blob: af6cac696d434c8ac3785427940c477e97408b1a [file] [log] [blame]
/*
* Copyright 2019 Red Hat Inc.
*
* 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 COPYRIGHT HOLDER(S) OR AUTHOR(S) 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 "priv.h"
#include <core/firmware.h>
#include <core/memory.h>
#include <subdev/mmu.h>
static struct nvkm_acr_hsf *
nvkm_acr_hsf_find(struct nvkm_acr *acr, const char *name)
{
struct nvkm_acr_hsf *hsf;
list_for_each_entry(hsf, &acr->hsf, head) {
if (!strcmp(hsf->name, name))
return hsf;
}
return NULL;
}
int
nvkm_acr_hsf_boot(struct nvkm_acr *acr, const char *name)
{
struct nvkm_subdev *subdev = &acr->subdev;
struct nvkm_acr_hsf *hsf;
int ret;
hsf = nvkm_acr_hsf_find(acr, name);
if (!hsf)
return -EINVAL;
nvkm_debug(subdev, "executing %s binary\n", hsf->name);
ret = nvkm_falcon_get(hsf->falcon, subdev);
if (ret)
return ret;
ret = hsf->func->boot(acr, hsf);
nvkm_falcon_put(hsf->falcon, subdev);
if (ret) {
nvkm_error(subdev, "%s binary failed\n", hsf->name);
return ret;
}
nvkm_debug(subdev, "%s binary completed successfully\n", hsf->name);
return 0;
}
static void
nvkm_acr_unload(struct nvkm_acr *acr)
{
if (acr->done) {
nvkm_acr_hsf_boot(acr, "unload");
acr->done = false;
}
}
static int
nvkm_acr_load(struct nvkm_acr *acr)
{
struct nvkm_subdev *subdev = &acr->subdev;
struct nvkm_acr_lsf *lsf;
u64 start, limit;
int ret;
if (list_empty(&acr->lsf)) {
nvkm_debug(subdev, "No LSF(s) present.\n");
return 0;
}
ret = acr->func->init(acr);
if (ret)
return ret;
acr->func->wpr_check(acr, &start, &limit);
if (start != acr->wpr_start || limit != acr->wpr_end) {
nvkm_error(subdev, "WPR not configured as expected: "
"%016llx-%016llx vs %016llx-%016llx\n",
acr->wpr_start, acr->wpr_end, start, limit);
return -EIO;
}
acr->done = true;
list_for_each_entry(lsf, &acr->lsf, head) {
if (lsf->func->boot) {
ret = lsf->func->boot(lsf->falcon);
if (ret)
break;
}
}
return ret;
}
static int
nvkm_acr_reload(struct nvkm_acr *acr)
{
nvkm_acr_unload(acr);
return nvkm_acr_load(acr);
}
static struct nvkm_acr_lsf *
nvkm_acr_falcon(struct nvkm_device *device)
{
struct nvkm_acr *acr = device->acr;
struct nvkm_acr_lsf *lsf;
if (acr) {
list_for_each_entry(lsf, &acr->lsf, head) {
if (lsf->func->bootstrap_falcon)
return lsf;
}
}
return NULL;
}
int
nvkm_acr_bootstrap_falcons(struct nvkm_device *device, unsigned long mask)
{
struct nvkm_acr_lsf *acrflcn = nvkm_acr_falcon(device);
struct nvkm_acr *acr = device->acr;
unsigned long id;
/* If there's no LS FW managing bootstrapping of other LS falcons,
* we depend on the HS firmware being able to do it instead.
*/
if (!acrflcn) {
/* Which isn't possible everywhere... */
if ((mask & acr->func->bootstrap_falcons) == mask) {
int ret = nvkm_acr_reload(acr);
if (ret)
return ret;
return acr->done ? 0 : -EINVAL;
}
return -ENOSYS;
}
if ((mask & acrflcn->func->bootstrap_falcons) != mask)
return -ENOSYS;
if (acrflcn->func->bootstrap_multiple_falcons) {
return acrflcn->func->
bootstrap_multiple_falcons(acrflcn->falcon, mask);
}
for_each_set_bit(id, &mask, NVKM_ACR_LSF_NUM) {
int ret = acrflcn->func->bootstrap_falcon(acrflcn->falcon, id);
if (ret)
return ret;
}
return 0;
}
bool
nvkm_acr_managed_falcon(struct nvkm_device *device, enum nvkm_acr_lsf_id id)
{
struct nvkm_acr *acr = device->acr;
if (acr) {
if (acr->managed_falcons & BIT_ULL(id))
return true;
}
return false;
}
static int
nvkm_acr_fini(struct nvkm_subdev *subdev, bool suspend)
{
nvkm_acr_unload(nvkm_acr(subdev));
return 0;
}
static int
nvkm_acr_init(struct nvkm_subdev *subdev)
{
if (!nvkm_acr_falcon(subdev->device))
return 0;
return nvkm_acr_load(nvkm_acr(subdev));
}
static void
nvkm_acr_cleanup(struct nvkm_acr *acr)
{
nvkm_acr_lsfw_del_all(acr);
nvkm_acr_hsfw_del_all(acr);
nvkm_firmware_put(acr->wpr_fw);
acr->wpr_fw = NULL;
}
static int
nvkm_acr_oneinit(struct nvkm_subdev *subdev)
{
struct nvkm_device *device = subdev->device;
struct nvkm_acr *acr = nvkm_acr(subdev);
struct nvkm_acr_hsfw *hsfw;
struct nvkm_acr_lsfw *lsfw, *lsft;
struct nvkm_acr_lsf *lsf;
u32 wpr_size = 0;
u64 falcons;
int ret, i;
if (list_empty(&acr->hsfw)) {
nvkm_debug(subdev, "No HSFW(s)\n");
nvkm_acr_cleanup(acr);
return 0;
}
/* Determine layout/size of WPR image up-front, as we need to know
* it to allocate memory before we begin constructing it.
*/
list_for_each_entry_safe(lsfw, lsft, &acr->lsfw, head) {
/* Cull unknown falcons that are present in WPR image. */
if (acr->wpr_fw) {
if (!lsfw->func) {
nvkm_acr_lsfw_del(lsfw);
continue;
}
wpr_size = acr->wpr_fw->size;
}
/* Ensure we've fetched falcon configuration. */
ret = nvkm_falcon_get(lsfw->falcon, subdev);
if (ret)
return ret;
nvkm_falcon_put(lsfw->falcon, subdev);
if (!(lsf = kmalloc(sizeof(*lsf), GFP_KERNEL)))
return -ENOMEM;
lsf->func = lsfw->func;
lsf->falcon = lsfw->falcon;
lsf->id = lsfw->id;
list_add_tail(&lsf->head, &acr->lsf);
acr->managed_falcons |= BIT_ULL(lsf->id);
}
/* Ensure the falcon that'll provide ACR functions is booted first. */
lsf = nvkm_acr_falcon(device);
if (lsf) {
falcons = lsf->func->bootstrap_falcons;
list_move(&lsf->head, &acr->lsf);
} else {
falcons = acr->func->bootstrap_falcons;
}
/* Cull falcons that can't be bootstrapped, or the HSFW can fail to
* boot and leave the GPU in a weird state.
*/
list_for_each_entry_safe(lsfw, lsft, &acr->lsfw, head) {
if (!(falcons & BIT_ULL(lsfw->id))) {
nvkm_warn(subdev, "%s falcon cannot be bootstrapped\n",
nvkm_acr_lsf_id(lsfw->id));
nvkm_acr_lsfw_del(lsfw);
}
}
if (!acr->wpr_fw || acr->wpr_comp)
wpr_size = acr->func->wpr_layout(acr);
/* Allocate/Locate WPR + fill ucode blob pointer.
*
* dGPU: allocate WPR + shadow blob
* Tegra: locate WPR with regs, ensure size is sufficient,
* allocate ucode blob.
*/
ret = acr->func->wpr_alloc(acr, wpr_size);
if (ret)
return ret;
nvkm_debug(subdev, "WPR region is from 0x%llx-0x%llx (shadow 0x%llx)\n",
acr->wpr_start, acr->wpr_end, acr->shadow_start);
/* Write WPR to ucode blob. */
nvkm_kmap(acr->wpr);
if (acr->wpr_fw && !acr->wpr_comp)
nvkm_wobj(acr->wpr, 0, acr->wpr_fw->data, acr->wpr_fw->size);
if (!acr->wpr_fw || acr->wpr_comp)
acr->func->wpr_build(acr, nvkm_acr_falcon(device));
acr->func->wpr_patch(acr, (s64)acr->wpr_start - acr->wpr_prev);
if (acr->wpr_fw && acr->wpr_comp) {
nvkm_kmap(acr->wpr);
for (i = 0; i < acr->wpr_fw->size; i += 4) {
u32 us = nvkm_ro32(acr->wpr, i);
u32 fw = ((u32 *)acr->wpr_fw->data)[i/4];
if (fw != us) {
nvkm_warn(subdev, "%08x: %08x %08x\n",
i, us, fw);
}
}
return -EINVAL;
}
nvkm_done(acr->wpr);
/* Allocate instance block for ACR-related stuff. */
ret = nvkm_memory_new(device, NVKM_MEM_TARGET_INST, 0x1000, 0, true,
&acr->inst);
if (ret)
return ret;
ret = nvkm_vmm_new(device, 0, 0, NULL, 0, NULL, "acr", &acr->vmm);
if (ret)
return ret;
acr->vmm->debug = acr->subdev.debug;
ret = nvkm_vmm_join(acr->vmm, acr->inst);
if (ret)
return ret;
/* Load HS firmware blobs into ACR VMM. */
list_for_each_entry(hsfw, &acr->hsfw, head) {
nvkm_debug(subdev, "loading %s fw\n", hsfw->name);
ret = hsfw->func->load(acr, hsfw);
if (ret)
return ret;
}
/* Kill temporary data. */
nvkm_acr_cleanup(acr);
return 0;
}
static void *
nvkm_acr_dtor(struct nvkm_subdev *subdev)
{
struct nvkm_acr *acr = nvkm_acr(subdev);
struct nvkm_acr_hsf *hsf, *hst;
struct nvkm_acr_lsf *lsf, *lst;
list_for_each_entry_safe(hsf, hst, &acr->hsf, head) {
nvkm_vmm_put(acr->vmm, &hsf->vma);
nvkm_memory_unref(&hsf->ucode);
kfree(hsf->imem);
list_del(&hsf->head);
kfree(hsf);
}
nvkm_vmm_part(acr->vmm, acr->inst);
nvkm_vmm_unref(&acr->vmm);
nvkm_memory_unref(&acr->inst);
nvkm_memory_unref(&acr->wpr);
list_for_each_entry_safe(lsf, lst, &acr->lsf, head) {
list_del(&lsf->head);
kfree(lsf);
}
nvkm_acr_cleanup(acr);
return acr;
}
static const struct nvkm_subdev_func
nvkm_acr = {
.dtor = nvkm_acr_dtor,
.oneinit = nvkm_acr_oneinit,
.init = nvkm_acr_init,
.fini = nvkm_acr_fini,
};
static int
nvkm_acr_ctor_wpr(struct nvkm_acr *acr, int ver)
{
struct nvkm_subdev *subdev = &acr->subdev;
struct nvkm_device *device = subdev->device;
int ret;
ret = nvkm_firmware_get(subdev, "acr/wpr", ver, &acr->wpr_fw);
if (ret < 0)
return ret;
/* Pre-add LSFs in the order they appear in the FW WPR image so that
* we're able to do a binary comparison with our own generator.
*/
ret = acr->func->wpr_parse(acr);
if (ret)
return ret;
acr->wpr_comp = nvkm_boolopt(device->cfgopt, "NvAcrWprCompare", false);
acr->wpr_prev = nvkm_longopt(device->cfgopt, "NvAcrWprPrevAddr", 0);
return 0;
}
int
nvkm_acr_new_(const struct nvkm_acr_fwif *fwif, struct nvkm_device *device,
enum nvkm_subdev_type type, int inst, struct nvkm_acr **pacr)
{
struct nvkm_acr *acr;
long wprfw;
if (!(acr = *pacr = kzalloc(sizeof(*acr), GFP_KERNEL)))
return -ENOMEM;
nvkm_subdev_ctor(&nvkm_acr, device, type, inst, &acr->subdev);
INIT_LIST_HEAD(&acr->hsfw);
INIT_LIST_HEAD(&acr->lsfw);
INIT_LIST_HEAD(&acr->hsf);
INIT_LIST_HEAD(&acr->lsf);
fwif = nvkm_firmware_load(&acr->subdev, fwif, "Acr", acr);
if (IS_ERR(fwif))
return PTR_ERR(fwif);
acr->func = fwif->func;
wprfw = nvkm_longopt(device->cfgopt, "NvAcrWpr", -1);
if (wprfw >= 0) {
int ret = nvkm_acr_ctor_wpr(acr, wprfw);
if (ret)
return ret;
}
return 0;
}