
405			PPC_BPF_LOAD_CPU(r_A);
406			break;
407			/*** Absolute loads from packet header/data ***/
408		case BPF_LD | BPF_W | BPF_ABS:
409			func = CHOOSE_LOAD_FUNC(K, sk_load_word);
410			goto common_load;
411		case BPF_LD | BPF_H | BPF_ABS:
412			func = CHOOSE_LOAD_FUNC(K, sk_load_half);
413			goto common_load;
414		case BPF_LD | BPF_B | BPF_ABS:
415			func = CHOOSE_LOAD_FUNC(K, sk_load_byte);
416		common_load:
417			/* Load from [K]. */
418			ctx->seen |= SEEN_DATAREF;
419			PPC_FUNC_ADDR(r_scratch1, func);
420			PPC_MTLR(r_scratch1);
421			PPC_LI32(r_addr, K);
422			PPC_BLRL();
423			/*
424			 * Helper returns 'lt' condition on error, and an
425			 * appropriate return value in r3
426			 */
427			PPC_BCC(COND_LT, exit_addr);
428			break;
429
430			/*** Indirect loads from packet header/data ***/
431		case BPF_LD | BPF_W | BPF_IND:
432			func = sk_load_word;
433			goto common_load_ind;
434		case BPF_LD | BPF_H | BPF_IND:
435			func = sk_load_half;
436			goto common_load_ind;
437		case BPF_LD | BPF_B | BPF_IND:
438			func = sk_load_byte;
439		common_load_ind:
440			/*
441			 * Load from [X + K].  Negative offsets are tested for
442			 * in the helper functions.
443			 */
444			ctx->seen |= SEEN_DATAREF | SEEN_XREG;
445			PPC_FUNC_ADDR(r_scratch1, func);
446			PPC_MTLR(r_scratch1);
447			PPC_ADDI(r_addr, r_X, IMM_L(K));
448			if (K >= 32768)
449				PPC_ADDIS(r_addr, r_addr, IMM_HA(K));
450			PPC_BLRL();
451			/* If error, cr0.LT set */
452			PPC_BCC(COND_LT, exit_addr);
453			break;
454
455		case BPF_LDX | BPF_B | BPF_MSH:
456			func = CHOOSE_LOAD_FUNC(K, sk_load_byte_msh);
457			goto common_load;
458			break;
459
460			/*** Jump and branches ***/
461		case BPF_JMP | BPF_JA:
462			if (K != 0)
463				PPC_JMP(addrs[i + 1 + K]);
464			break;
465
466		case BPF_JMP | BPF_JGT | BPF_K:
467		case BPF_JMP | BPF_JGT | BPF_X:
468			true_cond = COND_GT;
469			goto cond_branch;
470		case BPF_JMP | BPF_JGE | BPF_K:
471		case BPF_JMP | BPF_JGE | BPF_X:
472			true_cond = COND_GE;
473			goto cond_branch;
474		case BPF_JMP | BPF_JEQ | BPF_K:
475		case BPF_JMP | BPF_JEQ | BPF_X:
476			true_cond = COND_EQ;
477			goto cond_branch;
478		case BPF_JMP | BPF_JSET | BPF_K:
479		case BPF_JMP | BPF_JSET | BPF_X:
480			true_cond = COND_NE;
481			/* Fall through */
482		cond_branch:
483			/* same targets, can avoid doing the test :) */
484			if (filter[i].jt == filter[i].jf) {
485				if (filter[i].jt > 0)
486					PPC_JMP(addrs[i + 1 + filter[i].jt]);
487				break;
488			}
489
490			switch (code) {
491			case BPF_JMP | BPF_JGT | BPF_X:
492			case BPF_JMP | BPF_JGE | BPF_X:
493			case BPF_JMP | BPF_JEQ | BPF_X:
494				ctx->seen |= SEEN_XREG;
495				PPC_CMPLW(r_A, r_X);
496				break;
497			case BPF_JMP | BPF_JSET | BPF_X:
498				ctx->seen |= SEEN_XREG;
499				PPC_AND_DOT(r_scratch1, r_A, r_X);
500				break;
501			case BPF_JMP | BPF_JEQ | BPF_K:
502			case BPF_JMP | BPF_JGT | BPF_K:
503			case BPF_JMP | BPF_JGE | BPF_K:
504				if (K < 32768)
505					PPC_CMPLWI(r_A, K);
506				else {
507					PPC_LI32(r_scratch1, K);
508					PPC_CMPLW(r_A, r_scratch1);
509				}
510				break;
511			case BPF_JMP | BPF_JSET | BPF_K:
512				if (K < 32768)
513					/* PPC_ANDI is /only/ dot-form */
514					PPC_ANDI(r_scratch1, r_A, K);
515				else {
516					PPC_LI32(r_scratch1, K);
517					PPC_AND_DOT(r_scratch1, r_A,
518						    r_scratch1);
519				}
520				break;
521			}
522			/* Sometimes branches are constructed "backward", with
523			 * the false path being the branch and true path being
524			 * a fallthrough to the next instruction.
525			 */
526			if (filter[i].jt == 0)
527				/* Swap the sense of the branch */
528				PPC_BCC(true_cond ^ COND_CMP_TRUE,
529					addrs[i + 1 + filter[i].jf]);
530			else {
531				PPC_BCC(true_cond, addrs[i + 1 + filter[i].jt]);
532				if (filter[i].jf != 0)
533					PPC_JMP(addrs[i + 1 + filter[i].jf]);
534			}
535			break;
536		default:
537			/* The filter contains something cruel & unusual.
538			 * We don't handle it, but also there shouldn't be
539			 * anything missing from our list.
540			 */
541			if (printk_ratelimit())
542				pr_err("BPF filter opcode %04x (@%d) unsupported\n",
543				       filter[i].code, i);
544			return -ENOTSUPP;
545		}
546
547	}
548	/* Set end-of-body-code address for exit. */
549	addrs[i] = ctx->idx * 4;
550
551	return 0;
552}
553
554void bpf_jit_compile(struct bpf_prog *fp)
555{
556	unsigned int proglen;
557	unsigned int alloclen;
558	u32 *image = NULL;
559	u32 *code_base;
560	unsigned int *addrs;
561	struct codegen_context cgctx;
562	int pass;
563	int flen = fp->len;
564
565	if (!bpf_jit_enable)
566		return;
567
568	addrs = kzalloc((flen+1) * sizeof(*addrs), GFP_KERNEL);
569	if (addrs == NULL)
570		return;
571
572	/*
573	 * There are multiple assembly passes as the generated code will change
574	 * size as it settles down, figuring out the max branch offsets/exit
575	 * paths required.
576	 *
577	 * The range of standard conditional branches is +/- 32Kbytes.	Since
578	 * BPF_MAXINSNS = 4096, we can only jump from (worst case) start to
579	 * finish with 8 bytes/instruction.  Not feasible, so long jumps are
580	 * used, distinct from short branches.
581	 *
582	 * Current:
583	 *
584	 * For now, both branch types assemble to 2 words (short branches padded
585	 * with a NOP); this is less efficient, but assembly will always complete
586	 * after exactly 3 passes:
587	 *
588	 * First pass: No code buffer; Program is "faux-generated" -- no code
589	 * emitted but maximum size of output determined (and addrs[] filled
590	 * in).	 Also, we note whether we use M[], whether we use skb data, etc.
591	 * All generation choices assumed to be 'worst-case', e.g. branches all
592	 * far (2 instructions), return path code reduction not available, etc.
593	 *
594	 * Second pass: Code buffer allocated with size determined previously.
595	 * Prologue generated to support features we have seen used.  Exit paths
596	 * determined and addrs[] is filled in again, as code may be slightly
597	 * smaller as a result.
598	 *
599	 * Third pass: Code generated 'for real', and branch destinations
600	 * determined from now-accurate addrs[] map.
601	 *
602	 * Ideal:
603	 *
604	 * If we optimise this, near branches will be shorter.	On the
605	 * first assembly pass, we should err on the side of caution and
606	 * generate the biggest code.  On subsequent passes, branches will be
607	 * generated short or long and code size will reduce.  With smaller
608	 * code, more branches may fall into the short category, and code will
609	 * reduce more.
610	 *
611	 * Finally, if we see one pass generate code the same size as the
612	 * previous pass we have converged and should now generate code for
613	 * real.  Allocating at the end will also save the memory that would
614	 * otherwise be wasted by the (small) current code shrinkage.
615	 * Preferably, we should do a small number of passes (e.g. 5) and if we
616	 * haven't converged by then, get impatient and force code to generate
617	 * as-is, even if the odd branch would be left long.  The chances of a
618	 * long jump are tiny with all but the most enormous of BPF filter
619	 * inputs, so we should usually converge on the third pass.
620	 */
621
622	cgctx.idx = 0;
623	cgctx.seen = 0;
624	cgctx.pc_ret0 = -1;
625	/* Scouting faux-generate pass 0 */
626	if (bpf_jit_build_body(fp, 0, &cgctx, addrs))
627		/* We hit something illegal or unsupported. */
628		goto out;
629
630	/*
631	 * Pretend to build prologue, given the features we've seen.  This will
632	 * update ctgtx.idx as it pretends to output instructions, then we can
633	 * calculate total size from idx.
634	 */
635	bpf_jit_build_prologue(fp, 0, &cgctx);
636	bpf_jit_build_epilogue(0, &cgctx);
637
638	proglen = cgctx.idx * 4;
639	alloclen = proglen + FUNCTION_DESCR_SIZE;
640	image = module_alloc(alloclen);
641	if (!image)
642		goto out;
643
644	code_base = image + (FUNCTION_DESCR_SIZE/4);
645
646	/* Code generation passes 1-2 */
647	for (pass = 1; pass < 3; pass++) {
648		/* Now build the prologue, body code & epilogue for real. */
649		cgctx.idx = 0;
650		bpf_jit_build_prologue(fp, code_base, &cgctx);
651		bpf_jit_build_body(fp, code_base, &cgctx, addrs);
652		bpf_jit_build_epilogue(code_base, &cgctx);
653
654		if (bpf_jit_enable > 1)
655			pr_info("Pass %d: shrink = %d, seen = 0x%x\n", pass,
656				proglen - (cgctx.idx * 4), cgctx.seen);
657	}
658
659	if (bpf_jit_enable > 1)
660		/* Note that we output the base address of the code_base
661		 * rather than image, since opcodes are in code_base.
662		 */
663		bpf_jit_dump(flen, proglen, pass, code_base);
664
665	if (image) {
666		bpf_flush_icache(code_base, code_base + (proglen/4));
667#ifdef CONFIG_PPC64
668		/* Function descriptor nastiness: Address + TOC */
669		((u64 *)image)[0] = (u64)code_base;
670		((u64 *)image)[1] = local_paca->kernel_toc;
671#endif
672		fp->bpf_func = (void *)image;
673		fp->jited = true;
674	}
675out:
676	kfree(addrs);
677	return;
678}
679
680void bpf_jit_free(struct bpf_prog *fp)
681{
682	if (fp->jited)
683		module_memfree(fp->bpf_func);
684
685	bpf_prog_unlock_free(fp);
686}
687