android_kernel_samsung_sm86.../qcom/opensource/graphics-kernel/adreno_snapshot.c
David Wronek 880d405719 Add 'qcom/opensource/graphics-kernel/' from commit 'b4fdc4c04295ac59109ae19d64747522740c3f14'
git-subtree-dir: qcom/opensource/graphics-kernel
git-subtree-mainline: 992813d9c1
git-subtree-split: b4fdc4c042
Change-Id:
repo: https://git.codelinaro.org/clo/la/platform/vendor/qcom/opensource/graphics-kernel
tag: GRAPHICS.LA.14.0.r1-07700-lanai.0
2024-10-06 16:44:56 +02:00

1336 lines
37 KiB
C

// SPDX-License-Identifier: GPL-2.0-only
/*
* Copyright (c) 2012-2021, The Linux Foundation. All rights reserved.
* Copyright (c) 2022-2024, Qualcomm Innovation Center, Inc. All rights reserved.
*/
#include <linux/utsname.h>
#include "adreno.h"
#include "adreno_cp_parser.h"
#include "adreno_pm4types.h"
#include "adreno_snapshot.h"
/* Maintain a list of the objects we see during parsing */
#define SNAPSHOT_OBJ_BUFSIZE 64
/* Used to print error message if an IB has too many objects in it */
static int ib_max_objs;
struct snapshot_rb_params {
struct kgsl_snapshot *snapshot;
struct adreno_ringbuffer *rb;
};
/* Keep track of how many bytes are frozen after a snapshot and tell the user */
static size_t snapshot_frozen_objsize;
static struct kgsl_snapshot_object objbuf[SNAPSHOT_OBJ_BUFSIZE];
/* Pointer to the next open entry in the object list */
static unsigned int objbufptr;
static inline int adreno_rb_ctxtswitch(struct adreno_device *adreno_dev,
unsigned int *cmd)
{
return cmd[0] == cp_packet(adreno_dev, CP_NOP, 1) &&
cmd[1] == CONTEXT_TO_MEM_IDENTIFIER;
}
/* Push a new buffer object onto the list */
void kgsl_snapshot_push_object(struct kgsl_device *device,
struct kgsl_process_private *process,
uint64_t gpuaddr, uint64_t dwords)
{
int index;
struct kgsl_mem_entry *entry;
if (process == NULL || gpuaddr == 0)
return;
/*
* Sometimes IBs can be reused in the same dump. Because we parse from
* oldest to newest, if we come across an IB that has already been used,
* assume that it has been reused and update the list with the newest
* size.
*/
for (index = 0; index < objbufptr; index++) {
if (objbuf[index].gpuaddr == gpuaddr &&
objbuf[index].entry->priv == process) {
/*
* Check if newly requested size is within the
* allocated range or not, otherwise continue
* with previous size.
*/
if (!kgsl_gpuaddr_in_memdesc(
&objbuf[index].entry->memdesc,
gpuaddr, dwords << 2)) {
dev_err(device->dev,
"snapshot: gpuaddr 0x%016llX size is less than requested\n",
gpuaddr);
return;
}
objbuf[index].size = max_t(uint64_t,
objbuf[index].size,
dwords << 2);
return;
}
}
if (objbufptr == SNAPSHOT_OBJ_BUFSIZE) {
dev_err(device->dev, "snapshot: too many snapshot objects\n");
return;
}
entry = kgsl_sharedmem_find(process, gpuaddr);
if (entry == NULL) {
dev_err(device->dev,
"snapshot: Can't find entry for 0x%016llX\n", gpuaddr);
return;
}
/*
* In some gpu fault scenarios incorrect dword size resulting to return
* without putting IB addrs into the list. Hence update IB dword size
* within memdesc size to have IB dump in snapshot.
*/
if ((gpuaddr + (dwords << 2)) > (entry->memdesc.gpuaddr + entry->memdesc.size))
dwords = (entry->memdesc.size - (gpuaddr - entry->memdesc.gpuaddr)) >> 2;
if (!kgsl_gpuaddr_in_memdesc(&entry->memdesc, gpuaddr, dwords << 2)) {
dev_err(device->dev,
"snapshot: Mem entry 0x%016llX is too small\n",
gpuaddr);
kgsl_mem_entry_put(entry);
return;
}
/* Put it on the list of things to parse */
objbuf[objbufptr].gpuaddr = gpuaddr;
objbuf[objbufptr].size = dwords << 2;
objbuf[objbufptr++].entry = entry;
}
/*
* Returns index of the specified object is already on the list of buffers
* to be dumped
*/
static int find_object(uint64_t gpuaddr, struct kgsl_process_private *process)
{
int index;
for (index = 0; index < objbufptr; index++) {
if (objbuf[index].gpuaddr == gpuaddr &&
objbuf[index].entry->priv == process)
return index;
}
return -ENOENT;
}
/*
* snapshot_freeze_obj_list() - Take a list of ib objects and freeze their
* memory for snapshot
* @snapshot: The snapshot data.
* @process: The process to which the IB belongs
* @ib_obj_list: List of the IB objects
*
* Returns 0 on success else error code
*/
static int snapshot_freeze_obj_list(struct kgsl_snapshot *snapshot,
struct kgsl_process_private *process,
struct adreno_ib_object_list *ib_obj_list)
{
int ret = 0;
struct adreno_ib_object *ib_objs;
int i;
for (i = 0; i < ib_obj_list->num_objs; i++) {
int temp_ret;
int index;
int freeze = 1;
ib_objs = &(ib_obj_list->obj_list[i]);
/* Make sure this object is not going to be saved statically */
for (index = 0; index < objbufptr; index++) {
if ((objbuf[index].gpuaddr <= ib_objs->gpuaddr) &&
((objbuf[index].gpuaddr +
(objbuf[index].size)) >=
(ib_objs->gpuaddr + ib_objs->size)) &&
(objbuf[index].entry->priv == process)) {
freeze = 0;
objbuf[index].entry->memdesc.priv &=
~KGSL_MEMDESC_SKIP_RECLAIM;
break;
}
}
if (freeze) {
temp_ret = kgsl_snapshot_get_object(snapshot,
process, ib_objs->gpuaddr,
ib_objs->size,
ib_objs->snapshot_obj_type);
if (temp_ret < 0) {
if (ret >= 0)
ret = temp_ret;
} else {
snapshot_frozen_objsize += temp_ret;
}
}
}
return ret;
}
static inline bool adreno_ib_addr_overlap(u64 ibaddr, u64 gpuaddr, u64 dwords)
{
u64 size = dwords << 2;
if (size > (U64_MAX - gpuaddr))
return false;
return (ibaddr >= gpuaddr && ibaddr < (gpuaddr + size));
}
void adreno_parse_ib(struct kgsl_device *device,
struct kgsl_snapshot *snapshot,
struct kgsl_process_private *process,
uint64_t gpuaddr, uint64_t dwords)
{
struct adreno_ib_object_list *ib_obj_list;
/*
* Check the IB address - if it is either the last executed IB1
* then push it into the static blob otherwise put it in the dynamic
* list
*/
if (adreno_ib_addr_overlap(snapshot->ib1base, gpuaddr, dwords)) {
/*
* During restore after preemption, ib1base in the register
* can be updated by CP. In such scenarios, to dump complete
* IB1 in snapshot, we should consider ib1base from ringbuffer.
*/
snapshot->ib1base = gpuaddr;
snapshot->ib1size = dwords;
kgsl_snapshot_push_object(device, process, gpuaddr, dwords);
return;
}
if (kgsl_snapshot_have_object(snapshot, process,
gpuaddr, dwords << 2))
return;
if (-E2BIG == adreno_ib_create_object_list(device, process,
gpuaddr, dwords, snapshot->ib2base,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list)
kgsl_snapshot_add_ib_obj_list(snapshot, ib_obj_list);
}
void adreno_parse_ib_lpac(struct kgsl_device *device,
struct kgsl_snapshot *snapshot,
struct kgsl_process_private *process,
u64 gpuaddr, u64 dwords)
{
struct adreno_ib_object_list *ib_obj_list;
/*
* Check the IB address - if it is either the last executed IB1
* then push it into the static blob otherwise put it in the dynamic
* list
*/
if (adreno_ib_addr_overlap(snapshot->ib1base_lpac, gpuaddr, dwords)) {
/*
* During restore after preemption, ib1base in the register
* can be updated by CP. In such scenarios, to dump complete
* IB1 in snapshot, we should consider ib1base from ringbuffer.
*/
snapshot->ib1base_lpac = gpuaddr;
snapshot->ib1size_lpac = dwords;
kgsl_snapshot_push_object(device, process, gpuaddr, dwords);
return;
}
if (kgsl_snapshot_have_object(snapshot, process,
gpuaddr, dwords << 2))
return;
if (-E2BIG == adreno_ib_create_object_list(device, process,
gpuaddr, dwords, snapshot->ib2base_lpac,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list)
kgsl_snapshot_add_ib_obj_list(snapshot, ib_obj_list);
}
void adreno_snapshot_dump_all_ibs(struct kgsl_device *device,
unsigned int *rbptr, struct kgsl_snapshot *snapshot)
{
int index = 0;
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
struct kgsl_iommu *iommu = KGSL_IOMMU(device);
for (index = 0; index < KGSL_RB_DWORDS;) {
if (adreno_cmd_is_ib(adreno_dev, rbptr[index])) {
uint64_t ibaddr;
uint64_t ibsize;
if (ADRENO_LEGACY_PM4(adreno_dev)) {
ibaddr = rbptr[index + 1];
ibsize = rbptr[index + 2];
index += 3;
} else {
ibaddr = rbptr[index + 2];
ibaddr = ibaddr << 32 | rbptr[index + 1];
ibsize = rbptr[index + 3];
index += 4;
}
/* Don't parse known global IBs */
if (kgsl_gpuaddr_in_memdesc(iommu->setstate,
ibaddr, ibsize))
continue;
if (kgsl_gpuaddr_in_memdesc(adreno_dev->pwron_fixup,
ibaddr, ibsize))
continue;
adreno_parse_ib(device, snapshot, snapshot->process,
ibaddr, ibsize);
} else
index = index + 1;
}
}
/**
* snapshot_rb_ibs() - Dump rb data and capture the IB's in the RB as well
* @device: Pointer to a KGSL device
* @rb: The RB to dump
* @data: Pointer to memory where the RB data is to be dumped
* @snapshot: Pointer to information about the current snapshot being taken
*/
static void snapshot_rb_ibs(struct kgsl_device *device,
struct adreno_ringbuffer *rb,
struct kgsl_snapshot *snapshot)
{
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
unsigned int *rbptr, rptr = adreno_get_rptr(rb);
int index, i;
int parse_ibs = 0, ib_parse_start;
/*
* During IB parse, vmalloc is called which can sleep and
* should not be called from atomic context. Since IBs are not
* dumped during atomic snapshot, there is no need to parse it.
*/
if (device->snapshot_atomic)
return;
rbptr = rb->buffer_desc->hostptr;
/*
* KGSL tries to dump the active IB first. If it is not present, then
* only it dumps all the IBs. In few cases, non-active IBs may help to
* establish the flow and understand the hardware state better.
*/
if (device->dump_all_ibs) {
adreno_snapshot_dump_all_ibs(device, rbptr, snapshot);
return;
}
/*
* Figure out the window of ringbuffer data to dump. First we need to
* find where the last processed IB ws submitted. Start walking back
* from the rptr
*/
index = rptr;
do {
index--;
if (index < 0) {
if (ADRENO_LEGACY_PM4(adreno_dev))
index = KGSL_RB_DWORDS - 3;
else
index = KGSL_RB_DWORDS - 4;
/* We wrapped without finding what we wanted */
if (index < rb->wptr) {
index = rb->wptr;
break;
}
}
if (adreno_cmd_is_ib(adreno_dev, rbptr[index])) {
uint64_t ibaddr;
uint64_t ibsize;
if (ADRENO_LEGACY_PM4(adreno_dev)) {
ibaddr = rbptr[index + 1];
ibsize = rbptr[index + 2];
} else {
ibaddr = rbptr[index + 2];
ibaddr = ibaddr << 32 | rbptr[index + 1];
ibsize = rbptr[index + 3];
}
if (adreno_ib_addr_overlap(snapshot->ib1base,
ibaddr, ibsize)) {
/*
* During restore after preemption, ib1base in
* the register can be updated by CP. In such
* scenario, to dump complete IB1 in snapshot,
* we should consider ib1base from ringbuffer.
*/
snapshot->ib1base = ibaddr;
snapshot->ib1size = ibsize;
break;
}
}
} while (index != rb->wptr);
/*
* If the ib1 was not found, for example, if ib1base was restored
* incorrectly after preemption, then simply dump the entire
* ringbuffer along with all the IBs in the ringbuffer.
*/
if (index == rb->wptr) {
adreno_snapshot_dump_all_ibs(device, rb->buffer_desc->hostptr, snapshot);
return;
}
/*
* index points at the last submitted IB. We can only trust that the
* memory between the context switch and the hanging IB is valid, so
* the next step is to find the context switch before the submission
*/
while (index != rb->wptr) {
index--;
if (index < 0) {
index = KGSL_RB_DWORDS - 2;
/*
* Wrapped without finding the context switch. This is
* harmless - we should still have enough data to dump a
* valid state
*/
if (index < rb->wptr) {
index = rb->wptr;
break;
}
}
/* Break if the current packet is a context switch identifier */
if ((rbptr[index] == cp_packet(adreno_dev, CP_NOP, 1)) &&
(rbptr[index + 1] == CONTEXT_TO_MEM_IDENTIFIER))
break;
}
/*
* Index represents the start of the window of interest. We will try
* to dump all buffers between here and the rptr
*/
ib_parse_start = index;
/*
* Loop through the RB, looking for indirect buffers and MMU pagetable
* changes
*/
index = rb->wptr;
for (i = 0; i < KGSL_RB_DWORDS; i++) {
/*
* Only parse IBs between the start and the rptr or the next
* context switch, whichever comes first
*/
if (parse_ibs == 0 && index == ib_parse_start)
parse_ibs = 1;
else if (index == rptr || adreno_rb_ctxtswitch(adreno_dev,
&rbptr[index]))
parse_ibs = 0;
if (parse_ibs && adreno_cmd_is_ib(adreno_dev, rbptr[index])) {
struct kgsl_iommu *iommu = KGSL_IOMMU(device);
uint64_t ibaddr;
uint64_t ibsize;
if (ADRENO_LEGACY_PM4(adreno_dev)) {
ibaddr = rbptr[index + 1];
ibsize = rbptr[index + 2];
} else {
ibaddr = rbptr[index + 2];
ibaddr = ibaddr << 32 | rbptr[index + 1];
ibsize = rbptr[index + 3];
}
index = (index + 1) % KGSL_RB_DWORDS;
/* Don't parse known global IBs */
if (kgsl_gpuaddr_in_memdesc(iommu->setstate,
ibaddr, ibsize))
continue;
if (kgsl_gpuaddr_in_memdesc(adreno_dev->pwron_fixup,
ibaddr, ibsize))
continue;
adreno_parse_ib(device, snapshot, snapshot->process,
ibaddr, ibsize);
} else
index = (index + 1) % KGSL_RB_DWORDS;
}
}
/* Snapshot the ringbuffer memory */
static size_t snapshot_rb(struct kgsl_device *device, u8 *buf,
size_t remain, void *priv)
{
struct kgsl_snapshot_rb_v2 *header = (struct kgsl_snapshot_rb_v2 *)buf;
unsigned int *data = (unsigned int *)(buf + sizeof(*header));
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
struct snapshot_rb_params *snap_rb_params = priv;
struct kgsl_snapshot *snapshot = snap_rb_params->snapshot;
struct adreno_ringbuffer *rb = snap_rb_params->rb;
/*
* Dump the entire ringbuffer - the parser can choose how much of it to
* process
*/
if (remain < KGSL_RB_SIZE + sizeof(*header)) {
dev_err(device->dev,
"snapshot: Not enough memory for the rb section\n");
return 0;
}
/* Write the sub-header for the section */
header->start = 0;
header->end = KGSL_RB_DWORDS;
header->wptr = rb->wptr;
header->rptr = adreno_get_rptr(rb);
header->rbsize = KGSL_RB_DWORDS;
header->count = KGSL_RB_DWORDS;
adreno_rb_readtimestamp(adreno_dev, rb, KGSL_TIMESTAMP_QUEUED,
&header->timestamp_queued);
adreno_rb_readtimestamp(adreno_dev, rb, KGSL_TIMESTAMP_RETIRED,
&header->timestamp_retired);
header->gpuaddr = rb->buffer_desc->gpuaddr;
header->id = rb->id;
if (rb == adreno_dev->cur_rb)
snapshot_rb_ibs(device, rb, snapshot);
/* Just copy the ringbuffer, there are no active IBs */
memcpy(data, rb->buffer_desc->hostptr, KGSL_RB_SIZE);
/* Return the size of the section */
return KGSL_RB_SIZE + sizeof(*header);
}
static int _count_mem_entries(int id, void *ptr, void *data)
{
int *count = data;
*count = *count + 1;
return 0;
}
struct mem_entry {
uint64_t gpuaddr;
uint64_t size;
unsigned int type;
} __packed;
static size_t snapshot_capture_mem_list(struct kgsl_device *device,
u8 *buf, size_t remain, void *priv)
{
struct kgsl_snapshot_mem_list_v2 *header =
(struct kgsl_snapshot_mem_list_v2 *)buf;
int id, index = 0, ret = 0, num_mem = 0;
struct kgsl_process_private *process = priv;
struct mem_entry *m = (struct mem_entry *)(buf + sizeof(*header));
struct kgsl_mem_entry *entry;
/* we need a process to search! */
if (process == NULL)
return 0;
spin_lock(&process->mem_lock);
/* We need to know the number of memory objects that the process has */
idr_for_each(&process->mem_idr, _count_mem_entries, &num_mem);
if (num_mem == 0)
goto out;
if (remain < ((num_mem * sizeof(struct mem_entry)) + sizeof(*header))) {
dev_err(device->dev,
"snapshot: Not enough memory for the mem list\n");
goto out;
}
header->num_entries = num_mem;
header->ptbase = kgsl_mmu_pagetable_get_ttbr0(process->pagetable);
/*
* Walk through the memory list and store the
* tuples(gpuaddr, size, memtype) in snapshot
*/
idr_for_each_entry(&process->mem_idr, entry, id) {
m[index].gpuaddr = entry->memdesc.gpuaddr;
m[index].size = entry->memdesc.size;
m[index].type = kgsl_memdesc_get_memtype(&entry->memdesc);
index++;
}
ret = sizeof(*header) + (num_mem * sizeof(struct mem_entry));
out:
spin_unlock(&process->mem_lock);
return ret;
}
struct snapshot_ib_meta {
struct kgsl_snapshot *snapshot;
struct kgsl_snapshot_object *obj;
uint64_t ib1base;
uint64_t ib1size;
uint64_t ib2base;
uint64_t ib2size;
u64 ib1base_lpac;
u64 ib1size_lpac;
u64 ib2base_lpac;
u64 ib2size_lpac;
};
static void kgsl_snapshot_add_active_ib_obj_list(struct kgsl_device *device,
struct kgsl_snapshot *snapshot)
{
struct adreno_ib_object_list *ib_obj_list;
int index = -ENOENT;
if (!snapshot->ib1dumped)
index = find_object(snapshot->ib1base, snapshot->process);
/* only do this for IB1 because the IB2's are part of IB1 objects */
if ((index != -ENOENT) &&
(snapshot->ib1base == objbuf[index].gpuaddr)) {
if (-E2BIG == adreno_ib_create_object_list(device,
objbuf[index].entry->priv,
objbuf[index].gpuaddr,
objbuf[index].size >> 2,
snapshot->ib2base,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list) {
/* freeze the IB objects in the IB */
snapshot_freeze_obj_list(snapshot,
objbuf[index].entry->priv,
ib_obj_list);
adreno_ib_destroy_obj_list(ib_obj_list);
}
} else {
/* Get the IB2 index from parsed object */
index = find_object(snapshot->ib2base, snapshot->process);
if (index != -ENOENT)
adreno_parse_ib(device, snapshot, snapshot->process,
snapshot->ib2base, objbuf[index].size >> 2);
}
}
static void kgsl_snapshot_add_active_lpac_ib_obj_list(struct kgsl_device *device,
struct kgsl_snapshot *snapshot)
{
struct adreno_ib_object_list *ib_obj_list;
int index = -ENOENT;
if (!snapshot->ib1dumped_lpac)
index = find_object(snapshot->ib1base_lpac, snapshot->process_lpac);
/* only do this for IB1 because the IB2's are part of IB1 objects */
if ((index != -ENOENT) &&
(snapshot->ib1base_lpac == objbuf[index].gpuaddr)) {
if (-E2BIG == adreno_ib_create_object_list(device,
objbuf[index].entry->priv,
objbuf[index].gpuaddr,
objbuf[index].size >> 2,
snapshot->ib2base_lpac,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list) {
/* freeze the IB objects in the IB */
snapshot_freeze_obj_list(snapshot,
objbuf[index].entry->priv,
ib_obj_list);
adreno_ib_destroy_obj_list(ib_obj_list);
}
} else {
/* Get the IB2 index from parsed object */
index = find_object(snapshot->ib2base_lpac, snapshot->process_lpac);
if (index != -ENOENT)
adreno_parse_ib_lpac(device, snapshot, snapshot->process_lpac,
snapshot->ib2base_lpac, objbuf[index].size >> 2);
}
}
/*
* active_ib_is_parsed() - Checks if active ib is already parsed
* @gpuaddr: Active IB base address at the time of fault
* @size: Active IB size
* @process: The process to which the IB belongs
*
* Function returns true if the active is already is parsed
* else false
*/
static bool active_ib_is_parsed(uint64_t gpuaddr, uint64_t size,
struct kgsl_process_private *process)
{
int index;
/* go through the static list for gpuaddr is in list or not */
for (index = 0; index < objbufptr; index++) {
if ((objbuf[index].gpuaddr <= gpuaddr) &&
((objbuf[index].gpuaddr +
(objbuf[index].size)) >=
(gpuaddr + size)) &&
(objbuf[index].entry->priv == process))
return true;
}
return false;
}
/* Snapshot the memory for an indirect buffer */
static size_t snapshot_ib(struct kgsl_device *device, u8 *buf,
size_t remain, void *priv)
{
struct kgsl_snapshot_ib_v2 *header = (struct kgsl_snapshot_ib_v2 *)buf;
struct snapshot_ib_meta *metadata = priv;
unsigned int *src;
unsigned int *dst = (unsigned int *)(buf + sizeof(*header));
struct adreno_ib_object_list *ib_obj_list;
struct kgsl_snapshot *snapshot;
struct kgsl_snapshot_object *obj;
struct kgsl_memdesc *memdesc;
if (metadata == NULL || metadata->snapshot == NULL || metadata->obj == NULL) {
dev_err(device->dev, "snapshot: bad metadata\n");
return 0;
}
snapshot = metadata->snapshot;
obj = metadata->obj;
memdesc = &obj->entry->memdesc;
/* If size is zero get it from the medesc size */
if (!obj->size)
obj->size = (memdesc->size - (obj->gpuaddr - memdesc->gpuaddr));
if (remain < (obj->size + sizeof(*header))) {
dev_err(device->dev, "snapshot: Not enough memory for the ib\n");
return 0;
}
src = kgsl_gpuaddr_to_vaddr(memdesc, obj->gpuaddr);
if (src == NULL) {
dev_err(device->dev,
"snapshot: Unable to map GPU memory object 0x%016llX into the kernel\n",
obj->gpuaddr);
return 0;
}
/* only do this for IB1 because the IB2's are part of IB1 objects */
if (metadata->ib1base == obj->gpuaddr) {
snapshot->ib1dumped = active_ib_is_parsed(obj->gpuaddr,
obj->size, obj->entry->priv);
if (-E2BIG == adreno_ib_create_object_list(device,
obj->entry->priv,
obj->gpuaddr, obj->size >> 2,
snapshot->ib2base,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list) {
/* freeze the IB objects in the IB */
snapshot_freeze_obj_list(snapshot,
obj->entry->priv,
ib_obj_list);
adreno_ib_destroy_obj_list(ib_obj_list);
}
}
if (metadata->ib2base == obj->gpuaddr)
snapshot->ib2dumped = active_ib_is_parsed(obj->gpuaddr,
obj->size, obj->entry->priv);
if (metadata->ib1base_lpac == obj->gpuaddr) {
snapshot->ib1dumped_lpac = active_ib_is_parsed(obj->gpuaddr,
obj->size, obj->entry->priv);
if (-E2BIG == adreno_ib_create_object_list(device,
obj->entry->priv,
obj->gpuaddr, obj->size >> 2,
snapshot->ib2base_lpac,
&ib_obj_list))
ib_max_objs = 1;
if (ib_obj_list) {
/* freeze the IB objects in the IB */
snapshot_freeze_obj_list(snapshot,
obj->entry->priv,
ib_obj_list);
adreno_ib_destroy_obj_list(ib_obj_list);
}
}
if (metadata->ib2base_lpac == obj->gpuaddr)
snapshot->ib2dumped_lpac = active_ib_is_parsed(obj->gpuaddr,
obj->size, obj->entry->priv);
/* Write the sub-header for the section */
header->gpuaddr = obj->gpuaddr;
header->ptbase =
kgsl_mmu_pagetable_get_ttbr0(obj->entry->priv->pagetable);
header->size = obj->size >> 2;
/* Write the contents of the ib */
memcpy((void *)dst, (void *)src, (size_t) obj->size);
/* Write the contents of the ib */
return obj->size + sizeof(*header);
}
/* Dump another item on the current pending list */
static void dump_object(struct kgsl_device *device, int obj,
struct kgsl_snapshot *snapshot)
{
struct snapshot_ib_meta metadata;
metadata.snapshot = snapshot;
metadata.obj = &objbuf[obj];
metadata.ib1base = snapshot->ib1base;
metadata.ib1size = snapshot->ib1size;
metadata.ib2base = snapshot->ib2base;
metadata.ib2size = snapshot->ib2size;
metadata.ib1base_lpac = snapshot->ib1base_lpac;
metadata.ib1size_lpac = snapshot->ib1size_lpac;
metadata.ib2base_lpac = snapshot->ib2base_lpac;
metadata.ib2size_lpac = snapshot->ib2size_lpac;
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_IB_V2,
snapshot, snapshot_ib, &metadata);
if (objbuf[obj].entry) {
kgsl_memdesc_unmap(&(objbuf[obj].entry->memdesc));
kgsl_mem_entry_put(objbuf[obj].entry);
}
}
/* setup_fault process - Find kgsl_process_private struct that caused the fault
*
* Find the faulting process based what the dispatcher thinks happened and
* what the hardware is using for the current pagetable. The process struct
* will be used to look up GPU addresses that are encountered while parsing
* the GPU state.
*/
static struct kgsl_process_private *setup_fault_process(struct kgsl_device *device,
struct kgsl_snapshot *snapshot,
struct kgsl_context *context)
{
u64 hw_ptbase, proc_ptbase;
struct kgsl_process_private *process = context ? context->proc_priv : NULL;
if (process != NULL && !kgsl_process_private_get(process))
process = NULL;
/* Get the physical address of the MMU pagetable */
hw_ptbase = kgsl_mmu_get_current_ttbr0(&device->mmu, context);
/* if we have an input process, make sure the ptbases match */
if (process) {
int asid = kgsl_mmu_pagetable_get_asid(process->pagetable, context);
proc_ptbase = kgsl_mmu_pagetable_get_ttbr0(process->pagetable);
if (asid >= 0)
proc_ptbase |= FIELD_PREP(GENMASK_ULL(63, KGSL_IOMMU_ASID_START_BIT), asid);
/* agreement! No need to check further */
if (hw_ptbase == proc_ptbase)
goto done;
kgsl_process_private_put(process);
process = NULL;
dev_err(device->dev,
"snapshot: ptbase mismatch hw %llx sw %llx\n",
hw_ptbase, proc_ptbase);
}
/* try to find the right pagetable by walking the process list */
if (kgsl_mmu_is_perprocess(&device->mmu)) {
struct kgsl_process_private *tmp;
read_lock(&kgsl_driver.proclist_lock);
list_for_each_entry(tmp, &kgsl_driver.process_list, list) {
u64 pt_ttbr0;
pt_ttbr0 = kgsl_mmu_pagetable_get_ttbr0(tmp->pagetable);
if ((pt_ttbr0 == MMU_SW_PT_BASE(hw_ptbase))
&& kgsl_process_private_get(tmp)) {
process = tmp;
break;
}
}
read_unlock(&kgsl_driver.proclist_lock);
}
done:
return process;
}
/* Snapshot a global memory buffer */
size_t adreno_snapshot_global(struct kgsl_device *device, u8 *buf,
size_t remain, void *priv)
{
struct kgsl_memdesc *memdesc = priv;
struct kgsl_snapshot_gpu_object_v2 *header =
(struct kgsl_snapshot_gpu_object_v2 *)buf;
u8 *ptr = buf + sizeof(*header);
if (IS_ERR_OR_NULL(memdesc) || memdesc->size == 0)
return 0;
if (remain < (memdesc->size + sizeof(*header))) {
dev_err(device->dev,
"snapshot: Not enough memory for the memdesc\n");
return 0;
}
if (memdesc->hostptr == NULL) {
dev_err(device->dev,
"snapshot: no kernel mapping for global object 0x%016llX\n",
memdesc->gpuaddr);
return 0;
}
header->size = memdesc->size >> 2;
header->gpuaddr = memdesc->gpuaddr;
header->ptbase = MMU_DEFAULT_TTBR0(device);
header->type = SNAPSHOT_GPU_OBJECT_GLOBAL;
memcpy(ptr, memdesc->hostptr, memdesc->size);
return memdesc->size + sizeof(*header);
}
/* Snapshot IOMMU specific buffers */
static void adreno_snapshot_iommu(struct kgsl_device *device,
struct kgsl_snapshot *snapshot)
{
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
struct kgsl_iommu *iommu = KGSL_IOMMU(device);
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global, iommu->setstate);
if (adreno_is_preemption_enabled(adreno_dev))
kgsl_snapshot_add_section(device,
KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global, iommu->smmu_info);
}
static void adreno_snapshot_ringbuffer(struct kgsl_device *device,
struct kgsl_snapshot *snapshot, struct adreno_ringbuffer *rb)
{
struct snapshot_rb_params params = {
.snapshot = snapshot,
.rb = rb,
};
if (rb == NULL)
return;
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_RB_V2, snapshot,
snapshot_rb, &params);
}
static void adreno_snapshot_os(struct kgsl_device *device,
struct kgsl_snapshot *snapshot, struct kgsl_context *guilty,
struct kgsl_context *guilty_lpac)
{
struct kgsl_snapshot_section_header *sect =
(struct kgsl_snapshot_section_header *) snapshot->ptr;
struct kgsl_snapshot_linux_v4 *header = (struct kgsl_snapshot_linux_v4 *)
(snapshot->ptr + sizeof(*sect));
struct kgsl_context *context;
u32 remain;
void *mem;
int id;
if (snapshot->remain < (sizeof(*sect) + sizeof(*header))) {
SNAPSHOT_ERR_NOMEM(device, "OS");
return;
}
header->osid = KGSL_SNAPSHOT_OS_LINUX_V4;
strscpy(header->release, init_utsname()->release, sizeof(header->release));
strscpy(header->version, init_utsname()->version, sizeof(header->version));
header->seconds = ktime_get_real_seconds();
header->power_flags = device->pwrctrl.power_flags;
header->power_level = device->pwrctrl.active_pwrlevel;
header->power_interval_timeout = device->pwrctrl.interval_timeout;
header->grpclk = clk_get_rate(device->pwrctrl.grp_clks[0]);
/* Get the current PT base */
header->ptbase = kgsl_mmu_get_current_ttbr0(&device->mmu, guilty);
header->ptbase_lpac = kgsl_mmu_get_current_ttbr0(&device->mmu, guilty_lpac);
header->ctxtcount = 0;
/* If we know the guilty context then dump it */
if (guilty) {
header->current_context = guilty->id;
header->pid = guilty->tid;
strscpy(header->comm, guilty->proc_priv->comm,
sizeof(header->comm));
}
if (guilty_lpac) {
header->current_context_lpac = guilty_lpac->id;
header->pid_lpac = guilty_lpac->tid;
strscpy(header->comm_lpac, guilty_lpac->proc_priv->comm,
sizeof(header->comm_lpac));
}
remain = snapshot->remain - sizeof(*sect) + sizeof(*header);
mem = snapshot->ptr + sizeof(*sect) + sizeof(*header);
read_lock(&device->context_lock);
idr_for_each_entry(&device->context_idr, context, id) {
struct kgsl_snapshot_linux_context_v2 *c = mem;
if (remain < sizeof(*c))
break;
kgsl_readtimestamp(device, context, KGSL_TIMESTAMP_QUEUED,
&c->timestamp_queued);
kgsl_readtimestamp(device, context, KGSL_TIMESTAMP_CONSUMED,
&c->timestamp_consumed);
kgsl_readtimestamp(device, context, KGSL_TIMESTAMP_RETIRED,
&c->timestamp_retired);
header->ctxtcount++;
mem += sizeof(*c);
remain -= sizeof(*c);
}
read_unlock(&device->context_lock);
sect->magic = SNAPSHOT_SECTION_MAGIC;
sect->id = KGSL_SNAPSHOT_SECTION_OS;
sect->size = sizeof(*sect) + sizeof(*header) +
header->ctxtcount * sizeof(struct kgsl_snapshot_linux_context_v2);
snapshot->ptr += sect->size;
snapshot->remain -= sect->size;
snapshot->size += sect->size;
}
static void adreno_static_ib_dump(struct kgsl_device *device,
struct kgsl_process_private *process, u64 ib1base,
u32 ib1size, u64 ib2base, u32 ib2size)
{
if (process == NULL)
return;
/* Make sure that the last IB1 that was being executed is dumped.
* Since this was the last IB1 that was processed, we should have
* already added it to the list during the ringbuffer parse but we
* want to be double plus sure.
* The problem is that IB size from the register is the unprocessed size
* of the buffer not the original size, so if we didn't catch this
* buffer being directly used in the RB, then we might not be able to
* dump the whole thing. Try to dump the maximum possible size from the
* IB1 base address till the end of memdesc size so that we dont miss
* what we are interested in. Print a warning message so we can try to
* figure how often this really happens.
*/
if (ib1base && (-ENOENT == find_object(ib1base, process))) {
struct kgsl_mem_entry *entry;
u64 ibsize;
entry = kgsl_sharedmem_find(process, ib1base);
if (entry == NULL) {
dev_err(device->dev,
"Can't find a memory entry containing IB1BASE %16llx\n",
ib1base);
} else {
ibsize = entry->memdesc.size -
(ib1base - entry->memdesc.gpuaddr);
kgsl_mem_entry_put(entry);
kgsl_snapshot_push_object(device, process,
ib1base, ibsize >> 2);
dev_err(device->dev,
"CP_IB1_BASE %16llx is not found in the ringbuffer. Dumping %llx dwords of the buffer\n",
ib1base, ibsize >> 2);
}
}
/*
* Add the last parsed IB2 to the list. The IB2 should be found as we
* parse the objects below, but we try to add it to the list first, so
* it too can be parsed. Don't print an error message in this case - if
* the IB2 is found during parsing, the list will be updated with the
* correct size.
*/
if (ib2base && (-ENOENT == find_object(ib2base, process)))
kgsl_snapshot_push_object(device, process, ib2base, ib2size);
}
/* adreno_snapshot - Snapshot the Adreno GPU state
* @device - KGSL device to snapshot
* @snapshot - Pointer to the snapshot instance
* @context - context that caused the fault, if known by the driver
* This is a hook function called by kgsl_snapshot to snapshot the
* Adreno specific information for the GPU snapshot. In turn, this function
* calls the GPU specific snapshot function to get core specific information.
*/
void adreno_snapshot(struct kgsl_device *device, struct kgsl_snapshot *snapshot,
struct kgsl_context *context, struct kgsl_context *context_lpac)
{
unsigned int i;
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
const struct adreno_gpudev *gpudev = ADRENO_GPU_DEVICE(adreno_dev);
struct kgsl_snapshot_header *header = (struct kgsl_snapshot_header *)
snapshot->ptr;
/* Set up the master header */
header->magic = SNAPSHOT_MAGIC;
/* gpuid is deprecated so initialize it to an obviously wrong value */
header->gpuid = UINT_MAX;
header->chipid = adreno_dev->chipid;
snapshot->ptr += sizeof(*header);
snapshot->remain -= sizeof(*header);
snapshot->size += sizeof(*header);
/* Write the OS section */
adreno_snapshot_os(device, snapshot, context, context_lpac);
ib_max_objs = 0;
/* Reset the list of objects */
objbufptr = 0;
snapshot_frozen_objsize = 0;
snapshot->process = setup_fault_process(device, snapshot, context);
snapshot->process_lpac = setup_fault_process(device, snapshot, context_lpac);
/* Add GPU specific sections - registers mainly, but other stuff too */
if (gpudev->snapshot)
gpudev->snapshot(adreno_dev, snapshot);
snapshot->ib1dumped = false;
snapshot->ib2dumped = false;
snapshot->ib1dumped_lpac = false;
snapshot->ib2dumped_lpac = false;
adreno_snapshot_ringbuffer(device, snapshot, adreno_dev->cur_rb);
/* Dump the prev ringbuffer */
if (adreno_dev->prev_rb != adreno_dev->cur_rb)
adreno_snapshot_ringbuffer(device, snapshot,
adreno_dev->prev_rb);
if ((adreno_dev->next_rb != adreno_dev->prev_rb) &&
(adreno_dev->next_rb != adreno_dev->cur_rb))
adreno_snapshot_ringbuffer(device, snapshot,
adreno_dev->next_rb);
if (device->snapshot_atomic)
return;
/* Dump selected global buffers */
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global, device->memstore);
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global, device->scratch);
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global,
adreno_dev->pwron_fixup);
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_GPU_OBJECT_V2,
snapshot, adreno_snapshot_global,
adreno_dev->profile_buffer);
if (kgsl_mmu_get_mmutype(device) == KGSL_MMU_TYPE_IOMMU)
adreno_snapshot_iommu(device, snapshot);
/*
* Add a section that lists (gpuaddr, size, memtype) tuples of the
* hanging process
*/
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_MEMLIST_V2,
snapshot, snapshot_capture_mem_list, snapshot->process);
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_MEMLIST_V2,
snapshot, snapshot_capture_mem_list, snapshot->process_lpac);
adreno_static_ib_dump(device, snapshot->process,
snapshot->ib1base, snapshot->ib1size, snapshot->ib2base, snapshot->ib2size);
adreno_static_ib_dump(device, snapshot->process_lpac,
snapshot->ib1base_lpac, snapshot->ib1size_lpac,
snapshot->ib2base_lpac, snapshot->ib2size_lpac);
/*
* Go through the list of found objects and dump each one. As the IBs
* are parsed, more objects might be found, and objbufptr will increase
*/
for (i = 0; i < objbufptr; i++)
dump_object(device, i, snapshot);
/*
* Incase snapshot static blob is running out of memory, Add Active IB1
* and IB2 entries to obj_list so that active ib's can be dumped to
* snapshot dynamic blob.
*/
if (!snapshot->ib1dumped || !snapshot->ib2dumped)
kgsl_snapshot_add_active_ib_obj_list(device, snapshot);
if (ib_max_objs)
dev_err(device->dev, "Max objects found in GC IB\n");
if (!snapshot->ib1dumped_lpac || !snapshot->ib2dumped_lpac)
kgsl_snapshot_add_active_lpac_ib_obj_list(device, snapshot);
if (ib_max_objs)
dev_err(device->dev, "Max objects found in LPAC IB\n");
if (snapshot_frozen_objsize)
dev_err(device->dev,
"GPU snapshot froze %zdKb of GPU buffers\n",
snapshot_frozen_objsize / 1024);
}
void adreno_snapshot_registers(struct kgsl_device *device,
struct kgsl_snapshot *snapshot,
const unsigned int *regs, unsigned int count)
{
struct kgsl_snapshot_registers r;
r.regs = regs;
r.count = count;
kgsl_snapshot_add_section(device, KGSL_SNAPSHOT_SECTION_REGS, snapshot,
kgsl_snapshot_dump_registers, &r);
}
int adreno_snapshot_regs_count(const u32 *ptr)
{
unsigned int count = 0;
unsigned int group_count;
for ( ; ptr[0] != UINT_MAX; ptr += 2) {
group_count = REG_COUNT(ptr);
if (group_count == 1)
count += group_count + 1;
else
count += group_count + 2;
}
return count;
}
/*
* This is a new format for dumping the registers, where we dump just the first
* address of the register along with the count of the contiguous registers
* which we going to dump. This helps us save memory by not dumping the
* address for each register
*/
size_t adreno_snapshot_registers_v2(struct kgsl_device *device, u8 *buf,
size_t remain, void *priv)
{
const u32 *ptr = (const u32 *)priv;
unsigned int *data = (unsigned int *)buf;
int count = 0, k;
/* Figure out how many registers we are going to dump */
count = adreno_snapshot_regs_count(ptr);
if (remain < (count * 4)) {
SNAPSHOT_ERR_NOMEM(device, "REGISTERS");
return 0;
}
for (ptr = (const u32 *)priv; ptr[0] != UINT_MAX; ptr += 2) {
int cnt = REG_COUNT(ptr);
if (cnt == 1)
*data++ = BIT(31) | ptr[0];
else {
*data++ = ptr[0];
*data++ = cnt;
}
for (k = ptr[0]; k <= ptr[1]; k++) {
kgsl_regread(device, k, data);
data++;
}
}
/* Return the size of the section */
return (count * 4);
}
size_t adreno_snapshot_cx_misc_registers(struct kgsl_device *device,
u8 *buf, size_t remain, void *priv)
{
struct adreno_device *adreno_dev = ADRENO_DEVICE(device);
const u32 *ptr = (const u32 *)priv;
u32 *data = (unsigned int *)buf;
int count = 0, k;
/* Figure out how many registers we are going to dump */
count = adreno_snapshot_regs_count(ptr);
if (remain < (count * sizeof(u32))) {
SNAPSHOT_ERR_NOMEM(device, "CX_MISC REGISTERS");
return 0;
}
for (; ptr[0] != UINT_MAX; ptr += 2) {
int cnt = REG_COUNT(ptr);
if (cnt == 1)
*data++ = BIT(31) | ptr[0];
else {
*data++ = ptr[0];
*data++ = cnt;
}
for (k = ptr[0]; k <= ptr[1]; k++)
adreno_cx_misc_regread(adreno_dev,
k - adreno_dev->cx_misc_base, data++);
}
/* Return the size of the section */
return (count * sizeof(u32));
}