From a2085a540975a1e6660dbc3b0c092789f002703e Mon Sep 17 00:00:00 2001 From: Alan Modra Date: Sun, 8 Nov 2015 09:29:00 -0800 Subject: [PATCH 08/11] ELF unexec: Drive from PT_LOAD header rather than sections This rewrites bss handling in the ELF unexec code. Finding bss sections by name results in complicated code that - does not account for all names of possible bss sections, - assumes specific ordering of bss sections, - can wrongly choose a SHT_NOBITS section not in the bss segment, - incorrectly calculates bss size (no accounting for alignment gaps), - assumes .data and .bss are in the same segment. All of these problems and more are solved by finding the bss segment in PT_LOAD headers, ie. the address range included in p_memsz but not p_filesz of the last PT_LOAD header, then matching SHT_NOBITS sections in that address range. * unexelf.c: Delete old ppc comment. (OLD_PROGRAM_H): Define. (round_up): Delete. (unexec): Don't search for bss style sections by name. Instead, use the last PT_LOAD header address range covered by p_memsz but not p_filesz and match any SHT_NOBITS section in that address range. Simplify initialisation of section header vars. Don't assume that section headers are above bss segment. Move copying of bss area out of section loop. Align .data2 section to 1, since it now covers the entire bss area. For SHT_NOBITS sections in the bss segment, leave sh_addr and sh_addralign unchanged, but correct sh_offset. Clear memory corresponding to SHT_NOBITS .plt section. Delete comment and hacks for sections partly overlapping bss range now that the full range is properly calculated. Delete now dead .sbss code. (Bug#20614) --- src/unexelf.c | 290 ++++++++++++++++------------------------------------------ 1 file changed, 77 insertions(+), 213 deletions(-) diff --git a/src/unexelf.c b/src/unexelf.c index 15a4cde..4e9c50d 100644 --- a/src/unexelf.c +++ b/src/unexelf.c @@ -535,29 +535,6 @@ verify ((! TYPE_SIGNED (ElfW (Half)) /* Get the address of a particular section or program header entry, * accounting for the size of the entries. */ -/* - On PPC Reference Platform running Solaris 2.5.1 - the plt section is also of type NOBI like the bss section. - (not really stored) and therefore sections after the bss - section start at the plt offset. The plt section is always - the one just before the bss section. - Thus, we modify the test from - if (NEW_SECTION_H (nn).sh_offset >= new_data2_offset) - to - if (NEW_SECTION_H (nn).sh_offset >= - OLD_SECTION_H (old_bss_index-1).sh_offset) - This is just a hack. We should put the new data section - before the .plt section. - And we should not have this routine at all but use - the libelf library to read the old file and create the new - file. - The changed code is minimal and depends on prep set in m/prep.h - Erik Deumens - Quantum Theory Project - University of Florida - deumens@qtp.ufl.edu - Apr 23, 1996 - */ static void * entry_address (void *section_h, ptrdiff_t idx, ptrdiff_t entsize) @@ -570,23 +547,14 @@ entry_address (void *section_h, ptrdiff_t idx, ptrdiff_t entsize) (*(ElfW (Shdr) *) entry_address (old_section_h, n, old_file_h->e_shentsize)) #define NEW_SECTION_H(n) \ (*(ElfW (Shdr) *) entry_address (new_section_h, n, new_file_h->e_shentsize)) +#define OLD_PROGRAM_H(n) \ + (*(ElfW (Phdr) *) entry_address (old_program_h, n, old_file_h->e_phentsize)) #define NEW_PROGRAM_H(n) \ (*(ElfW (Phdr) *) entry_address (new_program_h, n, new_file_h->e_phentsize)) #define PATCH_INDEX(n) ((n) += old_bss_index <= (n)) typedef unsigned char byte; -/* Round X up to a multiple of Y. */ - -static ElfW (Addr) -round_up (ElfW (Addr) x, ElfW (Addr) y) -{ - ElfW (Addr) rem = x % y; - if (rem == 0) - return x; - return x - rem + y; -} - /* Return the index of the section named NAME. SECTION_NAMES, FILE_NAME and FILE_H give information about the file we are looking in. @@ -650,16 +618,15 @@ unexec (const char *new_name, const char *old_name) /* Point to the section name table in the old file. */ char *old_section_names; + ElfW (Phdr) *old_bss_seg, *new_bss_seg; ElfW (Addr) old_bss_addr, new_bss_addr; ElfW (Word) old_bss_size, new_data2_size; ElfW (Off) new_data2_offset; ElfW (Addr) new_data2_addr; ElfW (Off) old_bss_offset; - ElfW (Word) new_data2_incr; ptrdiff_t n, nn; - ptrdiff_t old_bss_index, old_sbss_index, old_plt_index; - ptrdiff_t old_data_index, new_data2_index; + ptrdiff_t old_bss_index, old_data_index; struct stat stat_buf; off_t old_file_size; @@ -703,54 +670,40 @@ unexec (const char *new_name, const char *old_name) old_section_names = (char *) old_base + OLD_SECTION_H (old_file_h->e_shstrndx).sh_offset; - /* Find the old .bss section. Figure out parameters of the new - data2 and bss sections. */ - - old_bss_index = find_section (".bss", old_section_names, - old_name, old_file_h, old_section_h, 0); - - old_sbss_index = find_section (".sbss", old_section_names, - old_name, old_file_h, old_section_h, 1); - if (old_sbss_index != -1) - if (OLD_SECTION_H (old_sbss_index).sh_type != SHT_NOBITS) - old_sbss_index = -1; - - /* PowerPC64 has .plt in the BSS section. */ - old_plt_index = find_section (".plt", old_section_names, - old_name, old_file_h, old_section_h, 1); - if (old_plt_index != -1) - if (OLD_SECTION_H (old_plt_index).sh_type != SHT_NOBITS) - old_plt_index = -1; - - if (old_sbss_index == -1 && old_plt_index == -1) - { - old_bss_addr = OLD_SECTION_H (old_bss_index).sh_addr; - old_bss_size = OLD_SECTION_H (old_bss_index).sh_size; - old_bss_offset = OLD_SECTION_H (old_bss_index).sh_offset; - new_data2_index = old_bss_index; - } - else if (old_plt_index != -1 - && (old_sbss_index == -1 - || (OLD_SECTION_H (old_sbss_index).sh_addr - > OLD_SECTION_H (old_plt_index).sh_addr))) + /* Find the PT_LOAD header covering the highest address. This + segment will be where bss sections are located, past p_filesz. */ + old_bss_seg = 0; + for (n = old_file_h->e_phnum; --n >= 0; ) { - old_bss_addr = OLD_SECTION_H (old_plt_index).sh_addr; - old_bss_size = OLD_SECTION_H (old_bss_index).sh_size - + OLD_SECTION_H (old_plt_index).sh_size; - if (old_sbss_index != -1) - old_bss_size += OLD_SECTION_H (old_sbss_index).sh_size; - old_bss_offset = OLD_SECTION_H (old_plt_index).sh_offset; - new_data2_index = old_plt_index; + ElfW (Phdr) *seg = &OLD_PROGRAM_H (n); + if (seg->p_type == PT_LOAD + && (old_bss_seg == 0 + || seg->p_vaddr > old_bss_seg->p_vaddr)) + old_bss_seg = seg; } - else + + /* Note that old_bss_addr may be lower than the first bss section + address, since the section may need aligning. */ + old_bss_addr = old_bss_seg->p_vaddr + old_bss_seg->p_filesz; + old_bss_offset = old_bss_seg->p_offset + old_bss_seg->p_filesz; + old_bss_size = old_bss_seg->p_memsz - old_bss_seg->p_filesz; + + /* Find the first bss style section in the bss segment range. */ + old_bss_index = -1; + for (n = old_file_h->e_shnum; --n > 0; ) { - old_bss_addr = OLD_SECTION_H (old_sbss_index).sh_addr; - old_bss_size = OLD_SECTION_H (old_bss_index).sh_size - + OLD_SECTION_H (old_sbss_index).sh_size; - old_bss_offset = OLD_SECTION_H (old_sbss_index).sh_offset; - new_data2_index = old_sbss_index; + ElfW (Shdr) *shdr = &OLD_SECTION_H (n); + if (shdr->sh_type == SHT_NOBITS + && shdr->sh_addr >= old_bss_addr + && shdr->sh_addr + shdr->sh_size <= old_bss_addr + old_bss_size + && (old_bss_index == -1 + || OLD_SECTION_H (old_bss_index).sh_addr > shdr->sh_addr)) + old_bss_index = n; } + if (old_bss_index == -1) + fatal ("no bss section found"); + /* Find the old .data section. Figure out parameters of the new data2 and bss sections. */ @@ -761,13 +714,7 @@ unexec (const char *new_name, const char *old_name) new_bss_addr = (ElfW (Addr)) new_break; new_data2_addr = old_bss_addr; new_data2_size = new_bss_addr - old_bss_addr; - new_data2_offset = OLD_SECTION_H (old_data_index).sh_offset - + (new_data2_addr - OLD_SECTION_H (old_data_index).sh_addr); - /* This is the amount by which the sections following the bss sections - must be shifted in the image. It can differ from new_data2_size if - the end of the old .data section (and thus the offset of the .bss - section) was unaligned. */ - new_data2_incr = new_data2_size + (new_data2_offset - old_bss_offset); + new_data2_offset = old_bss_offset; #ifdef UNEXELF_DEBUG fprintf (stderr, "old_bss_index %td\n", old_bss_index); @@ -778,7 +725,6 @@ unexec (const char *new_name, const char *old_name) DEBUG_LOG (new_data2_addr); DEBUG_LOG (new_data2_size); DEBUG_LOG (new_data2_offset); - DEBUG_LOG (new_data2_incr); #endif if (new_bss_addr < old_bss_addr + old_bss_size) @@ -792,7 +738,7 @@ unexec (const char *new_name, const char *old_name) if (new_file < 0) fatal ("Can't creat (%s): %s", new_name, strerror (errno)); - new_file_size = old_file_size + old_file_h->e_shentsize + new_data2_incr; + new_file_size = old_file_size + old_file_h->e_shentsize + new_data2_size; if (ftruncate (new_file, new_file_size)) fatal ("Can't ftruncate (%s): %s", new_name, strerror (errno)); @@ -811,15 +757,15 @@ unexec (const char *new_name, const char *old_name) /* Fix up file header. We'll add one section. Section header is further away now. */ - new_file_h->e_shoff += new_data2_incr; + if (new_file_h->e_shoff >= old_bss_offset) + new_file_h->e_shoff += new_data2_size; new_file_h->e_shnum += 1; /* Modify the e_shstrndx if necessary. */ PATCH_INDEX (new_file_h->e_shstrndx); - new_program_h = (ElfW (Phdr) *) ((byte *) new_base + old_file_h->e_phoff); - new_section_h = (ElfW (Shdr) *) - ((byte *) new_base + old_file_h->e_shoff + new_data2_incr); + new_program_h = (ElfW (Phdr) *) ((byte *) new_base + new_file_h->e_phoff); + new_section_h = (ElfW (Shdr) *) ((byte *) new_base + new_file_h->e_shoff); memcpy (new_program_h, old_program_h, old_file_h->e_phnum * old_file_h->e_phentsize); @@ -831,65 +777,21 @@ unexec (const char *new_name, const char *old_name) fprintf (stderr, "New section count %td\n", (ptrdiff_t) new_file_h->e_shnum); #endif - /* Fix up a new program header. Extend the writable data segment so - that the bss area is covered too. Find that segment by looking - for a segment that ends just before the .bss area. Make sure - that no segments are above the new .data2. Put a loop at the end - to adjust the offset and address of any segment that is above - data2, just in case we decide to allow this later. */ + /* Fix up program header. Extend the writable data segment so + that the bss area is covered too. */ - for (n = new_file_h->e_phnum; --n >= 0; ) - { - /* Compute maximum of all requirements for alignment of section. */ - ElfW (Word) alignment = (NEW_PROGRAM_H (n)).p_align; - if ((OLD_SECTION_H (old_bss_index)).sh_addralign > alignment) - alignment = OLD_SECTION_H (old_bss_index).sh_addralign; - -#ifdef __sgi - /* According to r02kar@x4u2.desy.de (Karsten Kuenne) - and oliva@gnu.org (Alexandre Oliva), on IRIX 5.2, we - always get "Program segment above .bss" when dumping - when the executable doesn't have an sbss section. */ - if (old_sbss_index != -1) -#endif /* __sgi */ - if (NEW_PROGRAM_H (n).p_vaddr + NEW_PROGRAM_H (n).p_filesz - > (old_sbss_index == -1 - ? old_bss_addr - : round_up (old_bss_addr, alignment))) - fatal ("Program segment above .bss in %s", old_name); - - if (NEW_PROGRAM_H (n).p_type == PT_LOAD - && (round_up ((NEW_PROGRAM_H (n)).p_vaddr - + (NEW_PROGRAM_H (n)).p_filesz, - alignment) - == round_up (old_bss_addr, alignment))) - break; - } - if (n < 0) - fatal ("Couldn't find segment next to .bss in %s", old_name); - - /* Make sure that the size includes any padding before the old .bss - section. */ - NEW_PROGRAM_H (n).p_filesz = new_bss_addr - NEW_PROGRAM_H (n).p_vaddr; - NEW_PROGRAM_H (n).p_memsz = NEW_PROGRAM_H (n).p_filesz; - -#if 0 /* Maybe allow section after data2 - does this ever happen? */ - for (n = new_file_h->e_phnum; --n >= 0; ) - { - if (NEW_PROGRAM_H (n).p_vaddr - && NEW_PROGRAM_H (n).p_vaddr >= new_data2_addr) - NEW_PROGRAM_H (n).p_vaddr += new_data2_size - old_bss_size; + new_bss_seg = new_program_h + (old_bss_seg - old_program_h); + new_bss_seg->p_filesz = new_bss_addr - new_bss_seg->p_vaddr; + new_bss_seg->p_memsz = new_bss_seg->p_filesz; - if (NEW_PROGRAM_H (n).p_offset >= new_data2_offset) - NEW_PROGRAM_H (n).p_offset += new_data2_incr; - } -#endif + /* Copy over what we have in memory now for the bss area. */ + memcpy (new_base + new_data2_offset, (caddr_t) old_bss_addr, new_data2_size); /* Fix up section headers based on new .data2 section. Any section whose offset or virtual address is after the new .data2 section - gets its value adjusted. .bss size becomes zero and new address - is set. data2 section header gets added by copying the existing - .data header and modifying the offset, address and size. */ + gets its value adjusted. .bss size becomes zero. data2 section + header gets added by copying the existing .data header and + modifying the offset, address and size. */ /* Walk through all section headers, insert the new data2 section right before the new bss section. */ @@ -900,9 +802,7 @@ unexec (const char *new_name, const char *old_name) ElfW (Shdr) *new_shdr = &NEW_SECTION_H (nn); /* If it is (s)bss section, insert the new data2 section before it. */ - /* new_data2_index is the index of either old_sbss or old_bss, that was - chosen as a section for new_data2. */ - if (n == new_data2_index) + if (n == old_bss_index) { /* Steal the data section header for this data2 section. */ memcpy (new_shdr, &OLD_SECTION_H (old_data_index), @@ -911,68 +811,43 @@ unexec (const char *new_name, const char *old_name) new_shdr->sh_addr = new_data2_addr; new_shdr->sh_offset = new_data2_offset; new_shdr->sh_size = new_data2_size; - /* Use the bss section's alignment. This will assure that the - new data2 section always be placed in the same spot as the old - bss section by any other application. */ - new_shdr->sh_addralign = old_shdr->sh_addralign; - - /* Now copy over what we have in the memory now. */ - memcpy (new_shdr->sh_offset + new_base, - (caddr_t) old_shdr->sh_addr, - new_data2_size); + new_shdr->sh_addralign = 1; nn++; new_shdr++; } memcpy (new_shdr, old_shdr, old_file_h->e_shentsize); - if (n == old_bss_index - /* The new bss and sbss section's size is zero, and its file offset - and virtual address should be off by NEW_DATA2_SIZE. */ - || n == old_sbss_index || n == old_plt_index - ) + if (new_shdr->sh_type == SHT_NOBITS + && new_shdr->sh_addr >= old_bss_addr + && (new_shdr->sh_addr + new_shdr->sh_size + <= old_bss_addr + old_bss_size)) { - /* NN should be `old_s?bss_index + 1' at this point. */ - new_shdr->sh_offset = new_data2_offset + new_data2_size; - new_shdr->sh_addr = new_data2_addr + new_data2_size; - /* Let the new bss section address alignment be the same as the - section address alignment followed the old bss section, so - this section will be placed in exactly the same place. */ - new_shdr->sh_addralign = OLD_SECTION_H (nn).sh_addralign; + /* SHT_NOBITS sections do not need a valid sh_offset, so it + might be incorrect. Write the correct value. */ + new_shdr->sh_offset = (new_shdr->sh_addr - new_bss_seg->p_vaddr + + new_bss_seg->p_offset); + + /* If this is was a SHT_NOBITS .plt section, then it is + probably a PowerPC PLT. If it is PowerPC64 ELFv1 then + glibc ld.so doesn't initialize the toc pointer word. A + non-zero toc pointer word can defeat Power7 thread safety + during lazy update of a PLT entry. This only matters if + emacs becomes multi-threaded. */ + if (strcmp (old_section_names + new_shdr->sh_name, ".plt") == 0) + memset (new_shdr->sh_offset + new_base, 0, new_shdr->sh_size); + + /* Set the new bss and sbss section's size to zero, because + we've already covered this address range by .data2. */ new_shdr->sh_size = 0; } else { /* Any section that was originally placed after the .bss - section should now be off by NEW_DATA2_INCR. If a - section overlaps the .bss section, consider it to be - placed after the .bss section. Overlap can occur if the - section just before .bss has less-strict alignment; this - was observed between .symtab and .bss on Solaris 2.5.1 - (sparc) with GCC snapshot 960602. + section should now be off by NEW_DATA2_SIZE. */ -> dump -h temacs - -temacs: - - **** SECTION HEADER TABLE **** -[No] Type Flags Addr Offset Size Name - Link Info Adralgn Entsize - -[22] 1 3 0x335150 0x315150 0x4 .data.rel.local - 0 0 0x4 0 - -[23] 8 3 0x335158 0x315158 0x42720 .bss - 0 0 0x8 0 - -[24] 2 0 0 0x315154 0x1c9d0 .symtab - 25 1709 0x4 0x10 - */ - - if (new_shdr->sh_offset >= old_bss_offset - || (new_shdr->sh_offset + new_shdr->sh_size - > new_data2_offset)) - new_shdr->sh_offset += new_data2_incr; + if (new_shdr->sh_offset >= old_bss_offset) + new_shdr->sh_offset += new_data2_size; /* Any section that was originally placed after the section header table should now be off by the size of one section @@ -992,23 +867,13 @@ temacs: && new_shdr->sh_type != SHT_DYNSYM) PATCH_INDEX (new_shdr->sh_info); - if (old_sbss_index != -1) - if (!strcmp (old_section_names + new_shdr->sh_name, ".sbss")) - { - new_shdr->sh_offset = - round_up (new_shdr->sh_offset, - new_shdr->sh_addralign); - new_shdr->sh_type = SHT_PROGBITS; - } - /* Now, start to copy the content of sections. */ if (new_shdr->sh_type == SHT_NULL || new_shdr->sh_type == SHT_NOBITS) continue; - /* Write out the sections. .data and .data1 (and data2, called - ".data" in the strings table) get copied from the current process - instead of the old file. */ + /* Some sections are copied from the current process instead of + the old file. */ if (!strcmp (old_section_names + new_shdr->sh_name, ".data") || !strcmp (old_section_names + new_shdr->sh_name, ".sdata") || !strcmp (old_section_names + new_shdr->sh_name, ".lit4") @@ -1037,8 +902,7 @@ temacs: || !strcmp (old_section_names + new_shdr->sh_name, ".got") #endif || !strcmp (old_section_names + new_shdr->sh_name, ".sdata1") - || !strcmp (old_section_names + new_shdr->sh_name, ".data1") - || !strcmp (old_section_names + new_shdr->sh_name, ".sbss")) + || !strcmp (old_section_names + new_shdr->sh_name, ".data1")) src = (caddr_t) old_shdr->sh_addr; else src = old_base + old_shdr->sh_offset; -- 2.7.4