Skip to contentSkip to conversation

[2.6.37] ftrace: C version of recordmcount

GIT PULL In review Steven Rostedt 4 participants 10 messages Oct 14, 2010 – Oct 29, 2010

Subsystems (4)
lore
Share & tools
lore

Steven Rostedt

author maintainer
#0 lore raw

Ingo,

Please pull the latest tip/perf/recordmcount tree, which can be found at:

git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-2.6-trace.git
tip/perf/recordmcount

John Reiser (1):
ftrace: Add C version of recordmcount compile time code

Steven Rostedt (2):
ftrace/x86: Add support for C version of recordmcount
ftrace: Remove duplicate code for 64 and 32 bit in recordmcount.c

----

 Makefile               |    6 +
 arch/x86/Kconfig       |    1 +
 kernel/trace/Kconfig   |    5 +
 scripts/Makefile       |    1 +
 scripts/Makefile.build |    4 +
 scripts/recordmcount.c |  345 +++++++++++++++++++++++++++++++++++++++++++++
 scripts/recordmcount.h |  366 ++++++++++++++++++++++++++++++++++++++++++++++++
 7 files changed, 728 insertions(+), 0 deletions(-)
author maintainer
#1 lore raw

PATCH 1/3 ftrace: Add C version of recordmcount compile time code

From: John Reiser <[email protected]>

Currently, the mcount callers are found with a perl script that does an objdump on every file in the kernel. This is a C version of that same code which should increase the performance time of compiling the kernel with dynamic ftrace enabled.

[ Updated the code to include .text.unlikely section as well as
changing the format to follow Linux coding style. ]

---
 scripts/recordmcount.c |  885 ++++++++++++++++++++++++++++++++++++++++++++++++
 1 files changed, 885 insertions(+), 0 deletions(-)
 create mode 100644 scripts/recordmcount.c
scripts/recordmcount.c new file +885 -0
new file mode 100644
index 0000000..34f32be
@@ -0,0 +1,885 @@
+/*
+ * recordmcount.c: construct a table of the locations of calls to 'mcount'
+ * so that ftrace can find them quickly.
+ * Copyright 2009 John F. Reiser <[email protected]>.  All rights reserved.
+ * Licensed under the GNU General Public License, version 2 (GPLv2).
+ *
+ * Restructured to fit Linux format, as well as other updates:
+ *  Copyright 2010 Steven Rostedt <[email protected]>, Red Hat Inc.
+ */
+
+/*
+ * Strategy: alter the .o file in-place.
+ *
+ * Append a new STRTAB that has the new section names, followed by a new array
+ * ElfXX_Shdr[] that has the new section headers, followed by the section
+ * contents for __mcount_loc and its relocations.  The old shstrtab strings,
+ * and the old ElfXX_Shdr[] array, remain as "garbage" (commonly, a couple
+ * kilobytes.)  Subsequent processing by /bin/ld (or the kernel module loader)
+ * will ignore the garbage regions, because they are not designated by the
+ * new .e_shoff nor the new ElfXX_Shdr[].  [In order to remove the garbage,
+ * then use "ld -r" to create a new file that omits the garbage.]
+ */
+
+#include <sys/types.h>
+#include <sys/mman.h>
+#include <sys/stat.h>
+#include <elf.h>
+#include <fcntl.h>
+#include <setjmp.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <unistd.h>
+
+static int fd_map;	/* File descriptor for file being modified. */
+static int mmap_failed; /* Boolean flag. */
+static void *ehdr_curr; /* current ElfXX_Ehdr *  for resource cleanup */
+static char gpfx;	/* prefix for global symbol name (sometimes '_') */
+static struct stat sb;	/* Remember .st_size, etc. */
+static jmp_buf jmpenv;	/* setjmp/longjmp per-file error escape */
+
+/* setjmp() return values */
+enum {
+	SJ_SETJMP = 0,  /* hardwired first return */
+	SJ_FAIL,
+	SJ_SUCCEED
+};
+
+/* Per-file resource cleanup when multiple files. */
+static void
+cleanup(void)
+{
+	if (!mmap_failed)
+		munmap(ehdr_curr, sb.st_size);
+	else
+		free(ehdr_curr);
+	close(fd_map);
+}
+
+static void __attribute__((noreturn))
+fail_file(void)
+{
+	cleanup();
+	longjmp(jmpenv, SJ_FAIL);
+}
+
+static void __attribute__((noreturn))
+succeed_file(void)
+{
+	cleanup();
+	longjmp(jmpenv, SJ_SUCCEED);
+}
+
+/* ulseek, uread, ...:  Check return value for errors. */
+
+static off_t
+ulseek(int const fd, off_t const offset, int const whence)
+{
+	off_t const w = lseek(fd, offset, whence);
+	if ((off_t)-1 == w) {
+		perror("lseek");
+		fail_file();
+	}
+	return w;
+}
+
+static size_t
+uread(int const fd, void *const buf, size_t const count)
+{
+	size_t const n = read(fd, buf, count);
+	if (n != count) {
+		perror("read");
+		fail_file();
+	}
+	return n;
+}
+
+static size_t
+uwrite(int const fd, void const *const buf, size_t const count)
+{
+	size_t const n = write(fd, buf, count);
+	if (n != count) {
+		perror("write");
+		fail_file();
+	}
+	return n;
+}
+
+static void *
+umalloc(size_t size)
+{
+	void *const addr = malloc(size);
+	if (0 == addr) {
+		fprintf(stderr, "malloc failed: %zu bytes\n", size);
+		fail_file();
+	}
+	return addr;
+}
+
+/*
+ * Get the whole file as a programming convenience in order to avoid
+ * malloc+lseek+read+free of many pieces.  If successful, then mmap
+ * avoids copying unused pieces; else just read the whole file.
+ * Open for both read and write; new info will be appended to the file.
+ * Use MAP_PRIVATE so that a few changes to the in-memory ElfXX_Ehdr
+ * do not propagate to the file until an explicit overwrite at the last.
+ * This preserves most aspects of consistency (all except .st_size)
+ * for simultaneous readers of the file while we are appending to it.
+ * However, multiple writers still are bad.  We choose not to use
+ * locking because it is expensive and the use case of kernel build
+ * makes multiple writers unlikely.
+ */
+static void *mmap_file(char const *fname)
+{
+	void *addr;
+
+	fd_map = open(fname, O_RDWR);
+	if (0 > fd_map || 0 > fstat(fd_map, &sb)) {
+		perror(fname);
+		fail_file();
+	}
+	if (!S_ISREG(sb.st_mode)) {
+		fprintf(stderr, "not a regular file: %s\n", fname);
+		fail_file();
+	}
+	addr = mmap(0, sb.st_size, PROT_READ|PROT_WRITE, MAP_PRIVATE,
+		    fd_map, 0);
+	mmap_failed = 0;
+	if (MAP_FAILED == addr) {
+		mmap_failed = 1;
+		addr = umalloc(sb.st_size);
+		uread(fd_map, addr, sb.st_size);
+	}
+	return addr;
+}
+
+/* w8rev, w8nat, ...: Handle endianness. */
+
+static uint64_t w8rev(uint64_t const x)
+{
+	return   ((0xff & (x >> (0 * 8))) << (7 * 8))
+	       | ((0xff & (x >> (1 * 8))) << (6 * 8))
+	       | ((0xff & (x >> (2 * 8))) << (5 * 8))
+	       | ((0xff & (x >> (3 * 8))) << (4 * 8))
+	       | ((0xff & (x >> (4 * 8))) << (3 * 8))
+	       | ((0xff & (x >> (5 * 8))) << (2 * 8))
+	       | ((0xff & (x >> (6 * 8))) << (1 * 8))
+	       | ((0xff & (x >> (7 * 8))) << (0 * 8));
+}
+
+static uint32_t w4rev(uint32_t const x)
+{
+	return   ((0xff & (x >> (0 * 8))) << (3 * 8))
+	       | ((0xff & (x >> (1 * 8))) << (2 * 8))
+	       | ((0xff & (x >> (2 * 8))) << (1 * 8))
+	       | ((0xff & (x >> (3 * 8))) << (0 * 8));
+}
+
+static uint32_t w2rev(uint16_t const x)
+{
+	return   ((0xff & (x >> (0 * 8))) << (1 * 8))
+	       | ((0xff & (x >> (1 * 8))) << (0 * 8));
+}
+
+static uint64_t w8nat(uint64_t const x)
+{
+	return x;
+}
+
+static uint32_t w4nat(uint32_t const x)
+{
+	return x;
+}
+
+static uint32_t w2nat(uint16_t const x)
+{
+	return x;
+}
+
+static uint64_t (*w8)(uint64_t);
+static uint32_t (*w)(uint32_t);
+static uint32_t (*w2)(uint16_t);
+
+/* Names of the sections that could contain calls to mcount. */
+static int
+is_mcounted_section_name(char const *const txtname)
+{
+	return 0 == strcmp(".text",          txtname) ||
+		0 == strcmp(".sched.text",    txtname) ||
+		0 == strcmp(".spinlock.text", txtname) ||
+		0 == strcmp(".irqentry.text", txtname) ||
+		0 == strcmp(".text.unlikely", txtname);
+}
+
+/* Append the new shstrtab, Elf32_Shdr[], __mcount_loc and its relocations. */
+static void append32(Elf32_Ehdr *const ehdr,
+		     Elf32_Shdr *const shstr,
+		     uint32_t const *const mloc0,
+		     uint32_t const *const mlocp,
+		     Elf32_Rel const *const mrel0,
+		     Elf32_Rel const *const mrelp,
+		     unsigned int const rel_entsize,
+		     unsigned int const symsec_sh_link)
+{
+	/* Begin constructing output file */
+	Elf32_Shdr mcsec;
+	char const *mc_name = (sizeof(Elf32_Rela) == rel_entsize)
+		? ".rela__mcount_loc"
+		:  ".rel__mcount_loc";
+	unsigned const old_shnum = w2(ehdr->e_shnum);
+	uint32_t const old_shoff = w(ehdr->e_shoff);
+	uint32_t const old_shstr_sh_size   = w(shstr->sh_size);
+	uint32_t const old_shstr_sh_offset = w(shstr->sh_offset);
+	uint32_t t = 1 + strlen(mc_name) + w(shstr->sh_size);
+	uint32_t new_e_shoff;
+
+	shstr->sh_size = w(t);
+	shstr->sh_offset = w(sb.st_size);
+	t += sb.st_size;
+	t += (3u & -t);  /* 4-byte align */
+	new_e_shoff = t;
+
+	/* body for new shstrtab */
+	ulseek(fd_map, sb.st_size, SEEK_SET);
+	uwrite(fd_map, old_shstr_sh_offset + (void *)ehdr, old_shstr_sh_size);
+	uwrite(fd_map, mc_name, 1 + strlen(mc_name));
+
+	/* old(modified) Elf32_Shdr table, 4-byte aligned */
+	ulseek(fd_map, t, SEEK_SET);
+	t += sizeof(Elf32_Shdr) * old_shnum;
+	uwrite(fd_map, old_shoff + (void *)ehdr,
+	       sizeof(Elf32_Shdr) * old_shnum);
+
+	/* new sections __mcount_loc and .rel__mcount_loc */
+	t += 2*sizeof(mcsec);
+	mcsec.sh_name = w((sizeof(Elf32_Rela) == rel_entsize) + strlen(".rel")
+		+ old_shstr_sh_size);
+	mcsec.sh_type = w(SHT_PROGBITS);
+	mcsec.sh_flags = w(SHF_ALLOC);
+	mcsec.sh_addr = 0;
+	mcsec.sh_offset = w(t);
+	mcsec.sh_size = w((void *)mlocp - (void *)mloc0);
+	mcsec.sh_link = 0;
+	mcsec.sh_info = 0;
+	mcsec.sh_addralign = w(4);
+	mcsec.sh_entsize = w(4);
+	uwrite(fd_map, &mcsec, sizeof(mcsec));
+
+	mcsec.sh_name = w(old_shstr_sh_size);
+	mcsec.sh_type = (sizeof(Elf32_Rela) == rel_entsize)
+		? w(SHT_RELA)
+		: w(SHT_REL);
+	mcsec.sh_flags = 0;
+	mcsec.sh_addr = 0;
+	mcsec.sh_offset = w((void *)mlocp - (void *)mloc0 + t);
+	mcsec.sh_size   = w((void *)mrelp - (void *)mrel0);
+	mcsec.sh_link = w(symsec_sh_link);
+	mcsec.sh_info = w(old_shnum);
+	mcsec.sh_addralign = w(4);
+	mcsec.sh_entsize = w(rel_entsize);
+	uwrite(fd_map, &mcsec, sizeof(mcsec));
+
+	uwrite(fd_map, mloc0, (void *)mlocp - (void *)mloc0);
+	uwrite(fd_map, mrel0, (void *)mrelp - (void *)mrel0);
+
+	ehdr->e_shoff = w(new_e_shoff);
+	ehdr->e_shnum = w2(2 + w2(ehdr->e_shnum));  /* {.rel,}__mcount_loc */
+	ulseek(fd_map, 0, SEEK_SET);
+	uwrite(fd_map, ehdr, sizeof(*ehdr));
+}
+
+/*
+ * append64 and append32 (and other analogous pairs) could be templated
+ * using C++, but the complexity is high.  (For an example, look at p_elf.h
+ * in the source for UPX, http://upx.sourceforge.net)  So: remember to make
+ * the corresponding change in the routine for the other size.
+ */
+static void append64(Elf64_Ehdr *const ehdr,
+		     Elf64_Shdr *const shstr,
+		     uint64_t const *const mloc0,
+		     uint64_t const *const mlocp,
+		     Elf64_Rel const *const mrel0,
+		     Elf64_Rel const *const mrelp,
+		     unsigned int const rel_entsize,
+		     unsigned int const symsec_sh_link)
+{
+	/* Begin constructing output file */
+	Elf64_Shdr mcsec;
+	char const *mc_name = (sizeof(Elf64_Rela) == rel_entsize)
+		? ".rela__mcount_loc"
+		:  ".rel__mcount_loc";
+	unsigned const old_shnum = w2(ehdr->e_shnum);
+	uint64_t const old_shoff = w8(ehdr->e_shoff);
+	uint64_t const old_shstr_sh_size   = w8(shstr->sh_size);
+	uint64_t const old_shstr_sh_offset = w8(shstr->sh_offset);
+	uint64_t t = 1 + strlen(mc_name) + w8(shstr->sh_size);
+	uint64_t new_e_shoff;
+
+	shstr->sh_size = w8(t);
+	shstr->sh_offset = w8(sb.st_size);
+	t += sb.st_size;
+	t += (7u & -t);  /* 8-byte align */
+	new_e_shoff = t;
+
+	/* body for new shstrtab */
+	ulseek(fd_map, sb.st_size, SEEK_SET);
+	uwrite(fd_map, old_shstr_sh_offset + (void *)ehdr, old_shstr_sh_size);
+	uwrite(fd_map, mc_name, 1 + strlen(mc_name));
+
+	/* old(modified) Elf64_Shdr table, 8-byte aligned */
+	ulseek(fd_map, t, SEEK_SET);
+	t += sizeof(Elf64_Shdr) * old_shnum;
+	uwrite(fd_map, old_shoff + (void *)ehdr,
+		sizeof(Elf64_Shdr) * old_shnum);
+
+	/* new sections __mcount_loc and .rel__mcount_loc */
+	t += 2*sizeof(mcsec);
+	mcsec.sh_name = w((sizeof(Elf64_Rela) == rel_entsize) + strlen(".rel")
+		+ old_shstr_sh_size);
+	mcsec.sh_type = w(SHT_PROGBITS);
+	mcsec.sh_flags = w8(SHF_ALLOC);
+	mcsec.sh_addr = 0;
+	mcsec.sh_offset = w8(t);
+	mcsec.sh_size = w8((void *)mlocp - (void *)mloc0);
+	mcsec.sh_link = 0;
+	mcsec.sh_info = 0;
+	mcsec.sh_addralign = w8(8);
+	mcsec.sh_entsize = w8(8);
+	uwrite(fd_map, &mcsec, sizeof(mcsec));
+
+	mcsec.sh_name = w(old_shstr_sh_size);
+	mcsec.sh_type = (sizeof(Elf64_Rela) == rel_entsize)
+		? w(SHT_RELA)
+		: w(SHT_REL);
+	mcsec.sh_flags = 0;
+	mcsec.sh_addr = 0;
+	mcsec.sh_offset = w8((void *)mlocp - (void *)mloc0 + t);
+	mcsec.sh_size   = w8((void *)mrelp - (void *)mrel0);
+	mcsec.sh_link = w(symsec_sh_link);
+	mcsec.sh_info = w(old_shnum);
+	mcsec.sh_addralign = w8(8);
+	mcsec.sh_entsize = w8(rel_entsize);
+	uwrite(fd_map, &mcsec, sizeof(mcsec));
+
+	uwrite(fd_map, mloc0, (void *)mlocp - (void *)mloc0);
+	uwrite(fd_map, mrel0, (void *)mrelp - (void *)mrel0);
+
+	ehdr->e_shoff = w8(new_e_shoff);
+	ehdr->e_shnum = w2(2 + w2(ehdr->e_shnum));  /* {.rel,}__mcount_loc */
+	ulseek(fd_map, 0, SEEK_SET);
+	uwrite(fd_map, ehdr, sizeof(*ehdr));
+}
+
+/*
+ * Look at the relocations in order to find the calls to mcount.
+ * Accumulate the section offsets that are found, and their relocation info,
+ * onto the end of the existing arrays.
+ */
+static uint32_t *sift32_rel_mcount(uint32_t *mlocp,
+				   unsigned const offbase,
+				   Elf32_Rel **const mrelpp,
+				   Elf32_Shdr const *const relhdr,
+				   Elf32_Ehdr const *const ehdr,
+				   unsigned const recsym,
+				   uint32_t const recval,
+				   unsigned const reltype)
+{
+	uint32_t *const mloc0 = mlocp;
+	Elf32_Rel *mrelp = *mrelpp;
+	Elf32_Shdr *const shdr0 = (Elf32_Shdr *)(w(ehdr->e_shoff)
+		+ (void *)ehdr);
+	unsigned const symsec_sh_link = w(relhdr->sh_link);
+	Elf32_Shdr const *const symsec = &shdr0[symsec_sh_link];
+	Elf32_Sym const *const sym0 = (Elf32_Sym const *)(w(symsec->sh_offset)
+		+ (void *)ehdr);
+
+	Elf32_Shdr const *const strsec = &shdr0[w(symsec->sh_link)];
+	char const *const str0 = (char const *)(w(strsec->sh_offset)
+		+ (void *)ehdr);
+
+	Elf32_Rel const *const rel0 = (Elf32_Rel const *)(w(relhdr->sh_offset)
+		+ (void *)ehdr);
+	unsigned rel_entsize = w(relhdr->sh_entsize);
+	unsigned const nrel = w(relhdr->sh_size) / rel_entsize;
+	Elf32_Rel const *relp = rel0;
+
+	unsigned mcountsym = 0;
+	unsigned t;
+
+	for (t = nrel; t; --t) {
+		if (!mcountsym) {
+			Elf32_Sym const *const symp =
+				&sym0[ELF32_R_SYM(w(relp->r_info))];
+
+			if (0 == strcmp((('_' == gpfx) ? "_mcount" : "mcount"),
+					&str0[w(symp->st_name)]))
+				mcountsym = ELF32_R_SYM(w(relp->r_info));
+		}
+		if (mcountsym == ELF32_R_SYM(w(relp->r_info))) {
+			uint32_t const addend = w(w(relp->r_offset) - recval);
+			mrelp->r_offset = w(offbase
+				+ ((void *)mlocp - (void *)mloc0));
+			mrelp->r_info = w(ELF32_R_INFO(recsym, reltype));
+			if (sizeof(Elf32_Rela) == rel_entsize) {
+				((Elf32_Rela *)mrelp)->r_addend = addend;
+				*mlocp++ = 0;
+			} else
+				*mlocp++ = addend;
+
+			mrelp = (Elf32_Rel *)(rel_entsize + (void *)mrelp);
+		}
+		relp = (Elf32_Rel const *)(rel_entsize + (void *)relp);
+	}
+	*mrelpp = mrelp;
+	return mlocp;
+}
+
+static uint64_t *sift64_rel_mcount(uint64_t *mlocp,
+				   unsigned const offbase,
+				   Elf64_Rel **const mrelpp,
+				   Elf64_Shdr const *const relhdr,
+				   Elf64_Ehdr const *const ehdr,
+				   unsigned const recsym,
+				   uint64_t const recval,
+				   unsigned const reltype)
+{
+	uint64_t *const mloc0 = mlocp;
+	Elf64_Rel *mrelp = *mrelpp;
+	Elf64_Shdr *const shdr0 = (Elf64_Shdr *)(w8(ehdr->e_shoff)
+		+ (void *)ehdr);
+	unsigned const symsec_sh_link = w(relhdr->sh_link);
+	Elf64_Shdr const *const symsec = &shdr0[symsec_sh_link];
+	Elf64_Sym const *const sym0 = (Elf64_Sym const *)(w8(symsec->sh_offset)
+		+ (void *)ehdr);
+
+	Elf64_Shdr const *const strsec = &shdr0[w(symsec->sh_link)];
+	char const *const str0 = (char const *)(w8(strsec->sh_offset)
+		+ (void *)ehdr);
+
+	Elf64_Rel const *const rel0 = (Elf64_Rel const *)(w8(relhdr->sh_offset)
+		+ (void *)ehdr);
+	unsigned rel_entsize = w8(relhdr->sh_entsize);
+	unsigned const nrel = w8(relhdr->sh_size) / rel_entsize;
+	Elf64_Rel const *relp = rel0;
+
+	unsigned mcountsym = 0;
+	unsigned t;
+
+	for (t = nrel; 0 != t; --t) {
+		if (!mcountsym) {
+			Elf64_Sym const *const symp =
+				&sym0[ELF64_R_SYM(w8(relp->r_info))];
+			char const *symname = &str0[w(symp->st_name)];
+
+			if ('.' == symname[0])
+				++symname;  /* ppc64 hack */
+			if (0 == strcmp((('_' == gpfx) ? "_mcount" : "mcount"),
+					symname))
+				mcountsym = ELF64_R_SYM(w8(relp->r_info));
+		}
+
+		if (mcountsym == ELF64_R_SYM(w8(relp->r_info))) {
+			uint64_t const addend = w8(w8(relp->r_offset) - recval);
+
+			mrelp->r_offset = w8(offbase
+				+ ((void *)mlocp - (void *)mloc0));
+			mrelp->r_info = w8(ELF64_R_INFO(recsym, reltype));
+			if (sizeof(Elf64_Rela) == rel_entsize) {
+				((Elf64_Rela *)mrelp)->r_addend = addend;
+				*mlocp++ = 0;
+			} else
+				*mlocp++ = addend;
+
+			mrelp = (Elf64_Rel *)(rel_entsize + (void *)mrelp);
+		}
+		relp = (Elf64_Rel const *)(rel_entsize + (void *)relp);
+	}
+	*mrelpp = mrelp;
+
+	return mlocp;
+}
+
+/*
+ * Find a symbol in the given section, to be used as the base for relocating
+ * the table of offsets of calls to mcount.  A local or global symbol suffices,
+ * but avoid a Weak symbol because it may be overridden; the change in value
+ * would invalidate the relocations of the offsets of the calls to mcount.
+ * Often the found symbol will be the unnamed local symbol generated by
+ * GNU 'as' for the start of each section.  For example:
+ *    Num:    Value  Size Type    Bind   Vis      Ndx Name
+ *      2: 00000000     0 SECTION LOCAL  DEFAULT    1
+ */
+static unsigned find32_secsym_ndx(unsigned const txtndx,
+				  char const *const txtname,
+				  uint32_t *const recvalp,
+				  Elf32_Shdr const *const symhdr,
+				  Elf32_Ehdr const *const ehdr)
+{
+	Elf32_Sym const *const sym0 = (Elf32_Sym const *)(w(symhdr->sh_offset)
+		+ (void *)ehdr);
+	unsigned const nsym = w(symhdr->sh_size) / w(symhdr->sh_entsize);
+	Elf32_Sym const *symp;
+	unsigned t;
+
+	for (symp = sym0, t = nsym; t; --t, ++symp) {
+		unsigned int const st_bind = ELF32_ST_BIND(symp->st_info);
+
+		if (txtndx == w2(symp->st_shndx)
+			/* avoid STB_WEAK */
+		    && (STB_LOCAL == st_bind || STB_GLOBAL == st_bind)) {
+			*recvalp = w(symp->st_value);
+			return symp - sym0;
+		}
+	}
+	fprintf(stderr, "Cannot find symbol for section %d: %s.\n",
+		txtndx, txtname);
+	fail_file();
+}
+
+static unsigned find64_secsym_ndx(unsigned const txtndx,
+				  char const *const txtname,
+				  uint64_t *const recvalp,
+				  Elf64_Shdr const *const symhdr,
+				  Elf64_Ehdr const *const ehdr)
+{
+	Elf64_Sym const *const sym0 = (Elf64_Sym const *)(w8(symhdr->sh_offset)
+		+ (void *)ehdr);
+	unsigned const nsym = w8(symhdr->sh_size) / w8(symhdr->sh_entsize);
+	Elf64_Sym const *symp;
+	unsigned t;
+
+	for (symp = sym0, t = nsym; t; --t, ++symp) {
+		unsigned int const st_bind = ELF64_ST_BIND(symp->st_info);
+
+		if (txtndx == w2(symp->st_shndx)
+			/* avoid STB_WEAK */
+		    && (STB_LOCAL == st_bind || STB_GLOBAL == st_bind)) {
+			*recvalp = w8(symp->st_value);
+			return symp - sym0;
+		}
+	}
+	fprintf(stderr, "Cannot find symbol for section %d: %s.\n",
+		txtndx, txtname);
+	fail_file();
+}
+
+/*
+ * Evade ISO C restriction: no declaration after statement in
+ * has32_rel_mcount.
+ */
+static char const *
+__has32_rel_mcount(Elf32_Shdr const *const relhdr,  /* is SHT_REL or SHT_RELA */
+		   Elf32_Shdr const *const shdr0,
+		   char const *const shstrtab,
+		   char const *const fname)
+{
+	/* .sh_info depends on .sh_type == SHT_REL[,A] */
+	Elf32_Shdr const *const txthdr = &shdr0[w(relhdr->sh_info)];
+	char const *const txtname = &shstrtab[w(txthdr->sh_name)];
+
+	if (0 == strcmp("__mcount_loc", txtname)) {
+		fprintf(stderr, "warning: __mcount_loc already exists: %s\n",
+			fname);
+		succeed_file();
+	}
+	if (SHT_PROGBITS != w(txthdr->sh_type) ||
+	    !is_mcounted_section_name(txtname))
+		return NULL;
+	return txtname;
+}
+
+static char const *has32_rel_mcount(Elf32_Shdr const *const relhdr,
+				    Elf32_Shdr const *const shdr0,
+				    char const *const shstrtab,
+				    char const *const fname)
+{
+	if (SHT_REL  != w(relhdr->sh_type) && SHT_RELA != w(relhdr->sh_type))
+		return NULL;
+	return __has32_rel_mcount(relhdr, shdr0, shstrtab, fname);
+}
+
+static char const *__has64_rel_mcount(Elf64_Shdr const *const relhdr,
+				      Elf64_Shdr const *const shdr0,
+				      char const *const shstrtab,
+				      char const *const fname)
+{
+	/* .sh_info depends on .sh_type == SHT_REL[,A] */
+	Elf64_Shdr const *const txthdr = &shdr0[w(relhdr->sh_info)];
+	char const *const txtname = &shstrtab[w(txthdr->sh_name)];
+
+	if (0 == strcmp("__mcount_loc", txtname)) {
+		fprintf(stderr, "warning: __mcount_loc already exists: %s\n",
+			fname);
+		succeed_file();
+	}
+	if (SHT_PROGBITS != w(txthdr->sh_type) ||
+	    !is_mcounted_section_name(txtname))
+		return NULL;
+	return txtname;
+}
+
+static char const *has64_rel_mcount(Elf64_Shdr const *const relhdr,
+				    Elf64_Shdr const *const shdr0,
+				    char const *const shstrtab,
+				    char const *const fname)
+{
+	if (SHT_REL  != w(relhdr->sh_type) && SHT_RELA != w(relhdr->sh_type))
+		return NULL;
+	return __has64_rel_mcount(relhdr, shdr0, shstrtab, fname);
+}
+
+static unsigned tot32_relsize(Elf32_Shdr const *const shdr0,
+			      unsigned nhdr,
+			      const char *const shstrtab,
+			      const char *const fname)
+{
+	unsigned totrelsz = 0;
+	Elf32_Shdr const *shdrp = shdr0;
+	for (; 0 != nhdr; --nhdr, ++shdrp) {
+		if (has32_rel_mcount(shdrp, shdr0, shstrtab, fname))
+			totrelsz += w(shdrp->sh_size);
+	}
+	return totrelsz;
+}
+
+static unsigned tot64_relsize(Elf64_Shdr const *const shdr0,
+			      unsigned nhdr,
+			      const char *const shstrtab,
+			      const char *const fname)
+{
+	unsigned totrelsz = 0;
+	Elf64_Shdr const *shdrp = shdr0;
+
+	for (; nhdr; --nhdr, ++shdrp) {
+		if (has64_rel_mcount(shdrp, shdr0, shstrtab, fname))
+			totrelsz += w8(shdrp->sh_size);
+	}
+	return totrelsz;
+}
+
+/* Overall supervision for Elf32 ET_REL file. */
+static void
+do32(Elf32_Ehdr *const ehdr, char const *const fname, unsigned const reltype)
+{
+	Elf32_Shdr *const shdr0 = (Elf32_Shdr *)(w(ehdr->e_shoff)
+		+ (void *)ehdr);
+	unsigned const nhdr = w2(ehdr->e_shnum);
+	Elf32_Shdr *const shstr = &shdr0[w2(ehdr->e_shstrndx)];
+	char const *const shstrtab = (char const *)(w(shstr->sh_offset)
+		+ (void *)ehdr);
+
+	Elf32_Shdr const *relhdr;
+	unsigned k;
+
+	/* Upper bound on space: assume all relevant relocs are for mcount. */
+	unsigned const totrelsz = tot32_relsize(shdr0, nhdr, shstrtab, fname);
+	Elf32_Rel *const mrel0 = umalloc(totrelsz);
+	Elf32_Rel *      mrelp = mrel0;
+
+	/* 2*sizeof(address) <= sizeof(Elf32_Rel) */
+	uint32_t *const mloc0 = umalloc(totrelsz>>1);
+	uint32_t *      mlocp = mloc0;
+
+	unsigned rel_entsize = 0;
+	unsigned symsec_sh_link = 0;
+
+	for (relhdr = shdr0, k = nhdr; k; --k, ++relhdr) {
+		char const *const txtname = has32_rel_mcount(relhdr, shdr0,
+			shstrtab, fname);
+		if (txtname) {
+			uint32_t recval = 0;
+			unsigned const recsym = find32_secsym_ndx(
+				w(relhdr->sh_info), txtname, &recval,
+				&shdr0[symsec_sh_link = w(relhdr->sh_link)],
+				ehdr);
+
+			rel_entsize = w(relhdr->sh_entsize);
+			mlocp = sift32_rel_mcount(mlocp,
+				(void *)mlocp - (void *)mloc0, &mrelp,
+				relhdr, ehdr, recsym, recval, reltype);
+		}
+	}
+	if (mloc0 != mlocp) {
+		append32(ehdr, shstr, mloc0, mlocp, mrel0, mrelp,
+			rel_entsize, symsec_sh_link);
+	}
+	free(mrel0);
+	free(mloc0);
+}
+
+static void
+do64(Elf64_Ehdr *const ehdr, char const *const fname, unsigned const reltype)
+{
+	Elf64_Shdr *const shdr0 = (Elf64_Shdr *)(w8(ehdr->e_shoff)
+		+ (void *)ehdr);
+	unsigned const nhdr = w2(ehdr->e_shnum);
+	Elf64_Shdr *const shstr = &shdr0[w2(ehdr->e_shstrndx)];
+	char const *const shstrtab = (char const *)(w8(shstr->sh_offset)
+		+ (void *)ehdr);
+
+	Elf64_Shdr const *relhdr;
+	unsigned k;
+
+	/* Upper bound on space: assume all relevant relocs are for mcount. */
+	unsigned const totrelsz = tot64_relsize(shdr0, nhdr, shstrtab, fname);
+	Elf64_Rel *const mrel0 = umalloc(totrelsz);
+	Elf64_Rel *      mrelp = mrel0;
+
+	/* 2*sizeof(address) <= sizeof(Elf64_Rel) */
+	uint64_t *const mloc0 = umalloc(totrelsz>>1);
+	uint64_t *      mlocp = mloc0;
+
+	unsigned rel_entsize = 0;
+	unsigned symsec_sh_link = 0;
+
+	for ((relhdr = shdr0), k = nhdr; k; --k, ++relhdr) {
+		char const *const txtname = has64_rel_mcount(relhdr, shdr0,
+			shstrtab, fname);
+		if (txtname) {
+			uint64_t recval = 0;
+			unsigned const recsym = find64_secsym_ndx(
+				w(relhdr->sh_info), txtname, &recval,
+				&shdr0[symsec_sh_link = w(relhdr->sh_link)],
+				ehdr);
+
+			rel_entsize = w8(relhdr->sh_entsize);
+			mlocp = sift64_rel_mcount(mlocp,
+				(void *)mlocp - (void *)mloc0, &mrelp,
+				relhdr, ehdr, recsym, recval, reltype);
+		}
+	}
+	if (mloc0 != mlocp) {
+		append64(ehdr, shstr, mloc0, mlocp, mrel0, mrelp,
+			rel_entsize, symsec_sh_link);
+	}
+	free(mrel0);
+	free(mloc0);
+}
+
+static void
+do_file(char const *const fname)
+{
+	Elf32_Ehdr *const ehdr = mmap_file(fname);
+	unsigned int reltype = 0;
+
+	ehdr_curr = ehdr;
+	w = w4nat;
+	w2 = w2nat;
+	w8 = w8nat;
+	switch (ehdr->e_ident[EI_DATA]) {
+		static unsigned int const endian = 1;
+	default: {
+		fprintf(stderr, "unrecognized ELF data encoding %d: %s\n",
+			ehdr->e_ident[EI_DATA], fname);
+		fail_file();
+	} break;
+	case ELFDATA2LSB: {
+		if (1 != *(unsigned char const *)&endian) {
+			/* main() is big endian, file.o is little endian. */
+			w = w4rev;
+			w2 = w2rev;
+			w8 = w8rev;
+		}
+	} break;
+	case ELFDATA2MSB: {
+		if (0 != *(unsigned char const *)&endian) {
+			/* main() is little endian, file.o is big endian. */
+			w = w4rev;
+			w2 = w2rev;
+			w8 = w8rev;
+		}
+	} break;
+	}  /* end switch */
+	if (0 != memcmp(ELFMAG, ehdr->e_ident, SELFMAG)
+	||  ET_REL != w2(ehdr->e_type)
+	||  EV_CURRENT != ehdr->e_ident[EI_VERSION]) {
+		fprintf(stderr, "unrecognized ET_REL file %s\n", fname);
+		fail_file();
+	}
+
+	gpfx = 0;
+	switch (w2(ehdr->e_machine)) {
+	default: {
+		fprintf(stderr, "unrecognized e_machine %d %s\n",
+			w2(ehdr->e_machine), fname);
+		fail_file();
+	} break;
+	case EM_386:	 reltype = R_386_32;                   break;
+	case EM_ARM:	 reltype = R_ARM_ABS32;                break;
+	case EM_IA_64:	 reltype = R_IA64_IMM64;   gpfx = '_'; break;
+	case EM_PPC:	 reltype = R_PPC_ADDR32;   gpfx = '_'; break;
+	case EM_PPC64:	 reltype = R_PPC64_ADDR64; gpfx = '_'; break;
+	case EM_S390:    /* reltype: e_class    */ gpfx = '_'; break;
+	case EM_SH:	 reltype = R_SH_DIR32;                 break;
+	case EM_SPARCV9: reltype = R_SPARC_64;     gpfx = '_'; break;
+	case EM_X86_64:	 reltype = R_X86_64_64;                break;
+	}  /* end switch */
+
+	switch (ehdr->e_ident[EI_CLASS]) {
+	default: {
+		fprintf(stderr, "unrecognized ELF class %d %s\n",
+			ehdr->e_ident[EI_CLASS], fname);
+		fail_file();
+	} break;
+	case ELFCLASS32: {
+		if (sizeof(Elf32_Ehdr) != w2(ehdr->e_ehsize)
+		||  sizeof(Elf32_Shdr) != w2(ehdr->e_shentsize)) {
+			fprintf(stderr,
+				"unrecognized ET_REL file: %s\n", fname);
+			fail_file();
+		}
+		if (EM_S390 == w2(ehdr->e_machine))
+			reltype = R_390_32;
+		do32(ehdr, fname, reltype);
+	} break;
+	case ELFCLASS64: {
+		Elf64_Ehdr *const ghdr = (Elf64_Ehdr *)ehdr;
+		if (sizeof(Elf64_Ehdr) != w2(ghdr->e_ehsize)
+		||  sizeof(Elf64_Shdr) != w2(ghdr->e_shentsize)) {
+			fprintf(stderr,
+				"unrecognized ET_REL file: %s\n", fname);
+			fail_file();
+		}
+		if (EM_S390 == w2(ghdr->e_machine))
+			reltype = R_390_64;
+		do64(ghdr, fname, reltype);
+	} break;
+	}  /* end switch */
+
+	cleanup();
+}
+
+int
+main(int argc, char const *argv[])
+{
+	int n_error = 0;  /* gcc-4.3.0 false positive complaint */
+	if (argc <= 1)
+		fprintf(stderr, "usage: recordmcount file.o...\n");
+	else  /* Process each file in turn, allowing deep failure. */
+	for (--argc, ++argv; 0 < argc; --argc, ++argv) {
+		int const sjval = setjmp(jmpenv);
+		switch (sjval) {
+		default: {
+			fprintf(stderr, "internal error: %s\n", argv[0]);
+			exit(1);
+		} break;
+		case SJ_SETJMP: {  /* normal sequence */
+			/* Avoid problems if early cleanup() */
+			fd_map = -1;
+			ehdr_curr = NULL;
+			mmap_failed = 1;
+			do_file(argv[0]);
+		} break;
+		case SJ_FAIL: {  /* error in do_file or below */
+			++n_error;
+		} break;
+		case SJ_SUCCEED: {  /* premature success */
+			/* do nothing */
+		} break;
+		}  /* end switch */
+	}
+	return !!n_error;
+}
+
+
Signature · 1 line
1.7.1
author maintainer
#2 lore raw

PATCH 2/3 ftrace/x86: Add support for C version of recordmcount

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

After verifying this works, other archs can add:

HAVE_C_MCOUNT_RECORD

in its Kconfig and it will use the C version of recordmcount instead of the perl version.

---
 Makefile               |    6 ++++++
 arch/x86/Kconfig       |    1 +
 kernel/trace/Kconfig   |    5 +++++
 scripts/Makefile       |    1 +
 scripts/Makefile.build |    4 ++++
 5 files changed, 17 insertions(+), 0 deletions(-)
Makefile +6 -0
index 534c09c..0dd3a8d 100644
@@ -568,6 +568,12 @@ endif
 
 ifdef CONFIG_FUNCTION_TRACER
 KBUILD_CFLAGS	+= -pg
+ifdef CONFIG_DYNAMIC_FTRACE
+	ifdef CONFIG_HAVE_C_MCOUNT_RECORD
+		BUILD_C_RECORDMCOUNT := y
+		export BUILD_C_RECORDMCOUNT
+	endif
+endif
 endif
 
 # We trigger additional mismatches with less inlining
arch/x86/Kconfig +1 -0
index c14d8b4..788b50e 100644
@@ -33,6 +33,7 @@ config X86
 	select HAVE_KRETPROBES
 	select HAVE_OPTPROBES
 	select HAVE_FTRACE_MCOUNT_RECORD
+	select HAVE_C_MCOUNT_RECORD
 	select HAVE_DYNAMIC_FTRACE
 	select HAVE_FUNCTION_TRACER
 	select HAVE_FUNCTION_GRAPH_TRACER
kernel/trace/Kconfig +5 -0
index 538501c..df00fbb 100644
@@ -49,6 +49,11 @@ config HAVE_SYSCALL_TRACEPOINTS
 	help
 	  See Documentation/trace/ftrace-design.txt
 
+config HAVE_C_MCOUNT_RECORD
+	bool
+	help
+	  C version of recordmcount available?
+
 config TRACER_MAX_TRACE
 	bool
 
scripts/Makefile +1 -0
index 842dbc2..2e08810 100644
@@ -11,6 +11,7 @@ hostprogs-$(CONFIG_KALLSYMS)     += kallsyms
 hostprogs-$(CONFIG_LOGO)         += pnmtologo
 hostprogs-$(CONFIG_VT)           += conmakehash
 hostprogs-$(CONFIG_IKCONFIG)     += bin2c
+hostprogs-$(BUILD_C_RECORDMCOUNT) += recordmcount
 
 always		:= $(hostprogs-y) $(hostprogs-m)
 
scripts/Makefile.build +4 -0
index a1a5cf9..4d03a7e 100644
@@ -209,12 +209,16 @@ cmd_modversions =								\
 endif
 
 ifdef CONFIG_FTRACE_MCOUNT_RECORD
+ifdef BUILD_C_RECORDMCOUNT
+cmd_record_mcount = $(srctree)/scripts/recordmcount "$(@)";
+else
 cmd_record_mcount = set -e ; perl $(srctree)/scripts/recordmcount.pl "$(ARCH)" \
 	"$(if $(CONFIG_CPU_BIG_ENDIAN),big,little)" \
 	"$(if $(CONFIG_64BIT),64,32)" \
 	"$(OBJDUMP)" "$(OBJCOPY)" "$(CC)" "$(LD)" "$(NM)" "$(RM)" "$(MV)" \
 	"$(if $(part-of-module),1,0)" "$(@)";
 endif
+endif
 
 define rule_cc_o_c
 	$(call echo-cmd,checksrc) $(cmd_checksrc)			  \
Signature · 1 line
1.7.1
#3 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

Quoted 4 lines · Steven Rostedt

From: Steven Rostedt <[email protected]>

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

I reported this recordmcount performance problem 2 years ago. Better later than never i guess.

arch/x86/Kconfig config X86 · line 36

Quoted 12 lines
Makefile +4 -0
+ifdef CONFIG_DYNAMIC_FTRACE
+	ifdef CONFIG_HAVE_C_MCOUNT_RECORD
+		BUILD_C_RECORDMCOUNT := y
+		export BUILD_C_RECORDMCOUNT
arch/x86/Kconfig +1 -0
@@ -33,6 +33,7 @@ config X86
 	select HAVE_KRETPROBES
 	select HAVE_OPTPROBES
 	select HAVE_FTRACE_MCOUNT_RECORD
+	select HAVE_C_MCOUNT_RECORD

The naming is inconsistent here - it should be HAVE_C_RECORDMCOUNT, like the build variable has, and like the utility is called. If we are going to add this flag to most architectures we should name it consistently.

Thanks,

Ingo

author maintainer
#4 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

Quoted 9 lines · Ingo Molnar
Quoted 4 lines · Steven Rostedt

From: Steven Rostedt <[email protected]>

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

I reported this recordmcount performance problem 2 years ago. Better later than never i guess.

And I also remember saying after I posted this code that it would have a compile time performance hit. Heck, it's a perl script running on every object file. It was obvious what was at issue here. But it's better to slow down the kernel build than to brick network cards. Also, perl was much easier to do.

That said, the embarrassing thing is not that I knew (or you reported it) about this performance problem. I'm actually quite embarrassed that I had this code sitting in my inbox for over a year. I just kept having other things that were more important coming up than lowering the compile time of the kernel. Although, I did work to get streamline config to offset this performance hit.

Finally, while at the End Users Summit, I decided to take a look at John's code, and I was quite impressed.

But as you said, better late than never.

Quoted 6 lines
Quoted 5 lines
Makefile +4 -0
+ifdef CONFIG_DYNAMIC_FTRACE
+	ifdef CONFIG_HAVE_C_MCOUNT_RECORD
+		BUILD_C_RECORDMCOUNT := y
+		export BUILD_C_RECORDMCOUNT
Quoted 11 lines

arch/x86/Kconfig config X86 · line 36

Quoted 7 lines
arch/x86/Kconfig +1 -0
@@ -33,6 +33,7 @@ config X86
 	select HAVE_KRETPROBES
 	select HAVE_OPTPROBES
 	select HAVE_FTRACE_MCOUNT_RECORD
+	select HAVE_C_MCOUNT_RECORD

The naming is inconsistent here - it should be HAVE_C_RECORDMCOUNT, like the build variable has, and like the utility is called. If we are going to add this flag to most architectures we should name it consistently.

Sure, want me to rebase it or just write a patch on top of it?

-- Steve

#5 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

Quoted 8 lines · Steven Rostedt
Quoted 7 lines · Ingo Molnar
Quoted 4 lines · Steven Rostedt

From: Steven Rostedt <[email protected]>

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

Quoted 7 lines
Quoted 2 lines

I reported this recordmcount performance problem 2 years ago. Better later than never i guess.

And I also remember saying after I posted this code that it would have a compile time performance hit. Heck, it's a perl script running on every object file. It was obvious what was at issue here. But it's better to slow down the kernel build than to brick network cards.

Well, it's even better to do neither!

Quoted 1 line

Also, perl was much easier to do.

Lets write the whole kernel in perl and forget about performance ;-)

Quoted 11 lines

That said, the embarrassing thing is not that I knew (or you reported it) about this performance problem. I'm actually quite embarrassed that I had this code sitting in my inbox for over a year. I just kept having other things that were more important coming up than lowering the compile time of the kernel. Although, I did work to get streamline config to offset this performance hit.

Finally, while at the End Users Summit, I decided to take a look at John's code, and I was quite impressed.

But as you said, better late than never.

Yeah. Note that as a maintainer i need to grumble when i see some not-so-good event - even if there's a happy resolution! Otherwise such cases would tend to creep up in frequency ;-)

Quoted 17 lines
Quoted 5 lines
Quoted 4 lines
Makefile +4 -0
+ifdef CONFIG_DYNAMIC_FTRACE
+	ifdef CONFIG_HAVE_C_MCOUNT_RECORD
+		BUILD_C_RECORDMCOUNT := y
+		export BUILD_C_RECORDMCOUNT
Quoted 11 lines

arch/x86/Kconfig config X86 · line 36

Quoted 7 lines
arch/x86/Kconfig +1 -0
@@ -33,6 +33,7 @@ config X86
 	select HAVE_KRETPROBES
 	select HAVE_OPTPROBES
 	select HAVE_FTRACE_MCOUNT_RECORD
+	select HAVE_C_MCOUNT_RECORD

The naming is inconsistent here - it should be HAVE_C_RECORDMCOUNT, like the build variable has, and like the utility is called. If we are going to add this flag to most architectures we should name it consistently.

Quoted 1 line

Sure, want me to rebase it or just write a patch on top of it?

Sure, patch on top would be fine.

Thanks,

Ingo

author maintainer
#6 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

Quoted 8 lines · Ingo Molnar
Quoted 4 lines · Steven Rostedt
Quoted 1 line · Ingo Molnar

* Steven Rostedt <[email protected]> wrote:

Also, perl was much easier to do.

Lets write the whole kernel in perl and forget about performance ;-)

Shhh, you're letting people know about my evil secret agenda!

Quoted 8 lines
Quoted 6 lines
Quoted 4 lines

The naming is inconsistent here - it should be HAVE_C_RECORDMCOUNT, like the build variable has, and like the utility is called. If we are going to add this flag to most architectures we should name it consistently.

Sure, want me to rebase it or just write a patch on top of it?

Sure, patch on top would be fine.

OK, I'll add a patch on top.

Thanks,

-- Steve

#7 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

In reply to Steven Rostedt
Quoted 5 lines · Steven Rostedt

From: Steven Rostedt <[email protected]>

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

Quoted 7 lines

After verifying this works, other archs can add:

HAVE_C_MCOUNT_RECORD

in its Kconfig and it will use the C version of recordmcount instead of the perl version.

While I haven't had the chance to debug this yet, turning it on for SH blows up immediately:

ftrace: allocating 15200 entries in 30 pages

------------[ cut here ]------------
WARNING: at /home/pmundt/devel/git/sh-2.6/kernel/trace/ftrace.c:1007
Modules linked in:

Pid : 0, Comm: swapper
CPU : 0 Not tainted (2.6.36-05622-g38ab134-dirty #508)

PC is at ftrace_bug+0x78/0x23c
PR is at ftrace_bug+0x74/0x23c
PC : 80064df4 SP : 8056ff70 SR : 400080f0 TEA : c0000004 R0 : 00000001 R1 : 00000001 R2 : 8064d862 R3 : 8056ff64 R4 : 805b47b4 R5 : 00000001 R6 : 00000000 R7 : 00000001 R8 : 803b15d8 R9 : 00000001 R10 : 9fc38be8 R11 : 00000000 R12 : 8064e88c R13 : 8064e880 R14 : 8056ff70
MACH: 00000000 MACL: 003d0900 GBR : 296e1678 PR : 80064df0

Call trace:
 [<80066a86>] ftrace_process_locs+0x15a/0x284
 [<803b15d8>] dns_query+0x0/0x26c
 [<805f6a1a>] ftrace_init+0x112/0x1a8
 [<801deec0>] strlen+0x0/0x58
 [<8008f098>] get_zeroed_page+0x0/0x34
 [<805f0918>] start_kernel+0x3e0/0x480
 [<801deec0>] strlen+0x0/0x58
 [<801eb388>] debug_smp_processor_id+0x0/0xe4
 [<80002132>] _stext+0x132/0x140

Code:

  80064dee:  mov       r9, r5
  80064df0:  tst       r9, r9
  80064df2:  bt        80064df6
->80064df4:  trapa     #62
  80064df6:  bra       80064ef2
  80064df8:  mov       r9, r5
  80064dfa:  mov.l     80064f68 <ftrace_bug+0x1ec/0x23c>, r1  ! 8064d862 <__warned.27604+0x0/0x1>
  80064dfc:  mov.b     r2, @r1
  80064dfe:  mov.l     80064f48 <ftrace_bug+0x1cc/0x23c>, r1  ! 8057021c <ftrace_disabled+0x0/0x4>
---[ end trace 4eaa2a86a8e2da22 ]---
ftrace failed to modify [<803b15d8>] dns_query+0x0/0x26c
 actual: 02:d1:22:4f

Testing tracer nop: PASSED

Suggestions?

#8 lore raw

Re: [PATCH 2/3] ftrace/x86: Add support for C version of recordmcount

[Note that I have trimmed the addressees for this reply.]

Quoted 28 lines · Paul Mundt
Quoted 2 lines

This patch adds the support for the C version of recordmcount and compile times show ~ 12% improvement.

While I haven't had the chance to debug this yet, turning it on for SH blows up immediately:

ftrace: allocating 15200 entries in 30 pages

------------[ cut here ]------------
WARNING: at /home/pmundt/devel/git/sh-2.6/kernel/trace/ftrace.c:1007
Modules linked in:

Pid : 0, Comm: swapper
CPU : 0 Not tainted (2.6.36-05622-g38ab134-dirty #508)

PC is at ftrace_bug+0x78/0x23c
PR is at ftrace_bug+0x74/0x23c

Call trace:
 [<80066a86>] ftrace_process_locs+0x15a/0x284
 [<803b15d8>] dns_query+0x0/0x26c

Code:

  80064dee:  mov       r9, r5
  80064df0:  tst       r9, r9
  80064df2:  bt        80064df6
->80064df4:  trapa     #62
  80064df6:  bra       80064ef2
Quoted 6 lines
---[ end trace 4eaa2a86a8e2da22 ]---
ftrace failed to modify [<803b15d8>] dns_query+0x0/0x26c
 actual: 02:d1:22:4f

Testing tracer nop: PASSED

Suggestions?

This looks like a disagreement over the interpretation of relocation type R_SH_DIR32 in Elf32_Rela on a 32-bit EM_SH (SuperH) machine.

arch/sh/kernel/module.c says:
-----
sym = (Elf32_Sym *)sechdrs[symindex].sh_addr
+ ELF32_R_SYM(rel[i].r_info);
relocation = sym->st_value + rel[i].r_addend;
...

                case R_SH_DIR32:
                        value = get_unaligned(location);
                        value += relocation;
                        put_unaligned(value, location);

-----

Paul Mundt sent to me an actual kernel object file main.o compiled -pg for EM_SH. According to 'readelf', the Perl script recordmcount.pl generates:
-----
Relocation section '.rela__mcount_loc' at offset 0x17ec8 contains 3 entries:
Offset Info Type Sym.Value Sym. Name + Addend

00000000  00006a01 R_SH_DIR32        00000000   do_one_initcall + 0
00000004  00006a01 R_SH_DIR32        00000000   do_one_initcall + 0
00000008  00006a01 R_SH_DIR32        00000000   do_one_initcall + 0

Symbol table '.symtab' contains 198 entries:
Num: Value Size Type Bind Vis Ndx Name
106: 00000000 364 FUNC GLOBAL DEFAULT 1 do_one_initcall
### (1==Ndx) ==> .text, so do_one_initcall is first in .text ### 106==0x6a

Section Headers:

  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 1] .text             PROGBITS        00000000 000034 00026c 00  AX  0   0  4
  [17] __mcount_loc      PROGBITS        00000000 001140 00000c 00   A  0   0  4
  [18] .rela__mcount_loc RELA            00000000 017ec8 000024 0c     46  17  4
  [46] .symtab           SYMTAB          00000000 022ac4 000c60 10     47  71  4

-----
where section 17, __mcount_loc, has three 32-bit words with values 0xc, 0x178, 0x1b0.

Also according to 'readelf', the C program recordmcount generates:
-----
Relocation section '.rela__mcount_loc' at offset 0x249f4 contains 3 entries:
Offset Info Type Sym.Value Sym. Name + Addend

00000000  00000201 R_SH_DIR32        00000000   .text + c
00000004  00000201 R_SH_DIR32        00000000   .text + 178
00000008  00000201 R_SH_DIR32        00000000   .text + 1b0

Symbol table '.symtab' contains 198 entries:
Num: Value Size Type Bind Vis Ndx Name

     2: 00000000     0 SECTION LOCAL  DEFAULT    1

### (1==Ndx) ==> .text

Section Headers:

  [Nr] Name              Type            Addr     Off    Size   ES Flg Lk Inf Al
  [ 1] .text             PROGBITS        00000000 000034 00026c 00  AX  0   0  4
  [44] .symtab           SYMTAB          00000000 017440 000c50 10     45  70  4
  [46] __mcount_loc      PROGBITS        00000000 0249e8 00000c 04   A  0   0  4
  [47] .rela__mcount_loc RELA            00000000 0249f4 000024 0c     44  46  4

-----
where section 46, __mcount_loc, has three 32-bit words with values 0, 0, 0.

Assuming the code in arch/sh/kernel/module.c, then I believe that those two descriptions are equivalent. The output of the Perl script calls for

*location = *location + do_one_initcall.st_value + .r_addend

with 0==.r_addend and initially *location==offset . The output of the C version calls for

*location = *location + .text.st_value + .r_addend

with offset==.r_addend and initially *location==0.

I did find in binutils-2.18.50.0.9/bfd/elf32-sh.c this ominous comment:
-----

      switch ((int) r_type)
        {
        final_link_relocate:
          /* COFF relocs don't use the addend. The addend is used for
             R_SH_DIR32 to be compatible with other compilers.  */
          r = _bfd_final_link_relocate (howto, input_bfd, input_section,
                                        contents, rel->r_offset,
                                        relocation, addend);

-----

The natural question is, "Does /bin/ld for EM_SH ignore the .r_addend, or does /bin/ld process .r_addend just like arch/sh/kernel/module.c ?" If /bin/ld ignores .r_addend for R_SH_DIR32, as hinted in the comment, then that would explain the crash.

Another question is, "Does the SECTION LOCAL symbol for .text have .st_value equal to the virtual address of the start of the section?" If a SECTION LOCAL symbol was not relocated like this, and thus had the value 0, then that also would explain the crash.

That's my analysis. If anyone can shed more light, then please do!

Signature · 1 line
John Reiser, [email protected]
author maintainer
#9 lore raw

PATCH 3/3 ftrace: Remove duplicate code for 64 and 32 bit in recordmcount.c

In reply to Steven Rostedt

From: Steven Rostedt <[email protected]>

The elf reader for recordmcount.c had duplicate functions for both 32 bit and 64 bit elf handling. This was due to the need of using the 32 and 64 bit elf structures.

This patch consolidates the two by using macros to define the 32 and 64 bit names in a recordmcount.h file, and then by just defining a RECORD_MCOUNT_64 macro and including recordmcount.h twice we create the funtions for both the 32 bit version as well as the 64 bit version using one code source.

---
 scripts/recordmcount.c |  548 +-----------------------------------------------
 scripts/recordmcount.h |  366 ++++++++++++++++++++++++++++++++
 2 files changed, 370 insertions(+), 544 deletions(-)
 create mode 100644 scripts/recordmcount.h
scripts/recordmcount.c +4 -544
index 34f32be..7f7f718 100644
@@ -212,550 +212,10 @@ is_mcounted_section_name(char const *const txtname)
 		0 == strcmp(".text.unlikely", txtname);
 }
 
-/* Append the new shstrtab, Elf32_Shdr[], __mcount_loc and its relocations. */
-static void append32(Elf32_Ehdr *const ehdr,
-		     Elf32_Shdr *const shstr,
-		     uint32_t const *const mloc0,
-		     uint32_t const *const mlocp,
-		     Elf32_Rel const *const mrel0,
-		     Elf32_Rel const *const mrelp,
-		     unsigned int const rel_entsize,
-		     unsigned int const symsec_sh_link)
-{
-	/* Begin constructing output file */
-	Elf32_Shdr mcsec;
-	char const *mc_name = (sizeof(Elf32_Rela) == rel_entsize)
-		? ".rela__mcount_loc"
-		:  ".rel__mcount_loc";
-	unsigned const old_shnum = w2(ehdr->e_shnum);
-	uint32_t const old_shoff = w(ehdr->e_shoff);
-	uint32_t const old_shstr_sh_size   = w(shstr->sh_size);
-	uint32_t const old_shstr_sh_offset = w(shstr->sh_offset);
-	uint32_t t = 1 + strlen(mc_name) + w(shstr->sh_size);
-	uint32_t new_e_shoff;
-
-	shstr->sh_size = w(t);
-	shstr->sh_offset = w(sb.st_size);
-	t += sb.st_size;
-	t += (3u & -t);  /* 4-byte align */
-	new_e_shoff = t;
-
-	/* body for new shstrtab */
-	ulseek(fd_map, sb.st_size, SEEK_SET);
-	uwrite(fd_map, old_shstr_sh_offset + (void *)ehdr, old_shstr_sh_size);
-	uwrite(fd_map, mc_name, 1 + strlen(mc_name));
-
-	/* old(modified) Elf32_Shdr table, 4-byte aligned */
-	ulseek(fd_map, t, SEEK_SET);
-	t += sizeof(Elf32_Shdr) * old_shnum;
-	uwrite(fd_map, old_shoff + (void *)ehdr,
-	       sizeof(Elf32_Shdr) * old_shnum);
-
-	/* new sections __mcount_loc and .rel__mcount_loc */
-	t += 2*sizeof(mcsec);
-	mcsec.sh_name = w((sizeof(Elf32_Rela) == rel_entsize) + strlen(".rel")
-		+ old_shstr_sh_size);
-	mcsec.sh_type = w(SHT_PROGBITS);
-	mcsec.sh_flags = w(SHF_ALLOC);
-	mcsec.sh_addr = 0;
-	mcsec.sh_offset = w(t);
-	mcsec.sh_size = w((void *)mlocp - (void *)mloc0);
-	mcsec.sh_link = 0;
-	mcsec.sh_info = 0;
-	mcsec.sh_addralign = w(4);
-	mcsec.sh_entsize = w(4);
-	uwrite(fd_map, &mcsec, sizeof(mcsec));
-
-	mcsec.sh_name = w(old_shstr_sh_size);
-	mcsec.sh_type = (sizeof(Elf32_Rela) == rel_entsize)
-		? w(SHT_RELA)
-		: w(SHT_REL);
-	mcsec.sh_flags = 0;
-	mcsec.sh_addr = 0;
-	mcsec.sh_offset = w((void *)mlocp - (void *)mloc0 + t);
-	mcsec.sh_size   = w((void *)mrelp - (void *)mrel0);
-	mcsec.sh_link = w(symsec_sh_link);
-	mcsec.sh_info = w(old_shnum);
-	mcsec.sh_addralign = w(4);
-	mcsec.sh_entsize = w(rel_entsize);
-	uwrite(fd_map, &mcsec, sizeof(mcsec));
-
-	uwrite(fd_map, mloc0, (void *)mlocp - (void *)mloc0);
-	uwrite(fd_map, mrel0, (void *)mrelp - (void *)mrel0);
-
-	ehdr->e_shoff = w(new_e_shoff);
-	ehdr->e_shnum = w2(2 + w2(ehdr->e_shnum));  /* {.rel,}__mcount_loc */
-	ulseek(fd_map, 0, SEEK_SET);
-	uwrite(fd_map, ehdr, sizeof(*ehdr));
-}
-
-/*
- * append64 and append32 (and other analogous pairs) could be templated
- * using C++, but the complexity is high.  (For an example, look at p_elf.h
- * in the source for UPX, http://upx.sourceforge.net)  So: remember to make
- * the corresponding change in the routine for the other size.
- */
-static void append64(Elf64_Ehdr *const ehdr,
-		     Elf64_Shdr *const shstr,
-		     uint64_t const *const mloc0,
-		     uint64_t const *const mlocp,
-		     Elf64_Rel const *const mrel0,
-		     Elf64_Rel const *const mrelp,
-		     unsigned int const rel_entsize,
-		     unsigned int const symsec_sh_link)
-{
-	/* Begin constructing output file */
-	Elf64_Shdr mcsec;
-	char const *mc_name = (sizeof(Elf64_Rela) == rel_entsize)
-		? ".rela__mcount_loc"
-		:  ".rel__mcount_loc";
-	unsigned const old_shnum = w2(ehdr->e_shnum);
-	uint64_t const old_shoff = w8(ehdr->e_shoff);
-	uint64_t const old_shstr_sh_size   = w8(shstr->sh_size);
-	uint64_t const old_shstr_sh_offset = w8(shstr->sh_offset);
-	uint64_t t = 1 + strlen(mc_name) + w8(shstr->sh_size);
-	uint64_t new_e_shoff;
-
-	shstr->sh_size = w8(t);
-	shstr->sh_offset = w8(sb.st_size);
-	t += sb.st_size;
-	t += (7u & -t);  /* 8-byte align */
-	new_e_shoff = t;
-
-	/* body for new shstrtab */
-	ulseek(fd_map, sb.st_size, SEEK_SET);
-	uwrite(fd_map, old_shstr_sh_offset + (void *)ehdr, old_shstr_sh_size);
-	uwrite(fd_map, mc_name, 1 + strlen(mc_name));
-
-	/* old(modified) Elf64_Shdr table, 8-byte aligned */
-	ulseek(fd_map, t, SEEK_SET);
-	t += sizeof(Elf64_Shdr) * old_shnum;
-	uwrite(fd_map, old_shoff + (void *)ehdr,
-		sizeof(Elf64_Shdr) * old_shnum);
-
-	/* new sections __mcount_loc and .rel__mcount_loc */
-	t += 2*sizeof(mcsec);
-	mcsec.sh_name = w((sizeof(Elf64_Rela) == rel_entsize) + strlen(".rel")
-		+ old_shstr_sh_size);
-	mcsec.sh_type = w(SHT_PROGBITS);
-	mcsec.sh_flags = w8(SHF_ALLOC);
-	mcsec.sh_addr = 0;
-	mcsec.sh_offset = w8(t);
-	mcsec.sh_size = w8((void *)mlocp - (void *)mloc0);
-	mcsec.sh_link = 0;
-	mcsec.sh_info = 0;
-	mcsec.sh_addralign = w8(8);
-	mcsec.sh_entsize = w8(8);
-	uwrite(fd_map, &mcsec, sizeof(mcsec));
-
-	mcsec.sh_name = w(old_shstr_sh_size);
-	mcsec.sh_type = (sizeof(Elf64_Rela) == rel_entsize)
-		? w(SHT_RELA)
-		: w(SHT_REL);
-	mcsec.sh_flags = 0;
-	mcsec.sh_addr = 0;
-	mcsec.sh_offset = w8((void *)mlocp - (void *)mloc0 + t);
-	mcsec.sh_size   = w8((void *)mrelp - (void *)mrel0);
-	mcsec.sh_link = w(symsec_sh_link);
-	mcsec.sh_info = w(old_shnum);
-	mcsec.sh_addralign = w8(8);
-	mcsec.sh_entsize = w8(rel_entsize);
-	uwrite(fd_map, &mcsec, sizeof(mcsec));
-
-	uwrite(fd_map, mloc0, (void *)mlocp - (void *)mloc0);
-	uwrite(fd_map, mrel0, (void *)mrelp - (void *)mrel0);
-
-	ehdr->e_shoff = w8(new_e_shoff);
-	ehdr->e_shnum = w2(2 + w2(ehdr->e_shnum));  /* {.rel,}__mcount_loc */
-	ulseek(fd_map, 0, SEEK_SET);
-	uwrite(fd_map, ehdr, sizeof(*ehdr));
-}
-
-/*
- * Look at the relocations in order to find the calls to mcount.
- * Accumulate the section offsets that are found, and their relocation info,
- * onto the end of the existing arrays.
- */
-static uint32_t *sift32_rel_mcount(uint32_t *mlocp,
-				   unsigned const offbase,
-				   Elf32_Rel **const mrelpp,
-				   Elf32_Shdr const *const relhdr,
-				   Elf32_Ehdr const *const ehdr,
-				   unsigned const recsym,
-				   uint32_t const recval,
-				   unsigned const reltype)
-{
-	uint32_t *const mloc0 = mlocp;
-	Elf32_Rel *mrelp = *mrelpp;
-	Elf32_Shdr *const shdr0 = (Elf32_Shdr *)(w(ehdr->e_shoff)
-		+ (void *)ehdr);
-	unsigned const symsec_sh_link = w(relhdr->sh_link);
-	Elf32_Shdr const *const symsec = &shdr0[symsec_sh_link];
-	Elf32_Sym const *const sym0 = (Elf32_Sym const *)(w(symsec->sh_offset)
-		+ (void *)ehdr);
-
-	Elf32_Shdr const *const strsec = &shdr0[w(symsec->sh_link)];
-	char const *const str0 = (char const *)(w(strsec->sh_offset)
-		+ (void *)ehdr);
-
-	Elf32_Rel const *const rel0 = (Elf32_Rel const *)(w(relhdr->sh_offset)
-		+ (void *)ehdr);
-	unsigned rel_entsize = w(relhdr->sh_entsize);
-	unsigned const nrel = w(relhdr->sh_size) / rel_entsize;
-	Elf32_Rel const *relp = rel0;
-
-	unsigned mcountsym = 0;
-	unsigned t;
-
-	for (t = nrel; t; --t) {
-		if (!mcountsym) {
-			Elf32_Sym const *const symp =
-				&sym0[ELF32_R_SYM(w(relp->r_info))];
-
-			if (0 == strcmp((('_' == gpfx) ? "_mcount" : "mcount"),
-					&str0[w(symp->st_name)]))
-				mcountsym = ELF32_R_SYM(w(relp->r_info));
-		}
-		if (mcountsym == ELF32_R_SYM(w(relp->r_info))) {
-			uint32_t const addend = w(w(relp->r_offset) - recval);
-			mrelp->r_offset = w(offbase
-				+ ((void *)mlocp - (void *)mloc0));
-			mrelp->r_info = w(ELF32_R_INFO(recsym, reltype));
-			if (sizeof(Elf32_Rela) == rel_entsize) {
-				((Elf32_Rela *)mrelp)->r_addend = addend;
-				*mlocp++ = 0;
-			} else
-				*mlocp++ = addend;
-
-			mrelp = (Elf32_Rel *)(rel_entsize + (void *)mrelp);
-		}
-		relp = (Elf32_Rel const *)(rel_entsize + (void *)relp);
-	}
-	*mrelpp = mrelp;
-	return mlocp;
-}
-
-static uint64_t *sift64_rel_mcount(uint64_t *mlocp,
-				   unsigned const offbase,
-				   Elf64_Rel **const mrelpp,
-				   Elf64_Shdr const *const relhdr,
-				   Elf64_Ehdr const *const ehdr,
-				   unsigned const recsym,
-				   uint64_t const recval,
-				   unsigned const reltype)
-{
-	uint64_t *const mloc0 = mlocp;
-	Elf64_Rel *mrelp = *mrelpp;
-	Elf64_Shdr *const shdr0 = (Elf64_Shdr *)(w8(ehdr->e_shoff)
-		+ (void *)ehdr);
-	unsigned const symsec_sh_link = w(relhdr->sh_link);
-	Elf64_Shdr const *const symsec = &shdr0[symsec_sh_link];
-	Elf64_Sym const *const sym0 = (Elf64_Sym const *)(w8(symsec->sh_offset)
-		+ (void *)ehdr);
-
-	Elf64_Shdr const *const strsec = &shdr0[w(symsec->sh_link)];
-	char const *const str0 = (char const *)(w8(strsec->sh_offset)
-		+ (void *)ehdr);
-
-	Elf64_Rel const *const rel0 = (Elf64_Rel const *)(w8(relhdr->sh_offset)
-		+ (void *)ehdr);
-	unsigned rel_entsize = w8(relhdr->sh_entsize);
-	unsigned const nrel = w8(relhdr->sh_size) / rel_entsize;
-	Elf64_Rel const *relp = rel0;
-
-	unsigned mcountsym = 0;
-	unsigned t;
-
-	for (t = nrel; 0 != t; --t) {
-		if (!mcountsym) {
-			Elf64_Sym const *const symp =
-				&sym0[ELF64_R_SYM(w8(relp->r_info))];
-			char const *symname = &str0[w(symp->st_name)];
-
-			if ('.' == symname[0])
-				++symname;  /* ppc64 hack */
-			if (0 == strcmp((('_' == gpfx) ? "_mcount" : "mcount"),
-					symname))
-				mcountsym = ELF64_R_SYM(w8(relp->r_info));
-		}
-
-		if (mcountsym == ELF64_R_SYM(w8(relp->r_info))) {
-			uint64_t const addend = w8(w8(relp->r_offset) - recval);
-
-			mrelp->r_offset = w8(offbase
-				+ ((void *)mlocp - (void *)mloc0));
-			mrelp->r_info = w8(ELF64_R_INFO(recsym, reltype));
-			if (sizeof(Elf64_Rela) == rel_entsize) {
-				((Elf64_Rela *)mrelp)->r_addend = addend;
-				*mlocp++ = 0;
-			} else
-				*mlocp++ = addend;
-
-			mrelp = (Elf64_Rel *)(rel_entsize + (void *)mrelp);
-		}
-		relp = (Elf64_Rel const *)(rel_entsize + (void *)relp);
-	}
-	*mrelpp = mrelp;
-
-	return mlocp;
-}
-
-/*
- * Find a symbol in the given section, to be used as the base for relocating
- * the table of offsets of calls to mcount.  A local or global symbol suffices,
- * but avoid a Weak symbol because it may be overridden; the change in value
- * would invalidate the relocations of the offsets of the calls to mcount.
- * Often the found symbol will be the unnamed local symbol generated by
- * GNU 'as' for the start of each section.  For example:
- *    Num:    Value  Size Type    Bind   Vis      Ndx Name
- *      2: 00000000     0 SECTION LOCAL  DEFAULT    1
- */
-static unsigned find32_secsym_ndx(unsigned const txtndx,
-				  char const *const txtname,
-				  uint32_t *const recvalp,
-				  Elf32_Shdr const *const symhdr,
-				  Elf32_Ehdr const *const ehdr)
-{
-	Elf32_Sym const *const sym0 = (Elf32_Sym const *)(w(symhdr->sh_offset)
-		+ (void *)ehdr);
-	unsigned const nsym = w(symhdr->sh_size) / w(symhdr->sh_entsize);
-	Elf32_Sym const *symp;
-	unsigned t;
-
-	for (symp = sym0, t = nsym; t; --t, ++symp) {
-		unsigned int const st_bind = ELF32_ST_BIND(symp->st_info);
-
-		if (txtndx == w2(symp->st_shndx)
-			/* avoid STB_WEAK */
-		    && (STB_LOCAL == st_bind || STB_GLOBAL == st_bind)) {
-			*recvalp = w(symp->st_value);
-			return symp - sym0;
-		}
-	}
-	fprintf(stderr, "Cannot find symbol for section %d: %s.\n",
-		txtndx, txtname);
-	fail_file();
-}
-
-static unsigned find64_secsym_ndx(unsigned const txtndx,
-				  char const *const txtname,
-				  uint64_t *const recvalp,
-				  Elf64_Shdr const *const symhdr,
-				  Elf64_Ehdr const *const ehdr)
-{
-	Elf64_Sym const *const sym0 = (Elf64_Sym const *)(w8(symhdr->sh_offset)
-		+ (void *)ehdr);
-	unsigned const nsym = w8(symhdr->sh_size) / w8(symhdr->sh_entsize);
-	Elf64_Sym const *symp;
-	unsigned t;
-
-	for (symp = sym0, t = nsym; t; --t, ++symp) {
-		unsigned int const st_bind = ELF64_ST_BIND(symp->st_info);
-
-		if (txtndx == w2(symp->st_shndx)
-			/* avoid STB_WEAK */
-		    && (STB_LOCAL == st_bind || STB_GLOBAL == st_bind)) {
-			*recvalp = w8(symp->st_value);
-			return symp - sym0;
-		}
-	}
-	fprintf(stderr, "Cannot find symbol for section %d: %s.\n",
-		txtndx, txtname);
-	fail_file();
-}
-
-/*
- * Evade ISO C restriction: no declaration after statement in
- * has32_rel_mcount.
- */
-static char const *
-__has32_rel_mcount(Elf32_Shdr const *const relhdr,  /* is SHT_REL or SHT_RELA */
-		   Elf32_Shdr const *const shdr0,
-		   char const *const shstrtab,
-		   char const *const fname)
-{
-	/* .sh_info depends on .sh_type == SHT_REL[,A] */
-	Elf32_Shdr const *const txthdr = &shdr0[w(relhdr->sh_info)];
-	char const *const txtname = &shstrtab[w(txthdr->sh_name)];
-
-	if (0 == strcmp("__mcount_loc", txtname)) {
-		fprintf(stderr, "warning: __mcount_loc already exists: %s\n",
-			fname);
-		succeed_file();
-	}
-	if (SHT_PROGBITS != w(txthdr->sh_type) ||
-	    !is_mcounted_section_name(txtname))
-		return NULL;
-	return txtname;
-}
-
-static char const *has32_rel_mcount(Elf32_Shdr const *const relhdr,
-				    Elf32_Shdr const *const shdr0,
-				    char const *const shstrtab,
-				    char const *const fname)
-{
-	if (SHT_REL  != w(relhdr->sh_type) && SHT_RELA != w(relhdr->sh_type))
-		return NULL;
-	return __has32_rel_mcount(relhdr, shdr0, shstrtab, fname);
-}
-
-static char const *__has64_rel_mcount(Elf64_Shdr const *const relhdr,
-				      Elf64_Shdr const *const shdr0,
-				      char const *const shstrtab,
-				      char const *const fname)
-{
-	/* .sh_info depends on .sh_type == SHT_REL[,A] */
-	Elf64_Shdr const *const txthdr = &shdr0[w(relhdr->sh_info)];
-	char const *const txtname = &shstrtab[w(txthdr->sh_name)];
-
-	if (0 == strcmp("__mcount_loc", txtname)) {
-		fprintf(stderr, "warning: __mcount_loc already exists: %s\n",
-			fname);
-		succeed_file();
-	}
-	if (SHT_PROGBITS != w(txthdr->sh_type) ||
-	    !is_mcounted_section_name(txtname))
-		return NULL;
-	return txtname;
-}
-
-static char const *has64_rel_mcount(Elf64_Shdr const *const relhdr,
-				    Elf64_Shdr const *const shdr0,
-				    char const *const shstrtab,
-				    char const *const fname)
-{
-	if (SHT_REL  != w(relhdr->sh_type) && SHT_RELA != w(relhdr->sh_type))
-		return NULL;
-	return __has64_rel_mcount(relhdr, shdr0, shstrtab, fname);
-}
-
-static unsigned tot32_relsize(Elf32_Shdr const *const shdr0,
-			      unsigned nhdr,
-			      const char *const shstrtab,
-			      const char *const fname)
-{
-	unsigned totrelsz = 0;
-	Elf32_Shdr const *shdrp = shdr0;
-	for (; 0 != nhdr; --nhdr, ++shdrp) {
-		if (has32_rel_mcount(shdrp, shdr0, shstrtab, fname))
-			totrelsz += w(shdrp->sh_size);
-	}
-	return totrelsz;
-}
-
-static unsigned tot64_relsize(Elf64_Shdr const *const shdr0,
-			      unsigned nhdr,
-			      const char *const shstrtab,
-			      const char *const fname)
-{
-	unsigned totrelsz = 0;
-	Elf64_Shdr const *shdrp = shdr0;
-
-	for (; nhdr; --nhdr, ++shdrp) {
-		if (has64_rel_mcount(shdrp, shdr0, shstrtab, fname))
-			totrelsz += w8(shdrp->sh_size);
-	}
-	return totrelsz;
-}
-
-/* Overall supervision for Elf32 ET_REL file. */
-static void
-do32(Elf32_Ehdr *const ehdr, char const *const fname, unsigned const reltype)
-{
-	Elf32_Shdr *const shdr0 = (Elf32_Shdr *)(w(ehdr->e_shoff)
-		+ (void *)ehdr);
-	unsigned const nhdr = w2(ehdr->e_shnum);
-	Elf32_Shdr *const shstr = &shdr0[w2(ehdr->e_shstrndx)];
-	char const *const shstrtab = (char const *)(w(shstr->sh_offset)
-		+ (void *)ehdr);
-
-	Elf32_Shdr const *relhdr;
-	unsigned k;
-
-	/* Upper bound on space: assume all relevant relocs are for mcount. */
-	unsigned const totrelsz = tot32_relsize(shdr0, nhdr, shstrtab, fname);
-	Elf32_Rel *const mrel0 = umalloc(totrelsz);
-	Elf32_Rel *      mrelp = mrel0;
-
-	/* 2*sizeof(address) <= sizeof(Elf32_Rel) */
-	uint32_t *const mloc0 = umalloc(totrelsz>>1);
-	uint32_t *      mlocp = mloc0;
-
-	unsigned rel_entsize = 0;
-	unsigned symsec_sh_link = 0;
-
-	for (relhdr = shdr0, k = nhdr; k; --k, ++relhdr) {
-		char const *const txtname = has32_rel_mcount(relhdr, shdr0,
-			shstrtab, fname);
-		if (txtname) {
-			uint32_t recval = 0;
-			unsigned const recsym = find32_secsym_ndx(
-				w(relhdr->sh_info), txtname, &recval,
-				&shdr0[symsec_sh_link = w(relhdr->sh_link)],
-				ehdr);
-
-			rel_entsize = w(relhdr->sh_entsize);
-			mlocp = sift32_rel_mcount(mlocp,
-				(void *)mlocp - (void *)mloc0, &mrelp,
-				relhdr, ehdr, recsym, recval, reltype);
-		}
-	}
-	if (mloc0 != mlocp) {
-		append32(ehdr, shstr, mloc0, mlocp, mrel0, mrelp,
-			rel_entsize, symsec_sh_link);
-	}
-	free(mrel0);
-	free(mloc0);
-}
-
-static void
-do64(Elf64_Ehdr *const ehdr, char const *const fname, unsigned const reltype)
-{
-	Elf64_Shdr *const shdr0 = (Elf64_Shdr *)(w8(ehdr->e_shoff)
-		+ (void *)ehdr);
-	unsigned const nhdr = w2(ehdr->e_shnum);
-	Elf64_Shdr *const shstr = &shdr0[w2(ehdr->e_shstrndx)];
-	char const *const shstrtab = (char const *)(w8(shstr->sh_offset)
-		+ (void *)ehdr);
-
-	Elf64_Shdr const *relhdr;
-	unsigned k;
-
-	/* Upper bound on space: assume all relevant relocs are for mcount. */
-	unsigned const totrelsz = tot64_relsize(shdr0, nhdr, shstrtab, fname);
-	Elf64_Rel *const mrel0 = umalloc(totrelsz);
-	Elf64_Rel *      mrelp = mrel0;
-
-	/* 2*sizeof(address) <= sizeof(Elf64_Rel) */
-	uint64_t *const mloc0 = umalloc(totrelsz>>1);
-	uint64_t *      mlocp = mloc0;
-
-	unsigned rel_entsize = 0;
-	unsigned symsec_sh_link = 0;
-
-	for ((relhdr = shdr0), k = nhdr; k; --k, ++relhdr) {
-		char const *const txtname = has64_rel_mcount(relhdr, shdr0,
-			shstrtab, fname);
-		if (txtname) {
-			uint64_t recval = 0;
-			unsigned const recsym = find64_secsym_ndx(
-				w(relhdr->sh_info), txtname, &recval,
-				&shdr0[symsec_sh_link = w(relhdr->sh_link)],
-				ehdr);
-
-			rel_entsize = w8(relhdr->sh_entsize);
-			mlocp = sift64_rel_mcount(mlocp,
-				(void *)mlocp - (void *)mloc0, &mrelp,
-				relhdr, ehdr, recsym, recval, reltype);
-		}
-	}
-	if (mloc0 != mlocp) {
-		append64(ehdr, shstr, mloc0, mlocp, mrel0, mrelp,
-			rel_entsize, symsec_sh_link);
-	}
-	free(mrel0);
-	free(mloc0);
-}
+/* 32 bit and 64 bit are very similar */
+#include "recordmcount.h"
+#define RECORD_MCOUNT_64
+#include "recordmcount.h"
 
 static void
 do_file(char const *const fname)
scripts/recordmcount.h new file +366 -0

366 lines not on this page · raw message

Signature · 1 line
1.7.1