s390/bpf: take advantage of stack_depth tracking

Make use of the "stack_depth" tracking feature introduced with
commit 8726679a0fa31 ("bpf: teach verifier to track stack depth") for the
s390 JIT, so that stack usage can be reduced.

Signed-off-by: Michael Holzheu <holzheu@linux.vnet.ibm.com>
Signed-off-by: Heiko Carstens <heiko.carstens@de.ibm.com>

arch/s390/net/bpf_jit.h

@@ -52,10 +52,13 @@ extern u8 sk_load_word[], sk_load_half[], sk_load_byte[];
  *
  * We get 160 bytes stack space from calling function, but only use
  * 12 * 8 byte for old backchain, r15..r6, and tail_call_cnt.
+ *
+ * The stack size used by the BPF program ("BPF stack" above) is passed
+ * via "aux->stack_depth".
  */
-#define STK_SPACE	(MAX_BPF_STACK + 8 + 8 + 4 + 4 + 160)
+#define STK_SPACE_ADD (8 + 8 + 4 + 4 + 160)
 #define STK_160_UNUSED	(160 - 12 * 8)
-#define STK_OFF		(STK_SPACE - STK_160_UNUSED)
+#define STK_OFF		(STK_SPACE_ADD - STK_160_UNUSED)
 #define STK_OFF_TMP	160	/* Offset of tmp buffer on stack */
 #define STK_OFF_HLEN	168	/* Offset of SKB header length on stack */
 #define STK_OFF_SKBP	176	/* Offset of SKB pointer on stack */

arch/s390/net/bpf_jit_comp.c

@@ -319,12 +319,12 @@ static void save_regs(struct bpf_jit *jit, u32 rs, u32 re)
 /*
  * Restore registers from "rs" (register start) to "re" (register end) on stack
  */
-static void restore_regs(struct bpf_jit *jit, u32 rs, u32 re)
+static void restore_regs(struct bpf_jit *jit, u32 rs, u32 re, u32 stack_depth)
 {
 	u32 off = STK_OFF_R6 + (rs - 6) * 8;
 
 	if (jit->seen & SEEN_STACK)
-		off += STK_OFF;
+		off += STK_OFF + stack_depth;
 
 	if (rs == re)
 		/* lg %rs,off(%r15) */
@@ -368,7 +368,7 @@ static int get_end(struct bpf_jit *jit, int start)
  * Save and restore clobbered registers (6-15) on stack.
  * We save/restore registers in chunks with gap >= 2 registers.
  */
-static void save_restore_regs(struct bpf_jit *jit, int op)
+static void save_restore_regs(struct bpf_jit *jit, int op, u32 stack_depth)
 {
 
 	int re = 6, rs;
@@ -381,7 +381,7 @@ static void save_restore_regs(struct bpf_jit *jit, int op)
 		if (op == REGS_SAVE)
 			save_regs(jit, rs, re);
 		else
-			restore_regs(jit, rs, re);
+			restore_regs(jit, rs, re, stack_depth);
 		re++;
 	} while (re <= 15);
 }
@@ -413,7 +413,7 @@ static void emit_load_skb_data_hlen(struct bpf_jit *jit)
  * Save registers and create stack frame if necessary.
  * See stack frame layout desription in "bpf_jit.h"!
  */
-static void bpf_jit_prologue(struct bpf_jit *jit)
+static void bpf_jit_prologue(struct bpf_jit *jit, u32 stack_depth)
 {
 	if (jit->seen & SEEN_TAIL_CALL) {
 		/* xc STK_OFF_TCCNT(4,%r15),STK_OFF_TCCNT(%r15) */
@@ -426,7 +426,7 @@ static void bpf_jit_prologue(struct bpf_jit *jit)
 	/* Tail calls have to skip above initialization */
 	jit->tail_call_start = jit->prg;
 	/* Save registers */
-	save_restore_regs(jit, REGS_SAVE);
+	save_restore_regs(jit, REGS_SAVE, stack_depth);
 	/* Setup literal pool */
 	if (jit->seen & SEEN_LITERAL) {
 		/* basr %r13,0 */
@@ -441,7 +441,7 @@ static void bpf_jit_prologue(struct bpf_jit *jit)
 		/* la %bfp,STK_160_UNUSED(%r15) (BPF frame pointer) */
 		EMIT4_DISP(0x41000000, BPF_REG_FP, REG_15, STK_160_UNUSED);
 		/* aghi %r15,-STK_OFF */
-		EMIT4_IMM(0xa70b0000, REG_15, -STK_OFF);
+		EMIT4_IMM(0xa70b0000, REG_15, -(STK_OFF + stack_depth));
 		if (jit->seen & SEEN_FUNC)
 			/* stg %w1,152(%r15) (backchain) */
 			EMIT6_DISP_LH(0xe3000000, 0x0024, REG_W1, REG_0,
@@ -458,7 +458,7 @@ static void bpf_jit_prologue(struct bpf_jit *jit)
 /*
  * Function epilogue
  */
-static void bpf_jit_epilogue(struct bpf_jit *jit)
+static void bpf_jit_epilogue(struct bpf_jit *jit, u32 stack_depth)
 {
 	/* Return 0 */
 	if (jit->seen & SEEN_RET0) {
@@ -470,7 +470,7 @@ static void bpf_jit_epilogue(struct bpf_jit *jit)
 	/* Load exit code: lgr %r2,%b0 */
 	EMIT4(0xb9040000, REG_2, BPF_REG_0);
 	/* Restore registers */
-	save_restore_regs(jit, REGS_RESTORE);
+	save_restore_regs(jit, REGS_RESTORE, stack_depth);
 	/* br %r14 */
 	_EMIT2(0x07fe);
 }
@@ -1018,7 +1018,7 @@ static noinline int bpf_jit_insn(struct bpf_jit *jit, struct bpf_prog *fp, int i
 		 */
 
 		if (jit->seen & SEEN_STACK)
-			off = STK_OFF_TCCNT + STK_OFF;
+			off = STK_OFF_TCCNT + STK_OFF + fp->aux->stack_depth;
 		else
 			off = STK_OFF_TCCNT;
 		/* lhi %w0,1 */
@@ -1046,7 +1046,7 @@ static noinline int bpf_jit_insn(struct bpf_jit *jit, struct bpf_prog *fp, int i
 		/*
 		 * Restore registers before calling function
 		 */
-		save_restore_regs(jit, REGS_RESTORE);
+		save_restore_regs(jit, REGS_RESTORE, fp->aux->stack_depth);
 
 		/*
 		 * goto *(prog->bpf_func + tail_call_start);
@@ -1272,7 +1272,7 @@ static int bpf_jit_prog(struct bpf_jit *jit, struct bpf_prog *fp)
 	jit->lit = jit->lit_start;
 	jit->prg = 0;
 
-	bpf_jit_prologue(jit);
+	bpf_jit_prologue(jit, fp->aux->stack_depth);
 	for (i = 0; i < fp->len; i += insn_count) {
 		insn_count = bpf_jit_insn(jit, fp, i);
 		if (insn_count < 0)
@@ -1280,7 +1280,7 @@ static int bpf_jit_prog(struct bpf_jit *jit, struct bpf_prog *fp)
 		/* Next instruction address */
 		jit->addrs[i + insn_count] = jit->prg;
 	}
-	bpf_jit_epilogue(jit);
+	bpf_jit_epilogue(jit, fp->aux->stack_depth);
 
 	jit->lit_start = jit->prg;
 	jit->size = jit->lit;