From 088e7af73a962fcc8883b7a6392544d8342553d6 Mon Sep 17 00:00:00 2001 From: Daisuke HATAYAMA Date: Fri, 5 Mar 2010 13:44:06 -0800 Subject: coredump: move dump_write() and dump_seek() into a header file My next patch will replace ELF_CORE_EXTRA_* macros by functions, putting them into other newly created *.c files. Then, each files will contain dump_write(), where each pair of binfmt_*.c and elfcore.c should be the same. So, this patch moves them into a header file with dump_seek(). Also, the patch deletes confusing DUMP_WRITE macros in each files. Signed-off-by: Daisuke HATAYAMA Cc: "Luck, Tony" Cc: Jeff Dike Cc: David Howells Cc: Greg Ungerer Cc: Roland McGrath Cc: Oleg Nesterov Cc: Ingo Molnar Cc: Alexander Viro Cc: Andi Kleen Cc: Alan Cox Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_elf.c | 52 ++++++++++++++-------------------------------------- 1 file changed, 14 insertions(+), 38 deletions(-) (limited to 'fs/binfmt_elf.c') diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index fd5b2ea5d299..0bcfbb05c32d 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -31,6 +31,7 @@ #include #include #include +#include #include #include #include @@ -1085,36 +1086,6 @@ out: * Modelled on fs/exec.c:aout_core_dump() * Jeremy Fitzhardinge */ -/* - * These are the only things you should do on a core-file: use only these - * functions to write out all the necessary info. - */ -static int dump_write(struct file *file, const void *addr, int nr) -{ - return file->f_op->write(file, addr, nr, &file->f_pos) == nr; -} - -static int dump_seek(struct file *file, loff_t off) -{ - if (file->f_op->llseek && file->f_op->llseek != no_llseek) { - if (file->f_op->llseek(file, off, SEEK_CUR) < 0) - return 0; - } else { - char *buf = (char *)get_zeroed_page(GFP_KERNEL); - if (!buf) - return 0; - while (off > 0) { - unsigned long n = off; - if (n > PAGE_SIZE) - n = PAGE_SIZE; - if (!dump_write(file, buf, n)) - return 0; - off -= n; - } - free_page((unsigned long)buf); - } - return 1; -} /* * Decide what to dump of a segment, part, all or none. @@ -1249,11 +1220,6 @@ static int writenote(struct memelfnote *men, struct file *file, } #undef DUMP_WRITE -#define DUMP_WRITE(addr, nr) \ - if ((size += (nr)) > cprm->limit || \ - !dump_write(cprm->file, (addr), (nr))) \ - goto end_coredump; - static void fill_elf_header(struct elfhdr *elf, int segs, u16 machine, u32 flags, u8 osabi) { @@ -1934,7 +1900,10 @@ static int elf_core_dump(struct coredump_params *cprm) fs = get_fs(); set_fs(KERNEL_DS); - DUMP_WRITE(elf, sizeof(*elf)); + size += sizeof(*elf); + if (size > cprm->limit || !dump_write(cprm->file, elf, sizeof(*elf))) + goto end_coredump; + offset += sizeof(*elf); /* Elf header */ offset += (segs + 1) * sizeof(struct elf_phdr); /* Program headers */ foffset = offset; @@ -1948,7 +1917,11 @@ static int elf_core_dump(struct coredump_params *cprm) fill_elf_note_phdr(&phdr, sz, offset); offset += sz; - DUMP_WRITE(&phdr, sizeof(phdr)); + + size += sizeof(phdr); + if (size > cprm->limit + || !dump_write(cprm->file, &phdr, sizeof(phdr))) + goto end_coredump; } dataoff = offset = roundup(offset, ELF_EXEC_PAGESIZE); @@ -1979,7 +1952,10 @@ static int elf_core_dump(struct coredump_params *cprm) phdr.p_flags |= PF_X; phdr.p_align = ELF_EXEC_PAGESIZE; - DUMP_WRITE(&phdr, sizeof(phdr)); + size += sizeof(phdr); + if (size > cprm->limit + || !dump_write(cprm->file, &phdr, sizeof(phdr))) + goto end_coredump; } #ifdef ELF_CORE_WRITE_EXTRA_PHDRS -- cgit v1.2.3 From 1fcccbac89f5bbc5e41aa72086960059fce372da Mon Sep 17 00:00:00 2001 From: Daisuke HATAYAMA Date: Fri, 5 Mar 2010 13:44:07 -0800 Subject: elf coredump: replace ELF_CORE_EXTRA_* macros by functions elf_core_dump() and elf_fdpic_core_dump() use #ifdef and the corresponding macro for hiding _multiline_ logics in functions. This patch removes #ifdef and replaces ELF_CORE_EXTRA_* by corresponding functions. For architectures not implemeonting ELF_CORE_EXTRA_*, we use weak functions in order to reduce a range of modification. This cleanup is for my next patches, but I think this cleanup itself is worth doing regardless of my firnal purpose. Signed-off-by: Daisuke HATAYAMA Cc: "Luck, Tony" Cc: Jeff Dike Cc: David Howells Cc: Greg Ungerer Cc: Roland McGrath Cc: Oleg Nesterov Cc: Ingo Molnar Cc: Alexander Viro Cc: Andi Kleen Cc: Alan Cox Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_elf.c | 14 +++++--------- 1 file changed, 5 insertions(+), 9 deletions(-) (limited to 'fs/binfmt_elf.c') diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index 0bcfbb05c32d..c1a499599b7d 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -1878,9 +1878,7 @@ static int elf_core_dump(struct coredump_params *cprm) * Please check DEFAULT_MAX_MAP_COUNT definition when you modify here. */ segs = current->mm->map_count; -#ifdef ELF_CORE_EXTRA_PHDRS - segs += ELF_CORE_EXTRA_PHDRS; -#endif + segs += elf_core_extra_phdrs(); gate_vma = get_gate_vma(current); if (gate_vma != NULL) @@ -1958,9 +1956,8 @@ static int elf_core_dump(struct coredump_params *cprm) goto end_coredump; } -#ifdef ELF_CORE_WRITE_EXTRA_PHDRS - ELF_CORE_WRITE_EXTRA_PHDRS; -#endif + if (!elf_core_write_extra_phdrs(cprm->file, offset, &size, cprm->limit)) + goto end_coredump; /* write out the notes section */ if (!write_note_info(&info, cprm->file, &foffset)) @@ -1999,9 +1996,8 @@ static int elf_core_dump(struct coredump_params *cprm) } } -#ifdef ELF_CORE_WRITE_EXTRA_DATA - ELF_CORE_WRITE_EXTRA_DATA; -#endif + if (!elf_core_write_extra_data(cprm->file, &size, cprm->limit)) + goto end_coredump; end_coredump: set_fs(fs); -- cgit v1.2.3 From 93eb211e6c9ff6054fcf9c5b9e344d8d9ad29175 Mon Sep 17 00:00:00 2001 From: Daisuke HATAYAMA Date: Fri, 5 Mar 2010 13:44:09 -0800 Subject: elf coredump: make offset calculation process and writing process explicit By the next patch, elf_core_dump() and elf_fdpic_core_dump() will support extended numbering and so will produce the corefiles with section header table in a special case. The problem is the process of writing a file header offset of the section header table into e_shoff field of the ELF header. ELF header is positioned at the beginning of the corefile, while section header at the end. So, we need to take which of the following ways: 1. Seek backward to retry writing operation for ELF header after writing process for a whole part 2. Make offset calculation process and writing process totally sequential The clause 1. is not always possible: one cannot assume that file system supports seek function. Consider the no_llseek case. Therefore, this patch adopts the clause 2. Signed-off-by: Daisuke HATAYAMA Cc: "Luck, Tony" Cc: Jeff Dike Cc: David Howells Cc: Greg Ungerer Cc: Roland McGrath Cc: Oleg Nesterov Cc: Ingo Molnar Cc: Alexander Viro Cc: Andi Kleen Cc: Alan Cox Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_elf.c | 27 ++++++++++++++++----------- 1 file changed, 16 insertions(+), 11 deletions(-) (limited to 'fs/binfmt_elf.c') diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index c1a499599b7d..6fc49b6ed936 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -1856,6 +1856,7 @@ static int elf_core_dump(struct coredump_params *cprm) loff_t offset = 0, dataoff, foffset; unsigned long mm_flags; struct elf_note_info info; + struct elf_phdr *phdr4note = NULL; /* * We no longer stop all VM operations. @@ -1898,28 +1899,22 @@ static int elf_core_dump(struct coredump_params *cprm) fs = get_fs(); set_fs(KERNEL_DS); - size += sizeof(*elf); - if (size > cprm->limit || !dump_write(cprm->file, elf, sizeof(*elf))) - goto end_coredump; - offset += sizeof(*elf); /* Elf header */ offset += (segs + 1) * sizeof(struct elf_phdr); /* Program headers */ foffset = offset; /* Write notes phdr entry */ { - struct elf_phdr phdr; size_t sz = get_note_info_size(&info); sz += elf_coredump_extra_notes_size(); - fill_elf_note_phdr(&phdr, sz, offset); - offset += sz; - - size += sizeof(phdr); - if (size > cprm->limit - || !dump_write(cprm->file, &phdr, sizeof(phdr))) + phdr4note = kmalloc(sizeof(*phdr4note), GFP_KERNEL); + if (!phdr4note) goto end_coredump; + + fill_elf_note_phdr(phdr4note, sz, offset); + offset += sz; } dataoff = offset = roundup(offset, ELF_EXEC_PAGESIZE); @@ -1931,6 +1926,15 @@ static int elf_core_dump(struct coredump_params *cprm) */ mm_flags = current->mm->flags; + size += sizeof(*elf); + if (size > cprm->limit || !dump_write(cprm->file, elf, sizeof(*elf))) + goto end_coredump; + + size += sizeof(*phdr4note); + if (size > cprm->limit + || !dump_write(cprm->file, phdr4note, sizeof(*phdr4note))) + goto end_coredump; + /* Write program headers for segments dump */ for (vma = first_vma(current, gate_vma); vma != NULL; vma = next_vma(vma, gate_vma)) { @@ -2004,6 +2008,7 @@ end_coredump: cleanup: free_note_info(&info); + kfree(phdr4note); kfree(elf); out: return has_dumped; -- cgit v1.2.3 From 8d9032bbe4671dc481261ccd4e161cd96e54b118 Mon Sep 17 00:00:00 2001 From: Daisuke HATAYAMA Date: Fri, 5 Mar 2010 13:44:10 -0800 Subject: elf coredump: add extended numbering support The current ELF dumper implementation can produce broken corefiles if program headers exceed 65535. This number is determined by the number of vmas which the process have. In particular, some extreme programs may use more than 65535 vmas. (If you google max_map_count, you can find some users facing this problem.) This kind of program never be able to generate correct coredumps. This patch implements ``extended numbering'' that uses sh_info field of the first section header instead of e_phnum field in order to represent upto 4294967295 vmas. This is supported by AMD64-ABI(http://www.x86-64.org/documentation.html) and Solaris(http://docs.sun.com/app/docs/doc/817-1984/). Of course, we are preparing patches for gdb and binutils. Signed-off-by: Daisuke HATAYAMA Cc: "Luck, Tony" Cc: Jeff Dike Cc: David Howells Cc: Greg Ungerer Cc: Roland McGrath Cc: Oleg Nesterov Cc: Ingo Molnar Cc: Alexander Viro Cc: Andi Kleen Cc: Alan Cox Cc: Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_elf.c | 66 ++++++++++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 63 insertions(+), 3 deletions(-) (limited to 'fs/binfmt_elf.c') diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index 6fc49b6ed936..78de530cfb02 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -1838,6 +1838,34 @@ static struct vm_area_struct *next_vma(struct vm_area_struct *this_vma, return gate_vma; } +static void fill_extnum_info(struct elfhdr *elf, struct elf_shdr *shdr4extnum, + elf_addr_t e_shoff, int segs) +{ + elf->e_shoff = e_shoff; + elf->e_shentsize = sizeof(*shdr4extnum); + elf->e_shnum = 1; + elf->e_shstrndx = SHN_UNDEF; + + memset(shdr4extnum, 0, sizeof(*shdr4extnum)); + + shdr4extnum->sh_type = SHT_NULL; + shdr4extnum->sh_size = elf->e_shnum; + shdr4extnum->sh_link = elf->e_shstrndx; + shdr4extnum->sh_info = segs; +} + +static size_t elf_core_vma_data_size(struct vm_area_struct *gate_vma, + unsigned long mm_flags) +{ + struct vm_area_struct *vma; + size_t size = 0; + + for (vma = first_vma(current, gate_vma); vma != NULL; + vma = next_vma(vma, gate_vma)) + size += vma_dump_size(vma, mm_flags); + return size; +} + /* * Actual dumper * @@ -1857,6 +1885,9 @@ static int elf_core_dump(struct coredump_params *cprm) unsigned long mm_flags; struct elf_note_info info; struct elf_phdr *phdr4note = NULL; + struct elf_shdr *shdr4extnum = NULL; + Elf_Half e_phnum; + elf_addr_t e_shoff; /* * We no longer stop all VM operations. @@ -1885,12 +1916,19 @@ static int elf_core_dump(struct coredump_params *cprm) if (gate_vma != NULL) segs++; + /* for notes section */ + segs++; + + /* If segs > PN_XNUM(0xffff), then e_phnum overflows. To avoid + * this, kernel supports extended numbering. Have a look at + * include/linux/elf.h for further information. */ + e_phnum = segs > PN_XNUM ? PN_XNUM : segs; + /* * Collect all the non-memory information about the process for the * notes. This also sets up the file header. */ - if (!fill_note_info(elf, segs + 1, /* including notes section */ - &info, cprm->signr, cprm->regs)) + if (!fill_note_info(elf, e_phnum, &info, cprm->signr, cprm->regs)) goto cleanup; has_dumped = 1; @@ -1900,7 +1938,7 @@ static int elf_core_dump(struct coredump_params *cprm) set_fs(KERNEL_DS); offset += sizeof(*elf); /* Elf header */ - offset += (segs + 1) * sizeof(struct elf_phdr); /* Program headers */ + offset += segs * sizeof(struct elf_phdr); /* Program headers */ foffset = offset; /* Write notes phdr entry */ @@ -1926,6 +1964,19 @@ static int elf_core_dump(struct coredump_params *cprm) */ mm_flags = current->mm->flags; + offset += elf_core_vma_data_size(gate_vma, mm_flags); + offset += elf_core_extra_data_size(); + e_shoff = offset; + + if (e_phnum == PN_XNUM) { + shdr4extnum = kmalloc(sizeof(*shdr4extnum), GFP_KERNEL); + if (!shdr4extnum) + goto end_coredump; + fill_extnum_info(elf, shdr4extnum, e_shoff, segs); + } + + offset = dataoff; + size += sizeof(*elf); if (size > cprm->limit || !dump_write(cprm->file, elf, sizeof(*elf))) goto end_coredump; @@ -2003,11 +2054,20 @@ static int elf_core_dump(struct coredump_params *cprm) if (!elf_core_write_extra_data(cprm->file, &size, cprm->limit)) goto end_coredump; + if (e_phnum == PN_XNUM) { + size += sizeof(*shdr4extnum); + if (size > cprm->limit + || !dump_write(cprm->file, shdr4extnum, + sizeof(*shdr4extnum))) + goto end_coredump; + } + end_coredump: set_fs(fs); cleanup: free_note_info(&info); + kfree(shdr4extnum); kfree(phdr4note); kfree(elf); out: -- cgit v1.2.3 From 30736a4d43f4af7f1a7836d6a266be17082195c4 Mon Sep 17 00:00:00 2001 From: Masami Hiramatsu Date: Fri, 5 Mar 2010 13:44:12 -0800 Subject: coredump: pass mm->flags as a coredump parameter for consistency Pass mm->flags as a coredump parameter for consistency. --- 1787 if (mm->core_state || !get_dumpable(mm)) { <- (1) 1788 up_write(&mm->mmap_sem); 1789 put_cred(cred); 1790 goto fail; 1791 } 1792 [...] 1798 if (get_dumpable(mm) == 2) { /* Setuid core dump mode */ <-(2) 1799 flag = O_EXCL; /* Stop rewrite attacks */ 1800 cred->fsuid = 0; /* Dump root private */ 1801 } --- Since dumpable bits are not protected by lock, there is a chance to change these bits between (1) and (2). To solve this issue, this patch copies mm->flags to coredump_params.mm_flags at the beginning of do_coredump() and uses it instead of get_dumpable() while dumping core. This copy is also passed to binfmt->core_dump, since elf*_core_dump() uses dump_filter bits in mm->flags. [akpm@linux-foundation.org: fix merge] Signed-off-by: Masami Hiramatsu Acked-by: Roland McGrath Cc: Hidehiro Kawai Cc: Oleg Nesterov Cc: Ingo Molnar Reviewed-by: KOSAKI Motohiro Signed-off-by: Andrew Morton Signed-off-by: Linus Torvalds --- fs/binfmt_elf.c | 14 +++----------- 1 file changed, 3 insertions(+), 11 deletions(-) (limited to 'fs/binfmt_elf.c') diff --git a/fs/binfmt_elf.c b/fs/binfmt_elf.c index 78de530cfb02..535e763ab1a6 100644 --- a/fs/binfmt_elf.c +++ b/fs/binfmt_elf.c @@ -1882,7 +1882,6 @@ static int elf_core_dump(struct coredump_params *cprm) struct vm_area_struct *vma, *gate_vma; struct elfhdr *elf = NULL; loff_t offset = 0, dataoff, foffset; - unsigned long mm_flags; struct elf_note_info info; struct elf_phdr *phdr4note = NULL; struct elf_shdr *shdr4extnum = NULL; @@ -1957,14 +1956,7 @@ static int elf_core_dump(struct coredump_params *cprm) dataoff = offset = roundup(offset, ELF_EXEC_PAGESIZE); - /* - * We must use the same mm->flags while dumping core to avoid - * inconsistency between the program headers and bodies, otherwise an - * unusable core file can be generated. - */ - mm_flags = current->mm->flags; - - offset += elf_core_vma_data_size(gate_vma, mm_flags); + offset += elf_core_vma_data_size(gate_vma, cprm->mm_flags); offset += elf_core_extra_data_size(); e_shoff = offset; @@ -1995,7 +1987,7 @@ static int elf_core_dump(struct coredump_params *cprm) phdr.p_offset = offset; phdr.p_vaddr = vma->vm_start; phdr.p_paddr = 0; - phdr.p_filesz = vma_dump_size(vma, mm_flags); + phdr.p_filesz = vma_dump_size(vma, cprm->mm_flags); phdr.p_memsz = vma->vm_end - vma->vm_start; offset += phdr.p_filesz; phdr.p_flags = vma->vm_flags & VM_READ ? PF_R : 0; @@ -2030,7 +2022,7 @@ static int elf_core_dump(struct coredump_params *cprm) unsigned long addr; unsigned long end; - end = vma->vm_start + vma_dump_size(vma, mm_flags); + end = vma->vm_start + vma_dump_size(vma, cprm->mm_flags); for (addr = vma->vm_start; addr < end; addr += PAGE_SIZE) { struct page *page; -- cgit v1.2.3