x86: remove iBCS support
[sfrench/cifs-2.6.git] / fs / binfmt_elf.c
index b1013f34085d00ac4b966e895ac58419f4ae3269..18ed6dd906c150208a4beb9230676a5a95de06d6 100644 (file)
 
 static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs);
 static int load_elf_library(struct file *);
-static unsigned long elf_map (struct file *, unsigned long, struct elf_phdr *, int, int);
+static unsigned long elf_map(struct file *, unsigned long, struct elf_phdr *,
+                               int, int, unsigned long);
 
 /*
  * If we don't support core dumping, then supply a NULL so we
  * don't even try.
  */
 #if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
-static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file);
+static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file, unsigned long limit);
 #else
 #define elf_core_dump  NULL
 #endif
@@ -150,6 +151,14 @@ create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
        struct task_struct *tsk = current;
        struct vm_area_struct *vma;
 
+       /*
+        * In some cases (e.g. Hyper-Threading), we want to avoid L1
+        * evictions by the processes running on the same package. One
+        * thing we can do is to shuffle the initial stack for them.
+        */
+
+       p = arch_align_stack(p);
+
        /*
         * If this architecture has a platform capability string, copy it
         * to userspace.  In some cases (Sparc), this info is impossible
@@ -160,14 +169,6 @@ create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
        if (k_platform) {
                size_t len = strlen(k_platform) + 1;
 
-               /*
-                * In some cases (e.g. Hyper-Threading), we want to avoid L1
-                * evictions by the processes running on the same package. One
-                * thing we can do is to shuffle the initial stack for them.
-                */
-
-               p = arch_align_stack(p);
-
                u_platform = (elf_addr_t __user *)STACK_ALLOC(p, len);
                if (__copy_to_user(u_platform, k_platform, len))
                        return -EFAULT;
@@ -175,6 +176,7 @@ create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
 
        /* Create the ELF interpreter info */
        elf_info = (elf_addr_t *)current->mm->saved_auxv;
+       /* update AT_VECTOR_SIZE_BASE if the number of NEW_AUX_ENT() changes */
 #define NEW_AUX_ENT(id, val) \
        do { \
                elf_info[ei_index++] = id; \
@@ -185,6 +187,8 @@ create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
        /* 
         * ARCH_DLINFO must come first so PPC can do its special alignment of
         * AUXV.
+        * update AT_VECTOR_SIZE_ARCH if the number of NEW_AUX_ENT() in
+        * ARCH_DLINFO changes
         */
        ARCH_DLINFO;
 #endif
@@ -295,33 +299,70 @@ create_elf_tables(struct linux_binprm *bprm, struct elfhdr *exec,
 #ifndef elf_map
 
 static unsigned long elf_map(struct file *filep, unsigned long addr,
-               struct elf_phdr *eppnt, int prot, int type)
+               struct elf_phdr *eppnt, int prot, int type,
+               unsigned long total_size)
 {
        unsigned long map_addr;
-       unsigned long pageoffset = ELF_PAGEOFFSET(eppnt->p_vaddr);
+       unsigned long size = eppnt->p_filesz + ELF_PAGEOFFSET(eppnt->p_vaddr);
+       unsigned long off = eppnt->p_offset - ELF_PAGEOFFSET(eppnt->p_vaddr);
+       addr = ELF_PAGESTART(addr);
+       size = ELF_PAGEALIGN(size);
 
-       down_write(&current->mm->mmap_sem);
        /* mmap() will return -EINVAL if given a zero size, but a
         * segment with zero filesize is perfectly valid */
-       if (eppnt->p_filesz + pageoffset)
-               map_addr = do_mmap(filep, ELF_PAGESTART(addr),
-                                  eppnt->p_filesz + pageoffset, prot, type,
-                                  eppnt->p_offset - pageoffset);
-       else
-               map_addr = ELF_PAGESTART(addr);
+       if (!size)
+               return addr;
+
+       down_write(&current->mm->mmap_sem);
+       /*
+       * total_size is the size of the ELF (interpreter) image.
+       * The _first_ mmap needs to know the full size, otherwise
+       * randomization might put this image into an overlapping
+       * position with the ELF binary image. (since size < total_size)
+       * So we first map the 'big' image - and unmap the remainder at
+       * the end. (which unmap is needed for ELF images with holes.)
+       */
+       if (total_size) {
+               total_size = ELF_PAGEALIGN(total_size);
+               map_addr = do_mmap(filep, addr, total_size, prot, type, off);
+               if (!BAD_ADDR(map_addr))
+                       do_munmap(current->mm, map_addr+size, total_size-size);
+       } else
+               map_addr = do_mmap(filep, addr, size, prot, type, off);
+
        up_write(&current->mm->mmap_sem);
        return(map_addr);
 }
 
 #endif /* !elf_map */
 
+static unsigned long total_mapping_size(struct elf_phdr *cmds, int nr)
+{
+       int i, first_idx = -1, last_idx = -1;
+
+       for (i = 0; i < nr; i++) {
+               if (cmds[i].p_type == PT_LOAD) {
+                       last_idx = i;
+                       if (first_idx == -1)
+                               first_idx = i;
+               }
+       }
+       if (first_idx == -1)
+               return 0;
+
+       return cmds[last_idx].p_vaddr + cmds[last_idx].p_memsz -
+                               ELF_PAGESTART(cmds[first_idx].p_vaddr);
+}
+
+
 /* This is much more generalized than the library routine read function,
    so we keep this separate.  Technically the library read function
    is only provided so that we can read a.out libraries that have
    an ELF header */
 
 static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
-               struct file *interpreter, unsigned long *interp_load_addr)
+               struct file *interpreter, unsigned long *interp_map_addr,
+               unsigned long no_base)
 {
        struct elf_phdr *elf_phdata;
        struct elf_phdr *eppnt;
@@ -329,6 +370,7 @@ static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
        int load_addr_set = 0;
        unsigned long last_bss = 0, elf_bss = 0;
        unsigned long error = ~0UL;
+       unsigned long total_size;
        int retval, i, size;
 
        /* First of all, some simple consistency checks */
@@ -367,6 +409,12 @@ static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
                goto out_close;
        }
 
+       total_size = total_mapping_size(elf_phdata, interp_elf_ex->e_phnum);
+       if (!total_size) {
+               error = -EINVAL;
+               goto out_close;
+       }
+
        eppnt = elf_phdata;
        for (i = 0; i < interp_elf_ex->e_phnum; i++, eppnt++) {
                if (eppnt->p_type == PT_LOAD) {
@@ -384,9 +432,14 @@ static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
                        vaddr = eppnt->p_vaddr;
                        if (interp_elf_ex->e_type == ET_EXEC || load_addr_set)
                                elf_type |= MAP_FIXED;
+                       else if (no_base && interp_elf_ex->e_type == ET_DYN)
+                               load_addr = -vaddr;
 
                        map_addr = elf_map(interpreter, load_addr + vaddr,
-                                          eppnt, elf_prot, elf_type);
+                                       eppnt, elf_prot, elf_type, total_size);
+                       total_size = 0;
+                       if (!*interp_map_addr)
+                               *interp_map_addr = map_addr;
                        error = map_addr;
                        if (BAD_ADDR(map_addr))
                                goto out_close;
@@ -452,8 +505,7 @@ static unsigned long load_elf_interp(struct elfhdr *interp_elf_ex,
                        goto out_close;
        }
 
-       *interp_load_addr = load_addr;
-       error = ((unsigned long)interp_elf_ex->e_entry) + load_addr;
+       error = load_addr;
 
 out_close:
        kfree(elf_phdata);
@@ -543,14 +595,14 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
        int load_addr_set = 0;
        char * elf_interpreter = NULL;
        unsigned int interpreter_type = INTERPRETER_NONE;
-       unsigned char ibcs2_interpreter = 0;
        unsigned long error;
        struct elf_phdr *elf_ppnt, *elf_phdata;
        unsigned long elf_bss, elf_brk;
        int elf_exec_fileno;
        int retval, i;
        unsigned int size;
-       unsigned long elf_entry, interp_load_addr = 0;
+       unsigned long elf_entry;
+       unsigned long interp_load_addr = 0;
        unsigned long start_code, end_code, start_data, end_data;
        unsigned long reloc_func_desc = 0;
        char passed_fileno[6];
@@ -660,14 +712,6 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
                        if (elf_interpreter[elf_ppnt->p_filesz - 1] != '\0')
                                goto out_free_interp;
 
-                       /* If the program interpreter is one of these two,
-                        * then assume an iBCS2 image. Otherwise assume
-                        * a native linux image.
-                        */
-                       if (strcmp(elf_interpreter,"/usr/lib/libc.so.1") == 0 ||
-                           strcmp(elf_interpreter,"/usr/lib/ld.so.1") == 0)
-                               ibcs2_interpreter = 1;
-
                        /*
                         * The early SET_PERSONALITY here is so that the lookup
                         * for the interpreter happens in the namespace of the 
@@ -687,7 +731,7 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
                         * switch really is going to happen - do this in
                         * flush_thread().      - akpm
                         */
-                       SET_PERSONALITY(loc->elf_ex, ibcs2_interpreter);
+                       SET_PERSONALITY(loc->elf_ex, 0);
 
                        interpreter = open_exec(elf_interpreter);
                        retval = PTR_ERR(interpreter);
@@ -730,6 +774,7 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
 
        /* Some simple consistency checks for the interpreter */
        if (elf_interpreter) {
+               static int warn;
                interpreter_type = INTERPRETER_ELF | INTERPRETER_AOUT;
 
                /* Now figure out which format our binary is */
@@ -741,6 +786,13 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
                if (memcmp(loc->interp_elf_ex.e_ident, ELFMAG, SELFMAG) != 0)
                        interpreter_type &= ~INTERPRETER_ELF;
 
+               if (interpreter_type == INTERPRETER_AOUT && warn < 10) {
+                       printk(KERN_WARNING "a.out ELF interpreter %s is "
+                               "deprecated and will not be supported "
+                               "after Linux 2.6.25\n", elf_interpreter);
+                       warn++;
+               }
+
                retval = -ELIBBAD;
                if (!interpreter_type)
                        goto out_free_dentry;
@@ -758,7 +810,7 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
                        goto out_free_dentry;
        } else {
                /* Executables without an interpreter also need a personality  */
-               SET_PERSONALITY(loc->elf_ex, ibcs2_interpreter);
+               SET_PERSONALITY(loc->elf_ex, 0);
        }
 
        /* OK, we are done with that, now set up the arg stuff,
@@ -792,7 +844,7 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
 
        /* Do this immediately, since STACK_TOP as used in setup_arg_pages
           may depend on the personality.  */
-       SET_PERSONALITY(loc->elf_ex, ibcs2_interpreter);
+       SET_PERSONALITY(loc->elf_ex, 0);
        if (elf_read_implies_exec(loc->elf_ex, executable_stack))
                current->personality |= READ_IMPLIES_EXEC;
 
@@ -814,9 +866,7 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
        current->mm->start_stack = bprm->p;
 
        /* Now we do a little grungy work by mmaping the ELF image into
-          the correct location in memory.  At this point, we assume that
-          the image should be loaded at fixed address, not at a variable
-          address. */
+          the correct location in memory. */
        for(i = 0, elf_ppnt = elf_phdata;
            i < loc->elf_ex.e_phnum; i++, elf_ppnt++) {
                int elf_prot = 0, elf_flags;
@@ -870,11 +920,15 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
                         * default mmap base, as well as whatever program they
                         * might try to exec.  This is because the brk will
                         * follow the loader, and is not movable.  */
+#ifdef CONFIG_X86
+                       load_bias = 0;
+#else
                        load_bias = ELF_PAGESTART(ELF_ET_DYN_BASE - vaddr);
+#endif
                }
 
                error = elf_map(bprm->file, load_bias + vaddr, elf_ppnt,
-                               elf_prot, elf_flags);
+                               elf_prot, elf_flags, 0);
                if (BAD_ADDR(error)) {
                        send_sig(SIGKILL, current, 0);
                        retval = IS_ERR((void *)error) ?
@@ -950,13 +1004,25 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
        }
 
        if (elf_interpreter) {
-               if (interpreter_type == INTERPRETER_AOUT)
+               if (interpreter_type == INTERPRETER_AOUT) {
                        elf_entry = load_aout_interp(&loc->interp_ex,
                                                     interpreter);
-               else
+               } else {
+                       unsigned long uninitialized_var(interp_map_addr);
+
                        elf_entry = load_elf_interp(&loc->interp_elf_ex,
                                                    interpreter,
-                                                   &interp_load_addr);
+                                                   &interp_map_addr,
+                                                   load_bias);
+                       if (!IS_ERR((void *)elf_entry)) {
+                               /*
+                                * load_elf_interp() returns relocation
+                                * adjustment
+                                */
+                               interp_load_addr = elf_entry;
+                               elf_entry += loc->interp_elf_ex.e_entry;
+                       }
+               }
                if (BAD_ADDR(elf_entry)) {
                        force_sig(SIGSEGV, current);
                        retval = IS_ERR((void *)elf_entry) ?
@@ -1010,6 +1076,12 @@ static int load_elf_binary(struct linux_binprm *bprm, struct pt_regs *regs)
        current->mm->end_data = end_data;
        current->mm->start_stack = bprm->p;
 
+#ifdef arch_randomize_brk
+       if (current->flags & PF_RANDOMIZE)
+               current->mm->brk = current->mm->start_brk =
+                       arch_randomize_brk(current->mm);
+#endif
+
        if (current->personality & MMAP_PAGE_ZERO) {
                /* Why this, you ask???  Well SVr4 maps page 0 as read-only,
                   and some applications "depend" upon this behavior.
@@ -1193,35 +1265,68 @@ static int dump_seek(struct file *file, loff_t off)
 }
 
 /*
- * Decide whether a segment is worth dumping; default is yes to be
- * sure (missing info is worse than too much; etc).
- * Personally I'd include everything, and use the coredump limit...
- *
- * I think we should skip something. But I am not sure how. H.J.
+ * Decide what to dump of a segment, part, all or none.
  */
-static int maydump(struct vm_area_struct *vma, unsigned long mm_flags)
+static unsigned long vma_dump_size(struct vm_area_struct *vma,
+                                  unsigned long mm_flags)
 {
        /* The vma can be set up to tell us the answer directly.  */
        if (vma->vm_flags & VM_ALWAYSDUMP)
-               return 1;
+               goto whole;
 
        /* Do not dump I/O mapped devices or special mappings */
        if (vma->vm_flags & (VM_IO | VM_RESERVED))
                return 0;
 
+#define FILTER(type)   (mm_flags & (1UL << MMF_DUMP_##type))
+
        /* By default, dump shared memory if mapped from an anonymous file. */
        if (vma->vm_flags & VM_SHARED) {
-               if (vma->vm_file->f_path.dentry->d_inode->i_nlink == 0)
-                       return test_bit(MMF_DUMP_ANON_SHARED, &mm_flags);
-               else
-                       return test_bit(MMF_DUMP_MAPPED_SHARED, &mm_flags);
+               if (vma->vm_file->f_path.dentry->d_inode->i_nlink == 0 ?
+                   FILTER(ANON_SHARED) : FILTER(MAPPED_SHARED))
+                       goto whole;
+               return 0;
        }
 
-       /* By default, if it hasn't been written to, don't write it out. */
-       if (!vma->anon_vma)
-               return test_bit(MMF_DUMP_MAPPED_PRIVATE, &mm_flags);
+       /* Dump segments that have been written to.  */
+       if (vma->anon_vma && FILTER(ANON_PRIVATE))
+               goto whole;
+       if (vma->vm_file == NULL)
+               return 0;
+
+       if (FILTER(MAPPED_PRIVATE))
+               goto whole;
 
-       return test_bit(MMF_DUMP_ANON_PRIVATE, &mm_flags);
+       /*
+        * If this looks like the beginning of a DSO or executable mapping,
+        * check for an ELF header.  If we find one, dump the first page to
+        * aid in determining what was mapped here.
+        */
+       if (FILTER(ELF_HEADERS) && vma->vm_file != NULL && vma->vm_pgoff == 0) {
+               u32 __user *header = (u32 __user *) vma->vm_start;
+               u32 word;
+               /*
+                * Doing it this way gets the constant folded by GCC.
+                */
+               union {
+                       u32 cmp;
+                       char elfmag[SELFMAG];
+               } magic;
+               BUILD_BUG_ON(SELFMAG != sizeof word);
+               magic.elfmag[EI_MAG0] = ELFMAG0;
+               magic.elfmag[EI_MAG1] = ELFMAG1;
+               magic.elfmag[EI_MAG2] = ELFMAG2;
+               magic.elfmag[EI_MAG3] = ELFMAG3;
+               if (get_user(word, header) == 0 && word == magic.cmp)
+                       return PAGE_SIZE;
+       }
+
+#undef FILTER
+
+       return 0;
+
+whole:
+       return vma->vm_end - vma->vm_start;
 }
 
 /* An ELF note in memory */
@@ -1281,7 +1386,8 @@ static int writenote(struct memelfnote *men, struct file *file,
        if (!dump_seek(file, (off))) \
                goto end_coredump;
 
-static void fill_elf_header(struct elfhdr *elf, int segs)
+static void fill_elf_header(struct elfhdr *elf, int segs,
+                           u16 machine, u32 flags, u8 osabi)
 {
        memcpy(elf->e_ident, ELFMAG, SELFMAG);
        elf->e_ident[EI_CLASS] = ELF_CLASS;
@@ -1291,12 +1397,12 @@ static void fill_elf_header(struct elfhdr *elf, int segs)
        memset(elf->e_ident+EI_PAD, 0, EI_NIDENT-EI_PAD);
 
        elf->e_type = ET_CORE;
-       elf->e_machine = ELF_ARCH;
+       elf->e_machine = machine;
        elf->e_version = EV_CURRENT;
        elf->e_entry = 0;
        elf->e_phoff = sizeof(struct elfhdr);
        elf->e_shoff = 0;
-       elf->e_flags = ELF_CORE_EFLAGS;
+       elf->e_flags = flags;
        elf->e_ehsize = sizeof(struct elfhdr);
        elf->e_phentsize = sizeof(struct elf_phdr);
        elf->e_phnum = segs;
@@ -1339,10 +1445,10 @@ static void fill_prstatus(struct elf_prstatus *prstatus,
        prstatus->pr_info.si_signo = prstatus->pr_cursig = signr;
        prstatus->pr_sigpend = p->pending.signal.sig[0];
        prstatus->pr_sighold = p->blocked.sig[0];
-       prstatus->pr_pid = p->pid;
-       prstatus->pr_ppid = p->parent->pid;
-       prstatus->pr_pgrp = process_group(p);
-       prstatus->pr_sid = process_session(p);
+       prstatus->pr_pid = task_pid_vnr(p);
+       prstatus->pr_ppid = task_pid_vnr(p->real_parent);
+       prstatus->pr_pgrp = task_pgrp_vnr(p);
+       prstatus->pr_sid = task_session_vnr(p);
        if (thread_group_leader(p)) {
                /*
                 * This is the record for the group leader.  Add in the
@@ -1385,10 +1491,10 @@ static int fill_psinfo(struct elf_prpsinfo *psinfo, struct task_struct *p,
                        psinfo->pr_psargs[i] = ' ';
        psinfo->pr_psargs[len] = 0;
 
-       psinfo->pr_pid = p->pid;
-       psinfo->pr_ppid = p->parent->pid;
-       psinfo->pr_pgrp = process_group(p);
-       psinfo->pr_sid = process_session(p);
+       psinfo->pr_pid = task_pid_vnr(p);
+       psinfo->pr_ppid = task_pid_vnr(p->real_parent);
+       psinfo->pr_pgrp = task_pgrp_vnr(p);
+       psinfo->pr_sid = task_session_vnr(p);
 
        i = p->state ? ffz(~p->state) + 1 : 0;
        psinfo->pr_state = i;
@@ -1403,6 +1509,238 @@ static int fill_psinfo(struct elf_prpsinfo *psinfo, struct task_struct *p,
        return 0;
 }
 
+static void fill_auxv_note(struct memelfnote *note, struct mm_struct *mm)
+{
+       elf_addr_t *auxv = (elf_addr_t *) mm->saved_auxv;
+       int i = 0;
+       do
+               i += 2;
+       while (auxv[i - 2] != AT_NULL);
+       fill_note(note, "CORE", NT_AUXV, i * sizeof(elf_addr_t), auxv);
+}
+
+#ifdef CORE_DUMP_USE_REGSET
+#include <linux/regset.h>
+
+struct elf_thread_core_info {
+       struct elf_thread_core_info *next;
+       struct task_struct *task;
+       struct elf_prstatus prstatus;
+       struct memelfnote notes[0];
+};
+
+struct elf_note_info {
+       struct elf_thread_core_info *thread;
+       struct memelfnote psinfo;
+       struct memelfnote auxv;
+       size_t size;
+       int thread_notes;
+};
+
+static int fill_thread_core_info(struct elf_thread_core_info *t,
+                                const struct user_regset_view *view,
+                                long signr, size_t *total)
+{
+       unsigned int i;
+
+       /*
+        * NT_PRSTATUS is the one special case, because the regset data
+        * goes into the pr_reg field inside the note contents, rather
+        * than being the whole note contents.  We fill the reset in here.
+        * We assume that regset 0 is NT_PRSTATUS.
+        */
+       fill_prstatus(&t->prstatus, t->task, signr);
+       (void) view->regsets[0].get(t->task, &view->regsets[0],
+                                   0, sizeof(t->prstatus.pr_reg),
+                                   &t->prstatus.pr_reg, NULL);
+
+       fill_note(&t->notes[0], "CORE", NT_PRSTATUS,
+                 sizeof(t->prstatus), &t->prstatus);
+       *total += notesize(&t->notes[0]);
+
+       /*
+        * Each other regset might generate a note too.  For each regset
+        * that has no core_note_type or is inactive, we leave t->notes[i]
+        * all zero and we'll know to skip writing it later.
+        */
+       for (i = 1; i < view->n; ++i) {
+               const struct user_regset *regset = &view->regsets[i];
+               if (regset->core_note_type &&
+                   (!regset->active || regset->active(t->task, regset))) {
+                       int ret;
+                       size_t size = regset->n * regset->size;
+                       void *data = kmalloc(size, GFP_KERNEL);
+                       if (unlikely(!data))
+                               return 0;
+                       ret = regset->get(t->task, regset,
+                                         0, size, data, NULL);
+                       if (unlikely(ret))
+                               kfree(data);
+                       else {
+                               if (regset->core_note_type != NT_PRFPREG)
+                                       fill_note(&t->notes[i], "LINUX",
+                                                 regset->core_note_type,
+                                                 size, data);
+                               else {
+                                       t->prstatus.pr_fpvalid = 1;
+                                       fill_note(&t->notes[i], "CORE",
+                                                 NT_PRFPREG, size, data);
+                               }
+                               *total += notesize(&t->notes[i]);
+                       }
+               }
+       }
+
+       return 1;
+}
+
+static int fill_note_info(struct elfhdr *elf, int phdrs,
+                         struct elf_note_info *info,
+                         long signr, struct pt_regs *regs)
+{
+       struct task_struct *dump_task = current;
+       const struct user_regset_view *view = task_user_regset_view(dump_task);
+       struct elf_thread_core_info *t;
+       struct elf_prpsinfo *psinfo;
+       struct task_struct *g, *p;
+       unsigned int i;
+
+       info->size = 0;
+       info->thread = NULL;
+
+       psinfo = kmalloc(sizeof(*psinfo), GFP_KERNEL);
+       fill_note(&info->psinfo, "CORE", NT_PRPSINFO, sizeof(*psinfo), psinfo);
+
+       if (psinfo == NULL)
+               return 0;
+
+       /*
+        * Figure out how many notes we're going to need for each thread.
+        */
+       info->thread_notes = 0;
+       for (i = 0; i < view->n; ++i)
+               if (view->regsets[i].core_note_type != 0)
+                       ++info->thread_notes;
+
+       /*
+        * Sanity check.  We rely on regset 0 being in NT_PRSTATUS,
+        * since it is our one special case.
+        */
+       if (unlikely(info->thread_notes == 0) ||
+           unlikely(view->regsets[0].core_note_type != NT_PRSTATUS)) {
+               WARN_ON(1);
+               return 0;
+       }
+
+       /*
+        * Initialize the ELF file header.
+        */
+       fill_elf_header(elf, phdrs,
+                       view->e_machine, view->e_flags, view->ei_osabi);
+
+       /*
+        * Allocate a structure for each thread.
+        */
+       rcu_read_lock();
+       do_each_thread(g, p)
+               if (p->mm == dump_task->mm) {
+                       t = kzalloc(offsetof(struct elf_thread_core_info,
+                                            notes[info->thread_notes]),
+                                   GFP_ATOMIC);
+                       if (unlikely(!t)) {
+                               rcu_read_unlock();
+                               return 0;
+                       }
+                       t->task = p;
+                       if (p == dump_task || !info->thread) {
+                               t->next = info->thread;
+                               info->thread = t;
+                       } else {
+                               /*
+                                * Make sure to keep the original task at
+                                * the head of the list.
+                                */
+                               t->next = info->thread->next;
+                               info->thread->next = t;
+                       }
+               }
+       while_each_thread(g, p);
+       rcu_read_unlock();
+
+       /*
+        * Now fill in each thread's information.
+        */
+       for (t = info->thread; t != NULL; t = t->next)
+               if (!fill_thread_core_info(t, view, signr, &info->size))
+                       return 0;
+
+       /*
+        * Fill in the two process-wide notes.
+        */
+       fill_psinfo(psinfo, dump_task->group_leader, dump_task->mm);
+       info->size += notesize(&info->psinfo);
+
+       fill_auxv_note(&info->auxv, current->mm);
+       info->size += notesize(&info->auxv);
+
+       return 1;
+}
+
+static size_t get_note_info_size(struct elf_note_info *info)
+{
+       return info->size;
+}
+
+/*
+ * Write all the notes for each thread.  When writing the first thread, the
+ * process-wide notes are interleaved after the first thread-specific note.
+ */
+static int write_note_info(struct elf_note_info *info,
+                          struct file *file, loff_t *foffset)
+{
+       bool first = 1;
+       struct elf_thread_core_info *t = info->thread;
+
+       do {
+               int i;
+
+               if (!writenote(&t->notes[0], file, foffset))
+                       return 0;
+
+               if (first && !writenote(&info->psinfo, file, foffset))
+                       return 0;
+               if (first && !writenote(&info->auxv, file, foffset))
+                       return 0;
+
+               for (i = 1; i < info->thread_notes; ++i)
+                       if (t->notes[i].data &&
+                           !writenote(&t->notes[i], file, foffset))
+                               return 0;
+
+               first = 0;
+               t = t->next;
+       } while (t);
+
+       return 1;
+}
+
+static void free_note_info(struct elf_note_info *info)
+{
+       struct elf_thread_core_info *threads = info->thread;
+       while (threads) {
+               unsigned int i;
+               struct elf_thread_core_info *t = threads;
+               threads = t->next;
+               WARN_ON(t->notes[0].data && t->notes[0].data != &t->prstatus);
+               for (i = 1; i < info->thread_notes; ++i)
+                       kfree(t->notes[i].data);
+               kfree(t);
+       }
+       kfree(info->psinfo.data);
+}
+
+#else
+
 /* Here is the structure in which status of each thread is captured. */
 struct elf_thread_status
 {
@@ -1411,7 +1749,7 @@ struct elf_thread_status
        elf_fpregset_t fpu;             /* NT_PRFPREG */
        struct task_struct *thread;
 #ifdef ELF_CORE_COPY_XFPREGS
-       elf_fpxregset_t xfpu;           /* NT_PRXFPREG */
+       elf_fpxregset_t xfpu;           /* ELF_CORE_XFPREG_TYPE */
 #endif
        struct memelfnote notes[3];
        int num_notes;
@@ -1446,8 +1784,8 @@ static int elf_dump_thread_status(long signr, struct elf_thread_status *t)
 
 #ifdef ELF_CORE_COPY_XFPREGS
        if (elf_core_copy_task_xfpregs(p, &t->xfpu)) {
-               fill_note(&t->notes[2], "LINUX", NT_PRXFPREG, sizeof(t->xfpu),
-                         &t->xfpu);
+               fill_note(&t->notes[2], "LINUX", ELF_CORE_XFPREG_TYPE,
+                         sizeof(t->xfpu), &t->xfpu);
                t->num_notes++;
                sz += notesize(&t->notes[2]);
        }
@@ -1455,6 +1793,176 @@ static int elf_dump_thread_status(long signr, struct elf_thread_status *t)
        return sz;
 }
 
+struct elf_note_info {
+       struct memelfnote *notes;
+       struct elf_prstatus *prstatus;  /* NT_PRSTATUS */
+       struct elf_prpsinfo *psinfo;    /* NT_PRPSINFO */
+       struct list_head thread_list;
+       elf_fpregset_t *fpu;
+#ifdef ELF_CORE_COPY_XFPREGS
+       elf_fpxregset_t *xfpu;
+#endif
+       int thread_status_size;
+       int numnote;
+};
+
+static int fill_note_info(struct elfhdr *elf, int phdrs,
+                         struct elf_note_info *info,
+                         long signr, struct pt_regs *regs)
+{
+#define        NUM_NOTES       6
+       struct list_head *t;
+       struct task_struct *g, *p;
+
+       info->notes = NULL;
+       info->prstatus = NULL;
+       info->psinfo = NULL;
+       info->fpu = NULL;
+#ifdef ELF_CORE_COPY_XFPREGS
+       info->xfpu = NULL;
+#endif
+       INIT_LIST_HEAD(&info->thread_list);
+
+       info->notes = kmalloc(NUM_NOTES * sizeof(struct memelfnote),
+                             GFP_KERNEL);
+       if (!info->notes)
+               return 0;
+       info->psinfo = kmalloc(sizeof(*info->psinfo), GFP_KERNEL);
+       if (!info->psinfo)
+               return 0;
+       info->prstatus = kmalloc(sizeof(*info->prstatus), GFP_KERNEL);
+       if (!info->prstatus)
+               return 0;
+       info->fpu = kmalloc(sizeof(*info->fpu), GFP_KERNEL);
+       if (!info->fpu)
+               return 0;
+#ifdef ELF_CORE_COPY_XFPREGS
+       info->xfpu = kmalloc(sizeof(*info->xfpu), GFP_KERNEL);
+       if (!info->xfpu)
+               return 0;
+#endif
+
+       info->thread_status_size = 0;
+       if (signr) {
+               struct elf_thread_status *tmp;
+               rcu_read_lock();
+               do_each_thread(g, p)
+                       if (current->mm == p->mm && current != p) {
+                               tmp = kzalloc(sizeof(*tmp), GFP_ATOMIC);
+                               if (!tmp) {
+                                       rcu_read_unlock();
+                                       return 0;
+                               }
+                               tmp->thread = p;
+                               list_add(&tmp->list, &info->thread_list);
+                       }
+               while_each_thread(g, p);
+               rcu_read_unlock();
+               list_for_each(t, &info->thread_list) {
+                       struct elf_thread_status *tmp;
+                       int sz;
+
+                       tmp = list_entry(t, struct elf_thread_status, list);
+                       sz = elf_dump_thread_status(signr, tmp);
+                       info->thread_status_size += sz;
+               }
+       }
+       /* now collect the dump for the current */
+       memset(info->prstatus, 0, sizeof(*info->prstatus));
+       fill_prstatus(info->prstatus, current, signr);
+       elf_core_copy_regs(&info->prstatus->pr_reg, regs);
+
+       /* Set up header */
+       fill_elf_header(elf, phdrs, ELF_ARCH, ELF_CORE_EFLAGS, ELF_OSABI);
+
+       /*
+        * Set up the notes in similar form to SVR4 core dumps made
+        * with info from their /proc.
+        */
+
+       fill_note(info->notes + 0, "CORE", NT_PRSTATUS,
+                 sizeof(*info->prstatus), info->prstatus);
+       fill_psinfo(info->psinfo, current->group_leader, current->mm);
+       fill_note(info->notes + 1, "CORE", NT_PRPSINFO,
+                 sizeof(*info->psinfo), info->psinfo);
+
+       info->numnote = 2;
+
+       fill_auxv_note(&info->notes[info->numnote++], current->mm);
+
+       /* Try to dump the FPU. */
+       info->prstatus->pr_fpvalid = elf_core_copy_task_fpregs(current, regs,
+                                                              info->fpu);
+       if (info->prstatus->pr_fpvalid)
+               fill_note(info->notes + info->numnote++,
+                         "CORE", NT_PRFPREG, sizeof(*info->fpu), info->fpu);
+#ifdef ELF_CORE_COPY_XFPREGS
+       if (elf_core_copy_task_xfpregs(current, info->xfpu))
+               fill_note(info->notes + info->numnote++,
+                         "LINUX", ELF_CORE_XFPREG_TYPE,
+                         sizeof(*info->xfpu), info->xfpu);
+#endif
+
+       return 1;
+
+#undef NUM_NOTES
+}
+
+static size_t get_note_info_size(struct elf_note_info *info)
+{
+       int sz = 0;
+       int i;
+
+       for (i = 0; i < info->numnote; i++)
+               sz += notesize(info->notes + i);
+
+       sz += info->thread_status_size;
+
+       return sz;
+}
+
+static int write_note_info(struct elf_note_info *info,
+                          struct file *file, loff_t *foffset)
+{
+       int i;
+       struct list_head *t;
+
+       for (i = 0; i < info->numnote; i++)
+               if (!writenote(info->notes + i, file, foffset))
+                       return 0;
+
+       /* write out the thread status notes section */
+       list_for_each(t, &info->thread_list) {
+               struct elf_thread_status *tmp =
+                               list_entry(t, struct elf_thread_status, list);
+
+               for (i = 0; i < tmp->num_notes; i++)
+                       if (!writenote(&tmp->notes[i], file, foffset))
+                               return 0;
+       }
+
+       return 1;
+}
+
+static void free_note_info(struct elf_note_info *info)
+{
+       while (!list_empty(&info->thread_list)) {
+               struct list_head *tmp = info->thread_list.next;
+               list_del(tmp);
+               kfree(list_entry(tmp, struct elf_thread_status, list));
+       }
+
+       kfree(info->prstatus);
+       kfree(info->psinfo);
+       kfree(info->notes);
+       kfree(info->fpu);
+#ifdef ELF_CORE_COPY_XFPREGS
+       kfree(info->xfpu);
+#endif
+}
+
+#endif
+
 static struct vm_area_struct *first_vma(struct task_struct *tsk,
                                        struct vm_area_struct *gate_vma)
 {
@@ -1488,32 +1996,17 @@ static struct vm_area_struct *next_vma(struct vm_area_struct *this_vma,
  * and then they are actually written out.  If we run out of core limit
  * we just truncate.
  */
-static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
+static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file, unsigned long limit)
 {
-#define        NUM_NOTES       6
        int has_dumped = 0;
        mm_segment_t fs;
        int segs;
        size_t size = 0;
-       int i;
        struct vm_area_struct *vma, *gate_vma;
        struct elfhdr *elf = NULL;
        loff_t offset = 0, dataoff, foffset;
-       unsigned long limit = current->signal->rlim[RLIMIT_CORE].rlim_cur;
-       int numnote;
-       struct memelfnote *notes = NULL;
-       struct elf_prstatus *prstatus = NULL;   /* NT_PRSTATUS */
-       struct elf_prpsinfo *psinfo = NULL;     /* NT_PRPSINFO */
-       struct task_struct *g, *p;
-       LIST_HEAD(thread_list);
-       struct list_head *t;
-       elf_fpregset_t *fpu = NULL;
-#ifdef ELF_CORE_COPY_XFPREGS
-       elf_fpxregset_t *xfpu = NULL;
-#endif
-       int thread_status_size = 0;
-       elf_addr_t *auxv;
        unsigned long mm_flags;
+       struct elf_note_info info;
 
        /*
         * We no longer stop all VM operations.
@@ -1531,52 +2024,6 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
        elf = kmalloc(sizeof(*elf), GFP_KERNEL);
        if (!elf)
                goto cleanup;
-       prstatus = kmalloc(sizeof(*prstatus), GFP_KERNEL);
-       if (!prstatus)
-               goto cleanup;
-       psinfo = kmalloc(sizeof(*psinfo), GFP_KERNEL);
-       if (!psinfo)
-               goto cleanup;
-       notes = kmalloc(NUM_NOTES * sizeof(struct memelfnote), GFP_KERNEL);
-       if (!notes)
-               goto cleanup;
-       fpu = kmalloc(sizeof(*fpu), GFP_KERNEL);
-       if (!fpu)
-               goto cleanup;
-#ifdef ELF_CORE_COPY_XFPREGS
-       xfpu = kmalloc(sizeof(*xfpu), GFP_KERNEL);
-       if (!xfpu)
-               goto cleanup;
-#endif
-
-       if (signr) {
-               struct elf_thread_status *tmp;
-               rcu_read_lock();
-               do_each_thread(g,p)
-                       if (current->mm == p->mm && current != p) {
-                               tmp = kzalloc(sizeof(*tmp), GFP_ATOMIC);
-                               if (!tmp) {
-                                       rcu_read_unlock();
-                                       goto cleanup;
-                               }
-                               tmp->thread = p;
-                               list_add(&tmp->list, &thread_list);
-                       }
-               while_each_thread(g,p);
-               rcu_read_unlock();
-               list_for_each(t, &thread_list) {
-                       struct elf_thread_status *tmp;
-                       int sz;
-
-                       tmp = list_entry(t, struct elf_thread_status, list);
-                       sz = elf_dump_thread_status(signr, tmp);
-                       thread_status_size += sz;
-               }
-       }
-       /* now collect the dump for the current */
-       memset(prstatus, 0, sizeof(*prstatus));
-       fill_prstatus(prstatus, current, signr);
-       elf_core_copy_regs(&prstatus->pr_reg, regs);
        
        segs = current->mm->map_count;
 #ifdef ELF_CORE_EXTRA_PHDRS
@@ -1587,42 +2034,16 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
        if (gate_vma != NULL)
                segs++;
 
-       /* Set up header */
-       fill_elf_header(elf, segs + 1); /* including notes section */
-
-       has_dumped = 1;
-       current->flags |= PF_DUMPCORE;
-
        /*
-        * Set up the notes in similar form to SVR4 core dumps made
-        * with info from their /proc.
+        * 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, signr, regs))
+               goto cleanup;
 
-       fill_note(notes + 0, "CORE", NT_PRSTATUS, sizeof(*prstatus), prstatus);
-       fill_psinfo(psinfo, current->group_leader, current->mm);
-       fill_note(notes + 1, "CORE", NT_PRPSINFO, sizeof(*psinfo), psinfo);
-       
-       numnote = 2;
-
-       auxv = (elf_addr_t *)current->mm->saved_auxv;
-
-       i = 0;
-       do
-               i += 2;
-       while (auxv[i - 2] != AT_NULL);
-       fill_note(&notes[numnote++], "CORE", NT_AUXV,
-                 i * sizeof(elf_addr_t), auxv);
-
-       /* Try to dump the FPU. */
-       if ((prstatus->pr_fpvalid =
-            elf_core_copy_task_fpregs(current, regs, fpu)))
-               fill_note(notes + numnote++,
-                         "CORE", NT_PRFPREG, sizeof(*fpu), fpu);
-#ifdef ELF_CORE_COPY_XFPREGS
-       if (elf_core_copy_task_xfpregs(current, xfpu))
-               fill_note(notes + numnote++,
-                         "LINUX", NT_PRXFPREG, sizeof(*xfpu), xfpu);
-#endif 
+       has_dumped = 1;
+       current->flags |= PF_DUMPCORE;
   
        fs = get_fs();
        set_fs(KERNEL_DS);
@@ -1635,12 +2056,7 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
        /* Write notes phdr entry */
        {
                struct elf_phdr phdr;
-               int sz = 0;
-
-               for (i = 0; i < numnote; i++)
-                       sz += notesize(notes + i);
-               
-               sz += thread_status_size;
+               size_t sz = get_note_info_size(&info);
 
                sz += elf_coredump_extra_notes_size();
 
@@ -1662,16 +2078,13 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
        for (vma = first_vma(current, gate_vma); vma != NULL;
                        vma = next_vma(vma, gate_vma)) {
                struct elf_phdr phdr;
-               size_t sz;
-
-               sz = vma->vm_end - vma->vm_start;
 
                phdr.p_type = PT_LOAD;
                phdr.p_offset = offset;
                phdr.p_vaddr = vma->vm_start;
                phdr.p_paddr = 0;
-               phdr.p_filesz = maydump(vma, mm_flags) ? sz : 0;
-               phdr.p_memsz = sz;
+               phdr.p_filesz = vma_dump_size(vma, 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;
                if (vma->vm_flags & VM_WRITE)
@@ -1688,36 +2101,23 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
 #endif
 
        /* write out the notes section */
-       for (i = 0; i < numnote; i++)
-               if (!writenote(notes + i, file, &foffset))
-                       goto end_coredump;
+       if (!write_note_info(&info, file, &foffset))
+               goto end_coredump;
 
        if (elf_coredump_extra_notes_write(file, &foffset))
                goto end_coredump;
 
-       /* write out the thread status notes section */
-       list_for_each(t, &thread_list) {
-               struct elf_thread_status *tmp =
-                               list_entry(t, struct elf_thread_status, list);
-
-               for (i = 0; i < tmp->num_notes; i++)
-                       if (!writenote(&tmp->notes[i], file, &foffset))
-                               goto end_coredump;
-       }
-
        /* Align to page */
        DUMP_SEEK(dataoff - foffset);
 
        for (vma = first_vma(current, gate_vma); vma != NULL;
                        vma = next_vma(vma, gate_vma)) {
                unsigned long addr;
+               unsigned long end;
 
-               if (!maydump(vma, mm_flags))
-                       continue;
+               end = vma->vm_start + vma_dump_size(vma, mm_flags);
 
-               for (addr = vma->vm_start;
-                    addr < vma->vm_end;
-                    addr += PAGE_SIZE) {
+               for (addr = vma->vm_start; addr < end; addr += PAGE_SIZE) {
                        struct page *page;
                        struct vm_area_struct *vma;
 
@@ -1725,7 +2125,7 @@ static int elf_core_dump(long signr, struct pt_regs *regs, struct file *file)
                                                &page, &vma) <= 0) {
                                DUMP_SEEK(PAGE_SIZE);
                        } else {
-                               if (page == ZERO_PAGE(addr)) {
+                               if (page == ZERO_PAGE(0)) {
                                        if (!dump_seek(file, PAGE_SIZE)) {
                                                page_cache_release(page);
                                                goto end_coredump;
@@ -1757,22 +2157,9 @@ end_coredump:
        set_fs(fs);
 
 cleanup:
-       while (!list_empty(&thread_list)) {
-               struct list_head *tmp = thread_list.next;
-               list_del(tmp);
-               kfree(list_entry(tmp, struct elf_thread_status, list));
-       }
-
        kfree(elf);
-       kfree(prstatus);
-       kfree(psinfo);
-       kfree(notes);
-       kfree(fpu);
-#ifdef ELF_CORE_COPY_XFPREGS
-       kfree(xfpu);
-#endif
+       free_note_info(&info);
        return has_dumped;
-#undef NUM_NOTES
 }
 
 #endif         /* USE_ELF_CORE_DUMP */