psymtab.c

Sun Solaris 10 中的 DTrace 组件的源代码。请参看: http://www.sun.com/software/solaris/observability.jsp

	(void) mutex_unlock(&sort_mtx);

	free(syms);
}

/*
 * Build the symbol table for the given mapped file.
 */
void
Pbuild_file_symtab(struct ps_prochandle *P, file_info_t *fptr)
{
	char objectfile[PATH_MAX];
	uint_t i;

	GElf_Ehdr ehdr;
	GElf_Sym s;

	Elf_Data *shdata;
	Elf_Scn *scn;
	Elf *elf;

	struct {
		GElf_Shdr c_shdr;
		Elf_Data *c_data;
		const char *c_name;
	} *cp, *cache = NULL, *dyn = NULL, *plt = NULL, *ctf = NULL;

	if (fptr->file_init)
		return;	/* We've already processed this file */

	/*
	 * Mark the file_info struct as having the symbol table initialized
	 * even if we fail below.  We tried once; we don't try again.
	 */
	fptr->file_init = 1;

	if (elf_version(EV_CURRENT) == EV_NONE) {
		dprintf("libproc ELF version is more recent than libelf\n");
		return;
	}

	if (P->state == PS_DEAD || P->state == PS_IDLE) {
		/*
		 * If we're a not live, we can't open files from the /proc
		 * object directory; we have only the mapping and file names
		 * to guide us.  We prefer the file_lname, but need to handle
		 * the case of it being NULL in order to bootstrap: we first
		 * come here during rd_new() when the only information we have
		 * is interpreter name associated with the AT_BASE mapping.
		 */
		(void) snprintf(objectfile, sizeof (objectfile), "%s",
		    fptr->file_lname ? fptr->file_lname : fptr->file_pname);
	} else {
		(void) snprintf(objectfile, sizeof (objectfile),
		    "/proc/%d/object/%s", (int)P->pid, fptr->file_pname);
	}

	/*
	 * Open the object file, create the elf file, and then get the elf
	 * header and .shstrtab data buffer so we can process sections by
	 * name. If anything goes wrong try to fake up an elf file from
	 * the in-core elf image.
	 */
	if ((fptr->file_fd = open(objectfile, O_RDONLY)) < 0) {
		dprintf("Pbuild_file_symtab: failed to open %s: %s\n",
		    objectfile, strerror(errno));

		if ((elf = fake_elf(P, fptr)) == NULL ||
		    elf_kind(elf) != ELF_K_ELF ||
		    gelf_getehdr(elf, &ehdr) == NULL ||
		    (scn = elf_getscn(elf, ehdr.e_shstrndx)) == NULL ||
		    (shdata = elf_getdata(scn, NULL)) == NULL) {
			dprintf("failed to fake up ELF file\n");
			return;
		}

	} else if ((elf = elf_begin(fptr->file_fd, ELF_C_READ, NULL)) == NULL ||
	    elf_kind(elf) != ELF_K_ELF ||
	    gelf_getehdr(elf, &ehdr) == NULL ||
	    (scn = elf_getscn(elf, ehdr.e_shstrndx)) == NULL ||
	    (shdata = elf_getdata(scn, NULL)) == NULL) {
		dprintf("failed to process ELF file %s: %s\n",
		    objectfile, elf_errmsg(elf_errno()));

		if ((elf = fake_elf(P, fptr)) == NULL ||
		    elf_kind(elf) != ELF_K_ELF ||
		    gelf_getehdr(elf, &ehdr) == NULL ||
		    (scn = elf_getscn(elf, ehdr.e_shstrndx)) == NULL ||
		    (shdata = elf_getdata(scn, NULL)) == NULL) {
			dprintf("failed to fake up ELF file\n");
			goto bad;
		}

	} else if (file_differs(P, elf, fptr)) {
		Elf *newelf;

		/*
		 * Before we get too excited about this elf file, we'll check
		 * its checksum value against the value we have in memory. If
		 * they don't agree, we try to fake up a new elf file and
		 * proceed with that instead.
		 */

		dprintf("ELF file %s (%lx) doesn't match in-core image\n",
		    fptr->file_pname,
		    (ulong_t)fptr->file_map->map_pmap.pr_vaddr);

		if ((newelf = fake_elf(P, fptr)) == NULL ||
		    elf_kind(newelf) != ELF_K_ELF ||
		    gelf_getehdr(newelf, &ehdr) == NULL ||
		    (scn = elf_getscn(newelf, ehdr.e_shstrndx)) == NULL ||
		    (shdata = elf_getdata(scn, NULL)) == NULL) {
			dprintf("failed to fake up ELF file\n");
		} else {
			(void) elf_end(elf);
			elf = newelf;

			dprintf("switched to faked up ELF file\n");
		}
	}

	if ((cache = malloc(ehdr.e_shnum * sizeof (*cache))) == NULL) {
		dprintf("failed to malloc section cache for %s\n", objectfile);
		goto bad;
	}

	dprintf("processing ELF file %s\n", objectfile);
	fptr->file_class = ehdr.e_ident[EI_CLASS];
	fptr->file_etype = ehdr.e_type;
	fptr->file_elf = elf;

	/*
	 * Iterate through each section, caching its section header, data
	 * pointer, and name.  We use this for handling sh_link values below.
	 */
	for (cp = cache + 1, scn = NULL; scn = elf_nextscn(elf, scn); cp++) {
		if (gelf_getshdr(scn, &cp->c_shdr) == NULL)
			goto bad; /* Failed to get section header */

		if ((cp->c_data = elf_getdata(scn, NULL)) == NULL)
			goto bad; /* Failed to get section data */

		if (cp->c_shdr.sh_name >= shdata->d_size)
			goto bad; /* Corrupt section name */

		cp->c_name = (const char *)shdata->d_buf + cp->c_shdr.sh_name;
	}

	/*
	 * Now iterate through the section cache in order to locate info
	 * for the .symtab, .dynsym, .dynamic, .plt, and .SUNW_ctf sections:
	 */
	for (i = 1, cp = cache + 1; i < ehdr.e_shnum; i++, cp++) {
		GElf_Shdr *shp = &cp->c_shdr;

		if (shp->sh_type == SHT_SYMTAB || shp->sh_type == SHT_DYNSYM) {
			sym_tbl_t *symp = shp->sh_type == SHT_SYMTAB ?
			    &fptr->file_symtab : &fptr->file_dynsym;

			/*
			 * It's possible that the we already got the symbol
			 * table from the core file itself. Either the file
			 * differs in which case our faked up elf file will
			 * only contain the dynsym (not the symtab) or the
			 * file matches in which case we'll just be replacing
			 * the symbol table we pulled out of the core file
			 * with an equivalent one. In either case, this
			 * check isn't essential, but it's a good idea.
			 */
			if (symp->sym_data == NULL) {
				symp->sym_data = cp->c_data;
				symp->sym_symn = shp->sh_size / shp->sh_entsize;
				symp->sym_strs =
				    cache[shp->sh_link].c_data->d_buf;
				symp->sym_strsz =
				    cache[shp->sh_link].c_data->d_size;
				symp->sym_hdr = cp->c_shdr;
				symp->sym_strhdr = cache[shp->sh_link].c_shdr;
			}

		} else if (shp->sh_type == SHT_DYNAMIC) {
			dyn = cp;

		} else if (strcmp(cp->c_name, ".plt") == 0) {
			plt = cp;

		} else if (strcmp(cp->c_name, ".SUNW_ctf") == 0) {
			/*
			 * Skip over bogus CTF sections so they don't come back
			 * to haunt us later.
			 */
			if (shp->sh_link == 0 ||
			    shp->sh_link > ehdr.e_shnum ||
			    (cache[shp->sh_link].c_shdr.sh_type != SHT_DYNSYM &&
			    cache[shp->sh_link].c_shdr.sh_type != SHT_SYMTAB)) {
				dprintf("Bad sh_link %d for "
				    "CTF\n", shp->sh_link);
				continue;
			}
			ctf = cp;
		}
	}

	/*
	 * At this point, we've found all the symbol tables we're ever going
	 * to find: the ones in the loop above and possibly the symtab that
	 * was included in the core file. Before we perform any lookups, we
	 * create sorted versions to optimize for lookups.
	 */
	optimize_symtab(&fptr->file_symtab);
	optimize_symtab(&fptr->file_dynsym);

	/*
	 * Fill in the base address of the text mapping for shared libraries.
	 * This allows us to translate symbols before librtld_db is ready.
	 */
	if (fptr->file_etype == ET_DYN) {
		fptr->file_dyn_base = fptr->file_map->map_pmap.pr_vaddr -
		    fptr->file_map->map_pmap.pr_offset;
		dprintf("setting file_dyn_base for %s to %p\n",
		    objectfile, (void *)fptr->file_dyn_base);
	}

	/*
	 * Record the CTF section information in the file info structure.
	 */
	if (ctf != NULL) {
		fptr->file_ctf_off = ctf->c_shdr.sh_offset;
		fptr->file_ctf_size = ctf->c_shdr.sh_size;
		if (ctf->c_shdr.sh_link != 0 &&
		    cache[ctf->c_shdr.sh_link].c_shdr.sh_type == SHT_DYNSYM)
			fptr->file_ctf_dyn = 1;
	}

	if (fptr->file_lo == NULL)
		goto done; /* Nothing else to do if no load object info */

	/*
	 * If the object is a shared library and we have a different rl_base
	 * value, reset file_dyn_base according to librtld_db's information.
	 */
	if (fptr->file_etype == ET_DYN &&
	    fptr->file_lo->rl_base != fptr->file_dyn_base) {
		dprintf("resetting file_dyn_base for %s to %p\n",
		    objectfile, (void *)fptr->file_lo->rl_base);
		fptr->file_dyn_base = fptr->file_lo->rl_base;
	}

	/*
	 * Fill in the PLT information for this file if a PLT symbol is found.
	 */
	if (sym_by_name(&fptr->file_dynsym, "_PROCEDURE_LINKAGE_TABLE_", &s,
	    NULL) != NULL) {
		fptr->file_plt_base = s.st_value + fptr->file_dyn_base;
		fptr->file_plt_size = (plt != NULL) ? plt->c_shdr.sh_size : 0;

		/*
		 * Bring the load object up to date; it is the only way the
		 * user has to access the PLT data. The PLT information in the
		 * rd_loadobj_t is not set in the call to map_iter() (the
		 * callback for rd_loadobj_iter) where we set file_lo.
		 */
		fptr->file_lo->rl_plt_base = fptr->file_plt_base;
		fptr->file_lo->rl_plt_size = fptr->file_plt_size;

		dprintf("PLT found at %p, size = %lu\n",
		    (void *)fptr->file_plt_base, (ulong_t)fptr->file_plt_size);
	}

	/*
	 * Fill in the PLT information.
	 */
	if (dyn != NULL) {
		uintptr_t dynaddr = dyn->c_shdr.sh_addr + fptr->file_dyn_base;
		size_t ndyn = dyn->c_shdr.sh_size / dyn->c_shdr.sh_entsize;
		GElf_Dyn d;

		for (i = 0; i < ndyn; i++) {
			if (gelf_getdyn(dyn->c_data, i, &d) != NULL &&
			    d.d_tag == DT_JMPREL) {
				fptr->file_jmp_rel =
				    d.d_un.d_ptr + fptr->file_dyn_base;
				break;
			}
		}

		dprintf("_DYNAMIC found at %p, %lu entries, DT_JMPREL = %p\n",
		    (void *)dynaddr, (ulong_t)ndyn, (void *)fptr->file_jmp_rel);
	}

done:
	free(cache);
	return;

bad:
	if (cache != NULL)
		free(cache);

	(void) elf_end(elf);
	fptr->file_elf = NULL;
	if (fptr->file_elfmem != NULL) {
		free(fptr->file_elfmem);
		fptr->file_elfmem = NULL;
	}
	(void) close(fptr->file_fd);
	fptr->file_fd = -1;
}

/*
 * Given a process virtual address, return the map_info_t containing it.
 * If none found, return NULL.
 */
map_info_t *
Paddr2mptr(struct ps_prochandle *P, uintptr_t addr)
{
	int lo = 0;
	int hi = P->map_count - 1;
	int mid;
	map_info_t *mp;

	while (lo <= hi) {

		mid = (lo + hi) / 2;
		mp = &P->mappings[mid];

		/* check that addr is in [vaddr, vaddr + size) */
		if ((addr - mp->map_pmap.pr_vaddr) < mp->map_pmap.pr_size)
			return (mp);

		if (addr < mp->map_pmap.pr_vaddr)
			hi = mid - 1;
		else
			lo = mid + 1;
	}

	return (NULL);
}

/*
 * Return the map_info_t for the executable file.
 * If not found, return NULL.
 */
static map_info_t *
exec_map(struct ps_prochandle *P)
{
	uint_t i;
	map_info_t *mptr;
	map_info_t *mold = NULL;
	file_info_t *fptr;
	uintptr_t base;

	for (i = 0, mptr = P->mappings; i < P->map_count; i++, mptr++) {
		if (mptr->map_pmap.pr_mapname[0] == '\0')
			continue;
		if (strcmp(mptr->map_pmap.pr_mapname, "a.out") == 0) {
			if ((fptr = mptr->map_file) != NULL &&
			    fptr->file_lo != NULL) {
				base = fptr->file_lo->rl_base;
				if (base >= mptr->map_pmap.pr_vaddr &&
				    base < mptr->map_pmap.pr_vaddr +
				    mptr->map_pmap.pr_size)	/* text space */
					return (mptr);
				mold = mptr;	/* must be the data */
				continue;
			}
			/* This is a poor way to test for text space */
			if (!(mptr->map_pmap.pr_mflags & MA_EXEC) ||
			    (mptr->map_pmap.pr_mflags & MA_WRITE)) {
				mold = mptr;
				continue;
			}
			return (mptr);
		}
	}

	return (mold);
}

/*
 * Given a shared object name, return the map_info_t for it.  If no matching
 * object is found, return NULL.  Normally, the link maps contain the full
 * object pathname, e.g. /usr/lib/libc.so.1.  We allow the object name to
 * take one of the following forms:
 *
 * 1. An exact match (i.e. a full pathname): "/usr/lib/libc.so.1"
 * 2. An exact basename match: "libc.so.1"
 * 3. An initial basename match up to a '.' suffix: "libc.so" or "libc"
 * 4. The literal string "a.out" is an alias for the executable mapping
 *
 * The third case is a convenience for callers and may not be necessary.
 *
 * As the exact same object name may be loaded on different link maps (see
 * dlmopen(3DL)), we also allow the caller to resolve the object name by
 * specifying a particular link map id.  If lmid is PR_LMID_EVERY, the
 * first matching name will be returned, regardless of the link map id.
 */
static map_info_t *
object_to_map(struct ps_prochandle *P, Lmid_t lmid, const char *objname)
{
	map_info_t *mp;
	file_info_t *fp;
	size_t objlen;
	uint_t i;

	/*
	 * First pass: look for exact matches of the entire pathname or
	 * basename (cases 1 and 2 above):
	 */
	for (i = 0, mp = P->mappings; i < P->map_count; i++, mp++) {

		if (mp->map_pmap.pr_mapname[0] == '\0' ||
		    (fp = mp->map_file) == NULL || fp->file_lname == NULL)
			continue;

		if (lmid != PR_LMID_EVERY &&
		    (fp->file_lo == NULL || lmid != fp->file_lo->rl_lmident))
			continue;

		/*
		 * If we match, return the primary text mapping; otherwise
		 * just return the mapping we matched.
		 */
		if (strcmp(fp->file_lname, objname) == 0 ||
		    strcmp(fp->file_lbase, objname) == 0)
			return (fp->file_map ? fp->file_map : mp);
	}

	objlen = strlen(objname);

	/*
	 * Second pass: look for partial matches (case 3 above):
	 */
	for (i = 0, mp = P->mappings; i < P->map_count; i++, mp++) {

		if (mp->map_pmap.pr_mapname[0] == '\0' ||
		    (fp = mp->map_file) == NULL || fp->file_lname == NULL)
			continue;

		if (lmid != PR_LMID_EVERY &&
		    (fp->file_lo == NULL || lmid != fp->file_lo->rl_lmident))
			continue;

		/*
		 * If we match, return the primary text mapping; otherwise
		 * just return the mapping we matched.
		 */
		if (strncmp(fp->file_lbase, objname, objlen) == 0 &&
		    fp->file_lbase[objlen] == '.')
			return (fp->file_map ? fp->file_map : mp);
	}

	/*
	 * One last check: we allow "a.out" to always alias the executable,
	 * assuming this name was not in use for something else.
	 */
	if ((lmid == PR_LMID_EVERY || lmid == LM_ID_BASE) &&
	    (strcmp(objname, "a.out") == 0))
		return (P->map_exec);

	return (NULL);
}

static map_info_t *
object_name_to_map(struct ps_prochandle *P, Lmid_t lmid, const char *name)
{
	map_info_t *mptr;

	if (!P->info_valid)
		Pupdate_maps(P);

	if (P->map_exec == NULL && ((mptr = Paddr2mptr(P,
	    Pgetauxval(P, AT_ENTRY))) != NULL || (mptr = exec_map(P)) != NULL))
		P->map_exec = mptr;

	if (P->map_ldso == NULL && (mptr = Paddr2mptr(P,
	    Pgetauxval(P, AT_BASE))) != NULL)
		P->map_ldso = mptr;

	if (name == PR_OBJ_EXEC)
		mptr = P->map_exec;
	else if (name == PR_OBJ_LDSO)
		mptr = P->map_ldso;
	else if (Prd_agent(P) != NULL || P->state == PS_IDLE)
		mptr = object_to_map(P, lmid, name);
	else
		mptr = NULL;

	return (mptr);
}

/*
 * When two symbols are found by address, decide which one is to be preferred.
 */
static GElf_Sym *
sym_prefer(GElf_Sym *sym1, char *name1, GElf_Sym *sym2, char *name2)
{
	/*
	 * Prefer the non-NULL symbol.
	 */
	if (sym1 == NULL)
		return (sym2);
	if (sym2 == NULL)
		return (sym1);

	/*
	 * Defer to the sort ordering...
	 */
	return (byaddr_cmp_common(sym1, name1, sym2, name2) <= 0 ? sym1 : sym2);
}

/*
 * Look up a symbol by address in the