nuttx/mm/iob/iob_alloc.c

412 lines
12 KiB
C
Raw Normal View History

2014-06-03 20:41:34 +02:00
/****************************************************************************
* mm/iob/iob_alloc.c
2014-06-03 20:41:34 +02:00
*
* 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
2014-06-03 20:41:34 +02:00
*
* http://www.apache.org/licenses/LICENSE-2.0
2014-06-03 20:41:34 +02:00
*
* 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.
2014-06-03 20:41:34 +02:00
*
****************************************************************************/
/****************************************************************************
* Included Files
****************************************************************************/
#include <nuttx/config.h>
#include <assert.h>
#include <errno.h>
#include <nuttx/irq.h>
#include <nuttx/arch.h>
#include <nuttx/sched.h>
#ifdef CONFIG_IOB_ALLOC
# include <nuttx/kmalloc.h>
#endif
#include <nuttx/mm/iob.h>
2014-06-03 20:41:34 +02:00
#include "iob.h"
/****************************************************************************
* Private Functions
2014-06-03 20:41:34 +02:00
****************************************************************************/
static clock_t iob_allocwait_gettimeout(clock_t start, unsigned int timeout)
{
sclock_t tick;
tick = clock_systime_ticks() - start;
if (tick >= MSEC2TICK(timeout))
{
tick = 0;
}
else
{
tick = MSEC2TICK(timeout) - tick;
}
return tick;
}
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
/****************************************************************************
* Name: iob_alloc_committed
*
* Description:
* Allocate an I/O buffer by taking the buffer at the head of the committed
* list.
*
****************************************************************************/
static FAR struct iob_s *iob_alloc_committed(void)
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
{
FAR struct iob_s *iob = NULL;
irqstate_t flags;
/* We don't know what context we are called from so we use extreme measures
* to protect the committed list: We disable interrupts very briefly.
*/
flags = spin_lock_irqsave(&g_iob_lock);
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
/* Take the I/O buffer from the head of the committed list */
iob = g_iob_committed;
if (iob != NULL)
{
/* Remove the I/O buffer from the committed list */
g_iob_committed = iob->io_flink;
/* Put the I/O buffer in a known state */
iob->io_flink = NULL; /* Not in a chain */
iob->io_len = 0; /* Length of the data in the entry */
iob->io_offset = 0; /* Offset to the beginning of data */
iob->io_pktlen = 0; /* Total length of the packet */
}
spin_unlock_irqrestore(&g_iob_lock, flags);
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
return iob;
}
/****************************************************************************
* Name: iob_allocwait
*
* Description:
* Allocate an I/O buffer, waiting if necessary. This function cannot be
* called from any interrupt level logic.
*
****************************************************************************/
static FAR struct iob_s *iob_allocwait(bool throttled, unsigned int timeout)
{
FAR struct iob_s *iob;
irqstate_t flags;
FAR sem_t *sem;
clock_t start;
int ret = OK;
#if CONFIG_IOB_THROTTLE > 0
/* Select the semaphore count to check. */
sem = (throttled ? &g_throttle_sem : &g_iob_sem);
#else
sem = &g_iob_sem;
#endif
/* The following must be atomic; interrupt must be disabled so that there
* is no conflict with interrupt level I/O buffer allocations. This is
* not as bad as it sounds because interrupts will be re-enabled while
* we are waiting for I/O buffers to become free.
*/
flags = enter_critical_section();
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
/* Try to get an I/O buffer. If successful, the semaphore count will be
* decremented atomically.
*/
start = clock_systime_ticks();
iob = iob_tryalloc(throttled);
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
while (ret == OK && iob == NULL)
{
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
/* If not successful, then the semaphore count was less than or equal
* to zero (meaning that there are no free buffers). We need to wait
* for an I/O buffer to be released and placed in the committed
* list.
*/
if (timeout == UINT_MAX)
{
ret = nxsem_wait_uninterruptible(sem);
}
else
{
ret = nxsem_tickwait_uninterruptible(sem,
iob_allocwait_gettimeout(start, timeout));
}
if (ret >= 0)
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
{
/* When we wake up from wait successfully, an I/O buffer was
* freed and we hold a count for one IOB.
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
*/
iob = iob_alloc_committed();
}
}
leave_critical_section(flags);
return iob;
}
#ifdef CONFIG_IOB_ALLOC
/****************************************************************************
* Name: iob_free_dynamic
*
* Description:
* Dummy free callback function, do nothing.
*
* Input Parameters:
* data -
*
****************************************************************************/
static void iob_free_dynamic(FAR void *data)
{
}
#endif
/****************************************************************************
* Public Functions
****************************************************************************/
/****************************************************************************
* Name: iob_timedalloc
*
* Description:
* Allocate an I/O buffer by taking the buffer at the head of the free list.
* This wait will be terminated when the specified timeout expires.
*
* Input Parameters:
* throttled - An indication of the IOB allocation is "throttled"
* timeout - Timeout value in milliseconds.
*
****************************************************************************/
FAR struct iob_s *iob_timedalloc(bool throttled, unsigned int timeout)
{
/* Were we called from the interrupt level? */
if (up_interrupt_context() || sched_idletask() || timeout == 0)
{
/* Yes, then try to allocate an I/O buffer without waiting */
return iob_tryalloc(throttled);
}
else
{
/* Then allocate an I/O buffer, waiting as necessary */
return iob_allocwait(throttled, timeout);
}
}
/****************************************************************************
* Name: iob_alloc
*
* Description:
* Allocate an I/O buffer by taking the buffer at the head of the free list.
*
****************************************************************************/
FAR struct iob_s *iob_alloc(bool throttled)
{
return iob_timedalloc(throttled, UINT_MAX);
}
/****************************************************************************
* Name: iob_tryalloc
*
* Description:
* Try to allocate an I/O buffer by taking the buffer at the head of the
* free list without waiting for a buffer to become free.
*
****************************************************************************/
FAR struct iob_s *iob_tryalloc(bool throttled)
{
FAR struct iob_s *iob;
irqstate_t flags;
#if CONFIG_IOB_THROTTLE > 0
FAR sem_t *sem;
#endif
#if CONFIG_IOB_THROTTLE > 0
/* Select the semaphore count to check. */
sem = (throttled ? &g_throttle_sem : &g_iob_sem);
#endif
/* We don't know what context we are called from so we use extreme measures
* to protect the free list: We disable interrupts very briefly.
*/
flags = spin_lock_irqsave(&g_iob_lock);
#if CONFIG_IOB_THROTTLE > 0
/* If there are free I/O buffers for this allocation */
if (sem->semcount > 0)
#endif
{
/* Take the I/O buffer from the head of the free list */
iob = g_iob_freelist;
There can be a failure in IOB allocation to some asynchronous behavior caused by the use of sem_post(). Consider this scenario: Task A holds an IOB.  There are no further IOBs.  The value of semcount is zero. Task B calls iob_alloc().  Since there are not IOBs, it calls sem_wait().  The v alue of semcount is now -1. Task A frees the IOB.  iob_free() adds the IOB to the free list and calls sem_post() this makes Task B ready to run and sets semcount to zero NOT 1.  There is one IOB in the free list and semcount is zero.  When Task B wakes up it would increment the sem_count back to the correct value. But an interrupt or another task runs occurs before Task B executes.  The interrupt or other tak takes the IOB off of the free list and decrements the semcount.  But since semcount is then < 0, this causes the assertion because that is an invalid state in the interrupt handler. So I think that the root cause is that there the asynchrony between incrementing the semcount. This change separates the list of IOBs: Currently there is only a free list of IOBs. The problem, I believe, is because of asynchronies due sem_post() post cause the semcount and the list content to become out of sync. This change adds a new 'committed' list: When there is a task waiting for an IOB, it will go into the committed list rather than the free list before the semaphore is posted. On the waiting side, when awakened from the semaphore wait, it will expect to find its IOB in the committed list, rather than free list. In this way, the content of the free list and the value of the semaphore count always remain in sync.
2017-05-16 19:03:35 +02:00
if (iob != NULL)
{
/* Remove the I/O buffer from the free list and decrement the
* counting semaphore(s) that tracks the number of available
* IOBs.
*/
g_iob_freelist = iob->io_flink;
/* Take a semaphore count. Note that we cannot do this in
* in the orthodox way by calling nxsem_wait() or nxsem_trywait()
* because this function may be called from an interrupt
* handler. Fortunately we know at at least one free buffer
* so a simple decrement is all that is needed.
*/
g_iob_sem.semcount--;
DEBUGASSERT(g_iob_sem.semcount >= 0);
#if CONFIG_IOB_THROTTLE > 0
/* The throttle semaphore is used to throttle the number of
* free buffers that are available. It is used to prevent
* the overrunning of the free buffer list. Please note that
* it can only be decremented to zero, which indicates no
* throttled buffers are available.
*/
if (g_throttle_sem.semcount > 0)
{
g_throttle_sem.semcount--;
}
#endif
spin_unlock_irqrestore(&g_iob_lock, flags);
/* Put the I/O buffer in a known state */
iob->io_flink = NULL; /* Not in a chain */
iob->io_len = 0; /* Length of the data in the entry */
iob->io_offset = 0; /* Offset to the beginning of data */
iob->io_pktlen = 0; /* Total length of the packet */
return iob;
}
}
spin_unlock_irqrestore(&g_iob_lock, flags);
return NULL;
}
#ifdef CONFIG_IOB_ALLOC
/****************************************************************************
* Name: iob_alloc_dynamic
*
* Description:
* Allocate an I/O buffer and playload from heap
*
* Input Parameters:
* size - The size of the io_data that is allocated.
*
* +---------+
* | IOB |
* | io_data |--+
* | buffer |<-+
* +---------+
*
****************************************************************************/
FAR struct iob_s *iob_alloc_dynamic(uint16_t size)
{
FAR struct iob_s *iob;
size_t alignsize;
alignsize = ROUNDUP(sizeof(struct iob_s), CONFIG_IOB_ALIGNMENT) + size;
iob = kmm_memalign(CONFIG_IOB_ALIGNMENT, alignsize);
if (iob)
{
iob->io_flink = NULL; /* Not in a chain */
iob->io_len = 0; /* Length of the data in the entry */
iob->io_offset = 0; /* Offset to the beginning of data */
iob->io_bufsize = size; /* Total length of the iob buffer */
iob->io_pktlen = 0; /* Total length of the packet */
iob->io_free = iob_free_dynamic; /* Customer free callback */
iob->io_data = (FAR uint8_t *)ROUNDUP((uintptr_t)(iob + 1),
CONFIG_IOB_ALIGNMENT);
}
return iob;
}
/****************************************************************************
* Name: iob_alloc_with_data
*
* Description:
* Allocate an I/O buffer from heap and attach the external payload
*
* Input Parameters:
* data - Make io_data point to a specific address, the caller is
* responsible for the memory management. The caller should
* ensure that the memory is not freed before the iob is freed.
*
* +---------+ +-->+--------+
* | IOB | | | data |
* | io_data |--+ +--------+
* +---------+
*
* size - The size of the data parameter
* free_cb - Notify the caller when the iob is freed. The caller can
* perform additional operations on the data before it is freed.
* The free_cb is called when the iob is freed.
*
****************************************************************************/
FAR struct iob_s *iob_alloc_with_data(FAR void *data, uint16_t size,
iob_free_cb_t free_cb)
{
FAR struct iob_s *iob;
DEBUGASSERT(free_cb != NULL);
iob = kmm_malloc(sizeof(struct iob_s));
if (iob)
{
iob->io_flink = NULL; /* Not in a chain */
iob->io_len = 0; /* Length of the data in the entry */
iob->io_offset = 0; /* Offset to the beginning of data */
iob->io_bufsize = size; /* Total length of the iob buffer */
iob->io_pktlen = 0; /* Total length of the packet */
iob->io_free = free_cb; /* Customer free callback */
iob->io_data = data;
}
return iob;
}
#endif