x86: sends the EOI later in the IRQ path
[akaros.git] / kern / src / monitor.c
index 71855a7..01b7c99 100644 (file)
@@ -8,6 +8,7 @@
 #include <arch/arch.h>
 #include <stab.h>
 #include <smp.h>
+#include <console.h>
 #include <arch/console.h>
 
 #include <stdio.h>
 #include <pmap.h>
 #include <kdebug.h>
 #include <testing.h>
-#include <kfs.h>
 #include <manager.h>
 #include <schedule.h>
-#include <resource.h>
+#include <kdebug.h>
+#include <syscall.h>
+#include <kmalloc.h>
+#include <elf.h>
+#include <event.h>
+#include <trap.h>
+#include <time.h>
 
 #include <ros/memlayout.h>
+#include <ros/event.h>
 
 #define CMDBUF_SIZE    80      // enough for one VGA text line
 
@@ -31,7 +38,7 @@ typedef struct command {
        const char *NTS name;
        const char *NTS desc;
        // return -1 to force monitor to exit
-       int (*func)(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf);
+       int (*func)(int argc, char **argv, struct hw_trapframe *hw_tf);
 } command_t;
 
 static command_t (RO commands)[] = {
@@ -39,20 +46,30 @@ static command_t (RO commands)[] = {
        { "kerninfo", "Display information about the kernel", mon_kerninfo },
        { "backtrace", "Dump a backtrace", mon_backtrace },
        { "reboot", "Take a ride to the South Bay", mon_reboot },
-       { "showmapping", "Shows VA->PA mappings between two virtual addresses (parameters)", mon_showmapping},
+       { "showmapping", "Shows VA->PA mappings", mon_showmapping},
        { "setmapperm", "Sets permissions on a VA->PA mapping", mon_setmapperm},
        { "cpuinfo", "Prints CPU diagnostics", mon_cpuinfo},
+       { "ps", "Prints process list", mon_ps},
        { "nanwan", "Meet Nanwan!!", mon_nanwan},
-       { "kfs_ls", "List files in KFS", mon_kfs_ls},
-       { "kfs_run", "Create and run a program from KFS", mon_kfs_run},
+       { "bin_ls", "List files in /bin", mon_bin_ls},
+       { "bin_run", "Create and run a program from /bin", mon_bin_run},
        { "manager", "Run the manager", mon_manager},
        { "procinfo", "Show information about processes", mon_procinfo},
+       { "exit", "Leave the monitor", mon_exit},
+       { "kfunc", "Run a kernel function directly (!!!)", mon_kfunc},
+       { "notify", "Notify a process.  Vcoreid will skip their prefs", mon_notify},
+       { "measure", "Run a specific measurement", mon_measure},
+       { "trace", "Run some tracing functions", mon_trace},
+       { "monitor", "Run the monitor on another core", mon_monitor},
+       { "fs", "Filesystem Diagnostics", mon_fs},
+       { "bb", "Try to run busybox (ash)", mon_bb},
+       { "alarm", "Alarm Diagnostics", mon_alarm},
 };
 #define NCOMMANDS (sizeof(commands)/sizeof(commands[0]))
 
 /***** Implementations of basic kernel monitor commands *****/
 
-int mon_help(int argc, char **argv, trapframe_t *tf)
+int mon_help(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
        int i;
 
@@ -61,15 +78,21 @@ int mon_help(int argc, char **argv, trapframe_t *tf)
        return 0;
 }
 
-int mon_kerninfo(int argc, char **argv, trapframe_t *tf)
+int mon_ps(int argc, char** argv, struct hw_trapframe *hw_tf)
+{
+       print_allpids();
+       return 0;
+}
+
+int mon_kerninfo(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
        extern char (RO SNT _start)[], (RO SNT etext)[], (RO SNT edata)[], (RO SNT end)[];
 
        cprintf("Special kernel symbols:\n");
-       cprintf("  _start %08x (virt)  %08x (phys)\n", _start, (uint32_t)(_start - KERNBASE));
-       cprintf("  etext  %08x (virt)  %08x (phys)\n", etext, (uint32_t)(etext - KERNBASE));
-       cprintf("  edata  %08x (virt)  %08x (phys)\n", edata, (uint32_t)(edata - KERNBASE));
-       cprintf("  end    %08x (virt)  %08x (phys)\n", end, (uint32_t)(end - KERNBASE));
+       cprintf("  _start %016x (virt)  %016x (phys)\n", _start, (uintptr_t)(_start - KERNBASE));
+       cprintf("  etext  %016x (virt)  %016x (phys)\n", etext, (uintptr_t)(etext - KERNBASE));
+       cprintf("  edata  %016x (virt)  %016x (phys)\n", edata, (uintptr_t)(edata - KERNBASE));
+       cprintf("  end    %016x (virt)  %016x (phys)\n", end, (uintptr_t)(end - KERNBASE));
        cprintf("Kernel executable memory footprint: %dKB\n",
                (uint32_t)(end-_start+1023)/1024);
        return 0;
@@ -102,15 +125,27 @@ static char RO* function_of(uint32_t address)
 }
 #endif
 
-int mon_backtrace(int argc, char **argv, trapframe_t *tf)
+int mon_backtrace(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
-       backtrace();
+       uintptr_t pc, fp;
+       if (argc == 1) {
+               backtrace();
+               return 0;
+       }
+       if (argc != 3) {
+               printk("Need either no arguments, or two (PC and FP) in hex\n");
+               return 1;
+       }
+       pc = strtol(argv[1], 0, 16);
+       fp = strtol(argv[2], 0, 16);
+       printk("Backtrace from instruction %p, with frame pointer %p\n", pc, fp);
+       backtrace_frame(pc, fp);
        return 0;
 }
 
-int mon_reboot(int argc, char **argv, trapframe_t *tf)
+int mon_reboot(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
-       cprintf("[Irish Accent]: She's goin' down, Cap'n!\n");
+       cprintf("[Scottish Accent]: She's goin' down, Cap'n!\n");
        reboot();
 
        // really, should never see this
@@ -118,17 +153,14 @@ int mon_reboot(int argc, char **argv, trapframe_t *tf)
        return 0;
 }
 
-int mon_showmapping(int argc, char **argv, trapframe_t *tf)
+int mon_showmapping(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
        if (argc < 2) {
-               cprintf("Shows virtual -> physical mappings for a virtual address range.\n");
+               cprintf("Shows virtual -> physical mappings for a virt addr range.\n");
                cprintf("Usage: showmapping START_ADDR [END_ADDR]\n");
                return 1;
        }
-       pde_t* pgdir = (pde_t*)vpd;
-       pte_t *pte, *pde;
-       page_t* page;
-       uintptr_t start, i;
+       uintptr_t start;
        size_t size;
        start = ROUNDDOWN(strtol(argv[1], 0, 16), PGSIZE);
        size = (argc == 2) ? 1 : strtol(argv[2], 0, 16) - start;
@@ -141,9 +173,9 @@ int mon_showmapping(int argc, char **argv, trapframe_t *tf)
        return 0;
 }
 
-int mon_setmapperm(int argc, char **argv, trapframe_t *tf)
+int mon_setmapperm(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
-#ifndef __i386__
+#ifndef CONFIG_X86_32
        cprintf("I don't support this call yet!\n");
        return 1;
 #else
@@ -167,13 +199,13 @@ int mon_setmapperm(int argc, char **argv, trapframe_t *tf)
        pde = &pgdir[PDX(va)];
        cprintf("   Virtual    Physical  Ps Dr Ac CD WT U W\n");
        cprintf("------------------------------------------\n");
-       cprintf("%08p  %08p  %1d  %1d  %1d  %1d  %1d  %1d %1d\n", va, page2pa(page),
+       cprintf("%p  %p  %1d  %1d  %1d  %1d  %1d  %1d %1d\n", va, page2pa(page),
               (*pte & PTE_PS) >> 7, (*pte & PTE_D) >> 6, (*pte & PTE_A) >> 5,
               (*pte & PTE_PCD) >> 4, (*pte & PTE_PWT) >> 3, (*pte & *pde & PTE_U) >> 2,
               (*pte & *pde & PTE_W) >> 1);
        *pte = PTE_ADDR(*pte) | (*pte & PTE_PS) |
               (PGOFF(strtol(argv[2], 0, 16)) & ~PTE_PS ) | PTE_P;
-       cprintf("%08p  %08p  %1d  %1d  %1d  %1d  %1d  %1d %1d\n", va, page2pa(page),
+       cprintf("%p  %p  %1d  %1d  %1d  %1d  %1d  %1d %1d\n", va, page2pa(page),
               (*pte & PTE_PS) >> 7, (*pte & PTE_D) >> 6, (*pte & PTE_A) >> 5,
               (*pte & PTE_PCD) >> 4, (*pte & PTE_PWT) >> 3, (*pte & *pde & PTE_U) >> 2,
               (*pte & *pde & PTE_W) >> 1);
@@ -181,29 +213,27 @@ int mon_setmapperm(int argc, char **argv, trapframe_t *tf)
 #endif
 }
 
-int mon_cpuinfo(int argc, char **argv, trapframe_t *tf)
+int mon_cpuinfo(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
        cprintf("Number of CPUs detected: %d\n", num_cpus);
        cprintf("Calling CPU's ID: 0x%08x\n", core_id());
 
-#ifdef __i386__
        if (argc < 2)
                smp_call_function_self(test_print_info_handler, NULL, 0);
        else
-               smp_call_function_single(strtol(argv[1], 0, 16),
+               smp_call_function_single(strtol(argv[1], 0, 10),
                                         test_print_info_handler, NULL, 0);
-#endif
        return 0;
 }
 
-int mon_manager(int argc, char** argv, trapframe_t *tf)
+int mon_manager(int argc, char** argv, struct hw_trapframe *hw_tf)
 {
        manager();
        panic("should never get here");
        return 0;
 }
 
-int mon_nanwan(int argc, char **argv, trapframe_t *tf)
+int mon_nanwan(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
        /* Borrowed with love from http://www.geocities.com/SoHo/7373/zoo.htm
         * (http://www.ascii-art.com/).  Slightly modified to make it 25 lines tall.
@@ -235,49 +265,72 @@ int mon_nanwan(int argc, char **argv, trapframe_t *tf)
        return 0;
 }
 
-int mon_kfs_ls(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
+int mon_bin_ls(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
-       printk("Files in KFS:\n-------------------------------\n");
-       for (int i = 0; i < MAX_KFS_FILES; i++)
-               if (kfs[i].name[0])
-                       printk("%s\n", kfs[i].name);
+       struct dirent dir = {0};
+       struct file *bin_dir;
+       int retval = 0;
+
+       bin_dir = do_file_open("/bin", 0, 0);
+       if (!bin_dir) {
+               printk("No /bin directory!\n");
+               return 1;
+       }
+       printk("Files in /bin:\n-------------------------------\n");
+       do {
+               retval = bin_dir->f_op->readdir(bin_dir, &dir); 
+               printk("%s\n", dir.d_name);
+       } while (retval == 1);
+       kref_put(&bin_dir->f_kref);
        return 0;
 }
 
-int mon_kfs_run(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
+int mon_bin_run(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
-       if (argc != 2) {
-               printk("Usage: kfs_run FILENAME\n");
+       if (argc < 2) {
+               printk("Usage: bin_run FILENAME\n");
                return 1;
        }
-       int kfs_inode = kfs_lookup_path(argv[1]);
-       if (kfs_inode < 0) {
-               printk("Bad filename!\n");
+       struct file *program;
+       int retval = 0;
+       char buf[6 + MAX_FILENAME_SZ] = "/bin/";        /* /bin/ + max + \0 */
+       strncpy(buf + 5, argv[1], MAX_FILENAME_SZ);
+       program = do_file_open(buf, 0, 0);
+       if (!program) {
+               printk("No such program!\n");
                return 1;
        }
-       struct proc *p = kfs_proc_create(kfs_inode);
-       // go from PROC_CREATED->PROC_RUNNABLE_S
-       spin_lock_irqsave(&p->proc_lock); // might not be necessary for a mon function
-       __proc_set_state(p, PROC_RUNNABLE_S);
-       schedule_proc(p);
-       spin_unlock_irqsave(&p->proc_lock);
-       // Should never return from schedule (env_pop in there)
-       // also note you may not get the process you created, in the event there
-       // are others floating around that are runnable
-       schedule();
+       char **p_argv = kmalloc(sizeof(char*) * argc, 0);       /* bin_run's argc */
+       for (int i = 0; i < argc - 1; i++)
+               p_argv[i] = argv[i + 1];
+       p_argv[argc - 1] = 0;
+       char *p_envp[] = {"LD_LIBRARY_PATH=/lib", 0};
+       struct proc *p = proc_create(program, p_argv, p_envp);
+       kfree(p_argv);
+       proc_wakeup(p);
+       proc_decref(p); /* let go of the reference created in proc_create() */
+       kref_put(&program->f_kref);
+       /* Make a scheduling decision.  You might not get the process you created,
+        * in the event there are others floating around that are runnable */
+       run_scheduler();
+       /* want to idle, so we un the process we just selected.  this is a bit
+        * hackish, but so is the monitor. */
+       smp_idle();
+       assert(0);
        return 0;
 }
 
-int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
+int mon_procinfo(int argc, char **argv, struct hw_trapframe *hw_tf)
 {
+       int8_t irq_state = 0;
        if (argc < 2) {
                printk("Usage: procinfo OPTION\n");
                printk("\tidlecores: show idle core map\n");
                printk("\tall: show all active pids\n");
-               printk("\trunnable: show proc_runnablelist\n");
+               printk("\tsched: scheduler diagnostic report\n");
                printk("\tresources: show resources wanted/granted for all procs\n");
                printk("\tpid NUM: show a lot of info for proc NUM\n");
-               printk("\tunlock NUM: unlock the lock for proc NUM (OMG!!!)\n");
+               printk("\tunlock: unlock the lock for the ADDR (OMG!!!)\n");
                printk("\tkill NUM: destroy proc NUM\n");
                return 1;
        }
@@ -285,8 +338,8 @@ int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
                print_idlecoremap();
        } else if (!strcmp(argv[1], "all")) {
                print_allpids();
-       } else if (!strcmp(argv[1], "runnable")) {
-               dump_proclist(&proc_runnablelist);
+       } else if (!strcmp(argv[1], "sched")) {
+               sched_diag();
        } else if (!strcmp(argv[1], "resources")) {
                print_all_resources();
        } else if (!strcmp(argv[1], "pid")) {
@@ -297,6 +350,17 @@ int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
                print_proc_info(strtol(argv[2], 0, 0));
        } else if (!strcmp(argv[1], "unlock")) {
                if (argc != 3) {
+                       printk("Gimme lock address!  Me want lock address!.\n");
+                       return 1;
+               }
+               spinlock_t *lock = (spinlock_t*)strtol(argv[2], 0, 16);
+               if (!lock) {
+                       printk("Null address...\n");
+                       return 1;
+               }
+               spin_unlock(lock);
+       } else if (!strcmp(argv[1], "kill")) {
+               if (argc != 3) {
                        printk("Give me a pid number.\n");
                        return 1;
                }
@@ -305,9 +369,133 @@ int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
                        printk("No such proc\n");
                        return 1;
                }
-               spin_unlock_irqsave(&p->proc_lock);
-       } else if (!strcmp(argv[1], "kill")) {
-               if (argc != 3) {
+               enable_irqsave(&irq_state);
+               proc_destroy(p);
+               disable_irqsave(&irq_state);
+               proc_decref(p);
+       } else {
+               printk("Bad option\n");
+               return 1;
+       }
+       return 0;
+}
+
+int mon_exit(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       return -1;
+}
+
+int mon_kfunc(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       void (*func)(void *arg, ...);
+
+       if (argc < 2) {
+               printk("Usage: kfunc FUNCTION [arg1] [arg2] [etc]\n");
+               printk("Arguments must be in hex.  Can take 6 args.\n");
+               return 1;
+       }
+       func = (void*)get_symbol_addr(argv[1]);
+       if (!func) {
+               printk("Function not found.\n");
+               return 1;
+       }
+       /* Not elegant, but whatever.  maybe there's a better syntax, or we can do
+        * it with asm magic. */
+       switch (argc) {
+               case 2: /* have to fake one arg */
+                       func((void*)0);
+                       break;
+               case 3: /* the real first arg */
+                       func((void*)strtol(argv[2], 0, 16));
+                       break;
+               case 4:
+                       func((void*)strtol(argv[2], 0, 16),
+                                   strtol(argv[3], 0, 16));
+                       break;
+               case 5:
+                       func((void*)strtol(argv[2], 0, 16),
+                                   strtol(argv[3], 0, 16),
+                                   strtol(argv[4], 0, 16));
+                       break;
+               case 6:
+                       func((void*)strtol(argv[2], 0, 16),
+                                   strtol(argv[3], 0, 16),
+                                   strtol(argv[4], 0, 16),
+                                   strtol(argv[5], 0, 16));
+                       break;
+               case 7:
+                       func((void*)strtol(argv[2], 0, 16),
+                                   strtol(argv[3], 0, 16),
+                                   strtol(argv[4], 0, 16),
+                                   strtol(argv[5], 0, 16),
+                                   strtol(argv[6], 0, 16));
+                       break;
+               case 8:
+                       func((void*)strtol(argv[2], 0, 16),
+                                   strtol(argv[3], 0, 16),
+                                   strtol(argv[4], 0, 16),
+                                   strtol(argv[5], 0, 16),
+                                   strtol(argv[6], 0, 16),
+                                   strtol(argv[7], 0, 16));
+                       break;
+               default:
+                       printk("Bad number of arguments.\n");
+                       return -1;
+       }
+       return 0;
+}
+
+/* Sending a vcoreid forces an event and an IPI/notification */
+int mon_notify(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       struct proc *p;
+       uint32_t vcoreid;
+       struct event_msg msg = {0};
+
+       if (argc < 3) {
+               printk("Usage: notify PID NUM [VCOREID]\n");
+               return 1;
+       }
+       p = pid2proc(strtol(argv[1], 0, 0));
+       if (!p) {
+               printk("No such proc\n");
+               return 1;
+       }
+       msg.ev_type = strtol(argv[2], 0, 0);
+       if (argc == 4) {
+               vcoreid = strtol(argv[3], 0, 0);
+               /* This will go to the private mbox */
+               post_vcore_event(p, &msg, vcoreid, EVENT_VCORE_PRIVATE);
+               proc_notify(p, vcoreid);
+       } else {
+               /* o/w, try and do what they want */
+               send_kernel_event(p, &msg, 0);
+       }
+       proc_decref(p);
+       return 0;
+}
+
+/* Micro-benchmarky Measurements.  This is really fragile code that probably
+ * won't work perfectly, esp as the kernel evolves. */
+int mon_measure(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       uint64_t begin = 0, diff = 0;
+       uint32_t end_refcnt = 0;
+       int8_t irq_state = 0;
+
+       if (argc < 2) {
+               printk("Usage: measure OPTION\n");
+               printk("\tkill PID : kill proc PID\n");
+               printk("\tpreempt PID : preempt proc PID (no delay)\n");
+               printk("\tpreempt PID [pcore] : preempt PID's pcore (no delay)\n");
+               printk("\tpreempt-warn PID : warn-preempt proc PID (pending)\n");
+               printk("\tpreempt-warn PID [pcore] : warn-preempt proc PID's pcore\n");
+               printk("\tpreempt-raw PID : raw-preempt proc PID\n");
+               printk("\tpreempt-raw PID [pcore] : raw-preempt proc PID's pcore\n");
+               return 1;
+       }
+       if (!strcmp(argv[1], "kill")) {
+               if (argc < 3) {
                        printk("Give me a pid number.\n");
                        return 1;
                }
@@ -316,11 +504,266 @@ int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
                        printk("No such proc\n");
                        return 1;
                }
+               begin = start_timing();
+#ifdef CONFIG_APPSERVER
+               printk("Warning: this will be inaccurate due to the appserver.\n");
+               end_refcnt = kref_refcnt(&p->p_kref) - p->procinfo->num_vcores - 1;
+#endif /* CONFIG_APPSERVER */
+               enable_irqsave(&irq_state);
                proc_destroy(p);
+               disable_irqsave(&irq_state);
+               proc_decref(p);
+#ifdef CONFIG_APPSERVER
+               /* Won't be that accurate, since it's not actually going through the
+                * __proc_free() path. */
+               spin_on(kref_refcnt(&p->p_kref) != end_refcnt); 
+#else
+               /* this is a little ghetto. it's not fully free yet, but we are also
+                * slowing it down by messing with it, esp with the busy waiting on a
+                * hyperthreaded core. */
+               spin_on(p->env_cr3);
+#endif /* CONFIG_APPSERVER */
+               /* No noticeable difference using stop_timing instead of read_tsc() */
+               diff = stop_timing(begin);
+       } else if (!strcmp(argv[1], "preempt")) {
+               if (argc < 3) {
+                       printk("Give me a pid number.\n");
+                       return 1;
+               }
+               struct proc *p = pid2proc(strtol(argv[2], 0, 0));
+               if (!p) {
+                       printk("No such proc\n");
+                       return 1;
+               }
+               if (argc == 4) { /* single core being preempted, warned but no delay */
+                       uint32_t pcoreid = strtol(argv[3], 0, 0);
+                       begin = start_timing();
+                       if (proc_preempt_core(p, pcoreid, 1000000)) {
+                               __sched_put_idle_core(p, pcoreid);
+                               /* done when unmapped (right before abandoning) */
+                               spin_on(p->procinfo->pcoremap[pcoreid].valid);
+                       } else {
+                               printk("Core %d was not mapped to proc\n", pcoreid);
+                       }
+                       diff = stop_timing(begin);
+               } else { /* preempt all cores, warned but no delay */
+                       end_refcnt = kref_refcnt(&p->p_kref) - p->procinfo->num_vcores;
+                       begin = start_timing();
+                       proc_preempt_all(p, 1000000);
+                       /* a little ghetto, implies no one is using p */
+                       spin_on(kref_refcnt(&p->p_kref) != end_refcnt);
+                       diff = stop_timing(begin);
+               }
+               proc_decref(p);
+       } else if (!strcmp(argv[1], "preempt-warn")) {
+               if (argc < 3) {
+                       printk("Give me a pid number.\n");
+                       return 1;
+               }
+               struct proc *p = pid2proc(strtol(argv[2], 0, 0));
+               if (!p) {
+                       printk("No such proc\n");
+                       return 1;
+               }
+               printk("Careful: if this hangs, then the process isn't responding.\n");
+               if (argc == 4) { /* single core being preempted-warned */
+                       uint32_t pcoreid = strtol(argv[3], 0, 0);
+                       spin_lock(&p->proc_lock);
+                       uint32_t vcoreid = p->procinfo->pcoremap[pcoreid].vcoreid;
+                       if (!p->procinfo->pcoremap[pcoreid].valid) {
+                               printk("Pick a mapped pcore\n");
+                               spin_unlock(&p->proc_lock);
+                               return 1;
+                       }
+                       begin = start_timing();
+                       __proc_preempt_warn(p, vcoreid, 1000000); // 1 sec
+                       spin_unlock(&p->proc_lock);
+                       /* done when unmapped (right before abandoning) */
+                       spin_on(p->procinfo->pcoremap[pcoreid].valid);
+                       diff = stop_timing(begin);
+               } else { /* preempt-warn all cores */
+                       printk("Warning, this won't work if they can't yield their "
+                              "last vcore, will stop at 1!\n");
+                       spin_lock(&p->proc_lock);
+                       begin = start_timing();
+                       __proc_preempt_warnall(p, 1000000);
+                       spin_unlock(&p->proc_lock);
+                       /* target cores do the unmapping / changing of the num_vcores */
+                       spin_on(p->procinfo->num_vcores > 1);
+                       diff = stop_timing(begin);
+               }
+               proc_decref(p);
+       } else if (!strcmp(argv[1], "preempt-raw")) {
+               if (argc < 3) {
+                       printk("Give me a pid number.\n");
+                       return 1;
+               }
+               struct proc *p = pid2proc(strtol(argv[2], 0, 0));
+               if (!p) {
+                       printk("No such proc\n");
+                       return 1;
+               }
+               if (argc == 4) { /* single core preempted, no warning or waiting */
+                       uint32_t pcoreid = strtol(argv[3], 0, 0);
+                       spin_lock(&p->proc_lock);
+                       if (!p->procinfo->pcoremap[pcoreid].valid) {
+                               printk("Pick a mapped pcore\n");
+                               spin_unlock(&p->proc_lock);
+                               return 1;
+                       }
+                       begin = start_timing();
+                       __proc_preempt_core(p, pcoreid);
+                       if (!p->procinfo->num_vcores)
+                               __proc_set_state(p, PROC_RUNNABLE_M);
+                       spin_unlock(&p->proc_lock);
+                       /* ghetto, since the ksched should be calling all of this */
+                       __sched_put_idle_core(p, pcoreid);
+                       /* done when unmapped (right before abandoning) */
+                       spin_on(p->procinfo->pcoremap[pcoreid].valid);
+                       diff = stop_timing(begin);
+               } else { /* preempt all cores, no warning or waiting */
+                       spin_lock(&p->proc_lock);
+                       uint32_t pc_arr[p->procinfo->num_vcores];
+                       uint32_t num_revoked;
+                       end_refcnt = kref_refcnt(&p->p_kref) - p->procinfo->num_vcores;
+                       begin = start_timing();
+                       num_revoked = __proc_preempt_all(p, pc_arr);
+                       __proc_set_state(p, PROC_RUNNABLE_M);
+                       spin_unlock(&p->proc_lock);
+                       if (num_revoked)
+                               __sched_put_idle_cores(p, pc_arr, num_revoked);
+                       /* a little ghetto, implies no one else is using p */
+                       spin_on(kref_refcnt(&p->p_kref) != end_refcnt);
+                       diff = stop_timing(begin);
+               }
+               proc_decref(p);
        } else {
                printk("Bad option\n");
                return 1;
        }
+       printk("[Tired Giraffe Accent] Took %llu usec (%llu nsec) to finish.\n",
+              tsc2usec(diff), tsc2nsec(diff));
+       return 0;
+}
+
+/* Used in various debug locations.  Not a kernel API or anything. */
+bool mon_verbose_trace = FALSE;
+
+int mon_trace(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       int core;
+       if (argc < 2) {
+               printk("Usage: trace OPTION\n");
+               printk("\tsyscall start [silent] [pid]: starts tracing\n");
+               printk("\tsyscall stop: stops tracing, prints if it was silent\n");
+               printk("\tcoretf COREID: prints PC, -1 for all cores, verbose => TF\n");
+               printk("\tpcpui [type [coreid]]: runs pcpui trace ring handlers\n");
+               printk("\tpcpui-reset [noclear]: resets/clears pcpui trace ring\n");
+               printk("\tverbose: toggles verbosity, depends on trace command\n");
+               return 1;
+       }
+       if (!strcmp(argv[1], "syscall")) {
+               if (argc < 3) {
+                       printk("Need a start or stop.\n");
+                       return 1;
+               }
+               if (!strcmp(argv[2], "start")) {
+                       bool all = TRUE;
+                       bool silent = FALSE;
+                       struct proc *p = NULL;
+                       if (argc >= 4) {
+                               silent = (bool)strtol(argv[3], 0, 0);
+                       }
+                       if (argc >= 5) {
+                               all = FALSE;
+                               p = pid2proc(strtol(argv[4], 0, 0));
+                               if (!p) {
+                                       printk("No such process\n");
+                                       return 1;
+                               }
+                       }
+                       systrace_start(silent);
+                       if (systrace_reg(all, p))
+                               printk("No room to trace more processes\n");
+               } else if (!strcmp(argv[2], "stop")) {
+                       /* Stop and print for all processes */
+                       systrace_stop();
+                       systrace_print(TRUE, 0);
+                       systrace_clear_buffer();
+               }
+       } else if (!strcmp(argv[1], "coretf")) {
+               if (argc != 3) {
+                       printk("Need a coreid, fool.\n");
+                       return 1;
+               }
+               core = strtol(argv[2], 0, 0);
+               if (core < 0) {
+                       printk("Sending NMIs to all cores:\n");
+                       for (int i = 0; i < num_cpus; i++)
+                               send_nmi(i);
+               } else {
+                       printk("Sending NMI core %d:\n", core);
+                       if (core >= num_cpus) {
+                               printk("No such core!  Maybe it's in another cell...\n");
+                               return 1;
+                       }
+                       send_nmi(core);
+               }
+               udelay(1000000);
+       } else if (!strcmp(argv[1], "pcpui")) {
+               int pcpui_type, pcpui_coreid;
+               if (argc >= 3)
+                       pcpui_type = strtol(argv[2], 0, 0);
+               else
+                       pcpui_type = 0;
+               printk("\nRunning PCPUI Trace Ring handlers for type %d\n", pcpui_type);
+               if (argc >= 4) {
+                       pcpui_coreid = strtol(argv[3], 0, 0); 
+                       pcpui_tr_foreach(pcpui_coreid, pcpui_type);
+               } else {
+                       pcpui_tr_foreach_all(pcpui_type);
+               }
+       } else if (!strcmp(argv[1], "pcpui-reset")) {
+               if (argc >= 3) {
+                       printk("\nResetting all PCPUI Trace Rings\n");
+                       pcpui_tr_reset_all();
+               } else {
+                       printk("\nResetting and clearing all PCPUI Trace Rings\n");
+                       pcpui_tr_reset_and_clear_all();
+               }
+       } else if (!strcmp(argv[1], "verbose")) {
+               if (mon_verbose_trace) {
+                       printk("Turning trace verbosity off\n");
+                       mon_verbose_trace = FALSE;
+               } else {
+                       printk("Turning trace verbosity on\n");
+                       mon_verbose_trace = TRUE;
+               }
+       } else if (!strcmp(argv[1], "opt2")) {
+               if (argc != 3) {
+                       printk("ERRRRRRRRRR.\n");
+                       return 1;
+               }
+               print_proc_info(strtol(argv[2], 0, 0));
+       } else {
+               printk("Bad option\n");
+               return 1;
+       }
+       return 0;
+}
+
+int mon_monitor(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       if (argc < 2) {
+               printk("Usage: monitor COREID\n");
+               return 1;
+       }
+       uint32_t core = strtol(argv[1], 0, 0);
+       if (core >= num_cpus) {
+               printk("No such core!  Maybe it's in another cell...\n");
+               return 1;
+       }
+       send_kernel_message(core, __run_mon, 0, 0, 0, KMSG_ROUTINE);
        return 0;
 }
 
@@ -329,7 +772,7 @@ int mon_procinfo(int argc, char *NTS *NT COUNT(argc) argv, trapframe_t *tf)
 #define WHITESPACE "\t\r\n "
 #define MAXARGS 16
 
-static int runcmd(char *NTS real_buf, trapframe_t *tf) {
+static int runcmd(char *NTS real_buf, struct hw_trapframe *hw_tf) {
        char * buf = NTEXPAND(real_buf);
        int argc;
        char *NTS argv[MAXARGS];
@@ -362,25 +805,154 @@ static int runcmd(char *NTS real_buf, trapframe_t *tf) {
                return 0;
        for (i = 0; i < NCOMMANDS; i++) {
                if (strcmp(argv[0], commands[i].name) == 0)
-                       return commands[i].func(argc, argv, tf);
+                       return commands[i].func(argc, argv, hw_tf);
        }
        cprintf("Unknown command '%s'\n", argv[0]);
        return 0;
 }
 
-void monitor(trapframe_t *tf) {
-       char *buf;
-
-       printk("Welcome to the ROS kernel monitor on core %d!\n", core_id());
+void monitor(struct hw_trapframe *hw_tf)
+{
+       #define MON_CMD_LENGTH 256
+       char buf[MON_CMD_LENGTH];
+       int cnt;
+       int coreid = core_id_early();
+
+       /* they are always disabled, since we have this irqsave lock */
+       if (irq_is_enabled())
+               printk("Entering Nanwan's Dungeon on Core %d (Ints on):\n", coreid);
+       else
+               printk("Entering Nanwan's Dungeon on Core %d (Ints off):\n", coreid);
        printk("Type 'help' for a list of commands.\n");
 
-       if (tf != NULL)
-               print_trapframe(tf);
+       if (hw_tf != NULL)
+               print_trapframe(hw_tf);
 
        while (1) {
-               buf = readline("K> ");
-               if (buf != NULL)
-                       if (runcmd(buf, tf) < 0)
+               /* on occasion, the kernel monitor can migrate (like if you run
+                * something that blocks / syncs and wakes up on another core) */
+               cmb();
+               cnt = readline(buf, MON_CMD_LENGTH, "ROS(Core %d)> ", core_id_early());
+               if (cnt > 0) {
+                       buf[cnt] = 0;
+                       if (runcmd(buf, hw_tf) < 0)
                                break;
+               }
+       }
+}
+
+int mon_fs(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       /* this assumes one mounted FS at the NS root */
+       struct super_block *sb;
+       struct file *file;
+       struct inode *inode;
+       struct dentry *dentry;
+       if (argc < 2) {
+               printk("Usage: fs OPTION\n");
+               printk("\topen: show all open files\n");
+               printk("\tinodes: show all inodes\n");
+               printk("\tdentries [lru|prune]: show all dentries, opt LRU/prune\n");
+               printk("\tls DIR: print the dir tree starting with DIR\n");
+               printk("\tpid: proc PID's fs crap placeholder\n");
+               return 1;
+       }
+       if (!strcmp(argv[1], "open")) {
+               printk("Open Files:\n----------------------------\n");
+               TAILQ_FOREACH(sb, &super_blocks, s_list) {
+                       printk("Superblock for %s\n", sb->s_name);
+                       TAILQ_FOREACH(file, &sb->s_files, f_list)
+                               printk("File: %p, %s, Refs: %d, Drefs: %d, Irefs: %d\n", file,
+                                      file_name(file), kref_refcnt(&file->f_kref),
+                                      kref_refcnt(&file->f_dentry->d_kref),
+                                      kref_refcnt(&file->f_dentry->d_inode->i_kref));
+               }
+       } else if (!strcmp(argv[1], "inodes")) {
+               printk("Mounted FS Inodes:\n----------------------------\n");
+               TAILQ_FOREACH(sb, &super_blocks, s_list) {
+                       printk("Superblock for %s\n", sb->s_name);
+                       TAILQ_FOREACH(inode, &sb->s_inodes, i_sb_list) {
+                               printk("Inode: %p, Refs: %d, Nlinks: %d, Size(B): %d\n",
+                                      inode, kref_refcnt(&inode->i_kref), inode->i_nlink,
+                                      inode->i_size);
+                               TAILQ_FOREACH(dentry, &inode->i_dentry, d_alias)
+                                       printk("\t%s: Dentry: %p, Refs: %d\n",
+                                              dentry->d_name.name, dentry,
+                                              kref_refcnt(&dentry->d_kref));
+                       }
+               }
+       } else if (!strcmp(argv[1], "dentries")) {
+               printk("Dentry Cache:\n----------------------------\n");
+               TAILQ_FOREACH(sb, &super_blocks, s_list) {
+                       printk("Superblock for %s\n", sb->s_name);
+                       printk("DENTRY     FLAGS      REFCNT NAME\n");
+                       printk("--------------------------------\n");
+                       /* Hash helper */
+                       void print_dcache_entry(void *item)
+                       {
+                               struct dentry *d_i = (struct dentry*)item;
+                               printk("%p %p %02d     %s\n", d_i, d_i->d_flags,
+                                      kref_refcnt(&d_i->d_kref), d_i->d_name.name);
+                       }
+                       hash_for_each(sb->s_dcache, print_dcache_entry);
+               }
+               if (argc < 3)
+                       return 0;
+               if (!strcmp(argv[2], "lru")) {
+                       printk("LRU lists:\n");
+                       TAILQ_FOREACH(sb, &super_blocks, s_list) {
+                               printk("Superblock for %s\n", sb->s_name);
+                               TAILQ_FOREACH(dentry, &sb->s_lru_d, d_lru)
+                                       printk("Dentry: %p, Name: %s\n", dentry,
+                                              dentry->d_name.name);
+                       }
+               } else if (!strcmp(argv[2], "prune")) {
+                       printk("Pruning unused dentries\n");
+                       TAILQ_FOREACH(sb, &super_blocks, s_list)
+                               dcache_prune(sb, FALSE);
+               }
+       } else if (!strcmp(argv[1], "ls")) {
+               if (argc != 3) {
+                       printk("Give me a dir.\n");
+                       return 1;
+               }
+               if (argv[2][0] != '/') {
+                       printk("Dear fellow giraffe lover, Use absolute paths.\n");
+                       return 1;
+               }
+               ls_dash_r(argv[2]);
+               /* whatever.  placeholder. */
+       } else if (!strcmp(argv[1], "pid")) {
+               if (argc != 3) {
+                       printk("Give me a pid number.\n");
+                       return 1;
+               }
+               /* whatever.  placeholder. */
+       } else {
+               printk("Bad option\n");
+               return 1;
+       }
+       return 0;
+}
+
+int mon_bb(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       char *l_argv[3] = {"", "busybox", "ash"};
+       return mon_bin_run(3, l_argv, hw_tf);
+}
+
+int mon_alarm(int argc, char **argv, struct hw_trapframe *hw_tf)
+{
+       if (argc < 2) {
+               printk("Usage: alarm OPTION\n");
+               printk("\tpcpu: print full alarm tchains from every core\n");
+               return 1;
        }
+       if (!strcmp(argv[1], "pcpu")) {
+               print_pcpu_chains();
+       } else {
+               printk("Bad option\n");
+               return 1;
+       }
+       return 0;
 }