From: Fabien Chouteau <chouteau@adacore.com>
To: Blue Swirl <blauwirbel@gmail.com>
Cc: qemu-devel@nongnu.org
Subject: Re: [Qemu-devel] [PATCH 2/6] [RFC] Emulation of GRLIB IRQMP as defined in GRLIB IP Core User's Manual.
Date: Tue, 07 Dec 2010 11:43:05 +0100 [thread overview]
Message-ID: <4CFE0FB9.5080000@adacore.com> (raw)
In-Reply-To: <AANLkTimw0D6DSgMbameqUj-KaxDnnjqk_UvUArdcw=0h@mail.gmail.com>
On 12/06/2010 06:25 PM, Blue Swirl wrote:
> On Mon, Dec 6, 2010 at 9:26 AM, Fabien Chouteau<chouteau@adacore.com> wrote:
>>
>> Signed-off-by: Fabien Chouteau<chouteau@adacore.com>
>> ---
>> hw/grlib_irqmp.c | 416 ++++++++++++++++++++++++++++++++++++++++++++++++++++++
>> 1 files changed, 416 insertions(+), 0 deletions(-)
>>
>> diff --git a/hw/grlib_irqmp.c b/hw/grlib_irqmp.c
>> new file mode 100644
>> index 0000000..69e1553
>> --- /dev/null
>> +++ b/hw/grlib_irqmp.c
>> @@ -0,0 +1,416 @@
>> +/*
>> + * QEMU GRLIB IRQMP Emulator
>> + *
>> + * (Multiprocessor and extended interrupt not supported)
>> + *
>> + * Copyright (c) 2010 AdaCore
>> + *
>> + * Permission is hereby granted, free of charge, to any person obtaining a copy
>> + * of this software and associated documentation files (the "Software"), to deal
>> + * in the Software without restriction, including without limitation the rights
>> + * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
>> + * copies of the Software, and to permit persons to whom the Software is
>> + * furnished to do so, subject to the following conditions:
>> + *
>> + * The above copyright notice and this permission notice shall be included in
>> + * all copies or substantial portions of the Software.
>> + *
>> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
>> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
>> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
>> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
>> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
>> + * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
>> + * THE SOFTWARE.
>> + */
>> +
>> +#include "sysbus.h"
>> +#include "cpu.h"
>> +
>> +#include "grlib.h"
>> +
>> +/* #define DEBUG_IRQ */
>> +
>> +#ifdef DEBUG_IRQ
>> +#define DPRINTF(fmt, ...) \
>> + do { printf("IRQMP: " fmt , ## __VA_ARGS__); } while (0)
>> +#else
>> +#define DPRINTF(fmt, ...)
>> +#endif
>> +
>> +#define IRQMP_MAX_CPU 16
>> +#define IRQMP_REG_SIZE 256 /* Size of memory mapped registers */
>> +
>> +/* Memory mapped register offsets */
>> +#define LEVEL_OFFSET 0x00
>> +#define PENDING_OFFSET 0x04
>> +#define FORCE0_OFFSET 0x08
>> +#define CLEAR_OFFSET 0x0C
>> +#define MP_STATUS_OFFSET 0x10
>> +#define BROADCAST_OFFSET 0x14
>> +#define MASK_OFFSET 0x40
>> +#define FORCE_OFFSET 0x80
>> +#define EXTENDED_OFFSET 0xC0
>> +
>> +typedef struct IRQMP
>> +{
>> + SysBusDevice busdev;
>> +
>> + CPUSPARCState *env;
>
> Devices should never access CPUState directly. Instead, board level
> should create CPU irqs and these should then be passed here.
>
This case is special, Leon3 is a System-On-Chip and some of the
components are very close to the processor.
IRQMP is not really a peripheral nor a part of the CPU, it's both...
>> +} IRQMP;
>> +
>> +typedef struct IRQMPState
>> +{
>> + uint32_t level;
>> + uint32_t pending;
>> + uint32_t clear;
>> + uint32_t broadcast;
>> +
>> + uint32_t mask[IRQMP_MAX_CPU];
>> + uint32_t force[IRQMP_MAX_CPU];
>> + uint32_t extended[IRQMP_MAX_CPU];
>> +
>> + IRQMP *parent;
>> +} IRQMPState;
>> +
>> +IRQMPState grlib_irqmp_state;
>
> Global state indicates poor design. Why separate IRQMP and IRQMPState?
I have to access IRQMPState in grlib_irqmp_ack and
grlib_irqmp_check_irqs, but I don't see how I can do it without a global
variable.
Again, I think that it's related to the special case of IRQMP.
>> +
>> +void grlib_irqmp_set_irq(void *opaque, int irq, int level);
>
> This should not be global. Again, creating qemu_irqs or moving some of
> the code to board level should help.
This one should be static indeed.
>> +
>> +DeviceState *grlib_irqmp_create(target_phys_addr_t base,
>> + CPUState *env,
>> + qemu_irq **cpu_irqs,
>> + uint32_t nr_irqs)
>> +{
>> + DeviceState *dev;
>> +
>> + assert(cpu_irqs != NULL);
>> +
>> + dev = qdev_create(NULL, "grlib,irqmp");
>> + qdev_prop_set_ptr(dev, "cpustate", env);
>> +
>> + if (qdev_init(dev)) {
>> + return NULL;
>> + }
>> +
>> + sysbus_mmio_map(sysbus_from_qdev(dev), 0, base);
>> +
>> + *cpu_irqs = qemu_allocate_irqs(grlib_irqmp_set_irq,
>> +&grlib_irqmp_state,
>> + nr_irqs);
>> +
>> + return dev;
>> +}
>> +
>> +static void grlib_irqmp_check_irqs(CPUState *env)
>> +{
>> + uint32_t pend = 0;
>> + uint32_t level0 = 0;
>> + uint32_t level1 = 0;
>> +
>> + assert(env != NULL);
>> +
>> + /* IRQ for CPU 0 (no SMP support) */
>> + pend = (grlib_irqmp_state.pending | grlib_irqmp_state.force[0])
>> +& grlib_irqmp_state.mask[0];
>> +
>> +
>> + level0 = pend& ~grlib_irqmp_state.level;
>> + level1 = pend& grlib_irqmp_state.level;
>> +
>> + DPRINTF("pend:0x%04x force:0x%04x mask:0x%04x lvl1:0x%04x lvl0:0x%04x\n",
>> + grlib_irqmp_state.pending, grlib_irqmp_state.force[0],
>> + grlib_irqmp_state.mask[0], level1, level0);
>
> The above should stay here, but code below should to go to board level
> (leon3.c). Then you need to separate device IRQ handling from CPU PIL
> handling.
If I want to use IRQMP for another machine I will have to duplicate the
code.
So I think it is the right place for this this code.
>> +
>> + /* Trigger level1 interrupt first and level0 if there is no level1 */
>> + if (level1 != 0) {
>> + env->pil_in = level1;
>> + } else {
>> + env->pil_in = level0;
>> + }
>> +
>> + if (env->pil_in&& (env->interrupt_index == 0 ||
>> + (env->interrupt_index& ~15) == TT_EXTINT)) {
>> + unsigned int i;
>> +
>> + for (i = 15; i> 0; i--) {
>> + if (env->pil_in& (1<< i)) {
>> + int old_interrupt = env->interrupt_index;
>> +
>> + env->interrupt_index = TT_EXTINT | i;
>> + if (old_interrupt != env->interrupt_index) {
>> + DPRINTF("Set CPU IRQ %d\n", i);
>> + cpu_interrupt(env, CPU_INTERRUPT_HARD);
>> + }
>> + break;
>> + }
>> + }
>> + } else if (!env->pil_in&& (env->interrupt_index& ~15) == TT_EXTINT) {
>> + DPRINTF("Reset CPU IRQ %d\n", env->interrupt_index& 15);
>> + env->interrupt_index = 0;
>> + cpu_reset_interrupt(env, CPU_INTERRUPT_HARD);
>> + }
>> +}
>> +
>> +void grlib_irqmp_ack(CPUSPARCState *env, int intno)
>> +{
>> + assert(env != NULL);
>> +
>> + uint32_t mask;
>> +
>> + intno&= 15;
>> + mask = 1<< intno;
>> +
>> + DPRINTF("grlib_irqmp_ack %d\n", intno);
>> +
>> + /* Clear registers */
>> + grlib_irqmp_state.pending&= ~mask;
>> + grlib_irqmp_state.force[0]&= ~mask; /* Only CPU 0 (No SMP support) */
>> +
>> + grlib_irqmp_check_irqs(env);
>> +}
>> +
>> +void grlib_irqmp_set_irq(void *opaque, int irq, int level)
>> +{
>> + IRQMPState *s = opaque;
>> + int i = 0;
>> +
>> + assert(opaque != NULL);
>> + assert(s->parent != NULL);
>> +
>> + if (level) {
>> + DPRINTF("Raise CPU IRQ %d\n", irq);
>> +
>> + if (s->broadcast& 1<< irq) {
>> + /* Broadcasted IRQ */
>> + for (i = 0; i< IRQMP_MAX_CPU; i++) {
>> + s->force[i] |= 1<< irq;
>> + }
>> + } else {
>> + s->pending |= 1<< irq;
>> + }
>> + grlib_irqmp_check_irqs(s->parent->env);
>> +
>> + } else {
>> +
>> + DPRINTF("Lower CPU IRQ %d\n", irq);
>> + if (s->broadcast& 1<< irq) {
>> + /* Broadcasted IRQ */
>> + for (i = 0; i< IRQMP_MAX_CPU; i++) {
>> + s->force[i]&= ~(1<< irq);
>> + }
>> + } else {
>> + s->pending&= ~(1<< irq);
>> + }
>> + grlib_irqmp_check_irqs(s->parent->env);
>> + }
>> +}
>> +
>> +static uint32_t grlib_irqmp_readl (void *opaque, target_phys_addr_t addr)
>> +{
>> + IRQMP *irqmp = opaque;
>> +
>> + assert(irqmp != NULL);
>> +
>> + addr&= 0xff;
>> +
>> + /* global registers */
>> + switch (addr)
>> + {
>> + case LEVEL_OFFSET:
>> + return grlib_irqmp_state.level;
>> +
>> + case PENDING_OFFSET:
>> + return grlib_irqmp_state.pending;
>> +
>> + case FORCE0_OFFSET:
>> + /* This register is an "alias" for the force register of CPU 0 */
>> + return grlib_irqmp_state.force[0];
>> +
>> + case CLEAR_OFFSET:
>> + case MP_STATUS_OFFSET:
>> + /* Always read as 0 */
>> + return 0;
>> +
>> + case BROADCAST_OFFSET:
>> + return grlib_irqmp_state.broadcast;
>> +
>> + default:
>> + break;
>> + }
>> +
>> + /* mask registers */
>> + if (addr>= MASK_OFFSET&& addr< FORCE_OFFSET) {
>> + int cpu = (addr - MASK_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + return grlib_irqmp_state.mask[cpu] ;
>> + }
>> +
>> + /* force registers */
>> + if (addr>= FORCE_OFFSET&& addr< EXTENDED_OFFSET) {
>> + int cpu = (addr - FORCE_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + return grlib_irqmp_state.force[cpu];
>> + }
>> +
>> + /* extended (not supported) */
>> + if (addr>= EXTENDED_OFFSET&& addr< IRQMP_REG_SIZE) {
>> + int cpu = (addr - EXTENDED_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + return grlib_irqmp_state.extended[cpu];
>> + }
>> +
>> + DPRINTF("read unknown register 0x%04x\n", (int)addr);
>> + return 0;
>> +}
>> +
>> +static void
>> +grlib_irqmp_writel (void *opaque, target_phys_addr_t addr, uint32_t value)
>> +{
>> + IRQMP *irqmp = opaque;
>> +
>> + assert(irqmp != NULL);
>> +
>> + addr&= 0xff;
>> +
>> + /* global registers */
>> + switch (addr)
>> + {
>> + case LEVEL_OFFSET:
>> + value&= 0xFFFF<< 1; /* clean up the value */
>> + grlib_irqmp_state.level = value;
>> + return;
>> +
>> + case PENDING_OFFSET:
>> + /* Read Only */
>> + return;
>> +
>> + case FORCE0_OFFSET:
>> + /* This register is an "alias" for the force register of CPU 0 */
>> +
>> + value&= 0xFFFE; /* clean up the value */
>> + grlib_irqmp_state.force[0] = value;
>> + grlib_irqmp_check_irqs(irqmp->env);
>> + return;
>> +
>> + case CLEAR_OFFSET:
>> + value&= ~1; /* clean up the value */
>> + grlib_irqmp_state.pending&= ~value;
>> + return;
>> +
>> + case MP_STATUS_OFFSET:
>> + /* Read Only (no SMP support) */
>> + return;
>> +
>> + case BROADCAST_OFFSET:
>> + value&= 0xFFFE; /* clean up the value */
>> + grlib_irqmp_state.broadcast = value;
>> + return;
>> +
>> + default:
>> + break;
>> + }
>> +
>> + /* mask registers */
>> + if (addr>= MASK_OFFSET&& addr< FORCE_OFFSET) {
>> + int cpu = (addr - MASK_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + value&= ~1; /* clean up the value */
>> + grlib_irqmp_state.mask[cpu] = value;
>> + grlib_irqmp_check_irqs(irqmp->env);
>> + return;
>> + }
>> +
>> + /* force registers */
>> + if (addr>= FORCE_OFFSET&& addr< EXTENDED_OFFSET) {
>> + int cpu = (addr - FORCE_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + uint32_t force = value& 0xFFFE;
>> + uint32_t clear = (value>> 16)& 0xFFFE;
>> + uint32_t old = grlib_irqmp_state.force[cpu];
>> +
>> + grlib_irqmp_state.force[cpu] = (old | force)& ~clear;
>> + grlib_irqmp_check_irqs(irqmp->env);
>> + return;
>> + }
>> +
>> + /* extended (not supported) */
>> + if (addr>= EXTENDED_OFFSET&& addr< IRQMP_REG_SIZE) {
>> + int cpu = (addr - EXTENDED_OFFSET) / 4;
>> + assert(cpu>= 0&& cpu< IRQMP_MAX_CPU);
>> +
>> + value&= 0xF; /* clean up the value */
>> + grlib_irqmp_state.extended[cpu] = value;
>> + return;
>> + }
>> +
>> + DPRINTF("write unknown register 0x%04x\n", (int)addr);
>
> Please use TARGET_FMT_plx, so the cast can be removed.
OK, I didn't know about that.
--
Fabien Chouteau
next prev parent reply other threads:[~2010-12-07 10:43 UTC|newest]
Thread overview: 51+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-12-06 9:26 [Qemu-devel] [PATCH 0/6] [RFC] New SPARC machine: Leon3 Fabien Chouteau
2010-12-06 9:26 ` [Qemu-devel] [PATCH 1/6] [RFC] Emulation of GRLIB GPTimer as defined in GRLIB IP Core User's Manual Fabien Chouteau
2010-12-06 17:12 ` Blue Swirl
2010-12-07 9:55 ` Fabien Chouteau
2010-12-08 8:30 ` Edgar E. Iglesias
2010-12-08 9:39 ` Fabien Chouteau
2010-12-08 21:02 ` Edgar E. Iglesias
2010-12-08 22:51 ` Edgar E. Iglesias
2010-12-09 10:04 ` Fabien Chouteau
2010-12-09 10:22 ` Edgar E. Iglesias
2010-12-06 9:26 ` [Qemu-devel] [PATCH 2/6] [RFC] Emulation of GRLIB IRQMP " Fabien Chouteau
2010-12-06 17:25 ` Blue Swirl
2010-12-07 10:43 ` Fabien Chouteau [this message]
2010-12-11 10:31 ` Blue Swirl
2010-12-13 16:23 ` Fabien Chouteau
2010-12-13 18:13 ` Blue Swirl
2010-12-09 10:32 ` Edgar E. Iglesias
2010-12-09 11:03 ` Fabien Chouteau
2010-12-09 11:06 ` Edgar E. Iglesias
2010-12-09 11:32 ` Fabien Chouteau
2010-12-06 9:26 ` [Qemu-devel] [PATCH 3/6] [RFC] Emulation of GRLIB APB UART " Fabien Chouteau
2010-12-06 17:29 ` Blue Swirl
2010-12-07 10:55 ` Fabien Chouteau
2010-12-06 9:26 ` [Qemu-devel] [PATCH 4/6] [RFC] Header file for the GRLIB components Fabien Chouteau
2010-12-06 17:31 ` Blue Swirl
2010-12-07 11:04 ` Fabien Chouteau
2010-12-06 9:26 ` [Qemu-devel] [PATCH 5/6] [RFC] Emulation of Leon3 Fabien Chouteau
2010-12-06 17:53 ` Blue Swirl
2010-12-07 11:40 ` Fabien Chouteau
2010-12-11 9:56 ` Blue Swirl
2010-12-13 15:51 ` Fabien Chouteau
2010-12-13 18:18 ` Blue Swirl
2010-12-15 17:47 ` Fabien Chouteau
2010-12-17 19:14 ` Blue Swirl
2010-12-20 6:46 ` Edgar E. Iglesias
2010-12-20 9:40 ` Fabien Chouteau
2010-12-20 20:09 ` Blue Swirl
2010-12-20 9:25 ` Fabien Chouteau
2010-12-20 19:27 ` Blue Swirl
2010-12-12 14:41 ` Andreas Färber
2010-12-13 17:00 ` Fabien Chouteau
2010-12-06 9:26 ` [Qemu-devel] [PATCH 6/6] [RFC] SPARCV8 asr17 register support Fabien Chouteau
2010-12-06 18:01 ` Blue Swirl
2010-12-07 11:51 ` Fabien Chouteau
2010-12-11 9:59 ` Blue Swirl
2010-12-13 17:01 ` Fabien Chouteau
2010-12-06 10:44 ` [Qemu-devel] [PATCH 0/6] [RFC] New SPARC machine: Leon3 Artyom Tarasenko
2010-12-06 15:07 ` Fabien Chouteau
2010-12-06 18:12 ` Blue Swirl
2010-12-07 17:43 ` Fabien Chouteau
2010-12-06 18:05 ` Blue Swirl
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=4CFE0FB9.5080000@adacore.com \
--to=chouteau@adacore.com \
--cc=blauwirbel@gmail.com \
--cc=qemu-devel@nongnu.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.