Merge projects/enhanced_coredumps (r204346) into HEAD:

Enhanced process coredump routines.

  This brings in the following features:
  1) Limit number of cores per process via the %I coredump formatter.
  Example:
    if corefilename is set to %N.%I.core AND num_cores = 3, then
    if a process "rpd" cores, then the corefile will be named
    "rpd.0.core", however if it cores again, then the kernel will
    generate "rpd.1.core" until we hit the limit of "num_cores".

    this is useful to get several corefiles, but also prevent filling
    the machine with corefiles.

  2) Encode machine hostname in core dump name via %H.

  3) Compress coredumps, useful for embedded platforms with limited space.
    A sysctl kern.compress_user_cores is made available if turned on.

    To enable compressed coredumps, the following config options need to be set:
    options COMPRESS_USER_CORES
    device zlib   # brings in the zlib requirements.
    device gzio   # brings in the kernel vnode gzip output module.

  4) Eventhandlers are fired to indicate coredumps in progress.

  5) The imgact sv_coredump routine has grown a flag to pass in more
  state, currently this is used only for passing a flag down to compress
  the coredump or not.

  Note that the gzio facility can be used for generic output of gzip'd
  streams via vnodes.

Obtained from: Juniper Networks
Reviewed by: kan

sys/conf/files

@@ -2053,6 +2053,7 @@ kern/kern_exec.c		standard
 kern/kern_exit.c		standard
 kern/kern_fail.c		standard
 kern/kern_fork.c		standard
+kern/kern_gzio.c		optional gzio
 kern/kern_idle.c		standard
 kern/kern_intr.c		standard
 kern/kern_jail.c		standard
@@ -2344,7 +2345,7 @@ net/slcompress.c		optional netgraph_vjc | sppp | \
 net/vnet.c			optional vimage
 net/zlib.c			optional crypto | geom_uzip | ipsec | \
 					 mxge | netgraph_deflate | \
-					 ddb_ctf
+					 ddb_ctf | zlib
 net80211/ieee80211.c		optional wlan
 net80211/ieee80211_acl.c	optional wlan wlan_acl
 net80211/ieee80211_action.c	optional wlan

sys/conf/options

@@ -71,6 +71,7 @@ COMPAT_FREEBSD5	opt_compat.h
 COMPAT_FREEBSD6	opt_compat.h
 COMPAT_FREEBSD7	opt_compat.h
 COMPILING_LINT	opt_global.h
+COMPRESS_USER_CORES opt_core.h
 CY_PCI_FASTINTR
 DEADLKRES	opt_watchdog.h
 DIRECTIO

sys/kern/imgact_elf.c

@@ -32,6 +32,7 @@
 __FBSDID("$FreeBSD$");
 
 #include "opt_compat.h"
+#include "opt_core.h"
 
 #include <sys/param.h>
 #include <sys/exec.h>
@@ -58,6 +59,10 @@ __FBSDID("$FreeBSD$");
 #include <sys/sysctl.h>
 #include <sys/sysent.h>
 #include <sys/vnode.h>
+#include <sys/syslog.h>
+#include <sys/eventhandler.h>
+
+#include <net/zlib.h>
 
 #include <vm/vm.h>
 #include <vm/vm_kern.h>
@@ -95,6 +100,12 @@ static boolean_t __elfN(check_note)(struct image_params *imgp,
 SYSCTL_NODE(_kern, OID_AUTO, __CONCAT(elf, __ELF_WORD_SIZE), CTLFLAG_RW, 0,
     "");
 
+#ifdef COMPRESS_USER_CORES
+static int compress_core(gzFile, char *, char *, unsigned int,
+    struct thread * td);
+#define CORE_BUF_SIZE	(16 * 1024)
+#endif
+
 int __elfN(fallback_brand) = -1;
 SYSCTL_INT(__CONCAT(_kern_elf, __ELF_WORD_SIZE), OID_AUTO,
     fallback_brand, CTLFLAG_RW, &__elfN(fallback_brand), 0,
@@ -1003,16 +1014,38 @@ static void cb_put_phdr(vm_map_entry_t, void *);
 static void cb_size_segment(vm_map_entry_t, void *);
 static void each_writable_segment(struct thread *, segment_callback, void *);
 static int __elfN(corehdr)(struct thread *, struct vnode *, struct ucred *,
-    int, void *, size_t);
+    int, void *, size_t, gzFile);
 static void __elfN(puthdr)(struct thread *, void *, size_t *, int);
 static void __elfN(putnote)(void *, size_t *, const char *, int,
     const void *, size_t);
 
+#ifdef COMPRESS_USER_CORES
+extern int compress_user_cores;
+extern int compress_user_cores_gzlevel;
+#endif
+
+static int
+core_output(struct vnode *vp, void *base, size_t len, off_t offset,
+    struct ucred *active_cred, struct ucred *file_cred,
+    struct thread *td, char *core_buf, gzFile gzfile) {
+
+	int error;
+	if (gzfile) {
+#ifdef COMPRESS_USER_CORES
+		error = compress_core(gzfile, base, core_buf, len, td);
+#else
+		panic("shouldn't be here");
+#endif
+	} else {
+		error = vn_rdwr_inchunks(UIO_WRITE, vp, base, len, offset,
+		    UIO_USERSPACE, IO_UNIT | IO_DIRECT, active_cred, file_cred,
+		    NULL, td);
+	}
+	return (error);
+}
+
 int
-__elfN(coredump)(td, vp, limit)
+__elfN(coredump)(struct thread *td, struct vnode *vp, off_t limit, int flags)
-	struct thread *td;
-	struct vnode *vp;
-	off_t limit;
 {
 	struct ucred *cred = td->td_ucred;
 	int error = 0;
@@ -1020,6 +1053,37 @@ __elfN(coredump)(td, vp, limit)
 	void *hdr;
 	size_t hdrsize;
 
+	gzFile gzfile = Z_NULL;
+	char *core_buf = NULL;
+#ifdef COMPRESS_USER_CORES
+	char gzopen_flags[8];
+	char *p;
+	int doing_compress = flags & IMGACT_CORE_COMPRESS;
+#endif
+
+	hdr = NULL;
+
+#ifdef COMPRESS_USER_CORES
+        if (doing_compress) {
+                p = gzopen_flags;
+                *p++ = 'w';
+                if (compress_user_cores_gzlevel >= 0 &&
+                    compress_user_cores_gzlevel <= 9)
+                        *p++ = '0' + compress_user_cores_gzlevel;
+                *p = 0;
+                gzfile = gz_open("", gzopen_flags, vp);
+                if (gzfile == Z_NULL) {
+                        error = EFAULT;
+                        goto done;
+                }
+                core_buf = malloc(CORE_BUF_SIZE, M_TEMP, M_WAITOK | M_ZERO);
+                if (!core_buf) {
+                        error = ENOMEM;
+                        goto done;
+                }
+        }
+#endif
+
 	/* Size the program segments. */
 	seginfo.count = 0;
 	seginfo.size = 0;
@@ -1044,7 +1108,8 @@ __elfN(coredump)(td, vp, limit)
 	if (hdr == NULL) {
 		return (EINVAL);
 	}
-	error = __elfN(corehdr)(td, vp, cred, seginfo.count, hdr, hdrsize);
+	error = __elfN(corehdr)(td, vp, cred, seginfo.count, hdr, hdrsize,
+	    gzfile);
 
 	/* Write the contents of all of the writable segments. */
 	if (error == 0) {
@@ -1055,17 +1120,28 @@ __elfN(coredump)(td, vp, limit)
 		php = (Elf_Phdr *)((char *)hdr + sizeof(Elf_Ehdr)) + 1;
 		offset = hdrsize;
 		for (i = 0; i < seginfo.count; i++) {
-			error = vn_rdwr_inchunks(UIO_WRITE, vp,
+			error = core_output(vp, (caddr_t)(uintptr_t)php->p_vaddr,
-			    (caddr_t)(uintptr_t)php->p_vaddr,
+			    php->p_filesz, offset, cred, NOCRED, curthread, core_buf, gzfile);
-			    php->p_filesz, offset, UIO_USERSPACE,
-			    IO_UNIT | IO_DIRECT, cred, NOCRED, NULL,
-			    curthread);
 			if (error != 0)
 				break;
 			offset += php->p_filesz;
 			php++;
 		}
 	}
+	if (error) {
+		log(LOG_WARNING,
+		    "Failed to write core file for process %s (error %d)\n",
+		    curproc->p_comm, error);
+	}
+
+#ifdef COMPRESS_USER_CORES
+done:
+#endif
+	if (core_buf)
+		free(core_buf, M_TEMP);
+	if (gzfile)
+		gzclose(gzfile);
+
 	free(hdr, M_TEMP);
 
 	return (error);
@@ -1189,13 +1265,14 @@ each_writable_segment(td, func, closure)
  * the page boundary.
  */
 static int
-__elfN(corehdr)(td, vp, cred, numsegs, hdr, hdrsize)
+__elfN(corehdr)(td, vp, cred, numsegs, hdr, hdrsize, gzfile)
 	struct thread *td;
 	struct vnode *vp;
 	struct ucred *cred;
 	int numsegs;
 	size_t hdrsize;
 	void *hdr;
+	gzFile gzfile;
 {
 	size_t off;
 
@@ -1204,10 +1281,26 @@ __elfN(corehdr)(td, vp, cred, numsegs, hdr, hdrsize)
 	off = 0;
 	__elfN(puthdr)(td, hdr, &off, numsegs);
 
-	/* Write it to the core file. */
+	if (!gzfile) {
-	return (vn_rdwr_inchunks(UIO_WRITE, vp, hdr, hdrsize, (off_t)0,
+		/* Write it to the core file. */
-	    UIO_SYSSPACE, IO_UNIT | IO_DIRECT, cred, NOCRED, NULL,
+		return (vn_rdwr_inchunks(UIO_WRITE, vp, hdr, hdrsize, (off_t)0,
-	    td));
+			UIO_SYSSPACE, IO_UNIT | IO_DIRECT, cred, NOCRED, NULL,
+			td));
+	} else {
+#ifdef COMPRESS_USER_CORES
+		if (gzwrite(gzfile, hdr, hdrsize) != hdrsize) {
+			log(LOG_WARNING,
+			    "Failed to compress core file header for process"
+			    " %s.\n", curproc->p_comm);
+			return (EFAULT);
+		}
+		else {
+			return (0);
+		}
+#else
+		panic("shouldn't be here");
+#endif
+	}
 }
 
 #if defined(COMPAT_IA32) && __ELF_WORD_SIZE == 32
@@ -1476,3 +1569,50 @@ static struct execsw __elfN(execsw) = {
 	__XSTRING(__CONCAT(ELF, __ELF_WORD_SIZE))
 };
 EXEC_SET(__CONCAT(elf, __ELF_WORD_SIZE), __elfN(execsw));
+
+#ifdef COMPRESS_USER_CORES
+/*
+ * Compress and write out a core segment for a user process.
+ *
+ * 'inbuf' is the starting address of a VM segment in the process' address
+ * space that is to be compressed and written out to the core file.  'dest_buf'
+ * is a buffer in the kernel's address space.  The segment is copied from 
+ * 'inbuf' to 'dest_buf' first before being processed by the compression
+ * routine gzwrite().  This copying is necessary because the content of the VM
+ * segment may change between the compression pass and the crc-computation pass
+ * in gzwrite().  This is because realtime threads may preempt the UNIX kernel.
+ */
+static int
+compress_core (gzFile file, char *inbuf, char *dest_buf, unsigned int len,
+    struct thread *td)
+{
+	int len_compressed;
+	int error = 0;
+	unsigned int chunk_len;
+
+	while (len) {
+		chunk_len = (len > CORE_BUF_SIZE) ? CORE_BUF_SIZE : len;
+		copyin(inbuf, dest_buf, chunk_len);
+		len_compressed = gzwrite(file, dest_buf, chunk_len);
+
+		EVENTHANDLER_INVOKE(app_coredump_progress, td, len_compressed);
+
+		if ((unsigned int)len_compressed != chunk_len) {
+			log(LOG_WARNING,
+			    "compress_core: length mismatch (0x%x returned, "
+			    "0x%x expected)\n", len_compressed, chunk_len);
+			EVENTHANDLER_INVOKE(app_coredump_error, td,
+			    "compress_core: length mismatch %x -> %x",
+			    chunk_len, len_compressed);
+			error = EFAULT;
+			break;
+		}
+		inbuf += chunk_len;
+		len -= chunk_len;
+		if (ticks - PCPU_GET(switchticks) >= hogticks)
+			uio_yield();
+	}
+
+	return (error);
+}
+#endif /* COMPRESS_USER_CORES */

sys/kern/kern_gzio.c

@@ -0,0 +1,406 @@
+/*
+ * $Id: kern_gzio.c,v 1.6 2008-10-18 22:54:45 lbazinet Exp $
+ *
+ * core_gzip.c -- gzip routines used in compressing user process cores
+ *
+ * This file is derived from src/lib/libz/gzio.c in FreeBSD.
+ */
+
+/* gzio.c -- IO on .gz files
+ * Copyright (C) 1995-1998 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ *
+ */
+
+/* @(#) $FreeBSD$ */
+
+#include <sys/param.h>
+#include <sys/proc.h>
+#include <sys/malloc.h>
+#include <sys/vnode.h>
+#include <sys/syslog.h>
+#include <sys/endian.h>
+#include <net/zutil.h>
+#include <sys/libkern.h>
+
+#include <sys/vnode.h>
+#include <sys/mount.h>
+
+#define GZ_HEADER_LEN	10
+
+#ifndef Z_BUFSIZE
+#  ifdef MAXSEG_64K
+#    define Z_BUFSIZE 4096 /* minimize memory usage for 16-bit DOS */
+#  else
+#    define Z_BUFSIZE 16384
+#  endif
+#endif
+#ifndef Z_PRINTF_BUFSIZE
+#  define Z_PRINTF_BUFSIZE 4096
+#endif
+
+#define ALLOC(size) malloc(size, M_TEMP, M_WAITOK | M_ZERO)
+#define TRYFREE(p) {if (p) free(p, M_TEMP);}
+
+static int gz_magic[2] = {0x1f, 0x8b}; /* gzip magic header */
+
+/* gzip flag byte */
+#define ASCII_FLAG   0x01 /* bit 0 set: file probably ascii text */
+#define HEAD_CRC     0x02 /* bit 1 set: header CRC present */
+#define EXTRA_FIELD  0x04 /* bit 2 set: extra field present */
+#define ORIG_NAME    0x08 /* bit 3 set: original file name present */
+#define COMMENT      0x10 /* bit 4 set: file comment present */
+#define RESERVED     0xE0 /* bits 5..7: reserved */
+
+typedef struct gz_stream {
+    z_stream stream;
+    int      z_err;   /* error code for last stream operation */
+    int      z_eof;   /* set if end of input file */
+    struct vnode *file; /* vnode pointer of .gz file */
+    Byte     *inbuf;  /* input buffer */
+    Byte     *outbuf; /* output buffer */
+    uLong    crc;     /* crc32 of uncompressed data */
+    char     *msg;    /* error message */
+    char     *path;   /* path name for debugging only */
+    int      transparent; /* 1 if input file is not a .gz file */
+    char     mode;    /* 'w' or 'r' */
+    long     startpos; /* start of compressed data in file (header skipped) */
+    off_t    outoff;  /* current offset in output file */
+    int	     flags;
+} gz_stream;
+
+
+local int do_flush        OF((gzFile file, int flush));
+local int    destroy      OF((gz_stream *s));
+local void   putU32      OF((gz_stream *file, uint32_t x));
+local void *gz_alloc      OF((void *notused, u_int items, u_int size));
+local void gz_free        OF((void *notused, void *ptr));
+
+/* ===========================================================================
+     Opens a gzip (.gz) file for reading or writing. The mode parameter
+   is as in fopen ("rb" or "wb"). The file is given either by file descriptor
+   or path name (if fd == -1).
+     gz_open return NULL if the file could not be opened or if there was
+   insufficient memory to allocate the (de)compression state; errno
+   can be checked to distinguish the two cases (if errno is zero, the
+   zlib error is Z_MEM_ERROR).
+*/
+gzFile gz_open (path, mode, vp)
+    const char *path;
+    const char *mode;
+    struct vnode *vp;
+{
+    int err;
+    int level = Z_DEFAULT_COMPRESSION; /* compression level */
+    int strategy = Z_DEFAULT_STRATEGY; /* compression strategy */
+    const char *p = mode;
+    gz_stream *s;
+    char fmode[80]; /* copy of mode, without the compression level */
+    char *m = fmode;
+    int resid;
+    int error;
+    char buf[GZ_HEADER_LEN + 1];
+
+    if (!path || !mode) return Z_NULL;
+
+    s = (gz_stream *)ALLOC(sizeof(gz_stream));
+    if (!s) return Z_NULL;
+
+    s->stream.zalloc = (alloc_func)gz_alloc;
+    s->stream.zfree = (free_func)gz_free;
+    s->stream.opaque = (voidpf)0;
+    s->stream.next_in = s->inbuf = Z_NULL;
+    s->stream.next_out = s->outbuf = Z_NULL;
+    s->stream.avail_in = s->stream.avail_out = 0;
+    s->file = NULL;
+    s->z_err = Z_OK;
+    s->z_eof = 0;
+    s->crc = 0;
+    s->msg = NULL;
+    s->transparent = 0;
+    s->outoff = 0;
+    s->flags = 0;
+
+    s->path = (char*)ALLOC(strlen(path)+1);
+    if (s->path == NULL) {
+        return destroy(s), (gzFile)Z_NULL;
+    }
+    strcpy(s->path, path); /* do this early for debugging */
+
+    s->mode = '\0';
+    do {
+        if (*p == 'r') s->mode = 'r';
+        if (*p == 'w' || *p == 'a') s->mode = 'w';
+        if (*p >= '0' && *p <= '9') {
+	    level = *p - '0';
+	} else if (*p == 'f') {
+	  strategy = Z_FILTERED;
+	} else if (*p == 'h') {
+	  strategy = Z_HUFFMAN_ONLY;
+	} else {
+	    *m++ = *p; /* copy the mode */
+	}
+    } while (*p++ && m != fmode + sizeof(fmode));
+
+    if (s->mode != 'w') {
+        log(LOG_ERR, "gz_open: mode is not w (%c)\n", s->mode);
+        return destroy(s), (gzFile)Z_NULL;
+    }
+    
+    err = deflateInit2(&(s->stream), level,
+                       Z_DEFLATED, -MAX_WBITS, DEF_MEM_LEVEL, strategy);
+    /* windowBits is passed < 0 to suppress zlib header */
+
+    s->stream.next_out = s->outbuf = (Byte*)ALLOC(Z_BUFSIZE);
+    if (err != Z_OK || s->outbuf == Z_NULL) {
+        return destroy(s), (gzFile)Z_NULL;
+    }
+
+    s->stream.avail_out = Z_BUFSIZE;
+    s->file = vp;
+
+    /* Write a very simple .gz header:
+     */
+    snprintf(buf, sizeof(buf), "%c%c%c%c%c%c%c%c%c%c", gz_magic[0],
+             gz_magic[1], Z_DEFLATED, 0 /*flags*/, 0,0,0,0 /*time*/,
+             0 /*xflags*/, OS_CODE);
+
+    if ((error = vn_rdwr(UIO_WRITE, s->file, buf, GZ_HEADER_LEN, s->outoff,
+                         UIO_SYSSPACE, IO_NODELOCKED|IO_UNIT, curproc->p_ucred,
+                         NOCRED, &resid, curthread))) {
+        s->outoff += GZ_HEADER_LEN - resid;
+        return destroy(s), (gzFile)Z_NULL;
+    }
+    s->outoff += GZ_HEADER_LEN;
+    s->startpos = 10L;
+    
+    return (gzFile)s;
+}
+
+
+ /* ===========================================================================
+ * Cleanup then free the given gz_stream. Return a zlib error code.
+   Try freeing in the reverse order of allocations.
+ */
+local int destroy (s)
+    gz_stream *s;
+{
+    int err = Z_OK;
+
+    if (!s) return Z_STREAM_ERROR;
+
+    TRYFREE(s->msg);
+
+    if (s->stream.state != NULL) {
+	if (s->mode == 'w') {
+	    err = deflateEnd(&(s->stream));
+	}
+    }
+    if (s->z_err < 0) err = s->z_err;
+
+    TRYFREE(s->inbuf);
+    TRYFREE(s->outbuf);
+    TRYFREE(s->path);
+    TRYFREE(s);
+    return err;
+}
+
+
+/* ===========================================================================
+     Writes the given number of uncompressed bytes into the compressed file.
+   gzwrite returns the number of bytes actually written (0 in case of error).
+*/
+int ZEXPORT gzwrite (file, buf, len)
+    gzFile file;
+    const voidp buf;
+    unsigned len;
+{
+    gz_stream *s = (gz_stream*)file;
+    off_t curoff;
+    size_t resid;
+    int error;
+    int vfslocked;
+
+    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
+
+    s->stream.next_in = (Bytef*)buf;
+    s->stream.avail_in = len;
+
+    curoff = s->outoff;
+    while (s->stream.avail_in != 0) {
+
+        if (s->stream.avail_out == 0) {
+
+            s->stream.next_out = s->outbuf;
+            vfslocked = VFS_LOCK_GIANT(s->file->v_mount);
+            error = vn_rdwr_inchunks(UIO_WRITE, s->file, s->outbuf, Z_BUFSIZE,
+                        curoff, UIO_SYSSPACE, IO_NODELOCKED|IO_UNIT,
+                        curproc->p_ucred, NOCRED, &resid, curthread);
+            VFS_UNLOCK_GIANT(vfslocked);
+            if (error) {
+                log(LOG_ERR, "gzwrite: vn_rdwr return %d\n", error);
+                curoff += Z_BUFSIZE - resid;
+                s->z_err = Z_ERRNO;
+                break;
+            }
+            curoff += Z_BUFSIZE;
+            s->stream.avail_out = Z_BUFSIZE;
+        }
+        s->z_err = deflate(&(s->stream), Z_NO_FLUSH);
+        if (s->z_err != Z_OK) {
+            log(LOG_ERR,
+                "gzwrite: deflate returned error %d\n", s->z_err);
+            break;
+        }
+    }
+
+    s->crc = ~crc32_raw(buf, len, ~s->crc);
+    s->outoff = curoff;
+
+    return (int)(len - s->stream.avail_in);
+}
+
+
+/* ===========================================================================
+     Flushes all pending output into the compressed file. The parameter
+   flush is as in the deflate() function.
+*/
+local int do_flush (file, flush)
+    gzFile file;
+    int flush;
+{
+    uInt len;
+    int done = 0;
+    gz_stream *s = (gz_stream*)file;
+    off_t curoff = s->outoff;
+    size_t resid;
+    int vfslocked = 0;
+    int error;
+
+    if (s == NULL || s->mode != 'w') return Z_STREAM_ERROR;
+
+    if (s->stream.avail_in) {
+        log(LOG_WARNING, "do_flush: avail_in non-zero on entry\n");
+    } 
+
+    s->stream.avail_in = 0; /* should be zero already anyway */
+
+    for (;;) {
+        len = Z_BUFSIZE - s->stream.avail_out;
+
+        if (len != 0) {
+            vfslocked = VFS_LOCK_GIANT(s->file->v_mount);
+            error = vn_rdwr_inchunks(UIO_WRITE, s->file, s->outbuf, len, curoff,
+                        UIO_SYSSPACE, IO_NODELOCKED|IO_UNIT, curproc->p_ucred,
+                        NOCRED, &resid, curthread);
+            VFS_UNLOCK_GIANT(vfslocked);
+	    if (error) {
+                s->z_err = Z_ERRNO;
+                s->outoff = curoff + len - resid;
+                return Z_ERRNO;
+            }
+            s->stream.next_out = s->outbuf;
+            s->stream.avail_out = Z_BUFSIZE;
+            curoff += len;
+        }
+        if (done) break;
+        s->z_err = deflate(&(s->stream), flush);
+
+	/* Ignore the second of two consecutive flushes: */
+	if (len == 0 && s->z_err == Z_BUF_ERROR) s->z_err = Z_OK;
+
+        /* deflate has finished flushing only when it hasn't used up
+         * all the available space in the output buffer: 
+         */
+        done = (s->stream.avail_out != 0 || s->z_err == Z_STREAM_END);
+ 
+        if (s->z_err != Z_OK && s->z_err != Z_STREAM_END) break;
+    }
+    s->outoff = curoff;
+
+    return  s->z_err == Z_STREAM_END ? Z_OK : s->z_err;
+}
+
+int ZEXPORT gzflush (file, flush)
+     gzFile file;
+     int flush;
+{
+    gz_stream *s = (gz_stream*)file;
+    int err = do_flush (file, flush);
+
+    if (err) return err;
+    return  s->z_err == Z_STREAM_END ? Z_OK : s->z_err;
+}
+
+
+/* ===========================================================================
+   Outputs a long in LSB order to the given file
+*/
+local void putU32 (s, x)
+    gz_stream *s;
+    uint32_t x;
+{
+    uint32_t xx;
+    off_t curoff = s->outoff;
+    int resid;
+
+#if BYTE_ORDER == BIG_ENDIAN
+    xx = bswap32(x);
+#else
+    xx = x;
+#endif
+    vn_rdwr(UIO_WRITE, s->file, (caddr_t)&xx, sizeof(xx), curoff,
+            UIO_SYSSPACE, IO_NODELOCKED|IO_UNIT, curproc->p_ucred,
+            NOCRED, &resid, curthread);
+    s->outoff += sizeof(xx) - resid;
+}
+
+
+/* ===========================================================================
+     Flushes all pending output if necessary, closes the compressed file
+   and deallocates all the (de)compression state.
+*/
+int ZEXPORT gzclose (file)
+    gzFile file;
+{
+    int err;
+    gz_stream *s = (gz_stream*)file;
+
+    if (s == NULL) return Z_STREAM_ERROR;
+
+    if (s->mode == 'w') {
+        err = do_flush (file, Z_FINISH);
+        if (err != Z_OK) {
+            log(LOG_ERR, "gzclose: do_flush failed (err %d)\n", err);
+            return destroy((gz_stream*)file);
+        }
+#if 0
+	printf("gzclose: putting crc: %lld total: %lld\n",
+	    (long long)s->crc, (long long)s->stream.total_in);
+	printf("sizeof uLong = %d\n", (int)sizeof(uLong));
+#endif
+        putU32 (s, s->crc);
+        putU32 (s, (uint32_t) s->stream.total_in);
+    }
+    return destroy((gz_stream*)file);
+}
+
+/*
+ * Space allocation and freeing routines for use by zlib routines when called
+ * from gzip modules.
+ */
+static void *
+gz_alloc(void *notused __unused, u_int items, u_int size)
+{
+    void *ptr;
+
+    MALLOC(ptr, void *, items * size, M_TEMP, M_NOWAIT | M_ZERO);
+    return ptr;
+}
+                                     
+static void
+gz_free(void *opaque __unused, void *ptr)
+{
+    FREE(ptr, M_TEMP);
+}
+

sys/kern/kern_sig.c

@@ -40,6 +40,7 @@ __FBSDID("$FreeBSD$");
 #include "opt_compat.h"
 #include "opt_kdtrace.h"
 #include "opt_ktrace.h"
+#include "opt_core.h"
 
 #include <sys/param.h>
 #include <sys/systm.h>
@@ -49,6 +50,7 @@ __FBSDID("$FreeBSD$");
 #include <sys/condvar.h>
 #include <sys/event.h>
 #include <sys/fcntl.h>
+#include <sys/imgact.h>
 #include <sys/kernel.h>
 #include <sys/ktr.h>
 #include <sys/ktrace.h>
@@ -78,6 +80,8 @@ __FBSDID("$FreeBSD$");
 #include <vm/vm_extern.h>
 #include <vm/uma.h>
 
+#include <sys/jail.h>
+
 #include <machine/cpu.h>
 
 #include <security/audit/audit.h>
@@ -98,7 +102,7 @@ SDT_PROBE_ARGTYPE(proc, kernel, , signal_discard, 1, "struct proc *");
 SDT_PROBE_ARGTYPE(proc, kernel, , signal_discard, 2, "int");
 
 static int	coredump(struct thread *);
-static char	*expand_name(const char *, uid_t, pid_t);
+static char	*expand_name(const char *, uid_t, pid_t, struct thread *, int);
 static int	killpg1(struct thread *td, int sig, int pgid, int all,
 		    ksiginfo_t *ksi);
 static int	issignal(struct thread *td, int stop_allowed);
@@ -2936,12 +2940,51 @@ childproc_exited(struct proc *p)
 	sigparent(p, reason, status);
 }
 
+/*
+ * We only have 1 character for the core count in the format
+ * string, so the range will be 0-9
+ */
+#define MAX_NUM_CORES 10
+static int num_cores = 5;
+
+static int
+sysctl_debug_num_cores_check (SYSCTL_HANDLER_ARGS)
+{
+	int error;
+	int new_val;
+	
+	error = sysctl_handle_int(oidp, &new_val, 0, req);
+	if (error != 0 || req->newptr == NULL)
+		return (error);
+	if (new_val > MAX_NUM_CORES)
+		new_val = MAX_NUM_CORES;
+	if (new_val < 0)
+		new_val = 0;
+	num_cores = new_val;
+	return (0);
+}
+SYSCTL_PROC(_debug, OID_AUTO, ncores, CTLTYPE_INT|CTLFLAG_RW, 
+	    0, sizeof(int), sysctl_debug_num_cores_check, "I", "");
+
+#if defined(COMPRESS_USER_CORES)
+int compress_user_cores = 1;
+SYSCTL_INT(_kern, OID_AUTO, compress_user_cores, CTLFLAG_RW,
+        &compress_user_cores, 0, "");
+
+int compress_user_cores_gzlevel = -1; /* default level */
+SYSCTL_INT(_kern, OID_AUTO, compress_user_cores_gzlevel, CTLFLAG_RW,
+    &compress_user_cores_gzlevel, -1, "user core gz compression level");
+
+#define GZ_SUFFIX	".gz"	
+#define GZ_SUFFIX_LEN	3	
+#endif
+
 static char corefilename[MAXPATHLEN] = {"%N.core"};
 SYSCTL_STRING(_kern, OID_AUTO, corefile, CTLFLAG_RW, corefilename,
 	      sizeof(corefilename), "process corefile name format string");
 
 /*
- * expand_name(name, uid, pid)
+ * expand_name(name, uid, pid, td, compress)
  * Expand the name described in corefilename, using name, uid, and pid.
  * corefilename is a printf-like string, with three format specifiers:
  *	%N	name of process ("name")
@@ -2952,20 +2995,21 @@ SYSCTL_STRING(_kern, OID_AUTO, corefile, CTLFLAG_RW, corefilename,
  * This is controlled by the sysctl variable kern.corefile (see above).
  */
 static char *
-expand_name(name, uid, pid)
+expand_name(const char *name, uid_t uid, pid_t pid, struct thread *td,
-	const char *name;
+    int compress)
-	uid_t uid;
-	pid_t pid;
 {
 	struct sbuf sb;
 	const char *format;
 	char *temp;
 	size_t i;
+	int indexpos;
+	char hostname[MAXHOSTNAMELEN];
 	
 	format = corefilename;
 	temp = malloc(MAXPATHLEN, M_TEMP, M_NOWAIT | M_ZERO);
 	if (temp == NULL)
 		return (NULL);
+	indexpos = -1;
 	(void)sbuf_new(&sb, temp, MAXPATHLEN, SBUF_FIXEDLEN);
 	for (i = 0; format[i]; i++) {
 		switch (format[i]) {
@@ -2975,6 +3019,15 @@ expand_name(name, uid, pid)
 			case '%':
 				sbuf_putc(&sb, '%');
 				break;
+			case 'H':	/* hostname */
+				getcredhostname(td->td_ucred, hostname,
+				    sizeof(hostname));
+				sbuf_printf(&sb, "%s", hostname);
+				break;
+			case 'I':       /* autoincrementing index */
+				sbuf_printf(&sb, "0");
+				indexpos = sbuf_len(&sb) - 1;
+				break;
 			case 'N':	/* process name */
 				sbuf_printf(&sb, "%s", name);
 				break;
@@ -2994,6 +3047,11 @@ expand_name(name, uid, pid)
 			sbuf_putc(&sb, format[i]);
 		}
 	}
+#ifdef COMPRESS_USER_CORES
+	if (compress) {
+		sbuf_printf(&sb, GZ_SUFFIX);
+	}
+#endif
 	if (sbuf_overflowed(&sb)) {
 		sbuf_delete(&sb);
 		log(LOG_ERR, "pid %ld (%s), uid (%lu): corename is too "
@@ -3003,6 +3061,53 @@ expand_name(name, uid, pid)
 	}
 	sbuf_finish(&sb);
 	sbuf_delete(&sb);
+
+	/*
+	 * If the core format has a %I in it, then we need to check
+	 * for existing corefiles before returning a name.
+	 * To do this we iterate over 0..num_cores to find a
+	 * non-existing core file name to use.
+	 */
+	if (indexpos != -1) {
+		struct nameidata nd;
+		int error, n;
+		int flags = O_CREAT | O_EXCL | FWRITE | O_NOFOLLOW;
+		int cmode = S_IRUSR | S_IWUSR | S_IRGRP | S_IWGRP;
+		int vfslocked;
+
+		for (n = 0; n < num_cores; n++) {
+			temp[indexpos] = '0' + n;
+			NDINIT(&nd, LOOKUP, NOFOLLOW | MPSAFE, UIO_SYSSPACE,
+			    temp, td); 
+			error = vn_open(&nd, &flags, cmode, NULL);
+			if (error) {
+				if (error == EEXIST) {
+					continue;
+				}
+				log(LOG_ERR,
+				    "pid %d (%s), uid (%u):  Path `%s' failed "
+                                    "on initial open test, error = %d\n",
+				    pid, name, uid, temp, error);
+				free(temp, M_TEMP);
+				return (NULL);
+			}
+			vfslocked = NDHASGIANT(&nd);
+			NDFREE(&nd, NDF_ONLY_PNBUF);
+			VOP_UNLOCK(nd.ni_vp, 0);
+			error = vn_close(nd.ni_vp, FWRITE, td->td_ucred, td);
+			VFS_UNLOCK_GIANT(vfslocked);
+			if (error) {
+				log(LOG_ERR,
+				    "pid %d (%s), uid (%u):  Path `%s' failed "
+                                    "on close after initial open test, "
+                                    "error = %d\n",
+				    pid, name, uid, temp, error);
+				free(temp, M_TEMP);
+				return (NULL);
+			}
+			break;
+		}
+	}
 	return (temp);
 }
 
@@ -3028,12 +3133,19 @@ coredump(struct thread *td)
 	char *name;			/* name of corefile */
 	off_t limit;
 	int vfslocked;
+	int compress;
 
+#ifdef COMPRESS_USER_CORES
+	compress = compress_user_cores;
+#else
+	compress = 0;
+#endif
 	PROC_LOCK_ASSERT(p, MA_OWNED);
 	MPASS((p->p_flag & P_HADTHREADS) == 0 || p->p_singlethread == td);
 	_STOPEVENT(p, S_CORE, 0);
 
-	name = expand_name(p->p_comm, td->td_ucred->cr_uid, p->p_pid);
+	name = expand_name(p->p_comm, td->td_ucred->cr_uid, p->p_pid, td,
+	    compress);
 	if (name == NULL) {
 		PROC_UNLOCK(p);
 #ifdef AUDIT
@@ -3124,7 +3236,7 @@ coredump(struct thread *td)
 	PROC_UNLOCK(p);
 
 	error = p->p_sysent->sv_coredump ?
-	  p->p_sysent->sv_coredump(td, vp, limit) :
+	  p->p_sysent->sv_coredump(td, vp, limit, compress ? IMGACT_CORE_COMPRESS : 0) :
 	  ENOSYS;
 
 	if (locked) {

sys/net/zlib.h

@@ -1010,6 +1010,13 @@ extern int EXPORT inflateInit2_ OF((z_streamp strm, int  windowBits,
 
 uLongf *get_crc_table OF((void)); /* can be used by asm versions of crc32() */
 
+#ifdef _KERNEL
+struct vnode;
+extern gzFile gz_open     OF((const char *path, const char *mode,
+	                              struct vnode *vp));
+#endif
+
+
 #ifdef __cplusplus
 }
 #endif

sys/net/zutil.h

@@ -0,0 +1,231 @@
+/* zutil.h -- internal interface and configuration of the compression library
+ * Copyright (C) 1995-1996 Jean-loup Gailly.
+ * For conditions of distribution and use, see copyright notice in zlib.h
+ */
+
+/* WARNING: this file should *not* be used by applications. It is
+   part of the implementation of the compression library and is
+   subject to change. Applications should only use zlib.h.
+ */
+
+/* From: zutil.h,v 1.16 1996/07/24 13:41:13 me Exp $ */
+/* $FreeBSD$ */
+
+#ifndef _Z_UTIL_H
+#define _Z_UTIL_H
+
+#define ZEXPORT
+
+#ifdef _KERNEL
+#include <net/zlib.h>
+#else
+#include "zlib.h"
+#endif
+
+#ifdef _KERNEL
+/* Assume this is a *BSD or SVR4 kernel */
+#include <sys/types.h>
+#include <sys/time.h>
+#include <sys/systm.h>
+#include <sys/module.h>
+#include <sys/errno.h>
+#include <sys/param.h>
+#include <sys/kernel.h>
+#  define HAVE_MEMCPY
+#  define memcpy(d, s, n)	bcopy((s), (d), (n))
+#  define memset(d, v, n)	bzero((d), (n))
+#  define memcmp		bcmp
+
+#else
+#if defined(__KERNEL__)
+/* Assume this is a Linux kernel */
+#include <linux/string.h>
+#define HAVE_MEMCPY
+
+#else /* not kernel */
+
+#if defined(MSDOS)||defined(VMS)||defined(CRAY)||defined(WIN32)||defined(RISCOS)
+#   include <stddef.h>
+#   include <errno.h>
+#else
+    extern int errno;
+#endif
+#ifdef STDC
+#  include <string.h>
+#  include <stdlib.h>
+#endif
+#endif /* __KERNEL__ */
+#endif /* _KERNEL */
+
+#ifndef local
+#  define local static
+#endif
+/* compile with -Dlocal if your debugger can't find static symbols */
+
+typedef unsigned char  uch;
+typedef uch FAR uchf;
+typedef unsigned short ush;
+typedef ush FAR ushf;
+typedef unsigned long  ulg;
+
+#define ERR_MSG(err) z_errmsg[Z_NEED_DICT-(err)]
+
+#define ERR_RETURN(strm,err) \
+  return (strm->msg = (const char*)ERR_MSG(err), (err))
+/* To be used only when the state is known to be valid */
+
+        /* common constants */
+
+#ifndef DEF_WBITS
+#  define DEF_WBITS MAX_WBITS
+#endif
+/* default windowBits for decompression. MAX_WBITS is for compression only */
+
+#if MAX_MEM_LEVEL >= 8
+#  define DEF_MEM_LEVEL 8
+#else
+#  define DEF_MEM_LEVEL  MAX_MEM_LEVEL
+#endif
+/* default memLevel */
+
+#define STORED_BLOCK 0
+#define STATIC_TREES 1
+#define DYN_TREES    2
+/* The three kinds of block type */
+
+#define MIN_MATCH  3
+#define MAX_MATCH  258
+/* The minimum and maximum match lengths */
+
+#define PRESET_DICT 0x20 /* preset dictionary flag in zlib header */
+
+        /* target dependencies */
+
+#ifdef MSDOS
+#  define OS_CODE  0x00
+#  ifdef __TURBOC__
+#    include <alloc.h>
+#  else /* MSC or DJGPP */
+#    include <malloc.h>
+#  endif
+#endif
+
+#ifdef OS2
+#  define OS_CODE  0x06
+#endif
+
+#ifdef WIN32 /* Window 95 & Windows NT */
+#  define OS_CODE  0x0b
+#endif
+
+#if defined(VAXC) || defined(VMS)
+#  define OS_CODE  0x02
+#  define FOPEN(name, mode) \
+     fopen((name), (mode), "mbc=60", "ctx=stm", "rfm=fix", "mrs=512")
+#endif
+
+#ifdef AMIGA
+#  define OS_CODE  0x01
+#endif
+
+#if defined(ATARI) || defined(atarist)
+#  define OS_CODE  0x05
+#endif
+
+#ifdef MACOS
+#  define OS_CODE  0x07
+#endif
+
+#ifdef __50SERIES /* Prime/PRIMOS */
+#  define OS_CODE  0x0F
+#endif
+
+#ifdef TOPS20
+#  define OS_CODE  0x0a
+#endif
+
+#if defined(_BEOS_) || defined(RISCOS)
+#  define fdopen(fd,mode) NULL /* No fdopen() */
+#endif
+
+        /* Common defaults */
+
+#ifndef OS_CODE
+#  define OS_CODE  0x03  /* assume Unix */
+#endif
+
+#ifndef FOPEN
+#  define FOPEN(name, mode) fopen((name), (mode))
+#endif
+
+         /* functions */
+
+#ifdef HAVE_STRERROR
+   extern char *strerror OF((int));
+#  define zstrerror(errnum) strerror(errnum)
+#else
+#  define zstrerror(errnum) ""
+#endif
+
+#if defined(pyr)
+#  define NO_MEMCPY
+#endif
+#if (defined(M_I86SM) || defined(M_I86MM)) && !defined(_MSC_VER)
+ /* Use our own functions for small and medium model with MSC <= 5.0.
+  * You may have to use the same strategy for Borland C (untested).
+  */
+#  define NO_MEMCPY
+#endif
+#if defined(STDC) && !defined(HAVE_MEMCPY) && !defined(NO_MEMCPY)
+#  define HAVE_MEMCPY
+#endif
+#ifdef HAVE_MEMCPY
+#  ifdef SMALL_MEDIUM /* MSDOS small or medium model */
+#    define zmemcpy _fmemcpy
+#    define zmemcmp _fmemcmp
+#    define zmemzero(dest, len) _fmemset(dest, 0, len)
+#  else
+#    define zmemcpy memcpy
+#    define zmemcmp memcmp
+#    define zmemzero(dest, len) memset(dest, 0, len)
+#  endif
+#else
+   extern void zmemcpy  OF((Bytef* dest, Bytef* source, uInt len));
+   extern int  zmemcmp  OF((Bytef* s1,   Bytef* s2, uInt len));
+   extern void zmemzero OF((Bytef* dest, uInt len));
+#endif
+
+/* Diagnostic functions */
+#ifdef DEBUG_ZLIB
+#  include <stdio.h>
+#  ifndef verbose
+#    define verbose 0
+#  endif
+   extern void z_error    OF((char *m));
+#  define Assert(cond,msg) {if(!(cond)) z_error(msg);}
+#  define Trace(x) fprintf x
+#  define Tracev(x) {if (verbose) fprintf x ;}
+#  define Tracevv(x) {if (verbose>1) fprintf x ;}
+#  define Tracec(c,x) {if (verbose && (c)) fprintf x ;}
+#  define Tracecv(c,x) {if (verbose>1 && (c)) fprintf x ;}
+#else
+#  define Assert(cond,msg)
+#  define Trace(x)
+#  define Tracev(x)
+#  define Tracevv(x)
+#  define Tracec(c,x)
+#  define Tracecv(c,x)
+#endif
+
+
+typedef uLong (*check_func) OF((uLong check, const Bytef *buf, uInt len));
+
+voidpf zcalloc OF((voidpf opaque, unsigned items, unsigned size));
+void   zcfree  OF((voidpf opaque, voidpf ptr));
+
+#define ZALLOC(strm, items, size) \
+           (*((strm)->zalloc))((strm)->opaque, (items), (size))
+#define ZFREE(strm, addr)  (*((strm)->zfree))((strm)->opaque, (voidpf)(addr))
+#define TRY_FREE(s, p) {if (p) ZFREE(s, p);}
+
+#endif /* _Z_UTIL_H */

sys/sys/eventhandler.h

@@ -211,7 +211,20 @@ EVENTHANDLER_DECLARE(process_exit, exitlist_fn);
 EVENTHANDLER_DECLARE(process_fork, forklist_fn);
 EVENTHANDLER_DECLARE(process_exec, execlist_fn);
 
+/*
+ * application dump event
+ */
 struct thread;
+typedef void (*app_coredump_start_fn)(void *, struct thread *, char *name);
+typedef void (*app_coredump_progress_fn)(void *, struct thread *td, int byte_count);
+typedef void (*app_coredump_finish_fn)(void *, struct thread *td);
+typedef void (*app_coredump_error_fn)(void *, struct thread *td, char *msg, ...);
+
+EVENTHANDLER_DECLARE(app_coredump_start, app_coredump_start_fn);
+EVENTHANDLER_DECLARE(app_coredump_progress, app_coredump_progress_fn);
+EVENTHANDLER_DECLARE(app_coredump_finish, app_coredump_finish_fn);
+EVENTHANDLER_DECLARE(app_coredump_error, app_coredump_error_fn);
+
 typedef void (*thread_ctor_fn)(void *, struct thread *);
 typedef void (*thread_dtor_fn)(void *, struct thread *);
 typedef void (*thread_fini_fn)(void *, struct thread *);

sys/sys/imgact.h

@@ -75,6 +75,8 @@ struct image_params {
 struct sysentvec;
 struct thread;
 
+#define IMGACT_CORE_COMPRESS	0x01
+
 int	exec_check_permissions(struct image_params *);
 register_t *exec_copyout_strings(struct image_params *);
 int	exec_new_vmspace(struct image_params *, struct sysentvec *);

sys/sys/imgact_aout.h

@@ -150,7 +150,8 @@ struct exec {
 struct thread;
 struct vnode;
 
-int	aout_coredump(struct thread *td, struct vnode *vp, off_t limit);
+int	aout_coredump(struct thread *td, struct vnode *vp, off_t limit,
+    int flags);
 #endif
 
 #endif /* !_IMGACT_AOUT_H_ */

sys/sys/imgact_elf.h

@@ -88,7 +88,7 @@ int	__elfN(brand_inuse)(Elf_Brandinfo *entry);
 int	__elfN(insert_brand_entry)(Elf_Brandinfo *entry);
 int	__elfN(remove_brand_entry)(Elf_Brandinfo *entry);
 int	__elfN(freebsd_fixup)(register_t **, struct image_params *);
-int	__elfN(coredump)(struct thread *, struct vnode *, off_t);
+int	__elfN(coredump)(struct thread *, struct vnode *, off_t, int);
 
 /* Machine specific function to dump per-thread information. */
 void	__elfN(dump_thread)(struct thread *, void *, size_t *);

sys/sys/sysent.h

@@ -87,7 +87,7 @@ struct sysentvec {
 	void		(*sv_prepsyscall)(struct trapframe *, int *, u_int *,
 			    caddr_t *);
 	char		*sv_name;	/* name of binary type */
-	int		(*sv_coredump)(struct thread *, struct vnode *, off_t);
+	int		(*sv_coredump)(struct thread *, struct vnode *, off_t, int);
 					/* function to dump core, or NULL */
 	int		(*sv_imgact_try)(struct image_params *);
 	int		sv_minsigstksz;	/* minimum signal stack size */