Merge branch 'bugfix/intr_free' into 'master'

fix(intr): fix the logic for allocating and freeing interrupts.

See merge request !1124
This commit is contained in:
Ivan Grokhotkov 2017-09-07 17:45:18 +08:00
commit b9dfb437ab
4 changed files with 251 additions and 71 deletions

View file

@ -197,6 +197,11 @@ esp_err_t esp_intr_alloc_intrstatus(int source, int flags, uint32_t intrstatusre
* Use an interrupt handle to disable the interrupt and release the resources
* associated with it.
*
* @note
* When the handler shares its source with other handlers, the interrupt status
* bits it's responsible for should be managed properly before freeing it. see
* ``esp_intr_disable`` for more details.
*
* @param handle The handle, as obtained by esp_intr_alloc or esp_intr_alloc_intrstatus
*
* @return ESP_ERR_INVALID_ARG if handle is invalid, or esp_intr_free runs on another core than
@ -227,8 +232,13 @@ int esp_intr_get_intno(intr_handle_t handle);
/**
* @brief Disable the interrupt associated with the handle
*
* @note For local interrupts (ESP_INTERNAL_* sources), this function has to be called on the
* CPU the interrupt is allocated on. Other interrupts have no such restriction.
* @note
* 1. For local interrupts (ESP_INTERNAL_* sources), this function has to be called on the
* CPU the interrupt is allocated on. Other interrupts have no such restriction.
* 2. When several handlers sharing a same interrupt source, interrupt status bits, which are
* handled in the handler to be disabled, should be masked before the disabling, or handled
* in other enabled interrupts properly. Miss of interrupt status handling will cause infinite
* interrupt calls and finally system crash.
*
* @param handle The handle, as obtained by esp_intr_alloc or esp_intr_alloc_intrstatus
*

View file

@ -168,7 +168,7 @@ struct non_shared_isr_arg_t {
};
//Linked list of vector descriptions, sorted by cpu.intno value
static vector_desc_t *vector_desc_head;
static vector_desc_t *vector_desc_head = NULL;
//This bitmask has an 1 if the int should be disabled when the flash is disabled.
static uint32_t non_iram_int_mask[portNUM_PROCESSORS];
@ -234,6 +234,32 @@ static vector_desc_t *get_desc_for_int(int intno, int cpu)
}
}
//Returns a vector_desc entry for an source, the cpu parameter is used to tell GPIO_INT and GPIO_NMI from different CPUs
static vector_desc_t * find_desc_for_source(int source, int cpu)
{
vector_desc_t *vd=vector_desc_head;
while(vd!=NULL) {
if ( !(vd->flags & VECDESC_FL_SHARED) ) {
if ( vd->source == source && cpu == vd->cpu ) break;
} else if ( vd->cpu == cpu ) {
// check only shared vds for the correct cpu, otherwise skip
bool found = false;
shared_vector_desc_t *svd = vd->shared_vec_info;
assert(svd != NULL );
while(svd) {
if ( svd->source == source ) {
found = true;
break;
}
svd = svd->next;
}
if ( found ) break;
}
vd=vd->next;
}
return vd;
}
esp_err_t esp_intr_mark_shared(int intno, int cpu, bool is_int_ram)
{
if (intno>31) return ESP_ERR_INVALID_ARG;
@ -286,11 +312,70 @@ static bool int_has_handler(int intr, int cpu)
return (_xt_interrupt_table[intr*portNUM_PROCESSORS+cpu].handler != xt_unhandled_interrupt);
}
static bool is_vect_desc_usable(vector_desc_t *vd, int flags, int cpu, int force)
{
//Check if interrupt is not reserved by design
int x = vd->intno;
if (int_desc[x].cpuflags[cpu]==INTDESC_RESVD) {
ALCHLOG("....Unusable: reserved");
return false;
}
if (int_desc[x].cpuflags[cpu]==INTDESC_SPECIAL && force==-1) {
ALCHLOG("....Unusable: special-purpose int");
return false;
}
//Check if the interrupt level is acceptable
if (!(flags&(1<<int_desc[x].level))) {
ALCHLOG("....Unusable: incompatible level");
return false;
}
//check if edge/level type matches what we want
if (((flags&ESP_INTR_FLAG_EDGE) && (int_desc[x].type==INTTP_LEVEL)) ||
(((!(flags&ESP_INTR_FLAG_EDGE)) && (int_desc[x].type==INTTP_EDGE)))) {
ALCHLOG("....Unusable: incompatible trigger type");
return false;
}
//check if interrupt is reserved at runtime
if (vd->flags&VECDESC_FL_RESERVED) {
ALCHLOG("....Unusable: reserved at runtime.");
return false;
}
//Ints can't be both shared and non-shared.
assert(!((vd->flags&VECDESC_FL_SHARED)&&(vd->flags&VECDESC_FL_NONSHARED)));
//check if interrupt already is in use by a non-shared interrupt
if (vd->flags&VECDESC_FL_NONSHARED) {
ALCHLOG("....Unusable: already in (non-shared) use.");
return false;
}
// check shared interrupt flags
if (vd->flags&VECDESC_FL_SHARED ) {
if (flags&ESP_INTR_FLAG_SHARED) {
bool in_iram_flag=((flags&ESP_INTR_FLAG_IRAM)!=0);
bool desc_in_iram_flag=((vd->flags&VECDESC_FL_INIRAM)!=0);
//Bail out if int is shared, but iram property doesn't match what we want.
if ((vd->flags&VECDESC_FL_SHARED) && (desc_in_iram_flag!=in_iram_flag)) {
ALCHLOG("....Unusable: shared but iram prop doesn't match");
return false;
}
} else {
//We need an unshared IRQ; can't use shared ones; bail out if this is shared.
ALCHLOG("...Unusable: int is shared, we need non-shared.");
return false;
}
} else if (int_has_handler(x, cpu)) {
//Check if interrupt already is allocated by xt_set_interrupt_handler
ALCHLOG("....Unusable: already allocated");
return false;
}
return true;
}
//Locate a free interrupt compatible with the flags given.
//The 'force' argument can be -1, or 0-31 to force checking a certain interrupt.
//When a CPU is forced, the INTDESC_SPECIAL marked interrupts are also accepted.
static int get_free_int(int flags, int cpu, int force)
static int get_available_int(int flags, int cpu, int force, int source)
{
int x;
int best=-1;
@ -299,69 +384,61 @@ static int get_free_int(int flags, int cpu, int force)
//Default vector desc, for vectors not in the linked list
vector_desc_t empty_vect_desc;
memset(&empty_vect_desc, 0, sizeof(vector_desc_t));
//Level defaults to any low/med interrupt
if (!(flags&ESP_INTR_FLAG_LEVELMASK)) flags|=ESP_INTR_FLAG_LOWMED;
ALCHLOG(TAG, "get_free_int: start looking. Current cpu: %d", cpu);
//Iterate over the 32 possible interrupts
ALCHLOG("get_available_int: try to find existing. Cpu: %d, Source: %d", cpu, source);
vector_desc_t *vd = find_desc_for_source(source, cpu);
if ( vd ) {
// if existing vd found, don't need to search any more.
ALCHLOG("get_avalible_int: existing vd found. intno: %d", vd->intno);
if ( force != -1 && force != vd->intno ) {
ALCHLOG("get_avalible_int: intr forced but not matach existing. existing intno: %d, force: %d", vd->intno, force);
} else if ( !is_vect_desc_usable(vd, flags, cpu, force) ) {
ALCHLOG("get_avalible_int: existing vd invalid.");
} else {
best = vd->intno;
}
return best;
}
if (force!=-1) {
ALCHLOG("get_available_int: try to find force. Cpu: %d, Source: %d, Force: %d", cpu, source, force);
//if force assigned, don't need to search any more.
vd = find_desc_for_int(force, cpu);
if (vd == NULL ) {
//if existing vd not found, just check the default state for the intr.
empty_vect_desc.intno = force;
vd = &empty_vect_desc;
}
if ( is_vect_desc_usable(vd, flags, cpu, force) ) {
best = vd->intno;
} else {
ALCHLOG("get_avalible_int: forced vd invalid.");
}
return best;
}
ALCHLOG("get_free_int: start looking. Current cpu: %d", cpu);
//No allocated handlers as well as forced intr, iterate over the 32 possible interrupts
for (x=0; x<32; x++) {
//Grab the vector_desc for this vector.
vector_desc_t *vd=find_desc_for_int(x, cpu);
if (vd==NULL) vd=&empty_vect_desc;
//See if we have a forced interrupt; if so, bail out if this is not it.
if (force!=-1 && force!=x) {
ALCHLOG(TAG, "Ignoring int %d: forced to %d", x, force);
continue;
vd=find_desc_for_int(x, cpu);
if (vd==NULL) {
empty_vect_desc.intno = x;
vd=&empty_vect_desc;
}
ALCHLOG(TAG, "Int %d reserved %d level %d %s hasIsr %d",
ALCHLOG("Int %d reserved %d level %d %s hasIsr %d",
x, int_desc[x].cpuflags[cpu]==INTDESC_RESVD, int_desc[x].level,
int_desc[x].type==INTTP_LEVEL?"LEVEL":"EDGE", int_has_handler(x, cpu));
//Check if interrupt is not reserved by design
if (int_desc[x].cpuflags[cpu]==INTDESC_RESVD) {
ALCHLOG(TAG, "....Unusable: reserved");
continue;
}
if (int_desc[x].cpuflags[cpu]==INTDESC_SPECIAL && force==-1) {
ALCHLOG(TAG, "....Unusable: special-purpose int");
continue;
}
//Check if the interrupt level is acceptable
if (!(flags&(1<<int_desc[x].level))) {
ALCHLOG(TAG, "....Unusable: incompatible level");
continue;
}
//check if edge/level type matches what we want
if (((flags&ESP_INTR_FLAG_EDGE) && (int_desc[x].type==INTTP_LEVEL)) ||
(((!(flags&ESP_INTR_FLAG_EDGE)) && (int_desc[x].type==INTTP_EDGE)))) {
ALCHLOG(TAG, "....Unusable: incompatible trigger type");
continue;
}
//Check if interrupt already is allocated by xt_set_interrupt_handler
if (int_has_handler(x, cpu) && !(vd->flags&VECDESC_FL_SHARED)) {
ALCHLOG(TAG, "....Unusable: already allocated");
continue;
}
//Ints can't be both shared and non-shared.
assert(!((vd->flags&VECDESC_FL_SHARED)&&(vd->flags&VECDESC_FL_NONSHARED)));
//check if interrupt is reserved at runtime
if (vd->flags&VECDESC_FL_RESERVED) {
ALCHLOG(TAG, "....Unusable: reserved at runtime.");
continue;
}
//check if interrupt already is in use by a non-shared interrupt
if (vd->flags&VECDESC_FL_NONSHARED) {
ALCHLOG(TAG, "....Unusable: already in (non-shared) use.");
continue;
}
if ( !is_vect_desc_usable(vd, flags, cpu, force) ) continue;
if (flags&ESP_INTR_FLAG_SHARED) {
//We're allocating a shared int.
bool in_iram_flag=((flags&ESP_INTR_FLAG_IRAM)!=0);
bool desc_in_iram_flag=((vd->flags&VECDESC_FL_INIRAM)!=0);
//Bail out if int is shared, but iram property doesn't match what we want.
if ((vd->flags&VECDESC_FL_SHARED) && (desc_in_iram_flag!=in_iram_flag)) {
ALCHLOG(TAG, "....Unusable: shared but iram prop doesn't match");
continue;
}
//See if int already is used as a shared interrupt.
if (vd->flags&VECDESC_FL_SHARED) {
//We can use this already-marked-as-shared interrupt. Count the already attached isrs in order to see
@ -377,9 +454,9 @@ static int get_free_int(int flags, int cpu, int force)
best=x;
bestSharedCt=no;
bestLevel=int_desc[x].level;
ALCHLOG(TAG, "...int %d more usable as a shared int: has %d existing vectors", x, no);
ALCHLOG("...int %d more usable as a shared int: has %d existing vectors", x, no);
} else {
ALCHLOG(TAG, "...worse than int %d", best);
ALCHLOG("...worse than int %d", best);
}
} else {
if (best==-1) {
@ -389,28 +466,23 @@ static int get_free_int(int flags, int cpu, int force)
if (bestLevel>int_desc[x].level) {
best=x;
bestLevel=int_desc[x].level;
ALCHLOG(TAG, "...int %d usable as a new shared int", x);
ALCHLOG("...int %d usable as a new shared int", x);
}
} else {
ALCHLOG(TAG, "...already have a shared int");
ALCHLOG("...already have a shared int");
}
}
} else {
//We need an unshared IRQ; can't use shared ones; bail out if this is shared.
if (vd->flags&VECDESC_FL_SHARED) {
ALCHLOG(TAG, "...Unusable: int is shared, we need non-shared.");
continue;
}
//Seems this interrupt is feasible. Select it and break out of the loop; no need to search further.
if (bestLevel>int_desc[x].level) {
best=x;
bestLevel=int_desc[x].level;
} else {
ALCHLOG(TAG, "...worse than int %d", best);
ALCHLOG("...worse than int %d", best);
}
}
}
ALCHLOG(TAG, "get_free_int: using int %d", best);
ALCHLOG("get_available_int: using int %d", best);
//Okay, by now we have looked at all potential interrupts and hopefully have selected the best one in best.
return best;
@ -508,7 +580,7 @@ esp_err_t esp_intr_alloc_intrstatus(int source, int flags, uint32_t intrstatusre
portENTER_CRITICAL(&spinlock);
int cpu=xPortGetCoreID();
//See if we can find an interrupt that matches the flags.
int intr=get_free_int(flags, cpu, force);
int intr=get_available_int(flags, cpu, force, source);
if (intr==-1) {
//None found. Bail out.
portEXIT_CRITICAL(&spinlock);
@ -720,15 +792,29 @@ esp_err_t IRAM_ATTR esp_intr_disable(intr_handle_t handle)
if (!handle) return ESP_ERR_INVALID_ARG;
portENTER_CRITICAL(&spinlock);
int source;
bool disabled = 1;
if (handle->shared_vector_desc) {
handle->shared_vector_desc->disabled=1;
source=handle->shared_vector_desc->source;
shared_vector_desc_t *svd=handle->vector_desc->shared_vec_info;
assert( svd != NULL );
while( svd ) {
if ( svd->source == source && svd->disabled == 0 ) {
disabled = 0;
break;
}
svd = svd->next;
}
} else {
source=handle->vector_desc->source;
}
if (source >= 0) {
//Disable using int matrix
intr_matrix_set(handle->vector_desc->cpu, source, INT_MUX_DISABLED_INTNO);
if ( disabled ) {
//Disable using int matrix
intr_matrix_set(handle->vector_desc->cpu, source, INT_MUX_DISABLED_INTNO);
}
} else {
//Disable using per-cpu regs
if (handle->vector_desc->cpu!=xPortGetCoreID()) {

View file

@ -228,3 +228,72 @@ TEST_CASE("Can allocate IRAM int only with an IRAM handler", "[esp32]")
err = esp_intr_free(ih);
TEST_ESP_OK(err);
}
#include "soc/spi_struct.h"
typedef struct {
bool flag1;
bool flag2;
bool flag3;
bool flag4;
} intr_alloc_test_ctx_t;
void IRAM_ATTR int_handler1(void* arg)
{
intr_alloc_test_ctx_t* ctx=(intr_alloc_test_ctx_t*)arg;
ets_printf("handler 1 called.\n");
if ( ctx->flag1 ) {
ctx->flag3 = true;
} else {
ctx->flag1 = true;
}
SPI2.slave.trans_done = 0;
}
void IRAM_ATTR int_handler2(void* arg)
{
intr_alloc_test_ctx_t* ctx = (intr_alloc_test_ctx_t*)arg;
ets_printf("handler 2 called.\n");
if ( ctx->flag2 ) {
ctx->flag4 = true;
} else {
ctx->flag2 = true;
}
}
TEST_CASE("allocate 2 handlers for a same source and remove the later one","[esp32]")
{
intr_alloc_test_ctx_t ctx = {false, false, false, false };
intr_handle_t handle1, handle2;
//enable spi
DPORT_SET_PERI_REG_MASK(DPORT_PERIP_CLK_EN_REG, DPORT_SPI_CLK_EN );
DPORT_CLEAR_PERI_REG_MASK(DPORT_PERIP_RST_EN_REG, DPORT_SPI_RST);
esp_err_t r;
r=esp_intr_alloc(ETS_SPI2_INTR_SOURCE, ESP_INTR_FLAG_SHARED, int_handler1, &ctx, &handle1);
TEST_ESP_OK(r);
//try an invalid assign first
r=esp_intr_alloc(ETS_SPI2_INTR_SOURCE, NULL, int_handler2, NULL, &handle2);
TEST_ASSERT_EQUAL_INT(r, ESP_ERR_NOT_FOUND );
//assign shared then
r=esp_intr_alloc(ETS_SPI2_INTR_SOURCE, ESP_INTR_FLAG_SHARED, int_handler2, &ctx, &handle2);
TEST_ESP_OK(r);
SPI2.slave.trans_inten = 1;
printf("trigger first time.\n");
SPI2.slave.trans_done = 1;
vTaskDelay(100);
TEST_ASSERT( ctx.flag1 && ctx.flag2 );
printf("remove intr 1.\n");
r=esp_intr_free(handle2);
printf("trigger second time.\n");
SPI2.slave.trans_done = 1;
vTaskDelay(500);
TEST_ASSERT( ctx.flag3 && !ctx.flag4 );
printf("test passed.\n");
}

View file

@ -83,6 +83,21 @@ It can also be useful to keep an interrupt handler in IRAM if it is called very
Refer to the :ref:`SPI flash API documentation <iram-safe-interrupt-handlers>` for more details.
Multiple Handlers Sharing A Source
----------------------------------
Several handlers can be assigned to a same source, given that all handlers are allocated using the ``ESP_INTR_FLAG_SHARED`` flag.
They'll be all allocated to the interrupt, which the source is attached to, and called sequentially when the source is active.
The handlers can be disabled and freed individually. The source is attached to the interrupt (enabled), if one or more handlers are enabled, otherwise detached.
A handler will never be called when disabled, while **its source may still be triggered** if any one of its handler enabled.
Sources attached to non-shared interrupt do not support this feature.
Though the framework support this feature, you have to use it *very carefully*. There usually exist 2 ways to stop a interrupt from being triggered: *disable the sourse* or *mask peripheral interrupt status*.
IDF only handles the enabling and disabling of the source itself, leaving status and mask bits to be handled by users. **Status bits should always be masked before the handler responsible for it is disabled,
or the status should be handled in other enabled interrupt properly**. You may leave some status bits unhandled if you just disable one of all the handlers without mask the status bits, which causes the interrupt being triggered infinitely,
and finally a system crash.
API Reference
-------------