nuttx/mm/mm_heap/mm_sem.c

266 lines
7.8 KiB
C
Raw Normal View History

/****************************************************************************
* mm/mm_heap/mm_sem.c
*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership. The
* ASF licenses this file to you under the Apache License, Version 2.0 (the
* "License"); you may not use this file except in compliance with the
* License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the
* License for the specific language governing permissions and limitations
* under the License.
*
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include <nuttx/config.h>
#include <unistd.h>
#include <errno.h>
#include <assert.h>
#include <debug.h>
#include <nuttx/semaphore.h>
#include <nuttx/mm/mm.h>
#include "mm_heap/mm.h"
/****************************************************************************
* Pre-processor Definitions
****************************************************************************/
/* This is a special value that indicates that there is no holder of the
* semaphore. The valid range of PIDs is 0-32767 and any value outside of
* that range could be used (except -ESRCH which is a special return value
* from getpid())
*/
#define NO_HOLDER ((pid_t)-1)
/* Define MONITOR_MM_SEMAPHORE to enable semaphore state monitoring */
#ifdef MONITOR_MM_SEMAPHORE
# define msemerr _err
# define msemwarn _warn
# define mseminfo _info
#else
# define msemerr _none
# define msemwarn _none
# define mseminfo _none
#endif
/****************************************************************************
* Public Functions
****************************************************************************/
/****************************************************************************
* Name: mm_seminitialize
*
* Description:
* Initialize the MM mutex
*
****************************************************************************/
void mm_seminitialize(FAR struct mm_heap_s *heap)
{
/* Initialize the MM semaphore to one (to support one-at-a-time access to
* private data sets).
*/
_SEM_INIT(&heap->mm_semaphore, 0, 1);
heap->mm_holder = NO_HOLDER;
heap->mm_counts_held = 0;
}
/****************************************************************************
* Name: mm_trysemaphore
*
* Description:
* Try to take the MM mutex. This is called only from the OS in certain
* conditions when it is necessary to have exclusive access to the memory
* manager but it is impossible to wait on a semaphore (e.g., the idle
* process when it performs its background memory cleanup).
*
****************************************************************************/
int mm_trysemaphore(FAR struct mm_heap_s *heap)
{
pid_t my_pid = getpid();
int ret;
/* getpid() returns the task ID of the task at the head of the ready-to-
* run task list. mm_trysemaphore() may be called during context
* switches. There are certain situations during context switching when
* the OS data structures are in flux and where the current task (i.e.,
* the task at the head of the ready-to-run task list) is not actually
* running. Granting the semaphore access in that case is known to result
* in heap corruption as in the following failure scenario.
*
2018-12-27 16:32:47 +01:00
* ---------------------------- -------------------------------
* TASK A TASK B
* ---------------------------- -------------------------------
* Begins memory allocation.
* - Holder is set to TASK B
* <---- Task B preempted, Task A runs
* Task A exits
* - Current task set to Task B
* Free tcb and stack memory
* - Since holder is Task B,
* memory manager is re-
* entered, and
* - Heap is corrupted.
* ---------------------------- -------------------------------
*
* This is handled by getpid(): If the case where Task B is not actually
* running, then getpid() will return the special value -ESRCH. That will
* avoid taking the fatal 'if' logic and will fall through to use the
* 'else', albeit with a nonsensical PID value.
*/
if (my_pid < 0)
{
ret = my_pid;
goto errout;
}
2018-12-27 16:32:47 +01:00
/* Does the current task already hold the semaphore? Is the current
* task actually running?
*/
if (heap->mm_holder == my_pid)
{
2018-12-27 16:32:47 +01:00
/* Yes, just increment the number of references held by the current
* task.
*/
heap->mm_counts_held++;
Merged in masayuki2009/nuttx.nuttx/fix_smp_bugs (pull request #615) Fix SMP related bugs * sched/sched: Fix a deadlock in SMP mode Two months ago, I introduced sched_tasklist_lock() and sched_tasklist_unlock() to protect tasklists in SMP mode. Actually, this change works pretty well for HTTP audio streaming aging test with lc823450-xgevk. However, I found a deadlock in the scheduler when I tried similar aging tests with DVFS autonomous mode where CPU clock speed changed based on cpu load. In this case, call sequences were as follows; cpu1: sched_unlock()->sched_mergepending()->sched_addreadytorun()->up_cpu_pause() cpu0: sched_lock()->sched_mergepending() To avoid this deadlock, I added sched_tasklist_unlock() when calling up_cpu_pause() and sched_addreadytorun(). Also, added sched_tasklist_lock() after the call. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * libc: Add critical section in lib_filesem.c for SMP To set my_pid into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * mm: Add critical section in mm_sem.c for SMP To set my_pid into mm_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * net: Add critical section in net_lock.c for SMP To set my pid (me) into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> Approved-by: Gregory Nutt <gnutt@nuttx.org>
2018-03-20 13:34:38 +01:00
ret = OK;
}
else
{
/* Try to take the semaphore */
ret = _SEM_TRYWAIT(&heap->mm_semaphore);
if (ret < 0)
{
ret = _SEM_ERRVAL(ret);
goto errout;
}
2018-12-27 16:32:47 +01:00
/* We have it. Claim the heap for the current task and return */
heap->mm_holder = my_pid;
heap->mm_counts_held = 1;
Merged in masayuki2009/nuttx.nuttx/fix_smp_bugs (pull request #615) Fix SMP related bugs * sched/sched: Fix a deadlock in SMP mode Two months ago, I introduced sched_tasklist_lock() and sched_tasklist_unlock() to protect tasklists in SMP mode. Actually, this change works pretty well for HTTP audio streaming aging test with lc823450-xgevk. However, I found a deadlock in the scheduler when I tried similar aging tests with DVFS autonomous mode where CPU clock speed changed based on cpu load. In this case, call sequences were as follows; cpu1: sched_unlock()->sched_mergepending()->sched_addreadytorun()->up_cpu_pause() cpu0: sched_lock()->sched_mergepending() To avoid this deadlock, I added sched_tasklist_unlock() when calling up_cpu_pause() and sched_addreadytorun(). Also, added sched_tasklist_lock() after the call. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * libc: Add critical section in lib_filesem.c for SMP To set my_pid into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * mm: Add critical section in mm_sem.c for SMP To set my_pid into mm_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * net: Add critical section in net_lock.c for SMP To set my pid (me) into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> Approved-by: Gregory Nutt <gnutt@nuttx.org>
2018-03-20 13:34:38 +01:00
ret = OK;
}
Merged in masayuki2009/nuttx.nuttx/fix_smp_bugs (pull request #615) Fix SMP related bugs * sched/sched: Fix a deadlock in SMP mode Two months ago, I introduced sched_tasklist_lock() and sched_tasklist_unlock() to protect tasklists in SMP mode. Actually, this change works pretty well for HTTP audio streaming aging test with lc823450-xgevk. However, I found a deadlock in the scheduler when I tried similar aging tests with DVFS autonomous mode where CPU clock speed changed based on cpu load. In this case, call sequences were as follows; cpu1: sched_unlock()->sched_mergepending()->sched_addreadytorun()->up_cpu_pause() cpu0: sched_lock()->sched_mergepending() To avoid this deadlock, I added sched_tasklist_unlock() when calling up_cpu_pause() and sched_addreadytorun(). Also, added sched_tasklist_lock() after the call. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * libc: Add critical section in lib_filesem.c for SMP To set my_pid into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * mm: Add critical section in mm_sem.c for SMP To set my_pid into mm_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> * net: Add critical section in net_lock.c for SMP To set my pid (me) into fs_folder atomically in SMP mode, critical section API must be used. Signed-off-by: Masayuki Ishikawa <Masayuki.Ishikawa@jp.sony.com> Approved-by: Gregory Nutt <gnutt@nuttx.org>
2018-03-20 13:34:38 +01:00
errout:
return ret;
}
/****************************************************************************
* Name: mm_takesemaphore
*
* Description:
* Take the MM mutex. This is the normal action before all memory
* management actions.
*
****************************************************************************/
void mm_takesemaphore(FAR struct mm_heap_s *heap)
{
pid_t my_pid = getpid();
2018-12-27 16:32:47 +01:00
/* Does the current task already hold the semaphore? */
if (heap->mm_holder == my_pid)
{
2018-12-27 16:32:47 +01:00
/* Yes, just increment the number of references held by the current
* task.
*/
heap->mm_counts_held++;
}
else
{
int ret;
/* Take the semaphore (perhaps waiting) */
mseminfo("PID=%d taking\n", my_pid);
do
{
ret = _SEM_WAIT(&heap->mm_semaphore);
/* The only case that an error should occur here is if the wait
* was awakened by a signal.
*/
if (ret < 0)
{
ret = _SEM_ERRVAL(ret);
DEBUGASSERT(ret == -EINTR || ret == -ECANCELED);
}
}
while (ret == -EINTR);
/* We have it (or some awful, unexpected error occurred). Claim
2018-12-27 16:32:47 +01:00
* the semaphore for the current task and return.
*/
heap->mm_holder = my_pid;
heap->mm_counts_held = 1;
}
mseminfo("Holder=%d count=%d\n", heap->mm_holder,
heap->mm_counts_held);
}
/****************************************************************************
* Name: mm_givesemaphore
*
* Description:
* Release the MM mutex when it is not longer needed.
*
****************************************************************************/
void mm_givesemaphore(FAR struct mm_heap_s *heap)
{
2018-12-27 16:32:47 +01:00
/* The current task should be holding at least one reference to the
* semaphore.
*/
DEBUGASSERT(heap->mm_holder == getpid());
2018-12-27 16:32:47 +01:00
/* Does the current task hold multiple references to the semaphore */
if (heap->mm_counts_held > 1)
{
/* Yes, just release one count and return */
heap->mm_counts_held--;
mseminfo("Holder=%d count=%d\n", heap->mm_holder,
heap->mm_counts_held);
}
else
{
2018-12-27 16:32:47 +01:00
/* Nope, this is the last reference held by the current task. */
mseminfo("PID=%d giving\n", getpid());
2014-05-28 22:09:58 +02:00
heap->mm_holder = NO_HOLDER;
heap->mm_counts_held = 0;
DEBUGVERIFY(_SEM_POST(&heap->mm_semaphore));
}
}