riscv: Fix text patching when IPI are used

For now, we use stop_machine() to patch the text and when we use IPIs for
remote icache flushes (which is emitted in patch_text_nosync()), the system
hangs.

So instead, make sure every CPU executes the stop_machine() patching
function and emit a local icache flush there.

Co-developed-by: Björn Töpel <bjorn@rivosinc.com>
Signed-off-by: Björn Töpel <bjorn@rivosinc.com>
Signed-off-by: Alexandre Ghiti <alexghiti@rivosinc.com>
Reviewed-by: Andrea Parri <parri.andrea@gmail.com>
Link: https://lore.kernel.org/r/20240229121056.203419-3-alexghiti@rivosinc.com
Signed-off-by: Palmer Dabbelt <palmer@rivosinc.com>

arch/riscv/include/asm/patch.h

@@ -6,6 +6,7 @@
 #ifndef _ASM_RISCV_PATCH_H
 #define _ASM_RISCV_PATCH_H
 
+int patch_insn_write(void *addr, const void *insn, size_t len);
 int patch_text_nosync(void *addr, const void *insns, size_t len);
 int patch_text_set_nosync(void *addr, u8 c, size_t len);
 int patch_text(void *addr, u32 *insns, int ninsns);

arch/riscv/kernel/ftrace.c

@@ -8,6 +8,7 @@
 #include <linux/ftrace.h>
 #include <linux/uaccess.h>
 #include <linux/memory.h>
+#include <linux/stop_machine.h>
 #include <asm/cacheflush.h>
 #include <asm/patch.h>
 
@@ -75,8 +76,7 @@ static int __ftrace_modify_call(unsigned long hook_pos, unsigned long target,
 		make_call_t0(hook_pos, target, call);
 
 	/* Replace the auipc-jalr pair at once. Return -EPERM on write error. */
-	if (patch_text_nosync
+	if (patch_insn_write((void *)hook_pos, enable ? call : nops, MCOUNT_INSN_SIZE))
-	    ((void *)hook_pos, enable ? call : nops, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;
@@ -88,7 +88,7 @@ int ftrace_make_call(struct dyn_ftrace *rec, unsigned long addr)
 
 	make_call_t0(rec->ip, addr, call);
 
-	if (patch_text_nosync((void *)rec->ip, call, MCOUNT_INSN_SIZE))
+	if (patch_insn_write((void *)rec->ip, call, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;
@@ -99,7 +99,7 @@ int ftrace_make_nop(struct module *mod, struct dyn_ftrace *rec,
 {
 	unsigned int nops[2] = {NOP4, NOP4};
 
-	if (patch_text_nosync((void *)rec->ip, nops, MCOUNT_INSN_SIZE))
+	if (patch_insn_write((void *)rec->ip, nops, MCOUNT_INSN_SIZE))
 		return -EPERM;
 
 	return 0;
@@ -134,6 +134,42 @@ int ftrace_update_ftrace_func(ftrace_func_t func)
 
 	return ret;
 }
+
+struct ftrace_modify_param {
+	int command;
+	atomic_t cpu_count;
+};
+
+static int __ftrace_modify_code(void *data)
+{
+	struct ftrace_modify_param *param = data;
+
+	if (atomic_inc_return(&param->cpu_count) == num_online_cpus()) {
+		ftrace_modify_all_code(param->command);
+		/*
+		 * Make sure the patching store is effective *before* we
+		 * increment the counter which releases all waiting CPUs
+		 * by using the release variant of atomic increment. The
+		 * release pairs with the call to local_flush_icache_all()
+		 * on the waiting CPU.
+		 */
+		atomic_inc_return_release(&param->cpu_count);
+	} else {
+		while (atomic_read(&param->cpu_count) <= num_online_cpus())
+			cpu_relax();
+	}
+
+	local_flush_icache_all();
+
+	return 0;
+}
+
+void arch_ftrace_update_code(int command)
+{
+	struct ftrace_modify_param param = { command, ATOMIC_INIT(0) };
+
+	stop_machine(__ftrace_modify_code, &param, cpu_online_mask);
+}
 #endif
 
 #ifdef CONFIG_DYNAMIC_FTRACE_WITH_REGS

arch/riscv/kernel/patch.c

@@ -188,7 +188,7 @@ int patch_text_set_nosync(void *addr, u8 c, size_t len)
 }
 NOKPROBE_SYMBOL(patch_text_set_nosync);
 
-static int patch_insn_write(void *addr, const void *insn, size_t len)
+int patch_insn_write(void *addr, const void *insn, size_t len)
 {
 	size_t patched = 0;
 	size_t size;
@@ -232,15 +232,23 @@ static int patch_text_cb(void *data)
 	if (atomic_inc_return(&patch->cpu_count) == num_online_cpus()) {
 		for (i = 0; ret == 0 && i < patch->ninsns; i++) {
 			len = GET_INSN_LENGTH(patch->insns[i]);
-			ret = patch_text_nosync(patch->addr + i * len,
+			ret = patch_insn_write(patch->addr + i * len, &patch->insns[i], len);
-						&patch->insns[i], len);
 		}
-		atomic_inc(&patch->cpu_count);
+		/*
+		 * Make sure the patching store is effective *before* we
+		 * increment the counter which releases all waiting CPUs
+		 * by using the release variant of atomic increment. The
+		 * release pairs with the call to local_flush_icache_all()
+		 * on the waiting CPU.
+		 */
+		atomic_inc_return_release(&patch->cpu_count);
 	} else {
 		while (atomic_read(&patch->cpu_count) <= num_online_cpus())
 			cpu_relax();
 	}
 
+	local_flush_icache_all();
+
 	return ret;
 }
 NOKPROBE_SYMBOL(patch_text_cb);