[PATCH] FDPIC: Add coredump capability for the ELF-FDPIC binfmt
authorDavid Howells <dhowells@redhat.com>
Mon, 10 Jul 2006 11:44:55 +0000 (04:44 -0700)
committerLinus Torvalds <torvalds@g5.osdl.org>
Mon, 10 Jul 2006 20:24:22 +0000 (13:24 -0700)
Add coredump capability for the ELF-FDPIC binfmt.

Signed-off-by: David Howells <dhowells@redhat.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>
arch/frv/kernel/process.c
fs/binfmt_elf_fdpic.c
include/asm-frv/elf.h
include/linux/elfcore.h

index ecdeafb2fdcee2e7cf5aecf38d54e2ad5ff6bbd3..f0c767a56c65be57e1ff46963176a75142669651 100644 (file)
@@ -371,3 +371,11 @@ int elf_check_arch(const struct elf32_hdr *hdr)
 
        return 1;
 }
+
+int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpregs)
+{
+       memcpy(fpregs,
+              &current->thread.user->f,
+              sizeof(current->thread.user->f));
+       return 1;
+}
index a4ff87389823dedacb6c1cc7abc16862148eb238..2f336582922940ffb9160d664f39acff70fe3ad1 100644 (file)
@@ -24,7 +24,9 @@
 #include <linux/file.h>
 #include <linux/fcntl.h>
 #include <linux/slab.h>
+#include <linux/pagemap.h>
 #include <linux/highmem.h>
+#include <linux/highuid.h>
 #include <linux/personality.h>
 #include <linux/ptrace.h>
 #include <linux/init.h>
@@ -48,6 +50,12 @@ typedef char *elf_caddr_t;
 #define kdebug(fmt, ...) do {} while(0)
 #endif
 
+#if 0
+#define kdcore(fmt, ...) printk("FDPIC "fmt"\n" ,##__VA_ARGS__ )
+#else
+#define kdcore(fmt, ...) do {} while(0)
+#endif
+
 MODULE_LICENSE("GPL");
 
 static int load_elf_fdpic_binary(struct linux_binprm *, struct pt_regs *);
@@ -70,10 +78,16 @@ static int elf_fdpic_map_file_constdisp_on_uclinux(struct elf_fdpic_params *,
 static int elf_fdpic_map_file_by_direct_mmap(struct elf_fdpic_params *,
                                             struct file *, struct mm_struct *);
 
+#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
+static int elf_fdpic_core_dump(long, struct pt_regs *, struct file *);
+#endif
+
 static struct linux_binfmt elf_fdpic_format = {
        .module         = THIS_MODULE,
        .load_binary    = load_elf_fdpic_binary,
-//     .core_dump      = elf_fdpic_core_dump,
+#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
+       .core_dump      = elf_fdpic_core_dump,
+#endif
        .min_coredump   = ELF_EXEC_PAGESIZE,
 };
 
@@ -87,7 +101,7 @@ static void __exit exit_elf_fdpic_binfmt(void)
        unregister_binfmt(&elf_fdpic_format);
 }
 
-module_init(init_elf_fdpic_binfmt);
+core_initcall(init_elf_fdpic_binfmt);
 module_exit(exit_elf_fdpic_binfmt);
 
 static int is_elf_fdpic(struct elfhdr *hdr, struct file *file)
@@ -1118,3 +1132,661 @@ static int elf_fdpic_map_file_by_direct_mmap(struct elf_fdpic_params *params,
 
        return 0;
 }
+
+/*****************************************************************************/
+/*
+ * ELF-FDPIC core dumper
+ *
+ * Modelled on fs/exec.c:aout_core_dump()
+ * Jeremy Fitzhardinge <jeremy@sw.oz.au>
+ *
+ * Modelled on fs/binfmt_elf.c core dumper
+ */
+#if defined(USE_ELF_CORE_DUMP) && defined(CONFIG_ELF_CORE)
+
+/*
+ * 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) {
+               if (file->f_op->llseek(file, off, SEEK_SET) != off)
+                       return 0;
+       } else {
+               file->f_pos = off;
+       }
+       return 1;
+}
+
+/*
+ * 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.
+ */
+static int maydump(struct vm_area_struct *vma)
+{
+       /* Do not dump I/O mapped devices or special mappings */
+       if (vma->vm_flags & (VM_IO | VM_RESERVED)) {
+               kdcore("%08lx: %08lx: no (IO)", vma->vm_start, vma->vm_flags);
+               return 0;
+       }
+
+       /* If we may not read the contents, don't allow us to dump
+        * them either. "dump_write()" can't handle it anyway.
+        */
+       if (!(vma->vm_flags & VM_READ)) {
+               kdcore("%08lx: %08lx: no (!read)", vma->vm_start, vma->vm_flags);
+               return 0;
+       }
+
+       /* Dump shared memory only if mapped from an anonymous file. */
+       if (vma->vm_flags & VM_SHARED) {
+               if (vma->vm_file->f_dentry->d_inode->i_nlink == 0) {
+                       kdcore("%08lx: %08lx: no (share)", vma->vm_start, vma->vm_flags);
+                       return 1;
+               }
+
+               kdcore("%08lx: %08lx: no (share)", vma->vm_start, vma->vm_flags);
+               return 0;
+       }
+
+#ifdef CONFIG_MMU
+       /* If it hasn't been written to, don't write it out */
+       if (!vma->anon_vma) {
+               kdcore("%08lx: %08lx: no (!anon)", vma->vm_start, vma->vm_flags);
+               return 0;
+       }
+#endif
+
+       kdcore("%08lx: %08lx: yes", vma->vm_start, vma->vm_flags);
+       return 1;
+}
+
+/* An ELF note in memory */
+struct memelfnote
+{
+       const char *name;
+       int type;
+       unsigned int datasz;
+       void *data;
+};
+
+static int notesize(struct memelfnote *en)
+{
+       int sz;
+
+       sz = sizeof(struct elf_note);
+       sz += roundup(strlen(en->name) + 1, 4);
+       sz += roundup(en->datasz, 4);
+
+       return sz;
+}
+
+/* #define DEBUG */
+
+#define DUMP_WRITE(addr, nr)   \
+       do { if (!dump_write(file, (addr), (nr))) return 0; } while(0)
+#define DUMP_SEEK(off) \
+       do { if (!dump_seek(file, (off))) return 0; } while(0)
+
+static int writenote(struct memelfnote *men, struct file *file)
+{
+       struct elf_note en;
+
+       en.n_namesz = strlen(men->name) + 1;
+       en.n_descsz = men->datasz;
+       en.n_type = men->type;
+
+       DUMP_WRITE(&en, sizeof(en));
+       DUMP_WRITE(men->name, en.n_namesz);
+       /* XXX - cast from long long to long to avoid need for libgcc.a */
+       DUMP_SEEK(roundup((unsigned long)file->f_pos, 4));      /* XXX */
+       DUMP_WRITE(men->data, men->datasz);
+       DUMP_SEEK(roundup((unsigned long)file->f_pos, 4));      /* XXX */
+
+       return 1;
+}
+#undef DUMP_WRITE
+#undef DUMP_SEEK
+
+#define DUMP_WRITE(addr, nr)   \
+       if ((size += (nr)) > limit || !dump_write(file, (addr), (nr))) \
+               goto end_coredump;
+#define DUMP_SEEK(off) \
+       if (!dump_seek(file, (off))) \
+               goto end_coredump;
+
+static inline void fill_elf_fdpic_header(struct elfhdr *elf, int segs)
+{
+       memcpy(elf->e_ident, ELFMAG, SELFMAG);
+       elf->e_ident[EI_CLASS] = ELF_CLASS;
+       elf->e_ident[EI_DATA] = ELF_DATA;
+       elf->e_ident[EI_VERSION] = EV_CURRENT;
+       elf->e_ident[EI_OSABI] = ELF_OSABI;
+       memset(elf->e_ident+EI_PAD, 0, EI_NIDENT-EI_PAD);
+
+       elf->e_type = ET_CORE;
+       elf->e_machine = ELF_ARCH;
+       elf->e_version = EV_CURRENT;
+       elf->e_entry = 0;
+       elf->e_phoff = sizeof(struct elfhdr);
+       elf->e_shoff = 0;
+       elf->e_flags = ELF_FDPIC_CORE_EFLAGS;
+       elf->e_ehsize = sizeof(struct elfhdr);
+       elf->e_phentsize = sizeof(struct elf_phdr);
+       elf->e_phnum = segs;
+       elf->e_shentsize = 0;
+       elf->e_shnum = 0;
+       elf->e_shstrndx = 0;
+       return;
+}
+
+static inline void fill_elf_note_phdr(struct elf_phdr *phdr, int sz, loff_t offset)
+{
+       phdr->p_type = PT_NOTE;
+       phdr->p_offset = offset;
+       phdr->p_vaddr = 0;
+       phdr->p_paddr = 0;
+       phdr->p_filesz = sz;
+       phdr->p_memsz = 0;
+       phdr->p_flags = 0;
+       phdr->p_align = 0;
+       return;
+}
+
+static inline void fill_note(struct memelfnote *note, const char *name, int type,
+               unsigned int sz, void *data)
+{
+       note->name = name;
+       note->type = type;
+       note->datasz = sz;
+       note->data = data;
+       return;
+}
+
+/*
+ * fill up all the fields in prstatus from the given task struct, except
+ * registers which need to be filled up seperately.
+ */
+static void fill_prstatus(struct elf_prstatus *prstatus,
+                         struct task_struct *p, long signr)
+{
+       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 = p->signal->session;
+       if (thread_group_leader(p)) {
+               /*
+                * This is the record for the group leader.  Add in the
+                * cumulative times of previous dead threads.  This total
+                * won't include the time of each live thread whose state
+                * is included in the core dump.  The final total reported
+                * to our parent process when it calls wait4 will include
+                * those sums as well as the little bit more time it takes
+                * this and each other thread to finish dying after the
+                * core dump synchronization phase.
+                */
+               cputime_to_timeval(cputime_add(p->utime, p->signal->utime),
+                                  &prstatus->pr_utime);
+               cputime_to_timeval(cputime_add(p->stime, p->signal->stime),
+                                  &prstatus->pr_stime);
+       } else {
+               cputime_to_timeval(p->utime, &prstatus->pr_utime);
+               cputime_to_timeval(p->stime, &prstatus->pr_stime);
+       }
+       cputime_to_timeval(p->signal->cutime, &prstatus->pr_cutime);
+       cputime_to_timeval(p->signal->cstime, &prstatus->pr_cstime);
+
+       prstatus->pr_exec_fdpic_loadmap = p->mm->context.exec_fdpic_loadmap;
+       prstatus->pr_interp_fdpic_loadmap = p->mm->context.interp_fdpic_loadmap;
+}
+
+static int fill_psinfo(struct elf_prpsinfo *psinfo, struct task_struct *p,
+                      struct mm_struct *mm)
+{
+       unsigned int i, len;
+
+       /* first copy the parameters from user space */
+       memset(psinfo, 0, sizeof(struct elf_prpsinfo));
+
+       len = mm->arg_end - mm->arg_start;
+       if (len >= ELF_PRARGSZ)
+               len = ELF_PRARGSZ - 1;
+       if (copy_from_user(&psinfo->pr_psargs,
+                          (const char __user *) mm->arg_start, len))
+               return -EFAULT;
+       for (i = 0; i < len; i++)
+               if (psinfo->pr_psargs[i] == 0)
+                       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 = p->signal->session;
+
+       i = p->state ? ffz(~p->state) + 1 : 0;
+       psinfo->pr_state = i;
+       psinfo->pr_sname = (i > 5) ? '.' : "RSDTZW"[i];
+       psinfo->pr_zomb = psinfo->pr_sname == 'Z';
+       psinfo->pr_nice = task_nice(p);
+       psinfo->pr_flag = p->flags;
+       SET_UID(psinfo->pr_uid, p->uid);
+       SET_GID(psinfo->pr_gid, p->gid);
+       strncpy(psinfo->pr_fname, p->comm, sizeof(psinfo->pr_fname));
+
+       return 0;
+}
+
+/* Here is the structure in which status of each thread is captured. */
+struct elf_thread_status
+{
+       struct list_head list;
+       struct elf_prstatus prstatus;   /* NT_PRSTATUS */
+       elf_fpregset_t fpu;             /* NT_PRFPREG */
+       struct task_struct *thread;
+#ifdef ELF_CORE_COPY_XFPREGS
+       elf_fpxregset_t xfpu;           /* NT_PRXFPREG */
+#endif
+       struct memelfnote notes[3];
+       int num_notes;
+};
+
+/*
+ * In order to add the specific thread information for the elf file format,
+ * we need to keep a linked list of every thread's pr_status and then create
+ * a single section for them in the final core file.
+ */
+static int elf_dump_thread_status(long signr, struct elf_thread_status *t)
+{
+       struct task_struct *p = t->thread;
+       int sz = 0;
+
+       t->num_notes = 0;
+
+       fill_prstatus(&t->prstatus, p, signr);
+       elf_core_copy_task_regs(p, &t->prstatus.pr_reg);
+
+       fill_note(&t->notes[0], "CORE", NT_PRSTATUS, sizeof(t->prstatus),
+                 &t->prstatus);
+       t->num_notes++;
+       sz += notesize(&t->notes[0]);
+
+       t->prstatus.pr_fpvalid = elf_core_copy_task_fpregs(p, NULL, &t->fpu);
+       if (t->prstatus.pr_fpvalid) {
+               fill_note(&t->notes[1], "CORE", NT_PRFPREG, sizeof(t->fpu),
+                         &t->fpu);
+               t->num_notes++;
+               sz += notesize(&t->notes[1]);
+       }
+
+#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);
+               t->num_notes++;
+               sz += notesize(&t->notes[2]);
+       }
+#endif
+       return sz;
+}
+
+/*
+ * dump the segments for an MMU process
+ */
+#ifdef CONFIG_MMU
+static int elf_fdpic_dump_segments(struct file *file, struct mm_struct *mm,
+                                  size_t *size, unsigned long *limit)
+{
+       struct vm_area_struct *vma;
+
+       for (vma = current->mm->mmap; vma; vma = vma->vm_next) {
+               unsigned long addr;
+
+               if (!maydump(vma))
+                       continue;
+
+               for (addr = vma->vm_start;
+                    addr < vma->vm_end;
+                    addr += PAGE_SIZE
+                    ) {
+                       struct vm_area_struct *vma;
+                       struct page *page;
+
+                       if (get_user_pages(current, current->mm, addr, 1, 0, 1,
+                                          &page, &vma) <= 0) {
+                               DUMP_SEEK(file->f_pos + PAGE_SIZE);
+                       }
+                       else if (page == ZERO_PAGE(addr)) {
+                               DUMP_SEEK(file->f_pos + PAGE_SIZE);
+                               page_cache_release(page);
+                       }
+                       else {
+                               void *kaddr;
+
+                               flush_cache_page(vma, addr, page_to_pfn(page));
+                               kaddr = kmap(page);
+                               if ((*size += PAGE_SIZE) > *limit ||
+                                   !dump_write(file, kaddr, PAGE_SIZE)
+                                   ) {
+                                       kunmap(page);
+                                       page_cache_release(page);
+                                       return -EIO;
+                               }
+                               kunmap(page);
+                               page_cache_release(page);
+                       }
+               }
+       }
+
+       return 0;
+
+end_coredump:
+       return -EFBIG;
+}
+#endif
+
+/*
+ * dump the segments for a NOMMU process
+ */
+#ifndef CONFIG_MMU
+static int elf_fdpic_dump_segments(struct file *file, struct mm_struct *mm,
+                                  size_t *size, unsigned long *limit)
+{
+       struct vm_list_struct *vml;
+
+       for (vml = current->mm->context.vmlist; vml; vml = vml->next) {
+       struct vm_area_struct *vma = vml->vma;
+
+               if (!maydump(vma))
+                       continue;
+
+               if ((*size += PAGE_SIZE) > *limit)
+                       return -EFBIG;
+
+               if (!dump_write(file, (void *) vma->vm_start,
+                               vma->vm_end - vma->vm_start))
+                       return -EIO;
+       }
+
+       return 0;
+}
+#endif
+
+/*
+ * Actual dumper
+ *
+ * This is a two-pass process; first we find the offsets of the bits,
+ * and then they are actually written out.  If we run out of core limit
+ * we just truncate.
+ */
+static int elf_fdpic_core_dump(long signr, struct pt_regs *regs,
+                              struct file *file)
+{
+#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;
+       struct elfhdr *elf = NULL;
+       loff_t offset = 0, dataoff;
+       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;
+#ifndef CONFIG_MMU
+       struct vm_list_struct *vml;
+#endif
+       elf_addr_t *auxv;
+
+       /*
+        * We no longer stop all VM operations.
+        *
+        * This is because those proceses that could possibly change map_count
+        * or the mmap / vma pages are now blocked in do_exit on current
+        * finishing this core dump.
+        *
+        * Only ptrace can touch these memory addresses, but it doesn't change
+        * the map_count or the pages allocated. So no possibility of crashing
+        * exists while dumping the mm->vm_next areas to the core file.
+        */
+
+       /* alloc memory for large data structures: too large to be on stack */
+       elf = kmalloc(sizeof(*elf), GFP_KERNEL);
+       if (!elf)
+               goto cleanup;
+       prstatus = kzalloc(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;
+               read_lock(&tasklist_lock);
+               do_each_thread(g,p)
+                       if (current->mm == p->mm && current != p) {
+                               tmp = kzalloc(sizeof(*tmp), GFP_ATOMIC);
+                               if (!tmp) {
+                                       read_unlock(&tasklist_lock);
+                                       goto cleanup;
+                               }
+                               INIT_LIST_HEAD(&tmp->list);
+                               tmp->thread = p;
+                               list_add(&tmp->list, &thread_list);
+                       }
+               while_each_thread(g,p);
+               read_unlock(&tasklist_lock);
+               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 */
+       fill_prstatus(prstatus, current, signr);
+       elf_core_copy_regs(&prstatus->pr_reg, regs);
+
+#ifdef CONFIG_MMU
+       segs = current->mm->map_count;
+#else
+       segs = 0;
+       for (vml = current->mm->context.vmlist; vml; vml = vml->next)
+           segs++;
+#endif
+#ifdef ELF_CORE_EXTRA_PHDRS
+       segs += ELF_CORE_EXTRA_PHDRS;
+#endif
+
+       /* Set up header */
+       fill_elf_fdpic_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.
+        */
+
+       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
+
+       fs = get_fs();
+       set_fs(KERNEL_DS);
+
+       DUMP_WRITE(elf, sizeof(*elf));
+       offset += sizeof(*elf);                         /* Elf header */
+       offset += (segs+1) * sizeof(struct elf_phdr);   /* Program headers */
+
+       /* 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;
+
+               fill_elf_note_phdr(&phdr, sz, offset);
+               offset += sz;
+               DUMP_WRITE(&phdr, sizeof(phdr));
+       }
+
+       /* Page-align dumped data */
+       dataoff = offset = roundup(offset, ELF_EXEC_PAGESIZE);
+
+       /* write program headers for segments dump */
+       for (
+#ifdef CONFIG_MMU
+               vma = current->mm->mmap; vma; vma = vma->vm_next
+#else
+                       vml = current->mm->context.vmlist; vml; vml = vml->next
+#endif
+            ) {
+               struct elf_phdr phdr;
+               size_t sz;
+
+#ifndef CONFIG_MMU
+               vma = vml->vma;
+#endif
+
+               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) ? sz : 0;
+               phdr.p_memsz = sz;
+               offset += phdr.p_filesz;
+               phdr.p_flags = vma->vm_flags & VM_READ ? PF_R : 0;
+               if (vma->vm_flags & VM_WRITE)
+                       phdr.p_flags |= PF_W;
+               if (vma->vm_flags & VM_EXEC)
+                       phdr.p_flags |= PF_X;
+               phdr.p_align = ELF_EXEC_PAGESIZE;
+
+               DUMP_WRITE(&phdr, sizeof(phdr));
+       }
+
+#ifdef ELF_CORE_WRITE_EXTRA_PHDRS
+       ELF_CORE_WRITE_EXTRA_PHDRS;
+#endif
+
+       /* write out the notes section */
+       for (i = 0; i < numnote; i++)
+               if (!writenote(notes + i, file))
+                       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))
+                               goto end_coredump;
+       }
+
+       DUMP_SEEK(dataoff);
+
+       if (elf_fdpic_dump_segments(file, current->mm, &size, &limit) < 0)
+               goto end_coredump;
+
+#ifdef ELF_CORE_WRITE_EXTRA_DATA
+       ELF_CORE_WRITE_EXTRA_DATA;
+#endif
+
+       if (file->f_pos != offset) {
+               /* Sanity check */
+               printk(KERN_WARNING
+                      "elf_core_dump: file->f_pos (%lld) != offset (%lld)\n",
+                      file->f_pos, offset);
+       }
+
+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
+       return has_dumped;
+#undef NUM_NOTES
+}
+
+#endif         /* USE_ELF_CORE_DUMP */
index 38656da00e4005ffe1f26647195595cf5fdd22cc..7df58a3e6e4a5537e69ea2fb97cc1be4c16c7368 100644 (file)
@@ -64,7 +64,7 @@ typedef unsigned long elf_greg_t;
 #define ELF_NGREG (sizeof(struct pt_regs) / sizeof(elf_greg_t))
 typedef elf_greg_t elf_gregset_t[ELF_NGREG];
 
-typedef struct fpmedia_struct elf_fpregset_t;
+typedef struct user_fpmedia_regs elf_fpregset_t;
 
 /*
  * This is used to ensure we don't load something for the wrong architecture.
@@ -116,6 +116,7 @@ do {                                                                                        \
 } while(0)
 
 #define USE_ELF_CORE_DUMP
+#define ELF_FDPIC_CORE_EFLAGS  EF_FRV_FDPIC
 #define ELF_EXEC_PAGESIZE      16384
 
 /* This is the location that an ET_DYN program is loaded if exec'ed.  Typical
@@ -125,9 +126,6 @@ do {                                                                                        \
 
 #define ELF_ET_DYN_BASE         0x08000000UL
 
-#define ELF_CORE_COPY_REGS(pr_reg, regs)                               \
-       memcpy(&pr_reg[0], &regs->sp, 31 * sizeof(uint32_t));
-
 /* This yields a mask that user programs can use to figure out what
    instruction set this cpu supports.  */
 
index 0cf0bea010fe2110417f98e83c3cbf60f94f2da5..9631dddae3486b337b628ce6f0a2986db2bc6172 100644 (file)
@@ -60,6 +60,16 @@ struct elf_prstatus
        long    pr_instr;               /* Current instruction */
 #endif
        elf_gregset_t pr_reg;   /* GP registers */
+#ifdef CONFIG_BINFMT_ELF_FDPIC
+       /* When using FDPIC, the loadmap addresses need to be communicated
+        * to GDB in order for GDB to do the necessary relocations.  The
+        * fields (below) used to communicate this information are placed
+        * immediately after ``pr_reg'', so that the loadmap addresses may
+        * be viewed as part of the register set if so desired.
+        */
+       unsigned long pr_exec_fdpic_loadmap;
+       unsigned long pr_interp_fdpic_loadmap;
+#endif
        int pr_fpvalid;         /* True if math co-processor being used.  */
 };