android_kernel_samsung_hero.../drivers/soc/qcom/peripheral-loader.c
2016-08-17 16:41:52 +08:00

1162 lines
29 KiB
C

/* Copyright (c) 2010-2016, The Linux Foundation. All rights reserved.
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 and
* only version 2 as published by the Free Software Foundation.
*
* 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.
*/
#include <linux/module.h>
#include <linux/string.h>
#include <linux/firmware.h>
#include <linux/io.h>
#include <linux/elf.h>
#include <linux/mutex.h>
#include <linux/memblock.h>
#include <linux/slab.h>
#include <linux/suspend.h>
#include <linux/rwsem.h>
#include <linux/sysfs.h>
#include <linux/workqueue.h>
#include <linux/jiffies.h>
#include <linux/err.h>
#include <linux/list.h>
#include <linux/list_sort.h>
#include <linux/idr.h>
#include <linux/interrupt.h>
#include <linux/of_gpio.h>
#include <linux/of_address.h>
#include <linux/io.h>
#include <linux/dma-mapping.h>
#include <soc/qcom/ramdump.h>
#include <soc/qcom/subsystem_restart.h>
#include <soc/qcom/secure_buffer.h>
#include <asm/uaccess.h>
#include <asm/setup.h>
#include <asm-generic/io-64-nonatomic-lo-hi.h>
#include "peripheral-loader.h"
#ifdef CONFIG_SEC_PERIPHERAL_SECURE_CHK
#include <linux/qcom/sec_debug.h>
#endif
#define pil_err(desc, fmt, ...) \
dev_err(desc->dev, "%s: " fmt, desc->name, ##__VA_ARGS__)
#define pil_info(desc, fmt, ...) \
dev_info(desc->dev, "%s: " fmt, desc->name, ##__VA_ARGS__)
#if defined(CONFIG_ARM)
#define pil_memset_io(d, c, count) memset(d, c, count)
#else
#define pil_memset_io(d, c, count) memset_io(d, c, count)
#endif
#define PIL_NUM_DESC 10
static void __iomem *pil_info_base;
/**
* proxy_timeout - Override for proxy vote timeouts
* -1: Use driver-specified timeout
* 0: Hold proxy votes until shutdown
* >0: Specify a custom timeout in ms
*/
static int proxy_timeout_ms = -1;
module_param(proxy_timeout_ms, int, S_IRUGO | S_IWUSR);
/**
* struct pil_mdt - Representation of <name>.mdt file in memory
* @hdr: ELF32 header
* @phdr: ELF32 program headers
*/
struct pil_mdt {
struct elf32_hdr hdr;
struct elf32_phdr phdr[];
};
/**
* struct pil_seg - memory map representing one segment
* @next: points to next seg mentor NULL if last segment
* @paddr: physical start address of segment
* @sz: size of segment
* @filesz: size of segment on disk
* @num: segment number
* @relocated: true if segment is relocated, false otherwise
*
* Loosely based on an elf program header. Contains all necessary information
* to load and initialize a segment of the image in memory.
*/
struct pil_seg {
phys_addr_t paddr;
unsigned long sz;
unsigned long filesz;
int num;
struct list_head list;
bool relocated;
};
/**
* struct pil_priv - Private state for a pil_desc
* @proxy: work item used to run the proxy unvoting routine
* @ws: wakeup source to prevent suspend during pil_boot
* @wname: name of @ws
* @desc: pointer to pil_desc this is private data for
* @seg: list of segments sorted by physical address
* @entry_addr: physical address where processor starts booting at
* @base_addr: smallest start address among all segments that are relocatable
* @region_start: address where relocatable region starts or lowest address
* for non-relocatable images
* @region_end: address where relocatable region ends or highest address for
* non-relocatable images
* @region: region allocated for relocatable images
* @unvoted_flag: flag to keep track if we have unvoted or not.
*
* This struct contains data for a pil_desc that should not be exposed outside
* of this file. This structure points to the descriptor and the descriptor
* points to this structure so that PIL drivers can't access the private
* data of a descriptor but this file can access both.
*/
struct pil_priv {
struct delayed_work proxy;
struct wakeup_source ws;
char wname[32];
struct pil_desc *desc;
struct list_head segs;
phys_addr_t entry_addr;
phys_addr_t base_addr;
phys_addr_t region_start;
phys_addr_t region_end;
void *region;
struct pil_image_info __iomem *info;
int id;
int unvoted_flag;
size_t region_size;
};
/**
* pil_do_ramdump() - Ramdump an image
* @desc: descriptor from pil_desc_init()
* @ramdump_dev: ramdump device returned from create_ramdump_device()
*
* Calls the ramdump API with a list of segments generated from the addresses
* that the descriptor corresponds to.
*/
int pil_do_ramdump(struct pil_desc *desc, void *ramdump_dev)
{
struct pil_priv *priv = desc->priv;
struct pil_seg *seg;
int count = 0, ret;
struct ramdump_segment *ramdump_segs, *s;
list_for_each_entry(seg, &priv->segs, list)
count++;
ramdump_segs = kcalloc(count, sizeof(*ramdump_segs), GFP_KERNEL);
if (!ramdump_segs)
return -ENOMEM;
if (desc->subsys_vmid > 0)
ret = pil_assign_mem_to_linux(desc, priv->region_start,
(priv->region_end - priv->region_start));
s = ramdump_segs;
list_for_each_entry(seg, &priv->segs, list) {
s->address = seg->paddr;
s->size = seg->sz;
s++;
}
ret = do_elf_ramdump(ramdump_dev, ramdump_segs, count);
kfree(ramdump_segs);
if (!ret && desc->subsys_vmid > 0)
ret = pil_assign_mem_to_subsys(desc, priv->region_start,
(priv->region_end - priv->region_start));
return ret;
}
EXPORT_SYMBOL(pil_do_ramdump);
int pil_assign_mem_to_subsys(struct pil_desc *desc, phys_addr_t addr,
size_t size)
{
int ret;
int srcVM[1] = {VMID_HLOS};
int destVM[1] = {desc->subsys_vmid};
int destVMperm[1] = {PERM_READ | PERM_WRITE};
ret = hyp_assign_phys(addr, size, srcVM, 1, destVM, destVMperm, 1);
if (ret)
pil_err(desc, "%s: failed for %pa address of size %zx - subsys VMid %d\n",
__func__, &addr, size, desc->subsys_vmid);
return ret;
}
EXPORT_SYMBOL(pil_assign_mem_to_subsys);
int pil_assign_mem_to_linux(struct pil_desc *desc, phys_addr_t addr,
size_t size)
{
int ret;
int srcVM[1] = {desc->subsys_vmid};
int destVM[1] = {VMID_HLOS};
int destVMperm[1] = {PERM_READ | PERM_WRITE | PERM_EXEC};
ret = hyp_assign_phys(addr, size, srcVM, 1, destVM, destVMperm, 1);
if (ret)
panic("%s: failed for %pa address of size %zx - subsys VMid %d. Fatal error.\n",
__func__, &addr, size, desc->subsys_vmid);
return ret;
}
EXPORT_SYMBOL(pil_assign_mem_to_linux);
int pil_assign_mem_to_subsys_and_linux(struct pil_desc *desc,
phys_addr_t addr, size_t size)
{
int ret;
int srcVM[1] = {VMID_HLOS};
int destVM[2] = {VMID_HLOS, desc->subsys_vmid};
int destVMperm[2] = {PERM_READ | PERM_WRITE, PERM_READ | PERM_WRITE};
ret = hyp_assign_phys(addr, size, srcVM, 1, destVM, destVMperm, 2);
if (ret)
pil_err(desc, "%s: failed for %pa address of size %zx - subsys VMid %d\n",
__func__, &addr, size, desc->subsys_vmid);
return ret;
}
EXPORT_SYMBOL(pil_assign_mem_to_subsys_and_linux);
int pil_reclaim_mem(struct pil_desc *desc, phys_addr_t addr, size_t size,
int VMid)
{
int ret;
int srcVM[2] = {VMID_HLOS, desc->subsys_vmid};
int destVM[1] = {VMid};
int destVMperm[1] = {PERM_READ | PERM_WRITE};
if (VMid == VMID_HLOS)
destVMperm[0] = PERM_READ | PERM_WRITE | PERM_EXEC;
ret = hyp_assign_phys(addr, size, srcVM, 2, destVM, destVMperm, 1);
if (ret)
panic("%s: failed for %pa address of size %zx - subsys VMid %d. Fatal error.\n",
__func__, &addr, size, desc->subsys_vmid);
return ret;
}
EXPORT_SYMBOL(pil_reclaim_mem);
/**
* pil_get_entry_addr() - Retrieve the entry address of a peripheral image
* @desc: descriptor from pil_desc_init()
*
* Returns the physical address where the image boots at or 0 if unknown.
*/
phys_addr_t pil_get_entry_addr(struct pil_desc *desc)
{
return desc->priv ? desc->priv->entry_addr : 0;
}
EXPORT_SYMBOL(pil_get_entry_addr);
static void __pil_proxy_unvote(struct pil_priv *priv)
{
struct pil_desc *desc = priv->desc;
desc->ops->proxy_unvote(desc);
notify_proxy_unvote(desc->dev);
__pm_relax(&priv->ws);
module_put(desc->owner);
}
static void pil_proxy_unvote_work(struct work_struct *work)
{
struct delayed_work *delayed = to_delayed_work(work);
struct pil_priv *priv = container_of(delayed, struct pil_priv, proxy);
__pil_proxy_unvote(priv);
}
static int pil_proxy_vote(struct pil_desc *desc)
{
int ret = 0;
struct pil_priv *priv = desc->priv;
if (desc->ops->proxy_vote) {
__pm_stay_awake(&priv->ws);
ret = desc->ops->proxy_vote(desc);
if (ret)
__pm_relax(&priv->ws);
}
if (desc->proxy_unvote_irq)
enable_irq(desc->proxy_unvote_irq);
notify_proxy_vote(desc->dev);
return ret;
}
static void pil_proxy_unvote(struct pil_desc *desc, int immediate)
{
struct pil_priv *priv = desc->priv;
unsigned long timeout;
if (proxy_timeout_ms == 0 && !immediate)
return;
else if (proxy_timeout_ms > 0)
timeout = proxy_timeout_ms;
else
timeout = desc->proxy_timeout;
if (desc->ops->proxy_unvote) {
if (WARN_ON(!try_module_get(desc->owner)))
return;
if (immediate)
timeout = 0;
if (!desc->proxy_unvote_irq || immediate)
schedule_delayed_work(&priv->proxy,
msecs_to_jiffies(timeout));
}
}
static irqreturn_t proxy_unvote_intr_handler(int irq, void *dev_id)
{
struct pil_desc *desc = dev_id;
struct pil_priv *priv = desc->priv;
pil_info(desc, "Power/Clock ready interrupt received\n");
if (!desc->priv->unvoted_flag) {
desc->priv->unvoted_flag = 1;
__pil_proxy_unvote(priv);
}
return IRQ_HANDLED;
}
static bool segment_is_relocatable(const struct elf32_phdr *p)
{
return !!(p->p_flags & BIT(27));
}
static phys_addr_t pil_reloc(const struct pil_priv *priv, phys_addr_t addr)
{
return addr - priv->base_addr + priv->region_start;
}
static struct pil_seg *pil_init_seg(const struct pil_desc *desc,
const struct elf32_phdr *phdr, int num)
{
bool reloc = segment_is_relocatable(phdr);
const struct pil_priv *priv = desc->priv;
struct pil_seg *seg;
if (!reloc && memblock_overlaps_memory(phdr->p_paddr, phdr->p_memsz)) {
pil_err(desc, "kernel memory would be overwritten [%#08lx, %#08lx)\n",
(unsigned long)phdr->p_paddr,
(unsigned long)(phdr->p_paddr + phdr->p_memsz));
return ERR_PTR(-EPERM);
}
if (phdr->p_filesz > phdr->p_memsz) {
pil_err(desc, "Segment %d: file size (%u) is greater than mem size (%u).\n",
num, phdr->p_filesz, phdr->p_memsz);
return ERR_PTR(-EINVAL);
}
seg = kmalloc(sizeof(*seg), GFP_KERNEL);
if (!seg)
return ERR_PTR(-ENOMEM);
seg->num = num;
seg->paddr = reloc ? pil_reloc(priv, phdr->p_paddr) : phdr->p_paddr;
seg->filesz = phdr->p_filesz;
seg->sz = phdr->p_memsz;
seg->relocated = reloc;
INIT_LIST_HEAD(&seg->list);
return seg;
}
#define segment_is_hash(flag) (((flag) & (0x7 << 24)) == (0x2 << 24))
static int segment_is_loadable(const struct elf32_phdr *p)
{
return (p->p_type == PT_LOAD) && !segment_is_hash(p->p_flags) &&
p->p_memsz;
}
static void pil_dump_segs(const struct pil_priv *priv)
{
struct pil_seg *seg;
phys_addr_t seg_h_paddr;
list_for_each_entry(seg, &priv->segs, list) {
seg_h_paddr = seg->paddr + seg->sz;
pil_info(priv->desc, "%d: %pa %pa\n", seg->num,
&seg->paddr, &seg_h_paddr);
}
}
/*
* Ensure the entry address lies within the image limits and if the image is
* relocatable ensure it lies within a relocatable segment.
*/
static int pil_init_entry_addr(struct pil_priv *priv, const struct pil_mdt *mdt)
{
struct pil_seg *seg;
phys_addr_t entry = mdt->hdr.e_entry;
bool image_relocated = priv->region;
if (image_relocated)
entry = pil_reloc(priv, entry);
priv->entry_addr = entry;
if (priv->desc->flags & PIL_SKIP_ENTRY_CHECK)
return 0;
list_for_each_entry(seg, &priv->segs, list) {
if (entry >= seg->paddr && entry < seg->paddr + seg->sz) {
if (!image_relocated)
return 0;
else if (seg->relocated)
return 0;
}
}
pil_err(priv->desc, "entry address %pa not within range\n", &entry);
pil_dump_segs(priv);
return -EADDRNOTAVAIL;
}
static int pil_alloc_region(struct pil_priv *priv, phys_addr_t min_addr,
phys_addr_t max_addr, size_t align)
{
void *region;
size_t size = max_addr - min_addr;
size_t aligned_size;
/* Don't reallocate due to fragmentation concerns, just sanity check */
if (priv->region) {
if (WARN(priv->region_end - priv->region_start < size,
"Can't reuse PIL memory, too small\n"))
return -ENOMEM;
return 0;
}
if (align > SZ_4M)
aligned_size = ALIGN(size, SZ_4M);
else
aligned_size = ALIGN(size, SZ_1M);
init_dma_attrs(&priv->desc->attrs);
dma_set_attr(DMA_ATTR_SKIP_ZEROING, &priv->desc->attrs);
dma_set_attr(DMA_ATTR_NO_KERNEL_MAPPING, &priv->desc->attrs);
region = dma_alloc_attrs(priv->desc->dev, aligned_size,
&priv->region_start, GFP_KERNEL,
&priv->desc->attrs);
if (region == NULL) {
pil_err(priv->desc, "Failed to allocate relocatable region of size %zx\n",
size);
return -ENOMEM;
}
priv->region = region;
priv->region_end = priv->region_start + size;
priv->base_addr = min_addr;
priv->region_size = aligned_size;
return 0;
}
static int pil_setup_region(struct pil_priv *priv, const struct pil_mdt *mdt)
{
const struct elf32_phdr *phdr;
phys_addr_t min_addr_r, min_addr_n, max_addr_r, max_addr_n, start, end;
size_t align = 0;
int i, ret = 0;
bool relocatable = false;
min_addr_n = min_addr_r = (phys_addr_t)ULLONG_MAX;
max_addr_n = max_addr_r = 0;
/* Find the image limits */
for (i = 0; i < mdt->hdr.e_phnum; i++) {
phdr = &mdt->phdr[i];
if (!segment_is_loadable(phdr))
continue;
start = phdr->p_paddr;
end = start + phdr->p_memsz;
if (segment_is_relocatable(phdr)) {
min_addr_r = min(min_addr_r, start);
max_addr_r = max(max_addr_r, end);
/*
* Lowest relocatable segment dictates alignment of
* relocatable region
*/
if (min_addr_r == start)
align = phdr->p_align;
relocatable = true;
} else {
min_addr_n = min(min_addr_n, start);
max_addr_n = max(max_addr_n, end);
}
}
/*
* Align the max address to the next 4K boundary to satisfy iommus and
* XPUs that operate on 4K chunks.
*/
max_addr_n = ALIGN(max_addr_n, SZ_4K);
max_addr_r = ALIGN(max_addr_r, SZ_4K);
if (relocatable) {
ret = pil_alloc_region(priv, min_addr_r, max_addr_r, align);
} else {
priv->region_start = min_addr_n;
priv->region_end = max_addr_n;
priv->base_addr = min_addr_n;
}
if (priv->info) {
__iowrite32_copy(&priv->info->start, &priv->region_start,
sizeof(priv->region_start) / 4);
writel_relaxed(priv->region_end - priv->region_start,
&priv->info->size);
}
return ret;
}
static int pil_cmp_seg(void *priv, struct list_head *a, struct list_head *b)
{
int ret = 0;
struct pil_seg *seg_a = list_entry(a, struct pil_seg, list);
struct pil_seg *seg_b = list_entry(b, struct pil_seg, list);
if (seg_a->paddr < seg_b->paddr)
ret = -1;
else if (seg_a->paddr > seg_b->paddr)
ret = 1;
return ret;
}
static int pil_init_mmap(struct pil_desc *desc, const struct pil_mdt *mdt)
{
struct pil_priv *priv = desc->priv;
const struct elf32_phdr *phdr;
struct pil_seg *seg;
int i, ret;
ret = pil_setup_region(priv, mdt);
if (ret)
return ret;
pil_info(desc, "loading from %pa to %pa\n", &priv->region_start,
&priv->region_end);
for (i = 0; i < mdt->hdr.e_phnum; i++) {
phdr = &mdt->phdr[i];
if (!segment_is_loadable(phdr))
continue;
seg = pil_init_seg(desc, phdr, i);
if (IS_ERR(seg))
return PTR_ERR(seg);
list_add_tail(&seg->list, &priv->segs);
}
list_sort(NULL, &priv->segs, pil_cmp_seg);
return pil_init_entry_addr(priv, mdt);
}
static void pil_release_mmap(struct pil_desc *desc)
{
struct pil_priv *priv = desc->priv;
struct pil_seg *p, *tmp;
u64 zero = 0ULL;
if (priv->info) {
__iowrite32_copy(&priv->info->start, &zero,
sizeof(zero) / 4);
writel_relaxed(0, &priv->info->size);
}
list_for_each_entry_safe(p, tmp, &priv->segs, list) {
list_del(&p->list);
kfree(p);
}
}
#define IOMAP_SIZE SZ_1M
struct pil_map_fw_info {
void *region;
struct dma_attrs attrs;
phys_addr_t base_addr;
struct device *dev;
};
static void *map_fw_mem(phys_addr_t paddr, size_t size, void *data)
{
struct pil_map_fw_info *info = data;
return dma_remap(info->dev, info->region, paddr, size,
&info->attrs);
}
static void unmap_fw_mem(void *vaddr, size_t size, void *data)
{
struct pil_map_fw_info *info = data;
dma_unremap(info->dev, vaddr, size);
}
static int pil_load_seg(struct pil_desc *desc, struct pil_seg *seg)
{
int ret = 0, count;
phys_addr_t paddr;
char fw_name[30];
int num = seg->num;
struct pil_map_fw_info map_fw_info = {
.attrs = desc->attrs,
.region = desc->priv->region,
.base_addr = desc->priv->region_start,
.dev = desc->dev,
};
void *map_data = desc->map_data ? desc->map_data : &map_fw_info;
if (seg->filesz) {
snprintf(fw_name, ARRAY_SIZE(fw_name), "%s.b%02d",
desc->fw_name, num);
ret = request_firmware_into_buf(fw_name, desc->dev, seg->paddr,
seg->filesz, desc->map_fw_mem,
desc->unmap_fw_mem, map_data);
if (ret < 0) {
pil_err(desc, "Failed to locate blob %s or blob is too big.\n",
fw_name);
return ret;
}
if (ret != seg->filesz) {
pil_err(desc, "Blob size %u doesn't match %lu\n",
ret, seg->filesz);
return -EPERM;
}
ret = 0;
}
/* Zero out trailing memory */
paddr = seg->paddr + seg->filesz;
count = seg->sz - seg->filesz;
while (count > 0) {
int size;
u8 __iomem *buf;
size = min_t(size_t, IOMAP_SIZE, count);
buf = desc->map_fw_mem(paddr, size, map_data);
if (!buf) {
pil_err(desc, "Failed to map memory\n");
return -ENOMEM;
}
pil_memset_io(buf, 0, size);
desc->unmap_fw_mem(buf, size, map_data);
count -= size;
paddr += size;
}
if (desc->ops->verify_blob) {
ret = desc->ops->verify_blob(desc, seg->paddr, seg->sz);
if (ret)
pil_err(desc, "Blob%u failed verification\n", num);
}
return ret;
}
static int pil_parse_devicetree(struct pil_desc *desc)
{
struct device_node *ofnode = desc->dev->of_node;
int clk_ready = 0;
if (!ofnode)
return -EINVAL;
if (of_property_read_u32(ofnode, "qcom,mem-protect-id",
&desc->subsys_vmid))
pr_debug("Unable to read the addr-protect-id for %s\n",
desc->name);
if (desc->ops->proxy_unvote && of_find_property(ofnode,
"qcom,gpio-proxy-unvote",
NULL)) {
clk_ready = of_get_named_gpio(ofnode,
"qcom,gpio-proxy-unvote", 0);
if (clk_ready < 0) {
dev_dbg(desc->dev,
"[%s]: Error getting proxy unvoting gpio\n",
desc->name);
return clk_ready;
}
clk_ready = gpio_to_irq(clk_ready);
if (clk_ready < 0) {
dev_err(desc->dev,
"[%s]: Error getting proxy unvote IRQ\n",
desc->name);
return clk_ready;
}
}
desc->proxy_unvote_irq = clk_ready;
return 0;
}
/* Synchronize request_firmware() with suspend */
static DECLARE_RWSEM(pil_pm_rwsem);
/**
* pil_boot() - Load a peripheral image into memory and boot it
* @desc: descriptor from pil_desc_init()
*
* Returns 0 on success or -ERROR on failure.
*/
int pil_boot(struct pil_desc *desc)
{
int ret;
char fw_name[30];
const struct pil_mdt *mdt;
const struct elf32_hdr *ehdr;
struct pil_seg *seg;
const struct firmware *fw;
struct pil_priv *priv = desc->priv;
bool mem_protect = false;
bool hyp_assign = false;
if (desc->shutdown_fail)
pil_err(desc, "Subsystem shutdown failed previously!\n");
/* Reinitialize for new image */
pil_release_mmap(desc);
down_read(&pil_pm_rwsem);
snprintf(fw_name, sizeof(fw_name), "%s.mdt", desc->fw_name);
ret = request_firmware(&fw, fw_name, desc->dev);
if (ret) {
pil_err(desc, "Failed to locate %s\n", fw_name);
goto out;
}
if (fw->size < sizeof(*ehdr)) {
pil_err(desc, "Not big enough to be an elf header\n");
ret = -EIO;
goto release_fw;
}
mdt = (const struct pil_mdt *)fw->data;
ehdr = &mdt->hdr;
if (memcmp(ehdr->e_ident, ELFMAG, SELFMAG)) {
pil_err(desc, "Not an elf header\n");
ret = -EIO;
goto release_fw;
}
if (ehdr->e_phnum == 0) {
pil_err(desc, "No loadable segments\n");
ret = -EIO;
goto release_fw;
}
if (sizeof(struct elf32_phdr) * ehdr->e_phnum +
sizeof(struct elf32_hdr) > fw->size) {
pil_err(desc, "Program headers not within mdt\n");
ret = -EIO;
goto release_fw;
}
ret = pil_init_mmap(desc, mdt);
if (ret)
goto release_fw;
desc->priv->unvoted_flag = 0;
ret = pil_proxy_vote(desc);
if (ret) {
pil_err(desc, "Failed to proxy vote\n");
goto release_fw;
}
if (desc->ops->init_image)
ret = desc->ops->init_image(desc, fw->data, fw->size);
if (ret) {
pil_err(desc, "Invalid firmware metadata\n");
#ifdef CONFIG_SEC_PERIPHERAL_SECURE_CHK
if (ret == -EINVAL && ((!strcmp(desc->name, "mba")) || (!strcmp(desc->name, "modem"))))
{
if (ret && desc->proxy_unvote_irq)
disable_irq(desc->proxy_unvote_irq);
pil_proxy_unvote(desc, ret);
release_firmware(fw);
up_read(&pil_pm_rwsem);
if (ret) {
if (priv->region) {
if (desc->subsys_vmid > 0 && !mem_protect &&
hyp_assign) {
pil_reclaim_mem(desc, priv->region_start,
(priv->region_end -
priv->region_start),
VMID_HLOS);
}
dma_free_attrs(desc->dev, priv->region_size,
priv->region, priv->region_start,
&desc->attrs);
priv->region = NULL;
}
pil_release_mmap(desc);
}
sec_peripheral_secure_check_fail();
}
#endif
goto err_boot;
}
if (desc->ops->mem_setup)
ret = desc->ops->mem_setup(desc, priv->region_start,
priv->region_end - priv->region_start);
if (ret) {
pil_err(desc, "Memory setup error\n");
goto err_deinit_image;
}
if (desc->subsys_vmid > 0) {
/* Make sure the memory is actually assigned to Linux. In the
* case where the shutdown sequence is not able to immediately
* assign the memory back to Linux, we need to do this here. */
ret = pil_assign_mem_to_linux(desc, priv->region_start,
(priv->region_end - priv->region_start));
if (ret)
pil_err(desc, "Failed to assign to linux, ret - %d\n",
ret);
ret = pil_assign_mem_to_subsys_and_linux(desc,
priv->region_start,
(priv->region_end - priv->region_start));
if (ret) {
pil_err(desc, "Failed to assign memory, ret - %d\n",
ret);
goto err_deinit_image;
}
hyp_assign = true;
}
list_for_each_entry(seg, &desc->priv->segs, list) {
ret = pil_load_seg(desc, seg);
if (ret)
goto err_deinit_image;
}
if (desc->subsys_vmid > 0) {
ret = pil_reclaim_mem(desc, priv->region_start,
(priv->region_end - priv->region_start),
desc->subsys_vmid);
if (ret) {
pil_err(desc, "Failed to assign %s memory, ret - %d\n",
desc->name, ret);
goto err_deinit_image;
}
hyp_assign = false;
}
ret = desc->ops->auth_and_reset(desc);
if (ret) {
pil_err(desc, "Failed to bring out of reset\n");
#ifdef CONFIG_SEC_PERIPHERAL_SECURE_CHK
if (ret == -EINVAL && ((!strcmp(desc->name, "mba")) || (!strcmp(desc->name, "modem")))) {
if (ret && desc->subsys_vmid > 0) {
pil_assign_mem_to_linux(desc, priv->region_start,
(priv->region_end - priv->region_start));
mem_protect = true;
}
if (ret && desc->ops->deinit_image)
desc->ops->deinit_image(desc);
if (ret && desc->proxy_unvote_irq)
disable_irq(desc->proxy_unvote_irq);
pil_proxy_unvote(desc, ret);
release_firmware(fw);
up_read(&pil_pm_rwsem);
if (ret) {
if (priv->region) {
if (desc->subsys_vmid > 0 && !mem_protect &&
hyp_assign) {
pil_reclaim_mem(desc, priv->region_start,
(priv->region_end -
priv->region_start),
VMID_HLOS);
}
dma_free_attrs(desc->dev, priv->region_size,
priv->region, priv->region_start,
&desc->attrs);
priv->region = NULL;
}
pil_release_mmap(desc);
}
sec_peripheral_secure_check_fail();
}
#endif
goto err_auth_and_reset;
}
pil_info(desc, "Brought out of reset\n");
err_auth_and_reset:
if (ret && desc->subsys_vmid > 0) {
pil_assign_mem_to_linux(desc, priv->region_start,
(priv->region_end - priv->region_start));
mem_protect = true;
}
err_deinit_image:
if (ret && desc->ops->deinit_image)
desc->ops->deinit_image(desc);
err_boot:
if (ret && desc->proxy_unvote_irq)
disable_irq(desc->proxy_unvote_irq);
pil_proxy_unvote(desc, ret);
release_fw:
release_firmware(fw);
out:
up_read(&pil_pm_rwsem);
if (ret) {
if (priv->region) {
if (desc->subsys_vmid > 0 && !mem_protect &&
hyp_assign) {
pil_reclaim_mem(desc, priv->region_start,
(priv->region_end -
priv->region_start),
VMID_HLOS);
}
dma_free_attrs(desc->dev, priv->region_size,
priv->region, priv->region_start,
&desc->attrs);
priv->region = NULL;
}
pil_release_mmap(desc);
}
return ret;
}
EXPORT_SYMBOL(pil_boot);
/**
* pil_shutdown() - Shutdown a peripheral
* @desc: descriptor from pil_desc_init()
*/
void pil_shutdown(struct pil_desc *desc)
{
struct pil_priv *priv = desc->priv;
if (desc->ops->shutdown) {
if (desc->ops->shutdown(desc))
desc->shutdown_fail = true;
else
desc->shutdown_fail = false;
}
if (desc->proxy_unvote_irq) {
disable_irq(desc->proxy_unvote_irq);
if (!desc->priv->unvoted_flag)
pil_proxy_unvote(desc, 1);
} else if (!proxy_timeout_ms)
pil_proxy_unvote(desc, 1);
else
flush_delayed_work(&priv->proxy);
}
EXPORT_SYMBOL(pil_shutdown);
/**
* pil_free_memory() - Free memory resources associated with a peripheral
* @desc: descriptor from pil_desc_init()
*/
void pil_free_memory(struct pil_desc *desc)
{
struct pil_priv *priv = desc->priv;
if (priv->region) {
if (desc->subsys_vmid > 0)
pil_assign_mem_to_linux(desc, priv->region_start,
(priv->region_end - priv->region_start));
dma_free_attrs(desc->dev, priv->region_size,
priv->region, priv->region_start, &desc->attrs);
priv->region = NULL;
}
}
EXPORT_SYMBOL(pil_free_memory);
static DEFINE_IDA(pil_ida);
/**
* pil_desc_init() - Initialize a pil descriptor
* @desc: descriptor to intialize
*
* Initialize a pil descriptor for use by other pil functions. This function
* must be called before calling pil_boot() or pil_shutdown().
*
* Returns 0 for success and -ERROR on failure.
*/
int pil_desc_init(struct pil_desc *desc)
{
struct pil_priv *priv;
int ret;
void __iomem *addr;
char buf[sizeof(priv->info->name)];
if (WARN(desc->ops->proxy_unvote && !desc->ops->proxy_vote,
"Invalid proxy voting. Ignoring\n"))
((struct pil_reset_ops *)desc->ops)->proxy_unvote = NULL;
priv = kzalloc(sizeof(*priv), GFP_KERNEL);
if (!priv)
return -ENOMEM;
desc->priv = priv;
priv->desc = desc;
priv->id = ret = ida_simple_get(&pil_ida, 0, PIL_NUM_DESC, GFP_KERNEL);
if (priv->id < 0)
goto err;
if (pil_info_base) {
addr = pil_info_base + sizeof(struct pil_image_info) * priv->id;
priv->info = (struct pil_image_info __iomem *)addr;
strncpy(buf, desc->name, sizeof(buf));
__iowrite32_copy(priv->info->name, buf, sizeof(buf) / 4);
}
ret = pil_parse_devicetree(desc);
if (ret)
goto err_parse_dt;
/* Ignore users who don't make any sense */
WARN(desc->ops->proxy_unvote && desc->proxy_unvote_irq == 0
&& !desc->proxy_timeout,
"Invalid proxy unvote callback or a proxy timeout of 0"
" was specified or no proxy unvote IRQ was specified.\n");
if (desc->proxy_unvote_irq) {
ret = request_threaded_irq(desc->proxy_unvote_irq,
NULL,
proxy_unvote_intr_handler,
IRQF_ONESHOT | IRQF_TRIGGER_RISING,
desc->name, desc);
if (ret < 0) {
dev_err(desc->dev,
"Unable to request proxy unvote IRQ: %d\n",
ret);
goto err;
}
disable_irq(desc->proxy_unvote_irq);
}
snprintf(priv->wname, sizeof(priv->wname), "pil-%s", desc->name);
wakeup_source_init(&priv->ws, priv->wname);
INIT_DELAYED_WORK(&priv->proxy, pil_proxy_unvote_work);
INIT_LIST_HEAD(&priv->segs);
/* Make sure mapping functions are set. */
if (!desc->map_fw_mem)
desc->map_fw_mem = map_fw_mem;
if (!desc->unmap_fw_mem)
desc->unmap_fw_mem = unmap_fw_mem;
return 0;
err_parse_dt:
ida_simple_remove(&pil_ida, priv->id);
err:
kfree(priv);
return ret;
}
EXPORT_SYMBOL(pil_desc_init);
/**
* pil_desc_release() - Release a pil descriptor
* @desc: descriptor to free
*/
void pil_desc_release(struct pil_desc *desc)
{
struct pil_priv *priv = desc->priv;
if (priv) {
ida_simple_remove(&pil_ida, priv->id);
flush_delayed_work(&priv->proxy);
wakeup_source_trash(&priv->ws);
}
desc->priv = NULL;
kfree(priv);
}
EXPORT_SYMBOL(pil_desc_release);
static int pil_pm_notify(struct notifier_block *b, unsigned long event, void *p)
{
switch (event) {
case PM_SUSPEND_PREPARE:
down_write(&pil_pm_rwsem);
break;
case PM_POST_SUSPEND:
up_write(&pil_pm_rwsem);
break;
}
return NOTIFY_DONE;
}
static struct notifier_block pil_pm_notifier = {
.notifier_call = pil_pm_notify,
};
static int __init msm_pil_init(void)
{
struct device_node *np;
struct resource res;
int i;
np = of_find_compatible_node(NULL, NULL, "qcom,msm-imem-pil");
if (!np) {
pr_warn("pil: failed to find qcom,msm-imem-pil node\n");
goto out;
}
if (of_address_to_resource(np, 0, &res)) {
pr_warn("pil: address to resource on imem region failed\n");
goto out;
}
pil_info_base = ioremap(res.start, resource_size(&res));
if (!pil_info_base) {
pr_warn("pil: could not map imem region\n");
goto out;
}
for (i = 0; i < resource_size(&res)/sizeof(u32); i++)
writel_relaxed(0, pil_info_base + (i * sizeof(u32)));
out:
return register_pm_notifier(&pil_pm_notifier);
}
device_initcall(msm_pil_init);
static void __exit msm_pil_exit(void)
{
unregister_pm_notifier(&pil_pm_notifier);
if (pil_info_base)
iounmap(pil_info_base);
}
module_exit(msm_pil_exit);
MODULE_LICENSE("GPL v2");
MODULE_DESCRIPTION("Load peripheral images and bring peripherals out of reset");