linux/arch/powerpc/platforms/cell/spu_base.c

694 lines
16 KiB
C
Raw Normal View History

/*
* Low-level SPU handling
*
* (C) Copyright IBM Deutschland Entwicklung GmbH 2005
*
* Author: Arnd Bergmann <arndb@de.ibm.com>
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2, or (at your option)
* any later version.
*
* This program is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
* GNU General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this program; if not, write to the Free Software
* Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
*/
#undef DEBUG
#include <linux/interrupt.h>
#include <linux/list.h>
#include <linux/module.h>
#include <linux/ptrace.h>
#include <linux/slab.h>
#include <linux/wait.h>
#include <linux/mm.h>
#include <linux/io.h>
#include <linux/mutex.h>
#include <linux/linux_logo.h>
#include <asm/spu.h>
#include <asm/spu_priv1.h>
#include <asm/xmon.h>
const struct spu_management_ops *spu_management_ops;
EXPORT_SYMBOL_GPL(spu_management_ops);
const struct spu_priv1_ops *spu_priv1_ops;
static struct list_head spu_list[MAX_NUMNODES];
static LIST_HEAD(spu_full_list);
static DEFINE_MUTEX(spu_mutex);
static DEFINE_SPINLOCK(spu_list_lock);
EXPORT_SYMBOL_GPL(spu_priv1_ops);
void spu_invalidate_slbs(struct spu *spu)
{
struct spu_priv2 __iomem *priv2 = spu->priv2;
if (spu_mfc_sr1_get(spu) & MFC_STATE1_RELOCATE_MASK)
out_be64(&priv2->slb_invalidate_all_W, 0UL);
}
EXPORT_SYMBOL_GPL(spu_invalidate_slbs);
/* This is called by the MM core when a segment size is changed, to
* request a flush of all the SPEs using a given mm
*/
void spu_flush_all_slbs(struct mm_struct *mm)
{
struct spu *spu;
unsigned long flags;
spin_lock_irqsave(&spu_list_lock, flags);
list_for_each_entry(spu, &spu_full_list, full_list) {
if (spu->mm == mm)
spu_invalidate_slbs(spu);
}
spin_unlock_irqrestore(&spu_list_lock, flags);
}
/* The hack below stinks... try to do something better one of
* these days... Does it even work properly with NR_CPUS == 1 ?
*/
static inline void mm_needs_global_tlbie(struct mm_struct *mm)
{
int nr = (NR_CPUS > 1) ? NR_CPUS : NR_CPUS + 1;
/* Global TLBIE broadcast required with SPEs. */
__cpus_setall(&mm->cpu_vm_mask, nr);
}
void spu_associate_mm(struct spu *spu, struct mm_struct *mm)
{
unsigned long flags;
spin_lock_irqsave(&spu_list_lock, flags);
spu->mm = mm;
spin_unlock_irqrestore(&spu_list_lock, flags);
if (mm)
mm_needs_global_tlbie(mm);
}
EXPORT_SYMBOL_GPL(spu_associate_mm);
static int __spu_trap_invalid_dma(struct spu *spu)
{
pr_debug("%s\n", __FUNCTION__);
spu->dma_callback(spu, SPE_EVENT_INVALID_DMA);
return 0;
}
static int __spu_trap_dma_align(struct spu *spu)
{
pr_debug("%s\n", __FUNCTION__);
spu->dma_callback(spu, SPE_EVENT_DMA_ALIGNMENT);
return 0;
}
static int __spu_trap_error(struct spu *spu)
{
pr_debug("%s\n", __FUNCTION__);
spu->dma_callback(spu, SPE_EVENT_SPE_ERROR);
return 0;
}
static void spu_restart_dma(struct spu *spu)
{
struct spu_priv2 __iomem *priv2 = spu->priv2;
if (!test_bit(SPU_CONTEXT_SWITCH_PENDING, &spu->flags))
out_be64(&priv2->mfc_control_RW, MFC_CNTL_RESTART_DMA_COMMAND);
}
static int __spu_trap_data_seg(struct spu *spu, unsigned long ea)
{
struct spu_priv2 __iomem *priv2 = spu->priv2;
struct mm_struct *mm = spu->mm;
u64 esid, vsid, llp;
int psize;
pr_debug("%s\n", __FUNCTION__);
if (test_bit(SPU_CONTEXT_SWITCH_ACTIVE, &spu->flags)) {
/* SLBs are pre-loaded for context switch, so
* we should never get here!
*/
printk("%s: invalid access during switch!\n", __func__);
return 1;
}
esid = (ea & ESID_MASK) | SLB_ESID_V;
switch(REGION_ID(ea)) {
case USER_REGION_ID:
[POWERPC] Introduce address space "slices" The basic issue is to be able to do what hugetlbfs does but with different page sizes for some other special filesystems; more specifically, my need is: - Huge pages - SPE local store mappings using 64K pages on a 4K base page size kernel on Cell - Some special 4K segments in 64K-page kernels for mapping a dodgy type of powerpc-specific infiniband hardware that requires 4K MMU mappings for various reasons I won't explain here. The main issues are: - To maintain/keep track of the page size per "segment" (as we can only have one page size per segment on powerpc, which are 256MB divisions of the address space). - To make sure special mappings stay within their allotted "segments" (including MAP_FIXED crap) - To make sure everybody else doesn't mmap/brk/grow_stack into a "segment" that is used for a special mapping Some of the necessary mechanisms to handle that were present in the hugetlbfs code, but mostly in ways not suitable for anything else. The patch relies on some changes to the generic get_unmapped_area() that just got merged. It still hijacks hugetlb callbacks here or there as the generic code hasn't been entirely cleaned up yet but that shouldn't be a problem. So what is a slice ? Well, I re-used the mechanism used formerly by our hugetlbfs implementation which divides the address space in "meta-segments" which I called "slices". The division is done using 256MB slices below 4G, and 1T slices above. Thus the address space is divided currently into 16 "low" slices and 16 "high" slices. (Special case: high slice 0 is the area between 4G and 1T). Doing so simplifies significantly the tracking of segments and avoids having to keep track of all the 256MB segments in the address space. While I used the "concepts" of hugetlbfs, I mostly re-implemented everything in a more generic way and "ported" hugetlbfs to it. Slices can have an associated page size, which is encoded in the mmu context and used by the SLB miss handler to set the segment sizes. The hash code currently doesn't care, it has a specific check for hugepages, though I might add a mechanism to provide per-slice hash mapping functions in the future. The slice code provide a pair of "generic" get_unmapped_area() (bottomup and topdown) functions that should work with any slice size. There is some trickiness here so I would appreciate people to have a look at the implementation of these and let me know if I got something wrong. Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org> Signed-off-by: Paul Mackerras <paulus@samba.org>
2007-05-08 06:27:27 +00:00
#ifdef CONFIG_PPC_MM_SLICES
psize = get_slice_psize(mm, ea);
#else
psize = mm->context.user_psize;
#endif
vsid = (get_vsid(mm->context.id, ea) << SLB_VSID_SHIFT) |
SLB_VSID_USER;
break;
case VMALLOC_REGION_ID:
if (ea < VMALLOC_END)
psize = mmu_vmalloc_psize;
else
psize = mmu_io_psize;
vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
SLB_VSID_KERNEL;
break;
case KERNEL_REGION_ID:
psize = mmu_linear_psize;
vsid = (get_kernel_vsid(ea) << SLB_VSID_SHIFT) |
SLB_VSID_KERNEL;
break;
default:
/* Future: support kernel segments so that drivers
* can use SPUs.
*/
pr_debug("invalid region access at %016lx\n", ea);
return 1;
}
llp = mmu_psize_defs[psize].sllp;
out_be64(&priv2->slb_index_W, spu->slb_replace);
out_be64(&priv2->slb_vsid_RW, vsid | llp);
out_be64(&priv2->slb_esid_RW, esid);
spu->slb_replace++;
if (spu->slb_replace >= 8)
spu->slb_replace = 0;
spu_restart_dma(spu);
spu->stats.slb_flt++;
return 0;
}
extern int hash_page(unsigned long ea, unsigned long access, unsigned long trap); //XXX
static int __spu_trap_data_map(struct spu *spu, unsigned long ea, u64 dsisr)
{
pr_debug("%s, %lx, %lx\n", __FUNCTION__, dsisr, ea);
/* Handle kernel space hash faults immediately.
User hash faults need to be deferred to process context. */
if ((dsisr & MFC_DSISR_PTE_NOT_FOUND)
&& REGION_ID(ea) != USER_REGION_ID
&& hash_page(ea, _PAGE_PRESENT, 0x300) == 0) {
spu_restart_dma(spu);
return 0;
}
if (test_bit(SPU_CONTEXT_SWITCH_ACTIVE, &spu->flags)) {
printk("%s: invalid access during switch!\n", __func__);
return 1;
}
spu->dar = ea;
spu->dsisr = dsisr;
mb();
spu->stop_callback(spu);
return 0;
}
static irqreturn_t
spu_irq_class_0(int irq, void *data)
{
struct spu *spu;
spu = data;
spu->class_0_pending = 1;
spu->stop_callback(spu);
return IRQ_HANDLED;
}
int
spu_irq_class_0_bottom(struct spu *spu)
{
unsigned long stat, mask;
unsigned long flags;
spu->class_0_pending = 0;
spin_lock_irqsave(&spu->register_lock, flags);
mask = spu_int_mask_get(spu, 0);
stat = spu_int_stat_get(spu, 0);
stat &= mask;
if (stat & 1) /* invalid DMA alignment */
__spu_trap_dma_align(spu);
if (stat & 2) /* invalid MFC DMA */
__spu_trap_invalid_dma(spu);
if (stat & 4) /* error on SPU */
__spu_trap_error(spu);
spu_int_stat_clear(spu, 0, stat);
spin_unlock_irqrestore(&spu->register_lock, flags);
return (stat & 0x7) ? -EIO : 0;
}
EXPORT_SYMBOL_GPL(spu_irq_class_0_bottom);
static irqreturn_t
spu_irq_class_1(int irq, void *data)
{
struct spu *spu;
unsigned long stat, mask, dar, dsisr;
spu = data;
/* atomically read & clear class1 status. */
spin_lock(&spu->register_lock);
mask = spu_int_mask_get(spu, 1);
stat = spu_int_stat_get(spu, 1) & mask;
dar = spu_mfc_dar_get(spu);
dsisr = spu_mfc_dsisr_get(spu);
if (stat & 2) /* mapping fault */
spu_mfc_dsisr_set(spu, 0ul);
spu_int_stat_clear(spu, 1, stat);
spin_unlock(&spu->register_lock);
pr_debug("%s: %lx %lx %lx %lx\n", __FUNCTION__, mask, stat,
dar, dsisr);
if (stat & 1) /* segment fault */
__spu_trap_data_seg(spu, dar);
if (stat & 2) { /* mapping fault */
__spu_trap_data_map(spu, dar, dsisr);
}
if (stat & 4) /* ls compare & suspend on get */
;
if (stat & 8) /* ls compare & suspend on put */
;
return stat ? IRQ_HANDLED : IRQ_NONE;
}
static irqreturn_t
spu_irq_class_2(int irq, void *data)
{
struct spu *spu;
unsigned long stat;
unsigned long mask;
spu = data;
spin_lock(&spu->register_lock);
stat = spu_int_stat_get(spu, 2);
mask = spu_int_mask_get(spu, 2);
/* ignore interrupts we're not waiting for */
stat &= mask;
/*
* mailbox interrupts (0x1 and 0x10) are level triggered.
* mask them now before acknowledging.
*/
if (stat & 0x11)
spu_int_mask_and(spu, 2, ~(stat & 0x11));
/* acknowledge all interrupts before the callbacks */
spu_int_stat_clear(spu, 2, stat);
spin_unlock(&spu->register_lock);
pr_debug("class 2 interrupt %d, %lx, %lx\n", irq, stat, mask);
if (stat & 1) /* PPC core mailbox */
spu->ibox_callback(spu);
if (stat & 2) /* SPU stop-and-signal */
spu->stop_callback(spu);
if (stat & 4) /* SPU halted */
spu->stop_callback(spu);
if (stat & 8) /* DMA tag group complete */
spu->mfc_callback(spu);
if (stat & 0x10) /* SPU mailbox threshold */
spu->wbox_callback(spu);
spu->stats.class2_intr++;
return stat ? IRQ_HANDLED : IRQ_NONE;
}
2006-07-03 11:36:01 +00:00
static int spu_request_irqs(struct spu *spu)
{
2006-07-03 11:36:01 +00:00
int ret = 0;
2006-07-03 11:36:01 +00:00
if (spu->irqs[0] != NO_IRQ) {
snprintf(spu->irq_c0, sizeof (spu->irq_c0), "spe%02d.0",
spu->number);
ret = request_irq(spu->irqs[0], spu_irq_class_0,
IRQF_DISABLED,
spu->irq_c0, spu);
if (ret)
goto bail0;
}
if (spu->irqs[1] != NO_IRQ) {
snprintf(spu->irq_c1, sizeof (spu->irq_c1), "spe%02d.1",
spu->number);
ret = request_irq(spu->irqs[1], spu_irq_class_1,
IRQF_DISABLED,
spu->irq_c1, spu);
if (ret)
goto bail1;
}
if (spu->irqs[2] != NO_IRQ) {
snprintf(spu->irq_c2, sizeof (spu->irq_c2), "spe%02d.2",
spu->number);
ret = request_irq(spu->irqs[2], spu_irq_class_2,
IRQF_DISABLED,
spu->irq_c2, spu);
if (ret)
goto bail2;
}
return 0;
2006-07-03 11:36:01 +00:00
bail2:
if (spu->irqs[1] != NO_IRQ)
free_irq(spu->irqs[1], spu);
bail1:
if (spu->irqs[0] != NO_IRQ)
free_irq(spu->irqs[0], spu);
bail0:
return ret;
}
2006-07-03 11:36:01 +00:00
static void spu_free_irqs(struct spu *spu)
{
2006-07-03 11:36:01 +00:00
if (spu->irqs[0] != NO_IRQ)
free_irq(spu->irqs[0], spu);
if (spu->irqs[1] != NO_IRQ)
free_irq(spu->irqs[1], spu);
if (spu->irqs[2] != NO_IRQ)
free_irq(spu->irqs[2], spu);
}
static void spu_init_channels(struct spu *spu)
{
static const struct {
unsigned channel;
unsigned count;
} zero_list[] = {
{ 0x00, 1, }, { 0x01, 1, }, { 0x03, 1, }, { 0x04, 1, },
{ 0x18, 1, }, { 0x19, 1, }, { 0x1b, 1, }, { 0x1d, 1, },
}, count_list[] = {
{ 0x00, 0, }, { 0x03, 0, }, { 0x04, 0, }, { 0x15, 16, },
{ 0x17, 1, }, { 0x18, 0, }, { 0x19, 0, }, { 0x1b, 0, },
{ 0x1c, 1, }, { 0x1d, 0, }, { 0x1e, 1, },
};
struct spu_priv2 __iomem *priv2;
int i;
priv2 = spu->priv2;
/* initialize all channel data to zero */
for (i = 0; i < ARRAY_SIZE(zero_list); i++) {
int count;
out_be64(&priv2->spu_chnlcntptr_RW, zero_list[i].channel);
for (count = 0; count < zero_list[i].count; count++)
out_be64(&priv2->spu_chnldata_RW, 0);
}
/* initialize channel counts to meaningful values */
for (i = 0; i < ARRAY_SIZE(count_list); i++) {
out_be64(&priv2->spu_chnlcntptr_RW, count_list[i].channel);
out_be64(&priv2->spu_chnlcnt_RW, count_list[i].count);
}
}
struct spu *spu_alloc_node(int node)
{
struct spu *spu = NULL;
mutex_lock(&spu_mutex);
if (!list_empty(&spu_list[node])) {
spu = list_entry(spu_list[node].next, struct spu, list);
list_del_init(&spu->list);
pr_debug("Got SPU %d %d\n", spu->number, spu->node);
}
mutex_unlock(&spu_mutex);
if (spu)
spu_init_channels(spu);
return spu;
}
EXPORT_SYMBOL_GPL(spu_alloc_node);
struct spu *spu_alloc(void)
{
struct spu *spu = NULL;
int node;
for (node = 0; node < MAX_NUMNODES; node++) {
spu = spu_alloc_node(node);
if (spu)
break;
}
return spu;
}
void spu_free(struct spu *spu)
{
mutex_lock(&spu_mutex);
list_add_tail(&spu->list, &spu_list[spu->node]);
mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_free);
static int spu_shutdown(struct sys_device *sysdev)
{
struct spu *spu = container_of(sysdev, struct spu, sysdev);
spu_free_irqs(spu);
spu_destroy_spu(spu);
return 0;
}
struct sysdev_class spu_sysdev_class = {
set_kset_name("spu"),
.shutdown = spu_shutdown,
};
int spu_add_sysdev_attr(struct sysdev_attribute *attr)
{
struct spu *spu;
mutex_lock(&spu_mutex);
list_for_each_entry(spu, &spu_full_list, full_list)
sysdev_create_file(&spu->sysdev, attr);
mutex_unlock(&spu_mutex);
return 0;
}
EXPORT_SYMBOL_GPL(spu_add_sysdev_attr);
int spu_add_sysdev_attr_group(struct attribute_group *attrs)
{
struct spu *spu;
mutex_lock(&spu_mutex);
list_for_each_entry(spu, &spu_full_list, full_list)
sysfs_create_group(&spu->sysdev.kobj, attrs);
mutex_unlock(&spu_mutex);
return 0;
}
EXPORT_SYMBOL_GPL(spu_add_sysdev_attr_group);
void spu_remove_sysdev_attr(struct sysdev_attribute *attr)
{
struct spu *spu;
mutex_lock(&spu_mutex);
list_for_each_entry(spu, &spu_full_list, full_list)
sysdev_remove_file(&spu->sysdev, attr);
mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_remove_sysdev_attr);
void spu_remove_sysdev_attr_group(struct attribute_group *attrs)
{
struct spu *spu;
mutex_lock(&spu_mutex);
list_for_each_entry(spu, &spu_full_list, full_list)
sysfs_remove_group(&spu->sysdev.kobj, attrs);
mutex_unlock(&spu_mutex);
}
EXPORT_SYMBOL_GPL(spu_remove_sysdev_attr_group);
static int spu_create_sysdev(struct spu *spu)
{
int ret;
spu->sysdev.id = spu->number;
spu->sysdev.cls = &spu_sysdev_class;
ret = sysdev_register(&spu->sysdev);
if (ret) {
printk(KERN_ERR "Can't register SPU %d with sysfs\n",
spu->number);
return ret;
}
sysfs_add_device_to_node(&spu->sysdev, spu->node);
return 0;
}
static int __init create_spu(void *data)
{
struct spu *spu;
int ret;
static int number;
unsigned long flags;
ret = -ENOMEM;
spu = kzalloc(sizeof (*spu), GFP_KERNEL);
if (!spu)
goto out;
spin_lock_init(&spu->register_lock);
mutex_lock(&spu_mutex);
spu->number = number++;
mutex_unlock(&spu_mutex);
ret = spu_create_spu(spu, data);
if (ret)
goto out_free;
spu_mfc_sdr_setup(spu);
spu_mfc_sr1_set(spu, 0x33);
ret = spu_request_irqs(spu);
if (ret)
goto out_destroy;
ret = spu_create_sysdev(spu);
if (ret)
goto out_free_irqs;
mutex_lock(&spu_mutex);
spin_lock_irqsave(&spu_list_lock, flags);
list_add(&spu->list, &spu_list[spu->node]);
list_add(&spu->full_list, &spu_full_list);
spin_unlock_irqrestore(&spu_list_lock, flags);
mutex_unlock(&spu_mutex);
spu->stats.utilization_state = SPU_UTIL_IDLE;
spu->stats.tstamp = jiffies;
goto out;
out_free_irqs:
spu_free_irqs(spu);
out_destroy:
spu_destroy_spu(spu);
out_free:
kfree(spu);
out:
return ret;
}
static const char *spu_state_names[] = {
"user", "system", "iowait", "idle"
};
static unsigned long long spu_acct_time(struct spu *spu,
enum spu_utilization_state state)
{
unsigned long long time = spu->stats.times[state];
if (spu->stats.utilization_state == state)
time += jiffies - spu->stats.tstamp;
return jiffies_to_msecs(time);
}
static ssize_t spu_stat_show(struct sys_device *sysdev, char *buf)
{
struct spu *spu = container_of(sysdev, struct spu, sysdev);
return sprintf(buf, "%s %llu %llu %llu %llu "
"%llu %llu %llu %llu %llu %llu %llu %llu\n",
spu_state_names[spu->stats.utilization_state],
spu_acct_time(spu, SPU_UTIL_USER),
spu_acct_time(spu, SPU_UTIL_SYSTEM),
spu_acct_time(spu, SPU_UTIL_IOWAIT),
spu_acct_time(spu, SPU_UTIL_IDLE),
spu->stats.vol_ctx_switch,
spu->stats.invol_ctx_switch,
spu->stats.slb_flt,
spu->stats.hash_flt,
spu->stats.min_flt,
spu->stats.maj_flt,
spu->stats.class2_intr,
spu->stats.libassist);
}
static SYSDEV_ATTR(stat, 0644, spu_stat_show, NULL);
static int __init init_spu_base(void)
{
int i, ret = 0;
for (i = 0; i < MAX_NUMNODES; i++)
INIT_LIST_HEAD(&spu_list[i]);
if (!spu_management_ops)
goto out;
/* create sysdev class for spus */
ret = sysdev_class_register(&spu_sysdev_class);
if (ret)
goto out;
ret = spu_enumerate_spus(create_spu);
if (ret < 0) {
printk(KERN_WARNING "%s: Error initializing spus\n",
__FUNCTION__);
goto out_unregister_sysdev_class;
}
if (ret > 0) {
/*
* We cannot put the forward declaration in
* <linux/linux_logo.h> because of conflicting session type
* conflicts for const and __initdata with different compiler
* versions
*/
extern const struct linux_logo logo_spe_clut224;
fb_append_extra_logo(&logo_spe_clut224, ret);
}
xmon_register_spus(&spu_full_list);
spu_add_sysdev_attr(&attr_stat);
return 0;
out_unregister_sysdev_class:
sysdev_class_unregister(&spu_sysdev_class);
out:
return ret;
}
module_init(init_spu_base);
MODULE_LICENSE("GPL");
MODULE_AUTHOR("Arnd Bergmann <arndb@de.ibm.com>");