From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gnu.wildebeest.org (gnu.wildebeest.org [45.83.234.184]) by sourceware.org (Postfix) with ESMTPS id C99683858404 for ; Wed, 10 Nov 2021 10:16:36 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org C99683858404 Authentication-Results: sourceware.org; dmarc=none (p=none dis=none) header.from=klomp.org Authentication-Results: sourceware.org; spf=pass smtp.mailfrom=klomp.org Received: from reform (deer0x16.wildebeest.org [172.31.17.152]) (using TLSv1.2 with cipher ADH-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by gnu.wildebeest.org (Postfix) with ESMTPSA id ED0A9302FBAB; Wed, 10 Nov 2021 11:16:31 +0100 (CET) Received: by reform (Postfix, from userid 1000) id 76BEE2E8070C; Wed, 10 Nov 2021 11:16:31 +0100 (CET) Date: Wed, 10 Nov 2021 11:16:31 +0100 From: Mark Wielaard To: John Mellor-Crummey Cc: elfutils-devel@sourceware.org, Jonathon Anderson , Xiaozhu Meng Subject: Re: [PATCH] (v2) read inlining info in an NVIDIA extended line map Message-ID: References: <7C166312-4876-444F-9B85-C7E30C8F4959@rice.edu> <37B899F3-FD5F-4F7B-9F1B-BE63CF84F554@rice.edu> <413A462A-D548-452E-9323-5193339C288A@rice.edu> MIME-Version: 1.0 Content-Type: multipart/mixed; boundary="f7R4E6TytW9gt3FW" Content-Disposition: inline Content-Transfer-Encoding: 8bit In-Reply-To: X-Spam-Status: No, score=-10.1 required=5.0 tests=BAYES_00, GIT_PATCH_0, KAM_DMARC_STATUS, KAM_SHORT, SPF_HELO_NONE, SPF_PASS, TXREP, WEIRD_PORT autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: elfutils-devel@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Elfutils-devel mailing list List-Unsubscribe: , List-Archive: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 10 Nov 2021 10:16:42 -0000 --f7R4E6TytW9gt3FW Content-Type: text/plain; charset=utf-8 Content-Disposition: inline Content-Transfer-Encoding: 8bit Hi John, On Thu, Nov 04, 2021 at 04:41:58PM -0500, John Mellor-Crummey via Elfutils-devel wrote: > Here I describe just the improvements to that patch that address Mark’s concerns: > > (1) all of the code for handling NVIDIA DWARF extensions is always > available; there is no special configuration switch needed. > (2) all changes are bracketed by comments that mark them NVIDIA > extensions > (3) the DWARF extended opcodes have been renamed with names that > include NVIDIA in them > (4) the two new API functions to surface the new information have > been improved to separate the interface result from the internal > representation (at Mark’s request) > (4a) the API for extracting the name of an inlined function in a > DWARF line now returns a const char * instead of a string > table index > (4b) the API for extracting an inline “context” now returns a > pointer to a DWARF line where the code is inlined rather > than returning an unsigned int (an index into the line table > that one could use to compute the pointer) > (5) there are test cases for readelf and libdw that use a binary > generated by NVIDIA’s compiler. the test cases include information > about how the binary was generated This is really nice. I did make a few tweaks: - Added your original overview as commit message because it contains all the relevant context and pointers to more information. - Added ChangeLog and NEWS entries, mainly for my own review. - I removed the bracketed comments, I think they cluttered the code and made it seem like we wanted to remove it or disable at some point. I think it should just be considered part of the normal code now. - I removed the NVIDIA_LINEMAP_INLINING_EXTENSIONS define from version.h. If people want they can have a configure check for the new dwarf_linecontext or dwarf_linefunctionname functions or the DW_LNE_NVIDIA_inlined_call or DW_LNE_NVIDIA_set_function_name constants. - I made dwarf_linefunctionname always return NULL on error (not the magic string "???", which is still used in readelf). - Changed the header check to be exactly 4 bytes, so we are sure to be able to read the str offset completely (if it is smaller or larger we cannot handle it). - The new dwarf_linecontext and dwarf_linefunctionname get their own new ELFUTILS_0.186 section in libdw.map because they are introduced with verion 0.186. - The new run-nvidia-extended-linemap-libdw.sh and run-nvidia-extended-linemap-readelf.sh sripts and testfile_nvidia_linemap.bz2 testfile were added to EXTRA_DIST so they show up in a dist tarball. Patch as committed attached. Hope you don't mind the cleanups. We still want to reduce the size of the Dwarf_Line_s and struct line_state (independent of these extensions). I opened a new bug for that: https://sourceware.org/bugzilla/show_bug.cgi?id=28574 Thanks, Mark --f7R4E6TytW9gt3FW Content-Type: text/x-diff; charset=utf-8 Content-Disposition: attachment; filename="0001-libdw-readelf-Read-inlining-info-in-NVIDIA-extended-.patch" Content-Transfer-Encoding: 8bit >From dbd44e96f101a81f387ca6ff46910ab74ed7b1dc Mon Sep 17 00:00:00 2001 From: John M Mellor-Crummey Date: Wed, 20 Oct 2021 18:19:38 -0500 Subject: [PATCH] libdw, readelf: Read inlining info in NVIDIA extended line map MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit As of CUDA 11.2, NVIDIA added extensions to the line map section of CUDA binaries to represent inlined functions. These extensions include - two new fields in a line table row to represent inline information: context, and functionname, - two new DWARF extended opcodes: DW_LNE_NVIDIA_inlined_call, DW_LNE_NVIDIA_set_function_name, - an additional word in the line table header that indicates the offset in the .debug_str function where the function names for this line table begin, and A line table row for an inlined function contains a non-zero "context" value. The “context” field indicates the index of the line table row that serves as the call site for an inlined context. The "functionname" field in a line table row is only meaningful if the "context" field of the row is non-zero. A meaningful "functionname" field contains an index into the .debug_str section relative to the base offset established in the line table header; the position in the .debug_str section indicates the name of the inlined function. These extensions resemble the proposed DWARF extensions (http://dwarfstd.org/ShowIssue.php?issue=140906.1) by Cary Coutant, but are not identical. This commit integrates support for handling NVIDIA's extended line maps into elfutil's libdw library, by adding two functions dwarf_linecontext and dwarf_linefunctionname, and the readelf --debug-dump=line command line utility. Signed-off-by: John M Mellor-Crummey Signed-off-by: Mark Wielaard --- ChangeLog | 4 + NEWS | 5 + libdw/ChangeLog | 19 +++ libdw/Makefile.am | 1 + libdw/dwarf.h | 4 + libdw/dwarf_getsrclines.c | 36 +++- libdw/dwarf_linecontext.c | 45 +++++ libdw/dwarf_linefunctionname.c | 52 ++++++ libdw/libdw.h | 9 + libdw/libdw.map | 6 + libdw/libdwP.h | 3 + src/ChangeLog | 6 + src/readelf.c | 62 +++++++ tests/.gitignore | 1 + tests/ChangeLog | 13 ++ tests/Makefile.am | 5 + tests/nvidia_extended_linemap_libdw.c | 166 +++++++++++++++++++ tests/run-nvidia-extended-linemap-libdw.sh | 60 +++++++ tests/run-nvidia-extended-linemap-readelf.sh | 120 ++++++++++++++ tests/testfile_nvidia_linemap.bz2 | Bin 0 -> 2365 bytes 20 files changed, 615 insertions(+), 2 deletions(-) create mode 100644 libdw/dwarf_linecontext.c create mode 100644 libdw/dwarf_linefunctionname.c create mode 100644 tests/nvidia_extended_linemap_libdw.c create mode 100755 tests/run-nvidia-extended-linemap-libdw.sh create mode 100755 tests/run-nvidia-extended-linemap-readelf.sh create mode 100644 tests/testfile_nvidia_linemap.bz2 diff --git a/ChangeLog b/ChangeLog index 6255fe61..d4b89f9c 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,7 @@ +2021-09-03 John Mellor-Crummey + + * NEWS: Read inlining info in NVIDIA extended line map + 2021-08-10 Adrian Ratiu * configure.ac (AC_CACHE_CHECK): Rework std=gnu99 check to allow clang. diff --git a/NEWS b/NEWS index b812b743..897b391d 100644 --- a/NEWS +++ b/NEWS @@ -16,6 +16,11 @@ debuginfod: Supply extra HTTP response headers, describing archive/file Limit the duration of groom ops roughly to rescan (-t) times. Several other performance improvements & prometheus metrics. +libdw: Support for the NVIDIA Cuda line map extensions. + DW_LNE_NVIDIA_inlined_call and DW_LNE_NVIDIA_set_function_name + are defined in dwarf.h. New functions dwarf_linecontext and + dwarf_linefunctionname + Version 0.185 debuginfod-client: Simplify curl handle reuse so downloads which diff --git a/libdw/ChangeLog b/libdw/ChangeLog index 38e6efb2..ca742e6b 100644 --- a/libdw/ChangeLog +++ b/libdw/ChangeLog @@ -1,3 +1,22 @@ +2021-10-20 John M Mellor-Crummey + + * dwarf_linecontext.c: New file. + * dwarf_linefunctionname.c: Likewise. + * Makefile.am (libdw_a_SOURCES): Add dwarf_linecontext.c and + dwarf_linefunctionname.c + * dwarf.h: Add DW_LNE_NVIDIA_inlined_call and + DW_LNE_NVIDIA_set_function_name. + * dwarf_getsrclines.c (struct line_state): Add context and + function_name fields. + (add_new_line): Set context and function_name. + (MAX_STACK_LINES): Reduce to MAX_STACK_ALLOC / 2. + (read_srclines): Initialize context and function_name. Try to + read debug_str_offset if available. Handle + DW_LNE_NVIDIA_inlined_call and DW_LNE_NVIDIA_set_function_name. + * libdw.h (dwarf_linecontext): New declaration. + (dwarf_linefunctionname): Likewise. + * libdw.map (ELFUTILS_0.186): New section. + 2021-11-08 Mark Wielaard * dwarf_begin_elf.c (scn_dwarf_type): New function. diff --git a/libdw/Makefile.am b/libdw/Makefile.am index 6b7834af..4fda33bd 100644 --- a/libdw/Makefile.am +++ b/libdw/Makefile.am @@ -63,6 +63,7 @@ libdw_a_SOURCES = dwarf_begin.c dwarf_begin_elf.c dwarf_end.c dwarf_getelf.c \ dwarf_linesrc.c dwarf_lineno.c dwarf_lineaddr.c \ dwarf_linecol.c dwarf_linebeginstatement.c \ dwarf_lineendsequence.c dwarf_lineblock.c \ + dwarf_linecontext.c dwarf_linefunctionname.c \ dwarf_lineprologueend.c dwarf_lineepiloguebegin.c \ dwarf_lineisa.c dwarf_linediscriminator.c \ dwarf_lineop_index.c dwarf_line_file.c \ diff --git a/libdw/dwarf.h b/libdw/dwarf.h index 19a4be96..3ce7f236 100644 --- a/libdw/dwarf.h +++ b/libdw/dwarf.h @@ -844,6 +844,10 @@ enum DW_LNE_set_discriminator = 4, DW_LNE_lo_user = 128, + + DW_LNE_NVIDIA_inlined_call = 144, + DW_LNE_NVIDIA_set_function_name = 145, + DW_LNE_hi_user = 255 }; diff --git a/libdw/dwarf_getsrclines.c b/libdw/dwarf_getsrclines.c index 8fc48e1d..2c1d7a40 100644 --- a/libdw/dwarf_getsrclines.c +++ b/libdw/dwarf_getsrclines.c @@ -93,6 +93,8 @@ struct line_state struct linelist *linelist; size_t nlinelist; unsigned int end_sequence; + unsigned int context; + unsigned int function_name; }; static inline void @@ -139,6 +141,8 @@ add_new_line (struct line_state *state, struct linelist *new_line) SET (epilogue_begin); SET (isa); SET (discriminator); + SET (context); + SET (function_name); #undef SET @@ -161,7 +165,7 @@ read_srclines (Dwarf *dbg, the stack. Stack allocate some entries, only dynamically malloc when more than MAX. */ #define MAX_STACK_ALLOC 4096 -#define MAX_STACK_LINES MAX_STACK_ALLOC +#define MAX_STACK_LINES (MAX_STACK_ALLOC / 2) #define MAX_STACK_FILES (MAX_STACK_ALLOC / 4) #define MAX_STACK_DIRS (MAX_STACK_ALLOC / 16) @@ -180,7 +184,9 @@ read_srclines (Dwarf *dbg, .prologue_end = false, .epilogue_begin = false, .isa = 0, - .discriminator = 0 + .discriminator = 0, + .context = 0, + .function_name = 0 }; /* The dirs normally go on the stack, but if there are too many @@ -648,6 +654,13 @@ read_srclines (Dwarf *dbg, } } + unsigned int debug_str_offset = 0; + if (unlikely (linep == header_start + header_length - 4)) + { + /* CUBINs contain an unsigned 4-byte offset */ + debug_str_offset = read_4ubyte_unaligned_inc (dbg, linep); + } + /* Consistency check. */ if (unlikely (linep != header_start + header_length)) { @@ -753,6 +766,8 @@ read_srclines (Dwarf *dbg, state.epilogue_begin = false; state.isa = 0; state.discriminator = 0; + state.context = 0; + state.function_name = 0; break; case DW_LNE_set_address: @@ -831,6 +846,23 @@ read_srclines (Dwarf *dbg, get_uleb128 (state.discriminator, linep, lineendp); break; + case DW_LNE_NVIDIA_inlined_call: + if (unlikely (linep >= lineendp)) + goto invalid_data; + get_uleb128 (state.context, linep, lineendp); + if (unlikely (linep >= lineendp)) + goto invalid_data; + get_uleb128 (state.function_name, linep, lineendp); + state.function_name += debug_str_offset; + break; + + case DW_LNE_NVIDIA_set_function_name: + if (unlikely (linep >= lineendp)) + goto invalid_data; + get_uleb128 (state.function_name, linep, lineendp); + state.function_name += debug_str_offset; + break; + default: /* Unknown, ignore it. */ if (unlikely ((size_t) (lineendp - (linep - 1)) < len)) diff --git a/libdw/dwarf_linecontext.c b/libdw/dwarf_linecontext.c new file mode 100644 index 00000000..84572e22 --- /dev/null +++ b/libdw/dwarf_linecontext.c @@ -0,0 +1,45 @@ +/* Return context in line. + This file is part of elfutils. + Written by John Mellor-Crummey , 2021. + + This file is free software; you can redistribute it and/or modify + it under the terms of either + + * the GNU Lesser General Public License as published by the Free + Software Foundation; either version 3 of the License, or (at + your option) any later version + + or + + * the GNU General Public License as published by the Free + Software Foundation; either version 2 of the License, or (at + your option) any later version + + or both in parallel, as here. + + elfutils is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received copies of the GNU General Public License and + the GNU Lesser General Public License along with this program. If + not, see . */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include "libdwP.h" + + +Dwarf_Line* +dwarf_linecontext (Dwarf_Lines* lines, Dwarf_Line *line) +{ + if (lines == NULL || line == NULL) + return NULL; + if (line->context == 0 || line->context >= lines->nlines) + return NULL; + + return lines->info + (line->context - 1); +} diff --git a/libdw/dwarf_linefunctionname.c b/libdw/dwarf_linefunctionname.c new file mode 100644 index 00000000..e194d212 --- /dev/null +++ b/libdw/dwarf_linefunctionname.c @@ -0,0 +1,52 @@ +/* Return function name in line. + This file is part of elfutils. + Written by John Mellor-Crummey , 2021. + + This file is free software; you can redistribute it and/or modify + it under the terms of either + + * the GNU Lesser General Public License as published by the Free + Software Foundation; either version 3 of the License, or (at + your option) any later version + + or + + * the GNU General Public License as published by the Free + Software Foundation; either version 2 of the License, or (at + your option) any later version + + or both in parallel, as here. + + elfutils is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU + General Public License for more details. + + You should have received copies of the GNU General Public License and + the GNU Lesser General Public License along with this program. If + not, see . */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include +#include "libdwP.h" + + +const char * +dwarf_linefunctionname (Dwarf *dbg, Dwarf_Line *line) +{ + if (dbg == NULL || line == NULL) + return NULL; + if (line->context == 0) + return NULL; + + Elf_Data *str_data = dbg->sectiondata[IDX_debug_str]; + if (str_data == NULL || line->function_name >= str_data->d_size + || memchr (str_data->d_buf + line->function_name, '\0', + str_data->d_size - line->function_name) == NULL) + return NULL; + + return (char *) str_data->d_buf + line->function_name; +} diff --git a/libdw/libdw.h b/libdw/libdw.h index 77174d28..64d1689a 100644 --- a/libdw/libdw.h +++ b/libdw/libdw.h @@ -701,6 +701,15 @@ extern int dwarf_linediscriminator (Dwarf_Line *line, unsigned int *discp) extern const char *dwarf_linesrc (Dwarf_Line *line, Dwarf_Word *mtime, Dwarf_Word *length); +/* Return the caller of this line if inlined. If not inlined, + return NULL. */ +extern Dwarf_Line *dwarf_linecontext (Dwarf_Lines *lines, Dwarf_Line *line); + +/* Return the function name in this line record. If this line is + inlined, this is the name of the function that was inlined. If this line + is not inlined, return NULL. */ +extern const char *dwarf_linefunctionname (Dwarf *dbg, Dwarf_Line *line); + /* Return file information. The returned string is NULL when an error occurred, or the file path. The file path is either absolute or relative to the compilation directory. See dwarf_decl_file. */ diff --git a/libdw/libdw.map b/libdw/libdw.map index 8ab0a2a0..4f530378 100644 --- a/libdw/libdw.map +++ b/libdw/libdw.map @@ -360,3 +360,9 @@ ELFUTILS_0.177 { # presume that NULL is only returned on error (otherwise ELF_K_NONE). dwelf_elf_begin; } ELFUTILS_0.175; + +ELFUTILS_0.186 { + global: + dwarf_linecontext; + dwarf_linefunctionname; +} ELFUTILS_0.177; diff --git a/libdw/libdwP.h b/libdw/libdwP.h index 48f3a943..360ad01a 100644 --- a/libdw/libdwP.h +++ b/libdw/libdwP.h @@ -303,6 +303,9 @@ struct Dwarf_Line_s unsigned int op_index:8; unsigned int isa:8; unsigned int discriminator:24; + /* These are currently only used for the NVIDIA extensions. */ + unsigned int context; + unsigned int function_name; }; struct Dwarf_Lines_s diff --git a/src/ChangeLog b/src/ChangeLog index 316bcb6d..05b2522d 100644 --- a/src/ChangeLog +++ b/src/ChangeLog @@ -1,3 +1,9 @@ +2021-10-20 John M Mellor-Crummey + + * readelf.c (print_debug_line_section): Try to read + debug_str_offset if available. Handle DW_LNE_NVIDIA_inlined_call + and DW_LNE_NVIDIA_set_function_name. + 2021-10-06 Mark Wielaard * elflint.c (check_sections): Don't dereference databits if bad. diff --git a/src/readelf.c b/src/readelf.c index 3a199068..c10038e3 100644 --- a/src/readelf.c +++ b/src/readelf.c @@ -8478,6 +8478,8 @@ print_debug_line_section (Dwfl_Module *dwflmod, Ebl *ebl, GElf_Ehdr *ehdr, header_length = read_8ubyte_unaligned_inc (dbg, linep); } + const unsigned char *header_start = linep; + /* Next the minimum instruction length. */ if ((size_t) (lineendp - linep) < 1) goto invalid_data; @@ -8761,6 +8763,13 @@ print_debug_line_section (Dwfl_Module *dwflmod, Ebl *ebl, GElf_Ehdr *ehdr, ++linep; } + unsigned int debug_str_offset = 0; + if (unlikely (linep == header_start + header_length - 4)) + { + /* CUBINs contain an unsigned 4-byte offset */ + debug_str_offset = read_4ubyte_unaligned_inc (dbg, linep); + } + if (linep == lineendp) { puts (_("\nNo line number statements.")); @@ -8909,6 +8918,59 @@ print_debug_line_section (Dwfl_Module *dwflmod, Ebl *ebl, GElf_Ehdr *ehdr, printf (_(" set discriminator to %u\n"), u128); break; + case DW_LNE_NVIDIA_inlined_call: + { + if (unlikely (linep >= lineendp)) + goto invalid_data; + + unsigned int context; + get_uleb128 (context, linep, lineendp); + + if (unlikely (linep >= lineendp)) + goto invalid_data; + + unsigned int function_name; + get_uleb128 (function_name, linep, lineendp); + function_name += debug_str_offset; + + Elf_Data *str_data = dbg->sectiondata[IDX_debug_str]; + char *function_str; + if (str_data == NULL || function_name >= str_data->d_size + || memchr (str_data->d_buf + function_name, '\0', + str_data->d_size - function_name) == NULL) + function_str = "???"; + else + function_str = (char *) str_data->d_buf + function_name; + + printf (_(" set inlined context %u," + " function name %s (0x%x)\n"), + context, function_str, function_name); + break; + } + + case DW_LNE_NVIDIA_set_function_name: + { + if (unlikely (linep >= lineendp)) + goto invalid_data; + + unsigned int function_name; + get_uleb128 (function_name, linep, lineendp); + function_name += debug_str_offset; + + Elf_Data *str_data = dbg->sectiondata[IDX_debug_str]; + char *function_str; + if (str_data == NULL || function_name >= str_data->d_size + || memchr (str_data->d_buf + function_name, '\0', + str_data->d_size - function_name) == NULL) + function_str = "???"; + else + function_str = (char *) str_data->d_buf + function_name; + + printf (_(" set function name %s (0x%x)\n"), + function_str, function_name); + } + break; + default: /* Unknown, ignore it. */ puts (_(" unknown opcode")); diff --git a/tests/.gitignore b/tests/.gitignore index d0e83da2..99d04819 100644 --- a/tests/.gitignore +++ b/tests/.gitignore @@ -85,6 +85,7 @@ /next-files /next-lines /next_cfi +/nvidia_extended_linemap_libdw /peel_type /rdwrmmap /read_unaligned diff --git a/tests/ChangeLog b/tests/ChangeLog index c5d00027..db8b13bd 100644 --- a/tests/ChangeLog +++ b/tests/ChangeLog @@ -1,3 +1,16 @@ +2021-10-20 John M Mellor-Crummey + + * nvidia_extended_linemap_libdw.c: New file. + * run-nvidia-extended-linemap-libdw.sh: New test. + * run-nvidia-extended-linemap-readelf.sh: Likewise. + * testfile_nvidia_linemap.bz2: New test file. + * .gitignore: Add nvidia_extended_linemap_libdw. + * Makefile.am (check_PROGRAMS): Add nvidia_extended_linemap_libdw. + (TESTS): Add run-nvidia-extended-linemap-libdw.sh and + run-nvidia-extended-linemap-readelf.sh + (EXTRA_DIST): Likewise and testfile_nvidia_linemap.bz2. + (nvidia_extended_linemap_libdw_LDADD): New variable. + 2021-11-08 Mark Wielaard * Makefile.am (TESTS): Add run-readelf-fat-lto.sh. diff --git a/tests/Makefile.am b/tests/Makefile.am index ccc4c052..6d3e75af 100644 --- a/tests/Makefile.am +++ b/tests/Makefile.am @@ -61,6 +61,7 @@ check_PROGRAMS = arextract arsymtest newfile saridx scnnames sectiondump \ dwelf_elf_e_machine_string \ getphdrnum leb128 read_unaligned \ msg_tst system-elf-libelf-test \ + nvidia_extended_linemap_libdw \ $(asm_TESTS) asm_TESTS = asm-tst1 asm-tst2 asm-tst3 asm-tst4 asm-tst5 \ @@ -189,6 +190,7 @@ TESTS = run-arextract.sh run-arsymtest.sh run-ar.sh newfile test-nlist \ leb128 read_unaligned \ msg_tst system-elf-libelf-test \ $(asm_TESTS) run-disasm-bpf.sh run-low_high_pc-dw-form-indirect.sh \ + run-nvidia-extended-linemap-libdw.sh run-nvidia-extended-linemap-readelf.sh \ run-readelf-dw-form-indirect.sh run-strip-largealign.sh if !BIARCH @@ -566,6 +568,8 @@ EXTRA_DIST = run-arextract.sh run-arsymtest.sh run-ar.sh \ run-getphdrnum.sh testfile-phdrs.elf.bz2 \ run-test-includes.sh run-low_high_pc-dw-form-indirect.sh \ run-readelf-dw-form-indirect.sh testfile-dw-form-indirect.bz2 \ + run-nvidia-extended-linemap-libdw.sh run-nvidia-extended-linemap-readelf.sh \ + testfile_nvidia_linemap.bz2 \ testfile-largealign.o.bz2 run-strip-largealign.sh @@ -739,6 +743,7 @@ dwelf_elf_e_machine_string_LDADD = $(libelf) $(libdw) getphdrnum_LDADD = $(libelf) $(libdw) leb128_LDADD = $(libelf) $(libdw) read_unaligned_LDADD = $(libelf) $(libdw) +nvidia_extended_linemap_libdw_LDADD = $(libelf) $(libdw) # We want to test the libelf header against the system elf.h header. # Don't include any -I CPPFLAGS. Except when we install our own elf.h. diff --git a/tests/nvidia_extended_linemap_libdw.c b/tests/nvidia_extended_linemap_libdw.c new file mode 100644 index 00000000..20d8d404 --- /dev/null +++ b/tests/nvidia_extended_linemap_libdw.c @@ -0,0 +1,166 @@ +/* Inspect nvidia extended linemap with dwarf_next_lines. + Copyright (C) 2002, 2004, 2018 Red Hat, Inc. + This file is part of elfutils. + + This file is free software; you can redistribute it and/or modify + it under the terms of the GNU General Public License as published by + the Free Software Foundation; either version 3 of the License, or + (at your option) any later version. + + elfutils is distributed in the hope that it will be useful, but + WITHOUT ANY WARRANTY; without even the implied warranty of + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + GNU General Public License for more details. + + You should have received a copy of the GNU General Public License + along with this program. If not, see . */ + +#ifdef HAVE_CONFIG_H +# include +#endif + +#include +#include +#include +#include ELFUTILS_HEADER(dw) +#include +#include +#include + + +int +main (int argc, char *argv[]) +{ + int result = 0; + int cnt; + + for (cnt = 1; cnt < argc; ++cnt) + { + int fd = open (argv[cnt], O_RDONLY); + + Dwarf *dbg = dwarf_begin (fd, DWARF_C_READ); + if (dbg == NULL) + { + printf ("%s not usable: %s\n", argv[cnt], dwarf_errmsg (-1)); + close (fd); + continue; + } + + Dwarf_Off off; + Dwarf_Off next_off = 0; + Dwarf_CU *cu = NULL; + Dwarf_Lines *lb; + size_t nlb; + int res; + while ((res = dwarf_next_lines (dbg, off = next_off, &next_off, &cu, + NULL, NULL, &lb, &nlb)) == 0) + { + printf ("off = %" PRIu64 "\n", off); + printf (" %zu lines\n", nlb); + + for (size_t i = 0; i < nlb; ++i) + { + Dwarf_Line *l = dwarf_onesrcline (lb, i); + if (l == NULL) + { + printf ("%s: cannot get individual line\n", argv[cnt]); + result = 1; + break; + } + + Dwarf_Addr addr; + if (dwarf_lineaddr (l, &addr) != 0) + addr = 0; + const char *file = dwarf_linesrc (l, NULL, NULL); + int line; + if (dwarf_lineno (l, &line) != 0) + line = 0; + + printf ("%" PRIx64 ": %s:%d:", (uint64_t) addr, + file ?: "???", line); + + /* Getting the file path through the Dwarf_Files should + result in the same path. */ + Dwarf_Files *files; + size_t idx; + if (dwarf_line_file (l, &files, &idx) != 0) + { + printf ("%s: cannot get file from line (%zd): %s\n", + argv[cnt], i, dwarf_errmsg (-1)); + result = 1; + break; + } + const char *path = dwarf_filesrc (files, idx, NULL, NULL); + if ((path == NULL && file != NULL) + || (path != NULL && file == NULL) + || (strcmp (file, path) != 0)) + { + printf ("%s: line %zd srcline (%s) != file srcline (%s)\n", + argv[cnt], i, file ?: "???", path ?: "???"); + result = 1; + break; + } + + int column; + if (dwarf_linecol (l, &column) != 0) + column = 0; + if (column >= 0) + printf ("%d:", column); + + bool is_stmt; + if (dwarf_linebeginstatement (l, &is_stmt) != 0) + is_stmt = false; + bool end_sequence; + if (dwarf_lineendsequence (l, &end_sequence) != 0) + end_sequence = false; + bool basic_block; + if (dwarf_lineblock (l, &basic_block) != 0) + basic_block = false; + bool prologue_end; + if (dwarf_lineprologueend (l, &prologue_end) != 0) + prologue_end = false; + bool epilogue_begin; + if (dwarf_lineepiloguebegin (l, &epilogue_begin) != 0) + epilogue_begin = false; + printf (" is_stmt:%s, end_seq:%s, bb:%s, prologue:%s, epilogue:%s\n", + is_stmt ? "yes" : "no", end_sequence ? "yes" : "no", + basic_block ? "yes" : "no", prologue_end ? "yes" : "no", + epilogue_begin ? "yes" : "no"); + + Dwarf_Line* callee_context = l; + Dwarf_Line* caller_context = dwarf_linecontext (lb, l); + unsigned int depth = 0; + while (caller_context != NULL) + { + depth++; + for (unsigned int x = 0; x < depth; x++) + printf (" "); + + const char *inlined_file = dwarf_linesrc (caller_context, + NULL, NULL); + int inlined_line; + if (dwarf_lineno (caller_context, &inlined_line) != 0) + inlined_line = 0; + + printf ("%s inlined at %s:%d\n", + dwarf_linefunctionname(dbg, callee_context), + inlined_file ?: "???", inlined_line); + + callee_context = caller_context; + caller_context = dwarf_linecontext (lb, callee_context); + } + } + } + + if (res < 0) + { + printf ("dwarf_next_lines failed: %s\n", dwarf_errmsg (-1)); + result = 1; + } + + dwarf_end (dbg); + close (fd); + } + + return result; +} diff --git a/tests/run-nvidia-extended-linemap-libdw.sh b/tests/run-nvidia-extended-linemap-libdw.sh new file mode 100755 index 00000000..d1df2cf3 --- /dev/null +++ b/tests/run-nvidia-extended-linemap-libdw.sh @@ -0,0 +1,60 @@ +# Copyright (C) 2011 Red Hat, Inc. +# This file is part of elfutils. +# +# This file is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 3 of the License, or +# (at your option) any later version. +# +# elfutils is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +. $srcdir/test-subr.sh + +# NOTE: +# the file testfile_nvidia_linemap is a CUDA binary for an NVIDIA A100 generated as follows using CUDA 11.2 +# nvcc -o main main.cu -Xcompiler "-g -fopenmp" -O3 -lineinfo -arch sm_80 -lcudart -lcuda -lstdc++ -lm +# cuobjdump -xelf all main +# mv main.sm_80.cubin testfile_nvidia_linemap + +testfiles testfile_nvidia_linemap +testrun_compare ${abs_top_builddir}/tests/nvidia_extended_linemap_libdw testfile_nvidia_linemap << EOF +off = 0 + 18 lines +0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:25:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +10: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:26:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +40: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:27:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +90: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:25:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +a0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:28:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +100: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:28:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +100: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:8:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + foo inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:28 +150: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:9:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + foo inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:28 +1e0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +1e0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:6:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +1e0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:8:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + foo inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:6 + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +220: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:9:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + foo inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:6 + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +2b0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:7:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +2f0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:8:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +2f0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:18:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + _Z1aPiS_S_ inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:8 + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +330: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:19:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no + _Z1aPiS_S_ inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/bar.h:8 + bar inlined at /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:31 +3c0: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:33:0: is_stmt:yes, end_seq:no, bb:no, prologue:no, epilogue:no +480: /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4/main.cu:33:0: is_stmt:yes, end_seq:yes, bb:no, prologue:no, epilogue:no +EOF diff --git a/tests/run-nvidia-extended-linemap-readelf.sh b/tests/run-nvidia-extended-linemap-readelf.sh new file mode 100755 index 00000000..1fa9b7b4 --- /dev/null +++ b/tests/run-nvidia-extended-linemap-readelf.sh @@ -0,0 +1,120 @@ +# Copyright (C) 2011 Red Hat, Inc. +# This file is part of elfutils. +# +# This file is free software; you can redistribute it and/or modify +# it under the terms of the GNU General Public License as published by +# the Free Software Foundation; either version 3 of the License, or +# (at your option) any later version. +# +# elfutils is distributed in the hope that it will be useful, but +# WITHOUT ANY WARRANTY; without even the implied warranty of +# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the +# GNU General Public License for more details. +# +# You should have received a copy of the GNU General Public License +# along with this program. If not, see . + +. $srcdir/test-subr.sh + +# NOTE: +# the file testfile_nvidia_linemap is a CUDA binary for an NVIDIA A100 generated as follows using CUDA 11.2 +# nvcc -o main main.cu -Xcompiler "-g -fopenmp" -O3 -lineinfo -arch sm_80 -lcudart -lcuda -lstdc++ -lm +# cuobjdump -xelf all main +# mv main.sm_80.cubin testfile_nvidia_linemap + +testfiles testfile_nvidia_linemap +testrun_compare ${abs_top_builddir}/src/readelf --debug-dump=line testfile_nvidia_linemap << EOF + +DWARF section [ 5] '.debug_line' at offset 0x3e0: + +Table at offset 0: + + Length: 253 + DWARF version: 2 + Prologue length: 111 + Address size: 8 + Segment selector size: 0 + Min instruction length: 1 + Max operations per instruction: 1 + Initial value if 'is_stmt': 1 + Line base: -5 + Line range: 14 + Opcode base: 10 + +Opcodes: + [1] 0 arguments + [2] 1 argument + [3] 1 argument + [4] 1 argument + [5] 1 argument + [6] 0 arguments + [7] 0 arguments + [8] 0 arguments + [9] 1 argument + +Directory table: + /home/johnmc/hpctoolkit-gpu-samples/nvidia_extended_linemap4 + +File name table: + Entry Dir Time Size Name + 1 1 1626104146 1819 main.cu + 2 1 1626104111 211 bar.h + +Line number statements: + [ 79] extended opcode 2: set address to 0 + [ 84] set file to 1 + [ 86] advance line by constant 24 to 25 + [ 88] copy + [ 89] special opcode 240: address+16 = 0x10 , line+1 = 26 + [ 8a] advance line by constant 1 to 27 + [ 8c] advance address by 48 to 0x40 + [ 8e] copy + [ 8f] advance line by constant -2 to 25 + [ 91] advance address by 80 to 0x90 + [ 94] copy + [ 95] special opcode 242: address+16 = 0xa0 , line+3 = 28 + [ 96] advance address by 96 to 0x100 + [ 99] copy + [ 9a] extended opcode 144: set inlined context 6, function name foo (0x0) + [ 9f] advance line by constant -20 to 8 + [ a1] copy + [ a2] advance line by constant 1 to 9 + [ a4] advance address by 80 to 0x150 + [ a7] copy + [ a8] extended opcode 144: set inlined context 0, function name foo (0x0) + [ ad] advance line by constant 22 to 31 + [ af] advance address by 144 to 0x1e0 + [ b2] copy + [ b3] set file to 2 + [ b5] extended opcode 144: set inlined context 9, function name bar (0x4) + [ ba] advance line by constant -25 to 6 + [ bc] copy + [ bd] set file to 1 + [ bf] extended opcode 144: set inlined context 10, function name foo (0x0) + [ c4] advance line by constant 2 to 8 + [ c6] copy + [ c7] advance line by constant 1 to 9 + [ c9] advance address by 64 to 0x220 + [ cc] copy + [ cd] set file to 2 + [ cf] extended opcode 144: set inlined context 9, function name bar (0x4) + [ d4] advance line by constant -2 to 7 + [ d6] advance address by 144 to 0x2b0 + [ d9] copy + [ da] advance line by constant 1 to 8 + [ dc] advance address by 64 to 0x2f0 + [ df] copy + [ e0] set file to 1 + [ e2] extended opcode 144: set inlined context 14, function name _Z1aPiS_S_ (0x8) + [ e7] advance line by constant 10 to 18 + [ e9] copy + [ ea] advance line by constant 1 to 19 + [ ec] advance address by 64 to 0x330 + [ ef] copy + [ f0] extended opcode 144: set inlined context 0, function name foo (0x0) + [ f5] advance line by constant 14 to 33 + [ f7] advance address by 144 to 0x3c0 + [ fa] copy + [ fb] advance address by 192 to 0x480 + [ fe] extended opcode 1: end of sequence +EOF diff --git a/tests/testfile_nvidia_linemap.bz2 b/tests/testfile_nvidia_linemap.bz2 new file mode 100644 index 0000000000000000000000000000000000000000..8a6d09fbd1419af2b4664e8751207be470198f1a GIT binary patch literal 2365 zcmV-D3BvY5T4*^jL0KkKSq)>6PXG%lfB*mg|L^|i|Nqm@%YXm>|9|bF;0gptplMZb z+z@oxek$MzA3=Le*sZ4QFf%trk<(lJVeQ+CXEdl4WbN< z003wJ222QOG|4JO$`Q^Ookzl7$XsrKr|R55rS!>07C#HCYdx~lTt}N z38AT?G}BEz6Gx~S0Av6%G&D2~0017KWB>pLfCE4P28bFNG}BB>fu@E=hD|gy8euXR zhCpD9Moj?FV2nlyri=j$0F0Vs(Sk(*1jt54hEjf{$kR~t(lp+b(p@T+?W=XfGcS&L5UbU0?35a#xs8L!r-U)TF1&F!__dKhBSQp!#+wj$%m9uMNR-G!K_M%dqZ6H; zp6s^6LtA42tC>wzChpmB6xue3L}H1b!5}H>R`jsiDJG5(on{65{32{>NRWayf(z@`1E`3y{ zl~q+m1VF!oZ8R)6HEO=U!CjHKL=i#lMp8Ugu|_5N>}!I}P<2H9jm?W|S`&9IL7fYl&9&Tajy_+U zco?v~b4#nESGuIQ`WznwnVcDcvISVY32sJ@;zMlYbDE<_Ci5dU5WJ7A^d;1t&%hFdjAKBEu4x*h0Wpf#4;~SW1h|NlaA3mRdio<% zGpjR0sd)xQTDw99PZxme&R|Hv~Sh-6jvSs*KAN5Z798xL3$; z9FnOe=%tyNnVINO5fKqInrWo~+BvML@!#9pbEQiVVg!noOb1AemW7{KqN380XnRY1 z6{JWGivo+!W+h-4Y3?|6%5MW0g;{{xNZkBE)Ja^Fk-a=Dk zS=wXMwrouPmhh`==%Vz67fDh|NyHUm+g!3v8g!M10QsMnf&`Qk+^IDn5Z;=Pj1~a5 zH&O@Dx5OLVo51JwnP)1TKRjw}r3i|mBC4vaL`EVgtVLB-h={~hRw5{4IBBKH!Y!!n8(v!LfgK&f~C6?3$*akx+tT+~` zQ&mt=O;aXEkO2drE@PBN-FLcr75GzJg+W0&@WT#_C_bl!Qmkb@|P;)P6=sVtW;lDLBoVZy4hQ58{8Mk=bRAgk!m5)flbxZQ3X+O!1lp@o(X z+SMs00cDV_Ba9V5-r^D{G#CoWD#*P+kpPw8M6zZih>HSCTG3%@-dffX$!=A)MiVAu zjx63v@7iqgAu&330$@?#DN z6NEditP%rEZN^JA8Y?;mfGz=a@`QwllSz_F08}3)mY~3kP&Ps_ZMmM0c!-Q49`B4# zcb+A-d|*WH1GzczT?Nf5cK!b{F_9@3pgiJJ3pWL7W>gA4 zGiDm}BpCOl$LYMDSzZy{NiuK^^pSeEbtH;V1z?N>)ah6m1~G-bIK(87XT3qK5)l9x z40mqd)(~-IuM57Vl3(pymwh1aNiBL67qdt&_v%@YFH69NQHiccgh3-{yq!_dYu|g@ z=@*8TomE+8s9{vLJ*R@(ZQtwkpx|qDhM_ZY0%9&mHga|(YR7p3b|5p@TdqM_Eg+-s zLL!iMrX_|_A~gsaL?&2rselIk+iLV;pl>wLF0}Gs9l=WS)!Yq_pXg3`#*M}(Q-g2@ z1`~`<3%;7Tx%_a8VyAZl!6_V$UBw`=4iMR4eXP@L*K7((7gR*|jKtwYBx*=67Nb8((Hm10 zTeLcBi18~G1vRWQDD}RujxL{FF1l1;K}-aDVK9GC32vRua)mg)2*#^e6+Jh`NDdNs zn#lrvt*a9C4j5hys&rMN&Pg;`(RCy8V3aTH7)tkQALtOK1(%p_BZdf^uW4^<3GX)mdW&zZuC%MyCDnGxGg7 j?