Files
u-boot-krane/drivers/pinctrl/pinctrl-single.c
T
Jordi Trepat MurandTom Rini 11f5a60ba7 pinctrl: single: parse gpio-range as a raw array (O(N^2) -> O(N))
single_add_gpio_func() calls ofnode_parse_phandle_with_args() once per
gpio-range entry. With a flat device tree, every call re-iterates the
property from index 0 and, because cellname is set, performs an
fdt_node_offset_by_phandle() (a full-FDT scan) at every step. The total
cost is therefore quadratic in the number of entries and proportional to
the size of the device tree.

The impact depends on when the pinctrl is probed. On the J722S EVM
defconfig, pinctrl@f4000 (main_pmx0, the SoC's stock 7-entry gpio-range)
is probed after relocation (caches on): single_add_gpio_func() takes
~18 ms there and ~1 ms with this change. When the node is probed before
relocation (caches off) the same parse takes hundreds of ms; on a board
that probes it pre-relocation and runs LPDDR4 clocked down it reached
723 ms, cut to 22 ms here. Any pinctrl-single user with a populated
gpio-range property pays this cost.

The phandle target is never dereferenced by this function (only the
argument cells are stored), so resolve it only for the first entry to
learn the per-entry cell count, then read the argument cells directly
with ofnode_read_u32_index() and allocate the ranges in one block. A
zero phandle still terminates the list, as in the original loop. This
follows the existing pinctrl-single,gpio-range usage observed in current
DTs, where entries use the same provider and therefore a uniform
per-entry cell count.

Signed-off-by: Jordi Trepat Mur <yordy1902@gmail.com>
Reviewed-by: Anshul Dalal <anshuld@ti.com>
Reviewed-by: Simon Glass <sjg@chromium.org>
2026-07-28 12:30:55 -06:00

692 lines
17 KiB
C

// SPDX-License-Identifier: GPL-2.0+
/*
* Copyright (C) EETS GmbH, 2017, Felix Brack <f.brack@eets.ch>
* Copyright (C) 2021 Dario Binacchi <dariobin@libero.it>
*/
#include <mapmem.h>
#include <dm.h>
#include <dm/device_compat.h>
#include <dm/devres.h>
#include <dm/of_access.h>
#include <dm/pinctrl.h>
#include <linux/libfdt.h>
#include <linux/list.h>
#include <asm/io.h>
#include <sort.h>
/**
* struct single_pdata - platform data
* @base: first configuration register
* @offset: index of last configuration register
* @mask: configuration-value mask bits
* @width: configuration register bit width
* @bits_per_mux: true if one register controls more than one pin
*/
struct single_pdata {
void *base;
int offset;
u32 mask;
u32 width;
u32 args_count;
bool bits_per_mux;
};
/**
* struct single_func - pinctrl function
* @node: list node
* @name: pinctrl function name
* @npins: number of entries in pins array
* @pins: pins array
*/
struct single_func {
struct list_head node;
const char *name;
unsigned int npins;
unsigned int *pins;
};
/**
* struct single_gpiofunc_range - pin ranges with same mux value of gpio fun
* @offset: offset base of pins
* @npins: number pins with the same mux value of gpio function
* @gpiofunc: mux value of gpio function
* @node: list node
*/
struct single_gpiofunc_range {
u32 offset;
u32 npins;
u32 gpiofunc;
struct list_head node;
};
/**
* struct single_priv - private data
* @bits_per_pin: number of bits per pin
* @npins: number of selectable pins
* @pin_name: temporary buffer to store the pin name
* @functions: list pin functions
* @gpiofuncs: list gpio functions
*/
struct single_priv {
#if (IS_ENABLED(CONFIG_SANDBOX))
u32 *sandbox_regs;
#endif
unsigned int bits_per_pin;
unsigned int npins;
char pin_name[PINNAME_SIZE];
struct list_head functions;
struct list_head gpiofuncs;
};
/**
* struct single_fdt_bits_cfg - pin configuration
*
* This structure is used for the pin configuration parameters in case
* the register controls more than one pin.
*
* @reg: configuration register offset
* @val: configuration register value
* @mask: configuration register mask
*/
struct single_fdt_bits_cfg {
fdt32_t reg;
fdt32_t val;
fdt32_t mask;
};
#if (!IS_ENABLED(CONFIG_SANDBOX))
static unsigned int single_read(struct udevice *dev, void *reg)
{
struct single_pdata *pdata = dev_get_plat(dev);
switch (pdata->width) {
case 8:
return readb(reg);
case 16:
return readw(reg);
default: /* 32 bits */
return readl(reg);
}
}
static void single_write(struct udevice *dev, unsigned int val, void *reg)
{
struct single_pdata *pdata = dev_get_plat(dev);
switch (pdata->width) {
case 8:
writeb(val, reg);
break;
case 16:
writew(val, reg);
break;
default: /* 32 bits */
writel(val, reg);
}
}
#else /* CONFIG_SANDBOX */
static unsigned int single_read(struct udevice *dev, void *reg)
{
struct single_priv *priv = dev_get_priv(dev);
return priv->sandbox_regs[map_to_sysmem(reg)];
}
static void single_write(struct udevice *dev, unsigned int val, void *reg)
{
struct single_priv *priv = dev_get_priv(dev);
priv->sandbox_regs[map_to_sysmem(reg)] = val;
}
#endif /* CONFIG_SANDBOX */
/**
* single_get_pin_by_offset() - get a pin based on the register offset
* @dev: single driver instance
* @offset: register offset from the base
*/
static int single_get_pin_by_offset(struct udevice *dev, unsigned int offset)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
if (offset > pdata->offset) {
dev_err(dev, "mux offset out of range: 0x%x (0x%x)\n",
offset, pdata->offset);
return -EINVAL;
}
if (pdata->bits_per_mux)
return (offset * BITS_PER_BYTE) / priv->bits_per_pin;
return offset / (pdata->width / BITS_PER_BYTE);
}
static int single_get_offset_by_pin(struct udevice *dev, unsigned int pin)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
unsigned int mux_bytes;
if (pin >= priv->npins)
return -EINVAL;
mux_bytes = pdata->width / BITS_PER_BYTE;
if (pdata->bits_per_mux) {
int byte_num;
byte_num = (priv->bits_per_pin * pin) / BITS_PER_BYTE;
return (byte_num / mux_bytes) * mux_bytes;
}
return pin * mux_bytes;
}
static const char *single_get_pin_function(struct udevice *dev,
unsigned int pin)
{
struct single_priv *priv = dev_get_priv(dev);
struct single_func *func;
int i;
list_for_each_entry(func, &priv->functions, node) {
for (i = 0; i < func->npins; i++) {
if (pin == func->pins[i])
return func->name;
if (pin < func->pins[i])
break;
}
}
return NULL;
}
static int single_get_pin_muxing(struct udevice *dev, unsigned int pin,
char *buf, int size)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
phys_addr_t phys_reg;
void *reg;
const char *fname;
unsigned int val;
int offset, pin_shift = 0;
offset = single_get_offset_by_pin(dev, pin);
if (offset < 0)
return offset;
reg = pdata->base + offset;
val = single_read(dev, reg);
phys_reg = map_to_sysmem(reg);
if (pdata->bits_per_mux)
pin_shift = pin % (pdata->width / priv->bits_per_pin) *
priv->bits_per_pin;
val &= (pdata->mask << pin_shift);
fname = single_get_pin_function(dev, pin);
snprintf(buf, size, "%pa 0x%08x %s", &phys_reg, val,
fname ? fname : "UNCLAIMED");
return 0;
}
static int single_request(struct udevice *dev, int pin, int flags)
{
struct single_priv *priv = dev_get_priv(dev);
struct single_pdata *pdata = dev_get_plat(dev);
struct single_gpiofunc_range *frange = NULL;
struct list_head *pos, *tmp;
void *reg;
int mux_bytes = 0;
u32 data;
/* If function mask is null, needn't enable it. */
if (!pdata->mask)
return -ENOTSUPP;
list_for_each_safe(pos, tmp, &priv->gpiofuncs) {
frange = list_entry(pos, struct single_gpiofunc_range, node);
if ((pin >= frange->offset + frange->npins) ||
pin < frange->offset)
continue;
mux_bytes = pdata->width / BITS_PER_BYTE;
reg = pdata->base + pin * mux_bytes;
data = single_read(dev, reg);
data &= ~pdata->mask;
data |= frange->gpiofunc;
single_write(dev, data, reg);
break;
}
return 0;
}
static struct single_func *single_allocate_function(struct udevice *dev,
unsigned int group_pins)
{
struct single_func *func;
func = devm_kmalloc(dev, sizeof(*func), GFP_KERNEL);
if (!func)
return ERR_PTR(-ENOMEM);
func->pins = devm_kmalloc(dev, sizeof(unsigned int) * group_pins,
GFP_KERNEL);
if (!func->pins) {
devm_kfree(dev, func);
return ERR_PTR(-ENOMEM);
}
return func;
}
static int single_pin_compare(const void *s1, const void *s2)
{
int pin1 = *(const unsigned int *)s1;
int pin2 = *(const unsigned int *)s2;
return pin1 - pin2;
}
/**
* single_configure_pins() - Configure pins based on FDT data
*
* @dev: Pointer to single pin configuration device which is the parent of
* the pins node holding the pin configuration data.
* @pins: Pointer to the first element of an array of register/value pairs
* of type 'u32'. Each such pair describes the pin to be configured
* and the value to be used for configuration.
* The value can either be a simple value if #pinctrl-cells = 1
* or a configuration value and a pin mux mode value if it is 2
* This pointer points to a 'pinctrl-single,pins' property in the
* device-tree.
* @size: Size of the 'pins' array in bytes.
* The number of cells in the array therefore equals to
* 'size / sizeof(u32)'.
* @fname: Function name.
*/
static int single_configure_pins(struct udevice *dev,
const u32 *pins,
int size, const char *fname)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
int stride = pdata->args_count + 1;
int n, pin, count = size / sizeof(u32);
struct single_func *func;
void *reg;
u32 offset, val, mux;
/* If function mask is null, needn't enable it. */
if (!pdata->mask)
return 0;
func = single_allocate_function(dev, count);
if (IS_ERR(func))
return PTR_ERR(func);
func->name = fname;
func->npins = 0;
for (n = 0; n < count; n += stride) {
offset = fdt32_to_cpu(pins[n]);
if (offset > pdata->offset) {
dev_err(dev, " invalid register offset 0x%x\n",
offset);
continue;
}
/* if the pinctrl-cells is 2 then the second cell contains the mux */
if (stride == 3)
mux = fdt32_to_cpu(pins[n + 2]);
else
mux = 0;
reg = pdata->base + offset;
val = (fdt32_to_cpu(pins[n + 1]) | mux) & pdata->mask;
pin = single_get_pin_by_offset(dev, offset);
if (pin < 0) {
dev_err(dev, " failed to get pin by offset %x\n",
offset);
continue;
}
single_write(dev, (single_read(dev, reg) & ~pdata->mask) | val,
reg);
dev_dbg(dev, " reg/val %pa/0x%08x\n", &reg, val);
func->pins[func->npins] = pin;
func->npins++;
}
qsort(func->pins, func->npins, sizeof(func->pins[0]),
single_pin_compare);
list_add(&func->node, &priv->functions);
return 0;
}
static int single_configure_bits(struct udevice *dev,
const struct single_fdt_bits_cfg *pins,
int size, const char *fname)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
int n, pin, count = size / sizeof(struct single_fdt_bits_cfg);
int npins_in_reg, pin_num_from_lsb;
struct single_func *func;
void *reg;
u32 offset, val, mask, bit_pos, val_pos, mask_pos, submask;
/* If function mask is null, needn't enable it. */
if (!pdata->mask)
return 0;
npins_in_reg = pdata->width / priv->bits_per_pin;
func = single_allocate_function(dev, count * npins_in_reg);
if (IS_ERR(func))
return PTR_ERR(func);
func->name = fname;
func->npins = 0;
for (n = 0; n < count; n++, pins++) {
offset = fdt32_to_cpu(pins->reg);
if (offset > pdata->offset) {
dev_dbg(dev, " invalid register offset 0x%x\n",
offset);
continue;
}
reg = pdata->base + offset;
pin = single_get_pin_by_offset(dev, offset);
if (pin < 0) {
dev_err(dev, " failed to get pin by offset 0x%pa\n",
&reg);
continue;
}
mask = fdt32_to_cpu(pins->mask);
val = fdt32_to_cpu(pins->val) & mask;
single_write(dev, (single_read(dev, reg) & ~mask) | val, reg);
dev_dbg(dev, " reg/val %pa/0x%08x\n", &reg, val);
while (mask) {
bit_pos = __ffs(mask);
pin_num_from_lsb = bit_pos / priv->bits_per_pin;
mask_pos = pdata->mask << bit_pos;
val_pos = val & mask_pos;
submask = mask & mask_pos;
if ((mask & mask_pos) == 0) {
dev_err(dev, "Invalid mask at 0x%x\n", offset);
break;
}
mask &= ~mask_pos;
if (submask != mask_pos) {
dev_warn(dev,
"Invalid submask 0x%x at 0x%x\n",
submask, offset);
continue;
}
func->pins[func->npins] = pin + pin_num_from_lsb;
func->npins++;
}
}
qsort(func->pins, func->npins, sizeof(func->pins[0]),
single_pin_compare);
list_add(&func->node, &priv->functions);
return 0;
}
static int single_set_state(struct udevice *dev,
struct udevice *config)
{
const u32 *prop;
const struct single_fdt_bits_cfg *prop_bits;
int len;
prop = dev_read_prop(config, "pinctrl-single,pins", &len);
if (prop) {
dev_dbg(dev, "configuring pins for %s\n", config->name);
if (len % sizeof(u32)) {
dev_dbg(dev, " invalid pin configuration in fdt\n");
return -FDT_ERR_BADSTRUCTURE;
}
single_configure_pins(dev, prop, len, config->name);
return 0;
}
/* pinctrl-single,pins not found so check for pinctrl-single,bits */
prop_bits = dev_read_prop(config, "pinctrl-single,bits", &len);
if (prop_bits) {
dev_dbg(dev, "configuring pins for %s\n", config->name);
if (len % sizeof(struct single_fdt_bits_cfg)) {
dev_dbg(dev, " invalid bits configuration in fdt\n");
return -FDT_ERR_BADSTRUCTURE;
}
single_configure_bits(dev, prop_bits, len, config->name);
return 0;
}
/* Neither 'pinctrl-single,pins' nor 'pinctrl-single,bits' were found */
return len;
}
static const char *single_get_pin_name(struct udevice *dev,
unsigned int selector)
{
struct single_priv *priv = dev_get_priv(dev);
if (selector >= priv->npins)
snprintf(priv->pin_name, PINNAME_SIZE, "Error");
else
snprintf(priv->pin_name, PINNAME_SIZE, "PIN%u", selector);
return priv->pin_name;
}
static int single_get_pins_count(struct udevice *dev)
{
struct single_priv *priv = dev_get_priv(dev);
return priv->npins;
}
static int single_add_gpio_func(struct udevice *dev)
{
struct single_priv *priv = dev_get_priv(dev);
const char *propname = "pinctrl-single,gpio-range";
const char *cellname = "#pinctrl-single,gpio-range-cells";
ofnode node = dev_ofnode(dev);
struct single_gpiofunc_range *range;
struct ofnode_phandle_args gpiospec;
int size, total_cells, cells_per_entry, entries, ret, i;
/*
* The original loop called ofnode_parse_phandle_with_args() once per
* entry. On a flat DT each call resolves the phandle, which scans the
* whole tree, so the repeated parse is quadratic in the number of
* entries and proportional to the tree size. This is very slow
* pre-relocation with caches off.
*
* The phandle target is not used here (only the argument cells are
* stored), so resolve it only for the first entry to learn the
* per-entry cell count, then read the argument cells directly. This
* follows the existing pinctrl-single,gpio-range usage observed in
* current DTs, where entries use the same provider and therefore a
* uniform per-entry cell count.
*/
ret = ofnode_parse_phandle_with_args(node, propname, cellname, 0, 0,
&gpiospec);
if (ret)
return 0;
if (gpiospec.args_count != 3)
return 0;
cells_per_entry = 1 + gpiospec.args_count;
size = ofnode_read_size(node, propname);
if (size < (int)sizeof(u32))
return 0;
total_cells = size / sizeof(u32);
entries = total_cells / cells_per_entry;
if (!entries)
return 0;
range = devm_kcalloc(dev, entries, sizeof(*range), GFP_KERNEL);
if (!range)
return -ENOMEM;
for (i = 0; i < entries; i++) {
int base = i * cells_per_entry;
u32 phandle;
ret = ofnode_read_u32_index(node, propname, base, &phandle);
if (ret)
break;
/* A zero phandle terminates the list, as in the original loop. */
if (!phandle)
break;
ret = ofnode_read_u32_index(node, propname, base + 1, &range->offset);
if (ret)
break;
ret = ofnode_read_u32_index(node, propname, base + 2, &range->npins);
if (ret)
break;
ret = ofnode_read_u32_index(node, propname, base + 3, &range->gpiofunc);
if (ret)
break;
list_add_tail(&range->node, &priv->gpiofuncs);
range++;
}
return 0;
}
static int single_probe(struct udevice *dev)
{
struct single_pdata *pdata = dev_get_plat(dev);
struct single_priv *priv = dev_get_priv(dev);
u32 size;
INIT_LIST_HEAD(&priv->functions);
INIT_LIST_HEAD(&priv->gpiofuncs);
size = pdata->offset + pdata->width / BITS_PER_BYTE;
#if (IS_ENABLED(CONFIG_SANDBOX))
priv->sandbox_regs =
devm_kzalloc(dev, size * sizeof(*priv->sandbox_regs),
GFP_KERNEL);
if (!priv->sandbox_regs)
return -ENOMEM;
#endif
/* looks like a possible divide by 0, but data->width avoids this */
priv->npins = size / (pdata->width / BITS_PER_BYTE);
if (pdata->bits_per_mux) {
if (!pdata->mask) {
dev_err(dev, "function mask needs to be non-zero\n");
return -EINVAL;
}
priv->bits_per_pin = fls(pdata->mask);
priv->npins *= (pdata->width / priv->bits_per_pin);
}
if (single_add_gpio_func(dev))
dev_dbg(dev, "gpio functions are not added\n");
dev_dbg(dev, "%d pins\n", priv->npins);
return 0;
}
static int single_of_to_plat(struct udevice *dev)
{
void *addr;
fdt_size_t size;
struct single_pdata *pdata = dev_get_plat(dev);
int ret;
ret = dev_read_u32(dev, "pinctrl-single,register-width", &pdata->width);
if (ret) {
dev_err(dev, "missing register width\n");
return ret;
}
switch (pdata->width) {
case 8:
case 16:
case 32:
break;
default:
dev_err(dev, "wrong register width\n");
return -EINVAL;
}
addr = dev_read_addr_size_index_ptr(dev, 0, &size);
if (!addr) {
dev_err(dev, "failed to get base register address\n");
return -EINVAL;
}
pdata->offset = size - pdata->width / BITS_PER_BYTE;
pdata->base = addr;
ret = dev_read_u32(dev, "pinctrl-single,function-mask", &pdata->mask);
if (ret) {
pdata->mask = 0;
dev_warn(dev, "missing function register mask\n");
}
pdata->bits_per_mux = dev_read_bool(dev, "pinctrl-single,bit-per-mux");
/* If no pinctrl-cells is present, default to old style of 2 cells with
* bits per mux and 1 cell otherwise.
*/
ret = dev_read_u32(dev, "#pinctrl-cells", &pdata->args_count);
if (ret)
pdata->args_count = pdata->bits_per_mux ? 2 : 1;
return 0;
}
const struct pinctrl_ops single_pinctrl_ops = {
.get_pins_count = single_get_pins_count,
.get_pin_name = single_get_pin_name,
.set_state = single_set_state,
.get_pin_muxing = single_get_pin_muxing,
.request = single_request,
};
static const struct udevice_id single_pinctrl_match[] = {
{ .compatible = "pinctrl-single" },
{ /* sentinel */ }
};
U_BOOT_DRIVER(single_pinctrl) = {
.name = "single-pinctrl",
.id = UCLASS_PINCTRL,
.of_match = single_pinctrl_match,
.ops = &single_pinctrl_ops,
.plat_auto = sizeof(struct single_pdata),
.priv_auto = sizeof(struct single_priv),
.of_to_plat = single_of_to_plat,
.probe = single_probe,
};