anolis: scripts: ftrace_init optimization about sorting

OpenAnolis Bug Tracker: 0000558

When the kernel starts, the initialization of fTrace takes
up a portion of the time (approximately 6~8ms) to sort mcount
addresses. We can save this time by moving mcount-sorting to
compile time.

Signed-off-by: Yinan Liu <[email protected]>
Reviewed-by: Shanpei Chen <[email protected]>

Author: Yinan Liu

kernel/trace/ftrace.c

@@ -5550,8 +5550,10 @@ static int ftrace_process_locs(struct module *mod,
 	if (!count)
 		return 0;
 
-	sort(start, count, sizeof(*start),
-	     ftrace_cmp_ips, NULL);
+	if (mod) {
+		sort(start, count, sizeof(*start),
+		     ftrace_cmp_ips, NULL);
+	}
 
 	start_pg = ftrace_allocate_pages(count);
 	if (!start_pg)

scripts/Makefile

@@ -22,14 +22,14 @@ hostprogs-$(CONFIG_SYSTEM_TRUSTED_KEYRING) += extract-cert
 hostprogs-$(CONFIG_SYSTEM_EXTRA_CERTIFICATE) += insert-sys-cert
 
 HOSTCFLAGS_sorttable.o = -I$(srctree)/tools/include
+HOSTLDLIBS_sorttable = -lpthread
 HOSTCFLAGS_asn1_compiler.o = -I$(srctree)/include
 HOSTLDLIBS_sign-file = -lcrypto
 HOSTLDLIBS_extract-cert = -lcrypto
 
 ifdef CONFIG_UNWINDER_ORC
 HOSTCFLAGS_sorttable.o += -I$(srctree)/tools/objtool/arch/$(SRCARCH)/include
 HOSTCFLAGS_sorttable.o += -DUNWINDER_ORC_ENABLED
-HOSTLDLIBS_sorttable = -lpthread
 endif
 
 always		:= $(hostprogs-y) $(hostprogs-m)

scripts/link-vmlinux.sh

@@ -295,6 +295,9 @@ fi
 info LD vmlinux
 vmlinux_link "${kallsymso}" vmlinux
 
+info SYSMAP System.map
+mksysmap vmlinux System.map
+
 if [ -n "${CONFIG_BUILDTIME_TABLE_SORT}" ]; then
 	info SORTTAB vmlinux
 	if ! sorttable vmlinux; then
@@ -303,9 +306,6 @@ if [ -n "${CONFIG_BUILDTIME_TABLE_SORT}" ]; then
 	fi
 fi
 
-info SYSMAP System.map
-mksysmap vmlinux System.map
-
 # step a (see comment above)
 if [ -n "${CONFIG_KALLSYMS}" ]; then
 	mksysmap ${kallsyms_vmlinux} .tmp_System.map

scripts/sorttable.c

@@ -30,6 +30,8 @@
 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>
+#include <errno.h>
+#include <pthread.h>
 
 #include <tools/be_byteshift.h>
 #include <tools/le_byteshift.h>

scripts/sorttable.h

@@ -21,6 +21,9 @@
 
 #undef extable_ent_size
 #undef compare_extable
+#undef get_mcount_loc
+#undef sort_mcount_loc
+#undef elf_mcount_loc
 #undef do_sort
 #undef Elf_Addr
 #undef Elf_Ehdr
@@ -43,6 +46,9 @@
 #ifdef SORTTABLE_64
 # define extable_ent_size	16
 # define compare_extable	compare_extable_64
+# define get_mcount_loc		get_mcount_loc_64
+# define sort_mcount_loc	sort_mcount_loc_64
+# define elf_mcount_loc		elf_mcount_loc_64
 # define do_sort		do_sort_64
 # define Elf_Addr		Elf64_Addr
 # define Elf_Ehdr		Elf64_Ehdr
@@ -64,6 +70,9 @@
 #else
 # define extable_ent_size	8
 # define compare_extable	compare_extable_32
+# define get_mcount_loc		get_mcount_loc_32
+# define sort_mcount_loc	sort_mcount_loc_32
+# define elf_mcount_loc		elf_mcount_loc_32
 # define do_sort		do_sort_32
 # define Elf_Addr		Elf32_Addr
 # define Elf_Ehdr		Elf32_Ehdr
@@ -86,8 +95,6 @@
 
 #if defined(SORTTABLE_64) && defined(UNWINDER_ORC_ENABLED)
 /* ORC unwinder only support X86_64 */
-#include <errno.h>
-#include <pthread.h>
 #include <asm/orc_types.h>
 
 #define ERRSTR_MAXSZ	256
@@ -193,6 +200,62 @@ static int compare_extable(const void *a, const void *b)
 		return 1;
 	return 0;
 }
+struct elf_mcount_loc {
+	Elf_Ehdr *ehdr;
+	Elf_Shdr *init_data_sec;
+	uint_t start_mcount_loc;
+	uint_t stop_mcount_loc;
+};
+
+/* Sort the addresses stored between __start_mcount_loc to __stop_mcount_loc in vmlinux */
+static void *sort_mcount_loc(void *arg)
+{
+	struct elf_mcount_loc *emloc = (struct elf_mcount_loc *)arg;
+	uint_t offset = emloc->start_mcount_loc - _r(&(emloc->init_data_sec)->sh_addr)
+					+ _r(&(emloc->init_data_sec)->sh_offset);
+	uint_t count = emloc->stop_mcount_loc - emloc->start_mcount_loc;
+	unsigned char *start_loc = (void *)emloc->ehdr + offset;
+
+	qsort(start_loc, count/sizeof(uint_t), sizeof(uint_t), compare_extable);
+	return NULL;
+}
+
+/* Get the address of __start_mcount_loc and __stop_mcount_loc in System.map */
+static void get_mcount_loc(uint_t *_start, uint_t *_stop)
+{
+	FILE *file_start, *file_stop;
+	char start_buff[20];
+	char stop_buff[20];
+	int len = 0;
+
+	file_start = popen(" grep start_mcount System.map | awk '{print $1}' ", "r");
+	if (!file_start) {
+		fprintf(stderr, "get start_mcount_loc error!");
+		return;
+	}
+
+	file_stop = popen(" grep stop_mcount System.map | awk '{print $1}' ", "r");
+	if (!file_stop) {
+		fprintf(stderr, "get stop_mcount_loc error!");
+		pclose(file_start);
+		return;
+	}
+
+	while (fgets(start_buff, sizeof(start_buff), file_start) != NULL) {
+		len = strlen(start_buff);
+		start_buff[len - 1] = '\0';
+	}
+	*_start = strtoul(start_buff, NULL, 16);
+
+	while (fgets(stop_buff, sizeof(stop_buff), file_stop) != NULL) {
+		len = strlen(stop_buff);
+		stop_buff[len - 1] = '\0';
+	}
+	*_stop = strtoul(stop_buff, NULL, 16);
+
+	pclose(file_start);
+	pclose(file_stop);
+}
 
 static int do_sort(Elf_Ehdr *ehdr,
 		   char const *const fname,
@@ -219,6 +282,10 @@ static int do_sort(Elf_Ehdr *ehdr,
 	int idx;
 	unsigned int shnum;
 	unsigned int shstrndx;
+	struct elf_mcount_loc mstruct;
+	uint_t _start_mcount_loc = 0;
+	uint_t _stop_mcount_loc = 0;
+	pthread_t mcount_sort_thread;
 #if defined(SORTTABLE_64) && defined(UNWINDER_ORC_ENABLED)
 	unsigned int orc_ip_size = 0;
 	unsigned int orc_size = 0;
@@ -255,6 +322,14 @@ static int do_sort(Elf_Ehdr *ehdr,
 			symtab_shndx = (Elf32_Word *)((const char *)ehdr +
 						      _r(&s->sh_offset));
 
+		/* locate the .init.data section in vmlinux */
+		if (!strcmp(secstrings + idx, ".init.data")) {
+			get_mcount_loc(&_start_mcount_loc, &_stop_mcount_loc);
+			mstruct.ehdr = ehdr;
+			mstruct.init_data_sec = s;
+			mstruct.start_mcount_loc = _start_mcount_loc;
+			mstruct.stop_mcount_loc = _stop_mcount_loc;
+		}
 #if defined(SORTTABLE_64) && defined(UNWINDER_ORC_ENABLED)
 		/* locate the ORC unwind tables */
 		if (!strcmp(secstrings + idx, ".orc_unwind_ip")) {
@@ -296,6 +371,21 @@ static int do_sort(Elf_Ehdr *ehdr,
 		goto out;
 	}
 #endif
+	if (!mstruct.init_data_sec || !_start_mcount_loc || !_stop_mcount_loc) {
+		fprintf(stderr,
+			"incomplete mcount's sort in file: %s\n",
+			fname);
+		goto out;
+	}
+
+	/* create thread to sort mcount_loc concurrently */
+	if (pthread_create(&mcount_sort_thread, NULL, &sort_mcount_loc, &mstruct)) {
+		fprintf(stderr,
+			"pthread_create mcount_sort_thread failed '%s': %s\n",
+			strerror(errno), fname);
+		goto out;
+	}
+
 	if (!extab_sec) {
 		fprintf(stderr,	"no __ex_table in file: %s\n", fname);
 		goto out;
@@ -366,17 +456,32 @@ static int do_sort(Elf_Ehdr *ehdr,
 		void *retval = NULL;
 		/* wait for ORC tables sort done */
 		rc = pthread_join(orc_sort_thread, &retval);
-		if (rc)
+		if (rc) {
 			fprintf(stderr,
 				"pthread_join failed '%s': %s\n",
 				strerror(errno), fname);
-		else if (retval) {
+		} else if (retval) {
 			rc = -1;
 			fprintf(stderr,
 				"failed to sort ORC tables '%s': %s\n",
 				(char *)retval, fname);
 		}
 	}
 #endif
+	if (mcount_sort_thread) {
+		void *retval = NULL;
+		/* wait for mcount sort done */
+		rc = pthread_join(mcount_sort_thread, &retval);
+		if (rc) {
+			fprintf(stderr,
+				"pthread_join failed '%s': %s\n",
+				strerror(errno), fname);
+		} else if (retval) {
+			rc = -1;
+			fprintf(stderr,
+				"failed to sort mcount '%s': %s\n",
+				(char *)retval, fname);
+		}
+	}
 	return rc;
 }