From dcfd7fed7088871f46d9da7183e485877fb2d81f Mon Sep 17 00:00:00 2001 From: "Frank Ch. Eigler" Date: Wed, 1 Apr 2009 22:50:47 -0400 Subject: PR10019: --skip-badvars to suppress run-time memory errors too * NEWS: Note this change. * hash.cxx (find_script_hash): Add s.skip_badvars into hash. * translate.cxx (translate_pass): Emit STP_SKIP_BADVARS. * runtime/loc2c-runtime.h (DEREF_FAULT, STORE_DEREF_FAULT): Provide dummy implementation if STP_SKIP_BADVARS. --- translate.cxx | 2 ++ 1 file changed, 2 insertions(+) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index 47fffd1e..9085349e 100644 --- a/translate.cxx +++ b/translate.cxx @@ -4955,6 +4955,8 @@ translate_pass (systemtap_session& s) s.op->newline() << "#define STP_OVERLOAD"; s.op->newline() << "#endif"; + s.op->newline() << "#define STP_SKIP_BADVARS " << (s.skip_badvars ? 1 : 0); + if (s.bulk_mode) s.op->newline() << "#define STP_BULKMODE"; -- cgit From 093421c6ba933754aa52cf6399aae5b8f86d86d9 Mon Sep 17 00:00:00 2001 From: William Cohen Date: Thu, 2 Apr 2009 17:41:20 -0400 Subject: Reorder includes so regs.c and regs-ia64.c included before task_finder.c. --- translate.cxx | 2 -- 1 file changed, 2 deletions(-) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index 9085349e..cab37487 100644 --- a/translate.cxx +++ b/translate.cxx @@ -4967,9 +4967,7 @@ translate_pass (systemtap_session& s) s.op->newline() << "#define STP_PERFMON"; s.op->newline() << "#include \"runtime.h\""; - s.op->newline() << "#include \"regs.c\""; s.op->newline() << "#include \"stack.c\""; - s.op->newline() << "#include \"regs-ia64.c\""; s.op->newline() << "#include \"stat.c\""; s.op->newline() << "#include "; s.op->newline() << "#include "; -- cgit From 3e3bd7b6b9dd2ba282990f39d60e3ad5ecfec023 Mon Sep 17 00:00:00 2001 From: Josh Stone Date: Mon, 6 Apr 2009 16:11:30 -0700 Subject: PR10026: Read marker/tracepoint args directly We already stash the context variables for markers and tracepoints into the locals for the probe body, but then we were using separate functions to read those locals for each particular probe body. This patch instead teaches the unparser how to emit the local name directly for those context variables. The resulting code from the translator is much simpler now. --- translate.cxx | 7 ++++++- 1 file changed, 6 insertions(+), 1 deletion(-) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index cab37487..c42097bb 100644 --- a/translate.cxx +++ b/translate.cxx @@ -918,6 +918,7 @@ c_unparser::emit_common_header () ostringstream oss; oss << "c->statp = & time_" << dp->basest()->name << ";" << endl; // -t anti-dupe oss << "# needs_global_locks: " << dp->needs_global_locks () << endl; + dp->print_dupe_stamp (oss); dp->body->print(oss); // NB: dependent probe conditions *could* be listed here, but don't need to be. // That's because they're only dependent on the probe body, which is already @@ -1507,6 +1508,7 @@ c_unparser::emit_probe (derived_probe* v) // be very different with or without -t. oss << "c->statp = & time_" << v->basest()->name << ";" << endl; + v->print_dupe_stamp (oss); v->body->print(oss); // Since the generated C changes based on whether or not the probe @@ -3488,7 +3490,10 @@ c_unparser_assignment::visit_symbol (symbol *e) void c_unparser::visit_target_symbol (target_symbol* e) { - throw semantic_error("cannot translate general target-symbol expression", e->tok); + if (!e->probe_context_var.empty()) + o->line() << "l->" << e->probe_context_var; + else + throw semantic_error("cannot translate general cast expression", e->tok); } -- cgit From dc31eb39ab70d9d6b81d1ab02fd49795a4d8f2d0 Mon Sep 17 00:00:00 2001 From: Mark Wielaard Date: Wed, 8 Apr 2009 12:35:01 +0200 Subject: Omit symbols that have suspicious addresses (before base) from symbol table. * translate.cxx (dump_unwindsyms): Filter out sym.st_value < base values. --- translate.cxx | 7 +++++-- 1 file changed, 5 insertions(+), 2 deletions(-) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index c42097bb..3442703d 100644 --- a/translate.cxx +++ b/translate.cxx @@ -4576,10 +4576,13 @@ dump_unwindsyms (Dwfl_Module *m, // PC's, so we omit undefined or "fake" absolute addresses. // These fake absolute addresses occur in some older i386 // kernels to indicate they are vDSO symbols, not real - // functions in the kernel. + // functions in the kernel. We also omit symbols that have + // suspicious addresses (before base). if ((GELF_ST_TYPE (sym.st_info) == STT_FUNC || GELF_ST_TYPE (sym.st_info) == STT_OBJECT) // PR10000: also need .data - && !(sym.st_shndx == SHN_UNDEF || sym.st_shndx == SHN_ABS)) + && !(sym.st_shndx == SHN_UNDEF + || sym.st_shndx == SHN_ABS + || sym.st_value < base)) { Dwarf_Addr sym_addr = sym.st_value; const char *secname = NULL; -- cgit From d6377d4452bb469ea97dc6747cd02cfb1df4af27 Mon Sep 17 00:00:00 2001 From: Mark Wielaard Date: Thu, 9 Apr 2009 23:18:46 +0200 Subject: Cleanup symbol table sanity checking. * translate.cxx (dump_unwindsyms): Get and check against module end, only check STT_FUNC && SHN_ABS for kernel, check shndxp for non-allocated section. --- translate.cxx | 45 +++++++++++++++++++++++++++++---------------- 1 file changed, 29 insertions(+), 16 deletions(-) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index 3442703d..a22e9a5b 100644 --- a/translate.cxx +++ b/translate.cxx @@ -4486,7 +4486,7 @@ dump_unwindsyms (Dwfl_Module *m, // In the future, we'll also care about data symbols. int syments = dwfl_module_getsymtab(m); - assert(syments); + dwfl_assert ("Getting symbol table for " + modname, syments >= 0); //extract build-id from debuginfo file int build_id_len = 0; @@ -4532,6 +4532,10 @@ dump_unwindsyms (Dwfl_Module *m, } } + // Use end as sanity check when resolving symbol addresses. + Dwarf_Addr end; + dwfl_module_info (m, NULL, NULL, &end, NULL, NULL, NULL, NULL); + // Look up the relocation basis for symbols int n = dwfl_module_relocations (m); @@ -4546,10 +4550,11 @@ dump_unwindsyms (Dwfl_Module *m, Dwarf_Addr extra_offset = 0; - for (int i = 1 /* XXX: why not 0? */ ; i < syments; ++i) + for (int i = 0; i < syments; ++i) { GElf_Sym sym; - const char *name = dwfl_module_getsym(m, i, &sym, NULL); + GElf_Word shndxp; + const char *name = dwfl_module_getsym(m, i, &sym, &shndxp); if (name) { // NB: Yey, we found the kernel's _stext value. @@ -4572,17 +4577,15 @@ dump_unwindsyms (Dwfl_Module *m, clog << "Found kernel _stext extra offset 0x" << hex << extra_offset << dec << endl; } - // We only need the function symbols to identify kernel-mode - // PC's, so we omit undefined or "fake" absolute addresses. - // These fake absolute addresses occur in some older i386 - // kernels to indicate they are vDSO symbols, not real - // functions in the kernel. We also omit symbols that have - // suspicious addresses (before base). + // We are only interested in "real" symbols. + // We omit symbols that have suspicious addresses (before base, + // or after end). if ((GELF_ST_TYPE (sym.st_info) == STT_FUNC || GELF_ST_TYPE (sym.st_info) == STT_OBJECT) // PR10000: also need .data - && !(sym.st_shndx == SHN_UNDEF - || sym.st_shndx == SHN_ABS - || sym.st_value < base)) + && !(sym.st_shndx == SHN_UNDEF // Value undefined, + || shndxp == (GElf_Word) -1 // in a non-allocated section, + || sym.st_value >= end // beyond current module, + || sym.st_value < base)) // before first section. { Dwarf_Addr sym_addr = sym.st_value; const char *secname = NULL; @@ -4607,6 +4610,16 @@ dump_unwindsyms (Dwfl_Module *m, { // This is a symbol within a (possibly relocatable) // kernel image. + + // We only need the function symbols to identify kernel-mode + // PC's, so we omit undefined or "fake" absolute addresses. + // These fake absolute addresses occur in some older i386 + // kernels to indicate they are vDSO symbols, not real + // functions in the kernel. We also omit symbols that have + if (GELF_ST_TYPE (sym.st_info) == STT_FUNC + && sym.st_shndx == SHN_ABS) + continue; + secname = "_stext"; // NB: don't subtract session.sym_stext, which could be inconveniently NULL. // Instead, sym_addr will get compensated later via extra_offset. @@ -4666,10 +4679,10 @@ dump_unwindsyms (Dwfl_Module *m, // There would be only a small benefit to warning. A user // likely can't do anything about this; backtraces for the // affected module would just get all icky heuristicy. -#if 0 - c->session.print_warning ("No unwind data for " + modname - + ", " + dwfl_errmsg (-1)); -#endif + // So only report in verbose mode. + if (c->session.verbose > 2) + c->session.print_warning ("No unwind data for " + modname + + ", " + dwfl_errmsg (-1)); } for (unsigned secidx = 0; secidx < seclist.size(); secidx++) -- cgit From 60ad8ebae62f472a9f089f51053a2d0d66c67a95 Mon Sep 17 00:00:00 2001 From: Mark Wielaard Date: Tue, 14 Apr 2009 19:30:56 +0200 Subject: Keep track of relocation section sizes. * runtime/sym.h (_stp_section): Add size field. * translate.cxx (dump_unwindsyms): Get start of module address space, turn seclist into vector of secname, size pairs, track relocation section size, or add module address range if no sections, output size in _stp_section list. --- translate.cxx | 58 +++++++++++++++++++++++++++++++++++++++++++--------------- 1 file changed, 43 insertions(+), 15 deletions(-) (limited to 'translate.cxx') diff --git a/translate.cxx b/translate.cxx index a22e9a5b..d81287ee 100644 --- a/translate.cxx +++ b/translate.cxx @@ -4532,9 +4532,14 @@ dump_unwindsyms (Dwfl_Module *m, } } - // Use end as sanity check when resolving symbol addresses. - Dwarf_Addr end; - dwfl_module_info (m, NULL, NULL, &end, NULL, NULL, NULL, NULL); + // Get the canonical path of the main file for comparison at runtime. + // When given directly by the user through -d or in case of the kernel + // name and path might differ. path should be used for matching. + // Use end as sanity check when resolving symbol addresses and to + // calculate size for .dynamic and .absolute sections. + const char *mainfile; + Dwarf_Addr start, end; + dwfl_module_info (m, NULL, &start, &end, NULL, NULL, &mainfile, NULL); // Look up the relocation basis for symbols int n = dwfl_module_relocations (m); @@ -4545,7 +4550,8 @@ dump_unwindsyms (Dwfl_Module *m, // XXX: unfortunate duplication with tapsets.cxx:emit_address() typedef map addrmap_t; // NB: plain map, sorted by address - vector seclist; // encountered relocation bases (section names) + vector > seclist; // encountered relocation bases + // (section names and sizes) map addrmap; // per-relocation-base sorted addrmap Dwarf_Addr extra_offset = 0; @@ -4588,11 +4594,11 @@ dump_unwindsyms (Dwfl_Module *m, || sym.st_value < base)) // before first section. { Dwarf_Addr sym_addr = sym.st_value; + Dwarf_Addr save_addr = sym_addr; const char *secname = NULL; if (n > 0) // only try to relocate if there exist relocation bases { - Dwarf_Addr save_addr = sym_addr; int ki = dwfl_module_relocate_address (m, &sym_addr); dwfl_assert ("dwfl_module_relocate_address", ki >= 0); secname = dwfl_module_relocation_info (m, ki, NULL); @@ -4645,10 +4651,31 @@ dump_unwindsyms (Dwfl_Module *m, // Compute our section number unsigned secidx; for (secidx=0; secidxsh_size; + } + seclist.push_back (make_pair(secname,size)); + } (addrmap[secidx])[sym_addr] = name; } @@ -4709,12 +4736,17 @@ dump_unwindsyms (Dwfl_Module *m, } c->output << "static struct _stp_section _stp_module_" << stpmod_idx<< "_sections[] = {\n"; + // For the kernel, executables (ET_EXEC) or shared libraries (ET_DYN) + // there is just one section that covers the whole address space of + // the module. For kernel modules (ET_REL) there can be multiple + // sections that get relocated separately. for (unsigned secidx = 0; secidx < seclist.size(); secidx++) { c->output << "{\n" - << ".name = " << lex_cast_qstring(seclist[secidx]) << ",\n" + << ".name = " << lex_cast_qstring(seclist[secidx].first) << ",\n" + << ".size = 0x" << hex << seclist[secidx].second << dec << ",\n" << ".symbols = _stp_module_" << stpmod_idx << "_symbols_" << secidx << ",\n" - << ".num_symbols = sizeof(_stp_module_" << stpmod_idx << "_symbols_" << secidx << ")/sizeof(struct _stp_symbol)\n" + << ".num_symbols = " << addrmap[secidx].size() << "\n" << "},\n"; } c->output << "};\n"; @@ -4722,11 +4754,6 @@ dump_unwindsyms (Dwfl_Module *m, c->output << "static struct _stp_module _stp_module_" << stpmod_idx << " = {\n"; c->output << ".name = " << lex_cast_qstring (modname) << ", \n"; - // Get the canonical path of the main file for comparison at runtime. - // When given directly by the user through -d or in case of the kernel - // name and path might differ. path should be used for matching. - const char *mainfile; - dwfl_module_info (m, NULL, NULL, NULL, NULL, NULL, &mainfile, NULL); mainfile = canonicalize_file_name(mainfile); c->output << ".path = " << lex_cast_qstring (mainfile) << ",\n"; @@ -4861,7 +4888,8 @@ emit_symbol_data (systemtap_session& s) { NULL, /* dwfl_linux_kernel_find_elf, */ dwfl_standard_find_debuginfo, - dwfl_offline_section_address, + NULL, /* ET_REL not supported for user space, only ET_EXEC and ET_DYN. + dwfl_offline_section_address, */ (char **) & debuginfo_path }; -- cgit