]> git.saurik.com Git - apple/xnu.git/blobdiff - bsd/kern/subr_log.c
xnu-4903.270.47.tar.gz
[apple/xnu.git] / bsd / kern / subr_log.c
index d39eccd5d040b3d4710ea12df77a9764d24f9e17..46475b683dd1e7aa179b0bab23f063867880e67c 100644 (file)
@@ -1,8 +1,8 @@
 /*
- * Copyright (c) 2000-2006 Apple Computer, Inc. All rights reserved.
+ * Copyright (c) 2000-2016 Apple, Inc. All rights reserved.
  *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_START@
- * 
+ *
  * This file contains Original Code and/or Modifications of Original Code
  * as defined in and that are subject to the Apple Public Source License
  * Version 2.0 (the 'License'). You may not use this file except in
  * unlawful or unlicensed copies of an Apple operating system, or to
  * circumvent, violate, or enable the circumvention or violation of, any
  * terms of an Apple operating system software license agreement.
- * 
+ *
  * Please obtain a copy of the License at
  * http://www.opensource.apple.com/apsl/ and read it before using this file.
- * 
+ *
  * The Original Code and all software distributed under the License are
  * distributed on an 'AS IS' basis, WITHOUT WARRANTY OF ANY KIND, EITHER
  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
@@ -22,7 +22,7 @@
  * FITNESS FOR A PARTICULAR PURPOSE, QUIET ENJOYMENT OR NON-INFRINGEMENT.
  * Please see the License for the specific language governing rights and
  * limitations under the License.
- * 
+ *
  * @APPLE_OSREFERENCE_LICENSE_HEADER_END@
  */
 /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
 #include <sys/systm.h>
 #include <sys/proc_internal.h>
 #include <sys/vnode.h>
+#include <stdbool.h>
+#include <firehose/tracepoint_private.h>
+#include <firehose/chunk_private.h>
+#include <firehose/ioctl_private.h>
+#include <os/firehose_buffer_private.h>
+
+#include <os/log_private.h>
 #include <sys/ioctl.h>
 #include <sys/msgbuf.h>
 #include <sys/file_internal.h>
 #include <sys/select.h>
 #include <sys/kernel.h>
 #include <kern/thread.h>
+#include <kern/sched_prim.h>
+#include <kern/simple_lock.h>
 #include <sys/lock.h>
 #include <sys/signalvar.h>
 #include <sys/conf.h>
 #include <sys/sysctl.h>
+#include <sys/queue.h>
 #include <kern/kalloc.h>
+#include <pexpert/pexpert.h>
+#include <mach/mach_port.h>
+#include <mach/mach_vm.h>
+#include <mach/vm_map.h>
+#include <vm/vm_kern.h>
+#include <kern/task.h>
+#include <kern/locks.h>
 
 /* XXX should be in a common header somewhere */
-extern void klogwakeup(void);
-extern void logwakeup(void);
+extern void logwakeup(struct msgbuf *);
+extern void oslogwakeup(void);
+extern void oslog_streamwakeup(void);
+static void oslog_streamwakeup_locked(void);
+vm_offset_t kernel_firehose_addr = 0;
+
+/* log message counters for streaming mode */
+uint32_t oslog_s_streamed_msgcount = 0;
+uint32_t oslog_s_dropped_msgcount  = 0;
+extern uint32_t oslog_s_error_count;
 
-#define LOG_RDPRI      (PZERO + 1)
+#define LOG_RDPRI       (PZERO + 1)
 
-#define LOG_NBIO       0x02
-#define LOG_ASYNC      0x04
-#define LOG_RDWAIT     0x08
+#define LOG_NBIO        0x02
+#define LOG_ASYNC       0x04
+#define LOG_RDWAIT      0x08
 
+/* All globals should be accessed under LOG_LOCK() */
+
+static char amsg_bufc[1024];
+static struct msgbuf aslbuf = {MSG_MAGIC, sizeof(amsg_bufc), 0, 0, amsg_bufc};
+struct msgbuf *aslbufp __attribute__((used)) = &aslbuf;
+
+/* logsoftc only valid while log_open=1 */
 struct logsoftc {
-       int     sc_state;               /* see above for possibilities */
-       struct  selinfo sc_selp;        /* thread waiting for select */
-       int     sc_pgid;                /* process/group for async I/O */
+       int     sc_state;               /* see above for possibilities */
+       struct  selinfo sc_selp;        /* thread waiting for select */
+       int     sc_pgid;                /* process/group for async I/O */
+       struct msgbuf *sc_mbp;
 } logsoftc;
 
-int    log_open;                       /* also used in log() */
-char smsg_bufc[MSG_BSIZE]; /* static buffer */
-struct msgbuf temp_msgbuf = {0,MSG_BSIZE,0,0,smsg_bufc};
-struct msgbuf *msgbufp;
-static int _logentrypend = 0;
-static int log_inited = 0;
-/* the following are implemented in osfmk/kern/printf.c  */
+static int log_open;
+char smsg_bufc[CONFIG_MSG_BSIZE]; /* static buffer */
+char oslog_stream_bufc[FIREHOSE_CHUNK_SIZE]; /* static buffer */
+struct firehose_chunk_s oslog_boot_buf = {
+       .fc_pos = {
+               .fcp_next_entry_offs = offsetof(struct firehose_chunk_s, fc_data),
+               .fcp_private_offs = FIREHOSE_CHUNK_SIZE,
+               .fcp_refcnt = 1, // indicate that there is a writer to this chunk
+               .fcp_stream = firehose_stream_persist,
+               .fcp_flag_io = 1, // for now, lets assume this is coming from the io bank
+       },
+}; /* static buffer */
+firehose_chunk_t firehose_boot_chunk = &oslog_boot_buf;
+struct msgbuf msgbuf = {MSG_MAGIC, sizeof(smsg_bufc), 0, 0, smsg_bufc};
+struct msgbuf oslog_stream_buf = {MSG_MAGIC, 0, 0, 0, NULL};
+struct msgbuf *msgbufp __attribute__((used)) = &msgbuf;
+struct msgbuf *oslog_streambufp __attribute__((used)) = &oslog_stream_buf;
+
+// List entries for keeping track of the streaming buffer
+static oslog_stream_buf_entry_t oslog_stream_buf_entries;
+
+#define OSLOG_NUM_STREAM_ENTRIES        64
+#define OSLOG_STREAM_BUF_SIZE           4096
+
+int oslog_open = 0;
+int os_log_wakeup = 0;
+int oslog_stream_open = 0;
+int oslog_stream_buf_bytesavail = 0;
+int oslog_stream_buf_size = OSLOG_STREAM_BUF_SIZE;
+int oslog_stream_num_entries = OSLOG_NUM_STREAM_ENTRIES;
+
+uint8_t __firehose_buffer_kernel_chunk_count = FIREHOSE_BUFFER_KERNEL_DEFAULT_CHUNK_COUNT;
+uint8_t __firehose_num_kernel_io_pages = FIREHOSE_BUFFER_KERNEL_DEFAULT_IO_PAGES;
+
+/* oslogsoftc only valid while oslog_open=1 */
+struct oslogsoftc {
+       int     sc_state;               /* see above for possibilities */
+       struct  selinfo sc_selp;        /* thread waiting for select */
+       int     sc_pgid;                /* process/group for async I/O */
+} oslogsoftc;
+
+struct oslog_streamsoftc {
+       int     sc_state;               /* see above for possibilities */
+       struct  selinfo sc_selp;        /* thread waiting for select */
+       int     sc_pgid;                /* process/group for async I/O */
+} oslog_streamsoftc;
+
+STAILQ_HEAD(, oslog_stream_buf_entry_s) oslog_stream_free_head =
+    STAILQ_HEAD_INITIALIZER(oslog_stream_free_head);
+STAILQ_HEAD(, oslog_stream_buf_entry_s) oslog_stream_buf_head =
+    STAILQ_HEAD_INITIALIZER(oslog_stream_buf_head);
+
+/* defined in osfmk/kern/printf.c  */
+extern void oslog_lock_init(void);
 extern void bsd_log_lock(void);
 extern void bsd_log_unlock(void);
-extern void bsd_log_init(void);
+
+/* defined for osfmk/kern/printf.c */
+void bsd_log_init(void);
+
+/*
+ * Ideally this file would define this lock, but bsd doesn't have the definition
+ * for lock groups.
+ */
+decl_lck_spin_data(extern, oslog_stream_lock)
+#define stream_lock() lck_spin_lock(&oslog_stream_lock)
+#define stream_unlock() lck_spin_unlock(&oslog_stream_lock)
 
 /* XXX wants a linker set so these can be static */
 extern d_open_t         logopen;
@@ -116,15 +206,44 @@ extern d_read_t         logread;
 extern d_ioctl_t        logioctl;
 extern d_select_t       logselect;
 
+/* XXX wants a linker set so these can be static */
+extern d_open_t         oslogopen;
+extern d_close_t        oslogclose;
+extern d_select_t       oslogselect;
+extern d_ioctl_t        oslogioctl;
+
+/* XXX wants a linker set so these can be static */
+extern d_open_t         oslog_streamopen;
+extern d_close_t        oslog_streamclose;
+extern d_read_t         oslog_streamread;
+extern d_ioctl_t        oslog_streamioctl;
+extern d_select_t       oslog_streamselect;
+
+void oslog_init(void);
+void oslog_setsize(int size);
+void oslog_streamwrite_locked(firehose_tracepoint_id_u ftid,
+    uint64_t stamp, const void *pubdata, size_t publen);
+void oslog_streamwrite_metadata_locked(oslog_stream_buf_entry_t m_entry);
+static oslog_stream_buf_entry_t oslog_stream_find_free_buf_entry_locked(void);
+static void oslog_streamwrite_append_bytes(const char *buffer, int buflen);
+
 /*
  * Serialize log access.  Note that the log can be written at interrupt level,
  * so any log manipulations that can be done from, or affect, another processor
  * at interrupt level must be guarded with a spin lock.
  */
 
-#define        LOG_LOCK() bsd_log_lock()
-#define        LOG_UNLOCK() bsd_log_unlock()
+#define LOG_LOCK() bsd_log_lock()
+#define LOG_UNLOCK() bsd_log_unlock()
+
+#if DEBUG
+#define LOG_SETSIZE_DEBUG(x...) kprintf(x)
+#else
+#define LOG_SETSIZE_DEBUG(x...) do { } while(0)
+#endif
 
+static int sysctl_kern_msgbuf(struct sysctl_oid *oidp,
+    void *arg1, int arg2, struct sysctl_req *req);
 
 /*ARGSUSED*/
 int
@@ -133,26 +252,24 @@ logopen(__unused dev_t dev, __unused int flags, __unused int mode, struct proc *
        LOG_LOCK();
        if (log_open) {
                LOG_UNLOCK();
-               return (EBUSY);
+               return EBUSY;
+       }
+       if (atm_get_diagnostic_config() & ATM_ENABLE_LEGACY_LOGGING) {
+               logsoftc.sc_mbp = msgbufp;
+       } else {
+               /*
+                * Support for messagetracer (kern_asl_msg())
+                * In this mode, /dev/klog exports only ASL-formatted messages
+                * written into aslbufp via vaddlog().
+                */
+               logsoftc.sc_mbp = aslbufp;
        }
+       logsoftc.sc_pgid = p->p_pid;            /* signal process only */
        log_open = 1;
-       logsoftc.sc_pgid = p->p_pid;            /* signal process only */
-       /*
-        * Potential race here with putchar() but since putchar should be
-        * called by autoconf, msg_magic should be initialized by the time
-        * we get here.
-        */
-       if (msgbufp->msg_magic != MSG_MAGIC) {
-               register int i;
 
-               msgbufp->msg_magic = MSG_MAGIC;
-               msgbufp->msg_bufx = msgbufp->msg_bufr = 0;
-               for (i=0; i < MSG_BSIZE; i++)
-                       msgbufp->msg_bufc[i] = 0;
-       }
        LOG_UNLOCK();
 
-       return (0);
+       return 0;
 }
 
 /*ARGSUSED*/
@@ -160,22 +277,160 @@ int
 logclose(__unused dev_t dev, __unused int flag, __unused int devtype, __unused struct proc *p)
 {
        LOG_LOCK();
-       log_open = 0;
+       logsoftc.sc_state &= ~(LOG_NBIO | LOG_ASYNC);
        selwakeup(&logsoftc.sc_selp);
        selthreadclear(&logsoftc.sc_selp);
+       log_open = 0;
+       LOG_UNLOCK();
+       return 0;
+}
+
+
+int
+oslogopen(__unused dev_t dev, __unused int flags, __unused int mode, struct proc *p)
+{
+       LOG_LOCK();
+       if (oslog_open) {
+               LOG_UNLOCK();
+               return EBUSY;
+       }
+       oslogsoftc.sc_pgid = p->p_pid;          /* signal process only */
+       oslog_open = 1;
+
+       LOG_UNLOCK();
+       return 0;
+}
+
+int
+oslogclose(__unused dev_t dev, __unused int flag, __unused int devtype, __unused struct proc *p)
+{
+       LOG_LOCK();
+       oslogsoftc.sc_state &= ~(LOG_NBIO | LOG_ASYNC);
+       selwakeup(&oslogsoftc.sc_selp);
+       selthreadclear(&oslogsoftc.sc_selp);
+       oslog_open = 0;
        LOG_UNLOCK();
-       return (0);
+       return 0;
+}
+
+int
+oslog_streamopen(__unused dev_t dev, __unused int flags, __unused int mode, struct proc *p)
+{
+       char *oslog_stream_msg_bufc = NULL;
+       oslog_stream_buf_entry_t entries = NULL;
+
+       stream_lock();
+       if (oslog_stream_open) {
+               stream_unlock();
+               return EBUSY;
+       }
+       stream_unlock();
+
+       // Allocate the stream buffer
+       oslog_stream_msg_bufc = kalloc(oslog_stream_buf_size);
+       if (!oslog_stream_msg_bufc) {
+               return ENOMEM;
+       }
+
+       /* entries to support kernel logging in stream mode */
+       entries = kalloc(oslog_stream_num_entries * sizeof(struct oslog_stream_buf_entry_s));
+       if (!entries) {
+               kfree(oslog_stream_msg_bufc, oslog_stream_buf_size);
+               return ENOMEM;
+       }
+
+       stream_lock();
+       if (oslog_stream_open) {
+               stream_unlock();
+               kfree(oslog_stream_msg_bufc, oslog_stream_buf_size);
+               kfree(entries, oslog_stream_num_entries * sizeof(struct oslog_stream_buf_entry_s));
+               return EBUSY;
+       }
+
+       assert(oslog_streambufp->msg_bufc == NULL);
+       oslog_streambufp->msg_bufc = oslog_stream_msg_bufc;
+       oslog_streambufp->msg_size = oslog_stream_buf_size;
+
+       oslog_stream_buf_entries = entries;
+
+       STAILQ_INIT(&oslog_stream_free_head);
+       STAILQ_INIT(&oslog_stream_buf_head);
+
+       for (int i = 0; i < oslog_stream_num_entries; i++) {
+               oslog_stream_buf_entries[i].type = oslog_stream_link_type_log;
+               oslog_stream_buf_entries[i].offset = 0;
+               oslog_stream_buf_entries[i].size = 0;
+               oslog_stream_buf_entries[i].timestamp = 0;
+               STAILQ_INSERT_TAIL(&oslog_stream_free_head, &oslog_stream_buf_entries[i], buf_entries);
+       }
+
+       /* there should be no pending entries in the stream */
+       assert(STAILQ_EMPTY(&oslog_stream_buf_head));
+       assert(oslog_streambufp->msg_bufx == 0);
+       assert(oslog_streambufp->msg_bufr == 0);
+
+       oslog_streambufp->msg_bufx = 0;
+       oslog_streambufp->msg_bufr = 0;
+       oslog_streamsoftc.sc_pgid = p->p_pid; /* signal process only */
+       oslog_stream_open = 1;
+       oslog_stream_buf_bytesavail = oslog_stream_buf_size;
+       stream_unlock();
+
+       return 0;
+}
+
+int
+oslog_streamclose(__unused dev_t dev, __unused int flag, __unused int devtype, __unused struct proc *p)
+{
+       oslog_stream_buf_entry_t next_entry = NULL;
+       char *oslog_stream_msg_bufc = NULL;
+       oslog_stream_buf_entry_t entries = NULL;
+
+       stream_lock();
+
+       if (oslog_stream_open == 0) {
+               stream_unlock();
+               return EBADF;
+       }
+
+       // Consume all log lines
+       while (!STAILQ_EMPTY(&oslog_stream_buf_head)) {
+               next_entry = STAILQ_FIRST(&oslog_stream_buf_head);
+               STAILQ_REMOVE_HEAD(&oslog_stream_buf_head, buf_entries);
+       }
+       oslog_streamwakeup_locked();
+       oslog_streamsoftc.sc_state &= ~(LOG_NBIO | LOG_ASYNC);
+       selwakeup(&oslog_streamsoftc.sc_selp);
+       selthreadclear(&oslog_streamsoftc.sc_selp);
+       oslog_stream_open = 0;
+       oslog_streambufp->msg_bufr = 0;
+       oslog_streambufp->msg_bufx = 0;
+       oslog_stream_msg_bufc = oslog_streambufp->msg_bufc;
+       oslog_streambufp->msg_bufc = NULL;
+       entries = oslog_stream_buf_entries;
+       oslog_stream_buf_entries = NULL;
+       oslog_streambufp->msg_size = 0;
+
+       stream_unlock();
+
+       // Free the stream buffer
+       kfree(oslog_stream_msg_bufc, oslog_stream_buf_size);
+       // Free the list entries
+       kfree(entries, oslog_stream_num_entries * sizeof(struct oslog_stream_buf_entry_s));
+
+       return 0;
 }
 
 /*ARGSUSED*/
 int
 logread(__unused dev_t dev, struct uio *uio, int flag)
 {
-       register long l;
+       int l;
        int error = 0;
+       struct msgbuf *mbp = logsoftc.sc_mbp;
 
        LOG_LOCK();
-       while (msgbufp->msg_bufr == msgbufp->msg_bufx) {
+       while (mbp->msg_bufr == mbp->msg_bufx) {
                if (flag & IO_NDELAY) {
                        error = EWOULDBLOCK;
                        goto out;
@@ -187,126 +442,346 @@ logread(__unused dev_t dev, struct uio *uio, int flag)
                logsoftc.sc_state |= LOG_RDWAIT;
                LOG_UNLOCK();
                /*
-                * If the wakeup is missed the ligtening bolt will wake this up 
-                * if there are any new characters. If that doesn't do it
-                * then wait for 5 sec and reevaluate 
+                * If the wakeup is missed
+                * then wait for 5 sec and reevaluate
                 */
-               if ((error = tsleep((caddr_t)msgbufp, LOG_RDPRI | PCATCH,
-                               "klog", 5 * hz)) != 0) {
+               if ((error = tsleep((caddr_t)mbp, LOG_RDPRI | PCATCH,
+                   "klog", 5 * hz)) != 0) {
                        /* if it times out; ignore */
-                       if (error != EWOULDBLOCK)
-                               return (error);
+                       if (error != EWOULDBLOCK) {
+                               return error;
+                       }
                }
                LOG_LOCK();
        }
        logsoftc.sc_state &= ~LOG_RDWAIT;
 
        while (uio_resid(uio) > 0) {
-               l = msgbufp->msg_bufx - msgbufp->msg_bufr;
-               if (l < 0)
-                       l = msgbufp->msg_size - msgbufp->msg_bufr;
+               int readpos;
+
+               l = mbp->msg_bufx - mbp->msg_bufr;
+               if (l < 0) {
+                       l = mbp->msg_size - mbp->msg_bufr;
+               }
                l = min(l, uio_resid(uio));
-               if (l == 0)
+               if (l == 0) {
                        break;
+               }
+
+               readpos = mbp->msg_bufr;
                LOG_UNLOCK();
-               error = uiomove((caddr_t)&msgbufp->msg_bufc[msgbufp->msg_bufr],
-                       (int)l, uio);
+               error = uiomove((caddr_t)&mbp->msg_bufc[readpos], l, uio);
                LOG_LOCK();
-               if (error)
+               if (error) {
                        break;
-               msgbufp->msg_bufr += l;
-               if (msgbufp->msg_bufr < 0 || msgbufp->msg_bufr >= msgbufp->msg_size)
-                       msgbufp->msg_bufr = 0;
+               }
+               mbp->msg_bufr = readpos + l;
+               if (mbp->msg_bufr >= mbp->msg_size) {
+                       mbp->msg_bufr = 0;
+               }
        }
 out:
        LOG_UNLOCK();
-       return (error);
+       return error;
+}
+
+/*ARGSUSED*/
+int
+oslog_streamread(__unused dev_t dev, struct uio *uio, int flag)
+{
+       int error = 0;
+       int copy_size = 0;
+       static char logline[FIREHOSE_CHUNK_SIZE];
+
+       stream_lock();
+
+       if (!oslog_stream_open) {
+               stream_unlock();
+               return EBADF;
+       }
+
+       while (STAILQ_EMPTY(&oslog_stream_buf_head)) {
+               assert(oslog_stream_buf_bytesavail == oslog_stream_buf_size);
+
+               if (flag & IO_NDELAY || oslog_streamsoftc.sc_state & LOG_NBIO) {
+                       stream_unlock();
+                       return EWOULDBLOCK;
+               }
+
+               oslog_streamsoftc.sc_state |= LOG_RDWAIT;
+               wait_result_t wr = assert_wait((event_t)oslog_streambufp,
+                   THREAD_INTERRUPTIBLE);
+               if (wr == THREAD_WAITING) {
+                       stream_unlock();
+                       wr = thread_block(THREAD_CONTINUE_NULL);
+                       stream_lock();
+               }
+
+               switch (wr) {
+               case THREAD_AWAKENED:
+               case THREAD_TIMED_OUT:
+                       break;
+               default:
+                       stream_unlock();
+                       return EINTR;
+               }
+       }
+
+       if (!oslog_stream_open) {
+               stream_unlock();
+               return EBADF;
+       }
+
+       int logpos = 0;
+       oslog_stream_buf_entry_t read_entry = NULL;
+       uint16_t rec_length;
+
+       read_entry = STAILQ_FIRST(&oslog_stream_buf_head);
+       assert(read_entry != NULL);
+       STAILQ_REMOVE_HEAD(&oslog_stream_buf_head, buf_entries);
+
+       // Copy the timestamp first
+       memcpy(logline + logpos, &read_entry->timestamp, sizeof(uint64_t));
+       logpos += sizeof(uint64_t);
+
+       switch (read_entry->type) {
+       /* Handle metadata messages */
+       case oslog_stream_link_type_metadata:
+       {
+               memcpy(logline + logpos,
+                   (read_entry->metadata), read_entry->size);
+               logpos += read_entry->size;
+
+               stream_unlock();
+
+               // Free the list entry
+               kfree(read_entry, (sizeof(struct oslog_stream_buf_entry_s) + read_entry->size));
+               break;
+       }
+       /* Handle log messages */
+       case oslog_stream_link_type_log:
+       {
+               /* ensure that the correct read entry was dequeued */
+               assert(read_entry->offset == oslog_streambufp->msg_bufr);
+               rec_length = read_entry->size;
+
+               // If the next log line is contiguous in the buffer, copy it out.
+               if (read_entry->offset + rec_length <= oslog_streambufp->msg_size) {
+                       memcpy(logline + logpos,
+                           oslog_streambufp->msg_bufc + read_entry->offset, rec_length);
+
+                       oslog_streambufp->msg_bufr += rec_length;
+                       if (oslog_streambufp->msg_bufr == oslog_streambufp->msg_size) {
+                               oslog_streambufp->msg_bufr = 0;
+                       }
+                       logpos += rec_length;
+               } else {
+                       // Otherwise, copy until the end of the buffer, and
+                       // copy the remaining bytes starting at index 0.
+                       int bytes_left = oslog_streambufp->msg_size - read_entry->offset;
+                       memcpy(logline + logpos,
+                           oslog_streambufp->msg_bufc + read_entry->offset, bytes_left);
+                       logpos += bytes_left;
+                       rec_length -= bytes_left;
+
+                       memcpy(logline + logpos, (const void *)oslog_streambufp->msg_bufc,
+                           rec_length);
+                       oslog_streambufp->msg_bufr = rec_length;
+                       logpos += rec_length;
+               }
+
+               oslog_stream_buf_bytesavail += read_entry->size;
+               assert(oslog_stream_buf_bytesavail <= oslog_stream_buf_size);
+
+               assert(oslog_streambufp->msg_bufr < oslog_streambufp->msg_size);
+               STAILQ_INSERT_TAIL(&oslog_stream_free_head, read_entry, buf_entries);
+
+               stream_unlock();
+               break;
+       }
+       default:
+       {
+               panic("Got unexpected log entry type: %hhu\n", read_entry->type);
+       }
+       }
+
+       copy_size = min(logpos, uio_resid(uio));
+       if (copy_size != 0) {
+               error = uiomove((caddr_t)logline, copy_size, uio);
+       }
+       (void)hw_atomic_add(&oslog_s_streamed_msgcount, 1);
+
+       return error;
 }
 
 /*ARGSUSED*/
 int
 logselect(__unused dev_t dev, int rw, void * wql, struct proc *p)
 {
-       switch (rw) {
+       const struct msgbuf *mbp = logsoftc.sc_mbp;
 
+       switch (rw) {
        case FREAD:
-               LOG_LOCK();     
-               if (msgbufp->msg_bufr != msgbufp->msg_bufx) {
+               LOG_LOCK();
+               if (mbp->msg_bufr != mbp->msg_bufx) {
                        LOG_UNLOCK();
-                       return (1);
+                       return 1;
                }
                selrecord(p, &logsoftc.sc_selp, wql);
                LOG_UNLOCK();
                break;
        }
-       return (0);
+       return 0;
+}
+
+int
+oslogselect(__unused dev_t dev, int rw, void * wql, struct proc *p)
+{
+       switch (rw) {
+       case FREAD:
+               LOG_LOCK();
+               if (os_log_wakeup) {
+                       LOG_UNLOCK();
+                       return 1;
+               }
+               selrecord(p, &oslogsoftc.sc_selp, wql);
+               LOG_UNLOCK();
+               break;
+       }
+       return 0;
+}
+
+int
+oslog_streamselect(__unused dev_t dev, int rw, void * wql, struct proc *p)
+{
+       int ret = 0;
+
+       stream_lock();
+
+       switch (rw) {
+       case FREAD:
+               if (STAILQ_EMPTY(&oslog_stream_buf_head)) {
+                       selrecord(p, &oslog_streamsoftc.sc_selp, wql);
+               } else {
+                       ret = 1;
+               }
+               break;
+       }
+
+       stream_unlock();
+       return ret;
 }
 
 void
-logwakeup(void)
+logwakeup(struct msgbuf *mbp)
 {
-       int pgid;
+       /* cf. r24974766 & r25201228*/
+       if (oslog_is_safe() == FALSE) {
+               return;
+       }
 
-       LOG_LOCK();     
+       LOG_LOCK();
        if (!log_open) {
                LOG_UNLOCK();
                return;
        }
+       if (NULL == mbp) {
+               mbp = logsoftc.sc_mbp;
+       }
+       if (mbp != logsoftc.sc_mbp) {
+               goto out;
+       }
        selwakeup(&logsoftc.sc_selp);
        if (logsoftc.sc_state & LOG_ASYNC) {
-               pgid = logsoftc.sc_pgid;
+               int pgid = logsoftc.sc_pgid;
                LOG_UNLOCK();
-               if (pgid < 0)
-                       gsignal(-pgid, SIGIO); 
-               else 
+               if (pgid < 0) {
+                       gsignal(-pgid, SIGIO);
+               } else {
                        proc_signal(pgid, SIGIO);
+               }
                LOG_LOCK();
        }
        if (logsoftc.sc_state & LOG_RDWAIT) {
-               wakeup((caddr_t)msgbufp);
+               wakeup((caddr_t)mbp);
                logsoftc.sc_state &= ~LOG_RDWAIT;
        }
+out:
        LOG_UNLOCK();
 }
 
 void
-klogwakeup(void)
+oslogwakeup(void)
 {
-       if (_logentrypend) {
-               _logentrypend = 0;
-               logwakeup();
+       LOG_LOCK();
+       if (!oslog_open) {
+               LOG_UNLOCK();
+               return;
+       }
+       selwakeup(&oslogsoftc.sc_selp);
+       os_log_wakeup = 1;
+       LOG_UNLOCK();
+}
+
+static void
+oslog_streamwakeup_locked(void)
+{
+       LCK_SPIN_ASSERT(&oslog_stream_lock, LCK_ASSERT_OWNED);
+       if (!oslog_stream_open) {
+               return;
+       }
+       selwakeup(&oslog_streamsoftc.sc_selp);
+       if (oslog_streamsoftc.sc_state & LOG_RDWAIT) {
+               wakeup((caddr_t)oslog_streambufp);
+               oslog_streamsoftc.sc_state &= ~LOG_RDWAIT;
        }
 }
 
+void
+oslog_streamwakeup(void)
+{
+       /* cf. r24974766 & r25201228*/
+       if (oslog_is_safe() == FALSE) {
+               return;
+       }
+
+       stream_lock();
+       oslog_streamwakeup_locked();
+       stream_unlock();
+}
+
 /*ARGSUSED*/
 int
 logioctl(__unused dev_t dev, u_long com, caddr_t data, __unused int flag, __unused struct proc *p)
 {
-       long l;
+       int l;
+       const struct msgbuf *mbp = logsoftc.sc_mbp;
 
        LOG_LOCK();
        switch (com) {
-
        /* return number of characters immediately available */
        case FIONREAD:
-               l = msgbufp->msg_bufx - msgbufp->msg_bufr;
-               if (l < 0)
-                       l += msgbufp->msg_size;
+               l = mbp->msg_bufx - mbp->msg_bufr;
+               if (l < 0) {
+                       l += mbp->msg_size;
+               }
                *(off_t *)data = l;
                break;
 
        case FIONBIO:
-               if (*(int *)data)
+               if (*(int *)data) {
                        logsoftc.sc_state |= LOG_NBIO;
-               else
+               } else {
                        logsoftc.sc_state &= ~LOG_NBIO;
+               }
                break;
 
        case FIOASYNC:
-               if (*(int *)data)
+               if (*(int *)data) {
                        logsoftc.sc_state |= LOG_ASYNC;
-               else
+               } else {
                        logsoftc.sc_state &= ~LOG_ASYNC;
+               }
                break;
 
        case TIOCSPGRP:
@@ -319,21 +794,139 @@ logioctl(__unused dev_t dev, u_long com, caddr_t data, __unused int flag, __unus
 
        default:
                LOG_UNLOCK();
-               return (-1);
+               return -1;
        }
        LOG_UNLOCK();
-       return (0);
+       return 0;
+}
+
+/*ARGSUSED*/
+int
+oslogioctl(__unused dev_t dev, u_long com, caddr_t data, __unused int flag, __unused struct proc *p)
+{
+       int ret = 0;
+       mach_vm_size_t buffer_size = (__firehose_buffer_kernel_chunk_count * FIREHOSE_CHUNK_SIZE);
+       firehose_buffer_map_info_t map_info = {0, 0};
+       firehose_buffer_t kernel_firehose_buffer = NULL;
+       mach_vm_address_t user_addr = 0;
+       mach_port_t mem_entry_ptr = MACH_PORT_NULL;
+
+       switch (com) {
+       /* return number of characters immediately available */
+
+       case LOGBUFFERMAP:
+               kernel_firehose_buffer = (firehose_buffer_t)kernel_firehose_addr;
+
+               ret = mach_make_memory_entry_64(kernel_map,
+                   &buffer_size,
+                   (mach_vm_offset_t) kernel_firehose_buffer,
+                   (MAP_MEM_VM_SHARE | VM_PROT_READ),
+                   &mem_entry_ptr,
+                   MACH_PORT_NULL);
+               if (ret == KERN_SUCCESS) {
+                       ret = mach_vm_map_kernel(get_task_map(current_task()),
+                           &user_addr,
+                           buffer_size,
+                           0,               /*  mask */
+                           VM_FLAGS_ANYWHERE,
+                           VM_MAP_KERNEL_FLAGS_NONE,
+                           VM_KERN_MEMORY_NONE,
+                           mem_entry_ptr,
+                           0,               /* offset */
+                           FALSE,               /* copy */
+                           VM_PROT_READ,
+                           VM_PROT_READ,
+                           VM_INHERIT_SHARE);
+               }
+
+               if (ret == KERN_SUCCESS) {
+                       map_info.fbmi_addr = (uint64_t) (user_addr);
+                       map_info.fbmi_size = buffer_size;
+                       bcopy(&map_info, data, sizeof(firehose_buffer_map_info_t));
+               }
+               break;
+       case LOGFLUSHED:
+               LOG_LOCK();
+               os_log_wakeup = 0;
+               LOG_UNLOCK();
+               __firehose_merge_updates(*(firehose_push_reply_t *)(data));
+               break;
+       default:
+               return -1;
+       }
+       return 0;
+}
+
+/*ARGSUSED*/
+int
+oslog_streamioctl(__unused dev_t dev, u_long com, caddr_t data, __unused int flag, __unused struct proc *p)
+{
+       int err = 0;
+
+       stream_lock();
+
+       switch (com) {
+       case FIONBIO:
+               if (data && *(int *)data) {
+                       oslog_streamsoftc.sc_state |= LOG_NBIO;
+               } else {
+                       oslog_streamsoftc.sc_state &= ~LOG_NBIO;
+               }
+               break;
+       case FIOASYNC:
+               if (data && *(int *)data) {
+                       oslog_streamsoftc.sc_state |= LOG_ASYNC;
+               } else {
+                       oslog_streamsoftc.sc_state &= ~LOG_ASYNC;
+               }
+               break;
+       default:
+               err = -1;
+               break;
+       }
+
+       stream_unlock();
+       return err;
 }
 
 void
 bsd_log_init(void)
 {
-       if (!log_inited) { 
-               msgbufp = &temp_msgbuf;
-               log_inited = 1;
-       }
+       /* After this point, we must be ready to accept characters */
 }
 
+void
+oslog_init(void)
+{
+       kern_return_t kr;
+       if (!PE_parse_boot_argn("firehose_chunk_count", &__firehose_buffer_kernel_chunk_count, sizeof(__firehose_buffer_kernel_chunk_count))) {
+               __firehose_buffer_kernel_chunk_count = FIREHOSE_BUFFER_KERNEL_DEFAULT_CHUNK_COUNT;
+       }
+       if (!PE_parse_boot_argn("firehose_io_pages", &__firehose_num_kernel_io_pages, sizeof(__firehose_num_kernel_io_pages))) {
+               __firehose_num_kernel_io_pages = FIREHOSE_BUFFER_KERNEL_DEFAULT_IO_PAGES;
+       }
+       if (!__firehose_kernel_configuration_valid(__firehose_buffer_kernel_chunk_count, __firehose_num_kernel_io_pages)) {
+               printf("illegal firehose configuration %u/%u, using defaults\n", __firehose_buffer_kernel_chunk_count, __firehose_num_kernel_io_pages);
+               __firehose_buffer_kernel_chunk_count = FIREHOSE_BUFFER_KERNEL_DEFAULT_CHUNK_COUNT;
+               __firehose_num_kernel_io_pages = FIREHOSE_BUFFER_KERNEL_DEFAULT_IO_PAGES;
+       }
+       vm_size_t size = __firehose_buffer_kernel_chunk_count * FIREHOSE_CHUNK_SIZE;
+
+       oslog_lock_init();
+
+       kr = kmem_alloc_flags(kernel_map, &kernel_firehose_addr,
+           size + (2 * PAGE_SIZE), VM_KERN_MEMORY_LOG,
+           KMA_GUARD_FIRST | KMA_GUARD_LAST);
+       if (kr != KERN_SUCCESS) {
+               panic("Failed to allocate memory for firehose logging buffer");
+       }
+       kernel_firehose_addr += PAGE_SIZE;
+       bzero((void *)kernel_firehose_addr, size);
+       /* register buffer with firehose */
+       kernel_firehose_addr = (vm_offset_t)__firehose_buffer_create((size_t *) &size);
+
+       printf("oslog_init completed, %u chunks, %u io pages\n", __firehose_buffer_kernel_chunk_count, __firehose_num_kernel_io_pages);
+}
 
 /*
  * log_putc_locked
@@ -351,29 +944,182 @@ bsd_log_init(void)
  *             SMP reentrancy.
  */
 void
-log_putc_locked(char c)
+log_putc_locked(struct msgbuf *mbp, char c)
+{
+       mbp->msg_bufc[mbp->msg_bufx++] = c;
+       if (mbp->msg_bufx >= mbp->msg_size) {
+               mbp->msg_bufx = 0;
+       }
+}
+
+static oslog_stream_buf_entry_t
+oslog_stream_find_free_buf_entry_locked(void)
 {
-       register struct msgbuf *mbp;
+       struct msgbuf *mbp;
+       oslog_stream_buf_entry_t buf_entry = NULL;
+
+       LCK_SPIN_ASSERT(&oslog_stream_lock, LCK_ASSERT_OWNED);
+
+       mbp = oslog_streambufp;
+
+       buf_entry = STAILQ_FIRST(&oslog_stream_free_head);
+       if (buf_entry) {
+               STAILQ_REMOVE_HEAD(&oslog_stream_free_head, buf_entries);
+       } else {
+               // If no list elements are available in the free-list,
+               // consume the next log line so we can free up its list element
+               oslog_stream_buf_entry_t prev_entry = NULL;
 
-       if (!log_inited) {
-               panic("bsd log is not inited");
+               buf_entry = STAILQ_FIRST(&oslog_stream_buf_head);
+               while (buf_entry->type == oslog_stream_link_type_metadata) {
+                       prev_entry = buf_entry;
+                       buf_entry = STAILQ_NEXT(buf_entry, buf_entries);
+               }
+
+               if (prev_entry == NULL) {
+                       STAILQ_REMOVE_HEAD(&oslog_stream_buf_head, buf_entries);
+               } else {
+                       STAILQ_REMOVE_AFTER(&oslog_stream_buf_head, prev_entry, buf_entries);
+               }
+
+               mbp->msg_bufr += buf_entry->size;
+               oslog_s_dropped_msgcount++;
+               if (mbp->msg_bufr >= mbp->msg_size) {
+                       mbp->msg_bufr = (mbp->msg_bufr % mbp->msg_size);
+               }
        }
 
-       mbp = msgbufp; 
-       if (mbp-> msg_magic != MSG_MAGIC) {
-               register int i;
+       return buf_entry;
+}
+
+void
+oslog_streamwrite_metadata_locked(oslog_stream_buf_entry_t m_entry)
+{
+       LCK_SPIN_ASSERT(&oslog_stream_lock, LCK_ASSERT_OWNED);
+       STAILQ_INSERT_TAIL(&oslog_stream_buf_head, m_entry, buf_entries);
+
+       return;
+}
+
+static void
+oslog_streamwrite_append_bytes(const char *buffer, int buflen)
+{
+       struct msgbuf *mbp;
+
+       LCK_SPIN_ASSERT(&oslog_stream_lock, LCK_ASSERT_OWNED);
+
+       assert(oslog_stream_buf_bytesavail >= buflen);
+       oslog_stream_buf_bytesavail -= buflen;
+       assert(oslog_stream_buf_bytesavail >= 0);
+
+       mbp = oslog_streambufp;
+       if (mbp->msg_bufx + buflen <= mbp->msg_size) {
+               /*
+                * If this will fit without needing to be split across the end
+                * of the buffer, copy it directly in one go.
+                */
+               memcpy((void *)(mbp->msg_bufc + mbp->msg_bufx), buffer, buflen);
+
+               mbp->msg_bufx += buflen;
+               if (mbp->msg_bufx == mbp->msg_size) {
+                       mbp->msg_bufx = 0;
+               }
+       } else {
+               /*
+                * Copy up to the end of the stream buffer, and then put what remains
+                * at the beginning.
+                */
+               int bytes_left = mbp->msg_size - mbp->msg_bufx;
+               memcpy((void *)(mbp->msg_bufc + mbp->msg_bufx), buffer, bytes_left);
+
+               buflen -= bytes_left;
+               buffer += bytes_left;
 
-               mbp->msg_magic = MSG_MAGIC;
-               mbp->msg_bufx = mbp->msg_bufr = 0;
-               for (i=0; i < MSG_BSIZE; i++)
-                       mbp->msg_bufc[i] = 0;
+               // Copy the remainder of the data from the beginning of stream
+               memcpy((void *)mbp->msg_bufc, buffer, buflen);
+               mbp->msg_bufx = buflen;
        }
-       mbp->msg_bufc[mbp->msg_bufx++] = c;
-       _logentrypend = 1;
-       if (mbp->msg_bufx < 0 || mbp->msg_bufx >= msgbufp->msg_size)
-               mbp->msg_bufx = 0;
+       return;
 }
 
+void
+oslog_streamwrite_locked(firehose_tracepoint_id_u ftid,
+    uint64_t stamp, const void *pubdata, size_t publen)
+{
+       struct msgbuf *mbp;
+       oslog_stream_buf_entry_t buf_entry = NULL;
+       oslog_stream_buf_entry_t next_entry = NULL;
+
+       uint16_t ft_size = offsetof(struct firehose_tracepoint_s, ft_data);
+       int ft_length = ft_size + publen;
+
+       LCK_SPIN_ASSERT(&oslog_stream_lock, LCK_ASSERT_OWNED);
+
+       mbp = oslog_streambufp;
+       if (ft_length > mbp->msg_size) {
+               (void)hw_atomic_add(&oslog_s_error_count, 1);
+               return;
+       }
+
+       // Ensure that we have a list element for this record
+       buf_entry = oslog_stream_find_free_buf_entry_locked();
+
+       assert(buf_entry != NULL);
+
+       while (ft_length > oslog_stream_buf_bytesavail) {
+               oslog_stream_buf_entry_t prev_entry = NULL;
+
+               next_entry = STAILQ_FIRST(&oslog_stream_buf_head);
+               assert(next_entry != NULL);
+               while (next_entry->type == oslog_stream_link_type_metadata) {
+                       prev_entry = next_entry;
+                       next_entry = STAILQ_NEXT(next_entry, buf_entries);
+               }
+
+               if (prev_entry == NULL) {
+                       STAILQ_REMOVE_HEAD(&oslog_stream_buf_head, buf_entries);
+               } else {
+                       STAILQ_REMOVE_AFTER(&oslog_stream_buf_head, prev_entry, buf_entries);
+               }
+
+               mbp->msg_bufr += next_entry->size;
+               if (mbp->msg_bufr >= mbp->msg_size) {
+                       mbp->msg_bufr = (mbp->msg_bufr % mbp->msg_size);
+               }
+
+               oslog_s_dropped_msgcount++;
+               oslog_stream_buf_bytesavail += next_entry->size;
+               assert(oslog_stream_buf_bytesavail <= oslog_stream_buf_size);
+
+               STAILQ_INSERT_TAIL(&oslog_stream_free_head, next_entry, buf_entries);
+       }
+
+       assert(ft_length <= oslog_stream_buf_bytesavail);
+
+       // Write the log line and update the list entry for this record
+       buf_entry->offset = mbp->msg_bufx;
+       buf_entry->size = ft_length;
+       buf_entry->timestamp = stamp;
+       buf_entry->type = oslog_stream_link_type_log;
+
+       // Construct a tracepoint
+       struct firehose_tracepoint_s fs = {
+               .ft_thread = thread_tid(current_thread()),
+               .ft_id.ftid_value = ftid.ftid_value,
+               .ft_length = publen
+       };
+
+       oslog_streamwrite_append_bytes((char *)&fs, sizeof(fs));
+       oslog_streamwrite_append_bytes(pubdata, publen);
+
+       assert(mbp->msg_bufr < mbp->msg_size);
+       // Insert the element to the buffer data list
+       STAILQ_INSERT_TAIL(&oslog_stream_buf_head, buf_entry, buf_entries);
+
+       return;
+}
+
+
 
 /*
  * log_putc
@@ -385,18 +1131,24 @@ log_putc_locked(char c)
  *
  * Returns:    (void)
  *
- * Notes:      This function is used for syingle byte output to the log.  It
+ * Notes:      This function is used for single byte output to the log.  It
  *             primarily exists to maintain binary backward compatibility.
  */
 void
 log_putc(char c)
 {
-       if (!log_inited) {
-               panic("bsd log is not inited");
-       }
+       int unread_count = 0;
        LOG_LOCK();
-       log_putc_locked(c);
+       log_putc_locked(msgbufp, c);
+       unread_count = msgbufp->msg_bufx - msgbufp->msg_bufr;
        LOG_UNLOCK();
+
+       if (unread_count < 0) {
+               unread_count = 0 - unread_count;
+       }
+       if (c == '\n' || unread_count >= (msgbufp->msg_size / 2)) {
+               logwakeup(msgbufp);
+       }
 }
 
 
@@ -406,59 +1158,170 @@ log_putc(char c)
  * to the kernel command line, and to read the current size using
  *   sysctl kern.msgbuf
  * If there is no parameter on the kernel command line, the buffer is
- * allocated statically and is MSG_BSIZE characters in size, otherwise
- * memory is dynamically allocated.
- * This function may only be called once, during kernel initialization.
- * Memory management must already be up. The buffer must not have
- * overflown yet.
+ * allocated statically and is CONFIG_MSG_BSIZE characters in size, otherwise
+ * memory is dynamically allocated. Memory management must already be up.
  */
-void
-log_setsize(long size) {
+int
+log_setsize(int size)
+{
        char *new_logdata;
-       if (msgbufp->msg_size!=MSG_BSIZE) {
-               printf("log_setsize: attempt to change size more than once\n");
-               return;
+       int new_logsize, new_bufr, new_bufx;
+       char *old_logdata;
+       int old_logsize, old_bufr, old_bufx;
+       int i, count;
+       char *p, ch;
+
+       if (size > MAX_MSG_BSIZE) {
+               return EINVAL;
        }
-       if (size==MSG_BSIZE)
-               return;
-       if (size<MSG_BSIZE) { /* we don't support reducing the log size */
-               printf("log_setsize: can't decrease log size\n");
-               return;
+
+       if (size <= 0) {
+               return EINVAL;
        }
+
+       new_logsize = size;
        if (!(new_logdata = (char*)kalloc(size))) {
                printf("log_setsize: unable to allocate memory\n");
-               return;
+               return ENOMEM;
        }
+       bzero(new_logdata, new_logsize);
+
        LOG_LOCK();
-       bcopy(smsg_bufc, new_logdata, MSG_BSIZE);
-       bzero(new_logdata+MSG_BSIZE, size - MSG_BSIZE);
-       /* this memory is now dead - clear it so that it compresses better
-          in case of suspend to disk etc. */
-       bzero(smsg_bufc, MSG_BSIZE);
-       msgbufp->msg_size = size;
+
+       old_logsize = msgbufp->msg_size;
+       old_logdata = msgbufp->msg_bufc;
+       old_bufr = msgbufp->msg_bufr;
+       old_bufx = msgbufp->msg_bufx;
+
+       LOG_SETSIZE_DEBUG("log_setsize(%d): old_logdata %p old_logsize %d old_bufr %d old_bufx %d\n",
+           size, old_logdata, old_logsize, old_bufr, old_bufx);
+
+       /* start "new_logsize" bytes before the write pointer */
+       if (new_logsize <= old_bufx) {
+               count = new_logsize;
+               p = old_logdata + old_bufx - count;
+       } else {
+               /*
+                * if new buffer is bigger, copy what we have and let the
+                * bzero above handle the difference
+                */
+               count = MIN(new_logsize, old_logsize);
+               p = old_logdata + old_logsize - (count - old_bufx);
+       }
+       for (i = 0; i < count; i++) {
+               if (p >= old_logdata + old_logsize) {
+                       p = old_logdata;
+               }
+
+               ch = *p++;
+               new_logdata[i] = ch;
+       }
+
+       new_bufx = i;
+       if (new_bufx >= new_logsize) {
+               new_bufx = 0;
+       }
+       msgbufp->msg_bufx = new_bufx;
+
+       new_bufr = old_bufx - old_bufr; /* how much were we trailing bufx by? */
+       if (new_bufr < 0) {
+               new_bufr += old_logsize;
+       }
+       new_bufr = new_bufx - new_bufr; /* now relative to oldest data in new buffer */
+       if (new_bufr < 0) {
+               new_bufr += new_logsize;
+       }
+       msgbufp->msg_bufr = new_bufr;
+
+       msgbufp->msg_size = new_logsize;
        msgbufp->msg_bufc = new_logdata;
+
+       LOG_SETSIZE_DEBUG("log_setsize(%d): new_logdata %p new_logsize %d new_bufr %d new_bufx %d\n",
+           size, new_logdata, new_logsize, new_bufr, new_bufx);
+
        LOG_UNLOCK();
-       printf("set system log size to %ld bytes\n", msgbufp->msg_size);
+
+       /* this memory is now dead - clear it so that it compresses better
+        *  in case of suspend to disk etc. */
+       bzero(old_logdata, old_logsize);
+       if (old_logdata != smsg_bufc) {
+               /* dynamic memory that must be freed */
+               kfree(old_logdata, old_logsize);
+       }
+
+       printf("set system log size to %d bytes\n", new_logsize);
+
+       return 0;
+}
+
+void
+oslog_setsize(int size)
+{
+       uint16_t scale = 0;
+       // If the size is less than the default stream buffer
+       // do nothing
+       if (size <= OSLOG_STREAM_BUF_SIZE) {
+               return;
+       }
+
+       scale = (uint16_t) (size / OSLOG_STREAM_BUF_SIZE);
+
+       oslog_stream_buf_size = size;
+       oslog_stream_num_entries = scale * OSLOG_NUM_STREAM_ENTRIES;
+       printf("oslog_setsize: new buffer size = %d, new num entries= %d\n", oslog_stream_buf_size, oslog_stream_num_entries);
+}
+
+SYSCTL_PROC(_kern, OID_AUTO, msgbuf,
+    CTLTYPE_INT | CTLFLAG_RW | CTLFLAG_LOCKED, 0, 0,
+    sysctl_kern_msgbuf, "I", "");
+
+static int
+sysctl_kern_msgbuf(struct sysctl_oid *oidp __unused,
+    void *arg1 __unused, int arg2 __unused, struct sysctl_req *req)
+{
+       int old_bufsize, bufsize;
+       int error;
+
+       LOG_LOCK();
+       old_bufsize = bufsize = msgbufp->msg_size;
+       LOG_UNLOCK();
+
+       error = sysctl_io_number(req, bufsize, sizeof(bufsize), &bufsize, NULL);
+       if (error) {
+               return error;
+       }
+
+       if (bufsize != old_bufsize) {
+               error = log_setsize(bufsize);
+       }
+
+       return error;
 }
 
-SYSCTL_LONG(_kern, OID_AUTO, msgbuf, CTLFLAG_RD, &temp_msgbuf.msg_size, "");
 
 /*
- * This should be called by single user mode /sbin/dmesg only.
+ * This should be called by /sbin/dmesg only via libproc.
  * It returns as much data still in the buffer as possible.
  */
 int
-log_dmesg(user_addr_t buffer, uint32_t buffersize, int32_t * retval) {
+log_dmesg(user_addr_t buffer, uint32_t buffersize, int32_t * retval)
+{
        uint32_t i;
-       uint32_t localbuff_size = (msgbufp->msg_size + 2);
+       uint32_t localbuff_size;
        int error = 0, newl, skip;
        char *localbuff, *p, *copystart, ch;
-       long copysize;  
+       size_t copysize;
 
+       LOG_LOCK();
+       localbuff_size = (msgbufp->msg_size + 2); /* + '\n' + '\0' */
+       LOG_UNLOCK();
+
+       /* Allocate a temporary non-circular buffer for copyout */
        if (!(localbuff = (char *)kalloc(localbuff_size))) {
                printf("log_dmesg: unable to allocate memory\n");
-               return (ENOMEM);
+               return ENOMEM;
        }
+
        /* in between here, the log could become bigger, but that's fine */
        LOG_LOCK();
 
@@ -468,33 +1331,38 @@ log_dmesg(user_addr_t buffer, uint32_t buffersize, int32_t * retval) {
         */
        p = msgbufp->msg_bufc + msgbufp->msg_bufx;
        for (i = newl = skip = 0; p != msgbufp->msg_bufc + msgbufp->msg_bufx - 1; ++p) {
-               if (p >= msgbufp->msg_bufc + msgbufp->msg_size)
+               if (p >= msgbufp->msg_bufc + msgbufp->msg_size) {
                        p = msgbufp->msg_bufc;
+               }
                ch = *p;
                /* Skip "\n<.*>" syslog sequences. */
                if (skip) {
-                       if (ch == '>')
+                       if (ch == '>') {
                                newl = skip = 0;
+                       }
                        continue;
                }
                if (newl && ch == '<') {
                        skip = 1;
                        continue;
                }
-               if (ch == '\0')
+               if (ch == '\0') {
                        continue;
-               newl = ch == '\n';
+               }
+               newl = (ch == '\n');
                localbuff[i++] = ch;
                /* The original version of this routine contained a buffer
                 * overflow. At the time, a "small" targeted fix was desired
                 * so the change below to check the buffer bounds was made.
                 * TODO: rewrite this needlessly convoluted routine.
                 */
-               if (i == (localbuff_size - 2))
+               if (i == (localbuff_size - 2)) {
                        break;
+               }
        }
-       if (!newl)
+       if (!newl) {
                localbuff[i++] = '\n';
+       }
        localbuff[i++] = 0;
 
        if (buffersize >= i) {
@@ -508,9 +1376,54 @@ log_dmesg(user_addr_t buffer, uint32_t buffersize, int32_t * retval) {
        LOG_UNLOCK();
 
        error = copyout(copystart, buffer, copysize);
-       if (!error)
+       if (!error) {
                *retval = copysize;
+       }
 
        kfree(localbuff, localbuff_size);
-       return (error);
+       return error;
 }
+
+#ifdef CONFIG_XNUPOST
+
+uint32_t find_pattern_in_buffer(char * pattern, uint32_t len, int expected_count);
+
+/*
+ * returns count of pattern found in systemlog buffer.
+ * stops searching further if count reaches expected_count.
+ */
+uint32_t
+find_pattern_in_buffer(char * pattern, uint32_t len, int expected_count)
+{
+       int match_count = 0;
+       int i           = 0;
+       int j           = 0;
+       int no_match    = 0;
+       int pos         = 0;
+       char ch         = 0;
+
+       if (pattern == NULL || len == 0 || expected_count == 0) {
+               return 0;
+       }
+
+       for (i = 0; i < msgbufp->msg_size; i++) {
+               no_match = 0;
+               for (j = 0; j < (int)len; j++) {
+                       pos = (msgbufp->msg_bufx + i + j) % msgbufp->msg_size;
+                       ch  = msgbufp->msg_bufc[pos];
+                       if (ch != pattern[j]) {
+                               no_match = 1;
+                               break;
+                       }
+               }
+               if (no_match == 0) {
+                       match_count++;
+                       if (match_count >= expected_count) {
+                               break;
+                       }
+               }
+       }
+       return match_count;
+}
+
+#endif