zircon/kernel/vm/pmm.cc - fuchsia - Git at Google

 // Copyright 2016 The Fuchsia Authors
 // Copyright (c) 2014 Travis Geiselbrecht
 //
 // Use of this source code is governed by a MIT-style
 // license that can be found in the LICENSE file or at
 // https://opensource.org/licenses/MIT

 #include "vm/pmm.h"

 #include <assert.h>
 #include <inttypes.h>
 #include <lib/boot-options/boot-options.h>
 #include <lib/console.h>
 #include <lib/counters.h>
 #include <lib/ktrace.h>
 #include <platform.h>
 #include <pow2.h>
 #include <stdlib.h>
 #include <string.h>
 #include <trace.h>
 #include <zircon/errors.h>
 #include <zircon/time.h>
 #include <zircon/types.h>

 #include <new>

 #include <fbl/auto_lock.h>
 #include <fbl/intrusive_double_list.h>
 #include <kernel/mp.h>
 #include <kernel/mutex.h>
 #include <kernel/timer.h>
 #include <ktl/algorithm.h>
 #include <lk/init.h>
 #include <vm/bootalloc.h>
 #include <vm/physmap.h>
 #include <vm/pmm_checker.h>
 #include <vm/scanner.h>
 #include <vm/vm.h>

 #include "pmm_arena.h"
 #include "pmm_node.h"
 #include "vm_priv.h"

 #include <ktl/enforce.h>

 #define LOCAL_TRACE VM_GLOBAL_TRACE(0)

 // Number of bytes available in the PMM after kernel init, but before userspace init.
 KCOUNTER(boot_memory_bytes, "boot.memory.post_init_free_bytes")

 // The (currently) one and only pmm node
 static PmmNode pmm_node;

 // Singleton
 static PhysicalPageBorrowingConfig ppb_config;

 // Check that if random should wait is requested that this is a debug build with assertions as it is
 // currently assumed that enabling this in a non-debug build would be a mistake that should be
 // caught.
 static void pmm_check_alloc_random_should_wait(uint level) {
   if (gBootOptions->pmm_alloc_random_should_wait) {
     ASSERT(DEBUG_ASSERT_IMPLEMENTED);
     printf("pmm: alloc-random-should-wait enabled\n");
   }
 }
 LK_INIT_HOOK(pmm_check_alloc_random_should_wait, &pmm_check_alloc_random_should_wait,
              LK_INIT_LEVEL_LAST)

 static void pmm_fill_free_pages(uint level) { pmm_node.FillFreePagesAndArm(); }
 LK_INIT_HOOK(pmm_fill, &pmm_fill_free_pages, LK_INIT_LEVEL_VM)

 vm_page_t* paddr_to_vm_page(paddr_t addr) { return pmm_node.PaddrToPage(addr); }

 zx_status_t pmm_add_arena(const pmm_arena_info_t* info) { return pmm_node.AddArena(info); }

 size_t pmm_num_arenas() { return pmm_node.NumArenas(); }

 zx_status_t pmm_get_arena_info(size_t count, uint64_t i, pmm_arena_info_t* buffer,
                                size_t buffer_size) {
   return pmm_node.GetArenaInfo(count, i, buffer, buffer_size);
 }

 zx_status_t pmm_alloc_page(uint alloc_flags, paddr_t* pa) {
   VM_KTRACE_DURATION(3, "pmm_alloc_page");
   return pmm_node.AllocPage(alloc_flags, nullptr, pa);
 }

 zx_status_t pmm_alloc_page(uint alloc_flags, vm_page_t** page) {
   VM_KTRACE_DURATION(3, "pmm_alloc_page");
   return pmm_node.AllocPage(alloc_flags, page, nullptr);
 }

 zx_status_t pmm_alloc_page(uint alloc_flags, vm_page_t** page, paddr_t* pa) {
   VM_KTRACE_DURATION(3, "pmm_alloc_page");
   return pmm_node.AllocPage(alloc_flags, page, pa);
 }

 zx_status_t pmm_alloc_pages(size_t count, uint alloc_flags, list_node* list) {
   VM_KTRACE_DURATION(3, "pmm_alloc_pages");
   return pmm_node.AllocPages(count, alloc_flags, list);
 }

 zx_status_t pmm_alloc_range(paddr_t address, size_t count, list_node* list) {
   VM_KTRACE_DURATION(3, "pmm_alloc_range");
   return pmm_node.AllocRange(address, count, list);
 }

 zx_status_t pmm_alloc_contiguous(size_t count, uint alloc_flags, uint8_t alignment_log2,
                                  paddr_t* pa, list_node* list) {
   VM_KTRACE_DURATION(3, "pmm_alloc_contiguous");
   // if we're called with a single page, just fall through to the regular allocation routine
   if (unlikely(count == 1 && alignment_log2 <= PAGE_SIZE_SHIFT)) {
     vm_page_t* page;
     zx_status_t status = pmm_node.AllocPage(alloc_flags, &page, pa);
     if (status != ZX_OK) {
       return status;
     }
     list_add_tail(list, &page->queue_node);
     return ZX_OK;
   }

   return pmm_node.AllocContiguous(count, alloc_flags, alignment_log2, pa, list);
 }

 void pmm_begin_loan(list_node* page_list) {
   VM_KTRACE_DURATION(3, "pmm_begin_loan");
   pmm_node.BeginLoan(page_list);
 }

 void pmm_cancel_loan(paddr_t address, size_t count) {
   VM_KTRACE_DURATION(3, "pmm_cancel_loan");
   pmm_node.CancelLoan(address, count);
 }

 void pmm_end_loan(paddr_t address, size_t count, list_node* page_list) {
   VM_KTRACE_DURATION(3, "pmm_end_loan");
   pmm_node.EndLoan(address, count, page_list);
 }

 void pmm_delete_lender(paddr_t address, size_t count) {
   VM_KTRACE_DURATION(3, "pmm_delete_lender");
   pmm_node.DeleteLender(address, count);
 }

 bool pmm_is_loaned(vm_page_t* page) {
   VM_KTRACE_DURATION(3, "pmm_is_loaned");
   return pmm_node.IsLoaned(page);
 }

 void pmm_free(list_node* list) {
   VM_KTRACE_DURATION(3, "pmm_free");
   pmm_node.FreeList(list);
 }

 void pmm_free_page(vm_page* page) {
   VM_KTRACE_DURATION(3, "pmm_free_page");
   pmm_node.FreePage(page);
 }

 uint64_t pmm_count_free_pages() { return pmm_node.CountFreePages(); }

 uint64_t pmm_count_loaned_free_pages() { return pmm_node.CountLoanedFreePages(); }

 uint64_t pmm_count_loaned_used_pages() { return pmm_node.CountLoanedNotFreePages(); }

 uint64_t pmm_count_loaned_pages() { return pmm_node.CountLoanedPages(); }

 uint64_t pmm_count_loan_cancelled_pages() { return pmm_node.CountLoanCancelledPages(); }

 uint64_t pmm_count_total_bytes() { return pmm_node.CountTotalBytes(); }

 PageQueues* pmm_page_queues() { return pmm_node.GetPageQueues(); }

 Evictor* pmm_evictor() { return pmm_node.GetEvictor(); }

 PhysicalPageBorrowingConfig* pmm_physical_page_borrowing_config() {
   // singleton
   return &ppb_config;
 }

 zx_status_t pmm_init_reclamation(const uint64_t* watermarks, uint8_t watermark_count,
                                  uint64_t debounce, void* context,
                                  mem_avail_state_updated_callback_t callback) {
   return pmm_node.InitReclamation(watermarks, watermark_count, debounce, context, callback);
 }

 zx_status_t pmm_wait_till_should_retry_single_alloc(const Deadline& deadline) {
   return pmm_node.WaitTillShouldRetrySingleAlloc(deadline);
 }

 void pmm_stop_returning_should_wait() { pmm_node.StopReturningShouldWait(); }

 void pmm_checker_check_all_free_pages() { pmm_node.CheckAllFreePages(); }

 #if __has_feature(address_sanitizer)
 void pmm_asan_poison_all_free_pages() { pmm_node.PoisonAllFreePages(); }
 #endif

 int64_t pmm_get_alloc_failed_count() { return PmmNode::get_alloc_failed_count(); }

 bool pmm_has_alloc_failed_no_mem() { return PmmNode::has_alloc_failed_no_mem(); }

 static void pmm_checker_enable(size_t fill_size, PmmChecker::Action action) {
   // We might be changing the fill size.  If we increase the fill size while the checker is active,
   // we might spuriously assert so disable the checker.
   pmm_node.DisableChecker();

   // Enable filling of pages going forward.
   pmm_node.EnableFreePageFilling(fill_size, action);

   // From this point on, pages will be filled when they are freed.  However, the free list may still
   // have a lot of unfilled pages so make a pass over them and fill them all.
   pmm_node.FillFreePagesAndArm();

   // All free pages have now been filled with |fill_size| and the checker is armed.
 }

 static void pmm_checker_disable() { pmm_node.DisableChecker(); }

 static bool pmm_checker_is_enabled() { return pmm_node.Checker()->IsArmed(); }

 static void pmm_checker_print_status() { pmm_node.Checker()->PrintStatus(stdout); }

 void pmm_checker_init_from_cmdline() {
   bool enabled = gBootOptions->pmm_checker_enabled;
   if (enabled) {
     size_t fill_size = gBootOptions->pmm_checker_fill_size;
     if (!PmmChecker::IsValidFillSize(fill_size)) {
       printf("PMM: value from %s is invalid (%lu), using PAGE_SIZE instead\n",
              kPmmCheckerFillSizeName.data(), fill_size);
       fill_size = PAGE_SIZE;
     }

     PmmChecker::Action action = PmmChecker::DefaultAction;
     const char* const action_string = gBootOptions->pmm_checker_action.data();
     if (action_string != nullptr) {
       if (auto opt_action = PmmChecker::ActionFromString(action_string)) {
         action = opt_action.value();
       } else {
         printf("PMM: value from %s is invalid (\"%s\"), using \"%s\" instead\n",
                kPmmCheckerActionName.data(), action_string, PmmChecker::ActionToString(action));
       }
     }

     pmm_node.EnableFreePageFilling(fill_size, action);
   }
 }

 static void pmm_dump_timer(Timer* t, zx_time_t now, void*) {
   zx_time_t deadline = zx_time_add_duration(now, ZX_SEC(1));
   t->SetOneshot(deadline, &pmm_dump_timer, nullptr);
   pmm_node.DumpFree();
 }

 LK_INIT_HOOK(
     pmm_boot_memory,
     [](unsigned int /*level*/) {
       // Track the amount of free memory available in the PMM after kernel init, but before
       // userspace starts.
       //
       // We record this in a kcounter to be tracked by build infrastructure over time.
       dprintf(INFO, "Free memory after kernel init: %" PRIu64 " bytes.\n",
               pmm_node.CountFreePages() * PAGE_SIZE);
       boot_memory_bytes.Set(pmm_node.CountFreePages() * PAGE_SIZE);
     },
     LK_INIT_LEVEL_USER - 1)

 static Timer dump_free_mem_timer;

 // Allocate memory until the PMM returns ZX_ERR_NO_MEMORY, then free it all.
 //
 // The idea is to cause the system to briefly dip into an out of memory state.
 static int cmd_oom_dip() {
   list_node list = LIST_INITIAL_VALUE(list);
   while (pmm_node.AllocPages(1, 0, &list) != ZX_ERR_NO_MEMORY) {
   }
   pmm_node.FreeList(&list);
   printf("allocated until ZX_ERR_NO_MEMORY (and then freed)\n");
   return ZX_OK;
 }

 static int cmd_oom(int argc, const cmd_args* argv) {
   uint64_t rate = 0;
   bool hard = false;
   if (argc > 2) {
     if (!strcmp(argv[2].str, "dip")) {
       return cmd_oom_dip();
     }

     if (!strcmp(argv[2].str, "signal")) {
       pmm_node.DebugMemAvailStateCallback(0);
       return ZX_OK;
     }
     if (!strcmp(argv[2].str, "hard")) {
       hard = true;
     } else {
       rate = strtoul(argv[2].str, nullptr, 0) * 1024 * 1024 / PAGE_SIZE;
     }
   }

   // When we reach the oom state the kernel may 'try harder' to reclaim memory and prevent us from
   // hitting oom. To avoid this we disable the scanner to prevent additional memory from becoming
   // classified as evictable, and then evict anything that is already considered.
   printf("Disabling VM scanner\n");
   scanner_push_disable_count();
   uint64_t pages_evicted = pmm_evictor()->EvictOneShotSynchronous(
       UINT64_MAX, Evictor::EvictionLevel::IncludeNewest, Evictor::Output::NoPrint);
   if (pages_evicted > 0) {
     printf("Leaked %" PRIu64 " pages from eviction\n", pages_evicted);
   }

   uint64_t pages_till_oom;
   // In case we are racing with someone freeing pages we will leak in a loop until we are sure
   // we have hit the oom state.
   while ((pages_till_oom = pmm_node.DebugNumPagesTillMemState(0)) > 0) {
     list_node list = LIST_INITIAL_VALUE(list);
     if (rate > 0) {
       uint64_t pages_leaked = 0;
       while (pages_leaked < pages_till_oom) {
         uint64_t alloc_pages = ktl::min(rate, pages_till_oom - pages_leaked);
         if (pmm_node.AllocPages(alloc_pages, 0, &list) == ZX_OK) {
           pages_leaked += alloc_pages;
           printf("Leaked %lu pages\n", pages_leaked);
         }
         Thread::Current::SleepRelative(ZX_SEC(1));
       }
     } else {
       if (pmm_node.AllocPages(pages_till_oom, 0, &list) == ZX_OK) {
         printf("Leaked %lu pages\n", pages_till_oom);
       }
     }
     // Ignore any errors under the assumption we had a racy allocation and try again next time
     // around the loop.
   }

   if (hard) {
     printf("Continuing to leak pages forever\n");
     // Keep leaking as fast possible.
     while (true) {
       vm_page_t* p;
       pmm_alloc_page(0, &p);
     }
   }

   return ZX_OK;
 }

 static int cmd_usage(const char* cmd_name, bool is_panic) {
   printf("usage:\n");
   printf("%s dump                                     : dump pmm info \n", cmd_name);
   if (!is_panic) {
     printf("%s free                                     : periodically dump free mem count\n",
            cmd_name);
     printf(
         "%s oom [<rate>]                             : leak memory until oom is triggered, "
         "optionally specify the rate at which to leak (in MB per second)\n",
         cmd_name);
     printf("%s oom dip                                  : allocate until no mem, then free\n",
            cmd_name);
     printf(
         "%s oom hard                                 : leak memory aggressively and keep on "
         "leaking\n",
         cmd_name);
     printf(
         "%s oom signal                               : trigger oom signal without leaking "
         "memory\n",
         cmd_name);
     printf("%s mem_avail_state info                     : dump memory availability state info\n",
            cmd_name);
     printf(
         "%s mem_avail_state [step] <state> [<nsecs>] : allocate memory to go to memstate "
         "<state>, hold the state for <nsecs> (10s by default). Only works if going to <state> "
         "from current state requires allocating memory, can't free up pre-allocated memory. In "
         "optional [step] mode, allocation pauses for 1 second at each intermediate memory "
         "availability state until <state> is reached.\n",
         cmd_name);
     printf("%s drop_user_pt                             : drop all user hardware page tables\n",
            cmd_name);
     printf("%s checker status                           : prints the status of the pmm checker\n",
            cmd_name);
     printf(
         "%s checker enable [<size>] [oops|panic]     : enables the pmm checker with optional "
         "fill size and optional action\n",
         cmd_name);
     printf("%s checker disable                          : disables the pmm checker\n", cmd_name);
     printf(
         "%s checker check                            : forces a check of all free pages in the "
         "pmm\n",
         cmd_name);
   }
   return ZX_ERR_INTERNAL;
 }

 static int cmd_pmm(int argc, const cmd_args* argv, uint32_t flags) {
   const bool is_panic = flags & CMD_FLAG_PANIC;
   const char* name = argv[0].str;

   if (argc < 2) {
     printf("not enough arguments\n");
     return cmd_usage(name, is_panic);
   }

   if (!strcmp(argv[1].str, "dump")) {
     pmm_node.Dump(is_panic);
   } else if (is_panic) {
     // No other operations will work during a panic.
     printf("Only the \"arenas\" command is available during a panic.\n");
     return cmd_usage(name, is_panic);
   } else if (!strcmp(argv[1].str, "free")) {
     static bool show_mem = false;

     if (!show_mem) {
       printf("pmm free: issue the same command to stop.\n");
       zx_time_t deadline = zx_time_add_duration(current_time(), ZX_SEC(1));
       const TimerSlack slack{ZX_MSEC(20), TIMER_SLACK_CENTER};
       const Deadline slackDeadline(deadline, slack);
       dump_free_mem_timer.Set(slackDeadline, &pmm_dump_timer, nullptr);
       show_mem = true;
     } else {
       dump_free_mem_timer.Cancel();
       show_mem = false;
     }
   } else if (!strcmp(argv[1].str, "oom")) {
     if (argc > 3) {
       return cmd_usage(name, is_panic);
     }
     return cmd_oom(argc, argv);
   } else if (!strcmp(argv[1].str, "mem_avail_state")) {
     if (argc < 3) {
       return cmd_usage(name, is_panic);
     }
     if (!strcmp(argv[2].str, "info")) {
       pmm_node.DumpMemAvailState();
     } else {
       bool step = false;
       int index = 2;
       if (!strcmp(argv[2].str, "step")) {
         step = true;
         index++;
       }

       uint8_t state = static_cast<uint8_t>(argv[index++].u);
       if (state > pmm_node.DebugMaxMemAvailState()) {
         printf("Invalid memstate %u. Specify a value between 0 and %u.\n", state,
                pmm_node.DebugMaxMemAvailState());
         return cmd_usage(name, is_panic);
       }

       uint64_t pages_to_alloc, pages_to_free = 0;
       list_node list = LIST_INITIAL_VALUE(list);

       if (step) {
         uint8_t s = pmm_node.DebugMaxMemAvailState();
         while (true) {
           // In case we are racing with someone freeing pages we will leak in a loop until we are
           // sure we have hit the required memory availability state.
           uint64_t pages_allocated = 0;
           while ((pages_to_alloc = pmm_node.DebugNumPagesTillMemState(s)) > 0) {
             if (pmm_node.AllocPages(pages_to_alloc, 0, &list) == ZX_OK) {
               printf("Leaked %lu pages\n", pages_to_alloc);
               pages_allocated += pages_to_alloc;
             }
           }
           pages_to_free += pages_allocated;
           if (s == state) {
             break;
           }
           s--;
           if (pages_allocated) {
             printf("Sleeping for 1 second...\n");
             Thread::Current::SleepRelative(ZX_SEC(1));
           }
         }
       } else {
         // In case we are racing with someone freeing pages we will leak in a loop until we are
         // sure we have hit the required memory availability state.
         while ((pages_to_alloc = pmm_node.DebugNumPagesTillMemState(state)) > 0) {
           if (pmm_node.AllocPages(pages_to_alloc, 0, &list) == ZX_OK) {
             printf("Leaked %lu pages\n", pages_to_alloc);
             pages_to_free += pages_to_alloc;
           }
         }
       }

       if (pages_to_free > 0) {
         uint64_t nsecs = 10;
         if (argc > index) {
           nsecs = argv[index].u;
         }
         printf("Sleeping for %lu seconds...\n", nsecs);
         Thread::Current::SleepRelative(ZX_SEC(nsecs));
         pmm_node.FreeList(&list);
         printf("Freed %lu pages\n", pages_to_free);
       }
     }
   } else if (!strcmp(argv[1].str, "drop_user_pt")) {
     VmAspace::DropAllUserPageTables();
   } else if (!strcmp(argv[1].str, "checker")) {
     if (argc < 3 || argc > 5) {
       return cmd_usage(name, is_panic);
     }
     if (!strcmp(argv[2].str, "status")) {
       pmm_checker_print_status();
     } else if (!strcmp(argv[2].str, "enable")) {
       size_t fill_size = PAGE_SIZE;
       PmmChecker::Action action = PmmChecker::DefaultAction;
       if (argc >= 4) {
         fill_size = argv[3].u;
         if (!PmmChecker::IsValidFillSize(fill_size)) {
           printf(
               "error: fill size must be a multiple of 8 and be between 8 and PAGE_SIZE, "
               "inclusive\n");
           return ZX_ERR_INTERNAL;
         }
       }
       if (argc == 5) {
         if (auto opt_action = PmmChecker::ActionFromString(argv[4].str)) {
           action = opt_action.value();
         } else {
           printf("error: invalid action\n");
           return ZX_ERR_INTERNAL;
         }
       }
       pmm_checker_enable(fill_size, action);
       // No need to print status as enabling automatically prints status.
     } else if (!strcmp(argv[2].str, "disable")) {
       pmm_checker_disable();
       pmm_checker_print_status();
     } else if (!strcmp(argv[2].str, "check")) {
       if (!pmm_checker_is_enabled()) {
         printf("error: pmm checker is not enabled\n");
         return ZX_ERR_INTERNAL;
       }
       printf("checking all free pages...\n");
       pmm_checker_check_all_free_pages();
       printf("done\n");
     } else {
       return cmd_usage(name, is_panic);
     }
   } else {
     printf("unknown command\n");

     return cmd_usage(name, is_panic);
   }

   return ZX_OK;
 }

 void pmm_print_physical_page_borrowing_stats() {
   uint64_t free_pages = pmm_count_free_pages();
   uint64_t loaned_free_pages = pmm_count_loaned_free_pages();
   uint64_t loaned_pages = pmm_count_loaned_pages();
   uint64_t loan_cancelled_pages = pmm_count_loan_cancelled_pages();
   uint64_t total_bytes = pmm_count_total_bytes();
   uint64_t used_loaned_pages = pmm_count_loaned_used_pages();
   printf(
       "PPB stats:\n"
       "  free pages: %" PRIu64 " free MiB: %" PRIu64
       "\n"
       "  loaned free pages: %" PRIu64 " loaned free MiB: %" PRIu64
       "\n"
       "  loaned pages: %" PRIu64 " loaned MiB: %" PRIu64
       "\n"
       "  used loaned pages: %" PRIu64 " used loaned MiB: %" PRIu64
       "\n"
       "  loan cancelled pages: %" PRIu64 " loan cancelled MIB: %" PRIu64
       "\n"
       "  total physical pages: %" PRIu64 " total MiB: %" PRIu64 "\n",
       free_pages, free_pages * PAGE_SIZE / MB, loaned_free_pages,
       loaned_free_pages * PAGE_SIZE / MB, loaned_pages, loaned_pages * PAGE_SIZE / MB,
       used_loaned_pages, used_loaned_pages * PAGE_SIZE / MB, loan_cancelled_pages,
       loan_cancelled_pages * PAGE_SIZE / MB, total_bytes / PAGE_SIZE, total_bytes / MB);
 }

 STATIC_COMMAND_START
 STATIC_COMMAND_MASKED("pmm", "physical memory manager", &cmd_pmm, CMD_AVAIL_ALWAYS)
 STATIC_COMMAND_END(pmm)
	// Copyright 2016 The Fuchsia Authors
	// Copyright (c) 2014 Travis Geiselbrecht
	//
	// Use of this source code is governed by a MIT-style
	// license that can be found in the LICENSE file or at
	// https://opensource.org/licenses/MIT

	#include "vm/pmm.h"

	#include <assert.h>
	#include <inttypes.h>
	#include <lib/boot-options/boot-options.h>
	#include <lib/console.h>
	#include <lib/counters.h>
	#include <lib/ktrace.h>
	#include <platform.h>
	#include <pow2.h>
	#include <stdlib.h>
	#include <string.h>
	#include <trace.h>
	#include <zircon/errors.h>
	#include <zircon/time.h>
	#include <zircon/types.h>

	#include <new>

	#include <fbl/auto_lock.h>
	#include <fbl/intrusive_double_list.h>
	#include <kernel/mp.h>
	#include <kernel/mutex.h>
	#include <kernel/timer.h>
	#include <ktl/algorithm.h>
	#include <lk/init.h>
	#include <vm/bootalloc.h>
	#include <vm/physmap.h>
	#include <vm/pmm_checker.h>
	#include <vm/scanner.h>
	#include <vm/vm.h>

	#include "pmm_arena.h"
	#include "pmm_node.h"
	#include "vm_priv.h"

	#include <ktl/enforce.h>

	#define LOCAL_TRACE VM_GLOBAL_TRACE(0)

	// Number of bytes available in the PMM after kernel init, but before userspace init.
	KCOUNTER(boot_memory_bytes, "boot.memory.post_init_free_bytes")

	// The (currently) one and only pmm node
	static PmmNode pmm_node;

	// Singleton
	static PhysicalPageBorrowingConfig ppb_config;

	// Check that if random should wait is requested that this is a debug build with assertions as it is
	// currently assumed that enabling this in a non-debug build would be a mistake that should be
	// caught.
	static void pmm_check_alloc_random_should_wait(uint level) {
	if (gBootOptions->pmm_alloc_random_should_wait) {
	ASSERT(DEBUG_ASSERT_IMPLEMENTED);
	printf("pmm: alloc-random-should-wait enabled\n");
	}
	}
	LK_INIT_HOOK(pmm_check_alloc_random_should_wait, &pmm_check_alloc_random_should_wait,
	LK_INIT_LEVEL_LAST)

	static void pmm_fill_free_pages(uint level) { pmm_node.FillFreePagesAndArm(); }
	LK_INIT_HOOK(pmm_fill, &pmm_fill_free_pages, LK_INIT_LEVEL_VM)

	vm_page_t* paddr_to_vm_page(paddr_t addr) { return pmm_node.PaddrToPage(addr); }

	zx_status_t pmm_add_arena(const pmm_arena_info_t* info) { return pmm_node.AddArena(info); }

	size_t pmm_num_arenas() { return pmm_node.NumArenas(); }

	zx_status_t pmm_get_arena_info(size_t count, uint64_t i, pmm_arena_info_t* buffer,
	size_t buffer_size) {
	return pmm_node.GetArenaInfo(count, i, buffer, buffer_size);
	}

	zx_status_t pmm_alloc_page(uint alloc_flags, paddr_t* pa) {
	VM_KTRACE_DURATION(3, "pmm_alloc_page");
	return pmm_node.AllocPage(alloc_flags, nullptr, pa);
	}

	zx_status_t pmm_alloc_page(uint alloc_flags, vm_page_t** page) {
	VM_KTRACE_DURATION(3, "pmm_alloc_page");
	return pmm_node.AllocPage(alloc_flags, page, nullptr);
	}

	zx_status_t pmm_alloc_page(uint alloc_flags, vm_page_t** page, paddr_t* pa) {
	VM_KTRACE_DURATION(3, "pmm_alloc_page");
	return pmm_node.AllocPage(alloc_flags, page, pa);
	}

	zx_status_t pmm_alloc_pages(size_t count, uint alloc_flags, list_node* list) {
	VM_KTRACE_DURATION(3, "pmm_alloc_pages");
	return pmm_node.AllocPages(count, alloc_flags, list);
	}

	zx_status_t pmm_alloc_range(paddr_t address, size_t count, list_node* list) {
	VM_KTRACE_DURATION(3, "pmm_alloc_range");
	return pmm_node.AllocRange(address, count, list);
	}

	zx_status_t pmm_alloc_contiguous(size_t count, uint alloc_flags, uint8_t alignment_log2,
	paddr_t* pa, list_node* list) {
	VM_KTRACE_DURATION(3, "pmm_alloc_contiguous");
	// if we're called with a single page, just fall through to the regular allocation routine
	if (unlikely(count == 1 && alignment_log2 <= PAGE_SIZE_SHIFT)) {
	vm_page_t* page;
	zx_status_t status = pmm_node.AllocPage(alloc_flags, &page, pa);
	if (status != ZX_OK) {
	return status;
	}
	list_add_tail(list, &page->queue_node);
	return ZX_OK;
	}

	return pmm_node.AllocContiguous(count, alloc_flags, alignment_log2, pa, list);
	}

	void pmm_begin_loan(list_node* page_list) {
	VM_KTRACE_DURATION(3, "pmm_begin_loan");
	pmm_node.BeginLoan(page_list);
	}

	void pmm_cancel_loan(paddr_t address, size_t count) {
	VM_KTRACE_DURATION(3, "pmm_cancel_loan");
	pmm_node.CancelLoan(address, count);
	}

	void pmm_end_loan(paddr_t address, size_t count, list_node* page_list) {
	VM_KTRACE_DURATION(3, "pmm_end_loan");
	pmm_node.EndLoan(address, count, page_list);
	}

	void pmm_delete_lender(paddr_t address, size_t count) {
	VM_KTRACE_DURATION(3, "pmm_delete_lender");
	pmm_node.DeleteLender(address, count);
	}

	bool pmm_is_loaned(vm_page_t* page) {
	VM_KTRACE_DURATION(3, "pmm_is_loaned");
	return pmm_node.IsLoaned(page);
	}

	void pmm_free(list_node* list) {
	VM_KTRACE_DURATION(3, "pmm_free");
	pmm_node.FreeList(list);
	}

	void pmm_free_page(vm_page* page) {
	VM_KTRACE_DURATION(3, "pmm_free_page");
	pmm_node.FreePage(page);
	}

	uint64_t pmm_count_free_pages() { return pmm_node.CountFreePages(); }

	uint64_t pmm_count_loaned_free_pages() { return pmm_node.CountLoanedFreePages(); }

	uint64_t pmm_count_loaned_used_pages() { return pmm_node.CountLoanedNotFreePages(); }

	uint64_t pmm_count_loaned_pages() { return pmm_node.CountLoanedPages(); }

	uint64_t pmm_count_loan_cancelled_pages() { return pmm_node.CountLoanCancelledPages(); }

	uint64_t pmm_count_total_bytes() { return pmm_node.CountTotalBytes(); }

	PageQueues* pmm_page_queues() { return pmm_node.GetPageQueues(); }

	Evictor* pmm_evictor() { return pmm_node.GetEvictor(); }

	PhysicalPageBorrowingConfig* pmm_physical_page_borrowing_config() {
	// singleton
	return &ppb_config;
	}

	zx_status_t pmm_init_reclamation(const uint64_t* watermarks, uint8_t watermark_count,
	uint64_t debounce, void* context,
	mem_avail_state_updated_callback_t callback) {
	return pmm_node.InitReclamation(watermarks, watermark_count, debounce, context, callback);
	}

	zx_status_t pmm_wait_till_should_retry_single_alloc(const Deadline& deadline) {
	return pmm_node.WaitTillShouldRetrySingleAlloc(deadline);
	}

	void pmm_stop_returning_should_wait() { pmm_node.StopReturningShouldWait(); }

	void pmm_checker_check_all_free_pages() { pmm_node.CheckAllFreePages(); }

	#if __has_feature(address_sanitizer)
	void pmm_asan_poison_all_free_pages() { pmm_node.PoisonAllFreePages(); }
	#endif

	int64_t pmm_get_alloc_failed_count() { return PmmNode::get_alloc_failed_count(); }

	bool pmm_has_alloc_failed_no_mem() { return PmmNode::has_alloc_failed_no_mem(); }

	static void pmm_checker_enable(size_t fill_size, PmmChecker::Action action) {
	// We might be changing the fill size. If we increase the fill size while the checker is active,
	// we might spuriously assert so disable the checker.
	pmm_node.DisableChecker();

	// Enable filling of pages going forward.
	pmm_node.EnableFreePageFilling(fill_size, action);

	// From this point on, pages will be filled when they are freed. However, the free list may still
	// have a lot of unfilled pages so make a pass over them and fill them all.
	pmm_node.FillFreePagesAndArm();

	// All free pages have now been filled with \|fill_size\| and the checker is armed.
	}

	static void pmm_checker_disable() { pmm_node.DisableChecker(); }

	static bool pmm_checker_is_enabled() { return pmm_node.Checker()->IsArmed(); }

	static void pmm_checker_print_status() { pmm_node.Checker()->PrintStatus(stdout); }

	void pmm_checker_init_from_cmdline() {
	bool enabled = gBootOptions->pmm_checker_enabled;
	if (enabled) {
	size_t fill_size = gBootOptions->pmm_checker_fill_size;
	if (!PmmChecker::IsValidFillSize(fill_size)) {
	printf("PMM: value from %s is invalid (%lu), using PAGE_SIZE instead\n",
	kPmmCheckerFillSizeName.data(), fill_size);
	fill_size = PAGE_SIZE;
	}

	PmmChecker::Action action = PmmChecker::DefaultAction;
	const char* const action_string = gBootOptions->pmm_checker_action.data();
	if (action_string != nullptr) {
	if (auto opt_action = PmmChecker::ActionFromString(action_string)) {
	action = opt_action.value();
	} else {
	printf("PMM: value from %s is invalid (\"%s\"), using \"%s\" instead\n",
	kPmmCheckerActionName.data(), action_string, PmmChecker::ActionToString(action));
	}
	}

	pmm_node.EnableFreePageFilling(fill_size, action);
	}
	}

	static void pmm_dump_timer(Timer* t, zx_time_t now, void*) {
	zx_time_t deadline = zx_time_add_duration(now, ZX_SEC(1));
	t->SetOneshot(deadline, &pmm_dump_timer, nullptr);
	pmm_node.DumpFree();
	}

	LK_INIT_HOOK(
	pmm_boot_memory,
	[](unsigned int /level/) {
	// Track the amount of free memory available in the PMM after kernel init, but before
	// userspace starts.
	//
	// We record this in a kcounter to be tracked by build infrastructure over time.
	dprintf(INFO, "Free memory after kernel init: %" PRIu64 " bytes.\n",
	pmm_node.CountFreePages() * PAGE_SIZE);
	boot_memory_bytes.Set(pmm_node.CountFreePages() * PAGE_SIZE);
	},
	LK_INIT_LEVEL_USER - 1)

	static Timer dump_free_mem_timer;

	// Allocate memory until the PMM returns ZX_ERR_NO_MEMORY, then free it all.
	//
	// The idea is to cause the system to briefly dip into an out of memory state.
	static int cmd_oom_dip() {
	list_node list = LIST_INITIAL_VALUE(list);
	while (pmm_node.AllocPages(1, 0, &list) != ZX_ERR_NO_MEMORY) {
	}
	pmm_node.FreeList(&list);
	printf("allocated until ZX_ERR_NO_MEMORY (and then freed)\n");
	return ZX_OK;
	}

	static int cmd_oom(int argc, const cmd_args* argv) {
	uint64_t rate = 0;
	bool hard = false;
	if (argc > 2) {
	if (!strcmp(argv[2].str, "dip")) {
	return cmd_oom_dip();
	}

	if (!strcmp(argv[2].str, "signal")) {
	pmm_node.DebugMemAvailStateCallback(0);
	return ZX_OK;
	}
	if (!strcmp(argv[2].str, "hard")) {
	hard = true;
	} else {
	rate = strtoul(argv[2].str, nullptr, 0) * 1024 * 1024 / PAGE_SIZE;
	}
	}

	// When we reach the oom state the kernel may 'try harder' to reclaim memory and prevent us from
	// hitting oom. To avoid this we disable the scanner to prevent additional memory from becoming
	// classified as evictable, and then evict anything that is already considered.
	printf("Disabling VM scanner\n");
	scanner_push_disable_count();
	uint64_t pages_evicted = pmm_evictor()->EvictOneShotSynchronous(
	UINT64_MAX, Evictor::EvictionLevel::IncludeNewest, Evictor::Output::NoPrint);
	if (pages_evicted > 0) {
	printf("Leaked %" PRIu64 " pages from eviction\n", pages_evicted);
	}

	uint64_t pages_till_oom;
	// In case we are racing with someone freeing pages we will leak in a loop until we are sure
	// we have hit the oom state.
	while ((pages_till_oom = pmm_node.DebugNumPagesTillMemState(0)) > 0) {
	list_node list = LIST_INITIAL_VALUE(list);
	if (rate > 0) {
	uint64_t pages_leaked = 0;
	while (pages_leaked < pages_till_oom) {
	uint64_t alloc_pages = ktl::min(rate, pages_till_oom - pages_leaked);
	if (pmm_node.AllocPages(alloc_pages, 0, &list) == ZX_OK) {
	pages_leaked += alloc_pages;
	printf("Leaked %lu pages\n", pages_leaked);
	}
	Thread::Current::SleepRelative(ZX_SEC(1));
	}
	} else {
	if (pmm_node.AllocPages(pages_till_oom, 0, &list) == ZX_OK) {
	printf("Leaked %lu pages\n", pages_till_oom);
	}
	}
	// Ignore any errors under the assumption we had a racy allocation and try again next time
	// around the loop.
	}

	if (hard) {
	printf("Continuing to leak pages forever\n");
	// Keep leaking as fast possible.
	while (true) {
	vm_page_t* p;
	pmm_alloc_page(0, &p);
	}
	}

	return ZX_OK;
	}

	static int cmd_usage(const char* cmd_name, bool is_panic) {
	printf("usage:\n");
	printf("%s dump : dump pmm info \n", cmd_name);
	if (!is_panic) {
	printf("%s free : periodically dump free mem count\n",
	cmd_name);
	printf(
	"%s oom [<rate>] : leak memory until oom is triggered, "
	"optionally specify the rate at which to leak (in MB per second)\n",
	cmd_name);
	printf("%s oom dip : allocate until no mem, then free\n",
	cmd_name);
	printf(
	"%s oom hard : leak memory aggressively and keep on "
	"leaking\n",
	cmd_name);
	printf(
	"%s oom signal : trigger oom signal without leaking "
	"memory\n",
	cmd_name);
	printf("%s mem_avail_state info : dump memory availability state info\n",
	cmd_name);
	printf(
	"%s mem_avail_state [step] <state> [<nsecs>] : allocate memory to go to memstate "
	"<state>, hold the state for <nsecs> (10s by default). Only works if going to <state> "
	"from current state requires allocating memory, can't free up pre-allocated memory. In "
	"optional [step] mode, allocation pauses for 1 second at each intermediate memory "
	"availability state until <state> is reached.\n",
	cmd_name);
	printf("%s drop_user_pt : drop all user hardware page tables\n",
	cmd_name);
	printf("%s checker status : prints the status of the pmm checker\n",
	cmd_name);
	printf(
	"%s checker enable [<size>] [oops\|panic] : enables the pmm checker with optional "
	"fill size and optional action\n",
	cmd_name);
	printf("%s checker disable : disables the pmm checker\n", cmd_name);
	printf(
	"%s checker check : forces a check of all free pages in the "
	"pmm\n",
	cmd_name);
	}
	return ZX_ERR_INTERNAL;
	}

	static int cmd_pmm(int argc, const cmd_args* argv, uint32_t flags) {
	const bool is_panic = flags & CMD_FLAG_PANIC;
	const char* name = argv[0].str;

	if (argc < 2) {
	printf("not enough arguments\n");
	return cmd_usage(name, is_panic);
	}

	if (!strcmp(argv[1].str, "dump")) {
	pmm_node.Dump(is_panic);
	} else if (is_panic) {
	// No other operations will work during a panic.
	printf("Only the \"arenas\" command is available during a panic.\n");
	return cmd_usage(name, is_panic);
	} else if (!strcmp(argv[1].str, "free")) {
	static bool show_mem = false;

	if (!show_mem) {
	printf("pmm free: issue the same command to stop.\n");
	zx_time_t deadline = zx_time_add_duration(current_time(), ZX_SEC(1));
	const TimerSlack slack{ZX_MSEC(20), TIMER_SLACK_CENTER};
	const Deadline slackDeadline(deadline, slack);
	dump_free_mem_timer.Set(slackDeadline, &pmm_dump_timer, nullptr);
	show_mem = true;
	} else {
	dump_free_mem_timer.Cancel();
	show_mem = false;
	}
	} else if (!strcmp(argv[1].str, "oom")) {
	if (argc > 3) {
	return cmd_usage(name, is_panic);
	}
	return cmd_oom(argc, argv);
	} else if (!strcmp(argv[1].str, "mem_avail_state")) {
	if (argc < 3) {
	return cmd_usage(name, is_panic);
	}
	if (!strcmp(argv[2].str, "info")) {
	pmm_node.DumpMemAvailState();
	} else {
	bool step = false;
	int index = 2;
	if (!strcmp(argv[2].str, "step")) {
	step = true;
	index++;
	}

	uint8_t state = static_cast<uint8_t>(argv[index++].u);
	if (state > pmm_node.DebugMaxMemAvailState()) {
	printf("Invalid memstate %u. Specify a value between 0 and %u.\n", state,
	pmm_node.DebugMaxMemAvailState());
	return cmd_usage(name, is_panic);
	}

	uint64_t pages_to_alloc, pages_to_free = 0;
	list_node list = LIST_INITIAL_VALUE(list);

	if (step) {
	uint8_t s = pmm_node.DebugMaxMemAvailState();
	while (true) {
	// In case we are racing with someone freeing pages we will leak in a loop until we are
	// sure we have hit the required memory availability state.
	uint64_t pages_allocated = 0;
	while ((pages_to_alloc = pmm_node.DebugNumPagesTillMemState(s)) > 0) {
	if (pmm_node.AllocPages(pages_to_alloc, 0, &list) == ZX_OK) {
	printf("Leaked %lu pages\n", pages_to_alloc);
	pages_allocated += pages_to_alloc;
	}
	}
	pages_to_free += pages_allocated;
	if (s == state) {
	break;
	}
	s--;
	if (pages_allocated) {
	printf("Sleeping for 1 second...\n");
	Thread::Current::SleepRelative(ZX_SEC(1));
	}
	}
	} else {
	// In case we are racing with someone freeing pages we will leak in a loop until we are
	// sure we have hit the required memory availability state.
	while ((pages_to_alloc = pmm_node.DebugNumPagesTillMemState(state)) > 0) {
	if (pmm_node.AllocPages(pages_to_alloc, 0, &list) == ZX_OK) {
	printf("Leaked %lu pages\n", pages_to_alloc);
	pages_to_free += pages_to_alloc;
	}
	}
	}

	if (pages_to_free > 0) {
	uint64_t nsecs = 10;
	if (argc > index) {
	nsecs = argv[index].u;
	}
	printf("Sleeping for %lu seconds...\n", nsecs);
	Thread::Current::SleepRelative(ZX_SEC(nsecs));
	pmm_node.FreeList(&list);
	printf("Freed %lu pages\n", pages_to_free);
	}
	}
	} else if (!strcmp(argv[1].str, "drop_user_pt")) {
	VmAspace::DropAllUserPageTables();
	} else if (!strcmp(argv[1].str, "checker")) {
	if (argc < 3 \|\| argc > 5) {
	return cmd_usage(name, is_panic);
	}
	if (!strcmp(argv[2].str, "status")) {
	pmm_checker_print_status();
	} else if (!strcmp(argv[2].str, "enable")) {
	size_t fill_size = PAGE_SIZE;
	PmmChecker::Action action = PmmChecker::DefaultAction;
	if (argc >= 4) {
	fill_size = argv[3].u;
	if (!PmmChecker::IsValidFillSize(fill_size)) {
	printf(
	"error: fill size must be a multiple of 8 and be between 8 and PAGE_SIZE, "
	"inclusive\n");
	return ZX_ERR_INTERNAL;
	}
	}
	if (argc == 5) {
	if (auto opt_action = PmmChecker::ActionFromString(argv[4].str)) {
	action = opt_action.value();
	} else {
	printf("error: invalid action\n");
	return ZX_ERR_INTERNAL;
	}
	}
	pmm_checker_enable(fill_size, action);
	// No need to print status as enabling automatically prints status.
	} else if (!strcmp(argv[2].str, "disable")) {
	pmm_checker_disable();
	pmm_checker_print_status();
	} else if (!strcmp(argv[2].str, "check")) {
	if (!pmm_checker_is_enabled()) {
	printf("error: pmm checker is not enabled\n");
	return ZX_ERR_INTERNAL;
	}
	printf("checking all free pages...\n");
	pmm_checker_check_all_free_pages();
	printf("done\n");
	} else {
	return cmd_usage(name, is_panic);
	}
	} else {
	printf("unknown command\n");

	return cmd_usage(name, is_panic);
	}

	return ZX_OK;
	}

	void pmm_print_physical_page_borrowing_stats() {
	uint64_t free_pages = pmm_count_free_pages();
	uint64_t loaned_free_pages = pmm_count_loaned_free_pages();
	uint64_t loaned_pages = pmm_count_loaned_pages();
	uint64_t loan_cancelled_pages = pmm_count_loan_cancelled_pages();
	uint64_t total_bytes = pmm_count_total_bytes();
	uint64_t used_loaned_pages = pmm_count_loaned_used_pages();
	printf(
	"PPB stats:\n"
	" free pages: %" PRIu64 " free MiB: %" PRIu64
	"\n"
	" loaned free pages: %" PRIu64 " loaned free MiB: %" PRIu64
	"\n"
	" loaned pages: %" PRIu64 " loaned MiB: %" PRIu64
	"\n"
	" used loaned pages: %" PRIu64 " used loaned MiB: %" PRIu64
	"\n"
	" loan cancelled pages: %" PRIu64 " loan cancelled MIB: %" PRIu64
	"\n"
	" total physical pages: %" PRIu64 " total MiB: %" PRIu64 "\n",
	free_pages, free_pages * PAGE_SIZE / MB, loaned_free_pages,
	loaned_free_pages * PAGE_SIZE / MB, loaned_pages, loaned_pages * PAGE_SIZE / MB,
	used_loaned_pages, used_loaned_pages * PAGE_SIZE / MB, loan_cancelled_pages,
	loan_cancelled_pages * PAGE_SIZE / MB, total_bytes / PAGE_SIZE, total_bytes / MB);
	}

	STATIC_COMMAND_START
	STATIC_COMMAND_MASKED("pmm", "physical memory manager", &cmd_pmm, CMD_AVAIL_ALWAYS)
	STATIC_COMMAND_END(pmm)